In mammals, mothers are essential for the early development of their infants since they provide postnatal care (Maestripieri and Mateo, 2009). Maternal loss in mammals reduces growth (Samuni et al., 2020), survival (Watts et al., 2009; Andres et al., 2013; Tung et al., 2016; Stanton et al., 2020), and long-term reproductive success (Andres et al., 2013; Strauss et al., 2020; Crockford et al., 2020; Zipple et al., 2021, reviewed in Clutton-Brock, 2016).

The biological embedding model (BEM; Power and Hertzman, 1997; Miller et al., 2011; Berens et al., 2017) posits that adversity experienced early in life, including exposure to severe stressors, can have deleterious consequences on an individual’s physiology and health across their lifespan. The BEM provides a promising conceptual framework to investigate the mechanisms underlying the fitness costs of maternal loss or other forms of early-life adversity.

Early-life adversity impacts several interconnected physiological pathways (reviewed in Berens et al., 2017) among which the hypothalamic-pituitary-adrenal (HPA) axis plays a central role (Miller et al., 2009; Miller et al., 2011; Taylor et al., 2011). The HPA axis is activated in response to internal physiological challenges and external stressors through a chain of reactions, known as the ‘stress response’ or ‘reactive homeostasis’ (Romero et al., 2009), which also results in the release of glucocorticoids (Sapolsky, 2002). Overall, this response is adaptive (Charmandari et al., 2005): it stimulates the release of energy out of storage in the form of glucose, and it increases cardiovascular circulation and respiratory rate, allowing organisms to respond to acute stressors such as predators (Sapolsky, 2002). Exposure to harsh social conditions during childhood, such as those that can result from maternal loss, may lead to repeated and prolonged activation of the HPA axis early in life. These activations provide an adaptive physiological response by mobilizing energy that helps children to cope with the immediate socioecological challenges but may result in long-term HPA axis dysfunction (i.e., either hypo- or hyper-responsiveness to stressor; Miller et al., 2011; Ehrlich et al., 2016; Berens et al., 2017). The HPA axis is considered to be at the core of the link between early-life adversity and fitness since repeated activation of the HPA axis over prolonged periods (chronic stress) and/or HPA axis malfunctioning can have detrimental effects on individual overall health (Sapolsky, 2002; Slavich and Cole, 2013). For instance, over- and/or prolonged activation of the HPA axis is known to suppress the immune system (Grossman, 1985; Setchell et al., 2010; Slavich and Cole, 2013) and reduce survival (Campos et al., 2021). The HPA axis also mediates some of the observed negative effects of early-life adversity on the immune response, such as elevated levels of inflammatory markers in the blood (Danese et al., 2011; Ehrlich et al., 2016; Rasmussen et al., 2019, reviewed in Berens et al., 2017). Assessing the consequences of traumatic early-life events, such as maternal loss, on the functioning of the HPA axis can provide insight into the mechanisms underlying the documented fitness costs of such events.

An alternative framework, the adaptive calibration model (ACM), proposes that intra- and interindividual changes in stress responsivity are mainly always adaptive, allowing individuals to adjust their physiology during development to respond to changes in the social and ecological environment (Del Giudice et al., 2011). According to the ACM, modification of the HPA axis activity in response to maternal loss should be sensitive to the phase of development at which maternal loss occurred, especially in long-lived species with an extended immature period. These modifications should then also be more or less long-lasting depending on the social and ecological environment faced by the developing individual after maternal loss (e.g., the amount of support provided by conspecifics).

In humans, maternal loss leads to short- and long-term alterations of the HPA axis functioning (Heim and Nemeroff, 2001; Sánchez et al., 2001). These effects are typically studied by investigating patterns of cortisol secretion, the main glucocorticoid circulating in mammals, including humans. In humans, cortisol levels follow a diurnal pattern characterized by an early morning peak (awakening response) and a regular decline throughout the day (Doman et al., 1986). These diurnal cortisol slopes, as well as the cortisol-awakening response specifically, serve as health markers. Deviations from stereotypical patterns of high morning and low evening cortisol levels (i.e., flatter diurnal slopes) are typically interpreted as indications of pathology and HPA axis dysregulation and/or a marker of chronic stress in human clinical studies (Pruessner et al., 1999; Sánchez et al., 2001; Clow et al., 2004; Kudielka et al., 2006; Miller et al., 2007). Flattening of diurnal cortisol slopes reflects a compression in the dynamic range of the HPA axis functioning (Karlamangla et al., 2019) that is indicative of a reduced ability to respond appropriately to stressors and to downregulate hormonal stress levels. Flatter diurnal cortisol slopes are even associated with direct fitness costs in humans such as reduced survival (Sephton et al., 2000).

To test conclusively the BEM and the ACM, which are diverging in some predictions (see above) but are not mutually exclusive, the physiological consequences of maternal loss must be investigated both during childhood and adulthood. Studies during childhood and in the time directly following maternal loss allow the assessment of the proximate responses of orphans in coping with new social challenges, while studies during adulthood allow the assessment of the long-lasting effects of early-life adversity. In humans, orphaned children typically exhibit lower cortisol-awakening responses (Carlson and Earls, 1997) or higher evening cortisol levels (Gunnar et al., 2001) than mother-reared children, leading to overall flatter diurnal cortisol slopes (Carlson and Earls, 1997; Tarullo and Gunnar, 2006). Flatter diurnal slopes and/or lower morning cortisol levels are also observed in children experiencing other forms of early-life adversity, such as maltreatment by parents, parental divorce, or placement in foster families (Kaufman, 1991; Hart et al., 1996; Meinlschmidt and Heim, 2005; Dozier et al., 2006; Bernard et al., 2015; McLachlan et al., 2016). Beyond diurnal cortisol patterns, maternal loss can also be associated with generally higher cortisol levels throughout the day (Gunnar et al., 2001).

In support of the BEM, several studies in humans document long-lasting effects of early-life adversity, including maternal loss, on children’s HPA axis functioning. Adults up to 64 years old who experienced mistreatment and/or the loss of one or both parents during childhood have, depending on the study, a lower (Meinlschmidt and Heim, 2005; Kawai et al., 2017) or higher cortisol-awakening response (Gonzalez et al., 2009; Butler et al., 2017), flatter diurnal cortisol slopes (Karlamangla et al., 2019), and generally higher cortisol levels throughout the day (Nicolson, 2004). However, the effects of parental loss on HPA axis activity are not long-lasting under every condition. In support of the ACM, human orphans adopted at an early age presented diurnal cortisol excretion profiles comparable to children raised by their biological parents (Gunnar et al., 2001). In contrast, orphans placed for extended periods in orphanages in which they were exposed to drastic food restriction and physical violence presented higher cortisol levels than mother-reared children, particularly in the evening (Gunnar et al., 2001).

Studies on captive non-human primates also reveal that the social environment can mediate the physiological effects of maternal loss. Orphan bonobo immatures raised in sanctuaries by surrogate human mothers presented similar cortisol levels to mother-reared immatures (Wobber and Hare, 2011), whereas nursery-reared orphan macaques, without surrogate mothers, presented blunted morning cortisol levels (Thomas et al., 1995). These studies on human and non-human primates highlight the flexible nature of the HPA axis functioning and how it can be reshaped by changes in the environment throughout development. However, the social support available to orphans in these captive studies is distinct from the potential support of conspecifics in the wild. Therefore, evaluating both the BEM and the ACM in wild long-lived animals with a slow life history is essential to understand the evolutionary roots of the human stress response and its role in developmental plasticity.

Wild long-lived mammals are adapted to the environment in which we typically observe them and in which selection may have favored mechanisms of rapid recovery from early-life traumatic events such as maternal loss to avoid long-term hyperactivation of the HPA axis (or chronic stress; Beehner and Bergman, 2017). Indeed, unlike humans, most long-lived mammalian species do not demonstrate alloparental care (Lukas and Clutton-Brock, 2012); therefore, in such species, individuals that lose their mothers during ontogeny may lack social support from their conspecifics following maternal loss.

A study on wild female baboons, using extensive long-term data, showed that simultaneous exposure to several forms of early-life adversity, and some isolated forms of adversity such as drought and low maternal rank, leads to an overall elevation in glucocorticoid levels in adulthood (Rosenbaum et al., 2020), offering support for the BEM. However, maternal loss in isolation did not lead to long-term elevation of glucocorticoid levels, suggesting that baboons may have buffering mechanisms to offset the effects of biological embedding for some forms of early-life adversity.

To our knowledge, this study on wild baboons constitutes the only test of the BEM in a wild long-lived non-human mammal. More studies are necessary to investigate if, or how extensively, the BEM and/or the ACM models apply to a wider range of long-lived species, both during development and in adulthood. In particular, it is important to test this model in long-lived species with a life history closer to that of humans. Baboons start reproducing 3–4 years after weaning, whereas humans and great apes, including chimpanzees, share an extended juvenile phase between weaning and first reproduction (Wittig and Boesch, 2019a). Furthermore, the study on baboons only assessed one marker of the HPA axis functioning (i.e., overall glucocorticoid levels) but did not investigate the impact on diurnal cortisol slopes. Assessments of diurnal cortisol slopes are important since these slopes are a marker of the HPA axis functioning (Karlamangla et al., 2019).

Using a long-term database, including demographic and urinary cortisol data, collected over a 19-year period on four wild Western chimpanzee communities (Pan troglodytes verus), we provide a rare test of the BEM and the ACM in a wild long-lived mammal. Specifically, our dataset allowed us to assess both the short- and long-term effects of maternal loss on the HPA axis activity in wild chimpanzees. We thereby investigated one of the potential physiological mechanisms explaining the fitness costs associated with maternal loss reported in wild chimpanzees such as reduced growth, survival, and reproductive success (Nakamura et al., 2014; Samuni et al., 2020; Stanton et al., 2020; Crockford et al., 2020). Furthermore, studying physiological effects using diurnal cortisol slopes is an underused paradigm in wild animal subjects despite its prevalence in the human health literature. In chimpanzees, these slopes are repeatable in adults (i.e., are consistent within a given individual over time, Sonnweber et al., 2018) but also show plasticity to physiological challenges such as disease outbreaks (Behringer et al., 2020) or aging (Emery Thompson et al., 2020).

For immatures, we investigated the effect of maternal loss in both sexes. For adult individuals, we focused on males, the philopatric sex in chimpanzees (Pusey, 1979; Boesch and Boesch-Achermann, 2000), since the early-life history of adult females who immigrated as adults into our study groups is often undocumented. For both age-class groups (male and female immatures and adult males), we first assessed whether average cortisol levels and the steepness of the diurnal cortisol slopes differed between orphan and non-orphaned individuals. We predicted that, as in humans, immature orphans would exhibit higher overall cortisol levels and flatter diurnal cortisol slopes than non-orphans. We also predicted that the overall effect of maternal loss on cortisol profiles would be more severe during the first years following maternal loss. That is because recently orphaned individuals have to adjust behaviorally and physiologically to a new social situation in which they do not benefit from maternal support and may have reduced access to food and socio-positive social interactions. Over time, orphans may adjust and develop compensatory strategies for this adversity, which could result in lower impacts on the HPA axis activity. Wild chimpanzees are known to provide social support to orphaned individuals, ranging from tolerance in feeding sites to full adoptions (i.e., daily consistent provisioning of care to the orphans such as carrying, grooming, food sharing; Uehara and Nyundo, 1983; Goodall, 1986; Wroblewski, 2008; Boesch et al., 2010; Hobaiter et al., 2014; Samuni et al., 2019a). This may provide similar social buffering to that observed in some human populations or in captive primate studies. Accordingly, following a peak in the modification of the orphan cortisol profile directly after maternal loss, we anticipated, as would be predicted by the ACM, some decline over time, but still for cortisol levels to remain elevated in orphans compared to non-orphans. Since the life history of chimpanzees, and especially the extended immature phase, resembles more the life history of humans than that of baboons, we predicted this to last even into adulthood, matching the patterns of long-lasting HPA activity alteration arising from maternal loss in humans.

Finally, in line with the ACM, which predicts differential adaptive physiological response at different stages of ontogeny, we predicted that immature orphans that lost their mothers earlier in their lives would have flatter diurnal cortisol slopes and overall higher cortisol levels than immatures who lost their mother at a later age due to a greater level of dependency on mothers in early ontogeny (Clark, 1977; Pusey, 1983; Boesch and Boesch-Achermann, 2000).

We used the long-term behavioral, demographic, and urine sample data of the Taï Chimpanzee Project (Wittig and Boesch, 2019b) collected on four communities of wild Western chimpanzees (East, North, Middle, and South) in the Taï National Park, Cote d’Ivoire (5°52′N, 7°20′E). The urine samples included in this study span over 19 years and were collected between 2000 and 2018 (see Table 1 for details about sample size).

We used a series of Bayesian linear mixed models (LMMs) to test our predictions regarding the effect of maternal loss on overall cortisol levels and diurnal slopes (jointly constituting the cortisol profile). We first tested these effects in socially immatures (i.e., males and females < 12 years of age because prior to 12 years, chimpanzees associate primarily with their mother; Reddy and Sandel, 2020). Secondly, we tested these effects in mature males (i.e., males ≥ 12 years of age).

For immatures, we first tested for differences in cortisol profiles between recently orphaned individuals (individuals that lost their mother for less than 2 years previously, N = 7), not-recently orphaned orphans (individuals that experienced maternal loss more than 2 years ago at the time of sampling, N = 16), and non-orphans (N = 36; all immature model). We used these three categories to test whether the potential effect of maternal loss on immature cortisol profiles is a short-term effect (i.e., only present in recently orphaned individuals) or endures throughout ontogeny. Note that immatures that have been sampled over several years can appear in two or three of the orphan categories in our dataset. We then tested, for immature orphans only (N = 17), the effect of two test predictors on their cortisol profiles, the age at which the orphan lost their mother, and the years since maternal loss (immature orphan model).

For mature males, we tested first for differences in cortisol profiles between orphans (individuals who were orphaned before reaching 12 years of age, N = 11) and non-orphans (N = 17, all adult male model). We then tested, for orphan mature males only (N = 11), the effect of the age at which the orphans lost their mother on their cortisol profiles (adult male orphan model). In all the models, each urine sample represented a data point, and the log-transformed cortisol concentration of the sample (expressed in ng/ml SG) was the response variable. We summarize below the test predictors we used in each model:

• In the all immature model, we used as a test predictor a categorical variable for orphan status with three levels: recently orphaned (within 2 years after maternal loss), not-recently orphaned (longer than 2 years since maternal loss), and non-orphans.

• In the immature orphan model, we used two test predictors: age when mother died, a continuous variable describing the age at which immature orphans lost their mother and years since maternal loss, a continuous variable describing the number of years since immature orphans lost their mother.

• In the all adult male model, we used as a test predictor a binary variable for orphan status, namely if the mature male had been orphaned before reaching 12 years of age yes/no.

• In the adult male orphan model, we used as a test predictor age when mother died, a continuous variable as in the immature orphan model.

