Menopause, hormone therapy and cognition: maximizing translation from preclinical research

Abstract

Menopause-associated and hormone-associated cognitive research has a rich history built from varied disciplines and species. This review discusses landmark rodent and human work addressing cognitive outcomes associated with varied experiences of menopause and hormone therapy. Critical variables in menopause and cognitive aging research are considered, including menopause etiology, background hormone milieu and parameters of exposure to estrogens and progestogens. Recent preclinical research has identified that menopause and ovarian hormone fluctuations across many neurobiological systems affect cognitive aging, mapping novel avenues for future research. Preclinical models provide insight into complex interdisciplinary relationships in a systematic and highly controlled fashion. We highlight that acknowledging the strengths and weaknesses for both preclinical and clinical research approaches is vital to accurate interpretation, optimal translation and the direction of future research. There is great value in collaboration and communication across preclinical and clinical realms, especially regarding reciprocal feedback of findings to advance preclinical models, improve experimental designs and enrich basic science translation to the clinic. In searching for biological mechanisms underlying the cognitive consequences of menopause and hormone therapies, it is noteworthy that clinical and preclinical scientists are grounded in the same fundamental goal of optimizing health outcomes for women across the lifespan.

Introduction

Accumulating basic science and human research across decades demonstrate that ovarian hormones exert widespread effects on the brain, including in regions that modulate learning and memory. Both endogenous and exogenous hormone fluctuations across the lifespan, as well as the cessation of regular ovarian function during aging, impact cognition. Yet data supporting this tenet are multifaceted and inconsistent. These seemingly discrepant findings reflect the complexity of hormone effects on the brain and cognition, especially as menopause ensues.

The menopause experience is complex and diverse, yielding individualized symptomology and variations in response to changing hormone milieus [1]. Differentiating menopause etiologies (i.e. surgical intervention or natural transition) is therefore important not only conceptually, but in experimental and clinical practice as well. During the menopause transition, some women experience minimal cognitive decline, while others have profound memory disturbances that are not reversed in postmenopause [2]. Variations in age at menopause transition onset and menopause etiology can impact health outcomes, spanning from cognitive decline and dementia risk to cardiovascular and metabolic disease [3–5]. Despite these findings, there remains a tendency to disregard this heterogeneity in menopause symptomology and type in human research, making it difficult to discern the specific drivers of menopause-related outcomes. Preclinical studies can be of great use in gaining fundamental understanding of the drivers of symptomology and biological mechanisms leading to the diverse outcomes observed with variations in menopause etiology and hormone therapy (HT) regimens.

Ovarian extracts have been used since at least the late nineteenth century to treat aging and menopausal symptoms [6]. In the 1927 monograph ‘Internal Secretions of the Ovary’, A. S. Parkes specified that:

… studying the internal secretions of the ovary is most satisfactorily sought by experiment, and since the lower mammals have to be used for this type of work, it is on their reactions that our knowledge of ovarian activity mainly depends … ovarian activity in the human subject must obey similar laws … work on the lower mammals may be made to throw much light on the problems associated with the human species. [7, p.1663]

Early research determined that ovary secretions drive many reproductive-related functions [8]. In 1947, Beach extended this animal work, showing that ovary secretions also impact sex behaviors, not just reproductive organs [9]. We have since learned that ovarian hormones impact nearly every bodily system and many non-reproductive behaviors including cognition; this holds true for animal models and humans. In fact, less than a decade after this surge in research on ovarian-driven effects on physiology and behavior, the first known controlled clinical assessment of 17β-estradiol (E2) was published; in 1952, Caldwell and Watson demonstrated memory enhancements in 75-year-old women after E2 administration [10].

Some of the most straight-forward clinical studies investigating surgical menopause and HT were done in the 1990s. For example, using a repeated-measures design, women receiving bilateral oophorectomy plus hysterectomy were randomly assigned to E2 or placebo post surgery. Women given E2 maintained or improved immediate and delayed verbal memory recall after surgery, while those who did not receive HT declined in verbal memory scores after surgery [11]. This study is especially notable because it systematically documented verbal memory declines concordant with drops in circulating ovarian hormone levels after surgical menopause, and that E2 alone can rescue this detriment. These landmark animal and human studies have driven many subsequent research focus areas. These foci include variations in menopause status, timing of HT, the specific estrogen used and inclusion of a progestogen. Each of these topics is considered in the following.

