DNA methylation as a window into female reproductive aging

Abstract

People with ovaries experience reproductive aging as their reproductive function and system declines. This has significant implications for both fertility and long-term health, with people experiencing an increased risk of cardiometabolic disorders after menopause. Reproductive aging can be assessed through markers of ovarian reserve, response to fertility treatment or molecular biomarkers, including DNA methylation. Changes in DNA methylation with age associate with poorer reproductive outcomes, and epigenome-wide studies can provide insight into genes and pathways involved. DNA methylation-based epigenetic clocks can quantify biological age in reproductive tissues and systemically. This review provides an overview of hallmarks and theories of aging in the context of the reproductive system, and then focuses on studies of DNA methylation in reproductive tissues.

Plain language summary

People with ovaries experience a natural decline in the function of their reproductive system as they age. This decline eventually leads to menopause, and after menopause, people have an increased risk of developing cardiovascular or other chronic diseases. In the clinic, it is hard to measure aging of the reproductive system, so other markers of the ovary's function, like the number of remaining eggs, are used. We can also measure reproductive aging using molecular biomarkers, which can help us determine when a person's molecular age is different from their chronological age. This review focuses on an overview of biological processes and theories associated with aging, and then focuses on what can be learned from molecular biomarkers.

Tweetable abstract

Biological and chronological aging of the reproductive system can occur at different rates and through different mechanisms. We can quantify consequences of reproductive aging on individual CpG sites or by using epigenetic clocks.

Aging is often defined as the time-dependent functional decline of living organisms [1]. While all body systems age, the functional consequences of aging are likely different based on tissue or organ system. One of the most prominent markers of aging is the loss of reproductive function, which has considerable consequences for fertility and future health. People with ovaries experience this natural decline in function, which leads to a decrease in the quantity and quality of oocytes until reproductive senescence occurs with the onset of menopause. After menopause, people with ovaries experience increased risk for age-related disorders including osteoporosis, memory problems and cardiovascular disease [2–4].

As people reach menopause at different chronological ages [5–7], there is evidence that biological aging of the reproductive system occurs at different rates. When menopause comes early, it has an even more marked effect on health and rates of all-cause mortality [8,9], making biological aging of the reproductive system a potential bellwether of future health. For this review, mechanisms contributing to or resulting from age-related biological decline of the ovaries are referred to as reproductive aging.

Reproductive aging as measured in reproductive tissues may be discrepant from both a person's chronological age and the biological age of their other systems. When a person's biological age is higher than their chronological age, this is referred to as age acceleration, and is generally associated with poor health outcomes. The reverse is also true; when a person's biological age is lower than their chronological age, they are experiencing age deceleration, which is generally associated with better states of health. In the case that reproductive aging is discrepant from systemic aging, operationalized as biological aging measured in blood, aging of the reproductive system may precede aging of other organ systems, which is supported by data linking age at menarche, reproductive capacity and menopause to lifespan and healthspan. Studies have shown that earlier menarche is associated with earlier menopause [10,11], potentially suggesting more allocation of resources to reproduction early on, leading to the end of a person's reproductive potential earlier in life [12]. In addition, a recent meta-analysis has shown that a shorter reproductive lifespan was associated with an increased risk of stroke [13], while other studies have shown earlier menopause is associated with decreased lifespan in humans [14]. Thus, identifying people who have a higher starting point or rate of reproductive aging could lead to the adoption of early intervention strategies. In mice, transplanting ovaries from a young mouse into an aged mouse can improve cognitive and sensory function [15] and extend lifespan [16,17], providing evidence that novel interventions may be able to extend healthspan and lifespan by addressing ovarian aging.

There are many factors that can impact biological aging, which must be taken into account. One method to address this is through assessing DNA methylation. DNA methylation refers to the addition of a methyl group to a cytosine–guanine dinucleotide (CpG site). It has several properties that make it an attractive candidate in the study of reproductive aging. First, DNA methylation at CpG sites helps to regulate gene expression across the life course [18] and is one mechanism through which gene expression responds to changes in the environment, hormone status or other physiologic processes. For example, reproductive aging follows a predictable hormonal pattern, which may be reflected in DNA methylation. DNA methylation also changes predictably with age at some CpG sites, and thus can be used to predict age [19]. This gives us a reliable metric with which to estimate biological age compared with chronological age. Finally, DNA methylation is tissue specific, which allows us to examine aging in different body systems. DNA methylation from blood can be used to estimate systemic biological age, while DNA methylation from components of the reproductive system – such as granulosa cells or oocytes – allow us to examine reproductive system-specific aging.

