Differential Effects of Western versus Mediterranean Diets and Psychosocial Stress on Ovarian Function in Female Monkeys (Macaca fascicularis)

Abstract

Ovarian dysfunction increases risk for chronic diseases of aging including cardiovascular disease, depression, cognitive impairment, as well as bone and muscle loss which promote frailty. Psychosocial stress can disrupt ovarian function, and recent observations suggest that consumption of a Western Diet may also. Determination of causal relationships among diet, psychosocial stress, and ovarian physiology is difficult in humans. Long-tailed (a.k.a. cynomolgus) macaques (Macaca fascicularis) are an excellent translational model for the study of diet and psychosocial effects on ovarian physiology and aging-related processes. They have 28-day menstrual cycles with hormonal fluctuations like those of women, and similar physiologic responses to alterations and/or cessation of cyclicity. We examined ovarian function in 38 middle-aged socially housed females fed either a Western or Mediterranean diet for 31 months (≈ a 9-year period for humans). During the last year, we examined cycle length and peak progesterone per cycle using blood sampling (3/week) and vaginal swabbing for menses (6/week). Repeated measures analysis revealed a circannual pattern consistent with increased menstrual cycle disturbance during the late Summer and early Fall (F(11,348)=4.05 p<0.001). In addition, both Western diet (F(1,34)=3.99; p=0.05) and the stress of low social status (F(1,34)=3.99; p=0.04) reduced mean progesterone levels. Thus, on average, subordinates in the Western group had the lowest average progesterone levels (10.02 ng/pl). Compared to Western diets, Mediterranean diets exhibited protective effects via menstrual cycle regularity. For dominant monkeys, consuming Mediterranean diets resulted in significantly greater likelihood of having regular menstrual cycles. Mediterranean diets also protected individuals from shorter than normal menstrual cycles. The relationships between diet and menstrual regularity were partially mediated by both adrenal reactivity and social isolation. This study demonstrates the additive negative effects of poor diet and psychosocial stress on ovarian physiology in mid-life and lays the groundwork for future investigations to uncover their impact on metabolic signatures of accelerated aging. The results also suggest that – compared to Western-style diets - a Mediterranean diet may exert a protective influence against ovarian dysfunction and its pathologic sequelae.

1. Introduction

Menopause represents a period of heightened risk for developing several chronic diseases of aging, including cardiovascular disease (CVD) (Atsma et al., 2006; Honigberg et al., 2019; Price et al., 2021), depression (Bromberger and Epperson, 2018), cognitive impairment (Rocca et al., 2009), osteopenia and osteoporosis (Li and Wang, 2018), and sarcopenia (Geraci et al., 2021). While the symptoms of such conditions often manifest during or after menopausal transitions, their origins may be premenopausal (Kaplan and Manuck, 2004, 2008). Ovarian disruption – including irregular cycles and hormonal insufficiencies – may accelerate the development of pathological post-menopausal conditions. For example, menstrual irregularity is associated with a nearly 50% increased risk of developing coronary heart disease (Solomon et al., 2002), and women with premature ovarian insufficiency, a condition in which women lose ovarian function before the age of 40, are at increased risk of developing CVD, osteopenia, and neurological disorders (Podfigurna-Stopa et al., 2016). Thus, even mild-to-moderate ovarian disruptions may increase disease risk in later life, and the premenopausal period may represent a critical window in the etiology of many chronic diseases of aging.

Several environmental factors are known to disrupt ovarian function, including contaminants (Woodruff et al., 2008), heat stress (Abdelnour et al., 2020), cigarette-smoking (Van Voorhis et al., 1996), and psychosocial stress (Albert et al., 2015; Kaplan, 2008; Roney and Simmons, 2015). Diets high in sugar, animal protein, saturated and trans fats, and other constituents also may harm ovarian health (Fontana and Torre, 2016; Jurczewska and Szostak-Węierek, 2022). Conversely, consumption of healthier diets may exert protective effects (Garruti et al., 2019; Kazemi et al., 2020). Diet therefore may provide a therapeutic strategy for improving ovarian function in reproductive-aged women, which, in turn, might delay or prevent the pathobiological sequelae of ovarian dysfunction.

Despite the need to better understand the role of diet in shaping ovarian health, significant gaps in knowledge exist due to several factors. First, most inferences have been drawn from retrospective self-reporting of menstruation (Solomon et al., 2002) and dietary patterns (Hodge et al., 2013). Long-term feeding trials with frequent menstrual diary check-ins would be helpful, however dietary interventions are costly, and study subjects may vary in their adherence to prescribed diets (Scotto et al., 2011). Participants also may inaccurately report lifestyle characteristics – e.g., physical activity, substance abuse, or trauma – that impact ovarian function. Study enrollment may fail to incorporate participants with limited access to healthcare settings, particularly those facing socioeconomic challenges (O’Neil et al., 2020; Saphner et al., 2021). Finally, disentangling the effects of diet versus associated phenotypes (e.g., obesity or hypercortisolism (Janssen, 2022)) on ovarian health can be challenging (Kazemi et al., 2020). Indeed, considerable variation exists in the occurrence of metabolic abnormalities across individuals considered obese (Blüher, 2020; Peeke and Chrousos, 1995).

