1. Developmental Biology
  2. Evolutionary Biology
Download icon

Early maternal loss leads to short- but not long-term effects on diurnal cortisol slopes in wild chimpanzees

  1. Cédric Girard-Buttoz  Is a corresponding author
  2. Patrick J Tkaczynski
  3. Liran Samuni
  4. Pawel Fedurek
  5. Cristina Gomes
  6. Therese Löhrich
  7. Virgile Manin
  8. Anna Preis
  9. Prince F Valé
  10. Tobias Deschner
  11. Roman M Wittig
  12. Catherine Crockford
  1. Department of Human Behavior, Ecology and Culture, Max Planck Institute for Evolutionary Anthropology, Germany
  2. Taï Chimpanzee Project, Centre Suisse de Recherches Scientifiques, Côte d'Ivoire
  3. Department of Primatology, Max Planck Institute for Evolutionary Anthropology, Germany
  4. Department of Human Evolutionary Biology, Harvard University, United States
  5. Division of Psychology, University of Stirling, United Kingdom
  6. Tropical Conservation Institute, Florida International University, United States
  7. World Wide Fund for Nature, Dzanga Sangha Protected Areas, Central African Republic
  8. Robert Koch Institute, Epidemiology of Highly Pathogenic Microorganisms, Germany
  9. Centre Suisse de Recherches Scientifiques en Côte d'Ivoire, Côte d'Ivoire
  10. Unité de Formation et de Recherche Biosciences, Université Félix Houphouët Boigny, Côte d'Ivoire
  11. Interim Group Primatology, Max Planck Institute for Evolutionary Anthropology, Germany
  12. Institut des Sciences Cognitives, CNRS, France
Research Article
  • Cited 0
  • Views 789
  • Annotations
Cite this article as: eLife 2021;10:e64134 doi: 10.7554/eLife.64134

Abstract

The biological embedding model (BEM) suggests that fitness costs of maternal loss arise when early-life experience embeds long-term alterations to hypothalamic-pituitary-adrenal (HPA) axis activity. Alternatively, the adaptive calibration model (ACM) regards physiological changes during ontogeny as short-term adaptations. Both models have been tested in humans but rarely in wild, long-lived animals. We assessed whether, as in humans, maternal loss had short- and long-term impacts on orphan wild chimpanzee urinary cortisol levels and diurnal urinary cortisol slopes, both indicative of HPA axis functioning. Immature chimpanzees recently orphaned and/or orphaned early in life had diurnal cortisol slopes reflecting heightened activation of the HPA axis. However, these effects appeared short-term, with no consistent differences between orphan and non-orphan cortisol profiles in mature males, suggesting stronger support for the ACM than the BEM in wild chimpanzees. Compensatory mechanisms, such as adoption, may buffer against certain physiological effects of maternal loss in this species.

Introduction

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).

Results

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).

Table 1
Sample size for immature and adult male orphans and non-orphans in each of the four study communities.
CommunityAge classOrphan statusN.IDNo. of samplesMean ± SE no. of sample per individualAge range (years)
Taï EastImmaturesNon-orphans8*11214.0 ± 3.83.8–11.9
Recently orphaned5*11623.2 ± 9.14.1–11.9
Non-recently orphaned9*13615.1 ± 6.46.1–11.8
Adult malesNon-orphans4456114.0 ± 56.213.8–40.7
Orphans3354118.0 ± 18.012.3–19.8
Taï MiddleImmaturesNon-orphans
Recently orphaned
Non-recently orphaned
Adult malesNon-orphans3175.7 ± 0.6717.2–33.7
Orphans
Taï NorthImmaturesNon-orphans1116815.3 ± 4.33.0–12.0
Recently orphaned
Non-recently orphaned273.5 ± 0.510.5–11.3
Adult malesNon-orphans39531.7 ± 9.112.3–20.8
Orphans26130.5 ± 3.512.1–20.4
Taï SouthImmaturesNon-orphans17*17310.2 ± 1.52.8–11.9
Recently orphaned2*189.0 ± 6.04.1–9.2
Non-recently orphaned5*11623.2 ± 7.76.1–11.9
Adult malesNon-orphans7858122.6 ± 18.112.1–45.3
Orphans634357.2 ± 31.512.1–21.9
  1. *

    50 immature individuals were included in this study but one immature in Taï East and two immatures in Taï South were sampled before and after their mother died (i.e., they are counted twice in the table, once as an orphan and once as a non-orphan). Six males were included in the study as both mature and immature individuals.

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
Results of the all immature and the immature orphan models.

These two models tested in immatures for the effect of orphan status (all immature model) and, for immature orphan only, the effect of age when mother died and years since maternal loss (orphan immature model) on cortisol levels and diurnal cortisol slopes. The results presented here are from reduced all immature and immature orphan models after removing the interactions for which the 90% credible interval (CI) overlapped 0. SE indicates the standard error of the estimate for each predictor. The coded level for each categorical predictor is indicated in parentheses. Control predictors are italicized. 95% CI low and 95% CI high indicate the lower and upper limits of the 95% CI. Likewise, 90% CI low and 90% CI high indicate the lower and upper limits of the 90% CI. CIs that do not overlap 0 are indicated in bold. Predictors for which the 95% CI did not overlap 0 are indicated with salmon shading, and predictors for which the 95% CI overlapped 0 but the 90% CI did not overlap 0 are indicated with yellow shading. LCMS: liquid chromatography mass spectrometry. Time of the day2: quadratic term for time of the day. Time of the day3: cubic term for time of the day.

