Psychosocial Stress Increases Risk for Type 2 Diabetes in Female Cynomolgus Macaques Consuming a Western Diet

Abstract

Chronic psychosocial stress is associated with increased risk of many chronic diseases including type 2 diabetes mellitus. However, it is difficult to establish a causal relationship between stress and diabetes in human studies because stressors often are self-reported and may be distant in time from metabolic consequences. Macaques are useful models of the effects of chronic psychosocial stress on health and may develop obesity and diabetes similar to human beings. Thus, we studied the relationships between social subordination stress – a well-validated psychological stressor in macaques – and body composition and carbohydrate metabolism in socially housed, middle-aged female cynomolgus monkeys (Macaca fascicularis; n=42). Following an 8-week baseline phase, the monkeys were fed a Western diet for 36 months (about equivalent to 10 human years). Social status was determined based on the outcomes of agonistic interactions ($\overline{X}=33.3$ observation hours/monkey). Phenotypes collected included plasma cortisol, body composition, circulating markers of glucose metabolism, activity levels, and heart rate variability measured as RMSSD (root of mean square of successive differences) and SDDN (standard deviation of beat to beat interval) after 1.5- and 3-years on diet. Mixed model analyses of variance revealed that aggression received, submissions sent, and cortisol were higher, and RMSSD and SDNN were lower in subordinates than dominants (social status: p<0.05). After 3 years of Western diet consumption, fasting triglyceride, glucose and insulin concentrations, calculated insulin resistance (HOMA-IR), body weight and body fat mass increased in all animals (time: all p’s<0.05); however, the increase in fasting glucose and HOMA-IR was significantly greater in subordinates than dominants (time × social status: p’s<0.05). Impaired glucose metabolism, (glucose > 100 mg/dl) incidence was significantly higher in subordinates (23%) than dominants (0%) (Fisher’s exact test, p<0.05). These findings suggest that chronic psychosocial stress, on a Western diet background, significantly increases type 2 diabetes risk in middle-aged female primates.

1. Introduction

Psychosocial stress is associated with increased risk of several prevalent chronic diseases including cardiometabolic diseases such as type 2 diabetes mellitus (Kuo et al., 2019; Shin and Kim, 2020). Social status is one of the clearest drivers of psychosocial stress. Low socioeconomic status (SES) appears stressful as indicated higher levels of cortisol, and catecholamines, and self-reported stress levels (Cohen et al., 2006; Damaske et al., 2016). Stress associated with low SES appears to mediate, in part, the relationship between SES and poor health outcomes; however other factors such as sustained work stress, early live adversity, stress management interventions, and healthy behaviors (diet, exercise, adherence to medications) have also been associated with obesity and diabetes (Hackett and Steptoe, 2017; Heraclides et al., 2009). Likely a combination of many risk factors influenced by chronicity, predictability, and availability of coping mechanisms, mediate and moderate this relationship (Chen and Miller, 2013; Torres and Nowson, 2007). Adverse health effects of psychosocial stress may also be directionally influenced and amplified by an unhealthy diet (Laugero et al., 2011). Americans report some of the highest perceived stress levels in the world (Gallup, 2019) and many consume a Western diet rich in animal protein, saturated fat, salt, and sugar which contribute to increased disease risk (Laugero et al., 2011; U.S. Department of Agriculture, 2016).

Chronic psychosocial stress could be responsible, in part, for the epidemic of obesity and metabolic syndrome (Moore and Cunningham, 2012). Stress activates the hypothalamic-pituitary-adrenal (HPA) axis and sympathetic nervous system (SNS), which mobilize energy through cortisol and epinephrine release, respectively (Harris, 2015). These mechanisms may become maladaptive under chronic stress conditions, resulting in persistent HPA and SNS activity, which may promote fat deposition, insulin resistance, and inflammatory processes (Kelly and Ismail, 2015), thereby increasing risk of obesity, type 2 diabetes, and cardiovascular disease (Brotman et al., 2007; Brunner, 2017; Chandola et al., 2010; Rosmond, 2005; Schneiderman et al., 2005; Steptoe and Kivimäki, 2012). Thus, there are plausible biological mechanisms through which chronic psychosocial stress may promote obesity, metabolic syndrome and type 2 diabetes (Moore and Cunningham, 2012).

