Neuro-pharmacological reinstatement of ovulation and associated neurobiology in a macaque model of functional hypothalamic amenorrhoea

Abstract

STUDY QUESTION

What is the underlying neuropathology in a cynomolgus macaque model of functional hypothalamic amenorrhoea (FHA) and can it be normalized to restore ovulation?

SUMMARY ANSWER

Anovulatory monkeys exhibited increased hypothalamic norepinephrine (NE), kisspeptin and gonadotropin-releasing hormone (GnRH) in the early follicular phase, but administration of the NE reuptake inhibitor (NRI), reboxetine (REB), restored ovulation during stress and normalized NE, kisspeptin and GnRH.

WHAT IS KNOWN ALREADY

Female cynomolgus macaques, like women, show individual reproductive sensitivity to modest psychosocial and metabolic stress. During stress, resilient females ovulate through two menstrual cycles whereas stress-sensitive (SS) macaques immediately cease ovulation. On Day 5 of a non-stressed menstrual cycle, resilient macaques have less NE synthesizing enzyme [dopamine β-hydroxylase (DBH)], kisspeptin and GnRH innervation of the medial basal hypothalamus but more endogenous serotonin than SS macaques. Stress increased DBH/NE, kisspeptin and GnRH but did not alter serotonin.

STUDY DESIGN, SIZE, DURATION

In a longitudinal design, 27 adult (7–13 years) female cynomolgus macaques (Macaca fascicularis) with three different levels of sensitivity to stress were monitored with daily vaginal swabs and frequent serum progesterone (P) measurements. Three 90-day experimental periods called ‘Cycle Sets’ were monitored. A Cycle Set consisted of one ovulatory menstrual cycle without stress, and two cycles, or 60 days, with modest stress. Each Cycle Set was followed by a rest period. During a Cycle Set, individuals were either untreated (placebo) or administered escitalopram (CIT) or REB. Ultimately, half of each sensitivity group was euthanized during stress with CIT or REB treatment and the hypothalamus was obtained. Neurobiological endpoints were compared between CIT and REB treatment groups in stress resilient and SS monkeys.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The monkeys were housed at the University of Pittsburgh primate facility for the duration of the experiments. Upon euthanasia, their brains and serum samples were shipped to the Oregon National Primate Research Center. The hypothalamus was examined with immunohistochemistry for the expression of DBH (a marker for NE axons), kisspeptin and GnRH. P was measured in the serum samples by radioimmunoassay.

MAIN RESULTS AND THE ROLE OF CHANCE

Daily administration of REB restored ovulation in 9 of 10 SS animals during stress. Of note, REB significantly increased P secretion during stress in the most sensitive group (P = 0.032), which indicates ovulation. CIT lacked efficacy. REB significantly reduced DBH/NE, kisspeptin and GnRH axon density in the hypothalamus relative to CIT treatment (P = 0.003. 0.018 and 0.0001, respectively) on Day 5 of the menstrual cycle in resilient and sensitive groups.

LARGE SCALE DATA

N/A.

LIMITATIONS, REASONS FOR CAUTION

The US FDA has not approved REB for human use, although it is used in Europe for the treatment of depression/anxiety as Edronax^TR. Whether REB could be useful for the treatment of FHA in women has not been determined.

WIDER IMPLICATIONS FOR THE FINDINGS

The use of an NRI to treat FHA is a novel approach and the potential reinstatement of ovulation could be straightforward compared to current treatment protocols. The underlying neurobiology provides a compelling case for treating the origin of the pathology, i.e. elevated NE, rather than circumventing the hypothalamus altogether with gonadotropins, which have associated risks such as hyperstimulation syndrome or multiple births.

STUDY FUNDING/COMPETING INTEREST(S)

Portions of this study were supported by NIH grant HD062864 to C.L.B., NIH grant HD62618 to J.L.C. and C.L.B. and 1P51 OD011092 for the operation of the Oregon National Primate Research Center. There were no competing interests.

Introduction

Chronic stress can lead to a number of secondary diseases such as mental illnesses like anxiety and/or depression, heart disease and immune suppression (McEwen, 2002). The reproductive system in women is also sensitive to stress (Chatterton, 1990; Xiao et al., 1998). Functional hypothalamic amenorrhoea (FHA) is a type of infertility that is defined as a sustained absence of normal menstrual cycles despite healthy reproductive organs with the ability to function normally (Reifenstein, 1946; Reindollar et al., 1986; Berga and Girton, 1989; Couzinet et al., 1999). Infertility of 12 months occurs in ∼9% of the global population (Boivin et al., 2007). In the USA, ∼30% of women presenting at infertility clinics are diagnosed with FHA, which more recently has been called stress-induced amenorrhoea (Marcus et al., 2001; Berga and Loucks, 2006). According to psychometric testing, women with FHA report an increased amount of psychosocial stress in their lives, despite experiencing comparable numbers of stressful life events as control populations, or women with other forms of infertility (Giles and Berga, 1993; Fioroni et al., 1994; Marcus et al., 2001). Women with FHA carefully watch their food intake and score higher on eating disorder inventories than controls, but not in the range of those with eating disorders. They also tend to exercise on a regular basis, often to control life stress (Giles and Berga, 1993; Fioroni et al., 1994; Laughlin et al., 1998; Marcus et al., 2001). Nonetheless, their weight remains in a normal range. Altogether, it appears that FHA results from exposure to a combination of common psychosocial and metabolic life stresses. Certainly, not all women who experience these stresses become infertile, suggesting that there is a range of sensitivity to developing stress-induced reproductive dysfunction in normal women.

Although the neural origin of FHA has been a tenet of gynaecology for decades (Berga et al., 1989), a clear understanding of the neurobiology and a simple neuro-pharmacological treatment are lacking. Rather, clinical treatments bypass the reproductive neural pathways in the hypothalamus altogether, providing supplementary stimulation of the pituitary or ovaries. Recognition and amelioration of stress sensitivity in women with FHA using cognitive behavioural therapy (CBT) has found success (Berga and Loucks, 2006), but it is a time-consuming process and not easily accessible in rural communities.