ModelResponsePredictorEstimateSE95% CI low95% CI high90% CI low90% CI high
All immature modelLog urinary cortisol levels (ng/ml SG)Intercept3.690.273.174.233.264.14
Time of the day–0.380.080.530.220.510.25
Time of the day2–0.010.08–0.170.14–0.140.11
Time of the day30.030.04–0.050.12–0.040.10
Orphan category (orphan for less than 2 years)0.010.19–0.360.38–0.300.32
Orphan category (orphan for more than 2 years)–0.140.22–0.580.32–0.500.23
Sex ratio0.070.07–0.060.2–0.040.18
Community size–0.120.06–0.240.010.220.01
Individual sex (male)–0.260.25–0.780.22–0.670.13
Individual age at sample–0.010.17–0.330.37–0.270.29
LCMS method (old)0.060.35–0.620.78–0.500.64
Sin(seasonDate)–0.070.04–0.140–0.13–0.02
Cos(seasonDate)0.010.04–0.060.08–0.050.07
Orphan category (less than 2 years): time of the day–0.220.100.410.030.380.07
Orphan category (more than 2 years): time of the day–0.110.08–0.270.03–0.240.01
Immature orphan modelLog urinary cortisol levels (ng/ml SG)Intercept3.590.263.084.113.174.02
Time of the day–0.440.170.750.050.70.14
Time of the day2–0.010.07–0.160.13–0.130.1
Time of the day30.030.11–0.180.27–0.140.22
Orphan’s age when mother died0.000.24–0.510.46–0.410.38
Years since maternal loss–0.160.16–0.480.14–0.420.09
Sex ratio–0.060.11–0.270.15–0.230.11
Community size–0.050.08–0.220.12–0.190.09
Individual sex (male)0.10.3–0.470.7–0.380.6
LCMS method (old)–0.040.32–0.660.64–0.550.51
Sin(seasonDate)0.020.05–0.090.12–0.070.1
Cos(seasonDate)0.040.05–0.070.14–0.050.12
Years since maternal loss: time of the day0.190.1–0.010.380.020.35
Years since maternal loss: time of the day2–0.050.05–0.150.05–0.140.03
Years since maternal loss: time of the day3–0.060.04–0.150.020.140.00
Age when mother died: time of the day0.080.09–0.090.25–0.060.22
Age when mother died: time of the day2–0.120.060.230.000.210.02
Effect of maternal loss on daily urinary cortisol level variations in immature chimpanzees.

Non-orphans are depicted by orange squares, recently orphans (orphaned for less than 2 years) by green triangles, `and non-recently orphaned (orphaned for more than 2 years) by blue circles. Each dot represents the average hourly cortisol level for all individuals of each orphan category. The size of the dot is proportional to the sample size (e.g., the number of data points) contributing to each dot. The orange solid line and green and blue dashed lines depict the model line prediction, and the orange, green, and blue area the 95% credible interval (CI) from the all immature model for non-orphans, recently orphans, and non-recently orphans, respectively. The model lines depict the consistent effect of the interaction between orphan status and the linear term for time of day in the all immature model (estimate: –0.22, 95% CI: [–0.03: –0.48]). The sample size for the all immature model was N = 846 samples and 50 individuals.

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).

Effect of time (in years) since being orphaned on daily urinary cortisol level variations of immature orphan chimpanzees.

Orange squares, blue circles, and green triangles depict cortisol levels for individuals who lost their mother less than 3 years ago, 3–6 years ago, and 6–9 years ago, respectively. Each dot represents the average hourly cortisol level of all individuals for each of the three categorical intervals of years since maternal loss. The size of the dot is proportional to the sample size (number of urine samples collected) for each hour of the day. The orange solid line, blue dashed line, and green dashed lines indicate model line predictions, and the orange, blue, and green light areas indicate the 95% credible interval (CI) for individuals who lost their mother less than 3 years ago, 3–6 years ago, and 6–9 years ago, respectively (immature orphan model). The model line depicts the interaction between years since maternal loss and the cubic term for time of day in the immature orphan model: (estimate and [90% CI]: –0.06 [-0.14: 0.00]), but this effect comprised a relatively large uncertainty since the 95% CI overlapped 0 (Table 2). Note that while the predictor ‘years since maternal loss’ was modeled as a continuous variable in the immature orphan model, for ease of interpretation we depict the results here for three categorical intervals of ‘years since maternal loss’. The sample size for the immature orphan model was N = 393 samples and 17 individuals.

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.

Effect of age at which immature orphans lost their mother on daily urinary cortisol level variations of immature orphan chimpanzees.