For each of the models, all these test predictors were included in interaction with the linear, the quadratic, and the cubic terms for time of sample collection to test the effect of these test predictors on diurnal cortisol slopes. The quadratic and cubic terms for time of sample collection were included here since a previous study using a large sample size showed that diurnal cortisol slopes in chimpanzees follow a curved cubic pattern (Emery Thompson et al., 2020). While our models differ in the test predictors that were included in each model, the control predictors were nearly identical for all models (except for the immature orphan model, see below). In all our models, we controlled for sex of the individual, community size, sex ratio of mature individuals in the community, age of the individual (except for the immature orphan model because of collinearity issue, see Materials and methods), the liquid chromatography mass spectrometry (LCMS) method used (‘old’ or ‘new’ method, see the Urine analysis section), and seasonal variation in ecological conditions (see Materials and methods). In addition, we controlled for repeated observations of the same individual over the same year by incorporating individual ID and year as random factors in each model. Finally, to control for the changes in cortisol diurnal slope with age, we built one slope per individual per year into each model by incorporating the dummy variable individual_year as a random factor. For each model, we first ran a full model comprising all the variables and interactions described above. When the 90% credible interval (CI) for the estimate of an interaction term overlapped 0, this indicated a large uncertainty and therefore that the effect of this interaction was not consistent in our data. Accordingly, we reran the models without those non-consistent interaction terms and present the results of the final reduced models. For each model, we report P + and P- as the percentage of the posterior distribution in support of the hypothesized positive or negative effect given the observed data. We also report the proportion of variance in the response explained by the fixed effects and the random effects (conditional R²) and the fixed effect only (marginal R²). Throughout the result section, we describe differences in cortisol levels in the morning, in the afternoon, or throughout the day, based on the visual inspection of the model line predictions and on the model output regarding diurnal cortisol slopes. We did not directly test differences in cortisol levels in morning or afternoon samples separately due to limitations in the sample size, but these differences can be assessed based on the diurnal cortisol slopes depicted in the figures.

Effect of maternal loss on immature cortisol profiles

For the all immature model (N = 846 samples and 50 individuals), assessing if immature individuals that were either recently orphaned, not-recently orphaned, and non-orphans differed in their cortisol profiles, the 90% CI for the estimate of the interaction terms between orphan status and the quadratic and the cubic term for time of the day overlapped with 0. In the reduced model not comprising these two interaction terms, we found that recently orphaned individuals had a consistently steeper linear slope (estimate for the interaction between orphan status and the linear term for time of the day: –0.22, 95% CI: [–0.03: –0.48], P = 98.9%, Table 2, Figure 1, and Appendix 1—figure 1) than non-orphans. On average, recently orphaned individuals had a diurnal cortisol slope 58% steeper than non-orphans (average slope and [95% CI] for recently orphaned: –0.60 [-0.79: –0.41], and non-orphans: –0.38 [-0.53: –0.22]). A visual inspection of the model line prediction (Figure 1) indicates that this difference in slopes may stem from higher early morning cortisol levels in recently orphaned immatures as compared to non-orphan immatures (Figure 1 and Appendix 1—figure 1). However, we found no consistent differences in the linear slopes of not-recently orphaned immatures and non-orphan immatures (90% CI: [–0.24: 0.01], P = 93.4%, Table 2, Figure 1, and Appendix 1—figure 1). Since our model comprised an interaction between orphan status and time of the day, the main effect of orphan status inherently corresponds only to the difference at midday. The 95% CI for the main effect of orphan status in the all immature model clearly overlapped 0, indicating that there was no consistent difference between recently orphaned, not-recently orphaned, and non-orphans in their average cortisol levels at midday. However, Figure 1 indicates that recently orphaned individuals had higher cortisol levels in the morning and lower cortisol levels in the afternoon than non-orphans. Conditional and marginal R² for the all immature model were 0.60 and 0.26, respectively.

Table 2

Model	Response	Predictor	Estimate	SE	95% CI low	95% CI high	90% CI low	90% CI high
All immature model	Log urinary cortisol levels (ng/ml SG)	Intercept	3.69	0.27	3.17	4.23	3.26	4.14
		Time of the day	–0.38	0.08	–0.53	–0.22	–0.51	–0.25
		Time of the day2	–0.01	0.08	–0.17	0.14	–0.14	0.11
		Time of the day3	0.03	0.04	–0.05	0.12	–0.04	0.10
		Orphan category (orphan for less than 2 years)	0.01	0.19	–0.36	0.38	–0.30	0.32
		Orphan category (orphan for more than 2 years)	–0.14	0.22	–0.58	0.32	–0.50	0.23
		Sex ratio	0.07	0.07	–0.06	0.2	–0.04	0.18
		Community size	–0.12	0.06	–0.24	0.01	–0.22	–0.01
		Individual sex (male)	–0.26	0.25	–0.78	0.22	–0.67	0.13
		Individual age at sample	–0.01	0.17	–0.33	0.37	–0.27	0.29
		LCMS method (old)	0.06	0.35	–0.62	0.78	–0.50	0.64
		Sin(seasonDate)	–0.07	0.04	–0.14	0	–0.13	–0.02
		Cos(seasonDate)	0.01	0.04	–0.06	0.08	–0.05	0.07
		Orphan category (less than 2 years): time of the day	–0.22	0.10	–0.41	–0.03	–0.38	–0.07
		Orphan category (more than 2 years): time of the day	–0.11	0.08	–0.27	0.03	–0.24	0.01
Immature orphan model	Log urinary cortisol levels (ng/ml SG)	Intercept	3.59	0.26	3.08	4.11	3.17	4.02
		Time of the day	–0.44	0.17	–0.75	–0.05	–0.7	–0.14
		Time of the day2	–0.01	0.07	–0.16	0.13	–0.13	0.1
		Time of the day3	0.03	0.11	–0.18	0.27	–0.14	0.22
		Orphan’s age when mother died	0.00	0.24	–0.51	0.46	–0.41	0.38
		Years since maternal loss	–0.16	0.16	–0.48	0.14	–0.42	0.09
		Sex ratio	–0.06	0.11	–0.27	0.15	–0.23	0.11
		Community size	–0.05	0.08	–0.22	0.12	–0.19	0.09
		Individual sex (male)	0.1	0.3	–0.47	0.7	–0.38	0.6
		LCMS method (old)	–0.04	0.32	–0.66	0.64	–0.55	0.51
		Sin(seasonDate)	0.02	0.05	–0.09	0.12	–0.07	0.1
		Cos(seasonDate)	0.04	0.05	–0.07	0.14	–0.05	0.12
		Years since maternal loss: time of the day	0.19	0.1	–0.01	0.38	0.02	0.35
		Years since maternal loss: time of the day2	–0.05	0.05	–0.15	0.05	–0.14	0.03
		Years since maternal loss: time of the day3	–0.06	0.04	–0.15	0.02	–0.14	0.00
		Age when mother died: time of the day	0.08	0.09	–0.09	0.25	–0.06	0.22
		Age when mother died: time of the day2	–0.12	0.06	–0.23	–0.00	–0.21	–0.02

Figure 1

Download asset Open asset

The second model (immature orphan model, N = 393 samples and 17 individuals) focusing on immature orphans revealed that orphans varied in their cortisol profiles depending on the age at which their mother died and on the length of time since they were orphaned. More specifically, and in line with the results of the all immature model, immature orphans whose mother had died several years before sampling had different diurnal cortisol slopes as compared to more recently orphaned immatures (estimate and [90% CI] for the interaction between years since maternal loss and the cubic term for time of day: –0.06 [-0.14: 0.00]). However, this difference comprised a relatively large uncertainty since the 90% CI comprised 0, the 95% CI overlapped 0 (Table 2), and P- = 93.8%. Immature orphans who lost their mother recently had a diurnal cortisol slope that curved upwards in the afternoon (orange squares in Figure 2). A visual inspection of the data and the model prediction lines reveals that cortisol levels of immature orphans who recently lost their mother had higher cortisol levels throughout the day (orange squares in Figure 2) as compared to immature orphans who lost their mother several years ago (green triangles in Figure 2). The difference in cortisol levels between recently and non-recently orphaned individuals was most evident during early morning and late afternoon (Figure 2).

Figure 2

Download asset Open asset

In addition, in the immature orphan model, we found a consistent effect of the interaction between age when mother died and the quadratic term of time of day (estimate and [95% CI]: –0.12 [-0.23: –0.00], P- = 97.8%, Table 2), indicating that the age at which immatures lost their mother consistently influenced their diurnal cortisol slopes and more specifically how it curved throughout the day. Immature individuals who lost their mother before 5 years of age (orange squares in Figure 3) had a diurnal cortisol slope that curved upwards. A visual inspection of the model prediction lines and the data points reveals that immatures orphaned at a younger age presented higher early morning and late afternoon cortisol levels than individuals who lost their mother at an older age. Those who lost their mother between 5 and 8 years of age (blue circles in Figure 3) had a relatively linear decrease in cortisol levels throughout the day. Finally, immature individuals orphaned between 8 and 12 years of age (green triangles in Figure 3) had a diurnal cortisol slope that curved downwards with lower early morning and late afternoon cortisol levels than individuals who lost their mother at a younger age. Please note that we depicted model line predictions for three age categories corresponding to three life history stages in chimpanzees (below 5 years: infancy, 5–8 years: juvenile period, 8–12 years: early adolescence) in Figure 3 for ease of interpretation of the effect, but the variable ‘age when mother died’ was incorporated in the immature orphan model as a continuous predictor. Conditional and marginal R² for the immature orphan model were 0.64 and 0.32, respectively.

Figure 3

Download asset Open asset

While the effect of maternal loss on wild animal survival and reproduction has been recently established (Foster et al., 2012; Andres et al., 2013; Tung et al., 2016; Walker et al., 2018; Surbeck et al., 2019; Crockford et al., 2020; Zipple et al., 2021), the mechanisms underlying these fitness costs remain understudied. Our study provides one of the rare empirical tests of the BEM (see also Rosenbaum et al., 2020) and ACM in wild long-lived mammals by assessing the short- and long-term physiological impacts of early maternal loss. While we found an effect of maternal loss on diurnal cortisol slopes in immature chimpanzees whose mothers died recently (all immature and immature orphan models), these effects were neither present in individuals who lost their mothers more than 2 years earlier (all immature and immature orphan models) nor in mature male chimpanzees (adult male orphan model). These results are in line with the absence of long-term effects of maternal loss alone on glucocorticoid levels in wild long-lived baboons (Rosenbaum et al., 2020). This suggests that the BEM (Power and Hertzman, 1997; Miller et al., 2011; Berens et al., 2017) may apply to long-lived wild mammals only following exposure to a combination of diverse sources of early-life adversity or to other sources of early-life adversity than maternal loss (see Rosenbaum et al., 2020). Our results also provide tentative support for the ACM. The ACM predicts that exposure to adversity leads to modification of the HPA axis activity, but that these modifications are more or less long-lasting depending on the social and ecological environment faced by the developing individual after exposure to adversity (Del Giudice et al., 2011). In the case of our study, amelioration in the social environment of immature chimpanzees, possibly in the form of buffering mechanisms ranging from minimal alloparental care to adoption (discussed below), could modify the HPA axis activity of orphans throughout ontogeny and eventually ameliorate long-term effects of early-life adversity. Alternatively, in our study, a survivorship bias may mean that chimpanzees with severely altered HPA axis activity following maternal loss did not survive to adulthood and thus were absent from our adult dataset. We also found that the age at maternal loss impacted the diurnal cortisol slopes of the orphans (immature orphan model). Orphans experiencing maternal loss at younger ages had a diurnal cortisol slope differing from the immatures orphaned when older, particularly in the quadratic term for time of the day (i.e., in how the slope curved). Orphans that lost their mother at younger ages had diurnal cortisol slopes that curved upwards towards the end of the day, whereas those that lost their mother later in ontogeny showed a more typical pattern observed in non-orphan immatures of more continual declines in their slopes. Visual inspection of the model lines and data points revealed that this difference in the slope led to higher early morning and late afternoon cortisol levels in orphan whom mother died early than in those of immatures orphaned when older. This latter finding is also in line with the ACM by highlighting potentially different physiological adaptive responses at different stages of ontogeny. In fact, diurnal cortisol slopes are indicative of the general functioning of the HPA axis (Karlamangla et al., 2019) and our results indicate that the diurnal cortisol slope of immature chimpanzees undergoes different levels of changes depending on the age at which they experience maternal loss, with more substantial deviations from mother-raised offspring pattern in immatures orphaned earlier in life.

In line with our prediction, orphans had different diurnal cortisol slopes as compared to non-orphans (all immature model). However, this difference was only present for orphans who recently lost their mother, that is, within 2 years at the time of sampling. Furthermore, the direction of the effect was opposite to the prediction derived from the clinical human literature (Kaufman, 1991; Hart et al., 1996; Meinlschmidt and Heim, 2005; Dozier et al., 2006; Bernard et al., 2015; McLachlan et al., 2016) in that recently orphaned immatures had a steeper rather than a flatter diurnal cortisol slopes compared to those of non-orphaned immatures. An inspection of the raw data (Figure 1 and Appendix 1—figure 1) reveals that this steeper slope is more likely to be driven by higher early morning cortisol levels rather than by lower late afternoon cortisol levels.

In humans, flatter diurnal cortisol slopes in individuals who experienced early-life adversity are often related to higher afternoon cortisol levels (Carlson and Earls, 1997; Gunnar et al., 2001; Tarullo and Gunnar, 2006). A failure to bring cortisol levels down in the afternoon has been interpreted as a fundamental dysregulation within the neuroendocrine system (Young et al., 1994). Such dysregulation might not apply to recently orphaned immature chimpanzees since, as compared to non-orphans, their diurnal cortisol slopes were steeper and their afternoon cortisol levels were not necessarily higher. The diurnal cortisol profile of recently orphaned immature chimpanzees better mirrors the cortisol profiles of individuals exposed to nutritional stress. Studies on humans and captive rats show that dietary restriction is associated with a rise in early morning cortisol levels (Goodwin et al., 1988; Garcia‐Belenguer et al., 1993). Nutritional stress in orphan chimpanzees may result from the lack of access to food sources that were initially provided by the mother (Pusey, 1983; Goldenberg and Wittemyer, 2017; Samuni et al., 2019a). Although chimpanzees in our study were orphaned after weaning age (i.e., after 4 years of age, Samuni et al., 2020; Appendix 1—table 3), these orphans may nonetheless be constrained in acquiring the amount of food necessary to maintain a positive energy balance. This may be because orphans lack socially facilitated access to food sources previously provided by their mother and also miss out on maternal sharing of high nutrient food, such as meat, nuts, and honey, which occurs intermittently throughout ontogeny (Samuni et al., 2019a). During the period directly following the loss of their mother, orphan chimpanzees might be exposed to acute stress, which modifies adaptively their HPA axis activity, allowing maintenance of homeostasis while developing and eventually survival until maturity, despite the potential lack of energy intake.