The purpose of this review is to examine reproductive aging through the lens of DNA methylation. We have included broad discussions of the hallmarks and theories of aging as they relate to reproductive aging, which has been reviewed in detail elsewhere [20,21]. In addition, we have focused primarily on clinical studies and have restricted included studies to either blood or ovarian tissue and cells, as they provide the most direct evidence of 1) how the decline of ovarian function impacts systemic aging in blood, and 2) how reproductive aging progresses in cells taken from the ovary.

Hallmarks of aging

In 2013, Lopez-Otin et al. proposed a framework describing nine hallmarks of aging: genomic instability, telomere shortening, epigenetic alteration, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion and altered intercellular communication [1]. This framework has recently been updated to include disabled macroautophagy, and to split altered intercellular communication into chronic inflammation and age-associated dysbiosis [22]. While a detailed review of each of these factors is out of scope of this review, each factor will be briefly discussed as it relates to reproductive aging and interactactions with DNA methylation (Figure 1). An in-depth review has been conducted by Zhu et al. [23].

Figure 1. . Young ovaries have a robust pool of follicles that diminishes in quantity and quality with time, largely due to the influence of the hallmarks of aging.

Primary hallmarks of aging are genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis and disabled macroautophagy. These hallmarks are central to aging and they result in the accumulation of damage over time. Antagonistic hallmarks including cellular senescence, mitochondrial dysfunction and deregulated nutrient sensing occur in reponse to this damage and display characteristics of antagonistic pleiotropy, described in the text. The integrative hallmarks, dysbiosis, chronic inflammation, altered cellular communication and stem cell exhaustion occur when accumulated damage cannot be ameliorated.

Described by Lopez-Otin et al. [22].

Genomic instability refers to the accumulation of damage to DNA and the age-related decline of the efficiency of DNA repair mechanisms [22]. One example of the impact of this hallmark in reproductive health has recently been shown in a genome-wide association study of more than 200,000 women. They found age at natural menopause was associated with 290 genetic variants, which were enriched for processes involved in DNA repair, including variants known to be expressed during follicular development. Loss-of-function alleles in CHEK2, which plays a role in regulating defective oocytes, were also highlighted; Chek2^-/- female mice experienced a slower depletion of ovarian reserve, which may point to a potential target for interventions aimed at improving in vitro fertilization success [24].

The second hallmark of aging is telomere shortening. Multiple studies have shown an association between shorter telomeres and poor reproductive outcomes, including lower pregnancy rates following in vitro fertilization [25,26]. While both telomere length and epigenetic clocks attempt to quantify biological aging, studies examining their correlation have returned mixed results [27–30]. Thus, they are likely capturing different aspects of aging.

Proteostasis refers to regulating a network of correctly folded proteins while avoiding the accumulation of misfolded proteins or protein aggregates [31]. Heat shock proteins are a key family in maintaining proteostasis, and their dysregulation has been linked with polcystic ovarian syndrome by increasing granulosa cell proliferation and decreasing apoptosis [32], making proteostasis an important process in ovarian function as well as aging.

Deregulated nutrient sensing, as it relates to the hallmarks of aging, occurs in four common pathways: IGF-1, mTOR, AMPK and Sirtuins [22]. There are numerous examples of nutrient sensing in the ovaries that relate to ovarian reserve and function [21]. Geroprotective interventions, such as calorie restriction, can modulate nutrient sensing. Calorie restriction can contribute to infertility, but has also been shown to increase ovarian reserve and result in better reproductive outcomes, which may suggest that calorie restriction induces protective mechanisms to preserve reproductive function while under stress [21]. Calorie restriction is particularly relevant to epigenetic aging as it is one of the few interventions associated with lower biological age [33,34].

Mitochondrial dysfunction also plays a key role in ovarian aging and function. Mitochondria play a vital role in follicular atresia and dysfunction could impact oocyte quality [23]. A recent study has also shown cumulus cells from people with premature ovarian failure and diminished ovarian reserve have lower mtDNA, indicating lower mitochondrial biogenesis [35].