Orange square, blue circles, and green triangles depict cortisol levels for individuals who lost their mother when they were less than 5 years, 5–8 years, and 8–12 years of age, respectively. Each dot represents the average hourly cortisol level of all individuals for each of the three categories of age when mother died. The size of the dot is proportional to the sample size (number of urine sample collected) for each hour of the day. The orange solid line and blue and green dashed lines indicate model line predictions, and the orange, blue, and green light areas the 95% credible interval (CI) for individuals who lost their mother when they were less than 5 years, 5–8 years, and 8–12 years of age, respectively (immature orphan model). The model lines depict the consistent effect of the interaction between age when mother died and the quadratic term of time of day in the immature orphan model (estimate and [95% CI]: –0.12 [0.00: –0.23]). Note that while ‘age at maternal loss’ was modeled as a continuous variable in the immature orphan model, for ease of interpretation the model is depicted here for three categorical intervals of ’age when mother died’. The sample size for the immature orphan model was N = 393 samples and 17 individuals.

Effect of maternal loss on cortisol slopes in mature male chimpanzees

In contrast to immature individuals, we did not detect a consistent effect of orphan status on mature males’ diurnal cortisol profiles and cortisol levels. All the 90% CIs for the estimate of the interaction terms between orphan status and the linear, quadratic, and cubic terms for time of day in the full all adult male model largely overlapped 0, and all P+ and P- were below 75% for the estimates of each interaction (N = 2184 samples and 28 individuals; Appendix 1—table 1). Furthermore, in a reduced model not comprising these interactions, the 90% CI for the estimate of orphan status largely overlapped 0 and P+ = 56% (Appendix 1—table 1). Conditional and marginal R2 for the all adult male model were 0.53 and 0.28, respectively.

We also did not detect a consistent effect of the age when mother died on orphan mature males’ diurnal cortisol slopes and cortisol levels. All the 90% CIs for the estimates of the interaction terms between age when mother died and the linear, quadratic, and cubic terms for time of day in the full adult male orphan model overlapped 0, and all P + and P- were below 92% (N = 769 samples and 10 individuals; Appendix 1—table 2). In a reduced model not comprising these interactions, the 90% CI for the estimate of age when mother died largely overlapped 0 and P- = 62% (Appendix 1—table 2). Conditional and marginal R2 for the adult male orphan model were 0.60 and 0.38, respectively.

Discussion

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.

Materials and methods

Ethics statement

Request a detailed protocol

Our study was purely observational and non-invasive. Observers followed the strict hygiene protocol of Taï Chimpanzee Project, which was adopted by IUCN as the best practice guideline for wild ape studies (Gilardi et al., 2015; Appendix 1). Observers quarantined for 5 days before following the chimpanzees. During follows, observers disinfected their hands and boots and changed clothes before leaving and entering camps. In the forest, observers wore face masks and kept a minimum distance of 8 m between themselves and the chimpanzees to avoid disease transmission from humans to chimpanzees, and to avoid disturbing the natural behavior of the observed individuals. The research presented here was approved by the ‘Ethikrat’ of the Max Planck Society on 04.08.2014.

Study communities

Request a detailed protocol

We used the long-term 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 behavioral observation of the chimpanzees started in 1982 and is still ongoing. The observation periods for each of the communities are as follows: North 1982–present; South, 1993–present; Middle, 1995–2004; East, 2000–present (Wittig, 2018). Urine samples were collected regularly in all communities from 2000 onwards, except for the East community where sample collection started in 2003.

Study subjects

Request a detailed protocol

For this study, we considered all immature individuals from both sexes (<12 years of age) and mature males (≥12 years of age) from whom urine samples were collected. The age ranges for immatures sampled in this study were 2.82–11.99 years for non-orphans and 4.10–11.99 years for orphans. Physical maturity may come later in male Taï chimpanzees, but 12 years is the age at which chimpanzees range predominantly independently of their mother (Taï Chimpanzee Project, unpublished data) and are fully integrated in the male hierarchy (Mielke et al., 2018). We excluded mature females from the analysis since most females in our study immigrated from unhabituated communities into the study communities, which meant we had no knowledge of the presence or absence of their mothers during their immature years. We excluded orphans for whom the date of death of the mother was unknown (e.g., occurred before habituation of the study community). For individual samples, we excluded outliers (i.e., samples with very low or very high hormonal measures, see details in Appendix 1) and samples collected when the individuals were sick. We also ensured that the final dataset comprised at least three data points per individual per year, and that the earliest and the latest samples were separated by at least 6 hr, to ensure a meaningful evaluation of the diurnal cortisol slope of each individual (see details in Appendix 1). In total, we used 846 samples from 50 immatures, including 17 orphans (N samples per individual mean ± SE = 16.9 ± 2.2) and 2184 samples from 28 mature males, including 11 orphans (N samples per individual mean ± SE = 78 ± 13.5). For immatures, we used an average of nine samples per individual per year, and half of the individual_year comprised at least seven samples. For adult males, we used an average of 19 samples per individual per year, and half of the individual_year comprised at least 10 samples.