Dietary patterns are also not equally represented across socioeconomic groups, which may vary in their exposure to psychologic stressors. Individuals with low incomes are more likely to consume unhealthy diets (Appelhans et al., 2012; Darmon and Drewnowski, 2008; Pechey et al., 2013; Pechey and Monsivais, 2016) and experience the highest levels psychological stress (Baum et al., 1999; Cundiff et al., 2022). Such psychogenic stressors may interact with dietary patterns to disrupt ovarian physiology, further complicating the study of diet effects on ovarian health. Short-term, crossover studies suggest that Westernized-style diets intensify physiologic stress responses (Jakulj et al., 2007), whereas adherence to healthier diets may provide resilience against stressful stimuli (Bonaccio et al., 2018; Crichton et al., 2013; Hodge et al., 2013). Reactivity to stressful stimuli may alter function to both the hypothalamic-pituitary-adrenal (HPA) as well as the hypothalamic-pituitary-gonadal (HPG) axes.

Randomized preclinical studies provide important opportunities to overcome some of the obstacles facing clinical research. Long-tailed (a.k.a. cynomolgus) macaques (Macaca fascicularis) are excellent nonhuman primate (NHP) models to study the effects of diet and psychosocial stress on ovarian physiology. Similar to women, long-tailed macaques have year- round 28-day (± 2 days) menstrual cycles, fluctuations of gonadotropin and sex hormones, and relative premenopausal resistance to cardiovascular disease and osteoporosis (Kaplan et al., 1984; Kaplan and Manuck, 2004). Macaques respond to Western diets like humans, developing coronary and carotid atherogenesis, elevated sympathetic activity, higher anxiety, blunted HR responses to acute stressors, elevated cortisol reactivity, increased body fat and insulin resistance, and hepatosteatosis (Johnson et al., 2022; Shively et al., 2020; Shively et al., 2019; Shively and Clarkson, 1994; Shively et al., 1989; Shively et al., 2009b). Macaque social environments, like humans, are characterized by social inequity (Shively and Day, 2015), with low-ranking individuals exhibiting relatively greater degrees of physiological stress responses compared to their high-ranking counterparts (Shively et al., 2020). The psychosocial stress of social subordination causes ovarian dysfunction in these animals (Kaplan et al., 2010; Shively and Clarkson, 1994). However, many studies of psychosocial stress effects on NHP ovarian function have been conducted on a background of laboratory chow diet, a diet for which there is no human equivalent (Supplemental Table 1). Thus, the findings of such work may not translate with great fidelity to humans.

Here we determined the relationships between psychosocial stress, diet, and ovarian physiology in a well-established model of women’s health, the long-tailed macaque (M. fascicularis) (Clarkson and Mehaffey, 2009; Cline and Wood, 2009; Kaplan et al., 2009; Register, 2009; Shively and Clarkson, 2009; Shively and Day, 2015; Shively et al., 2009a; Shively et al., 2009b). Since both Western diets and psychosocial stress may instigate multi-system disturbances (Frye et al., 2020; Johnson et al., 2020; Newman et al., 2021; Shively et al., 2020; Shively et al., 2019; Shively and Day, 2015), which result in ovarian dysfunction, we also disentangled the effects of diet and status from obesity, adrenocortical reactivity, and social isolation via mediation analyses. Given previous work in this model (Shively and Day, 2015), we hypothesized that socially subordinate monkeys would exhibit more ovarian dysfunction than socially dominant monkeys. We also hypothesized that consumption of a Mediterranean-like diet would provide protection from psychosocial stress effects relative to a Western-like diet, and that psychosocial stress measures would mediate the effect of diet on ovarian function. We confirmed that the stress of low social status disrupted ovarian function, discovered that subjects consuming Mediterranean diets exhibited healthier ovarian profiles than those consuming Western diets, and that these diet-associated differences were partially mediated by adrenal reactivity and social isolation. This study demonstrates the additive negative effects of poor diet and psychosocial stress on ovarian health, which, in turn, may increase the risk of aging-related chronic diseases later in life.