Cellular senescence is an irreversible state of growth arrest that may be triggered by a range of signals including some of the prior hallmarks of aging, such as DNA damage, changes in mitochondrial function and telomere shortening. The accumulation of senescent cells results in aging, and some studies suggest that removing senescent cells can attenuate these effects [36]. Reproductive senescence can be beneficial to preserve reproductive wellbeing, but is detrimental in older age, leading to ovarian and uterine dysfunction [37].

Like cellular senescence, stem cell function is also highly influenced by other hallmarks of aging [22]. Stem cells are vital in allowing tissues to recover from damage, with differentiated cells being able to re-acquire plasticity as needed after injury [22]. Somatic cells can be reprogrammed to induced pluripotent stem cells by the addition of four transcription factors: OCT4, SOX2, KLF4 and MYC. This process results in a seemingly rejuevenated cell and can even reverse epigenetic aging [22]. Owing to this property, one caveat of using epigenetic clocks is that they are currently calculated on a population of cells from a given tissue, which may include stem cells. Thus, stem cell exhaustion may be partially responsible for changes in epigenetic age [38]. Recent studies are attempting to delay or treat ovarian aging using stem cells, but this field is still developing [20].

Chronic inflammation often co-occurs with other hallmarks of aging, but may also drive aging on its own [22]. Inflammation is a crucial part of nonpathogenic ovarian function, especially during follicle rupture for ovulation and subsequent repair. However, inflammation in the context of obesity or polcystic ovarian syndrome can be harmful for ovarian physiology [39–41], which may have implications for accelerating ovarian aging. Accelerated epigenetic aging has been associated with systemic inflammation in blood [42].

Macroautophagy refers to a cellular recycling process that delivers cellular components to lysoymes for degredation [43]. Disabled macroautophagy has been implicated in poor granulosa cell function, and genes associated with autophagy were downregulated in patients with premature ovarian insufficiency [44], suggesting a potential role of macroautophagy in ovarian aging.

Dysbiosis refers to a disruption in the gut microbiome, which usually plays a role in nutrient absorption, metabolite production and defense against potential pathogens. Microbiome compositions may change with age, but current studies are inconclusive [22]. In relation to reproductive health, the gut microbiome modulates estrogen levels, which may contribute to poor health outcomes including sex-hormone driven cancers and endometriosis [45]. In addition, fecal transplantation from young to aged mice was associated with improved ovarian function [46]. Thus far, no studies have examined the relationship between epigenetic age and gut microbiome composition, although this is a promising area of future research.

The final hallmark of aging, epigenetic alterations, is the focus of this review. Epigenetics broadly refers to structural modifications that regulate gene expression without changing the underlying DNA sequence including DNA methylation, histone modifications, chromatin remodeling and noncoding RNAs. While DNA methylation is the focus of this review and is described in detail below, previously listed epigenetic mechanisms are also associated with ovarian function and aging [47–49].

Theories of aging

The antagonistic pleiotropy and disposable soma theories of aging provide useful context around reproduction and menopause. The antagonistic pleiotropy theory posits that the organism makes a trade-off between increased longevity and reproductive fitness [50]. Menopause and its associated hypo-estrogenic state have been proposed as an example of antagonistic pleiotropy. Successful reproductive function, with the exception of lactation, requires an estrogenic environment [51]. Estrogen depletion during lactation allows maternal energy to be focused on caring for her newborn until suckling decreases to conserve energy prior to subsequent reproductive efforts [51]. This hypoestrogenic state in menopause, however, is associated with cardiovascular and musculoskeletal negative health outcomes [51]. Thus, there is a trade-off between reproductive fitness and later health. A special case of antagonistic pleiotropy is the disposable soma theory (accumulation of somatic genetic or epigenetic mutations throughout life due to imperfect DNA repair/maintenance) [50]. This may also play into reproduction, as older parents have likely experienced more deleterious mutations as they reach the end of their reproductive periods than younger parents. Several studies have also shown age-related changes in epigenetic areas important for the maintenance of epigenetic patterns, including the formation of senescence-associated heterochromatin foci and decreased activity of DNA methyltransferases, further supporting this theory [52].