Demographic and behavioral data collection

Request a detailed protocol

In Taï, each chimpanzee community is followed daily by a joint effort of local and international assistants and researchers (Wittig and Boesch, 2019b). Each day, the observers conducted focal follows from and to sleeping sites. The focal individual was either followed all day (i.e., 12 hr) or the identity of the focal changed around 12h30 and two different individuals were followed each day one after the other (i.e., 6 hr focal follow each). The observer recorded detailed focal and ad libitum behavioral data (Altmann, 1974). Observers recorded all social interactions such as aggression and submissive behaviors, which we then used to build a dominance hierarchy (see below). In addition, each day, the observers recorded the presence of all individual chimpanzees they encountered, which provides a detailed account of the demography of each community. Specifically, we obtained detailed information on individuals’ date of birth, immigration/emigration, and death or disappearance. This information was used to determine the early-life history of the study subject, namely if their mother died before they reached 12 years of age, and, if so, the age of the subject when its mother died.

Assessment of dominance hierarchy in mature males

Request a detailed protocol

Since dominance rank may correlate with the cortisol levels of adult male chimpanzees (e.g., Muller and Wrangham, 2004), we wanted to control for this parameter in our analysis of cortisol patterns in adult males. We calculated the dominance hierarchy for mature males in each of the study communities using a modified version of the Elo-rating method (Neumann et al., 2011) developed by Foerster et al., 2016. In this modified version, the k parameters and the starting score of each individual are optimized using maximum likelihood approximation (Mielke et al., 2018, see details in Appendix 1). We used all of the long-term data available on unidirectional submissive pant-grunt vocalizations, given by the lower ranking of the two individuals towards the higher ranking (Bygott, 1979). We used 9189 pant-grunt recorded for males in Taï South, 3952 in Taï East, 5784 in Taï North, and 111 in Taï Middle. All Elo-rating scores were standardized between 0 and 1 with 1 being the highest-ranking individual and 0 the lowest ranking on any given day. We then extracted the Elo-rating score of each individual on the day when each urine sample was collected.

Urine sample collection and analysis

Request a detailed protocol

During chimpanzee follows, we collected urine samples opportunistically from known individuals. Directly after urination, we collected the urine from leaves and/or the ground into a 2 ml cryo vial using a disposable plastic pipette. Within 12 hr of collection, we placed these vials in liquid nitrogen. Subsequently, the samples were shipped on dry ice to the Endocrinology Laboratory of the Max Plank Institute for Evolutionary Anthropology in Leipzig, Germany, and stored at ≤–20°C until analysis. We usedLCMS ( Hauser et al., 2008; Murtagh et al., 2013) and MassLynx (version 4.1; QuanLynx-Software) to quantify cortisol concentrations in each sample. For all samples analyzed, we used either prednisolone (hereafter the ‘old’ method; i.e., most samples analyzed prior to July 2016) or cortisol d4 (hereafter the ‘new’ method; i.e., all samples analyzed after September 2016) as our internal standard for the cortisol measurements. To adjust for water content in the urine (i.e., urine concentration), we measured, for each sample, its specific gravity (SG) using a refractometer (TEC, Ober-Ramstadt, Germany). We corrected our cortisol concentration for urine water content in each sample using the following formula provided by Miller et al., 2004:

SG corrected cortisol = raw cortisol concentration(ng/ml) × (SGpopulation mean  1.0)(SGsample  1.0)

where SGpopulation mean is the mean SG value average across all the samples used in this study (SGpopulation mean = 1.022 in our study), and SGsample is the SG value of each given sample. In the article, all cortisol concentrations are reported as ng/ml SG.

Statistical analysis

Request a detailed protocol

We used a series of Bayesian 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 from both sexes (all immature and immature orphan models). Secondly, we tested these effects in mature males in the all adult male and adult male orphan models (i.e., four models in total). In all the models, each urine sample represented a data point and the cortisol concentration of the sample (expressed in ng/ml SG) was the response variable. We log-transformed the cortisol values to achieve a symmetric distribution of the response.

Effect of maternal loss on cortisol profiles in immatures

Request a detailed protocol

In the all immature model, we tested the prediction that immature orphans, especially those recently orphaned, will have higher overall cortisol levels and a flatter slope of diurnal cortisol slope when compared to non-orphans. We fitted a LMM with a categorical variable for orphan status with three levels (recently orphaned, non-recently orphaned, and non-orphans) as our test predictor to test for the short- and long-term effects of maternal loss on overall cortisol levels in immatures. In addition, to test for the effect of this orphan status on the diurnal cortisol variation, we incorporated the linear, quadratic, and cubic terms for time of sample collection as test predictors as well as their interaction with orphan status to test whether the diurnal cortisol slope differed between recently orphaned, non-recently orphaned, and non-orphans. The time of sample collection was expressed in minutes with 0 being midnight and 720 being noon. In addition, in this model, we used the following control predictors: the LCMS method used (‘old’ or ‘new’ method, see the Urine analysis section), sex of the individual, community size, sex ratio of mature individuals in the community and age of the individual, since these factors can all influence the cortisol profiles of immature and/or mature chimpanzees (Muller and Wrangham, 2004; Emery Thompson et al., 2020; Tkaczynski et al., 2020). Community ID was not included as a control predictor in our analysis since it was highly correlated with community size. In addition, we accounted for seasonal variation in ecological conditions (e.g. rainfall, temperature, food availability) that can affect cortisol levels in chimpanzees (Wessling et al., 2018; Preis et al., 2019; Samuni et al., 2019b) by converting the Julian date at which samples were collected into a circular variable and including the sine and cosine of this variable in our model (Wessling et al., 2018).