Reproductive tracts and cyclicity of the aging female rodent and human

While female rodents and humans experience similar patterns of endocrine fluctuations across the reproductive cycle during the fertile life stage, a key difference is that rodents have a 4–5-day estrous cycle and humans a 28-day menstrual cycle [12] (Figure 1). In midlife, rodents experience hypothalamic–pituitary–ovarian disruption, termed estropause, rather than complete ovarian follicle depletion as experienced in menopausal women [13,14]. With estropause, there are prolonged periods of relatively elevated estrogens and progesterone [15,16]; in old age, complete acyclicity can occur [17]. Thus, in estropausal rodents, circulating ovarian hormone levels remain detectable, but regular cyclicity halts.

Figure 1.

Relative shifts in 17β-estradiol (E2) and progesterone levels in female rodents and humans during critical adult time points, including reproductive cyclicity, pregnancy and reproductive senescence. During the rodent estrous cycle, E2 and progesterone levels peak at proestrus and decline at estrus, with progesterone levels at baseline and E2 levels moderate to low before fully declining in metestrus. Progesterone levels at metestrus briefly increase before returning to baseline in diestrus, before proestrus is again initiated. Comparably, in women, E2 levels peak at the periovulatory phase following a rise through the follicular phase. In the luteal phase, progesterone levels rise and E2 is maintained at a moderate level before both return to baseline in the early follicular phase. With pregnancy, both species have robust E2 and progesterone increases (dual bars on the relative y axis indicate markedly higher levels during this period). In the rodent, progesterone levels decline during the final days of pregnancy; in humans, this decline is sharp and occurs just before parturition. Ovarian hormone profiles across species are most distinct during reproductive senescence; in rodents, estropause results in moderate to high E2 levels and moderate progesterone levels. In humans, the menopause transition is characterized by fluctuating E2 levels that gradually decline to near-undetectable levels after menopause, and irregular progesterone levels that likewise show a decline into postmenopausal life.

In humans, persistent ovarian follicular atresia throughout life leads to estrogen and progesterone decline, altering hypothalamic–pituitary–ovarian axis feedback; after menopause, estrogen and progesterone levels become very low or undetectable, yielding considerable rises in pituitary-derived follicle stimulating hormone and luteinizing hormone [18]. Hence, while rodent endocrine profiles shift between the fertile period and estropause, this shift does not specifically mimic menopause in women. Therefore, an unaltered, ovary-intact aging rodent is not an optimal model for investigating the human menopause transition and post-reproductive life stage. Systematic manipulations of the rodent reproductive tract, however, can produce models that recapitulate specific human menopause profiles, permitting interpretive dissection of biological contributors to outcomes from that particular manipulation, while simultaneously allowing strict experimental control.

Modeling variations in type of menopause

Mounting human and animal data reveal that menopause type is key to neurocognitive outcomes. There are numerous variations in gynecological surgery, including, but not limited to, oophorectomy, hysterectomy or oophorectomy plus hysterectomy (Figure 2). Preclinical research evaluating menopause, HT and resulting behavior has primarily been conducted using the ovariectomy (Ovx; surgical rodent procedure corresponding to oophorectomy) model. For Ovx, the ovaries and oviducts (an analog to the fallopian tubes) are removed, while uterine tissue remains [19]. Ovx models surgical ovary removal in women, resulting in rapid declines in circulating ovarian hormones. Oophorectomy, particularly prior to natural menopause, has been associated with negative health outcomes including poorer memory, increased risk of affective disorders, neurodegenerative disease, cardiovascular disease, osteoporosis, diabetes, cancer and all-cause mortality [3,20–25]. Although fewer than 15% of women undergo oophorectomy, the Ovx model, from a basic science perspective, is fundamental to understanding the consequences of complete ovarian hormone loss. Moreover, exogenous HT in an Ovx background allows assessments of a given hormone without interpretive concerns regarding interactions with other ovarian-derived hormones.

Figure 2.

Variations in surgical menopause in women and in rodent models, noting comparative anatomical designations and respective surgical procedures and protocols.

In the USA, up to 600,000 women undergo hysterectomy each year, with about half including oophorectomy [26–28]. Until recently, hysterectomy effects on the brain and behavior have been largely disregarded, given the dogma that the non-pregnant uterus is dormant and ‘useless’ without ovarian hormone stimulation [29,30]. Methodical evaluation of hysterectomy in the context of memory and aging has historically been limited in preclinical and clinical research; for the latter, many studies collapse women with ovaries into one category, whether or not their uterus was maintained.