Prospective studies suggest that chronic stress is an important risk factor for the metabolic syndrome (Chandola et al., 2006) and type 2 diabetes (Kelly and Ismail, 2015) particularly in middle-aged women (Harris et al., 2017; Heraclides et al., 2009; Williams et al., 2013). While obesity is an established risk factor for type 2 diabetes, this relationship appears strongest in low SES individuals suggesting that stress may increase the likelihood of disease development in obesity (Volaco et al., 2018). However, literature addressing the role of psychosocial stress in type 2 diabetes is limited and contradictory (Pouwer et al., 2010). Furthermore, it is difficult to establish a causal relationship between stress and diabetes in human studies because stressors are self-reported, often retrospectively, often confounded by race, diet, education and health behaviors, and may be distant in time from metabolic impacts.

The use of a NHP model disentangles many of these confounding influences of metabolic disease in low SES individuals as all animals have equal nutrition, access to veterinary care, and housing. Several nonhuman primate (NHPs) species are useful models of psychosocial stress effects on cardiometabolic health (Harris, 2015; Shively et al., 2009b). In female cynomolgus macaques (Macaca fascicularis) consuming a Western diet, low social status is accompanied by behavioral and physiological indicators of stress and increases in visceral fat deposition, resulting in characteristics of the metabolic syndrome, e.g. dyslipidemia and hyperinsulinemia, with perturbed autonomic and HPA function (Jayo et al., 1993; Shively et al., 2009b), and exacerbated coronary artery atherosclerosis (Adams et al., 1985; Kaplan et al., 1984; Kaplan et al., 1982; Shively and Clarkson, 1994; Shively et al., 2009a). Likewise, macaques, as well as other Old World monkeys, have been widely studies as models of type 2 diabetes mellitus (Wagner et al., 2006). Like humans, the disease is most common in older, obese NHPs. Before developing overt diabetes, NHPs exhibit a period of insulin resistance that is initially met with compensatory insulin secretion. When either insulin resistance or a deficiency in pancreatic insulin production occurs, fasting glucose concentrations begin to rise and other diabetic signs, such as increases in circulating triglycerides, begin to emerge (Wagner et al., 2006). Pathological changes in pancreatic islets are also similar to those seen in human diabetics (Reaven, 1988). However, whether and how social stress increases incidence of type 2 diabetes is unknown. Given these characteristics of the NHP model, we sought to determine the relationships between the psychosocial stress and cardiometabolic profiles. We hypothesized that socially subordinate middle-aged female cynomolgus macaques would demonstrate increased obesity, perturbed autonomic function, and impaired glucose metabolism consistent with early type 2 diabetes, compared to their dominant counterparts.

2. Materials and Methods

2.1. Animal Subjects

Forty-two middle-aged female cynomolgus monkeys, imported from Indonesia (Institut, Pertanian Bogor, Bogor, Indonesia), were quarantined for one month and then randomly assigned to social groups of n=4–5, and housed in indoor pens (3.05m × 3.05m × 3.05m) with 12/12 light/dark cycle (0600–1800h) and water available ad libitum. Average age was estimated by dentition to be 15.7 (± 0.3) years, which is approximately equivalent to a human age of 45 years. All animal manipulations were performed according to the guidelines of state and federal laws, the US Department of Health and Human Services, and the Animal Care and Use Committee of Wake Forest University School of Medicine. Wake Forest University is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

2.2. Experimental Design (Supplementary Figure 1)

These animals were subjects of a randomized preclinical trial to determine the effects of selective serotonin reuptake inhibitor (SSRI) treatment on depressive behavior and coronary artery atherosclerosis. After a 2-month baseline phase, stratified randomization was used to assign the monkeys by social group [dominant (n = 21), subordinate (n = 21)] as previously described (Shively et al., 2015). SSRI effects on depressive behavior, metabolism, and coronary artery atherosclerosis were reported previously (Shively and Day, 2015; Silverstein-Metzler et al., 2016). Social status effects have not been previously reported. There were no significant interaction effects of social status and SSRI treatment on any outcomes. Thus, here we focus on social subordination stress, stress physiology, and metabolic characteristics relevant to type 2 diabetes.