Ovarian reserve, fertility & reproductive aging

The concept of reproductive aging is a framework to better understand how and why people of the same chronological age have different outcomes related to their reproductive system, including response to fertility treatments, age at menopause and long-term health. These distinctions, however, may not always be clear cut; some women will be able to conceive a pregnancy at the same age as other women experience menopause, suggesting they have different reproductive ages [53,54]. There is not currently a clinically accepted test to measure reproductive aging. However, it is possible to clinically estimate ovarian reserve and a rough approximation of time to menopause, which are at least partially linked with reproductive aging.

Ovarian reserve testing most often occurs as people seek treatment for infertility or fertility preservation, both of which are becoming more common as increasing numbers of people delay childbearing until later in life [55–57]. One of the most common markers of ovarian reserve is anti-Müllerian hormone (AMH). AMH is produced by specialized cells inside ovarian follicles called granulosa cells. Higher serum AMH levels correspond to a higher estimated number of follicles in the ovaries [58]. Multiple studies have demonstrated associations between AMH levels and in vitro fertilization (IVF) success, and others have investigated its association with time to menopause, but this is less precise [59–62]. Another informative marker of ovarian reserve is an ultrasound-based antral follicle count. Antral follicles are visible during the later stages of folliculogenesis and correlate with the number of oocytes that could be stimulated to mature during the controlled ovarian hyperstimulation phase of IVF. Of note, while ovarian reserve markers may predict IVF success, they are not indicative of natural fertility [63], indicating that other factors also impact reproductive aging.

In addition to AMH, studies have demonstrated that follicle-stimulating hormone (FSH) and estradiol levels follow a predictable pattern in the years approaching menopause [6,7]. The Study of Women's Health Across the Nation showed that FSH begins to increase approximately 6 years prior to menopause, and estradiol levels decreased up to 2 years prior to menopause [64]. Although studies have not used FSH and estradiol to directly predict age at menopause, these hormones have been shown to play a role in some post-menopausal health outcomes, including osteoporosis and cardiovascular disease [65].

Although ovarian reserve and time to menopause markers are correlated with chronological age, they show considerable variability within age groups. Some people have an estimated ovarian reserve that is discordant with their chronological age, indicating that their chronological age and the biological age of their reproductive system may not be the same. As shown in Figure 2, people with ovaries with increased biological age compared with their chronological age will likely experience menopause earlier, as their oocyte pool is either depleted more quickly or was lower at birth. Furthermore, it is possible that such a difference in biological age may influence systemic health.

Figure 2. . The oocyte pool is depleted throughout the lifespan.

Prior to birth, the oocyte pool is reduced to approximately 1 million eggs. Between birth and puberty, oocyte loss continues so that approximately 300,000 oocytes are present at menarche. The oocyte pool continues to decline steadily after menarche until a more rapid rate of loss begins around 35 years of age. At the time of menopause, the oocyte pool is depleted. As women with the same chronological age can have differences in their estimated ovarian reserve, women who have a higher ovarian reserve than expected for their age are shown in green. Women who have a lower ovarian reserve than expected for their age are shown in red. The color gradient shows the degree of age differences, with darker colors representing more extreme values.

This observation is supported by studies showing that surgical menopause (bilateral oophorectomy) results in a higher risk of cardiovascular disease, cognitive impairment and mortality [66,67]. In a mouse model, Mason et al. demonstrated that transplantation of ovaries from a young mouse into an older mouse resulted in an increased lifespan [16]. Thus, estimating biological age may be an important emerging diagnostic tool when assessing fertility and long-term health.

Epigenetic mechanisms of ovarian aging

DNA methylation can be used to assess mechanisms of ovarian aging by: 1) identifying specific genes and pathways associated with this process; and 2) using epigenetic clocks to assess concordance between chronological age, systemic epigenetic age and reproductive epigenetic age. Genes associated with ovarian aging can be identified through epigenome-wide association studies (EWAS) that result in the identification of individual CpG sites within genes. The genes associated with these CpG sites can then be used in a gene ontology analysis that identifies overrepresented biological pathways. Epigenetic clocks can be calculated using multiple methods, but their underlying principle is that they are based on a weighted average of a pre-identified set of CpG sites. The resulting epigenetic age can be compared with chronological age to determine the presence of age acceleration, when epigenetic age exceeds chronological age, and age deceleration, when epigenetic age is younger than chronological age.