The weight of relevant literature indicates that impairment of the FSH rise during the luteal-follicular transition is the cause of FHA (de Koning et al., 2004). However, ovarian insensitivity has also been noted (Jonard et al., 2005). A great deal of experimental focus has been devoted to the lack of ovulatory gonadotropin-releasing hormone (GnRH) pulses in this disease, and indeed GnRH supplementation has been used to treat FHA (Hurley et al., 1983; Miller et al., 1983). Nevertheless, the midcycle GnRH deficit is due to FSH dysfunction 14 days earlier with the ensuing lack of follicle development and estradiol (E) production. Under experimental conditions, low GnRH pulse frequency favours FSH over LH secretion and low GnRH pulse frequency was used in clinical treatment protocols (Stern et al., 2017; Stamatiades and Kaiser, 2018). Clearly, without FSH stimulation, ovarian follicles fail to grow past 2–5 mm, fail to form a dominant follicle and fail to secrete E needed for succeeding events such as the ovulatory LH surge. Thus, to treat the neural basis of FHA, it is necessary to recognize that the dysfunction begins with FSH and that the lack of an ovulatory surge of GnRH/LH after 14 days is a consequence of insufficient follicle development, not the proximate cause of FHA. It is also important to keep these very different, but linked, phases of the menstrual cycle ideologically separate to understand the data and discussion that follow. We recognize that there is no GnRH or LH surge in FHA, but the endogenous menstrual cycle normally progresses from crucial events in the luteal-follicular transition.

To further our understanding of FHA, we modelled the psychosoical stress, diet and exercise habits reported by women in a species of old world macaque, the cynomolgus macaque (Macaca fascicularis). Cynomolgus macaques, like women, display monthly menstrual cycles with no seasonal variation. As in women, we found that mild psychosocial stress (i.e. relocation to a new housing area while living in a single cage), combined with a mild diet and exercise regimen, suppressed reproductive function in about one-quarter to one-third of cynomolgus monkeys, which then reversed upon stress removal. That is, stress-resilient (SR) monkeys continue to display ovulatory menstrual cycles and cyclic patterns of reproductive hormone secretion for two menstrual cycles when exposed to the combined stress paradigm. However, stress-sensitive (SS) monkeys fail to develop a dominant follicle, ovulate, or mense (exhibit amenorrhoea), when left in the mild stress condition for 2 cycles, or the equivalent of 60 days (Cameron, 2000; Bethea et al., 2005, 2008; Williams et al., 2007). Within populations of cynomolgus females, there was a normal distribution of sensitivity of the reproductive axis to stress. The model has been highly relevant to infertility, quantifiable by menstrual cycles and hormone levels and reproducible with small populations of wild caught cynomolgus macaques, thereby fulfilling the need for accuracy, authenticity and fidelity in a biological model.

In neurobiology studies, we found that SS animals exhibited lower serotonin in the midbrain and hypothalamus than highly stress-resilient (HSR) animals, but application of stress had no effect on serotonin (Bethea et al., 2014). In contrast, norepinephrine (NE) increased with stress in the midbrain and hypothalamus in both HSR and SS animals. We speculated that the HSR animals continued to ovulate through two stress cycles because their elevated NE occurred in conjunction with a highly functional serotonin system. In comparison, SS animals ceased ovulation and did not mense again after the application of stress due to elevated NE with a lower functioning serotonin system (Bethea et al., 2013, 2014).

In addition, increased KISS1 mRNA and GnRH were observed in the hypothalamus of SS animals relative to HSR animals; and elevated GnRH correlated with increased serum LH on Day 5 of the follicular phase (Bethea et al., 2014). The differential secretion of LH and FSH by gonadotropes requires select factors in the microenvironment (McNeilly et al., 2003). Based upon extensive literature, we hypothesized that slightly elevated GnRH in the early follicular phase could compromise FSHβ subunit production or processing by the gonadotrope, thereby restricting folliculogenesis. In other studies, women with an inadequate luteal phase had more frequent LH pulses, and presumably GnRH pulses, during the follicular phase compared to normal women (Soules et al., 1989; Suh and Betz, 1993; De Souza et al., 1998; de Koning et al., 2004). Moreover, models indicate that the apparent frequency-dependent pattern of FSHβ expression actually resulted from variations in average GnRH concentration (Stern et al., 2017). These data support the hypothesis that elevated GnRH during the luteal-follicular transition, regardless of the pattern of secretion, is detrimental to FSH secretion.

Other evidence regarding NE unified our hypothesis. SS animals are more anxious than HSR animals (S.M. Herod and J.L. Cameron, unpublished work) and elevated NE plays a role in human anxiety or vigilance in non-human primates (NHPs) (Aston-Jones et al., 1991; Sullivan et al., 1999; Cameron et al., 2004; Itoi and Sugimoto, 2010). There was also ample evidence that stress increased NE in humans and other pre-clinical models (Strawn and Geracioti, 2008; McCall et al., 2015; Winklewski et al., 2017). Furthermore, a significant body of work indicated that NE increases LH secretion (Clifton and Sawyer, 1979; Martinez de la Escalera et al., 1992; Barraclough, 1994; Herbison, 1997; Spies et al., 1997; Anselmo-Franci et al., 1999; Szawka et al., 2013).

Therefore, we hypothesized that stress increased NE, which led to increased kisspeptin and, in turn, GnRH in the early follicular phase. Elevated GnRH contributed to a change in the microenvironment of the gonadotrope, which altered the FSHβ:LHβ ratio (McNeilly et al., 2003; Bilezikjian et al., 2004; Das and Kumar, 2018). Consequently, inhibition of NE would restore ovulation in SS animals during stress. In our search for an effective neuro-pharmacological agent that would inhibit only NE, we ruled out enzyme inhibitors and tricyclic antidepressants due to broad effects. However, it was recognized that NE reuptake inhibitors (NRIs) had the counter-intuitive effect of decreasing NE neuronal firing, possibly due to feedback from a synaptic autoreceptor (Szabo and Blier, 2001; Hussain and Maani, 2019). The antidepressant medication reboxetine (REB) exhibited significantly greater affinity for the norepinephrine transporter (NET) than the serotonin transporter (SERT) and initial characterization of REB discussed its merits as potentially the first selective NRI (Wong et al., 2000; Hajos et al., 2006). Although REB was approved in Europe for the treatment of anxiety in humans (Edronax^TR) (Burrows et al., 1998), it never gained approval in the USA by the FDA. However, its use in this study established proof of concept.