Effect of age at mother’s death and time since mother died on cortisol profile in orphan immatures

Request a detailed protocol

In the immature orphan model, we focused on immatures who were orphans at the time of sampling in order to investigate more specifically the effects of the age at maternal loss and years since maternal loss on cortisol levels. We tested the predictions that (a) immature individuals who were orphaned at a younger age would have higher cortisol levels and flatter diurnal cortisol slopes and (b) that, if some form of recovery occurs, these effects will be weaker the more time has passed since an individual lost its mother. We incorporated two test predictors in the immature orphan model, the age at which the individuals have been orphaned (in days since their date of birth), and the time since the individuals have been orphaned (in days since the date their mother died). As in the all immature model, we incorporated the six interaction terms between these two test predictors and the linear, quadratic, and cubic terms for time of sample collection. As before, we also incorporated individual sex, community size, sex ratio, and LCMS method, and the sine and cosine of the Julian date as control fixed effect. Initially, we also wanted to incorporate the age of the individual at the time of sampling into the immature orphan model. This was however not possible due to collinearity between the individual age at sample and both the age at which the individual was orphaned and the time since its mother died (i.e., the model did not run due to collinearity issues). However, since age at sample did not have a consistent effect in our immature sample in the all immature model (90% CI: –0.27: 0.29) we decided to not include age at sample in the immature orphan model, allowing us to test our variable of interest, namely years since maternal loss and age when mother died.

Effect of maternal loss on cortisol profiles in mature males

Request a detailed protocol

In the all adult male model, we tested whether mature males (≥12 years) who were orphaned as immatures (i.e., before 12 years of age) had overall higher cortisol levels and a flatter diurnal cortisol slope than mature males who did not lose their mother before 12 years of age. We fitted a LMM with the early-life orphan status (i.e., ‘no’, if the mother of the individual was still alive when the individual reached 12 years of age, and ‘yes’ if the mother died before the individual was 12 years of age) as our test predictor. As in the all immature model, we incorporated three interaction terms between orphan status and the linear, quadratic, and cubic terms for time of sample collection. As in the other models, we used community size, the sex ratio, the age of the individual, the LCMS method, and the sine and cosine of the Julian date as control fixed effects. In addition, we controlled for the dominance rank of the individual by adding the standardized Elo-rating score of each individual on the day the sample was collected as a fixed effect into the model.

Effect of age at mother’s death on cortisol profile in orphan immatures

Request a detailed protocol

In the adult male orphan model, we assessed whether the age at which the orphan adult male chimpanzees lost their mother impacted the diurnal cortisol levels and slopes of mature males (i.e., whether the potential effect of early-life adversity continued into adulthood). Accordingly, we fitted a LMM with ‘age at which mother died’ as a test predictor and its interaction with the linear, quadratic, and cubic terms for time of day. In this model, we used only samples collected from mature males who lost their mothers before they were 12 years of age. As previously, we used community size, the sex ratio, the age of the individual, the LCMS method, and the sine and cosine of the Julian date as control fixed effects.

In addition to the fixed effects, in all of the LMMs we included individual identity as a random factor to avoid pseudoreplication. To control for the changes in cortisol diurnal slope with age, we built one slope per individual per year into each model by incorporating as random factor a dummy variable ‘individual_year’. In addition, since certain years might have particularly harsh or favorable ecological conditions, and since this can affect cortisol levels in primates (e.g., Young et al., 2019), we also included year as a random factor in each model. Finally, our hormonal dataset included samples collected by different observers with different research interests (hereafter project). Whilst all projects followed a similar design to collect at least one urine sample from focal individuals throughout the day, some projects conducted additional targeted sampling of specific behaviors such as aggressions or affiliations. 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.

All analyses were conducted in R 4.0.3 (R core Team 2020) using the function brm from the package ‘brms’ (Bürkner, 2018). In each model, we included the maximal random slope structure between each fixed predictor (test and control) and each random effect (Baayen et al., 2008; Barr et al., 2013) and the correlation between intercept and slopes. In particular, the linear, quadratic, and cubic terms for time of day were included as random slopes within each of the random effects. For each model, we extracted both the 95% and 90% CIs for each fixed effect and for each random effect from the posterior distribution. In all the models, we used weakly regularizing priors for the fixed effects (Normal (0,1)) and the priors given by default by the function ‘get_prior’ of the package ‘brms’ for the random effects (i.e., Student t (3, 0, 2.5) for the random intercepts and slopes and lkj (1) for their correlation). We chose weakly regularizing 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).