Epigenome-wide association studies

EWAS results in the identification of individual CpG sites associated with a particular phenotype. However, there have been very few EWAS related to ovarian aging. First, Yu et al. examined DNA methylation and gene expression in granulosa cells from young, high responders to ovarian stimulation and older, poor responders to assess methylation changes with reproductive aging. Transcriptomic analysis identified 3397 genes that were differentially expressed between the two groups, including genes known to be related to ovarian function. Using MethylCap-seq and reduced representation bisulfite sequencing, they identified >16,000 differentially methylated sites between the young, high responder and older, poor responder groups. These sites tended to demonstrate shifts towards either very high or very low methylation levels. In addition, they showed that methylation-associated changes in gene expression were enriched in genes where the 3′-end was GC rich, and cite the AMH gene as an example of this effect [68]. The strengths of this study are that it was the first to show an age-related drift of methylation towards extreme values, and was able to correlate age-related changes in DNA methylation to transcript levels to specific gene regions. However, this study only included a small number of pooled samples for DNA methylation analysis (two sets of ten pooled samples from each group) and only 12 samples for transcriptomic analysis. In addition, this study did not examine pathways enriched in the differentially methylated genes alone.

Next, Olsen et al. examined the relationship between AMH and ovarian reserve classification (low, medium, high) in 63 granulosa cell samples [69]. They identified 4199 differentially methylated CpG sites between women with low and normal ovarian reserve, and 447 differences between normal and high ovarian reserve at a false discovery rate (FDR) < 0.05. Gene ontology analysis found few enriched pathways, but did identify enrichment in pathways related to cell-to-cell adhesion [69]. When looking at variability in DNA methylation, the authors identified increased variability in genes involved in reproduction and folliculogenesis [69]. They also performed these analyses in 118 buffy coat samples, which contain a concentrated layer of lymphocytes, monocytes, granulocytes and platelets, but report few associations with ovarian reserve status [69]. This study, although underpowered in some analyses, provides evidence that epigenetic differences in granulosa cells may result in age-related decline in ovarian reserve.

The animal literature provides other examples of the relationship between DNA methylation and ovarian aging. In a bovine model, DNA methylation data was generated from 80 oocyte donors and 277 blood donors. An EWAS identified >500 CpG sites associated with chronological age in blood, while only 141 CpG sites were associated with chronological age in oocytes, but there was only a very weak correlation between CpG sites in the different tissues [70]. Furthermore, in a porcine ovaries, using whole-genome bisulfite sequencing, Xi et al. identified 422 differentially methylated regions associated with young versus old pigs. Hypermethylated regions (n = 303) were associated with genes related to protein binding and apoptotic signaling. Hypomethylated regions (n = 119) showed enrichment for embryonic development and immune system regulation. In addition, hypomethylated genes were also associated with apoptotic signaling [71]. Limitations of this study include that whole ovaries containing multiple cell types were used and that samples only came from the beginning (180 days old) and end (8 years old) of the porcine reproductive life [71]. Overall, all of these EWAS had limited sample size and resolution, but suggest that there are systemic differences in DNA methylation between young and aged oocytes and ovaries.

Epigenetic clocks

Epigenetic clocks have become a widely used and valuable tool to evaluate aging in a range of tissues and health conditions. The first epigenetic clock, credited to Steve Horvath, consisted of 353 CpG sites and was designed to predict age accurately in a range of cell and tissue types [19]. Other first-generation clocks were developed for use in specific tissues, including blood, cord blood, skin, muscle, granulosa cells and placenta [72–78]. These clocks also allowed for the calculation of epigenetic age acceleration, defined as the residual between a person's chronological and predicted age. Follow-up studies found that increased predicted age compared with chronological age (epigenetic age acceleration) was associated with a wide range of negative health outcomes including cardiovascular disease, cancer and all-cause mortality [79,80]. Overall, the defining characteristic of first-generation clocks is that they are designed to predict chronological age as accurately as possible. Each of these clocks uses a weighted average of a set of CpG sites to predict age, although some apply a transformation. However, very few CpG sites are present in more than one clock. Therefore, each clock can have different associations with different aspects of aging [81].

Second-generation epigenetic clocks that take clinical information into account have since been developed with the purpose of assessing healthspan-associated outcomes. For example, the GrimAge clock was trained using a two-stage approach that first developed DNA methylation-based estimators of plasma proteins and smoking, and then combined these markers into a composite score (GrimAge). GrimAge acceleration has been shown to be predictive of time to cancer, time to coronary heart disease and all-cause mortality, which may make it a better predictor of healthspan than first-generation clocks trained to predict only chronological age [82]. Other second-generation clocks take a similar staged approach, but as with first-generation clocks, very few CpG sites are common between multiple clocks and they may associate with different aspects of aging [81].