2. Material and methods

2.1. Subjects

The study sample consisted of 38 adult premenopausal (age: mean = 9.0, range = 8.2-10.4 years, estimated by dentition) female long-tailed (a.k.a. cynomolgus) macaques (Macaca fascicularis) living in small (N = 4-5) social groups at Wake Forest University School of Medicine (Winston-Salem, NC). Monkeys were housed indoors in 3m X 3m X 3m enclosures with exposure to sunlight and a 12/12 light/dark cycle (light period: 0600–1800). During the Pre-treatment phase (~7 months), all subjects consumed standard monkey chow diet (120kcal/kg/day) (Supplemental Table 1) and water ad libitum. After the Pre-treatment phase, individuals were randomly assigned to either the Western or Mediterranean experimental diet groups for 31 months (Supplemental Figure 1). The diet groups were balanced on biomarkers relevant to overall health: body weight, body mass index, circulating basal cortisol, and fasting triglyceride concentrations (Shively et al., 2019). Use of the randomized trial design, particularly when matched at baseline on key characteristics as was done here, is a rigorous approach that allows for the assumption of similarity at baseline and thus causal inference. 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, and complied with the American Society of Primatologists Principles for the Ethical Treatment of Non-Human Primates. All experiments complied with the ARRIVE guidelines. Given the nature of the study, all researchers involved in the experiment were aware of the group allocation at each stage of the experiment (i.e., data collection, management, and analyses).

2.2. Behavior

All instances of agonistic and affiliative behavior were recorded in 10 min focal animal observations (Altmann, 1974) beginning in the third month of the Treatment phase as previously described (Shively et al., 2020). Focal observations were conducted twice per week, balanced for time of day. This produced 31 hours of behavioral observation for each monkey over the course of the experiment. Social isolation was defined as time spent alone, out of arm’s reach of another monkey. Social status was determined based on the outcomes of agonistic interactions (see (Shively, 1998; Shively and Day, 2015; Shively and Kaplan, 1991; Shively et al., 2009b) for details). Using these data, we generated a metric (ranging from 0 to 1) representing the proportion of social partners that were submissive to each animal. Females in the top half of their hierarchy – i.e., those that submitted to less than half of conspecifics – were considered dominant and all others subordinate. Social subordination appears stressful, as subordinates in this study received more aggression, were groomed less, spent more time fearfully scanning, had higher systolic blood pressure, and had higher sympathetic nervous system activity than did dominants (Shively et al., 2020). Thus, we used social status as a proxy for psychosocial stress.

2.3. Experimental Diets

Following Pre-treatment (i.e., chow diet), subjects consumed either a Western-like (hereafter “Western”) or Mediterranean-like (hereafter “Mediterranean”) diet for 31 months. As previously described (Shively et al., 2019), these diets were developed to mimic Western (U.S. Department of Agriculture, 2016) and Mediterranean (Estruch et al., 2018) dietary patterns observed in humans (see Supplemental Table 1 for diet compositions). The experimental diets were formulated to be isocaloric with respect to macromolecules (protein, fat, and carbohydrate) and were identical in the content of cholesterol (~320 mg/2,000 kcal/day). However, they differed in composition; fats and protein were primarily derived from animal sources in the Western diet and from plant sources in the Mediterranean diet. Thus, the Mediterranean diet was higher in monounsaturated fats and had a lower omega-6: omega-3 fatty acid ratio than the Western diet, whereas the Western diet had relatively high levels of saturated animal fats, sodium, and refined sugars. Johnson et al. (2022) and Shively et al. (2019) provide comprehensive accounts regarding feeding protocols. Individuals received 120 kcal per kg of bodyweight each day (120 kcal/kg/day).

2.4. Ovarian Function

During months 16-27 of the Treatment phase, we determined the peak luteal-phase progesterone concentration (hereafter peak progesterone (ng/ml)), cycle length (days), and cycle regularity for each menstrual cycle by vaginal swabbing for menses (6/week) and blood sampling (3/week) for progesterone determinations. For sample collections, study subjects were trained to cooperate with vaginal swabbing and femoral venipuncture by voluntary presentation (Shively and Clarkson, 1994). We evaluated ovulatory, anovulatory cycles, and progesterone-deficient cycles using previously described methods (Wilks et al., 1976). Progesterone concentrations were assessed using a commercially available radioimmunoassay kit (DiaSource Progesterone RIA-CT KIP1458, Louvain-la-Neuve, Belgium).

2.5. Adrenocortical Activity and Body Mass Index

Adrenocortical reactivity (i.e., cortisol) has demonstrated inhibitory effects on the hypothalamic-pituitary-gonadal (HPG) axis, including suppression of estrogen and progesterone secretion and increased target tissue resistance to these hormones (Chrousos et al., 1998; Kalantaridou et al., 2004; Magiakou et al., 1997). Thus, we sought to determine whether adrenocortical reactivity mediated the effects of diet and/or psychosocial stress on HPG physiology. We measured adrenocortical reactivity with an adrenocorticotropin (ACTH) challenge (Shively et al., 2020) during Treatment Month 30. Briefly, animals were fasted overnight, and then were administered dexamethasone (0.5 mg/kg). This dose suppressed HPA activity. Four hours following dexamethasone, study subjects were sedated with ketamine hydrocholoride (15 mg/kg). The initial (baseline) blood sample was collected <9 minutes after personnel entered the building. Study subjects then received a dose of adrenocorticotropic hormone (ACTH) (Cortrosyn^®, Organon, Inc., 10 ng/kg IV), followed by repeated blood sampling at 15- and 30-minutes post ACTH. The cortisol response was calculated as the cortisol area under the curve between baseline and 30 minutes post challenge.