Such modifications, and in particular early morning cortisol elevations, may mediate other fitness-related traits. For instance, in the same population, orphans lose out on growth compared to non-orphans (Samuni et al., 2020), which might result from reallocation of energy towards more vital functions (e.g., thermoregulation, muscle activity, general organismal functioning) than growth during energetically challenging periods. Early morning cortisol levels can be indicative of such re-allocation. In contrast, nutritional stress might be less prominent in sampled human orphans, which are predominantly studied in Western countries, where, even in orphanages, children are likely to receive sufficient amounts of food. This might explain why, unlike human orphans that have lower morning cortisol levels (Meinlschmidt and Heim, 2005; Dozier et al., 2006), recently orphaned chimpanzees being under nutritional stress have higher early morning cortisol levels than non-orphans.

The results of our orphan/non-orphan comparison in immature chimpanzees (all immature model) were partially confirmed by our second analysis that included only samples from immature orphans (immature orphan model). This allowed us to assess the effect of years since maternal loss as a continuous variable while accounting for the age at which each orphan lost their mother. In this analysis, we found that recently orphaned immatures and immatures who lost their mother at a younger age had higher early morning cortisol levels and higher late afternoon cortisol levels as compared to other orphans. However, we note that there was some uncertainty in the strength and direction of this effect, which is likely due to the limited sample size for this analysis and the variable nature of experiences among orphans following maternal loss (see below). Nevertheless, this result implies that recently orphaned immatures and orphans who lost their mothers at a younger age are exposed to more stressors during the time when chimpanzees are the most socially active (i.e., early morning and late afternoon). This could reflect higher nutritional or energetic stress for orphans when groups are most active. These early morning and late afternoon peaks could equally reflect exposure to social stressors when other individuals are most active. Play bouts involving orphan chimpanzees are shorter and more frequently escalate into aggression when compared to play bouts between non-orphans (Botero et al., 2013; Leeuwen et al., 2014). This could, in turn, lead to increased cortisol levels since in chimpanzees and other primates aggression generally increases cortisol levels (Girard-Buttoz et al., 2009; Emery Thompson et al., 2010; Wittig et al., 2015; but see Preis et al., 2019).

Mothers are the main social partners of immature chimpanzees Reddy and Sandel, 2020; therefore, orphan chimpanzees lack access to a key source of social buffering from environmental stressors (Young et al., 2014; Wittig et al., 2016). Orphans may later form relationships that provide such buffering, but in the short term, new orphans and/or those who lost their mother at a young age may be more exposed to cumulative social and psychological stressors without access to social buffering mechanisms. The social factors affecting chimpanzee orphans may be similar to those impacting human orphans who are frequent targets of assault due to a lack of social support from a caregiver (e.g., Frank et al., 1996).

As in humans, maternal loss impacts the physiology of immature chimpanzees. However, the effects of maternal loss on chimpanzees’ physiology differ strongly from that of humans, in that they do not persist into adulthood. More strikingly, immatures that were orphaned for more than 2 years had cortisol excretion profiles that were not consistently different from the profiles of non-orphans. Adult humans up to 64 years old, which experienced mistreatment and/or the loss of one or both parents during childhood, still present alteration of their diurnal cortisol slopes and overall cortisol levels (Nicolson, 2004; Meinlschmidt and Heim, 2005; Gonzalez et al., 2009; Kawai et al., 2017; Butler et al., 2017; Karlamangla et al., 2019). The re-establishment of a normal functioning of the HPA axis in mature male chimpanzees but also in immatures orphaned for more than 2 years may reflect a form of recovery in those individuals. In turn, this potential recovery is in line with the ACM, which predicts that the HPA axis activity has been selected to be adaptively flexible throughout ontogeny, allowing the individuals to readjust HPA axis activity following critical developmental challenges (in the case of our study, maternal loss). These changes in HPA axis activity are conditional adaptations to match the social and physical environment (Del Giudice et al., 2011). The lack of apparent long-term effects of maternal loss on immature chimpanzee physiology could thus be indicative of ameliorations in the environment of these orphans in the years following maternal loss, possibly in terms of improved access to social support and food.

Adoption by conspecifics is one such environmental modification that, in humans, is associated with cortisol profiles returning to similar levels as those of non-orphans (Gunnar et al., 2001). In chimpanzees, adoption is a relatively common phenomenon (Uehara and Nyundo, 1983; Goodall, 1986; Wroblewski, 2008; Boesch et al., 2010; Hobaiter et al., 2014; Samuni et al., 2019a). Adopters typically provide surrogate care for the orphan in the form of grooming, limited provisioning through food sharing, agonistic support, nest sharing, and even carrying (Samuni et al., 2019a). As a result, adoption may also increase the survival probability of the orphan (Hobaiter et al., 2014). It is possible that the adoption of immature orphan chimpanzees also alleviates some of the effects of maternal loss on cortisol excretion profiles as observed in humans (Gunnar et al., 2001). Unfortunately, data were insufficient to evaluate effectively the effect of adoption on cortisol excretion profiles in this sample (adoption status is available for 9 out of 17 orphans; Appendix 1—table 3). Most orphans for whom we have data were adopted at some point during their immature years (Appendix 1—table 3), but the level of care provided by the adopter varies greatly across orphans (Samuni et al., 2019a). Nevertheless, we lack detailed information on the alloparental care provided to most orphans in our sample to test the potential effect of the intensity of alloparental care on cortisol profiles.

While the ACM highlights that changes of the HPA axis activity during ontogeny when facing adversity are likely adaptive, it also emphasizes that exposure to extreme social situations, such as complete parental loss, may generate maladaptive phenotypes, especially if those events occur early during ontogeny (Del Giudice et al., 2011). Our results indicate that the age at maternal loss indeed has an impact on the orphan stress physiology in chimpanzees, but the subjects in our study may not have been orphaned at a young enough age to lead to long-term irreversible modifications of their physiology.

Long-term alteration of the HPA axis functioning related to early-life adversity is explained, at least partly, by epigenetic mechanisms (Weaver et al., 2004) and in particular by a hyper-methylation of DNA in regions coding for the glucocorticoid receptors (GRs) in the brain (Liu et al., 1997; Weaver et al., 2007, reviewed in Zhang et al., 2013). In rodents tested in laboratory conditions, these alterations of the GR in the brain take place early in life; a lower density of GR reduces the effectiveness of the glucocorticoid-negative feedback loop and ultimately results in prolonged elevated cortisol levels (Zhang et al., 2013). All orphan adult males in our study were orphaned after they were 4 years of age, an age at which the epigenetic effect on GR in the brain may be reduced or absent. Sampling individuals who lost their mother before weaning represents a methodological challenge since they often die soon after maternal loss (Nakamura et al., 2014; Stanton et al., 2020).

In conclusion, our study provides evidence of an effect of early-life adversity on diurnal cortisol slopes in immatures of a wild mammal population. Interestingly, our study provides contrasts to studies on humans by showing that the modifications of the HPA axis activity following maternal loss are not long-lasting. Even though modifications of the HPA axis activation may have some fitness consequences in wild mammalian populations (Campos et al., 2021), this mechanism is unlikely to explain the lower reproductive fitness and survival observed in adult male orphan chimpanzees (Nakamura et al., 2014; Crockford et al., 2020). In fact, in our study male orphans who reach adulthood do not present cortisol excretion profiles that differ from those of non-orphans. In wild baboons, maternal loss also did not have in itself long-term consequences on cortisol levels. Physiological embedding of maternal loss for this particular trait may thus not apply to, at least some, wild long-lived species. Prolonged alteration of the HPA axis functioning during ontogeny may not be viable in these species (Boonstra and Fox, 2013; Dantzer et al., 2016; Beehner and Bergman, 2017), and/or, as predicted by the ACM, improvement in the social environment, such as access to buffering mechanisms, such as alloparental care or adoption, allows the HPA axis activity to return to a regular level.

While modifications of HPA axis activity are unlikely to directly impact chimpanzee fitness in adult individuals, the modification in the HPA axis activity during the two first years following maternal loss found in our study might mediate fitness-related costs such as slowed growth (Samuni et al., 2020) rather than directly resulting in negative fitness outcomes. In general, future studies on wild mammals should link the effects of early-life adversity on an individual’s physiology to long-term fitness consequences in order to gain a clearer understanding of the selective forces at play. An investigation of the physiological and social differences, between orphans who do survive and those who do not reach maturity, as well as identifying and quantifying the effects of the buffering mechanisms that contribute to these differences, will be key in this process.

The data have been deposited on Dryad: https://doi.org/10.5061/dryad.gtht76hk2.

1. 1. Girard-Buttoz C
   2. Tkaczynski PJ
   3. Samuni L
   4. Fedurek P
   5. Gomes C
   6. Löhrich T
   7. Manin V
   8. Preis A
   9. Valé PF
   10. Deschner T
   11. Wittig RM
   12. Crockford C

   (2020) Dryad Digital Repository

   Data from: Early maternal loss affects diurnal cortisol slopes in immature but not mature wild chimpanzees.

   https://doi.org/10.5061/dryad.gtht76hk2

1. 1. Altmann J

   (1974) Observational study of behavior: Sampling methods

   Behaviour 49:227–267.

   https://doi.org/10.1163/156853974x00534

   • PubMed
   • Google Scholar
2. 1. Andres D
   2. Clutton-Brock TH
   3. Kruuk LEB
   4. Pemberton JM
   5. Stopher K
   6. Ruckstuhl KE

   (2013) Sex differences in the consequences of maternal loss in a long-lived mammal, the red deer (Cervus elaphus

   Behavioral Ecology and Sociobiology 67:1249–1258.

   https://doi.org/10.1007/s00265-013-1552-3

   • Google Scholar
3. 1. Baayen RH
   2. Davidson DJ
   3. Bates DM

   (2008) Mixed-effects modeling with crossed random effects for subjects and items

   Journal of Memory and Language 59:390–412.

   https://doi.org/10.1016/j.jml.2007.12.005

   • Google Scholar
4. 1. Barr DJ
   2. Levy R
   3. Scheepers C
   4. Tily HJ

   (2013) Random effects structure for confirmatory hypothesis testing: Keep it maximal

   Journal of Memory and Language 68:255–278.

   https://doi.org/10.1016/j.jml.2012.11.001

   • Google Scholar
5. 1. Beehner JC
   2. Bergman TJ

   (2017) The next step for stress research in primates: To identify relationships between glucocorticoid secretion and fitness

   Hormones and Behavior 91:68–83.

   https://doi.org/10.1016/j.yhbeh.2017.03.003

   • PubMed
   • Google Scholar
6. 1. Behringer V
   2. Preis A
   3. Wu DF
   4. Crockford C
   5. Leendertz FH
   6. Wittig RM
   7. Deschner T

   (2020) Urinary Cortisol Increases During a Respiratory Outbreak in Wild Chimpanzees

   Frontiers in Veterinary Science 7:485.

   https://doi.org/10.3389/fvets.2020.00485

   • PubMed
   • Google Scholar
7. 1. Berens AE
   2. Jensen SKG
   3. Nelson CA

   (2017) Biological embedding of childhood adversity: from physiological mechanisms to clinical implications

   BMC Medicine 15:135.

   https://doi.org/10.1186/s12916-017-0895-4

   • PubMed
   • Google Scholar
8. 1. Bernard K
   2. Zwerling J
   3. Dozier M

   (2015) Effects of early adversity on young children’s diurnal cortisol rhythms and externalizing behavior

   Developmental Psychobiology 57:935–947.

   https://doi.org/10.1002/dev.21324

   • PubMed
   • Google Scholar
9. Book

   1. Boesch C
   2. Boesch-Achermann H

   (2000)

   The Chimpanzees of the Taï Forest: Behavioural Ecology and Evolution

   Oxford: Oxford University Press.

   • Google Scholar
10. 1. Boesch C
    2. Bole C
    3. Eckhardt N
    4. Boesch H

    (2010) Altruism in forest chimpanzees: The case of adoption

    PLOS ONE 5:e8901.

    https://doi.org/10.1371/journal.pone.0008901

    • Google Scholar
11. 1. Boonstra R
    2. Fox C

    (2013) Reality as the leading cause of stress: Rethinking the impact of chronic stress in nature

    Functional Ecology 27:11–23.

    https://doi.org/10.1111/1365-2435.12008

    • Google Scholar
12. 1. Botero M
    2. MacDonald SE
    3. Miller RS

    (2013) Anxiety-related behavior of orphan chimpanzees (Pan troglodytes schweinfurthii) at Gombe National Park, Tanzania

    Primates; Journal of Primatology 54:21–26.

    https://doi.org/10.1007/s10329-012-0327-1

    • PubMed
    • Google Scholar
13. 1. Bürkner PC

    (2018) Advanced bayesian multilevel modeling with the R package brms

    The R Journal 10:395.

    https://doi.org/10.32614/RJ-2018-017

    • Google Scholar
14. 1. Butler K
    2. Klaus K
    3. Edwards L
    4. Pennington K

    (2017) Elevated cortisol awakening response associated with early life stress and impaired executive function in healthy adult males

    Hormones and Behavior 95:13–21.

    https://doi.org/10.1016/j.yhbeh.2017.07.013

    • PubMed
    • Google Scholar
15. Book

    1. Bygott J

    (1979)

    Agonistic behaviour, dominance, and social structure in wild chimpanzees of the Gombe National Park

    In: Bygott J, editors. The Great Apes. Benjamin-Cummings. pp. 405–428.

    • Google Scholar
16. 1. Campos FA
    2. Archie EA
    3. Gesquiere LR
    4. Tung J
    5. Altmann J
    6. Alberts SC

    (2021) Glucocorticoid exposure predicts survival in female baboons

    Science Advances 7:eabf6759.

    https://doi.org/10.1126/sciadv.abf6759

    • Google Scholar
17. 1. Carlson M
    2. Earls F

    (1997) Psychological and neuroendocrinological sequelae of early social deprivation in institutionalized children in romania

    Annals of the New York Academy of Sciences 807:419–428.

    https://doi.org/10.1111/j.1749-6632.1997.tb51936.x

    • PubMed
    • Google Scholar
18. 1. Charmandari E
    2. Tsigos C
    3. Chrousos G

    (2005) Endocrinology of the Stress Response

    Annual Review of Physiology 67:259–284.

    https://doi.org/10.1146/annurev.physiol.67.040403.120816

    • PubMed
    • Google Scholar
19. Software

    1. Clark C

    (1977)

    A preliminary report on weaning among chimpanzees of the Gombe National Park, Tanzania

    Primate Bio-Social Development.
20. 1. Clow A
    2. Thorn L
    3. Evans P
    4. Hucklebridge F

    (2004) The awakening cortisol response: Methodological issues and significance

    Stress 7:29–37.

    https://doi.org/10.1080/10253890410001667205

    • PubMed
    • Google Scholar
21. Book

    1. Clutton-Brock TH

    (2016)

    Mammal Societies

    Chichester, UK: John Whiley & Sons.