To test these two hypotheses, we applied our stress paradigm to a population of 27 imported cynomolgus macaques. Animals were monitored for sensitivity status through a ‘no stress-no drug’ menstrual cycle and two ‘stress-no drug cycles’, or 60 days if menses were not detected, with combined diet, exercise and psychosocial stress. Then, the same protocol was repeated with the administration of either escitalopram (CIT) or REB for the last 14 days of the ‘no stress’ cycle and through two stress cycles, or 60 days. Serum progesterone (P) concentrations were monitored for the evidence of ovulation. Hypothalamic function was determined with immunohistochemistry (IHC). By employing a monkey model of FHA, this study demonstrates the efficacy of brief treatment with a selective NRI restores ovulation in SS individuals under stress and provides further evidence that elevated NE plays a central role in the neural cascade leading to FHA.

Materials and methods

Animals and treatments

This experiment was approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh and conducted in accordance with the 2011 Eighth Edition of the National Institute of Health Guide for the Care and Use of Laboratory Animals.

The physiological studies described herein were conducted at the University of Pittsburg, Pittsburgh, PA. Twenty-seven adult female cynomolgus monkeys (Macaca fascicularis) were imported for study. The animals were 7–13 years of age. After quarantine, the animals were housed in single cages in large rooms with other monkeys. Movement of monkeys within the room, and in and out of the room, was limited so that the monkeys had a stable social environment for 4 months. The monkeys were fed commercially available monkey chow (12 pellets of Purina High Protein monkey chow) once a day at ∼1100 h supplemented with fruit. Water was available ad libitum.

Cynomolgus macaques form social hierarchies in grouped housing and subordinates receive more aggression. With single cages, the rank is not a variable. Housing, diet, food intake, weighing, daily vaginal swabs for detecting menstruation and blood sampling for P measurements have been described previously (Williams et al., 2001, 2007). The monkeys were similar in weight, and there were no body weight changes throughout the characterization of stress sensitivity (Herod et al., 2011).

Assessment of stress sensitivity

For each monkey, sensitivity of the reproductive axis to stress was categorized by assessing changes in menstrual cycle length (presence of blood on vaginal swab) and ovulation as determined with P secretion when monkeys were exposed to mild psychosocial and metabolic stressors. Characterization of sensitivity was performed after each monkey had been living in its home cage surrounded by familiar monkeys for several months and exhibiting regular menstrual cycles (mense every 25–38 days) as determined with vaginal swabs. In our early studies, the stressors consisted of moving the monkeys to a novel room surrounded by unfamiliar conspecifics, running 1 h on a treadmill 5 days/week and reducing the available chow by 20% (Bethea et al., 2008). However, in this study, food intake was decreased 25% and monkeys ran 30 min/day 5 days a week because several treadmills broke (beyond repair) and no funds were available for the purchase of new treadmills. The changes were calculated to cause a comparable decrease in calorie availability.

Animals were categorized as HSR if they presented a normal ovulatory menstrual cycle [mense every 25–38 days; peak P > 2 ng/ml in luteal phase] in Stress Cycle 1 and again in Stress Cycle 2. Medium stress-resilient (MSR) animals were defined as those animals that presented a normal ovulatory menstrual cycle in response to Stress Cycle 1 but failed to mense by Day 38 of Stress Cycle 2. Animals that immediately suppressed normal menstrual cyclicity upon exposure to stress (i.e. they failed to mense within 60 days and failed to ovulate as indicated by P secretion) were categorized as SS. All animals were allowed to recover normal menstrual cyclicity before further experimentation.

Study protocol

The animals were moved to their homeroom, provided with full rations and stopped running until full regular menstruation resumed as determined with vaginal swabs. Then, monitoring started for a non-stressed cycle. Drug treatment was initiated with either CIT (1.2–4.8 mg/kg B.I.D.) or REB (2 mg, B.I.D.) on Day 14 of the non-stress cycle and continued through two additional stress cycles or equivalent days in the absence of menstruation (Fig. 1 and Supplementary Fig. S1). Serum samples were obtained every 3 days. Based upon menstrual cycle records or days beyond the last mense, serum samples from the luteal window were submitted for P RIA. The highest serum P concentration was called ‘peak P’, and it was used for further data compilation. For data analysis, animals were grouped according to the initial characterization of their stress status.

Figure 1.

Illustration of the experimental protocol. Each animal was exposed to three Cycle Sets with rest periods in between that enabled restoration of normal menstrual cycles. (A) Each Cycle Set was composed of one full menstrual cycle with no stress and two menstrual cycles or 60 days with our combined metabolic and psychosocial stress regimen. (B) After quarantine, the animals were housed in large rooms where they adjusted to their housing and resumed menstrual cycles. The first Cycle Set was performed with no drug administration for characterization of the stress sensitivity of each animal. This assigned sensitivity label [highly stress resilient (HSR), medium stress resilient (MSR) or stress sensitive (SS)] was retained throughout the protocol. If an animal did not mense after the ‘no stress’ cycle, then she was monitored for at least 60 days. After Cycle Set 1 and a rest period, the animals were started on Cycle Set 2. During the ‘no stress’ cycle, the animals were administered escitalopram (CIT) starting on Day 14 and continuing through Stress Cycles 1 and 2. After Cycle Set 2 and a rest period, the animals were started on Cycle Set 3. During the ‘no stress’ cycle, the animals were administered reboxetine (REB) starting on Day 14 and continuing through Stress Cycles 1 and 2.

Terminal protocol

After menstrual cycles resumed during a rest period, the terminal protocol was initiated. The three sensitivity groups were split and half were treated with CIT and the other half with REB starting on Day 14 of a non-stressed cycle and continued for 3 weeks. Stress was applied for the last 5–7 days of drug treatment and the animals were euthanized (Supplementary Fig. S2). Altogether, drug treatment equalled 19–21 days until euthanasia.

Hormone assay

The Endocrine Technology and Support Laboratory, Oregon National Primate Research Center (Beaverton, OR, USA) conducted the P RIA on collected serum. P RIA was performed using a Roche Cobas e411 automatic clinical platform assay instrument. The assay has been validated for macaque plasma in comparison with traditional RIA. Within-assay coefficients of variation were <5% and between-assay coefficients of variation were <10%.

Euthanasia and tissue preparation

The monkeys were euthanized according to the procedures recommended by the Panel on Euthanasia of the American Veterinary Association. Each animal was sedated with ketamine, given an overdose of pentobarbital (25 mg/kg, intravascular) and exsanguinated by severance of the descending aorta. The brain was perfused with 1 l of heparinized saline followed by 7 l of 4% formaldehyde, removed from the cranium and shipped overnight to the Oregon National Primate Research Center (ONPRC) in perfusion buffer. The details of the brain perfusion, dissection, post-fixation, cryoprotection, sectioning, mounting and storage were previously described (Bethea et al., 2013). For this study, 25-µm-thick sections were obtained from the hypothalamic blocks yielding eight representative sets containing 24 sections at 200-µm intervals. All sections were stored floating in cryoprotectant at −20°C.