Before fitting each model, we tested for collinearity issues between our predictor variables by computing the variance inflation factor (VIF) using the function vif from the package ‘car’ (Fox and Weisberg, 2011). Collinearity was not an issue in any of the final models (VIF of all predictor variables < 3.6). Sampling diagnostics (Rhat <1.1) and trace plots confirmed chain convergence for all models. Effective sample sizes (all >1000) confirmed no issues with autocorrelation of sampling for all models. 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). After running the models, we processed to a posterior predictive check using the function ‘pp_check’ from the package ‘brms’ (Appendix 1—figure 2, Appendix 1—figure 3, Appendix 1—figure 4, Appendix 1—figure 5).

Appendix 1

Data preparation

The initial dataset comprised 4604 samples (1518 from immature individuals and 3086 from mature males) from 46 female and 48 male immature individuals and 34 mature males. We applied a suite of selection criteria to subset our dataset to samples collected from individuals from whom all demographic and social data needed were available. We excluded all individuals for whom we could not assess if the mother died before they were 12 years of age or the age they were when their mother died. We also excluded samples for whom the cortisol concentration could not be measured or was excessively low (<0.1 ng/ml SG). We excluded samples with very low SG (SG <1.003). Very low SG values are a sign of over-diluted samples that reflect potential contamination with rain water and can, in turn, inflate cortisol concentration measurements. We also excluded samples collected from individuals on days when they displayed injuries or symptoms of sickness (as assessed by the on-site veterinary staff) since injury and sickness lead to extremely elevated cortisol levels in primates (e.g., Barton 1987; Muehlenbein & Watts 2010; Behringer et al., 2020). Finally, since a large part of our analysis focused on circadian cortisol variation, we excluded all samples for which we did not have a precise time of collection recorded. For the same reason, we limited our dataset for each individual to years when at least three samples were collected from this specific individual, and in years in which the earliest and the latest sample collections were separated in time by at least 6 hr. This criterion was applied in order to be able to calculate, in our statistical model, a meaningful circadian slope for each individual each year with time variation representing at least half of the active time of the chimpanzees (i.e., at least 6 hr out of 12 hr). The three samples could have been collected on different days, but ‘time of sample collection’ was used to define the 6 hr criteria. Following this selection process, we were left with 849 samples from 50 immatures (including 17 orphans) and 2184 samples from 28 mature males (including 11 orphans).

Appendix 1—table 1
Results of the all adult male model testing the effect of maternal loss on cortisol profiles in all mature males.

’All adult male model full’ refers to the full model ran with all the interactions considered. ‘All adult male model reduced’ refers to the reduced model after removing the interactions for which the 90% credible interval (CI) overlapped 0. SE indicates the standard error of the estimate for each predictor. The coded level for each categorical predictor is indicated in parentheses. Control predictors are italicized. 95% CI low and 95% CI high indicate the lower and upper limits of the 95% CI. Likewise, 90% CI low and 90% CI high indicate the lower and upper limits of the 90% CI. CIs that do not overlap 0 are indicated in bold. LCMS: liquid chromatography mass spectrometry. Time of the day2: quadratic term for time of the day. Time of the day3: cubic term for time of the day.

ModelResponsePredictorEstimateSE95% CI low95% CI high90% CI low90% CI high
All adult male model fullLog urinary cortisol levels (ng/ml SG)Intercept3.970.253.444.443.554.36
Time of the day–0.480.09–0.67–0.32–0.63–0.35
Time of the day2–0.070.05–0.170.03–0.150.01
Time of the day30.030.04–0.050.12–0.040.1
Orphan status (yes, orphan)0.020.19–0.360.38–0.290.32
Individual age at sample0.190.14–0.070.49–0.030.42
LCMS method (old)0.170.31–0.410.82–0.310.69
Community size0.060.09–0.130.24–0.10.21
Sex ratio0.040.07–0.110.18–0.090.16
Dominance rank0.040.08–0.120.19–0.090.17
Sin(seasonDate)–0.040.02–0.080–0.070
Cos(seasonDate)–0.010.02–0.050.03–0.040.03
Orphan status (yes): time of the day0.020.04–0.050.1–0.040.09
Orphan status (yes): time of the day20.010.04–0.080.09–0.060.08
Orphan status (yes): time of the day3–0.030.08–0.180.13–0.150.1
All adult male model reducedLog urinary cortisol levels (ng/ml SG)Intercept3.970.243.454.433.554.35
Time of the day–0.490.08–0.67–0.33–0.64–0.36
Time of the day2–0.060.04–0.150.03–0.140.01
Time of the day30.040.04–0.030.12–0.020.11
Orphan status (yes, orphan)0.030.18–0.330.38–0.280.32
Individual age at sample0.190.14–0.070.48–0.020.42
LCMS method (old)0.170.3–0.410.8–0.30.68
Community size0.050.09–0.130.24–0.10.21
Sex ratio0.030.07–0.110.17–0.090.15
Dominance rank0.040.08–0.120.19–0.090.17
Sin(seasonDate)–0.040.02–0.080–0.070
Cos(seasonDate)–0.010.02–0.050.03–0.040.03
Appendix 1—table 2
Results of the adult male orphan model testing, in orphan mature males only, the effect of age at maternal loss on cortisol profiles.