During the 3-year study, monkeys were fed a semi-synthetic diet designed to mimic a typical Western diet, containing 44% of calories from fat and 0.29 mg/Cal cholesterol, which is comparable to a human consumption of 500 mg cholesterol/2000 calories, and roughly comparable to 2 eggs per day (Groban et al., 2014). After 1.5-years of Western diet consumption the monkeys were randomized by social group to either placebo (n=21) or SSRI (n=21) treatment for 1.5 years as previously described (Shively et al., 2015).

2.3. Behavior Observations

Behavior was recorded during 10 minute focal animal observations, 6–8 times per month, counterbalanced for time of day, for two 12-month periods, once during the first 1.5-years, and once during the last 1.5 years (an average of 33.3 hours/monkey total) (Shively et al., 1986).

2.3.1. Social Behavior

Aggressive behaviors recorded included open mouth threat, stare threat, yawn threat, chase, bite, slap/grab, displace, and cage shake display. Submissive behaviors included fear grimace, lip smack, move away, crouch, flee, present hindquarters, scream, and scream threat (Shively et al., 1986). Average rates of these behaviors were determined for the first 1.5-years on diet and from 1.5 to 3-years on diet and used in analysis.

Each month, the outcomes of aggressive interactions were used to determine social status. The number of animals from which each female elicited submissive responses was determined. To account for differences in group size, relative ranks, which ranged from 0–1, were calculated as: (number of animals that submit)/ (number of animals in social group-1). Relative ranks were then averaged for each monkey for the entire experimental period. As observed previously, social status was stable over the course of the experiment (Shively and Kaplan, 1991). Those monkeys having an average relative rank greater than 0.5 were designated as dominants and the rest were considered subordinate.

2.4. Cortisol Response

Dexamethasone suppression tests were performed at baseline and after 14 and 34 months on the Western diet (Shively et al., 2005). A morning blood sample was taken for basal serum cortisol determinations, an evening dose of dexamethasone (0.13 mg/kg) was administered, and a second blood sample was obtained the next morning for cortisol concentration determination. Blood samples were taken within nine minutes of capture and sedation with ketamine hydrochloride (10–15 mg/kg). Percent suppression, an indicator of sensitivity to negative feedback, was calculated as the difference between the two cortisol measures divided by the basal value and multiplied by 100 (Shively, 1998). Basal cortisol and percent suppression were used as dependent variables (Kaplan et al., 1986).

2.5. Anthropometrics

Body weight (BW) and trunk length were measured at baseline and after 15 and 35 months on the Western diet. Body mass index (BMI) was estimated as the ratio of BW to the square of trunk length measured from the suprasternal notch to the pubic symphysis (in kg/m²) (Shively et al., 2009b).

2.6. Body Composition and Fat Distribution

Body composition and fat distribution were determined at baseline and after 16 and 35 months on the Western diet. Body fat and lean mass were determined using Dual X-ray Absorptiometry (DEXA whole-body scans, Hologic Discovery A Dual Xray Bone Densitometer, Bedford, MA) of anesthetized animals with manufacturer developed protocols. Percent fat and lean mass were calculated as a percent of whole body mass (Shively et al., 2019). Subcutaneous and visceral abdominal fat volumes were determined from whole body computed tomography (CT) scans collected with a 32 slice multi-detector CT scanner (Toshiba America Medical Systems Inc. Tustin, CA) as previously described (Silverstein-Metzler et al., 2016).