Animal selection for neurobiology

The sensitivity status was based upon the initial characterization. Animals that exhibited ovulation with REB treatment (according to P secretion) were used for the neurobiology studies. MSR and SS animals were screened as a combined group due to the small numbers of animals in each of these sensitivity categories. This yielded four HSR plus four SS/MSR animals that were euthanized with CIT treatment, and four HSR plus four SS/MSR animals that were euthanized with REB treatment. In the neurobiology analysis, the SS/MSR group is referred to as ‘SS’ for brevity. This approach optimized the potential differences between groups and further enabled a 2 × 2 block design with a technically feasible number of sections in each IHC assay.

IHC for NE, kisspeptin and GnRH

NE innervation of the hypothalamus was inferred from detection of dopamine β-hydroxylase (DBH) in the arcuate (ARC) nucleus. DBH is the final enzyme in the pathway for NE synthesis. Six sections at 200-µm intervals through the rostral ARC nucleus from each animal were stained for DBH. The primary antibody to DBH (rabbit anti-DBH, Immunostar, Hudson, WI; formerly Incstar, product ID 22806) was used at a concentration of 1/4000. Vector biotinylated goat anti-rabbit secondary antibody, Vector Avidin–Biotin Complex (ABC), and Vector diaminobenzidine (DAB) solutions were used for final signal development according to the recommendations of the manufacturer with multiple rinses between incubations (Vector Laboratories, Inc., Burlingame, CA). The DBH IHC assay was characterized across a range of titres with positive and negative controls, and details of buffers and incubation times are published (Bethea et al., 2014, 2015).

With the angle of sectioning that we employ, kisspeptin neurons appeared in the ARC, which in primates, enlarges and extends into the infundibular nucleus or infundibulum (INF). The INF recess of the third ventricle demarcates the INF. This morphology was illustrated in Figure 1 of Bethea et al. (2014) with in situ hybridization (Bethea et al., 2014). The morphology of the KISS1 mRNA population in the rostral ARC was inconsistent between animals and added more variance to the analysis. Therefore, we focused on the INF for the kisspeptin IHC analysis in this study. Hypothalamic sections of the INF from each animal were immunostained for kisspeptin with a polyclonal anti-kisspeptin antibody raised in sheep and used at a concentration of 1/60 000 (GQ2; a gift of Dr. Stephen Bloom, Imperial College, London, UK). Vector biotinylated secondary rabbit anti-sheep antibody, Vector ABC and Vector DAB solutions were used for final development according to the recommendations of the manufacturer with multiple rinses between incubations. Six sections at 200-µm intervals of the ARC–INF were stained for kisspeptin and 3–6 sections from each animal (similar to sections 2, 3 and 4 in Figure 1 of Bethea et al., 2014) were subjected to image analysis. Details of IHC assay for kisspeptin are published including absorption controls (Ramaswamy et al., 2008; Alcin et al., 2013).

Hypothalamic GnRH axon densities entering the median eminence (ME) were compared between animals following GnRH IHC. Four sections of the ME at 50 µm intervals proximal to the ARC from each animal were stained for GnRH with rabbit polyclonal anti-GnRH antibody HU60F (generously provided by Dr. Henryk Urbanski, ONPRC, Beaverton, OR, USA) at a dilution of 1/2000 as published (Centeno et al., 2007; Bethea et al., 2014). Vector biotinylated goat anti-rabbit antibody, Vector ABC, and Vector DAB solutions were used for final signal development according to the recommendations of the manufacturer with multiple rinses between incubations.

Image capture and analysis

Images containing the region of interest (ROI) were scanned and captured at 200× under an upright Olympus BX6IVS microscope with VS-ASW imaging software (Olympus Americas). Colour images were captured and converted to black and white for image analysis with Fiji (ImageJ freeware from NIH) using the Bioformats plugin. The final ROI was cropped from the larger scanned image.

When needed, the third ventricle was spliced out of the image, thereby joining the left and right sides of the ARC nucleus. This provided a larger ROI for analysis. The background was subtracted from the ROI and the contrast was enhanced. The positive and negative pixels in the ROI were segmented using the threshold function to a similar degree with each image. When needed in the ME, disjointed ROIs were analysed and the average was used as the section value. The variation in the size of the ROI from animal to animal required expression of the positive pixel area in microns and expression of the individual section data as positive pixel area per square micron. This normalization enabled direct comparisons between treatments and sensitivity groups. The section values were averaged to obtain the animal value. The animal values were used for statistical analysis.

Statistics

Progesterone

The peak serum P concentration was recorded from each animal during each cycle. Since blood samples were obtained from the same animals across the series of menstrual cycles, a repeated measures analysis of variance (RM-ANOVA) was required. However, the traditional repeated measures ANOVA cannot tolerate a missing data point, and there was one missing data point in the MSR group. A mixed effects ANOVA is a repeated measures analysis that allows a missing data point. Therefore, this type of analysis was applied across all cycles or 30-day periods in amenorrheic animals within each sensitivity group. Post hoc pairwise comparisons between all treatment cycles within a sensitivity group were also performed with Tukey’s, as recommended.

Neurobiology

The animal values were grouped by treatment and sensitivity for two-way ANOVA followed by Tukey’s post hoc pairwise comparison. Thus, the variation around the mean represents animal-to-animal variation. The power is dependent upon the animal number, which was low. Thus, differences between the groups needed to be large to reach significance. However, a lack of difference between groups could be a Type 2 error. Prism 8.0 from Graph Pad was used for all analyses with alpha equal to 0.05.

Results

Progesterone production

Peak luteal phase P concentrations in each stress group and during each Cycle Set are illustrated in Fig. 2 and Supplementary Fig. S3. Serum P concentrations are shown on the Y-axis. The drug treatments are shown on the X-axis. The small case ‘x’ means the presence of stress or drug during the cycle or 30-day period. Different letters indicate a significant post hoc difference between the average ‘peak P’ values of the specific treatments.

Figure 2.