In addition to their associations with future health, epigenetic clocks may be able to reflect many of the hallmarks of aging. A recent cell culture-based study found that epigenetic age was increased with time in culture, but the rate of epigenetic aging was not impacted by preventing telomere attrition. They also saw that irradiation did not impact epigenetic age, making the ticking of the clock a separate process from genomic instability. Epigenetic age was, however, impacted by disturbing the mTOR nutrient sensing pathway, reducing mitochondrial activity and differences in stem cell fractions. In addition, they showed that while embryonic stem cells and induced pluripotent stem cells do not experience epigenetic aging, differentiation triggers epigenetic aging [83]. Thus, many, but not all hallmarks of aging are reflected in epigenetic clocks.

Several studies have applied these clocks to the study of reproductive aging, both systemically and in local reproductive tissues. Systemic studies have examined the relationship between age acceleration from blood samples and indicators of reproductive aging including parity, fertility, and timing and symptoms of menopause (Table 1).

Table 1. . Overview of key studies related to epigenetic aging.

First author	Clock	Tissue (n)	Population	Main findings	Ref.
Parity
Ryan	Horvath	Blood (397)	USA	Age acceleration increased as number of pregnancies increased	[52]
Kresovich	Horvath, Hannum, and PhenoAge	Blood (2356)	Philippines	Increased age acceleration, based on the Hannum and PhenoAge clocks, was associated with a higher number of live births	[51]
Nishitani	Horvath, Skin & Blood	Saliva (51)	Japan	Decreased age acceleration in both clocks associated with parity status and number of deliveries	[53]
Ovarian reserve
Morin	Horvath	Blood (77), cumulus cells (77)	USA	DNA methylation age of cumulus cells was lower than in blood. Within age groups, DNA methylation age was not associated with ovarian response to stimulation	[54]
Hanson	Horvath	Blood (175), cumulus cells (171)	USA	In a subgroup of women <38 years old, increased age acceleration was associated with poor ovarian response to stimulation in WBCs. A cumulus cell clock could not be constructed	[55]
Olsen	Horvath, Skin & Blood, Granulosa Cell	Blood (118) granulosa cells (59)	Sweden	DNA methylation age (Horvath) in granulosa cells was not correlated with chronological age. The Skin & Blood clock was weakly correlated with chronological age in granulosa cells, and the granulosa cell clock was correlated with chronological age	[57]
Monseur	Horvath, modified with 500 additional targeted loci	Blood (39)	USA	Higher age acceleration was associated with markers of ovarian reserve (AMH, AFC) and oocyte yield, maturity, and blastocyst formation	[56]
Knight	GrimAge†	Granulosa cells (70)	USA	Increased GrimAge acceleration was associated with ovarian reserve (AMH, AFC) and oocyte yield and maturity	[58]
Menopause
Levine	Horvath	Blood (2320), saliva (113), buccal epithelium (790)	USA, Italy, UK	Increased age acceleration in blood is negatively correlated with age at menopause. Buccal and saliva age acceleration were not associated with age at menopause	[61]
Thurston	PhenoAge, GrimAge†	Blood (1194)	USA	Increased PhenoAge was associated with severe hot flashes at enrollment and both increased PhenoAge and GrimAge were associated with later onset of vasomotor symptoms	[62]

^†These studies evaluated other epigenetic clocks, the table represents main findings.

AFC: Antral follicle count; AMH: Anti-Müllerian hormone; WBC: White blood cell.

Parity, or the number of times a person with ovaries has given birth, has been associated with longevity. An epidemiological study from the Women's Health Initiative found that women with a later age at first delivery had increased longevity compared to women with a first delivery before 25 years of age, as did women with two term pregnancies, compared to nulliparous women [84]. Other studies have found that women with two to four term deliveries had the lowest risk of all-cause mortality, compared to both women who had fewer than two and greater than four deliveries [85,86]. Despite these findings, the impact of parity on epigenetic age acceleration is unclear. Ryan et al. and Kresovich and et al. found that increasing parity and gravidity, respectively, were associated with age acceleration in blood samples from young Filipino women, with Ryan examining only the Horvath clock and Kresovich examining the Horvath, Hannum and PhenoAge clocks. However, associations with parity were attenuated after covariate adjustment [87,88]. Nishitani et al. found an increase in parity was associated with age deceleration using the Horvath and Skin & Blood clocks and years of motherhood was associated with age deceleration in the Horvath clock in saliva from Japanese women [89]. These differences could be due to the tissue-specific nature of DNA methylation, the use of different epigenetic clocks or differences based on the populations studied; for example, women in Ryan and Kresovich's studies were in their early twenties while Nishitani's study included women 27–46 years of age.