2.7. Carbohydrate and Lipid Metabolism

After an overnight fast, blood samples were collected at baseline and after 15 and 35 months on the Western diet to determine fasting insulin, glucose, and triglyceride concentrations. Glucose and triglyceride concentrations were determined by colorimetric assay using reagents (ACE-GLU and ACE-TG) and instrumentation (ACE ALERA autoanalyzer) from Alfa Wasserman Diagnostic Technologies (West Caldwell, NJ), and insulin was determined by enzyme-linked immunosorbent assay as previously described (Shively et al., 2015; Silverstein-Metzler et al., 2016). The homeostasis model assessment (HOMA-IR = [mg/dL fasting glucose × mIU/L fasting insulin]/405) was used as a measure of insulin resistance (Bonora et al., 2000). In monkeys, fasting glucose concentrations are about 20 to 30 mg/dL lower than humans’ (average human basal glucose 85 mg/dL; (Merimee and Tyson, 1977); thus impaired glucose metabolism indicative of diabetes was defined as a fasting glucose > 100 mg/dl (Wagner et al., 2006).

2.8. Activity by Actigraphy

At baseline and after 15 and 35 months on the Western diet, activity was assessed by recording movement via accelerometry as previously described (ActiGraph GT3X Triaxial Activity Monitor and ACTILIFE-Desktop Software, Pensacola, FL) (Shively, 1998; Silverstein-Metzler et al., 2016). Average 24-hour activity and night-time activity (2400–400h), a surrogate for sleep disruption, were determined for analysis.

2.9. Autonomic Function

We used heart rate telemetry to determine heart rate variability (HRV) at baseline and after 15 and 35 months on the Western diet. We measured HRV at night (0100–0300) during the nadir of activity for these diurnal NHPs. Moreover, the study subjects were undisturbed during this period; thus, their physiology was relatively independent of extraneous factors that might influence heart rate, such as locomotion or the presence of caretakers. We used Nevrokard-HRV software (Nevrokard Kiauta, d.o.o. Izola Slovenia) to determine six HRV parameters defined in Table 1, representing both the time (SDNN and RMSSD) and frequency domains (VLF, LF, HF, and LF/HF) (Shively et al., 2020). Time-domain measurements indicate variability in amount of time between successive heart beats (a.k.a. inter-beat intervals (IBI)) after artifacts have been removed, whereas the frequency-domain indices reflect the relative power (i.e., signal energy) of spectral frequency bands (Shaffer and Ginsberg, 2017). Time-domain HRV measurements are inversely correlated to heart rate (Kazmi et al., 2016), and we assessed heart rates (beats per minute) to confirm this relationship. The standard deviation of successive IBIs (SDNN) from short-term, resting recordings reflects parasympathetic nervous system (PNS) activity (Shaffer and Ginsberg, 2017), whereas the root-mean square of successive differences (RMSSD) is primarily used to estimate heart rate changes mediated by the vagal nerve (Shaffer et al., 2014). While the very low frequency (VLF) band is modulated primarily by the SNS, the low frequency (LF) band may reflect a combination of baroreflex (i.e., vagal) and sympathetic activity (Shaffer et al., 2014). The high frequency (HF) band reflects the activity of the PNS. LF/HF ratio represents sympathovagal balance (Shaffer and Ginsberg, 2017; Shaffer et al., 2014; Shively et al., 2007).

Table 1.

Heart rate variability (HRV) parameters (Shaffer and Ginsberg, 2017; Shaffer et al., 2014).