    • Google Scholar
22. 1. Crockford C
    2. Samuni L
    3. Vigilant L
    4. Wittig RM

    (2020) Postweaning maternal care increases male chimpanzee reproductive success

    Science Advances 6:eaaz5746.

    https://doi.org/10.1126/sciadv.aaz5746

    • PubMed
    • Google Scholar
23. 1. Danese A
    2. Caspi A
    3. Williams B
    4. Ambler A
    5. Sugden K
    6. Mika J
    7. Werts H
    8. Freeman J
    9. Pariante CM
    10. Moffitt TE
    11. Arseneault L

    (2011) Biological embedding of stress through inflammation processes in childhood

    Molecular Psychiatry 16:244–246.

    https://doi.org/10.1038/mp.2010.5

    • PubMed
    • Google Scholar
24. 1. Dantzer B
    2. Westrick SE
    3. Kesteren F

    (2016) Relationships between endocrine traits and life histories in wild animals: Insights, problems, and potential pitfalls

    Integrative and Comparative Biology 56:185–197.

    https://doi.org/10.1093/icb/icw051

    • Google Scholar
25. 1. Del Giudice M
    2. Ellis BJ
    3. Shirtcliff EA

    (2011) The Adaptive Calibration Model of stress responsivity

    Neuroscience & Biobehavioral Reviews 35:1562–1592.

    https://doi.org/10.1016/j.neubiorev.2010.11.007

    • Google Scholar
26. 1. Doman J
    2. Thompson S
    3. Grochocinski V
    4. Jarrett D
    5. Kupfer DJ

    (1986) A computer algorithm to determine the nadir and rise time in nocturnal cortisol secretion

    Psychoneuroendocrinology 11:359–366.

    https://doi.org/10.1016/0306-4530(86)90022-3

    • PubMed
    • Google Scholar
27. 1. Dozier M
    2. Manni M
    3. Gordon MK
    4. Peloso E
    5. Gunnar MR
    6. Stovall-McClough KC
    7. Eldreth D
    8. Levine S

    (2006) Foster children’s diurnal production of cortisol: An exploratory study

    Child Maltreatment 11:189–197.

    https://doi.org/10.1177/1077559505285779

    • PubMed
    • Google Scholar
28. 1. Ehrlich KB
    2. Ross KM
    3. Chen E
    4. Miller GE

    (2016) Testing the biological embedding hypothesis: Is early life adversity associated with a later proinflammatory phenotype

    Development and Psychopathology 28:1273–1283.

    https://doi.org/10.1017/S0954579416000845

    • PubMed
    • Google Scholar
29. 1. Emery Thompson M
    2. Muller MN
    3. Kahlenberg SM
    4. Wrangham RW

    (2010) Dynamics of social and energetic stress in wild female chimpanzees

    Hormones and Behavior 58:440–449.

    https://doi.org/10.1016/j.yhbeh.2010.05.009

    • PubMed
    • Google Scholar
30. 1. Emery Thompson M
    2. Fox SA
    3. Berghänel A
    4. Sabbi KH
    5. Phillips-Garcia S
    6. Enigk DK
    7. Otali E
    8. Machanda ZP
    9. Wrangham RW
    10. Muller MN

    (2020) Wild chimpanzees exhibit humanlike aging of glucocorticoid regulation

    PNAS 117:8424–8430.

    https://doi.org/10.1073/pnas.1920593117

    • Google Scholar
31. 1. Foerster S
    2. Franz M
    3. Murray CM
    4. Gilby IC
    5. Feldblum JT
    6. Walker KK
    7. Pusey AE

    (2016) Chimpanzee females queue but males compete for social status

    Scientific Reports 6:35404.

    https://doi.org/10.1038/srep35404

    • PubMed
    • Google Scholar
32. 1. Foster EA
    2. Franks DW
    3. Mazzi S
    4. Darden SK
    5. Balcomb KC
    6. Ford JKB
    7. Croft DP

    (2012) Adaptive prolonged postreproductive life span in killer whales

    Science 337:1313.

    https://doi.org/10.1126/science.1224198

    • PubMed
    • Google Scholar
33. Book

    1. Fox J
    2. Weisberg S

    (2011)

    An (r) Companion to Applied Regression (Second Edition)

    Thousand Oaks CA: Sage.

    • Google Scholar
34. 1. Frank DA
    2. Klass PE
    3. Earls F
    4. Eisenberg L

    (1996) Infants and young children in orphanages: One view from pediatrics and child psychiatry

    Pediatrics 97:569–578.

    https://doi.org/10.1542/peds.2005-2943

    • Google Scholar
35. 1. Garcia‐Belenguer S
    2. Oliver C
    3. Morméde P

    (1993) Facilitation and feedback in the Hypothalamo-Pituitary-Adrenal axis during food restriction in rats

    Journal of Neuroendocrinology 5:663–668.

    https://doi.org/10.1111/j.1365-2826.1993.tb00537.x

    • PubMed
    • Google Scholar
36. Book

    1. Gilardi K
    2. Gillespie TR
    3. Leendertz FH
    4. Macfie EJ
    5. Travis DA
    6. Whittier CA
    7. Williamson EA

    (2015)

    Best Practice Guidelines for Health Monitoring and Disease Control in Great Ape Populations, Occasional Papers of the IUCN Species Survival Commission

    IUCN SSC Primate Specialist Group.

    • Google Scholar
37. 1. Girard-Buttoz C
    2. Heistermann M
    3. Krummel S
    4. Engelhardt A

    (2009) Seasonal and social influences on fecal androgen and glucocorticoid excretion in wild male long-tailed macaques (Macaca fascicularis

    Physiology & Behavior 98:168–175.

    https://doi.org/10.1016/j.physbeh.2009.05.005

    • Google Scholar
38. 1. Goldenberg SZ
    2. Wittemyer G

    (2017) Orphaned female elephant social bonds reflect lack of access to mature adults

    Scientific Reports 7:1–7.

    https://doi.org/10.1038/s41598-017-14712-2

    • Google Scholar
39. 1. Gonzalez A
    2. Jenkins JM
    3. Steiner M
    4. Fleming AS

    (2009) The relation between early life adversity, cortisol awakening response and diurnal salivary cortisol levels in postpartum women

    Psychoneuroendocrinology 34:76–86.

    https://doi.org/10.1016/j.psyneuen.2008.08.012

    • PubMed
    • Google Scholar
40. Book

    1. Goodall J

    (1986)

    The Chimpanzees of Gombe: Patterns of Behavior

    Harvard University Press.

    • Google Scholar
41. 1. Goodwin GM
    2. Fairburn CG
    3. Keenan JC
    4. Cowen PJ

    (1988) The effects of dieting and weight loss upon the stimulation of thyrotropin (TSH) by thyrotropin-releasing hormone (TRH) and suppression of cortisol secretion by dexamethasone in men and women

    Journal of Affective Disorders 14:137–144.

    https://doi.org/10.1016/0165-0327(88)90056-0

    • PubMed
    • Google Scholar
42. 1. Grossman CJ

    (1985) Interactions between the gonadal steroids and the immune system

    Science 227:257–261.

    https://doi.org/10.1126/science.3871252

    • PubMed
    • Google Scholar
43. 1. Gunnar MR
    2. Morison SJ
    3. Chisholm K
    4. Schuder M

    (2001) Salivary cortisol levels in children adopted from Romanian orphanages

    Development and Psychopathology 13:611–628.

    https://doi.org/10.1017/s095457940100311x

    • PubMed
    • Google Scholar
44. 1. Hart J
    2. Gunnar M
    3. Cicchetti D

    (1996) Altered neuroendocrine activity in maltreated children related to symptoms of depression

    Development and Psychopathology 8:201–214.

    https://doi.org/10.1017/S0954579400007045

    • Google Scholar
45. 1. Hauser B
    2. Deschner T
    3. Boesch C

    (2008) Development of a liquid chromatography–tandem mass spectrometry method for the determination of 23 endogenous steroids in small quantities of primate urine

    Journal of Chromatography B 862:100–112.

    https://doi.org/10.1016/j.jchromb.2007.11.009

    • Google Scholar
46. 1. Heim C
    2. Nemeroff CB

    (2001) The role of childhood trauma in the neurobiology of mood and anxiety disorders: preclinical and clinical studies

    Biological Psychiatry 49:1023–1039.

    https://doi.org/10.1016/s0006-3223(01)01157-x

    • PubMed
    • Google Scholar
47. 1. Hobaiter C
    2. Schel AM
    3. Langergraber K
    4. Zuberbühler K

    (2014) Adoption’ by maternal siblings in wild chimpanzees

    PLOS ONE 9:e103777.

    https://doi.org/10.1371/journal.pone.0103777

    • PubMed
    • Google Scholar
48. 1. Karlamangla AS
    2. Merkin SS
    3. Almeida DM
    4. Friedman EM
    5. Mogle JA
    6. Seeman TE

    (2019) Early-life adversity and dysregulation of adult diurnal cortisol rhythm

    The Journals of Gerontology: Series B 74:160–169.

    https://doi.org/10.1093/geronb/gby097

    • Google Scholar
49. 1. Kaufman J

    (1991) Depressive disorders in maltreated children

    Journal of the American Academy of Child & Adolescent Psychiatry 30:257–265.

    https://doi.org/10.1097/00004583-199103000-00014

    • Google Scholar
50. 1. Kawai T
    2. Kuwano Y
    3. Masuda K
    4. Fujita K
    5. Tanaka H
    6. Nishikawa T
    7. Rokutan K
    8. Nishida K

    (2017) Adverse parenting is associated with blunted salivary cortisol awakening response and altered expression of glucocorticoid receptor β and β2-adrenergic receptor mrnas in leukocytes in japanese medical students

    Stress 20:159–166.

    https://doi.org/10.1080/10253890.2017.1297415

    • PubMed
    • Google Scholar
51. 1. Kudielka BM
    2. Bellingrath S
    3. Hellhammer DH

    (2006)

    Cortisol in burnout and vital exhaustion: an overview

    Giornale Italiano Di Medicina Del Lavoro Ed Ergonomia 28:34–42.

    • Google Scholar
52. 1. Leeuwen EJC
    2. Mulenga IC
    3. Chidester DL

    (2014) Early social deprivation negatively affects social skill acquisition in chimpanzees (Tropan

    Animal Cognition 17:407–414.

    https://doi.org/10.1007/s10071-013-0672-5

    • Google Scholar
53. 1. Lemoine NP

    (2019) Moving beyond noninformative priors: why and how to choose weakly informative priors in Bayesian analyses

    Oikos 128:912–928.

    https://doi.org/10.1111/oik.05985

    • Google Scholar
54. 1. Liu D
    2. Diorio J
    3. Tannenbaum B
    4. Caldji C
    5. Francis D
    6. Freedman A
    7. Sharma S
    8. Pearson D
    9. Plotsky PM
    10. Meaney MJ

    (1997) Maternal care, hippocampal glucocorticoid receptors, and Hypothalamic-Pituitary-Adrenal responses to stress

    Science 277:1659–1662.

    https://doi.org/10.1126/science.277.5332.1659

    • PubMed
    • Google Scholar
55. 1. Lukas D
    2. Clutton-Brock T

    (2012) Cooperative breeding and monogamy in mammalian societies

    Proceedings. Biological Sciences 279:2151–2156.

    https://doi.org/10.1098/rspb.2011.2468

    • PubMed
    • Google Scholar
56. Book

    1. Maestripieri D
    2. Mateo JM

    (2009) Maternal Effects in Mammals

    University of Chicago Press.

    https://doi.org/10.7208/chicago/9780226501222.001.0001

    • Google Scholar
57. 1. McLachlan K
    2. Rasmussen C
    3. Oberlander TF
    4. Loock C
    5. Pei J
    6. Andrew G
    7. Reynolds J
    8. Weinberg J

    (2016) Dysregulation of the cortisol diurnal rhythm following prenatal alcohol exposure and early life adversity

    Alcohol 53:9–18.

    https://doi.org/10.1016/j.alcohol.2016.03.003

    • PubMed
    • Google Scholar
58. 1. Meinlschmidt G
    2. Heim C

    (2005) Decreased cortisol awakening response after early loss experience

    Psychoneuroendocrinology 30:568–576.

    https://doi.org/10.1016/j.psyneuen.2005.01.006

    • PubMed
    • Google Scholar
59. 1. Mielke A
    2. Samuni L
    3. Preis A
    4. Gogarten JF
    5. Crockford C
    6. Wittig RM

    (2018) Bystanders intervene to impede grooming in Western chimpanzees and sooty mangabeys

    Royal Society Open Science 4:171296.

    https://doi.org/10.1098/rsos.171296

    • Google Scholar
60. 1. Miller RC
    2. Brindle E
    3. Holman DJ
    4. Shofer J
    5. Klein NA
    6. Soules MR
    7. O’Connor KA

    (2004) Comparison of specific gravity and creatinine for normalizing urinary reproductive hormone concentrations

    Clinical Chemistry 50:924–932.

    https://doi.org/10.1373/clinchem.2004.032292

    • PubMed
    • Google Scholar
61. 1. Miller GE
    2. Chen E
    3. Zhou ES

    (2007) If it goes up, must it come down? Chronic stress and the hypothalamic-pituitary-adrenocortical axis in humans

    Psychological Bulletin 133:25–45.

    https://doi.org/10.1037/0033-2909.133.1.25

    • PubMed
    • Google Scholar
62. 1. Miller GE
    2. Chen E
    3. Fok AK
    4. Walker H
    5. Lim A
    6. Nicholls EF
    7. Cole S
    8. Kobor MS

    (2009) Low early-life social class leaves a biological residue manifested by decreased glucocorticoid and increased proinflammatory signaling

    PNAS 106:14716–14721.

    https://doi.org/10.1073/pnas.0902971106

    • PubMed
    • Google Scholar
63. 1. Miller GE
    2. Chen E
    3. Parker KJ

    (2011) Psychological stress in childhood and susceptibility to the chronic diseases of aging: Moving toward a model of behavioral and biological mechanisms

    Psychological Bulletin 137:959–997.

    https://doi.org/10.1037/a0024768

    • PubMed
    • Google Scholar
64. 1. Muller MN
    2. Wrangham RW

    (2004) Dominance, cortisol and stress in wild chimpanzees (Pan troglodytes schweinfurthii)

    Behavioral Ecology and Sociobiology 55:332–340.

    https://doi.org/10.1007/s00265-003-0713-1

    • Google Scholar
65. 1. Murtagh R
    2. Behringer V
    3. Deschner T

    (2013)

    LC-MS as a method for non-invasive measurement of steroid hormones and their metabolites in urine and faeces of animals

    Veterinary Medicine Austria 100:247–254.