Obesity also has been well-established as a disruptor of HPG function, with obese women being more likely to suffer ovarian disruptions manifesting as missed ovulations, hormonal perturbations, and infertility (Brewer and Balen, 2010; Hartz et al., 1979). We therefor sought to determine whether body mass index mediated the effects of diet and/or stress on ovarian physiology. Body length and body weight were measured during Treatment Month 27, and body mass index (BMI) was calculated as BW/(body length [meters])² (Jayo et al., 1993).

2.6. Statistics

Statistical analysis was performed with jamovi (The jamovi project, 2021) and R version 4.1.3 (R Core Team, 2022). Unless otherwise noted, data are presented as mean ± SEM. A value of p≤0.05 was considered statistically significant. We used general linear mixed models (GLMM) (repeated measures design with random intercepts) to assess the effects of diet (Western versus Mediterranean), social status (dominant versus subordinate), and time (Treatment month) on peak progesterone and cycle length during the Treatment phase. Post hoc comparisons were made with Bonferroni corrections. Due to the high number of potential comparisons, post hoc paired comparisons were not made on 3-way interactions. We used a mixed-effects logistic regression to assess whether social status and/or dietary group impacted the likelihood that an individual would experience an irregular menstrual cycle. The mean cycle length was 29 days; those shorter than 25 days or longer than 33 days were considered irregular cycles.

We followed the statistical approach outlined by VanderWeele (Vanderweele, 2013) to determine whether body mass index, adrenal reactivity, or social behavior mediated the effects of diet and social status on ovarian function. BMI and the area under the cortisol curve in response to ACTH challenge were used as mediating variables. Given that the Mediterranean diet reduced social isolation in these monkeys (Johnson et al., 2022), and social isolation may disrupt ovarian physiology (Rice et al., 2022), we used average rates of social isolation during Treatment months 13-24 as the mediating variable of social behavior. We investigated the following dependent variables during the Treatment phase: 1) the average peak progesterone, 2) the average cycle length, and 3) the number of irregular cycles. Poisson distributions were used in the models to analyze numbers of irregular cycles. We did not adjust analyses for multiple testing as the main objective of this portion of the study was to generate hypotheses concerning the mechanistic underpinnings of diet and/or stress effects on ovarian function.

Each mediation analysis consisted of three steps. First, we determine whether there were significant bivariate relationships between each predictor variable (diet and social status) and the average measures of ovarian physiology (i.e., average peak progesterone, average cycle length, or number of irregular cycles) during the Treatment phase. If a significant bivariate relationship was detected, we then specified two statistical models: a) the mediator model for the conditional distribution of the mediators (i.e., BMI, cortisol AUC, and social isolation) given the treatment (diet) and b) the outcome model for the conditional distribution of the outcome given the treatment and mediator. Third, we ran the mediation analysis using mediate function in R (Tingley et al., 2014). We report two decompositions of the total effect: the natural indirect effect and natural direct effect. The natural direct effect is the effect of the predictor on the outcome, absent the mediator. The natural indirect – a.k.a. causal mediation effect – represents the effect of the predictor on the outcome that is transmitted through the mediator.

3. Results

3.1. Peak Luteal Phase Progesterone Concentrations

Repeated measures GLMMs revealed significant main effects of time (F(11,348)=4.05; p<0.001), diet (F(1,34)=3.99; p=0.05), and social status (F(1,34)=4.32; p=0.04) on mean peak luteal-phase progesterone levels (Figure 1). In addition to a circannual pattern, typified by a late-summer/early-fall seasonal menstrual disturbances (i.e., the lowest luteal phase progesterone concentrations), dominant monkeys showed significantly higher mean peak progesterone levels compared to subordinates. We conducted pairwise comparisons of the estimate marginal means, correcting for multiple comparisons via the Holm method, to compare the progesterone levels of each of the four experimental groups – Mediterranean/Dominant, Mediterranean/Subordinate, Western/Dominant, and Western/Subordinate. Dominant monkeys consuming the Mediterranean diet had the highest progesterone levels, and these were significantly greater than subordinate monkeys that consumed the Western diet (t=3.02, p=0.03). The experimental means for each group: Mediterranean/Dominant=17.60 ng/ml; Mediterranean/Subordinate=12.50 ng/ml; Western/Dominant=12.50 ng/ml; Western/Subordinate=10.02 ng/ml (Supplemental Figure 2).

Figure 1.