HRV Parameter	Description	Interpretation
Time Domain Measures
SDNN	Standard deviation of successive inter-beat intervals of normal sinus beats	Short-term, resting recordings reflect parasympathetic nervous system activity
RMSSD	Root mean square of successive differences between normal heartbeats	Estimates heart rate changes mediated by the vagal nerve
Frequency Domain Measures
VLF	Relative power of the very low frequency band (range: <0.01Hz)	Primarily modulated by the sympathetic nervous system
LF	Relative power of the low frequency band (range: 0.01–0.2 Hz)	Reflects a combination of baroreflex (i.e., vagal) and sympathetic activity
HF	Relative power of the high frequency band (range: 0.20–0.80 Hz)	Reflects parasympathetic nervous system (PNS) activity
LF/HF	Ratio of LF-to-HF power	Represents sympathovagal balance

3.0. Statistical Analysis

Baseline characteristics were assessed via Student’s T-test to confirm no differences in pre-experimental metabolic outcomes. To determine social status effects over time, variables were analyzed using 2 (dominant, subordinate) × 2 (placebo, SSRI) × 2 (time) mixed model analysis of variance (ANOVA). Log transformation was used to equalize between group variances. As SSRI effects have been previously reported (Shively et al., 2015; Silverstein-Metzler et al., 2016), and there were no significant interactions of social status with SSRI treatment on any outcomes. Social status and social status × time effects are reported here. In the case of significant status or status × time interactions, Fisher’s protected least significant difference tests for post hoc comparisons were used to determine social status differences at each time point. All p-values <0.1 are reported with statistical significance set at p ≤ 0.05 (Wasserstein and Lazar, 2016). The raw data are shown as dot-plots with bars indicating the mean and SEM and are accompanied by bar graphs of the adjusted means after back transformation to original units.

3. Results

3.1. Baseline Characteristics (Supplementary Table 1)

Baseline characteristics were assessed via Student’s T-test to confirm no differences in pre-experimental metabolic outcomes. As expected, only indicators of social subordination stress - basal cortisol (T(40)=−2.45, p=0.02), submissions sent (T(40)=−4.76, p<0.001), and aggression received (T(40)=−5.23, p<0.001) were significantly higher on subordinates compared to dominant animals (Shively and Kaplan, 1984; Shively, 1998; Shively et al., 2009a; Shively et al., 2009b). All other p-values were >0.05.

3.2. Social Status Differences in Behavior, Cortisol, and Nighttime Activity (Figure 1)

Figure 1.

Social Status Differences in Behavioral and Physiological Indicators of Stress. Subordinates were more submissive (A), received more aggression (B) and had significantly higher morning cortisol levels (C) than dominants. Nighttime activity (D), a surrogate for sleep quality, increased with time in subordinate animals only resulting in a significant status × time effect. Left panels depict the distribution of raw data with bars indicating mean and SEM; right panels depict means. Submissions sent, aggression received, and nighttime activity were log transformed for analysis and back-transformed to original units for graph. *p<0.05, ** p≤0.01.

Subordinates had higher rates of submissive behavior (Figure 1A: status F[1,38]=100.0, p<0.001) and received more aggression (Figure 1B: status F[1,38]=86.1, p<0.001) than dominants. There were no effects of time or status × time on submissive or aggressive behavior (p’s>0.05), although aggression in subordinates tended to decrease over time (time × status F[1,38]=3.03, p=0.062). Subordinates also had significantly higher morning cortisol levels than dominants (Figure 1C: status F[1,38]=9.74, p=0.003). There were no effects of time or status × time on basal cortisol (p’s>0.05). There were no effects of status, time, or their interaction on cortisol percent suppression following dexamethasone administration (p’s>0.05; not shown). There was no significant effect of status or time on night time activity (Figure 1D); however, there was a significant status × time interaction (F[1,38]=7.23, p=0.011). After 3 years on the Western diet, subordinates had more sleep disturbances, as indicated by night-time activity, than dominants (p=0.025). There were no status effects on 24-hour activity (p’s>0.05; not shown).