Our laboratory recently addressed this knowledge gap, assessing variations in surgical menopause using an adult rat model [31] (Figure 2). For hysterectomy with ovarian conservation, uterine horns were ligated below the oviduct and the uterine body was removed at the uterocervical junction. For Ovx plus hysterectomy, the ovaries and oviducts were excised along with the uterus. Ovx was completed by ligating the uterine horn below the oviducts, and removing the ovaries and oviducts. Recognizing comparative anatomy, for human hysterectomy with ovarian conservation, fallopian tubes are often removed with the uterus [32]; however, the anatomical positioning of the rodent oviducts prevented their removal in the case of hysterectomy to avoid damaging ovarian tissues.

Results from this novel animal model showed that hysterectomy with ovarian conservation yielded a unique, detrimental impact on working memory 6 weeks after surgery compared to all other treatment groups. There were some ovarian hormone profile alterations, but ovarian follicle morphology was largely indistinguishable between shamcontrol and hysterectomized rats. This suggests that hysterectomy-induced cognitive impairments were not secondary to ovarian change [31], and different cognitive phenotypes emerged depending on the specific surgical menopause variant. Such variations in cognitive capacities due to menopause etiology require further investigation, especially regarding long-term outcomes during aging and effects of HT within these different models of varied menopause baselines.

While surgical menopause models provide valuable insight into brain and behavioral outcomes, most women undergo a natural menopause transition and the reproductive tract remains intact. Even after follicular depletion, the postmenopausal ovary continues to produce androgens [33,34]. Thus, the presence of follicle-deplete ovarian tissue influences systemic steroid hormone levels, and could impact how HT affects the brain and cognition. Given that a key difference in human versus rodent reproductive senescence is that rodents do not experience total follicular depletion, the ovatoxin 4-vinylcyclohexene-diepoxide (VCD) has been utilized to induce accelerated ovarian follicular atresia in rodents to model transitional menopause [35,36]. VCD targets the finite ovarian follicle pool via apoptotic processes, leading to a significantly reduced ovarian follicle reserve [37,38], with endocrine profiles more similar to those of transitionally menopausal women than Ovx or intact aged female rodent models [34,36,39–41]. With VCD, researchers gain experimental control of follicular depletion induction, and this can be leveraged to ask specific questions regarding timing of transitional menopause and critical windows for HT, while controlling for advanced aging and the presence of an intact, but follicle-deplete, reproductive tract.

Capitalizing on both surgical and transitional menopause models, our laboratory demonstrated that rats receiving Ovx prior to VCD-induced follicular depletion performed worse than rats receiving Ovx following VCD-induced follicular depletion on a delayed memory-retention task. These findings indicate that an ovarian follicle depletion transition before Ovx could benefit cognition as compared to no follicle depletion transition [39]. Moreover, adult rats undergoing transitional menopause showed spatial memory impairments early in follicular depletion [41]. Such findings complement observations that women who experience early menopause due to surgical intervention or premature ovarian failure tend to be more vulnerable to cognitive deficits [20,42,43]. Based on these collective findings, we and others have postulated that a key window of brain reorganization occurs during menopause, and the experience of early or premature menopause disrupts this sensitive window, resulting in susceptibility to cognitive dysfunction [44,45].

Timing of hormone therapy relative to menopause and age: insights from preclinical research

The idea of a critical or sensitive window of hormone effects is rooted in the sexual differentiation literature, wherein there is a limited time frame where steroid hormones organize reproductive, brain and behavioral outcomes in a sex-specific manner. This framing has been extended to other indications, including the critical window around ovarian hormone loss. An early example of this is a study where rats were given different regimens of estrogen or estrogen plus progesterone either immediately, 3 months or 10 months after Ovx. Twelve months later, immediate or early treatment with estrogen or estrogen plus progesterone benefited cognition, while HT at 10 months post surgery yielded no cognitive benefits [46]. Furthermore, Daniel and colleagues performed an elegant series of experiments highlighting the importance of HT timing and duration, demonstrating that immediate, but not delayed, E2 treatment benefits attentional processes [47], and that transient E2 exposure in midlife can have long-lasting memory benefits and enhance hippocampal estrogen receptor expression, even after E2 treatment is discontinued in an Ovx model [48,49]. It is important to note that preclinical cognitive and neurobiological effects of estrogen are not always consistent, even when given in this critical window; menopause type, HT type and behavior testing account for some variations in outcomes [44,50–58].