Effects of diet, social status, and time on luteal phase progesterone over the course of the dietary intervention. Progesterone significantly varied over time, including a seasonal depression of progesterone profiles during the months of August, September, and October (Time: F(11,348)=4.05; p<0.001). Socially dominant monkeys had higher levels of progesterone than socially subordinate monkeys (Status: F(1,34)=4.32; p=0.04). Monkeys consuming the Mediterranean diets had higher progesterone profiles compared to monkeys consuming the Western diet (Diet: F(1,34)=3.99; p=0.05). Thus, dominant monkeys consuming the Mediterranean diet had the highest progesterone levels whereas subordinate monkeys consuming the Western diet had the lowest (Holm-Corrected Pairwise Comparison: t=3.02, p=0.03).

3.2. Cycle Length

There was a significant three-way interaction between time, experimental diet, and social status (F(11,343)=1.83; p=0.05; Supplemental Figure 3). Visual inspection suggested that subordinate cycle lengths were more variable than dominants, but that Western and Mediterranean diets differentially affected this relationship. We explored this three-way effect in more detail by characterizing cycle lengths as regular (29 ± 4 days) or irregular.

3.3. Cycle Regularity (Figure 2)

Figure 2.

Interaction of diet and social status on menstrual cycle regularity (Diet x Status: β=1.98 p=0.02). A) Dominant monkeys consuming the Mediterranean diet were significantly more likely to exhibit a regular menstrual cycle (25-33 days) compared to dominants consuming Western diets (p_bonf=0.02). B & C) For both dominants (Fisher’s Exact Test: p= 0.02) and subordinates (Fisher’s Exact Test: p= 0.01), consumption of the Mediterranean diet significantly reduced the number of short menstrual cycles. Sample Sizes: Western diet N=21 (10 subordinate, 11 dominant); Mediterranean N=17 (7 subordinate, 10 dominant)

On average, members of the Western group experienced 3.14 irregular cycles, compared to 1.76 in the Mediterranean group (β=−0.58; p=0.009). Using the repeated measures analyses, we detected a significant interaction effect of diet and social status on cycle regularity (Diet x Status: β=1.98; p=0.02). Pairwise post hoc comparisons with Bonferroni corrections indicated that dominant monkeys consuming Mediterranean diets were less likely to have irregular cycles compared to dominants consuming Western diets (p_bonf=0.02). All other pairwise post hoc comparisons were not significant (Figure 2A).

Next, we assessed the dominant and subordinate groups individually to determine whether diet composition influenced each group’s tendency to exhibit short (< 25 days) versus long (> 33 days) menstrual cycles using Fisher’s Exact Tests (Figures 2B & 2C). For both dominants (p= 0.02) and subordinates (p= 0.01), consumption of the Mediterranean resulted in significantly fewer short rather than longer-than normal menstrual cycles (Supplemental Figure 4). That is, for monkeys in the Mediterranean diet group, we only observed 2 short cycles for dominant monkeys and 1 short cycle in subordinate monkeys.

3.4. Bivariate Relationships Between Diet and Social Status, and Mediator Variables (Table 1)

Table 1.

Bivariate relationships between diet and social status with all outcome and mediator variables. Cortisol area under the curve (AUC) has been previously reported (Shively et al., 2020). Sample sizes: Western diet N=21 (10 subordinate, 11 dominant); Mediterranean N=17 (7 subordinate, 10 dominant)

	Diet Effects		Status Effects
	β (±SE)	p	β (±SE)	p
Outcome Variables
Average Peak Progesterone	−4.10 (±1.82)	0.03	2.82 (±1.89)	0.14
Average Cycle Length	0.83 (±2.26)	0.72	3.09 (±2.20)	0.17
Number of Irregular Cycles	0.58 (±0.22)	0.01	0.002 (±0.21)	0.99
Mediating Variables
Body Mass Index	7.35 (±3.28)	0.03	−4.09 (±3.44)	0.24
Cortisol Area Under the Curve	119.88 (±58.67)	0.05	−57.16 (±61.24)	0.34
Average Rates of Social Isolation	11.79 (±5.39)	0.04	2.29 (±5.72)	0.69

There were no statistically significant relationships between social status and the condensed measures of ovarian physiology (i.e., average peak progesterone, average cycle length, or number of irregular cycles) during the Treatment phase. Social status was also unrelated to any of the three mediators (Table 1). Thus, we did not pursue any additional mediation analyses with social status as a predictor variable or covariate. There were significant bivariate effects of diet on average peak progesterone (β=−4.10; p=0.03) and the number of irregular cycles (β=0.58; p=0.01). However, diet was unrelated to the average cycle length (β=0.83; p=0.72), so we did not conduct mediation analyses on this outcome. Regarding the mediators, the Western group had significantly higher BMIs ($\overline{\mathrm{x}}$ = 49.42) than the Mediterranean group ($\overline{\mathrm{x}}$ = 42.07) (β=7.35; p=0.03), significantly greater adrenal reactivity (Western cortisol AUC $\overline{\mathrm{x}}$ = 650.65; Mediterranean cortisol AUC $\overline{\mathrm{x}}$ = 530.77) (β=119.88; p=0.05) and spent significantly more time alone ($\overline{\mathrm{x}}$ = 38.39) compared to those consuming the Mediterranean diet ($\overline{\mathrm{x}}$ = 26.60) (β=11.79; p=0.04).