3.3. Body Weight and Composition (Supplementary Figures 2 and 3)

After 3 years, subordinates had the highest BWs, whole body fat mass, and abdominal fat volumes, on average; however, there were no significant effects of status, or status × time interactions, on BW, BMI, whole body fat mass, whole body lean mass (Supplementary Figure 2), volumes of total abdominal fat, visceral abdominal fat, subcutaneous abdominal fat, or the ratio of visceral to subcutaneous fat (Supplementary Figure 3; all p’s>0.05). In all animals, whole body fat mass increased (time F[1,38]=28.9, p<0.01), while whole body lean mass decreased (time F[1,38]=14.1, p<0.01) over time. There were no effects of time on BW, BMI, volume of total abdominal fat, visceral abdominal fat, subcutaneous abdominal fat, or the ratio of visceral to subcutaneous fat (all p’s>0.05).

3.4. Social Status Differences in Carbohydrate Metabolism (Figure 2)

Figure 2.

Social Status Differences in Carbohydrate Metabolism. After 3-years of Western diet consumption, triglyceride levels (A) significantly increased; while concentrations were higher in subordinates, the status × time interaction did not reach significance (p=0.07). Fasting glucose concentrations (B) significantly increased over time in all animals, and this increase was most pronounced in subordinates (significant main effect of time and status × time interaction). Fasting insulin (C) did not differ by social status, time, or their interaction; however, HOMA-IR (D) was significantly higher in subordinates and increased over time (significant effects of status, time, and status × time). After 3-years of Western diet consumption, the incidence of impaired glucose metabolism, defined as a fasting glucose > 100 mg/dl, was greater in subordinates (23%) than dominants (0%) (E). A-D variables were log transformed for analysis. Left panels represent raw data, right panels represent means back-transformed to original units. *p<0.05, **p≤0.01.

There was no main effect of status (p>0.05) on fasting triglycerides (Figure 2A), however, triglyceride concentrations increased in all animals over time (main effect of time: F[1,38]=12.4, p<0.01). There was a trend toward a or status × time interaction (F[1,38]=3.52, p=0.068) suggesting that subordinates drove the time effect. At the 3-year time point, triglyceride concentrations were about 50% higher, on average, in subordinates than dominants. There was a main effect of time (F[1,38]=7.72, p<0.01) on fasting glucose (Figure 2B). Subordinates tended to have higher fasting glucose concentrations, but the main effect of status did not reach significance (F[1,38]=3.80, p=0.059); however, there was a significant status × time interaction (F[1,38]=5.87, p=0.02). Fasting glucose concentrations were about 25% higher in subordinates than dominants after 3 years on diet. There were no significant effects of time, status, or their interaction on fasting insulin concentrations (Figure 2C; all p’s>0.05). However, HOMA-IR (Figure 2D), a calculated measure of insulin resistance, was significantly greater in subordinates (status: F[1,38]=4.24, p=0.046). HOMA-IR significantly increased over time (F[1,38]=4.72, p=0.036) but the time × status interaction was not significant (p>0.05). The incidence of impaired glucose metabolism (fasting glucose > 100 mg/dl) at study end was significantly greater in subordinates (23%) than dominants (0%) (Fisher’s exact test, p=0.047, Figure 2E).

3.5. Social Status Differences in Heart Rate Variability (Figure 3)

Figure 3.

Social Status Differences in Heart Rate Variability. Subordinates had significantly lower SDNN (A) and RMSSD (B) than dominants. Left panels depict the distribution of raw data with bars indicating mean and SEM. Variables were log transformed for analysis. Right panels depict means back-transformed to original units. *p<0.05, **p≤0.01.

The heart rate telemetry signal failed in two animals. Thus, the final sample was N=40. There were main effects of time for all measures in the frequency and time domains (VLF: F[1,36]=51.5, p<0.001; LF: F[1,36]=31.1, p<0.001; HF: F[1,36]=93.6, p<0.001; LF/HF: F[1,36]=78.4, p<0.01; SDNN: F[1,36]=5.83, p=0.021; RMSSD: F[1,36]=21.4, p<0.01). VLF, LF and the LF/HF ratio increased, whereas HF, SDNN, and RMSSD decreased with time, suggesting a relative increase in sympathetic and decrease in parasympathetic inputs, and decreased HRV with time. Dominants did not differ from subordinates in the frequency-domain parameters (Supplementary Figure 4; all p’s>0.05). There were significant main effects of status for SDNN (Figure 3A; F[1,36]=10.645, p=0.002) and RMSSD (Figure 3B; F[1,36]=6.004, p=0.019) such that dominants had higher time domain measures than subordinates. Dominants also had lower heart rates, on average, than subordinates, although this difference was not statistically significant (F[1,36]=2.889, p=0.098). There were no significant interaction effects (all p’s>0.05).