Use of E2 or conjugated equine estrogens as the estrogenic component in hormone therapy

Human and animal studies have tested the cognitive effects of different estrogens. These include (but are not limited to) E2 and conjugated equine estrogens (CEE), the latter of which is composed of >50% estrone sulfate, more than 10 other estrogen sulfates and trace E2 levels. Human work extending from five decades ago indicates that CEE-containing HT enhances memory through self-report [59], randomized assessments [60] and case studies [61]. The Women’s Health Initiative Memory Study (WHIMS) assessed the effect of HT on the risk of probable dementia, cognitive impairment and global cognition in women aged 65 years and older. CEE was used in the WHIMS since it was the most common estrogenic component of HT during this time. Outcomes indicated an increased risk of probable dementia with CEE plus medroxyprogesterone acetate (MPA), challenging the precept that estrogen-containing HT consistently yields beneficial cognitive outcomes [62].

While there has been much discussion regarding outcomes, it should be noted that the WHIMS invigorated women’s health research, emphasizing the importance of studying menopause, cognitive health and aging. More recent assessments have focused on E2-containing HT. Notable findings include that menopausal women at risk for Alzheimer’s disease taking E2 performed better on a verbal memory task than women taking CEE, suggesting more potent benefits in at-risk populations with E2 specifically [63]. However, in recently menopausal healthy women, neither transdermal E2 nor oral CEE impacted memory in the Kronos Early Estrogen Prevention Study (Cognitive and Affective sub study) (KEEPS Cog) [64]. While it is important to note that each estrogen HT formulation in KEEPS Cog included micronized progesterone, which could impact E2-related outcomes, no differences in cognitive performance were observed on estrogen treatment days versus estrogen plus progestogen treatment days [64].

Rodent work has systematically compared various estrogens for cognitive efficacy. In Ovx rats, E2 afforded greater cognitive benefits than CEE [65]. Based on our findings from testing spatial learning and memory in rodents, the hormones predominantly responsible for the effects of CEE are likely estrone and, after estrone conversion, E2. Other estrogens in CEE have also shown biological and behavioral impacts in rodents, including for cognition in the Ovx rat [66–69]. Our laboratory has also systematically tested E2 HT in surgical (Ovx) and transitional (VCD) rodent menopause models, whereby strong working memory benefits were observed with E2 treatment following surgical menopause [70]. In contrast, in the transitional model of menopause, both learning benefits and retention deficits were observed with E2 treatment [54]. These divergent cognitive effects depending upon menopause history have also been observed with CEE. Indeed, CEE administration benefited spatial working memory with surgical menopause (Ovx), but impaired spatial working memory with transitional menopause (VCD) in rats [71].

Use of progestogens

In women with a uterus, a progestogen must be included in HT to counter endometrial hyperplasia from unopposed estrogens [72,73]. Thus, studying estrogens and progestogens independently and interactively is imperative since they both endogenously circulate across the female reproductive lifespan, and are each HT components. All synthetic progestins bind to progesterone receptors and exert anti-estrogenic effects, resulting in the intended antiproliferative and secretory transformation effects on uterine tissues [74]. Progestins also interact with androgen, glucocorticoid and mineralocorticoid receptors to varying degrees [75], resulting in distinct effects of different progestins.

We have demonstrated that progestins in three different classes had varied cognitive effects using an aging Ovx rat model. Specifically, MPA and norethindrone acetate impaired memory, while levonorgestrel enhanced memory, when given alone [76]. Recently, using an Ovx model, we replicated levonorgestrel-induced memory benefits and E2-induced memory benefits, but found that, when levonorgestrel and E2 were administered together, the mnemonic benefits were obviated [77]. Some of the Women’s Health Initiative and ancillary data have indicated a negative impact of combined HT as well; CEE plus MPA significantly increased dementia risk while CEE only yielded a non-significant increased risk, and CEE versus CEE plus MPA yielded different verbal memory effects [62,78–81]. MPA impairs memory in rodent models across the adult lifespan, showing effects that are long-lasting and potentially non-reversible [82–85]. Detrimental cellular effects with MPA administration have also been observed, including exacerbation of glutamate-induced excitotoxicity, attenuation of estrogen-induced neuroprotection in hippocampal neuronal cell culture [86,87] and altered glutamic acid decarboxylase (GAD)65+67 expression in the hippocampus and entorhinal cortex of middle-aged Ovx rats [82].

Natural progesterone is associated with cognitive impairments in healthy women and with the ‘maternal amnesia’ phenomenon [88,89]. Progesterone administration impairs cognition in young and aged Ovx rodents, reverses beneficial effects of tonic and cyclic E2 HT on spatial memory in middle-aged rats, and abates E2-induced growth factor increases in the rodent entorhinal cortex [82,90–92]. In mice, some literature indicates positive effects of acute progesterone regimens on novelty-related memory, suggesting that, much like estrogens, the progesterone regimen and memory domain impact outcomes [93].