Progesterone secretion across the three Cycle Sets. Histograms depicting the average of peak serum P concentrations ± SEM in each sensitivity group across the three Cycle Sets of the experimental protocol. The animals were assigned to a sensitivity group based upon their menses and P data from the initial ‘no drug’ Cycle Set. There was a significant difference in serum P concentrations between the treatments. (A) Serum P concentrations in the HSR animals were indicative of corpus luteum function in the presence and absence of stress in the ‘no drug’ Cycle Set. No P data are available from the ‘CIT-no stress’ cycle, but CIT appeared to decrease P secretion in HSR animals under stress. Treatment with REB in Cycle Set 3 returned P secretion to levels observed in the ‘no drug’ Cycle Set 1. (B) Serum P concentrations in the MSR group in Cycle Set 1 were consistent with previous observations. That is, P secretion from the corpus luteum was robust during the ‘no stress-no drug’ cycle and then reduced in the following two cycles with stress and no drug. CIT did not improve corpus luteum function under stress in Cycle Set 2. REB treatment in Cycle Set 3 improved P secretion in Stress Cycle 1, but less so in Stress Cycle 2, although all five animals ovulated in both stress cycles. (C) The SS animals exhibited poor, if any, ovulation in Cycle Set 1, even in the absence of stress, and no ovulation in the two stress cycles; thus, serum P concentrations were very low. There was a small increase in P secretion in the CIT + stress cycles that represented ovulation in two of five animals. REB treatment significantly increased P secretion during Cycle Set 3. All five animals ovulated with robust corpus luteum function in the first ‘REB + stress’ cycle. In the second ‘REB + stress’ cycle, three of five animals clearly ovulated. (D) Progesterone secretion in three animals that became unwell during the protocol but did not exhibit overt physical symptoms. Different letters indicate a significant difference between the treatments with Tukey’s post hoc pairwise comparison. Open bars indicate the no stress cycle. Hatched bars indicate Stress Cycle 1. Black bars indicate Stress Cycle 2. The drug treatments of the Cycle Sets are defined under the X-axis. Each bar is further defined with a small case ‘x’ indicating the presence of stress and the presence of drug treatment.

As shown in Fig. 2A and Supplementary Fig. S3 A, there was a difference in peak P values between treatments in the HSR group (F [7,91] = 3.26; P = 0.004, RM-ANOVA). During the no drug Cycle Set 1 (left column), HSR animals exhibited robust serum P concentrations during the ‘no stress’ cycle as well as during Stress Cycles 1 and 2. In the middle column, the serum P concentrations are illustrated for the CIT-treated Cycle Set 2. CIT was administered on Day 14 of the ‘no stress’ cycle and continued through two stress cycles, or 60 days in the absence of menstruation. Serum P measurements were not provided to the senior author during the CIT-treated ‘no stress' cycle, but our previous paper showed that CIT had no effect in non-stressed HSR animals (Bethea et al., 2011). However, P secretion was reduced in HSR animals that were stressed and treated with CIT compared to the ‘no drug’ Cycle Set. The right column illustrates serum P concentrations in the REB-treated Cycle Set 3. The HSR group with REB treatment exhibited robust P secretion during the ‘no stress’ cycle and during Stress Cycles 1 and 2 in a manner similar to the ‘no drug’ Cycle Set. Statistically, serum P was higher in the HSR group with REB treatment during the ‘no stress’ cycle than during either of the stress cycles with CIT treatment (P = 0.015 and 0.04, respectively, Tukey’s).

As shown in Fig. 2B and Supplementary Fig. S3B, there was a significant difference in peak P values between treatments in the MSR group (F [8,35] = 6.70; P < 0.0001, RM-ANOVA). Of note, in Cycle Set 1 (left column) with no drug treatment, the MSR group exhibited robust serum P concentrations during the ‘no stress’ cycle, then modest P secretion during Stress Cycle 1 (which ended with menstruation) and low or non-ovulatory P secretion during Stress Cycle 2, which was monitored for >30 days due to the lack of menstruation. Thus, CIT did not improve P secretion in MSR animals under stress. However, REB increased P secretion (indicating ovulation) in Stress Cycles 1 and 2 (right column) compared to ‘no drug’ Stress Cycles 1 and 2 (left column) of the MSR group. There was no difference in P secretion between the ‘no stress–no drug’ cycle and the ‘no stress + REB’ cycle in the MSR group. Significantly higher serum P concentrations were found in the MSR group with ‘no drug–no stress’ and with REB treatment during Stress Cycles 1 and 2 than in ‘no drug’ Stress Cycles 1 and 2 and during Stress Cycles 1 and 2 with CIT treatment (Tukey’s, P ≤ 0.03, all).

The SS animals are believed to be a model for women with FHA; and their P secretory data are shown in Fig. 2C and Supplementary Fig. S3C. There was a significant difference in peak P values between treatments in the SS group (F [8, 32] = 3.785; P = 0.0032, RM-ANOVA). In Cycle Set 1 with no drug (left column), the SS animals exhibited low serum P concentrations during the ‘no stress’ cycle suggesting formation of a poor corpus luteum, and non-ovulatory P secretion profiles for two anovulatory cycles, or 60 days. In this study, CIT marginally increased P secretion in SS animals under stress to concentrations that may have indicated small corpus luteum development. REB markedly improved luteal P production in SS animals in the absence and presence of stress. In the right column, administration of REB on Day 14 of the SS ‘no stress’ cycle increased serum P concentrations compared to the ‘no stress, no drug’ cycle in the left column. In addition, REB increased luteal P secretion in SS animals during Stress Cycles 1 and 2 (right column), compared to the ‘no drug’ Stress Cycles 1 and 2 in the left column. The peak P concentrations indicate that ovulation occurred and menstruation marked the end of each cycle. P was statistically higher in the SS group during Stress Cycle 1 with REB treatment than in the entire Cycle Set 1 with no drug (Tukey’s, P ≤ 0.03, all).

Figure 2D illustrates three animals that we labelled ‘non-responders’ because they ceased ovulation shortly after initiation of the protocol and never ovulated again. Subsequent examination revealed a significant leucocyte infiltration of the ME of one animal, suggesting that these animals may have developed an underlying illness that was asymptomatic during clinical observations.