4. Discussion

This is the most comprehensive, controlled longitudinal assessment of the effects of psychosocial stress on obesity, carbohydrate metabolism, and autonomic function in primates available to date. The animals consumed a diet similar to that consumed in Western countries, lived in stable social groups, and experienced mild, chronic psychosocial stress due to social subordination. Our assessment of psychosocial stress in this study was comprehensive, including quantification of stress experienced (aggression), and behavioral (submission), and physiological (autonomic nervous system, HPA function, and sleep disturbance) responses to that stress. After 3 years consumption of a Western diet, which approximates a 10 year period in humans, cardiometabolic disease risk factors were increased in socially stressed subordinates compared to dominants; including higher fasting glucose concentrations, greater insulin resistance (HOMA-IR), and a higher incidence of clinically impaired carbohydrate metabolism.

These results were achieved using a well-studied model of social subordination stress. Previously, we have shown that, compared to dominant female macaques, subordinates receive more aggression, are groomed less, respond to a standardized stressor with higher heart rates, have higher basal cortisol levels, secrete more cortisol in response to an adrenocorticotropin challenge, and have heavier adrenal glands than dominants – all behavioral and physiological indicators of stress (Shively and Kaplan, 1984; Shively, 1998; Shively et al., 2009a; Shively et al., 2009b). Several of these behavioral and physiological indicators of stress were recapitulated in this study. Similar to previous studies, we did not find a social status difference in body weight or BMI (Kaplan et al., 2010; Kaplan et al., 2002), likely because the feeding protocol in this and previous studies was designed to minimize differential access to food by feeding in multiple sites and to 10% excess. However, subordinates received more aggression, displayed more submission, and had higher basal cortisol levels compared to dominants (Shively and Willard, 2012). Disturbances in the HPA axis are also characteristic of people reporting high levels of psychosocial stress (Kunz-Ebrecht et al., 2004; Sandström et al., 2012; Sjörs et al., 2014).

Likewise, psychosocial stress is associated with disturbed sleep. Cross-sectional clinical studies have shown that stress is associated with shortened sleep, sleep fragmentation, and possibly disruption of normal sleep stages (Âkerstedt, 2006). Sleep is essential for physiological processes that support long-term physical and mental health. Impaired sleep is associated with elevated cortisol levels (Morgan et al., 2017) and temporal differences in salivary cortisol concentrations (Castro-Diehl et al., 2015), which may perpetuate the effects of chronic stress on adverse health outcomes. Indeed, prior sleep disturbances predict the development of diabetes and cardiovascular disease in longitudinal studies in humans (Âkerstedt, 2006; Vgontzas et al., 2008) and macaques (Zhao et al., 2021). Sleep architecture is difficult to capture in socially housed NHPs. However, accelerometry has been used as a non-invasive index of sleep, inferring sleep from periods of inactivity tends to overestimate the amount of time slept during the night and underestimate sleep disruptions (Qin et al., 2020; Sadeh, 2011). Here we observed that after 3 years consuming a Western diet, sleep quality was perturbed in subordinates, but not dominants, between 2400 and 0400h, the nadir of the heart rate and activity circadian rhythm in NHPs on a 0600 to1800 light/dark cycle (Shively, 1998). Interestingly, sleep quality was not perturbed at 1.5 years suggesting that sleep disruption may be due to the cumulative physiological effects of chronic psychosocial stress, on a Western diet background.