    • Google Scholar
66. 1. Nakamura M
    2. Hayaki H
    3. Hosaka K
    4. Itoh N
    5. Zamma K

    (2014) Brief communication: Orphaned male chimpanzees die young even after weaning

    American Journal of Physical Anthropology 153:139–143.

    https://doi.org/10.1002/ajpa.22411

    • PubMed
    • Google Scholar
67. 1. Neumann C
    2. Duboscq J
    3. Dubuc C
    4. Ginting A
    5. Irwan AM
    6. Agil M
    7. Widdig A
    8. Engelhardt A

    (2011) Assessing dominance hierarchies: validation and advantages of progressive evaluation with Elo-rating

    Animal Behaviour 82:911–921.

    https://doi.org/10.1016/j.anbehav.2011.07.016

    • Google Scholar
68. 1. Nicolson NA

    (2004) Childhood parental loss and cortisol levels in adult men

    Psychoneuroendocrinology 29:1012–1018.

    https://doi.org/10.1016/j.psyneuen.2003.09.005

    • PubMed
    • Google Scholar
69. 1. Power C
    2. Hertzman C

    (1997) Social and biological pathways linking early life and adult disease

    British Medical Bulletin 53:210–221.

    https://doi.org/10.1093/oxfordjournals.bmb.a011601

    • PubMed
    • Google Scholar
70. 1. Preis A
    2. Samuni L
    3. Deschner T
    4. Crockford C
    5. Wittig RM

    (2019) Urinary cortisol, aggression, dominance and competition in wild, West African male chimpanzees

    Frontiers in Ecology and Evolution 7:107.

    https://doi.org/10.3389/fevo.2019.00107

    • Google Scholar
71. 1. Pruessner JC
    2. Hellhammer DH
    3. Kirschbaum C

    (1999) Burnout, perceived stress, and cortisol responses to awakening

    Psychosomatic Medicine 61:197.

    https://doi.org/10.1097/00006842-199903000-00012

    • PubMed
    • Google Scholar
72. Book

    1. Pusey AE

    (1979)

    Intercommunity transfer of chimpanzees in Gombe National Park

    In: Hamburg D, McCown E, editors. The Great Apes. Cummings. pp. 465–479.

    • Google Scholar
73. 1. Pusey AE

    (1983) Mother-offspring relationships in chimpanzees after weaning

    Animal Behaviour 31:363–377.

    https://doi.org/10.1016/S0003-3472(83)80055-4

    • Google Scholar
74. 1. Rasmussen LJH
    2. Moffitt TE
    3. Eugen-Olsen J
    4. Belsky DW
    5. Danese A
    6. Harrington H
    7. Houts RM
    8. Poulton R
    9. Sugden K
    10. Williams B
    11. Caspi A

    (2019) Cumulative childhood risk is associated with a new measure of chronic inflammation in adulthood

    Journal of Child Psychology and Psychiatry 60:199–208.

    https://doi.org/10.1111/jcpp.12928

    • PubMed
    • Google Scholar
75. 1. Reddy RB
    2. Sandel AA

    (2020) Social relationships between chimpanzee sons and mothers endure but change during adolescence and adulthood

    Behavioral Ecology and Sociobiology 74:150.

    https://doi.org/10.1007/s00265-020-02937-7

    • Google Scholar
76. 1. Romero LM
    2. Dickens MJ
    3. Cyr NE

    (2009) The reactive scope model - A new model integrating homeostasis, allostasis, and stress

    Hormones and Behavior 55:375–389.

    https://doi.org/10.1016/j.yhbeh.2008.12.009

    • PubMed
    • Google Scholar
77. 1. Rosenbaum S
    2. Zeng S
    3. Campos FA
    4. Gesquiere LR
    5. Altmann J
    6. Alberts SC
    7. Li F
    8. Archie EA

    (2020) Social bonds do not mediate the relationship between early adversity and adult glucocorticoids in wild baboons

    PNAS 117:20052–20062.

    https://doi.org/10.1073/pnas.2004524117

    • Google Scholar
78. Book

    1. Samuni L
    2. Wittig RM
    3. Crockford C

    (2019a) Adoption in the Taï chimpanzees: Costs, benefits and strong social relationships

    In: Boesch RM, Wittig RM, Crockford C, editors. The Chimpanzees of the Taï Forest: 40 Years of Research. CA, UK: Cambridge University Press. pp. 141–158.

    https://doi.org/10.1017/9781108674218.011

    • Google Scholar
79. 1. Samuni L
    2. Preis A
    3. Deschner T
    4. Wittig RM
    5. Crockford C

    (2019b) Cortisol and oxytocin show independent activity during chimpanzee intergroup conflict

    Psychoneuroendocrinology 104:165–173.

    https://doi.org/10.1016/j.psyneuen.2019.02.007

    • PubMed
    • Google Scholar
80. 1. Samuni L
    2. Tkaczynski P
    3. Deschner T
    4. Löhrrich T
    5. Wittig RM
    6. Crockford C

    (2020) Maternal effects on offspring growth indicate post-weaning juvenile dependence in chimpanzees (Pan troglodytes verus

    Frontiers in Zoology 17:1.

    https://doi.org/10.1186/s12983-019-0343-8

    • PubMed
    • Google Scholar
81. 1. Sánchez MM
    2. Ladd CO
    3. Plotsky PM

    (2001) Early adverse experience as a developmental risk factor for later psychopathology: Evidence from rodent and primate models

    Development and Psychopathology 13:419–449.

    https://doi.org/10.1017/s0954579401003029

    • PubMed
    • Google Scholar
82. Book

    1. Sapolsky RM

    (2002)

    Endocrinology of the stress response

    In: Becker JB, Breedlove SM, Crews D, McCarthy MM, editors. Behavioural Endocrinology (2nd edition). Cambridge, Massachusets: MIT Press. pp. 409–450.

    • Google Scholar
83. 1. Sephton SE
    2. Sapolsky RM
    3. Kraemer HC
    4. Spiegel D

    (2000) Diurnal cortisol rhythm as a predictor of breast cancer survival

    JNCI: Journal of the National Cancer Institute 92:994–1000.

    https://doi.org/10.1093/jnci/92.12.994

    • Google Scholar
84. 1. Setchell JM
    2. Smith T
    3. Wickings EJ
    4. Knapp LA

    (2010) Stress, social behaviour, and secondary sexual traits in a male primate

    Hormones and Behavior 58:720–728.

    https://doi.org/10.1016/j.yhbeh.2010.07.004

    • PubMed
    • Google Scholar
85. 1. Slavich GM
    2. Cole SW

    (2013) The Emerging Field of Human Social Genomics

    Clinical Psychological Science 1:331–348.

    https://doi.org/10.1177/2167702613478594

    • PubMed
    • Google Scholar
86. 1. Sonnweber R
    2. Araya-Ajoy YG
    3. Behringer V
    4. Deschner T
    5. Tkaczynski P
    6. Fedurek P
    7. Preis A
    8. Samuni L
    9. Zommers Z
    10. Gomes C
    11. Zuberbühler K
    12. Wittig RM
    13. Crockford C

    (2018) Circadian rhythms of urinary cortisol levels vary between individuals in wild male chimpanzees: A reaction norm approach

    Frontiers in Ecology and Evolution 6:17.

    https://doi.org/10.3389/fevo.2018.00085

    • Google Scholar
87. 1. Stanton MA
    2. Lonsdorf E
    3. Murray CM
    4. Pusey AE

    (2020) Consequences of maternal loss before and after weaning in male and female wild chimpanzees

    Behavioral Ecology and Sociobiology 74:22.

    https://doi.org/10.1007/s00265-020-2804-7

    • Google Scholar
88. 1. Strauss ED
    2. Shizuka D
    3. Holekamp KE

    (2020) Juvenile rank acquisition is associated with fitness independent of adult rank

    Proceedings of the Royal Society B: Biological Sciences 287:20192969.

    https://doi.org/10.1098/rspb.2019.2969

    • Google Scholar
89. 1. Surbeck M
    2. Boesch C
    3. Crockford C
    4. Thompson ME
    5. Furuichi T
    6. Fruth B
    7. Hohmann G
    8. Ishizuka S
    9. Machanda Z
    10. Muller MN
    11. Pusey A
    12. Sakamaki T
    13. Tokuyama N
    14. Walker K
    15. Wrangham R
    16. Wroblewski E
    17. Zuberbühler K
    18. Vigilant L
    19. Langergraber K

    (2019) Males with a mother living in their group have higher paternity success in bonobos but not chimpanzees

    Current Biology 29:R354.

    https://doi.org/10.1016/j.cub.2019.03.040

    • Google Scholar
90. 1. Tarullo AR
    2. Gunnar MR

    (2006) Child maltreatment and the developing HPA axis

    Hormones and Behavior 50:632–639.

    https://doi.org/10.1016/j.yhbeh.2006.06.010

    • PubMed
    • Google Scholar
91. 1. Taylor SE
    2. Way BM
    3. Seeman TE

    (2011) Early adversity and adult health outcomes

    Development and Psychopathology 23:939–954.

    https://doi.org/10.1017/S0954579411000411

    • PubMed
    • Google Scholar
92. 1. Thomas BW
    2. Champoux M
    3. Suomi SJ
    4. Gunnar MR

    (1995) Salivary cortisol in nursery-reared rhesus monkeys: Reactivity to peer interactions and altered circadian activity

    Developmental Psychobiology 28:257–267.

    https://doi.org/10.1002/dev.420280502

    • Google Scholar
93. 1. Tkaczynski PJ
    2. Behringer V
    3. Ackermann CY
    4. Fedurek P
    5. Fruth B
    6. Girard-Buttoz C
    7. Hobaiter C
    8. Lee SM
    9. Löhrich T
    10. Preis A
    11. Samuni L
    12. Zommers Z
    13. Zuberbühler K
    14. Deschner T
    15. Wittig RM
    16. Hohmann G
    17. Crockford C

    (2020) Patterns of urinary cortisol levels during ontogeny appear population specific rather than species specific in wild chimpanzees and bonobos

    Journal of Human Evolution 147:32866765.

    https://doi.org/10.1016/j.jhevol.2020.102869

    • Google Scholar
94. 1. Tung J
    2. Archie EA
    3. Altmann J
    4. Alberts SC

    (2016) Cumulative early life adversity predicts longevity in wild baboons

    Nature Communications 7:11181.

    https://doi.org/10.1038/ncomms11181

    • PubMed
    • Google Scholar
95. 1. Uehara S
    2. Nyundo R

    (1983) One observed case of temporary adoption of an infant by unrelated nulliparous females among wild chimpanzees in the Mahale Mountains, Tanzania

    Primates; Journal of Primatology 24:456–466.

    https://doi.org/10.1007/BF02381679

    • Google Scholar
96. 1. Walker KK
    2. Walker CS
    3. Goodall J
    4. Pusey AE

    (2018) Maturation is prolonged and variable in female chimpanzees

    Journal of Human Evolution 114:131–140.

    https://doi.org/10.1016/j.jhevol.2017.10.010

    • PubMed
    • Google Scholar
97. 1. Watts HE
    2. Tanner JB
    3. Lundrigan BL
    4. Holekamp KE

    (2009) Post-weaning maternal effects and the evolution of female dominance in the spotted hyena

    Proceedings. Biological Sciences 276:2291–2298.

    https://doi.org/10.1098/rspb.2009.0268

    • PubMed
    • Google Scholar
98. 1. Weaver ICG
    2. Cervoni N
    3. Champagne FA
    4. D’Alessio AC
    5. Sharma S
    6. Seckl JR
    7. Dymov S
    8. Szyf M
    9. Meaney MJ

    (2004) Epigenetic programming by maternal behavior

    Nature Neuroscience 7:847–854.

    https://doi.org/10.1038/nn1276

    • PubMed
    • Google Scholar
99. 1. Weaver ICG
    2. D’Alessio AC
    3. Brown SE
    4. Hellstrom IC
    5. Dymov S
    6. Sharma S
    7. Szyf M
    8. Meaney MJ

    (2007) The transcription factor nerve growth factor-inducible protein a mediates epigenetic programming: Altering epigenetic marks by immediate-early genes

    The Journal of Neuroscience 27:1756–1768.

    https://doi.org/10.1523/JNEUROSCI.4164-06.2007

    • PubMed
    • Google Scholar
100. 1. Wessling E
     2. Deschner T
     3. Mundry R
     4. Pruetz J
     5. Wittig R
     6. Kühl H

     (2018) Seasonal variation in physiology challenges the notion of chimpanzees (Pan troglodytes verus) as a forest-adapted species

     Frontiers in Ecology and Evolution 6:60.

     https://doi.org/10.3389/fevo.2018.00060

     • Google Scholar
101. 1. Wittig RM
     2. Crockford C
     3. Weltring A
     4. Deschner T
     5. Zuberbuehler K

     (2015) Single aggressive interactions increase urinary glucocorticoid levels in wild male chimpanzees

     PLOS ONE 10:e0118695.

     https://doi.org/10.1371/journal.pone.0118695

     • Google Scholar
102. 1. Wittig RM
     2. Crockford C
     3. Weltring A
     4. Langergraber KE
     5. Deschner T
     6. Zuberbühler K

     (2016) Social support reduces stress hormone levels in wild chimpanzees across stressful events and everyday affiliations

     Nature Communications 7:1–8.

     https://doi.org/10.1038/ncomms13361

     • PubMed
     • Google Scholar
103. Book

     1. Wittig RM

     (2018) Tai chimpanzees

     In: Vonk J, Shackelford TK, editors. Encyclopedia of Animal Cognition and Behavior. Cham: Springer International Publishing. pp. 1403–1404.

     https://doi.org/10.1007/978-3-319-47829-6

     • Google Scholar
104. Book

     1. Wittig RM
     2. Boesch C

     (2019a) Demography and life history of five chimpanzee communities in Taï National Park

     In: Boesch C, Wittig RM, Deschner VT, editors. The Chimpanzees of the Taï Forest: 40 Years of Research. Cambridge, UK: Cambridge University Press. pp. 125–140.

     https://doi.org/10.1017/9781108674218.010

     • Google Scholar
105. Book

     1. Wittig RM
     2. Boesch C

     (2019b) Observation protocol and long-term data collection in Tai

     In: Boesch C, editors. The Chimpanzees of the Taï Forest: 40 Years of Research. Cambridge University Press. pp. 44–57.

     https://doi.org/10.1017/9781108674218

     • Google Scholar
106. 1. Wobber V
     2. Hare B

     (2011) Psychological Health of Orphan Bonobos and Chimpanzees in African Sanctuaries

     PLOS ONE 6:e17147.

     https://doi.org/10.1371/journal.pone.0017147

     • PubMed
     • Google Scholar
107. 1. Wroblewski EE

     (2008) An unusual incident of adoption in a wild chimpanzee (Pan troglodytes) population at Gombe National Park

     American Journal of Primatology 70:995–998.

     https://doi.org/10.1002/ajp.20582

     • PubMed
     • Google Scholar
108. 1. Young EA
     2. Haskett RF
     3. Grunhaus L
     4. Pande A
     5. Weinberg VM
     6. Watson SJ
     7. Akil H

     (1994) Increased Evening Activation of the Hypothalamic-Pituitary-Adrenal Axis in Depressed Patients

     Archives of General Psychiatry 51:701–707.

     https://doi.org/10.1001/archpsyc.1994.03950090033005

     • PubMed
     • Google Scholar
109. 1. Young C
     2. Majolo B
     3. Heistermann M
     4. Schülke O
     5. Ostner J

     (2014) Responses to social and environmental stress are attenuated by strong male bonds in wild macaques

     PNAS 111:18195–18200.

     https://doi.org/10.1073/pnas.1411450111

     • Google Scholar
110. 1. Young C
     2. Bonnell TR
     3. Brown LR
     4. Dostie MJ
     5. Ganswindt A
     6. Kienzle S
     7. McFarland R
     8. Henzi SP
     9. Barrett L

     (2019) Climate induced stress and mortality in vervet monkeys

     Royal Society Open Science 6:191078.

     https://doi.org/10.1098/rsos.191078

     • PubMed
     • Google Scholar
111. 1. Zhang TY
     2. Labonté B
     3. Wen XL
     4. Turecki G
     5. Meaney MJ

     (2013) Epigenetic mechanisms for the early environmental regulation of hippocampal glucocorticoid receptor gene expression in rodents and humans

     Neuropsychopharmacology 38:111–123.

     https://doi.org/10.1038/npp.2012.149

     • PubMed
     • Google Scholar
112. 1. Zipple MN
     2. Altmann J
     3. Campos FA
     4. Cords M
     5. Fedigan LM
     6. Lawler RR
     7. Lonsdorf E
     8. Perry S
     9. Pusey AE
     10. Stoinski TS
     11. Strier KB
     12. Alberts SC

     (2021) Maternal death and offspring fitness in multiple wild primates

     PNAS 118:e2015317118.

     https://doi.org/10.1073/pnas.2015317118

     • Google Scholar

Acceptance summary:

This study examined the relationship between maternal loss and cortisol excretion – a proxy for stress, among wild chimpanzees in the Tai Forest of Cote d'Ivoire. It tests whether and how the effect of early maternal loss is reflected in individual cortisol levels. The authors found that diurnal cortisol slopes across the day differed between immature chimps who lost their mothers early and those who did not, but that this difference is not visible later in life, suggesting a short-term alteration of the hypothalamic-pituitary-adrenal axis – a finding consistent with the adaptive calibration model rather than the biological embedding model that proposes a long-term alteration of the hypothalamic-adrenal-axis due to early maternal loss. The study is one of the rare empirical tests of the biological embedding and the adaptive calibration models on wild long-lived mammals and opens opportunities for further investigation of the impact of early life experience on the resilience of non-human primates which are especially threatened by hunting and habitat destruction. Future studies should, however, address the possibility that the lack of a later life association between maternal loss and cortisol levels may be due to selective early mortality of individuals with high cortisol levels. Social and environmental factors that may buffer the effect of early maternal loss should also be considered.