Of note, the box diagrams of serum P concentrations in Supplementary Fig. S3 indicate the variable ranges of P values for each treatment. Even in treatments with marked improvement in the average peak P concentration, the bottom range of P secretion indicated that an animal within that treatment might not have ovulated. Individual variation is always an issue with NHPs, and it should be expected in women as well. For example, one SS animal did not appear to ovulate during her ‘no stress + REB’ cycle (P = 0.5 ng/ml), and then she ovulated during the first and second ‘stress + REB’ cycles with serum P concentrations of 6.83 and 10.5 ng/ml, respectively. Another SS animal ovulated during her ‘no stress + REB’ cycle (P = 2.6 ng/ml) and ovulated with a robust corpus luteum in the first ‘stress + REB’ cycle (P = 13.48 ng/ml) but failed to ovulate at all in the second ‘stress + REB’ cycle (P = 0.33 ng/ml).

Neurobiology

Norepinephrine

NE was examined in the rostral ARC nucleus by immunostaining for DBH. This enzyme converts dopamine (DA) to NE, and it localizes throughout the neuronal soma, axons and dendrites. An increase in the density of DBH axons in the ARC nucleus is proposed to be an indicator of increased NE production impinging upon the HPG axis. Figure 3 contains photomicrographs of NE immunostaining in the ARC nucleus of a typical animal from low-to-high magnification as would be performed for image capture and analysis. Figure 3A and B shows the gross morphology of the medial basal hypothalamus (MBH) with ARC (darker area) and ME. This is an example of a section in which the left and right ARC were cropped and joined for analysis (removed the third ventricle). Figure 3C and D shows the synaptic boutons that are highlighted in red after application of the threshold subroutine. Figure 3F and G shows individual neurons in the ARC with typical high-density NE synaptic boutons (black dots) in contact with the neuronal soma. In addition, Figure 3E contains a histogram illustrating the quantitative analysis of NE density across the groups examined. There was a significant difference in NE axon/bouton density between the groups (F [3, 44] = 5.510; P = 0.003, two-way ANOVA). There was a post hoc difference between HSR + CIT and HSR + REB (P = 0.008, Tukey’s) and between SS + CIT and SS + REB (P = 0.016, Tukey’s).

Figure 3.

Norepinephrine innervation of arcuate nucleus. (A) Photomicrograph of the intact medial basal hypothalamus (MBH) of a typical cynomolgus macaque showing the trapezoid shaped arcuate (ARC) nucleus merging into the median eminence of the hypophysial stalk. The dopamine β-hydroxylase (DBH) immunostaining gives the ARC a shaded appearance relative to the rest of the MBH. (B) Magnification of the ARC nucleus. (C) Dense concentration of DBH-stained axons. The higher magnification insert reveals individual boutons, which indicate synaptic spine contacts. (D) The same field in panel (C), with the boutons highlighted in red. The higher magnification insert reveals individual boutons highlighted in red that contributed to the quantitation of positive pixel area. This type of image is obtained with the ‘threshold’ mechanism of Fiji, which enables the highlighting of stained areas. (E) Histogram illustrating the average ± SEM density of norepinephrine (NE) in the ARC nucleus expressed as the ratio of positive (stained) microns to total micron area measured. There was a significant difference between HSR + CIT and HSR + REB and between SS + CIT and SS + REB. (F and G) Individual neurons in the ARC with dense NE boutons in synaptic contact with the soma.

Kisspeptin

Kisspeptin axon density was examined at the interface of the INF and ME due to the consistent density in the smaller area as compared to the entire INF. The INF also had neuronal somas, which were not desired for the analysis. Figure 4A–D contains photomicrographs of the kisspeptin staining. Figure 4A and B illustrates the gross morphology of the INF/ME area that was examined. Figure 4C and D shows high magnifications of the kisspeptin staining. Figure 4D contains the same field as Figure 4C, but with the axons and boutons highlighted in red after threshold discrimination of positive and negative pixels. Figure 4E illustrates the quantitative analysis of kisspeptin axon density in histogram format. There was a significant difference in kisspeptin axon/bouton density between the groups (F [3, 39] = 6.037; P = 0.0018, two-way ANOVA). There was also a post hoc difference between HSR + CIT and HSR + REB (P = 0.017, Tukey’s) and between SS + CIT and SS + REB (P = 0.004, Tukey’s).

Figure 4.

Kisspeptin innervation of the median eminence. (A and B) Photomicrographs of the infundibular (INF) area of a typical cynomolgus macaque showing the neural area merging into the median eminence of the hypophysial stalk. (C) Region of interest (ROI) at a higher magnification that reveals individual boutons. (D) Threshold application of enlarged ROI of panel (C) showing boutons highlighted in red that contributed to the quantitation of positive pixel area. (E) Histogram illustrating the average ± SEM density of kisspeptin in the INF–median eminence (ME) expressed as the ratio of positive (stained) microns to total micron area measured. There was a significant difference between HSR + CIT and HSR + REB and between SS + CIT and SS + REB.

Gonadotropin releasing hormone

Images of GnRH axon density were captured from the ME, sometimes from multiple areas of the same section. If multiple areas were analysed from a single section, then the average of the areas was obtained to represent that particular section. Figure 5A–D shows photomicrographs at increasing magnifications of the ARC–ME–pituitary stalk that were immunostained for GnRH. In this instance, the images in Figures 5A, B and Figures 5C, Dwere obtained from different sections. Figure 5D shows the axons and boutons in 5C highlighted in red after the threshold discrimination of positive and negative pixels. Figure 5E illustrates the quantitative analysis of GnRH axon density in histogram format. There was a significant difference in GnRH axon density between the groups (F [3, 30] = 12.60; P < 0.0001). There were post hoc differences between HSR + CIT and HSR + REB (P = 0.0004, Tukey’s), between SS + CIT and SS + REB (P < 0.0001, Tukey’s) and of note, between HSR + CIT and SS + CIT (P = 0.012, Tukey’s).

Figure 5.

Gonadotropin-releasing hormone innervation of the arcuate–median eminence. (A and B) Photomicrographs of the ARC–ME of a typical cynomolgus macaque showing the neural area merging into the median eminence of the hypophysial stalk (rostral to the INF–ME of Fig. 4). The pars tuberalis is also evident as a rim of pituitary cells around the edge of the ME. (C and D) Enlarged photomicrographs of the gonadotropin-releasing hormone (GnRH) axons with stained boutons in the ME without (C) and with (D) threshold highlighting (red) of the GnRH-positive pixels. (E) Histogram representing the GnRH analysis. GnRH was highest in the HSR + CIT group, but it was significantly reduced in the HSR + REB group. The SS + CIT group exhibited an intermediate density of GnRH, but REB treatment further reduced GnRH axon density in the ME at the level of the ARC nucleus. There was a significant difference between HSR + CIT and HSR + REB, HSR + CIT and SS + CIT, and between SS + CIT and SS + REB.