HRV is also sensitive to psychosocial stress and may represent a key mechanism linking psychosocial stress and cardiometabolic risk. Dominant monkeys had higher HRV (SDNN and RMSSD), and tended to have lower heart rates, than subordinates. This inverse relationship between HRV and heart rate is also observed in humans (Kazmi et al., 2016). Parasympathetic cardiovascular modulation is stronger during deep stages of sleep, including non-rapid eye movement sleep (NREM), than during lighter sleep stages and wakefulness (Boudreau et al., 2013). Thus, lower nighttime HRV is consistent with sleep disruption or failure to achieve deeper sleep stages. Dominant females, having higher HRV, therefore likely had higher parasympathetic activation at night. These results may have clinical implications as low nighttime HRV is associated with elevated cardiovascular disease risk (Jarczok et al., 2019). Our results suggest that lower nighttime HRV may mediate the relationship between psychosocial stress and elevated cardiometabolic disease risk.

Triglyceride and glucose concentrations, and HOMA-IR all increased over time likely reflected aging and cumulative exposure to a maladaptive diet in these middle-aged monkeys. The 1.5 year interval between first and second measures is about equal to a 5 year interval in humans.

Similar to middle-aged humans, we observed decreased lean mass and parasympathetic activity, and increased body fat, fasting glucose and triglyceride concentrations, insulin resistance, and sympathetic activity as the animals grew older. These observations support the use of middle-aged NHPs in studies of aging. Notably, increases in glucose, and HOMA-IR were greater in subordinates. The higher HOMA-IR observed in subordinates indicates that they were relatively insulin resistant. Finally, impaired glucose metabolism, defined as fasting glucose > 100 mg/dl (Wagner et al., 2006), was more prevalent in subordinates (23%) than dominants (0%) after three years of Western diet consumption. The emergence of impaired carbohydrate metabolism after 3 years in subordinate animals is especially interesting given the lack of statistically significant status differences in adiposity. The concept of metabolically healthy obesity is supported by the observation that some individuals have excess adiposity in the absence of metabolic perturbation (Smith et al., 2019). Observations made here suggest that metabolically healthy obesity may be observed in non-stressed individuals, whereas psychosocial stress may promote metabolically unhealthy obesity (Kavanagh et al., 2017).

This is the first report of a significant social status difference in carbohydrate metabolism in NHPs. Here we present evidence for a cumulative effect of social subordination stress and prolonged exposure to a Western diet on several cardiometabolic risk factors. Despite the comprehensive analysis of cardiometabolic risk factors relevant to type 2 diabetes in a highly translational model of middle-aged women’s health, there are limitations to this study. First, the cohort of monkeys used for this study were subjects of a randomized preclinical trial to determine the effects of selective serotonin reuptake inhibitor (SSRI) treatment on depressive behavior and coronary artery atherosclerosis. To control of this, SSRI-treatment was included in statistical analysis and no status × treatment effects were identified. All animals ate the same diet, designed to recapitulate key characteristics of the human Western diet. Thus, interactions of diet composition and social status were also not addressed. Likewise, middle-aged females were studied because, among adults, the overall prevalence and number of new diabetic cases increases with age (Control and Prevention, 2014; Geiss et al., 2018), and because stress sensitivity and stress effects on physiology and health are more widely reported in females than males (Moisan, 2021). Thus, age and sex differences could not be addressed. Finally, social status was observed, not manipulated; thus differences between dominants and subordinates reported here are associations. Finally, further exploration is needed of the novel finding of perturbed sleep in subordinates and its role in the development of cardiometabolic risk factors and type 2 diabetes.

5. Conclusion

Impaired carbohydrate metabolism in socially subordinate NHPs was observed in the absence of significant increases in body weight or abdominal obesity. These findings suggest that chronic psychosocial stress on a Western diet background may promote impaired carbohydrate metabolism and type 2 diabetes risk in middle-aged female primates.

Highlights.

• Chronic psychosocial stress significantly increased type 2 diabetes risk

• Psychosocial stress may promote metabolic unhealthy obesity

• Chronic psychosocial stress and western diet may impair parasympathetic activation