Decision letter after peer review:

Thank you for submitting your article "Early maternal loss affects diurnal cortisol slopes in immature but not mature wild chimpanzees" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and George Perry as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

This paper tests the biological embedding model by asking whether and how early maternal loss effects cortisol levels and diurnal cortisol slopes among wild chimpanzees at the Tai Forest, Cote d'Ivoire. The results suggest that maternal loss alters the HPA stress axis in wild chimpanzees, but these effects are not visible later in life. Authors suggest that the lack of a later life association between maternal loss and cortisol levels may be due to selective early mortality of individuals with high cortisol levels but did not provide any survival or behavioural data to show that orphans and non-orphans differ in any fitness-related traits other than cortisol. Furthermore, the association between cortisol and the HPA axis is in the opposite direction to that observed in humans and there seems to be no significant increase in cortisol in orphans compared to non-orphans. Overall, the study is the result of extensive fieldwork and the number of samples collected is impressive. The subject is very interesting, and we generally agree that with an extensive reworking of the entire framework and analyses, it could be a good fit for eLife.

The analyses will benefit greatly if the authors use effect sizes and confidence intervals for inferences instead of p-values. This may solve the significance threshold issues. Moreover, the reliance on p-values seem to limit the value of the data. For example, authors suggest that results from model 1 should be treated with caution because the full model is not significantly different from the null model, but by relying on it as the key finding of the study without exploring effect sizes, it does not seem that they did exercise sufficient caution.

Please find more specific comments below:

Essential revisions:

1. Present results as effect sizes with confidence intervals and make inferences along the line of the percentage (or ng/ml) by which orphans differ from non-orphans and over time. This effect sizes can be more easily compared with results from human studies on cortisol. Please communicate findings more clearly and discuss exactly why the pattern in this Chimp population may be different from that in humans. Pay attention to the following comment from reviewers:

a. Despite acknowledging that the "significance of these predictors should be interpreted with caution" because model 1a did not reach significance, the authors make very strong claims about the results in the discussion- and also feature the finding of that model in the title of the paper. That seems problematic to me- especially because the insignificant model results (more intense diurnal slopes among immature orphans) diverge from the expectations set forth by other works in humans and non-humans. The finding that this is to do with higher-than-expected morning cortisol is puzzling given that evening levels are generally considered more responsive or plastic. However, this could also be an artefact of fitting the models without the third-order term for time.

2. The lack of significance could be due to insufficient sampling or a true lack of predictive power. Reviewers provide specific suggestions on how to reanalyse the data given the difficulty of collecting additional samples currently.

a. Model 1B and figure 2 demonstrate that the cortisol response to maternal loss declines over time and that after 2 years it is no longer detectably different from non-orphans. The authors do not account for age since maternal loss in model 1A. If a considerable proportion of samples were from orphans that lost their mothers more than 2 years ago, this would reduce the likelihood of detecting a significant difference between orphans and non-orphans and potentially explain the lack of significance in the overall model. Crucially I think if model 1a was adjusted to separate out recent orphans from those that lost their mothers less recently this could enable the authors to better back up their claims at least in relation to changes in overall cortisol levels.

b. The truth is that cortisol data are very messy and even though 300+ samples from 50 individuals might seem like a lot, it might turn out that it isn't enough to detect a signal. At other sites, cortisol levels and diurnal slopes shift with age- and this is true for humans as well. However, the slope should be more susceptible than more average levels so the authors might be able to make stronger conclusions based on average or time-corrected cortisol rather than focusing so much on a slope. Either way- though improved modelling to access slope or by setting slope aside and focusing on average cortisol- the data here certainly have a path to publication

c. My principal concerns with this paper, as written, revolve around the methods/results. First and foremost, I am not convinced that the authors have a sufficient sample size to evaluate the predictions/hypotheses outlined in the introduction. While 849 urine samples are a large number, and again, their efforts here should be commended, the sample spread is quite thin once it is spliced up into appropriate categories, especially considering how many samples were collected per individual year, on average. As the authors indicate throughout and especially when describing their modelling approach, cortisol is inherently a very noisy hormone impacted by a myriad of factors- including age in at least one other densely sampled chimpanzee community. I am also surprised that time of day was modelled quadratically. It is my understanding that humans, other populations of chimpanzees, and other mammals follow a sigmoidal curve which should be modelled with a third-order term as well. For these reasons, it is difficult to tell whether model 1A is not significant because of the insufficient sample or a true lack of predictive power. Additionally, I am concerned that the paper seems to focus so much on the results from a single model term in a model that did not reach significance.

d. It would be useful to see some of the raw data- especially a plot showing cortisol values across the time of day. Regarding that third-order term- it isn't out of the question that T + T^2 would be sufficient, however, I do believe it's important to rule out that including T^3 does a better job.

e. L449 As slopes are calculated for each individual each year, what is the mean number of samples per individual per year? 3 minimum seems very small for calculating a slope but if the average is considerably higher, then perhaps it is not an issue.

f. Please include more information about model results, sample size, etc. in figure captions. When denoting sample sizes, it would be useful to know both the number of urine samples and the number of unique individuals that contributed to the dataset.

g. I would like to see more transparency about sample size, concerning the number of samples, from x individuals, in y study groups.

3. It would be great if authors can provide additional data that show possible differences in survival and/or behaviour between orphaned and non-orphaned immatures and further incorporate the reason for the maternal loss and the age at which mothers died into the analyses. An aged mother dying could have different effects compared to a prime-age mother dying. It is hugely surprising that behavioural data is completely excluded from the study after claiming to have followed individual animals for 6 or 12 hours per day. The manuscript would be quite a bit different from what we have reviewed here but tying the cortisol data to some concrete behavioural and survival observations would help contextualize the results. See specific comment from a reviewer below:

a. L294 briefly mentions survivorship bias. I would like to see a more thorough discussion of this. Did any individuals that were orphaned subsequently die? How were these handled? Are there enough to compare them to those that survived?

b. I wondered throughout the manuscript whether and how post-weaning survival could be included more directly to bring clarity to the role of cortisol/HPA regulation in fitness. I am not exactly sure what to suggest- but I think that directly discussing how differences in survival/reproduction may be related to HPA functioning in this population of chimpanzees, even if they are limited to qualitative comparisons, could improve the manuscript quite a lot.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Early maternal loss leads to short- but not long-term effects on diurnal cortisol slopes in wild chimpanzees" for further consideration by eLife. Your revised article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and George Perry as the Senior Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Your study examines the relationship between maternal loss and cortisol excretion among wild chimpanzees. It tests ideas from multiple fields to determine whether and how maternal loss is reflected in the alteration of the HPA axis. Three reviewers commented on your initial submission and we requested a revision of the manuscript.

You have clearly put a great deal of additional work into this revision and the manuscript is much improved from the prior submission. We find your responses to our previous concerns satisfying.

There are, however, a few areas that would still require work/clarification including some typos that should be fixed before we can make a decision on your submission. Note that this does not amount to a partial acceptance of your manuscript.

Introduction: We greatly appreciate the care that you have taken in integrating the suggestions of all reviewers and recrafting the introduction. This section does a wonderful job of setting up the study and the edits you made make for a very impactful series of arguments.

Methods/Results: There are still some spots that are difficult to follow in the results, and this section might require a bit more work than others. Consider the specific comments below:

L266-294: This section describing the models is still a bit confusing, especially the section about predictor variables and how they differ between models 1a/b, 2a/b. Later in the methods, the authors mention that they ran models for each sex separately, but that is not mentioned here.

I was slightly confused by this list of predictors at first thinking all predictors were used in each model. Is it possible to make it a little clearer that this is not the case? Perhaps something like "Each model contained one or more of the following test predictor variables" in line 283/284. In line 287 I think you mean model 1a rather than model 1b – this is probably the root of my confusion as it makes it seem like both years since maternal loss and orphan status as a categorical variable with three levels were included in the same model.

I don't have much specific advice to solve the problem, other than to say that it was difficult to follow each thread. Perhaps if each section was more simply dedicated to each model (like the longer methods section) rather than going back and forth between the things that were the same across models versus different? It's a lot to keep track of, so redundancy might be better for ease of interpretation in this case??

L315-16: I am not very familiar with Bayesian approaches and this section is unclear to me. In frequentist statistics effects sizes and variance explained are not the same things – could the authors clarify what they are reporting here and what it means?

L329-331: Did the authors directly test for categorical differences in morning cortisol or evening cortisol or are all of the comparisons here based on slope?

L362-365: The wording here "in particular in the early morning and the afternoon" is confusing given that the take-away is that cortisol had an upward slope and was, therefore, higher in the afternoon compared to the morning.

L374-389: The authors jump back and forth between describing life-history-based age categories (under 5 y.o. = infants, 5-8 = juveniles, 8-12 = adolescents) and referring to specific ages ("who lost their mother at 4 y.o."). That makes it difficult to parse whether and where they are using continuous versus categorical age predictions. It is especially difficult because the text describes things as one way or both ways, but the figures describe something firmly in the middle. Please revise these sections to make them clearer.

Discussion: L444-447: Higher morning and higher evening cortisol does not necessarily mean anything about slope (i.e. the am and pm increases could be equally leading to similar slopes, but higher average cortisol). I think it is important to specify exactly what the authors mean here- are orphans experiencing higher am, higher pm, and different slopes? If so how are the slopes different in layman's terms and which point is contributing to that difference in slope? It looks like the answer comes later (lines 457-58), but it still isn't so clear throughout the paragraph which parts of the results correspond to what theoretical models/predictions, and how. For instance, in L447-448: could the authors be more specific about how this finding aligns with the ACM?

L471-476: How often does food sharing happen with mothers and weaned offspring? It seems like the authors are asserting that calories from food sharing make up a significant portion of the juvenile chimpanzee diet. Is this the case? If so, that would seem different from other sites.

L532: One thing to be careful about in discussing adaptive calibration is that the model is more focused on the plasticity of the HPA axis than a change in the environmental conditions. In other words, a return to normal could reflect that the environment has adjusted-but the ACM predicts that the HPA readjusts itself during critical developmental/life history timepoints (e.g. adrenarche, puberty, pregnancy/parenthood) to account for environmental conditions. So that return to normal could be the HPA readjusting itself to essentially make what it previously considered a stressful environment led to less of a stress response kind of like making it a new normal.

Methods: Can authors add a bit more detail about the choices that they made in creating these models? This will be instructive for helping other scholars follow and match their methodology. For instance (L872-875), what is the difference between a regularizing prior and any other type of prior?

One general question: because I'm not so familiar with Bayesian LMM/GLMM, are there any guidelines or rules for limiting the number of predictor/control terms included in the models? The authors have clearly gone to great pains to control for things, so the concern would just be that including so many terms would exhaust the degrees of freedom for the number of individuals included in the study.

L772-774: If the models are fit separately for males and females, does that mean that 1a/b are four models? 1aMale, 1aFemale, 1bMale, 1bFemale? Were there any differences in results for males versus females?

L861-865: Is this a standard control for this field? It's unclear how including project as a random effect would account for things that aren't already controlled for using the other factors mentioned here: individual, year, individual_year, etc.

L888: Does this mean that all of the actual sample sizes were > 1000? Or something else? My understanding is that there were models, e.g. those with immatures only, that included fewer than 1000 samples?

Summary:

This paper tests the biological embedding model by asking whether and how early maternal loss effects cortisol levels and diurnal cortisol slopes among wild chimpanzees at the Tai Forest, Cote d'Ivoire. The results suggest that maternal loss alters the HPA stress axis in wild chimpanzees, but these effects are not visible later in life. Authors suggest that the lack of a later life association between maternal loss and cortisol levels may be due to selective early mortality of individuals with high cortisol levels but did not provide any survival or behavioural data to show that orphans and non-orphans differ in any fitness-related traits other than cortisol.

We acknowledge that a robust analysis assessing the effect of modification in cortisol profiles on survival would be, as suggested, a strong addition to our study. Limited by only 9 confirmed deaths of immatures before maturity in our sample, we conducted a descriptive analysis.

In order to link the diurnal cortisol slopes and cortisol levels to fitness, we compared the average cortisol levels (random intercepts in Model 1a) and linear diurnal cortisol slopes (random linear slope for time of the day in Model 1a) of immatures who survived until maturity (i.e. until 12 years of age) to those of individuals who did not survive until maturity. We compared the diurnal cortisol slopes and cortisol levels of immatures who survived until maturity (i.e. until 12 years of age) (N = 23) and the ones who did not (N = 9). For 18 of the 50 immatures we do not have information on whether or not they reached maturity since they either did not reach maturity yet or may have emigrated to another group before maturity. The comparison revealed that the average intercept (cortisol levels) and linear term for time of day (i.e. cortisol diurnal linear slope) of immatures who survived until maturity, whether they were orphaned or not, were very similar (mean log cortisol levels ± SE = 3.686±0.005 vs. 3.687±0.008 and mean linear diurnal cortisol slope ± SE = -0.381±0.004 vs. -0.384±0.004). This lack of difference did not arise from a lack of variation in individual random intercepts or slopes which ranged from 3.539 to 3.761 and from -0.446 to -0.307 respectively. Superficially, this descriptive analysis above supports our conclusion that effect of maternal loss on circadian cortisol patterns in adult chimpanzees indicates a recovery over time rather than a sampling bias due to the death of orphans with altered cortisol profiles.