Discussion

Cynomolgus monkeys have menstrual cycles like women without regard to season. Also like women, individual cynomolgus macaques differ in their sensitivity to stress that manifests in reproductive function. The application of psychosocial and metabolic stress, with no weight loss, was based on, and designed to duplicate, clinical observations and psychometric test results in women with FHA. Our protocol caused different degrees of anovulation in these animals, which for study purposes were labelled HSR, MSR or SS. Yet, these divisions are more convenient than absolute. With the study of many animals, it became apparent that the differences in individual stress sensitivity fell along a continuum or spectrum, rather than into absolute classifications. We reported that ∼1/3 of each imported population was represented in each stress category. However, in this study, there were 14 animals in the HSR category compared to 5 in each of the MSR and SS categories. We attribute this to less running due to broken treadmills in Pittsburgh that could not be replaced. Although food was reduced to compensate calorically, it was not as effective as hoped. Nonetheless, by comparing the neurobiology of the animals at each end of the spectrum, we were able to document differences that ultimately suggested a relatively coherent, but still sparse, understanding of FHA and a simple intervention in a model that is relevant to women’s health.

Normally, an elevation in FSH at the luteal-follicular transition recruits a cohort of follicles (Mihm et al., 2011) and dysfunction in FSH secretion is regarded as the primary pathology of most cases of FHA. Thus, understanding the regulation of FSH was required to imagine a potential method for normalization in FHA. Variation in GnRH pulses has been accepted as a sufficient mechanism for the differential regulation of FSH and LH, in which slower pulses favour FSH and faster pulses favour LH, although models indicate that the average concentration of GnRH matters most (Pau et al., 1991; Stern et al., 2017; Coss, 2018; Stamatiades and Kaiser, 2018). Nevertheless, other factors such as activin, inhibin and follistatin are also indispensable (Fujii et al., 2002; McNeilly et al., 2003; Bilezikjian et al., 2012) and molecular mechanisms have been further elucidated (Pernasetti et al., 2001; Das and Kumar, 2018). Indeed, the regulation of FSH has emerged as a very complex interplay of hypothalamic, pituitary and ovarian factors operating in the microenvironment of the gonadotrope. A multitude of factors control various intercellular signalling pathways that enable gonadotropes to fluctuate the synthesis of the specific glycoprotein chains, FSHβ and LHβ, as needed for follicle development and luteinization, respectively.

Not all clinical studies found lower FSH in FHA patients compared with controls, but women with luteal phase defect (LPD) exhibit significantly lower serum FSH concentrations during the early follicular phase compared to controls (Soules et al., 1989; Jonard et al., 2005), as well as an increase in LH pulse frequency and amplitude (Suh and Betz, 1993). Clearly, a robust corpus luteum requires a robust follicle, which depends on FSH (Wilks et al., 1976), and a finely tuned FSH:LH ratio is critical (Wilks et al., 1976; Kim et al., 2016). It is attractive to speculate that LPD might be a less severe subtype of FHA in which a poor dominant follicle produces a poorly functioning corpus luteum.

In SS animals on Day 5 of the menstrual cycle, we showed that stress increased NE in the hypothalamic ARC nucleus, which correlated with increased KISS1 mRNA. The increase in KISS1 mRNA occurred with a concomitant increase in GnRH in the MBH, and this in turn correlated with slightly increased serum LH (Bethea et al., 2014). At this time of the menstrual cycle, KISS1 mRNA and kisspeptin in the ARC nucleus should be entirely inhibited by the follicular production of E (Goodman and Lehman, 2012), as demonstrated in HSR animals euthanized on Day 5 in the absence of stress (Bethea et al., 2014).

Therefore, stress increased NE and blocked the efficacy of E negative feedback; in other words, stress and NE decreased sensitivity to E negative feedback. Dissimilarly, in an ovariectomized rhesus macaque model with E replacement, subordinate macaques showed earlier suppression of average LH secretion than dominant macaques (Michopoulos et al., 2009). This was interpreted as increased sensitivity to E negative feedback due to stress. However, the differences between our ovarian-intact cynomolgus model of FHA with psychosocial and metabolic stressors versus long-term ovariectomized rhesus macaques with only psychosocial stress are numerous. It may be moot to compare negative feedback in an FHA model and a menopause model due to the presence and absence of ovaries, as even the quiescent ovaries in FHA produce inhibin. Also, importantly, cynomolgus macaques are overall more sensitive to stress than rhesus macaques. How the feedback processes differ on a cellular or molecular level in these divergent models remains to be discovered.

Stress-induced NE occurs in both HSR and SS animals, but HSR animals have more serotonin, which might ameliorate some of the consequences of stress. Individuals that stopped ovulating immediately upon initiation of our stress regimen had lower endogenous serotonin compared to animals that continue to ovulate through two cycles of stress (Bethea et al., 2005, 2014). Our attempt to elevate serotonin with CIT raised more questions than it solved. Previously, administration of CIT to non-stressed animals increased E and P secretion in SS individuals but had no effect in HSR individuals (Bethea et al., 2011). With stress, CIT was somewhat protective of ovulation in SS animals, but it had a detrimental effect in the HSR group relative to no drug- or REB-Cycle Sets (this study). While the ability of a selective serotonin reuptake inhibitor (SSRI) to affect the ovarian steroid secretion is fascinating, the mechanism needs a great deal more experimental investigation before rational discussion is possible. Clinical results from SSRI administration to women with FHA were also disappointing (Sylvester et al., 2019).

Therefore, we explored the possibility of blocking the stress-induced increase in NE. The pathological elevation of NE was recognized in human anxiety disorders for years, and the accumulated experience in anxiety treatment benefitted our study of FHA. REB, a selective NRI, was approved in Europe (Edronax^TR) and the pharmacological profile indicated that it blocked NET with a 68-fold higher potency than it blocked SERT. The Ki for the DA transporter and other serotonin, adrenergic, dopaminergic, acetylcholine and the H1 receptors was 500–>10 000 nM (Wong et al., 2000; Hajos et al., 2006). Hence, REB was acceptable for the proof of concept in NHPs and side effects reported in humans were not discernable in NHPs at the dose utilized (Burrows et al., 1998; Versiani et al., 1999).