’Adult male orphan model full’ refers to the full model ran with all the interactions considered. ‘Adult male orphan model reduced’ refers to the reduced model after removing the interactions for which the 90% credible interval (CI) overlapped 0.SE indicates the standard error of the estimate for each predictor. The coded level for each categorical predictor is indicated in parentheses. Control predictors are italicized. 95% CI low and 95% CI high indicate the lower and upper limits of the 95% CI. Likewise, 90% CI low and 90% CI high indicate the lower and upper limits of the 90% CI. CIs that do not overlap 0 are indicated in bold. LCMS: liquid chromatography mass spectrometry. Time of the day2: quadratic term for time of the day. Time of the day3: cubic term for time of the day.

ModelResponsePredictorEstimateSE95% CI low95% CI high90% CI low90% CI high
Adult male orphan fullLog urinary cortisol levels (ng/ml SG)Intercept4.070.333.414.713.534.58
Time of the day–0.50.13–0.77–0.24–0.72–0.29
Time of the day2–0.080.07–0.210.05–0.190.03
Time of the day30.040.06–0.090.17–0.060.14
Orphan’s age when mother died–0.120.25–0.620.37–0.530.29
Community size0.040.21–0.340.48–0.270.4
Sex ratio0.020.17–0.310.34–0.250.29
Dominance rank–0.070.18–0.410.3–0.360.24
Orphan’s age at sample0.430.37–0.231.24–0.121.08
LCMS method (old)0.030.31–0.550.66–0.450.54
Sin(seasonDate)–0.010.03–0.080.05–0.070.04
Cos(seasonDate)–0.010.03–0.080.06–0.070.05
Orphan’s age when mother died: time of the day0.030.07–0.110.18–0.090.04
Orphan’s age when mother died: time of the day20.060.04–0.020.14–0.010.12
Orphan’s age when mother died: time of the day3–0.020.04–0.110.06–0.090.15
Adult male orphan reducedLog urinary cortisol levels (ng/ml SG)Time of the day4.070.333.434.723.554.6
Time of the day2–0.510.13–0.77–0.26–0.72–0.3
Time of the day3–0.070.07–0.210.06–0.190.04
Orphan’s age when mother died0.040.06–0.080.17–0.050.14
Community size–0.080.25–0.590.41–0.50.33
Sex ratio0.040.21–0.360.49–0.280.39
Dominance rank0.020.17–0.330.33–0.270.28
Orphan’s age at sample–0.070.18–0.420.31–0.360.24
LCMS method (old)0.430.38–0.241.26–0.121.09
Sin(seasonDate)0.020.31–0.580.66–0.470.54
Cos(seasonDate)–0.010.03–0.080.06–0.070.05
Time of the day–0.010.03–0.080.06–0.070.04
Appendix 1—table 3
List of orphan immatures in the study and information about the adoption by adult individuals in the community.
IdentitySexCommunityAge when mother died (years)Adopted?*Duration of adoption*No. of samples in the studyAge when sampled for the study
BeatriceFEast4.9Yes>3 years385.3–8.6
EmmaFEast4.1Yes>2 years174.1–6.1
EolosMEast3.9Yes>3 years77.0–7.3
ErasmusMEast8.7Unknown318.8–11.8
FatimaFEast6.6Unknown87.8–10.1
GiaFEast2.6Yes17 months810.6–11.9
MaimounaFEast4.7No206.7–8.9
QuarantineFEast5.5Yes>2 years47.6–7.9
RichelieuMEast5.4Unknown6410.7–11.8
WillyMEast10.5Unknown5710.5–11.9
BalooFSouth3.8yes1 year227.3–8.5
CaramelMSouth7.3Unknown48.7 – 9.3
MohanFSouth4.1Yes1.5 year344.1 – 6.5
OscarMSouth4.7Yes>2 years498.1 – 11.9
WalaFSouth4.9Unknown258.1 – 10.9
RoxaneFNorth4.7Unknown410.5 – 11.3
VoltaFNorth3.8Unknown310.7 – 10.7
  1. *

    Data taken from Samuni et al., 2019a.

Appendix 1—figure 1
Diurnal cortisol level variation in immature chimpanzees.

Non-orphans are depicted by orange squares, recently orphans (orphaned for less than 2 years) by green triangles, and non-recently orphaned (orphaned for more than 2 years) in blue circles. Each dot represents one sample. As for Figure 1, the orange solid line and green and blue dashed lines depict the model prediction lines from the all immature model Model 1a for non-orphans, recently orphans, and non-recently orphans, respectively. The model lines depict the consistent effect of the interaction between orphan status and the linear term for time of day in the all immature model Model 1a (estimate: –0.22, 95% credible interval [CI]: [–0.03: –0.48]). The sample size for the all immature model Model 1a was N = 846 samples and 50 individuals.

Appendix 1—figure 2
Posterior predictive check for the all immature model.
Appendix 1—figure 3
Posterior predictive check for the immature orphan model.
Appendix 1—figure 4
Posterior predictive check for the all adult male model.
Appendix 1—figure 5
Posterior predictive check for the adult male orphan model.