Beyond reproduction: menopause and ovarian hormones impact cognition and brain areas vulnerable to aging

As scientific technology and knowledge have advanced, we have learned that the female brain is not in a fixed state. Females show marked responsiveness to the activational effects of fluctuating ovarian hormones across the lifespan, impacting neurobiology, neurochemistry and behavior. Each of the systems and signaling molecules discussed here have distinct roles in learning and memory, and are altered with aging and the ovarian hormone milieu. The cholinergic system has an integral role in memory, aging and Alzheimer’s disease, and has also been of focus in neuroendocrinology as estrogens impact this system and yield robust behavioral effects [94,95]. Nearly three decades ago, landmark studies from the McEwen laboratory revealed that ovarian hormones impact cholinergic receptors, as well as dendritic spine density, in particular hippocampal subregions [96,97]. Specifically, hippocampal dendritic spine density fluctuates with the rat estrous cycle, with peak elevations during proestrus when E2 levels were highest. Additional work in Ovx rats has demonstrated that E2 treatment increased hippocampal spine density, and progesterone supplementation modulated these effects [98–102]. Moreover, there are links between estrogens, spine density and memory, with work showing that E2 enhances spatial memory parallel to the time frame whereby E2 facilitates hippocampal dendritic spine increases [56,103,104].

There is emerging evidence that serotonin, dopamine, glutamate and GABAergic systems are involved in cognition and neurodegeneration, and are affected by ovarian hormones. For example, in rodents, serotonin modulation in the hippocampus impacts cognition [105], and brain serotonin levels fluctuate across the estrous cycle and with estrogen and progesterone treatment after Ovx [96]. In addition, glutamatergic excitotoxicity and aberrant GABAergic signaling tightly coupled with cognitive decline and neurodegeneration [106–108], and estrogens and progestogens impact GABAergic functioning [82,109,110]. Thus, while less is understood about hormone effects on serotonergic, dopaminergic, glutamatergic and GABAergic systems, there is mounting evidence that each of these are affected by hormone exposure [53,65,111–114].

Studying complex systems using animal models: know your model and what you are modeling

Comparative medicine builds a bridge between animal and human health and science, and has contributed tremendously to the advancement of medicine. There are both drawbacks and benefits of using rodent models to discover insights into human systems, diseases and putative treatments. Drawbacks in animal research include the understanding that, while rodent models afford greater experimental control over key factors, they do not seamlessly recapitulate human conditions, diseases or pathologies. Researchers utilizing animal models must be willing to recognize these limitations in their own work in seeking to better understand the human experience, taking care to implement measured and thoughtful study designs with clearly planned controls. Clinical scientists must also recognize limitations in determining causal relationships from correlational and observational studies, particularly given that menopause is typically highly confounded with age and cannot be randomized in human studies. It is also critical to carefully consider the appropriate sample size and statistical method to ensure adequate power to evaluate interactive effects of key variables (e.g. type of menopause and type of intervention) for both preclinical and clinical designs. The translation of findings from preclinical models to clinical populations is improved when basic scientists and clinical researchers consider the strengths and limitations of each approach.

There are also notable and significant benefits of rodent models. Mice and rats have a short gestation period of about 21 days, reach sexual maturity by 2 months of age and have a comparatively short lifespan of 2–3 years, with 2 weeks of a rat life estimated to be about 1 year of a human life [115]. Thus, longitudinal and aging rodent studies can be completed more rapidly than human studies. Additional benefits include environmental and diet control, systematic and precise experimental manipulations, and control of factors that can impact dependent variables which are difficult or impossible to control in humans. For menopause-related evaluations, clinically significant controllable factors include age at menopause, timing of HT manipulation and history of hormone exposures. Importantly, each of these factors has been shown to affect menopause-related or HT-related health outcomes in women, underscoring the powerful impact rodent models can have on discovering and isolating factors related to human health.

Concluding remarks: leveraging cross-talk across species and disciplines to maximize translation

Optimizing pathways from preclinical research to clinical application necessitates reciprocal feedback between these respective domains. Building bridges to facilitate communication and integrate findings across species will improve the design of preclinical models and experiments, in turn permitting greater translational value since the tested factors were directly identified by clinical work. It is especially noteworthy that, in searching for biological mechanisms underlying the cognitive consequences of menopause and HT, clinical and preclinical scientists are grounded in the same fundamental goal of optimizing health outcomes for women across the lifespan.