3.5. Mediation Analyses of Ovarian Function (Table 2)

Table 2.

Summary of mediation analyses. In the regressions, the main predictor variable was diet; the mediator variables were Treatment Phase body mass index (Treatment Month 27), area under the curve during adrenocorticotropin challenge (ACTH AUC) (Treatment Month 30), and social isolation (Average Rate from Treatment Months 13 to 24); and the outcome variables were average peak progesterone, average cycle length, and the number of irregular cycles during the Treatment phase. The total effect represents the effect of diet on the outcome. We report two decompositions of the total effect: 1) The direct effect, which is the effect of the predictor on the outcome, absent the mediator; and 2) the indirect effect (a.k.a. causal mediation effect) which represents the effect of the predictor on the outcome that is transmitted through the mediator. Sample Sizes: Western diet N=21 (10 subordinate, 11 dominant); Mediterranean N=17 (7 subordinate, 10 dominant) p≤0.05*

	Diet Effects
	Body Mass Index	ACTH AUC	Social Isolation
	β	β	β
Average Peak Progesterone
Natural Indirect Effect	−1.01	−0.82	−1.14
Natural Direct Effect	−3.21	−3.25	−2.90
Total Effect	−4.21*	−4.08*	−4.04*
Number of Irregular Cycles
Natural Indirect Effect	0.44	0.53*	0.45*
Natural Direct Effect	0.94	0.83	0.80
Total Effect	1.38*	1.36*	1.25*

3.5.1. Body Mass Index Was Not a Significant Mediator of the Effects of Diet on Ovarian Function

Body mass index did not mediate the relationship between diet and mean peak progesterone (Indirect Effect: β=−1.01; p=0.16) or the number of irregular cycles (Indirect Effect: β=0.44; p=0.13).

3.5.2. Mediation Effect of Adrenal Reactivity

Adrenal reactivity did not mediate the relationship between diet and mean peak progesterone (Indirect Effect: β=−0.82; p=0.21). Adrenal reactivity did partially mediate the relationship between diet and the number of irregular menstrual cycles (Figure 3A; Indirect Effect: β=0.53; p=0.04). That is, of the increase in likelihood of having an irregular cycle due to consuming a Western diet (Total Effect=1.36), an estimated 39% of that increase was due to elevated adrenal reactivity.

Figure 3.

Forest plot of mediation analyses for the partial mediation effects of A) adrenocortical reactivity and B) rate of time spent alone on the effects of diet on the number of irregular menstrual cycles during the Treatment period. Of the increased in likelihood of having an irregular cycle due to Western diet, ~39% was due to adrenocortical reactivity and ~36% of the increases were due to adrenocortical reactivity and rates of time spent alone, respectively. Sample Sizes: Western diet N=21 (10 subordinate, 11 dominant); Mediterranean N=17 (7 subordinate, 10 dominant)

3.5.3. Effects of Diet & Social Isolation

The rates of time spent alone did not significantly mediate the relationship between diet and mean peak progesterone (Indirect Effect: β=−1.14; p=0.09). However, like adrenal reactivity, the rates of time spent alone mediated effects of diet on the number of irregular cycles (β=0.45; p=0.03; Figure 3B). That is, of the increase in likelihood of having an irregular cycle due to Western diet (Total Effect: 1.25), an estimated 36% of that increase was due to time spent alone.

4. Discussion

Our findings confirm previous studies identifying the adverse effects of psychosocial stress on ovarian physiology, as subordinate females exhibited significantly lower luteal phase progesterone levels compared to their dominant counterparts in repeated measures analysis. We also provide several novel insights. This study is the first to report that female long-tailed macaques exhibited seasonal menstrual disturbances, with the lowest luteal phase progesterone concentrations clustering in the late summer and early fall in North Carolina, USA. We also found that individuals consuming a Western-like diet had 1) blunted progesterone levels and 2) within dominant monkeys, more irregular cycles compared to those consuming a Mediterranean-like diet. Our data suggest that HPA function and social isolation may mediate the relationships between diet and ovarian function. The observation that the Mediterranean diet group had higher mean peak progesterone concentrations than the Western diet group is consistent with a similar recent report in pair-living rhesus macaques that Western diet consumers had lower progesterone levels than chow consumers (Bishop et al., 2021). In that study, dietary fat content in the Western diet was twice that of the chow diet. Here we present the first evidence of protective effects of a Mediterranean diet, containing the same proportion of calories from fat as the Western diet, on luteal phase progesterone secretion. In our study, both diets were formulated to recapitulate diets currently consumed widely in human populations. Altogether, this work showcases the negative effects of poor diet and psychosocial stress on ovarian physiology in premenopausal females and suggests that a Mediterranean diet may promote ovarian health.