Data availability

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

The following data sets were generated
    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

References

  1. Book
    1. Boesch C
    2. Boesch-Achermann H
    (2000)
    The Chimpanzees of the Taï Forest: Behavioural Ecology and Evolution
    Oxford: Oxford University Press.
  2. 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.
  3. Software
    1. Clark C
    (1977)
    A preliminary report on weaning among chimpanzees of the Gombe National Park, Tanzania
    Primate Bio-Social Development.
  4. Book
    1. Clutton-Brock TH
    (2016)
    Mammal Societies
    Chichester, UK: John Whiley & Sons.
  5. Book
    1. Fox J
    2. Weisberg S
    (2011)
    An (r) Companion to Applied Regression (Second Edition)
    Thousand Oaks CA: Sage.
  6. 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.
  7. Book
    1. Goodall J
    (1986)
    The Chimpanzees of Gombe: Patterns of Behavior
    Harvard University Press.
    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
    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.
    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.
  8. 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.
  9. 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
  10. 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.
  11. 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
  12. 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
  13. 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

Decision letter

  1. Chima Nwaogu
    Reviewing Editor; University of Cape Town, South Africa
  2. George H Perry
    Senior Editor; Pennsylvania State University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

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?

https://doi.org/10.7554/eLife.64134.sa1

Author response

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).

https://doi.org/10.7554/eLife.64134.sa2

Article and author information

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

Funding

H2020 European Research Council (679787)

  • Catherine Crockford

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are very grateful to Christophe Boesch for his years of dedication to building the Taï Chimpanzee Project and amassing impressive long-term data, and for engaging in massive and critical conservation efforts to ensure the ongoing survival of West African Chimpanzees. We thank the Ministère de l’Enseignement Supérieur et de la Recherche Scientifique and the Ministère de Eaux et Fôrets in Côte d’Ivoire, and the Office Ivoirien des Parcs et Réserves for permitting the study. We are grateful to the Centre Suisse de Recherches Scientifiques en Côte d’Ivoire and to Tatiana Bortolato and Lara Southern and to the staff members of the Taï Chimpanzee Project for their support and collecting the data. We thank three anonymous reviewers and the editor for very constructive comments on a previous version of this manuscript. This study was funded by the Max Planck Society and the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program awarded to CC (grant agreement no. 679787). Core funding for the Taï Chimpanzee Project has been provided by the Max Planck Society since 1997.

Ethics

The research presented here was non-invasive and did not comprise experimental work. Our work comprised only behavioral observations from a distance and non-invasive collection of urine samples. This work was approved by the 'Ethikrat' of the Max Planck Society and the European Research Council ethics board under grant agreement no. 679787.

Senior Editor

  1. George H Perry, Pennsylvania State University, United States

Reviewing Editor

  1. Chima Nwaogu, University of Cape Town, South Africa

Publication history

  1. Received: October 19, 2020
  2. Accepted: May 19, 2021
  3. Version of Record published: June 16, 2021 (version 1)

Copyright

© 2021, Girard-Buttoz et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 789
    Page views
  • 58
    Downloads
  • 0
    Citations

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

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Developmental Biology
    Zhaoyang Liu et al.
    Research Article Updated

    Adolescent idiopathic scoliosis (AIS) is the most common spine disorder affecting children worldwide, yet little is known about the pathogenesis of this disorder. Here, we demonstrate that genetic regulation of structural components of the axial skeleton, the intervertebral discs, and dense connective tissues (i.e., ligaments and tendons) is essential for the maintenance of spinal alignment. We show that the adhesion G protein-coupled receptor ADGRG6, previously implicated in human AIS association studies, is required in these tissues to maintain typical spine alignment in mice. Furthermore, we show that ADGRG6 regulates biomechanical properties of tendon and stimulates CREB signaling governing gene expression in cartilaginous tissues of the spine. Treatment with a cAMP agonist could mirror aspects of receptor function in culture, thus defining core pathways for regulating these axial cartilaginous and connective tissues. As ADGRG6 is a key gene involved in human AIS, these findings open up novel therapeutic opportunities for human scoliosis.

    1. Developmental Biology
    2. Genetics and Genomics
    Yukiko Tando et al.
    Research Article

    Exposure to environmental factors during fetal development may lead to epigenomic modifications in fetal germ cells, altering gene expression and promoting diseases in successive generations. In mouse, maternal exposure to di(2-ethylhexyl) phthalate (DEHP) is known to induce defects in spermatogenesis in successive generations, but the mechanism(s) of impaired spermatogenesis are unclear. Here, we showed that maternal DEHP exposure results in DNA hypermethylation of promoters of spermatogenesis-related genes in fetal testicular germ cells in F1 mice, and hypermethylation of Hist1h2ba, Sycp1, and Taf7l, which are crucial for spermatogenesis, persisted from fetal testicular cells to adult spermatogonia, resulting in the downregulation of expression of these genes. Forced methylation of these gene promoters silenced expression of these loci in a reporter assay. These results suggested that maternal DEHP exposure-induced hypermethylation of Hist1h2ba, Sycp1, and Taf7l results in downregulation of these genes in spermatogonia and subsequent defects in spermatogenesis, at least in the F1 generation.