Nonetheless, we feel that this analysis is flawed for two reasons. First, due to the stochastic nature of death in the Tai population, sudden death such as from predation or anthrax, may not be associated with cortisol patterns in the preceding months, limiting any interpretation of these descriptive results. Second, given that it is usually not known when an individual will die, sampling within a few months of death was very rare. Thus, what we capture here is a comparison of longer term patterns (using samples across several years), which as we see from Model 1a and b, can change over time, especially in orphans. Also, the sample was too small to control for potentially influential factors such as dominance and orphan status. Hence, although we have tried to follow the reviewer’s advice here, unfortunately we do not have a robust enough dataset to include such an analysis in the paper, but hope to build a stronger dataset over the next years.

In addition, we also investigated descriptively the survival of orphans and non-orphans in our sample until maturity and found that the overall likelihood to survive until maturity for immature orphans was not substantially different from the one of non-orphans (69.2% for orphans versus 73.7% for non-orphans). At face value, additional information, albeit based on only 9 deaths, support our conclusion that the absence of effect of maternal loss in adult chimpanzees indicates a recovery over time rather than a sampling bias due to the death of orphans with altered cortisol profiles. However, we would not be comfortable to publish this result based on such a small sample (for survival analyses), especially as papers from two other chimpanzee populations show a different result, that orphan status does impact on survival (Nakamura et al. 2014; Stanton et al. 2020).

Furthermore, the association between cortisol and the HPA axis is in the opposite direction to that observed in humans and there seems to be no significant increase in cortisol in orphans compared to non-orphans.

We were also surprised that the direction of the effect on diurnal cortisol slopes for the orphans was opposite to that in most human studies. We are nevertheless confident in the meaningfulness of this effect since we could confirm this result using a Bayesian framework allowing us to investigate in more depth the level of uncertainty in the effect found (see details below).

Overall, the study is the result of extensive fieldwork and the number of samples collected is impressive. The subject is very interesting, and we generally agree that with an extensive reworking of the entire framework and analyses, it could be a good fit for eLife.

We thank the editor and the reviewer for this positive evaluation of our manuscript.

The analyses will benefit greatly if the authors use effect sizes and confidence intervals for inferences instead of p-values. This may solve the significance threshold issues. Moreover, the reliance on p-values seem to limit the value of the data. For example, authors suggest that results from model 1 should be treated with caution because the full model is not significantly different from the null model, but by relying on it as the key finding of the study without exploring effect sizes, it does not seem that they did exercise sufficient caution.

We agree that the reliance on p-value only does not allow full exploration of the degree of uncertainty or consistency in the effects investigated. Therefore, we reran the entirety of our analysis using a Bayesian framework that allows for more accurate estimations of uncertainty in effect estimates when using models with a complex random effect structure, such as in our study.

In the new analysis we also included, as suggested by the reviewers, a cubic term for time of the day and a categorical variable for orphan status with three levels (orphan for less than 2 years, orphan for more than 2 years, non-orphan). This new analysis highlighted that orphans who lost their mother less than two years ago (but not other orphans), had a consistently steeper diurnal cortisol slope (on average 58% steeper) than non-orphans. It also confirmed the effect of the age at maternal loss on immature diurnal cortisol slopes with immatures that lost their mother earlier in life having upward curving slopes, characterized by higher morning and afternoon levels. Finally, the new analysis using a cubic term for time of the day revealed that the diurnal cortisol slopes of immatures were affected by the time that passed since their mother died.

We report now 95% and 90% credible intervals for the estimates as well as the percentage of the posterior distribution which supports the direction of the estimate for each test predictor in all our models to add extra information for the reader on the degree of uncertainty of each effect. Following reviewers’ suggestions, we also report effect sizes whenever possible.

Please find more specific comments below:

Essential revisions:

1. Present results as effect sizes with confidence intervals and make inferences along the line of the percentage (or ng/ml) by which orphans differ from non-orphans and over time. This effect sizes can be more easily compared with results from human studies on cortisol. Please communicate findings more clearly and discuss exactly why the pattern in this Chimp population may be different from that in humans. Pay attention to the following comment from reviewers:

We thank the reviewers for this suggestion which improves the clarity of our manuscript. We have rewritten the entire result section based on the results of our new analyses (see above) and provide more information on the degree of uncertainty and magnitude of our results by systematically reporting now marginal R² and conditional R² for each model, the 95% and 90% credible intervals for each estimate, the percentage of the posterior distribution which confirms the effect of the estimate for each test predictor as well as, whenever possible, effect sizes. In particular, as mentioned above, we found that recently orphaned immature chimpanzees (i.e. individuals who lost their mother for less than 2 years) had diurnal cortisol slopes which were on average 58% steeper than non-orphans. We have clarified the discussion on why this effect differs in direction from most findings in humans.

a. Despite acknowledging that the "significance of these predictors should be interpreted with caution" because model 1a did not reach significance, the authors make very strong claims about the results in the discussion- and also feature the finding of that model in the title of the paper. That seems problematic to me- especially because the insignificant model results (more intense diurnal slopes among immature orphans) diverge from the expectations set forth by other works in humans and non-humans. The finding that this is to do with higher-than-expected morning cortisol is puzzling given that evening levels are generally considered more responsive or plastic. However, this could also be an artefact of fitting the models without the third-order term for time.

2. The lack of significance could be due to insufficient sampling or a true lack of predictive power. Reviewers provide specific suggestions on how to reanalyse the data given the difficulty of collecting additional samples currently.

a. Model 1B and figure 2 demonstrate that the cortisol response to maternal loss declines over time and that after 2 years it is no longer detectably different from non-orphans. The authors do not account for age since maternal loss in model 1A. If a considerable proportion of samples were from orphans that lost their mothers more than 2 years ago, this would reduce the likelihood of detecting a significant difference between orphans and non-orphans and potentially explain the lack of significance in the overall model. Crucially I think if model 1a was adjusted to separate out recent orphans from those that lost their mothers less recently this could enable the authors to better back up their claims at least in relation to changes in overall cortisol levels.

We are grateful to these constructive suggestions which improved the accuracy of our analysis. Following reviewers’ comments on how to modify the structure of model 1a (comments 1a and 2a) by incorporating a cubic term for time of the day and categorizing orphans between recently and non-recently orphans we have rerun a modified version of Model 1a using a Bayesian framework. This new analysis allowed us to report more precisely on the degree of uncertainty of the results (see above). We found that recently orphaned immature chimpanzees had consistently steeper diurnal cortisol slopes than non-orphan immatures. The difference in slope was however not consistent between orphans who had lost their mother for longer than 2 years and non-orphans. Therefore, it is possible, as the reviewer pointed out, that the lack of categorization of orphans into the two categories and of the non-inclusion of a cubic term for time of the day led to the lack of significance in our previous analysis.

b. The truth is that cortisol data are very messy and even though 300+ samples from 50 individuals might seem like a lot, it might turn out that it isn't enough to detect a signal. At other sites, cortisol levels and diurnal slopes shift with age- and this is true for humans as well. However, the slope should be more susceptible than more average levels so the authors might be able to make stronger conclusions based on average or time-corrected cortisol rather than focusing so much on a slope. Either way- though improved modelling to access slope or by setting slope aside and focusing on average cortisol- the data here certainly have a path to publication

We have modified our models to include a cubic term for time of the day and thus improve our modelling of the diurnal cortisol slopes. Using a Bayesian framework, we are now able to assess better the uncertainty in our results and report on it. We would like to point out that we have 846 samples in our first model (Model 1a) from 50 individuals not only 300.

In our analysis we set a threshold at three samples per year per individual since with only 2 samples an unusually high or low value would lead to dramatic changes in the slope steepness and direction. Mathematically, three sample is the bare minimum to calculate a meaningful slope and that is why we chose this threshold. However, it is important to note that the number of samples per individual per year is well above such a threshold, therefore we are confident that the slopes are estimated with accuracy. In fact, in model 1a we have on average 9 sample per year per individual. This information has been added to the manuscript (Lines 523-526).

c. My principal concerns with this paper, as written, revolve around the methods/results. First and foremost, I am not convinced that the authors have a sufficient sample size to evaluate the predictions/hypotheses outlined in the introduction. While 849 urine samples are a large number, and again, their efforts here should be commended, the sample spread is quite thin once it is spliced up into appropriate categories, especially considering how many samples were collected per individual year, on average. As the authors indicate throughout and especially when describing their modelling approach, cortisol is inherently a very noisy hormone impacted by a myriad of factors- including age in at least one other densely sampled chimpanzee community.

We agree that 849 samples may appear rather thin as a sample size. However, since, as mentioned above, the average number of sample collected per year per individual was 9, we are confident that our dataset allows for modelling meaningful slopes. Furthermore, we now revert to a Bayesian approach which provide more accurate estimates of the effect for models with limited sample size and complex random slope structures as ours.

Please note that, while reanalysing the data we realized that one individual in the year that it was orphaned, had less than three samples collected before and after the date that he was orphaned. Since the individual_year random slope structure in Model 1a is also separating the year when the individual is orphaned and the year when he is not (even within the same year) we had to exclude samples from that individual that year. Our final sample size still comprises 50 individuals but was slightly reduced to 846 samples.

I am also surprised that time of day was modelled quadratically. It is my understanding that humans, other populations of chimpanzees, and other mammals follow a sigmoidal curve which should be modelled with a third-order term as well. For these reasons, it is difficult to tell whether model 1A is not significant because of the insufficient sample or a true lack of predictive power. Additionally, I am concerned that the paper seems to focus so much on the results from a single model term in a model that did not reach significance.

We thank the reviewer for this suggestion allowing us to model our data more accurately. Following the reviewers’ suggestion, we have now incorporated a cubic term for time of the day in all our statistical models. That addition plus the new categorical variable for orphan status in model 1a separating orphans into recently and non-recently orphaned provide clear evidence for a robust and consistent effect of being recently orphaned on diurnal cortisol slopes: recently orphaned immatures had a diurnal cortisol slope on average 58% steeper than non-orphans. Since the 95% credible interval for this effect does not overlap 0 we can discuss this result with more certainty than the result of the previous model in the first version of the manuscript.

d. It would be useful to see some of the raw data- especially a plot showing cortisol values across the time of day. Regarding that third-order term- it isn't out of the question that T + T^2 would be sufficient, however, I do believe it's important to rule out that including T^3 does a better job.

We agree that it is important to show the raw cortisol data across the time of the day. We have now added a plot (Appendix 1-figure 1) showing the cortisol value for each sample in immature chimpanzees.

Please note that Figures 1, 2 and 3 show cortisol data for each hour of the day and also allow the reader to visualize raw cortisol variation throughout the day. In these Figures we decided to plot only 1 data point for each orphan category for each hour since it allows for better visualization of the daily variation and of the CI.

e. L449 As slopes are calculated for each individual each year, what is the mean number of samples per individual per year? 3 minimum seems very small for calculating a slope but if the average is considerably higher, then perhaps it is not an issue.

We are confident that our sample size is sufficient to evaluate individual diurnal cortisol slope yearly since the average number of samples collected for each individual each year is three times our cut-off criteria of three samples for the immatures (9 sample on average per individual per year and 50% of the ID_year comprise over 7 samples) and more than six times our cutoff criteria for the adult males (19 sample on average per individual per year and 50% of the levels for the ID_year variable comprise over 10 samples). This information has been added to the manuscript (Lines 523-526).

Please note that consistent inter-individual differences in diurnal cortisol slopes can be found using a smaller sample size than ours per individual per year (Sonnweber et al. 2018) so that we are confident that our sample size is large enough to provide meaningful results.

f. Please include more information about model results, sample size, etc. in figure captions. When denoting sample sizes, it would be useful to know both the number of urine samples and the number of unique individuals that contributed to the dataset.

As suggested by the reviewer we added information on model results and sample sizes in the figure caption for each figure.

g. I would like to see more transparency about sample size, concerning the number of samples, from x individuals, in y study groups.

We added information on the number of samples and of individuals for each model and for each orphan categories in the main text throughout the method and result sections. In addition, we added extra information on sample size to Table 1 (i.e. mean ± SE number of sample per individual for each orphan category for immature and mature males for each community).

3. It would be great if authors can provide additional data that show possible differences in survival and/or behaviour between orphaned and non-orphaned immatures and further incorporate the reason for the maternal loss and the age at which mothers died into the analyses. An aged mother dying could have different effects compared to a prime-age mother dying. It is hugely surprising that behavioural data is completely excluded from the study after claiming to have followed individual animals for 6 or 12 hours per day. The manuscript would be quite a bit different from what we have reviewed here but tying the cortisol data to some concrete behavioural and survival observations would help contextualize the results. See specific comment from a reviewer below:

Incorporating information on survival and maternal health and age is very interesting, and a future goal in our research. Unfortunately, our sample in this paper does not permit a meaningful analysis of this, as only 9 of the immatures died. As described above we have conducted a descriptive analysis of the cortisol slopes and cortisol levels of immatures surviving or not until maturity in our samples. However, as explained in details above, we do not feel confident to publish such an analysis since it is not robust enough.

As for the behavioural data, unfortunately most of the samples for immatures collected for this study were collected opportunistically alongside focal observation of mature individuals. Our research group only recently began collecting behavioural samples for immatures. Therefore, we cannot causally link behaviour to hormonal patterns in the context of this paper.

a. L294 briefly mentions survivorship bias. I would like to see a more thorough discussion of this. Did any individuals that were orphaned subsequently die? How were these handled? Are there enough to compare them to those that survived?

As mentioned above there was no obvious survival bias of non-orphan as compared to orphan immatures in this sample (although we say this cautiously as there was not a sufficient sample to run a survival analysis and other chimpanzee sites do show that orphans lose out on survival, Nakamura et al. 2014; Stanton et al. 2020), nor an obvious difference in diurnal cortisol slopes or average cortisol levels between immatures that survived until maturity and immatures that did not. As mentioned above we did not include this information into the manuscript since it may be misleading for the reader.

Orphans that died very soon after their mother died are not included in the sample since we were unable to collect at least 3 samples from them following maternal loss, and thus unable to determine their circadian cortisol pattern. That is probably why the youngest orphan in our sample was 4.1 years of age. We are aware that this may have constitute a bias and discuss this issue in the manuscript: specifically, we address that there may be a sensitive window for mammals during immaturity in which long-term permanent modification of the HPA axis functioning occurs (Lines 461-471).

b. I wondered throughout the manuscript whether and how post-weaning survival could be included more directly to bring clarity to the role of cortisol/HPA regulation in fitness. I am not exactly sure what to suggest- but I think that directly discussing how differences in survival/reproduction may be related to HPA functioning in this population of chimpanzees, even if they are limited to qualitative comparisons, could improve the manuscript quite a lot.

A recent publication of our working group on the same population of chimpanzees found that individuals that lost their mother before reaching maturity suffer some fitness cost in the form of lower reproductive success as an adult (Crockford et al. 2020). While our analysis suggests that loss of fitness in these orphans is probably not directly linked to HPA regulation, the finding of our study may provide some possible mechanism through which individuals are affected in their growth (Samuni et al. 2020) via short-term alteration to HPA axis functioning which may lead to later loss in reproductive success. As for survival, as explained above, there was no indication that orphans survived less than non-orphans nor that the individuals with altered diurnal cortisol slopes or higher cortisol levels where less likely to survive until maturity. Yet this is based on a descriptive investigation of the data and a proper analysis needs to be conducted with a larger sample.