REB administration reinstated ovulation in SS animals under stress as demonstrated by luteal phase P secretion and each cycle was marked by menstruation. Improved ovulation frequency and luteal P secretion were achieved by starting REB on Day 14 of a ‘no stress’ cycle. This time was chosen to ensure consistency in the experimental protocol. In women with FHA and erratic or no ovulation, an NRI could probably be started anytime.

When 14 days of REB treatment preceded and then continued through 5–7 days of stress prior to euthanasia, there was a significant decrease in NE, kisspeptin and GnRH in the ARC–ME–INF complex compared to CIT treatment. Indeed, the density of NE, kisspeptin and GnRH axons in the MBH/ME was the inverse of serum P concentrations. Together, the data support our hypothesis that the lack of FSH-induced follicle development in FHA is a consequence of stress-induced NE in individuals with a compromised serotonin system and that NE causes an increase in kisspeptin and GnRH, thereby detrimentally altering the biological FSH:LH ratio needed by the ovary in the luteal-follicular transition period.

It may be noted that the expression of NE, kisspeptin and GnRH do not precisely follow from each other. That is, while NE and kisspeptin are higher with CIT than REB in a similar pattern in both HSR and SS groups, GnRH expression varied from this pattern. GnRH expression in the ME was decreased in the SS + CIT group compared to the HSR + CIT group, which may merit further discussion. This disconnection suggests that the effect of CIT on GnRH in the SS group may have had some independence from the upstream NE–kisspeptin circuit. How or where this occurred is unknown. The different overall function of the serotonin system between the two groups may play a role, but the mechanism is not readily apparent. Also, confounding the issue could be a difference in measuring GnRH in the MBH versus the ME. Our measurements of GnRH in the MBH have consistently documented higher levels of GnRH in SS than HSR animals in the absence or presence of stress (Centeno et al., 2007; Bethea et al., 2014). However, initial measurements of GnRH in the ME did not agree with GnRH levels in the MBH in non-stressed animals and we have no adequate explanation (Centeno et al., 2007). Herein, we aimed for dense areas of GnRH axons without cell bodies in the ME, but an important caveat of this study is that measurement of GnRH in the MBH may have provided a pattern more consistent with NE/DBH and kisspeptin analysis. Reports of GnRH neuronal remodelling (Prevot et al., 2010) and multiple sites/types of input to each GnRH neuron (Kiss and Halasz, 1985; Dudas and Merchenthaler, 2006; Ojeda et al., 2008; Borsay et al., 2014) point to interesting avenues for future investigation.

At this time, evidence has accumulated that different populations of kisspeptin and GnRH neurons are likely involved in positive and negative feedback of E. That is, one set of kisspeptin or GnRH neurons is inhibited with E feedback (negative feedback) and a different set is activated with elevated E secretion (positive feedback) for the ovulatory surge of LH (Boukhliq et al., 1999; Herbison, 2008; Oka, 2009; Goodman and Lehman, 2012; Smith, 2013; Urbanski, 2015). This study concerns the set of kisspeptin and GnRH neurons that are involved in the negative feedback of E. Still, the negative-feedback set must work properly for follicle development and ultimately for E-activation of the positive-feedback set of neurons responsible for the LH surge. If, as we postulate for FHA, the negative-feedback set of kisspeptin/GnRH neurons is not inhibited by low E due to elevated NE during the early follicular phase, then the microenvironment needed for FSHβ synthesis becomes scrambled, and FSH secretion is compromised. Along this line of reasoning, we propose that NE can determine the sensitivity of the negative-feedback neurons. That is, during the normal menstrual cycle, NE would be low in the early follicular phase when low E effectively inhibits kisspeptin, and in turn GnRH. In the late follicular phase, NE would increase and block the negative feedback, thereby enabling positive feedback to occur. This concept is a minor twist on the long-held paradigm that NE stimulates GnRH for the ovulatory surge of LH, and it applies nicely to actions via two sets of reproductive neurons with differential responses to NE. In primates, the positive-feedback set of GnRH neurons is distinguished by their use of the GnRH-II isoform (Urbanski, 2015).

Another aspect of FSH functionality that we may not adequately appreciate is the post-translational processing required, including proper folding and glycosylation. It is conceivable that the microenvironment impacts FSH processing in ways that we do not comprehend and that this renders secreted FSH less functional. Investigation and further insight into this question in patients with FHA could be fruitful. In truth, there are many more processes in the gonadotrope that are differentially affected by intercellular signalling pathways and that are impacted by the microenvironment but collectively are too vast for this discussion.

We believe that the efficacy of REB in this study is due to its greater action at NET than at SERT. Serotonin and NE uptake inhibitors, such as venlafaxine or levomilnacipran, have higher affinity for SERT than NET (Sorensen et al., 2012) and may not have the same effect in FHA as REB or selective NRIs. In NHPs, funding to show that REB is capable of enabling pregnancy with short treatment duration has not materialized. Therefore, clinical trials of available NRIs are needed to establish the usefulness of this type of treatment in women with FHA and to answer the many questions that arise from bench to bedside. REB (Edronax^TR) is not an available medication for humans in the USA, although another NRI, atomoxetine (Strattera^TR) is marketed for ADHD. An off-label trial of different doses of Edronax for treatment of FHA in the EU could provide important information.

In conclusion, the data in this paper refine the general notion that ‘too little GnRH’ causes FHA (Gordon et al., 2017). We hypothesized that in FHA, NE–kisspeptin–GnRH are elevated enough at the wrong time of the cycle to sabotage the precise signals needed for the gonadotrope to produce adequate or biologically active FSH for follicle development. While monkeys are not identical to humans in all respects, their use in reproductive neuroendocrinology is predicated on the understanding that the regulation of their menstrual cycle is identical to that of women. The neurobiology of HSR and SS macaques as described herein supports the contention that stress-induced NE blocks E negative feedback, which leads to an increase in kisspeptin, and in turn GnRH, which tilts gonadotropes away from proper FSHβ secretion at the critical period of follicle development. In turn, this causes a cascade of events that render the secretion of midcycle ovulatory surges of GnRH and LH impossible. Like all of the other factors in the regulation of ovulation, timing is everything. In addition, we show that ovulation with robust corpus luteum function can be restored during stress in a majority of SS macaques by administration of a selective NRI with higher affinity at NET than at SERT. This treatment decreased NE, kisspeptin and GnRH in the hypothalamus after 5 days of stress thereby providing further support of our hypothesis. It is hoped that this working hypothesis will inspire further investigation into potentially treating the underlying neuropathology of FHA.

Data availability

All data are incorporated into the article and its Supplementary material.