Our discovery of a pattern of seasonal disturbance in ovarian cycling is contrary to some published reports (Dang, 1977; Weinbauer et al., 2008) of an absence of seasonality in the reproductive cycles of captive long-tailed macaques. These differences might be explained by a more thorough characterization of menstrual cyclicity and quality in the present study, or by other differences in husbandry techniques or methods for assessing ovarian physiology. Like our study, Dang et al. (Dang, 1977) studied the menstrual cycles of females housed with access to natural light conditions (Paris, France). However, unlike the macaques studied here, the females in Dang et al. maintained visual contact with male monkeys and had mating opportunities during the study. Furthermore, 49 of the 55 females described by Dang et al. received progesterone treatments prior to the study to synchronize menstrual cycles. Weinbauer et al. (2008) review data obtained using vaginal smears to detect menstruation, rather than repeated blood sampling, the technique used here. Interestingly, Li et al. (Li et al., 2022) report a birth peak in late summer an early fall, which temporally aligns with our observed pattern of ovarian disturbance. The results reported here suggest that ovarian physiology in the long-tailed macaque has a circannual rhythm which has also been reported in some populations of women (Sundararaj et al., 1978).

Inadequate progesterone profiles, like those reported in the Western dietary group, reflect luteal phase aberrations, which may increase risk of infertility and early miscarriage (Arredondo and Noble, 2006). Our findings contribute to a growing literature demonstrating the potential protective effects associated with consumption of healthier diets, like the Mediterranean diet (Garruti et al., 2019; Kazemi et al., 2020). For example, in the Nurses’ Health study, a subset of women consuming a “fertility diet” consisting of monounsaturated fats, plant proteins, low glycemic carbohydrates, iron supplements, and multivitamins, experienced lower rates of infertility due to self-reported ovarian disorders (Chavarro et al., 2007). Conversely, diets high in sugar, animal protein, saturated and trans fats, and other constituents also may harm ovarian health (Fontana and Torre, 2016; Jurczewska and Szostak-Węierek, 2022). Longitudinal preclinical studies would be useful to better understand the effects of these dietary patterns on fertility and ovarian health.

The clinical sequelae of poor ovarian function are extensive, the most pervasive being infertility, osteoporosis, and CVD. Premenopausal women with comparatively low progesterone profiles also may experience reductions in bone mineral density (Prior, 2018; Prior et al., 1990). Progesterone promotes bone mineralization by stimulating mesenchymal stem cells to differentiate into osteoblasts, and stimulating osteoblast activity (Seifert-Klauss and Prior, 2010). Those with high luteal progesterone are more likely to have higher follicular estradiol concentrations, which also beneficially effects bone density by preserving physiologic bone modeling and preventing estrogen deficiency-related osteoclast activity and bone resorption (Adams et al., 1985). Thus, higher ovarian steroid levels build greater bone reserves before the menopausal transition, a critical period for osteopenia in women (Karlamangla et al., 2018). Long-tailed macaques with poor ovarian function develop more extensive coronary artery atherosclerosis (Kaplan and Manuck, 2017). Women with irregular menstrual cycles or premature ovarian insufficiency are far more likely to develop postmenopausal CVD (Solomon et al., 2002) (Podfigurna-Stopa et al., 2016).

Diet and psychosocial stress interacted to mediate menstrual cycle regularity. Here, menstrual cycles were considered “irregular” if they were either abnormally short (<25 days) or long (>33 days). In the Western diet group, dominants and subordinates differed in their propensities for irregular cycles; dominants experienced more short cycles and fewer long cycles, whereas subordinates had fewer short cycles and more long cycles. In contrast, status was not associated with differences in cycle regularity in the Mediterranean diet group. Thus, Western versus Mediterranean diets may exert different physiologic effects depending on whether individuals are psychosocially stressed. These data add to a growing body of evidence demonstrating that Western diet exacerbates the deleterious health effects of psychosocial stress via multi-system disturbances, (Frye et al., 2020; Gonzalez-Armenta et al., 2019; Johnson et al., 2020; Newman et al., 2021; Shively et al., 2020; Shively et al., 2019; Shively and Day, 2015).