In addition to gravidity and parity, studies of overall fertility and IVF outcomes are key to understanding ovarian aging. Most human studies are conducted in blood or proxy reproductive tissues such as granulosa cells to avoid the destruction of a potentially viable oocyte. These cells are collected along with the oocyte during an oocyte retrieval procedure. A recent bovine study, however, found no correlation between epigenetic age acceleration in blood and oocytes from the same animal [70], supporting differences between systemic aging in blood and reproductive aging.

There have been several key studies examining age acceleration in cumulus granulosa cells. Morin et al. found that the predicted age using the Horvath clock of cumulus granulosa cells was younger than predicted age in blood samples from the same people (n = 77), but did not show an association between predicted DNA methylation age and response to ovarian hyperstimulation in either tissue [90]. However, another recent study using the Horvath clock did show a relationship between poor ovarian response, defined as less than or equal to five oocytes retrieved, in women less than 38 years of age (n = 146) and age acceleration [91]. This study also attempted to develop a cumulus cell-specific epigenetic clock, but models were unable to reliably predict age [91]. Monseur et al. did report associations between age acceleration and markers of ovarian reserve, including AMH and oocyte yield, from 39 blood samples using the Horvath clock. However, they defined age acceleration in a nontraditional manner as having a DNA methylation-based age that is more than 2 years older than their chronological age [92]. In line with previous findings, Olsen and Hanson showed that granulosa cells had lower predicted ages than blood samples [78,91]. These studies, however, relied on first-generation epigenetic clocks, which may not be applicable to granulosa cells. Thus, Olsen et al. developed a Granulosa Cell clock, but did not evaluate associations with markers of fertility [78]. Finally, we examined granulosa cell age acceleration from 70 participants in the Granulosa Cell clock, the original Horvath clock, and the second-generation GrimAge and PhenoAge clocks [93]. We found associations between markers of ovarian reserve, including AMH levels, antral follicle counts and oocyte yield, among others, and age acceleration based on the GrimAge clock, PhenoAge clock and the Horvath clock [93]. The Granulosa Cell clock age acceleration, however, was only associated with AMH levels [93]. The limitations of this study include that using samples from IVF patients may have limited generalizability, as women with infertility do not necessarily represent the general population. These studies demonstrate that age acceleration, measured using DNA methylation, is associated with ovarian reserve, and that assessment of reproductive aging may require specialized tissue rather than blood samples.

Reproductive aging potentially affects more than fertility as the risks for many age-related disorders increase at the onset of menopause, and later menopause is associated with better long-term health [8,94,95]. Two key studies have examined how reproductive aging relates to epigenetic aging. First, in a large study of more than 12,000 Dutch women, Levine and et al. found that early menopause, surgical menopause and a longer timeframe since menopause were associated with increased epigenetic age acceleration [96]. While the authors suggest several potential causative models, they hypothesize that the directionality of the association implies that menopause increases epigenetic age acceleration. A subsequent study with 1206 participants from the Women's Health Initiative and Observational Study examined the relationship between age acceleration and vasomotor symptoms commonly experienced during menopause, which showed that severe and/or late-occurring symptoms were associated with age acceleration using the PhenoAge and GrimAge clocks [97]. Overall, these studies support using age acceleration as a marker of health during the peri-menopausal period.