References:

Crockford, C., Samuni, L., Vigilant, L. and Wittig, R. M. 2020. Postweaning maternal care increases male chimpanzee reproductive success. Science Advances, 6, eaaz5746.

Nakamura, M., Hayaki, H., Hosaka, K., Itoh, N. and Zamma, K. 2014. Brief communication: Orphaned male chimpanzees die young even after weaning. American Journal of Physical Anthropology, 153, 139–143.

Samuni, L., Tkaczynski, P., Deschner, T., Löhrrich, T., Wittig, R. M. and Crockford, C. 2020. Maternal effects on offspring growth indicate post-weaning juvenile dependence in chimpanzees (Pan troglodytes verus). Frontiers in Zoology, 17, 1.

Sonnweber, R., Araya-Ajoy, Y. G., Behringer, V., Deschner, T., Tkaczynski, P., Fedurek, P., Preis, A., Samuni, L., Zommers, Z., Gomes, C., Zuberbühler, K., Wittig, R. M. and Crockford, C. 2018. Circadian rhythms of urinary cortisol levels vary between individuals in wild male chimpanzees: A reaction norm approach. Frontiers in Ecology and Evolution, 6,

Stanton, M. A., Lonsdorf, E. V., Murray, C. M. and Pusey, A. E. 2020. Consequences of maternal loss before and after weaning in male and female wild chimpanzees. Behavioral Ecology and Sociobiology, 74, 22.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

[…] There are, however, a few areas that would still require work/clarification including some typos that should be fixed before we can make a decision on your submission. Note that this does not amount to a partial acceptance of your manuscript.

Introduction: We greatly appreciate the care that you have taken in integrating the suggestions of all reviewers and recrafting the introduction. This section does a wonderful job of setting up the study and the edits you made make for a very impactful series of arguments.

Thank you very much for the positive evaluation of our introduction. We are happy to hear that you appreciate the way we incorporated your previous comments. Thank you again for giving us the tools to strengthen the manuscript.

We have reworked the unclear parts of the manuscript, redrawn all the figures and corrected the typos to improve clarity and readability, and explain this in detail below.

Methods/Results: There are still some spots that are difficult to follow in the results, and this section might require a bit more work than others. Consider the specific comments below:

L266-294: This section describing the models is still a bit confusing, especially the section about predictor variables and how they differ between models 1a/b, 2a/b. Later in the methods, the authors mention that they ran models for each sex separately, but that is not mentioned here.

We apologize for the confusion. We agree that the phrasing was misleading. All our immature models include samples from both males and females, whereas only males were included in the mature models. To clarify the statistical analyses further, we have changed the model names to provide an intuitive description of each model. Model 1a and 1b have been replaced by “all immature” and “immature orphan”, and models 2a and 2b have been replaced by “all adult male” and “adult male orphan” models.

We have edited the text accordingly: “We used a series of Bayesian Linear Mixed Models (LMMs) to test our predictions regarding the effect of maternal loss on overall cortisol levels and diurnal slopes (jointly constituting the cortisol profile). We first tested these effects in socially immatures (i.e. males and females <12 years of age because prior to 12 years, chimpanzees associate primarily with their mother, Reddy and Sandel 2020). Secondly, we tested these effects in mature males (i.e. males >= 12 years of age)” (Line 238-242).

We have also modified the related paragraph in the result section (Lines 821-828).

I was slightly confused by this list of predictors at first thinking all predictors were used in each model. Is it possible to make it a little clearer that this is not the case? Perhaps something like "Each model contained one or more of the following test predictor variables" in line 283/284. In line 287 I think you mean model 1a rather than model 1b – this is probably the root of my confusion as it makes it seem like both years since maternal loss and orphan status as a categorical variable with three levels were included in the same model.

We have rephrased this section to make the presentation of our models and of the different predictor variables clearer. We present now the set of predictor variables for each model separately (Lines 273-285).

I don't have much specific advice to solve the problem, other than to say that it was difficult to follow each thread. Perhaps if each section was more simply dedicated to each model (like the longer methods section) rather than going back and forth between the things that were the same across models versus different? It's a lot to keep track of, so redundancy might be better for ease of interpretation in this case??

We modified our presentation of the model parameters to clarify which variables are included in each model (Lines 273-285). As mentioned above we also changed the name of models to improve clarity (i.e. we replaced Model 1a, 1b, 2a, 2b by “all immature”, “immature orphan”, “all adult male” and “adult male orphan” models).

L315-16: I am not very familiar with Bayesian approaches and this section is unclear to me. In frequentist statistics effects sizes and variance explained are not the same things – could the authors clarify what they are reporting here and what it means?

We report the percentage of the variance explained by the random and fixed effect. You are correct that it is not the same thing as the effect size. Effect sizes are reported elsewhere in the text (e.g. “On average, recently orphaned individuals had a diurnal cortisol slope 58% steeper than non-orphans” Lines 366-367). We have removed mention of effect size in this sentence and simply mention that we report the percentage of variance explained by the fixed and random effects.

L329-331: Did the authors directly test for categorical differences in morning cortisol or evening cortisol or are all of the comparisons here based on slope?

All the comparisons are based on the meaningful differences in slopes and on visual assessment based on the figures depicting the model predictions. We have clarified that in the text (Lines 345-349).

L362-365: The wording here "in particular in the early morning and the afternoon" is confusing given that the take-away is that cortisol had an upward slope and was, therefore, higher in the afternoon compared to the morning.

We have modified our phrasing to clarify our point: “A visual inspection of the data and the model prediction lines reveals that cortisol levels of immature orphans who lost their mother recently had higher cortisol levels throughout the day (orange squares in Figure 2) as compared to immature orphans who lost their mother several years ago (green triangles in Figure 2). The difference in cortisol levels between recently and non-recently orphaned individuals was most evident during early morning and late afternoon (Figure 2).“ (Lines 396-402).

L374-389: The authors jump back and forth between describing life-history-based age categories (under 5 y.o. = infants, 5-8 = juveniles, 8-12 = adolescents) and referring to specific ages ("who lost their mother at 4 y.o."). That makes it difficult to parse whether and where they are using continuous versus categorical age predictions. It is especially difficult because the text describes things as one way or both ways, but the figures describe something firmly in the middle. Please revise these sections to make them clearer.

We modified all the figures to include age categorical intervals that correspond to the life history stages mentioned in the text and are in line with the main text.

Discussion: L444-447: Higher morning and higher evening cortisol does not necessarily mean anything about slope (i.e. the am and pm increases could be equally leading to similar slopes, but higher average cortisol). I think it is important to specify exactly what the authors mean here- are orphans experiencing higher am, higher pm, and different slopes? If so how are the slopes different in layman's terms and which point is contributing to that difference in slope? It looks like the answer comes later (lines 457-58), but it still isn't so clear throughout the paragraph which parts of the results correspond to what theoretical models/predictions, and how. For instance, in L447-448: could the authors be more specific about how this finding aligns with the ACM?

We have edited most of the paragraph to be more specific as to how our results correspond with the predictions of the ACM, and to improve the description of the results found across orphan categories. We specify that the higher early or afternoon cortisol levels for immatures orphaned early as compared to other orphans are derived from visual inspection of the model lines:

“Orphans experiencing maternal loss at younger ages had a diurnal cortisol slope differing from the immatures orphaned when older, in particular in the quadratic term for time of the day (i.e. in how the slope curved). […] In fact, diurnal cortisol slopes are indicative of the general functioning of the HPA axis (Karlamangla et al. 2019) and our results indicate that the diurnal cortisol slope of immature chimpanzees undergo different levels of changes depending on the age at which they experience maternal loss, with more substantial deviations from mother-raised offspring pattern in immatures orphaned earlier in life.” (Lines 501-526).

L471-476: How often does food sharing happen with mothers and weaned offspring? It seems like the authors are asserting that calories from food sharing make up a significant portion of the juvenile chimpanzee diet. Is this the case? If so, that would seem different from other sites.

We have modified the sentence to clarify that the food sharing between the mother and weaned offspring is occasional (Line 557).

L532: One thing to be careful about in discussing adaptive calibration is that the model is more focused on the plasticity of the HPA axis than a change in the environmental conditions. In other words, a return to normal could reflect that the environment has adjusted-but the ACM predicts that the HPA readjusts itself during critical developmental/life history timepoints (e.g. adrenarche, puberty, pregnancy/parenthood) to account for environmental conditions. So that return to normal could be the HPA readjusting itself to essentially make what it previously considered a stressful environment led to less of a stress response kind of like making it a new normal.

We have modified our argumentation relating our results to the ACM model in the mentioned paragraph and in other sections of the discussion. We have rephrased the specific paragraph as follows:

“The re-establishment of a normal functioning of the HPA axis in mature male chimpanzees but also in immatures orphaned for more than two years may reflect a form of recovery in those individuals. […] The lack of apparent long-term effects of maternal loss on immature chimpanzee physiology could thus be indicative of ameliorations in the environment of these orphans in the years following maternal loss, possibly in terms of improved access to social support and food.” (Lines 617-626).

Methods: Can authors add a bit more detail about the choices that they made in creating these models? This will be instructive for helping other scholars follow and match their methodology. For instance (L872-875), what is the difference between a regularizing prior and any other type of prior?

We added information regarding our choice of priors to the method section:

“We chose weakly regularising prior for the fixed effects since they give less weight to outlier data points and therefore help constrain model predictions to biologically meaningful estimates and CI (Lemoine 2019).” (Lines 946-948).

One general question: because I'm not so familiar with Bayesian LMM/GLMM, are there any guidelines or rules for limiting the number of predictor/control terms included in the models? The authors have clearly gone to great pains to control for things, so the concern would just be that including so many terms would exhaust the degrees of freedom for the number of individuals included in the study.

The general guideline is to have at least 10 data points per degree of freedom in the model (i.e. per levels of predictor variable). Our model with the smallest sample size and the largest number of predictor variables is Model 1b (now called “immature orphan model”) in which we use 393 data points for 18 predictors (including the interaction terms) resulting in over 20 data points per predictor, well above the recommended threshold. We therefore do not think that model complexity is a concern for our analyses.

L772-774: If the models are fit separately for males and females, does that mean that 1a/b are four models? 1aMale, 1aFemale, 1bMale, 1bFemale? Were there any differences in results for males versus females?

Apologies for the confusion. We conducted two sets of analyses, either including samples from immature individuals (previously named model 1a and 1b) or mature individuals (previously named model 2a and 2b). The set of analyses of immatures included samples from both males and females in the same analysis, whereas the set of analyses of mature individuals included only males. We have clarified this point in the text:

“We used a series of Bayesian Linear Mixed Models (LMMs) to test our predictions regarding the effect of maternal loss on overall cortisol levels and diurnal slopes (jointly constituting the cortisol profile). […] Secondly, we tested these effects in mature males in the all adult male and adult male orphan models (i.e. four models in total).” (Lines 821-825).

L861-865: Is this a standard control for this field? It's unclear how including project as a random effect would account for things that aren't already controlled for using the other factors mentioned here: individual, year, individual_year, etc.

In this study, we used a longitudinal dataset spanning 20 years, meaning that our data reflects the combined research effort of several observers, each having different research questions (e.g. some observers targeted urine sample collection after aggression while others collected samples opportunistically). Therefore, the sample collection is not fully random but may be biased by the question under investigation. To account for this variation, we have added the “project” as a random effect in our analyses. We have clarified this aspect in the manuscript:

“Finally, our hormonal dataset included samples collected by different observers with different research interests (hereafter project). […] Thus, to account for potential variation in cortisol levels that may be a result of inter-observer project bias, we added the ‘project’ type as an additional random factor.” (Lines 930-936).

L888: Does this mean that all of the actual sample sizes were > 1000? Or something else? My understanding is that there were models, e.g. those with immatures only, that included fewer than 1000 samples?

In Bayesian analysis, the effective sample size reflects the number of independent samples that were drawn from the posterior distribution to calculate our estimates using MCMC processes. Effective sample size here reflects the amount of autocorrelation within the chains. It is not related to sample size (i.e. it is not related to the number of actual data points that we used in our models). We have specified the later point in the manuscript:

“Please note that the effective sample size is a measure of autocorrelation and does not correspond to the number of data points that were used for each model (namely 393, 846, 2184, and 769 for the all immature, the orphan immature, the all adult male and the adult male orphan models respectively)” (Lines 985-988).

Author details

1. Cédric Girard-Buttoz

   1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
   2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft

   For correspondence

   cedric_girard@eva.mpg.de

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1742-4400

2. Patrick J Tkaczynski

   1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
   2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire

   Contribution

   Conceptualization, Data curation, Methodology, writing-review-and-editing

   Competing interests

   No competing interests declared

3. Liran Samuni

   1. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire
   2. Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
   3. Department of Human Evolutionary Biology, Harvard University, Cambridge, United States

   Contribution

   Data curation, writing-review-and-editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-7957-6050

4. Pawel Fedurek

   Division of Psychology, University of Stirling, Stirling, United Kingdom

   Contribution

   writing-review-and-editing

   Competing interests

   No competing interests declared

5. Cristina Gomes

   Tropical Conservation Institute, Florida International University, Miami, United States

   Contribution

   Data curation, writing-review-and-editing

   Competing interests

   No competing interests declared

6. Therese Löhrich

   1. World Wide Fund for Nature, Dzanga Sangha Protected Areas, Bangui, Central African Republic
   2. Robert Koch Institute, Epidemiology of Highly Pathogenic Microorganisms, Berlin, Germany

   Contribution

   Data curation

   Competing interests

   No competing interests declared

7. Virgile Manin

   1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
   2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire

   Contribution

   Data curation

   Competing interests

   No competing interests declared

8. Anna Preis

   Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

   Contribution

   Data curation, writing-review-and-editing

   Competing interests

   No competing interests declared

9. Prince F Valé

   1. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire
   2. Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
   3. Centre Suisse de Recherches Scientifiques en Côte d'Ivoire, Abidjan, Côte d'Ivoire
   4. Unité de Formation et de Recherche Biosciences, Université Félix Houphouët Boigny, Abidjan, Côte d'Ivoire

   Contribution

   Data curation

   Competing interests

   No competing interests declared

10. Tobias Deschner

    Interim Group Primatology, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany

    Contribution

    Data curation, Methodology, resources, writing-review-and-editing

    Competing interests

    No competing interests declared

11. Roman M Wittig

    1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
    2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire

    Contribution

    Conceptualization, Data curation, funding-acquisition, Methodology, project-administration, resources, supervision, writing-review-and-editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-6490-4031

12. Catherine Crockford

    1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany
    2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Abidjan, Côte d'Ivoire
    3. Institut des Sciences Cognitives, CNRS, Lyon, France

    Contribution

    Conceptualization, Data curation, funding-acquisition, Investigation, Methodology, project-administration, supervision, writing-review-and-editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-6597-5106

• 1,232

  Page views

• 125

  Downloads

• 11

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.