The most dramatic effects on cycle regularity were the protective effects of Mediterranean diets against shorter-than-normal menstrual cycles. We recorded only three short cycles in the Mediterranean group versus 33 short cycles in the Western group. This pattern was consistent across dominant and subordinate monkeys. These data agree well with observational studies of reproductive-aged women. In a cross-sectional study of Spanish university women, low adherence to a Mediterranean-style diet was associated with irregular menstrual cycles compared to women with higher degrees of adherence (Onieva-Zafra et al., 2020). Clinical data have shown that individuals with short menstrual cycles (defined as < 26 days) also exhibit endocrine profiles reflective of advanced reproductive aging, including elevated follicle stimulating hormone, early bursts of estradiol in the follicular phase, low progesterone in the luteal phase, and low ovarian reserve as reflected in anti-Mullerian hormone or antral follicle count (MacNaughton et al., 1992; Meyer et al., 2007; Mumford et al., 2012; Younis et al., 2020). These observations suggest that the Mediterranean diet may protect against premature reproductive aging. Support for this hypothesis comes from a recent study showing that a high intake of oily fish and legumes was associated with a delay in the onset of menopause (Dunneram et al., 2018). Thus, the health benefits of Mediterranean diet in women may in part be due to the beneficial effects on ovarian function. Future longitudinal studies are needed to better understand the interrelationships between diet composition, ovarian function, and aging-related chronic disease risk in females. Further studies of the impact of diet composition on the milieu of endocrine signals associated with reproductive aging in NHPs would be helpful.

The pathophysiologic sequelae of Western diet consumption may be partly explained by elevated adrenal reactivity. Previously we reported that in this study the Western diets perturbed HPA physiology (Shively et al., 2020). Here we report that adrenal reactivity partially mediated the relationship between diet and the number of irregular menstrual cycles. That is, a significant proportion of the increased risk of having an irregular cycle due to Western diet consumption was due to elevated adrenocortical reactivity. Cortisol is known to have inhibitory effects on the HPG axis, including suppression of estrogen and progesterone secretion and increased target tissue resistance to these hormones (Chrousos et al., 1998; Kalantaridou et al., 2004; Magiakou et al., 1997). Here, it appears that the Western diet instigated cortisol-related physiologic cascades which disrupted reproductive physiology. There also may be a bidirectional relationship between HPG and HPA physiology, such that reproductive physiology may have impacted HPA responses (Kirschbaum et al., 1999; Toufexis et al., 2014; Viau, 2002). Future studies are needed to understand the mechanistic underpinnings and directionality of this relationship. Such insights may have clinical implications given that reproductive dysfunctions, such as menstrual cycle irregularity, may themselves result in an increased risk for pathologic aging (Kaplan, 2008).

In addition to adrenal reactivity, social isolation also partially mediated the effects of diet on cycle regularity. Cross-sectional clinical data from the Whitehall II cohort (ages 45-59 years) has shown that greater total cortisol output during waking hours is associated with more self-reported social isolation (Grant et al., 2009). However, cross-sectional studies do not allow for determination of causality. Our study design allows for determination of diet effects on ovarian function and behavior. Interestingly, our measures of social isolation and adrenal reactivity were unrelated. This may be due to the nature of our dependent variables, or to alternative mechanisms underlying the relationships between diet, social isolation, and ovarian physiology.

As previously reported (Johnson et al., 2022), consumption of the Mediterranean diet in this study resulted in less anxiety-like behaviors, more affiliation, and less time spent alone compared to Western diet consumption, and that these diet-alter behaviors mediated monocyte inflammatory gene expression (Johnson et al., 2021). Proinflammatory profiles have been shown to perturb ovarian function (Boots and Jungheim, 2015). Thus, the behavioral effects on ovarian function reported here may be due to behavioral effects on inflammation. This study adds to the growing body of work describing behavioral mechanisms through which Mediterranean diets may protect women’s health. It also suggests a role of pro-inflammatory pathways in shaping ovarian physiology. Therapeutic efforts to increase social integration may be facilitated by the inclusion of anti-inflammatory dietary patterns, such as a Mediterranean diet, that reduce social isolation.

This study has some limitations. First, inter-individual variation in reproductive physiology may have obfuscated some of the patterns in the data in this relatively small sample. Second, we were also limited by the number of repeated measures for our mediating variables. While we had weekly assessments of ovarian function, we were limited to individual timepoints for our mediating variables. Thus, we may have missed important patterns by using Treatment phase averages for progesterone and cycle lengths. Lastly, while we tested three specific hypotheses regarding the potential mechanisms underlying diet-ovarian relationships, ovarian and HPA function are clearly linked through complex mechanisms which limit firm conclusions as to directionality of those interactions. Likewise, other diet and stress related candidate mechanisms such as inflammation or direct hormone effects need to be explored to better understand the molecular pathways that may increase risk for pathologic aging trajectories.

5. Conclusions

Mediterranean diet consumption results in a healthier ovarian function profile compared to Western diet consumption, and diet effects on ovarian function may be mediated by both adrenal reactivity and social isolation. This study demonstrates the protective effects of the healthful Mediterranean diet on ovarian health, which may vary with an individual’s exposure to psychogenic stressors. Such work is important for developing interventions to mitigate the risk of aging-related chronic diseases later in life.

Highlights.

• Subjects exhibited circannual patterns consistent with menstrual cycle disturbances

• Poor diet & psychosocial stress exert additive negative effects on ovarian function

• Mediterranean diets may protect against ovarian dysfunction and pathologic sequelae

Data statement

The data for this manuscript are available upon reasonable request from the corresponding author, Dr. Carol Shively.