As epigenetic age acceleration is associated with many indicators of both reproductive aging and long-term health, it is crucial to investigate whether epigenetic age acceleration can be reversed or slowed. Several studies have shown that a healthy lifestyle including calorie restriction, exercise and stress management are associated with lower age acceleration [33,98,99], but few studies have attempted to reverse age acceleration, which could provide substantial benefits to people with ovaries seeking pregnancy and those experiencing a highly symptomatic menopausal transition. The first clinical trial with the explicit aim of reversing age acceleration was the Thymus Regeneration, Immunorestoration, and Insulin Mitigation (TRIIM) trial. Fahy et al. showed that a combination of recombinant human growth hormone, dehydroepiandrosterone and metformin aimed at regenerating the thymus resulted in decreased age acceleration, which partially persisted for 6 months after treatment discontinuation [100]. The exact mechanism is unknown, but the authors postulate that it is related to cell composition and hematopoiesis [100]. In the animal literature, several studies have shown that caloric restriction and rapamycin can extend lifespan, improve ovarian reserve and result in a lower epigenetic age [33,101–103]. While reversing epigenetic aging is an attractive frontier, the field is in its infancy and much more work will be done to evaluate its clinical consequences.

Conclusion

The function of the reproductive system declines with age, having consequences for both fertility and long-term health. Hallmarks of aging, which can act synergistically, are evident in reproductive tissues. Reproductive aging provides an informative model to study evolutionary theories of aging, especially related to antagonist pleiotropy, as the costs of reproduction may impact longevity. Studying reproductive aging in humans has been hindered by the lack of a reliable clinical test. Currently, AMH levels and antral follicles counts serve as imperfect surrogates to estimate ovarian reserve; these test results do not predict live birth rates and have only modest correlations with development of age-related chronic disease. DNA methylation studies provide many promising avenues to study reproductive aging, both systemically and in reproductive tissues. Several studies have used epigenetic clocks to try to better quantify reproductive aging, with pilot studies showing promising results. The knowledge to be gained from studies of DNA methylation and reproductive aging has the potential to revolutionize our understanding the theories and hallmarks of aging, as well as potential clinical applications.

Future perspective

DNA methylation likely plays a crucial role in controlling the course and timing of reproductive aging. This review has shown that there are strong associations between reproductive aging and age acceleration captured by epigenetic clocks. However, it is unknown if age acceleration is a cause or consequence of advancing age. Recent studies are beginning to address this issue using a Mendelian Randomization approach, which allows for the investigation of causality. Levine et al. performed this type of analysis in the Women's Health Initiative cohort examining the relationship between single-nucleotide polymorphisms associated with age at menopause and age acceleration. They concluded that menopause was likely causal in increasing age acceleration [96]. These types of studies can further our understanding of the sequence of events leading up to advancing reproductive aging and may allow us to gauge whether infertility is a cause or consequence of accelerated aging. We have also identified gaps in the literature related to epigenome-wide studies of reproductive aging in both systemic and proxy tissues (e.g., granulosa cells and ovaries). A promising area for future research would include the development of a second-generation clock for reproductive aging that takes laboratory values and follicle counts into consideration. This would allow for better predictions of age acceleration that are clinically relevant in the context of ovarian aging and fertility. In addition, age acceleration could be used to identify people at high risk for age-related disease and test antiaging interventions and adjunctive therapies as part of IVF. Overall, future studies should focus on including larger, more diverse populations, and further explore the clinical relevance of epigenetic changes related to advancing reproductive age.

Executive summary.

Background

• People with ovaries experience a natural decline in their ovarian function as they age.

• The reproductive system may age at a different rate than other body systems, which may have consequences for long-term health.

• DNA methylation can be used to predict reproductive aging.

Hallmarks of aging

• Each hallmark of aging, as proposed by Lopez-Otin, has the potential to impact reproductive aging.

Theories of aging

• Antagonistic pleiotropy refers to a trade-off between reproductive fitness and longevity.

• Reproduction requires an estrogenic environment. A hypoestrogenic state, seen in menopause, has negative cardiovascular and musculoskeletal consequences for longevity.

Ovarian reserve, fertility & reproductive aging

• Reproductive aging occurs at different rates in people with the same chronological age.

• Ovarian reserve testing likely reflects some aspects of reproductive aging.

Epigenetic mechanisms of ovarian aging

• Genes and pathways associated with reproductive aging can be identified through epigenome-wide association studies.

• Epigenetic age acceleration is based on a weighted average of a set of CpG sites.

EWAS

• Multiple studies have identified differentially methylated sites based on ovarian reserve and between young and aged reproductive tissues.

Epigenetic clocks

• Epigenetic clocks have been developed with the goals of predicting both chronological age and morbidity and mortality.

• Epigenetic clocks have been associated with a variety of reproductive outcomes, such as IVF success, parity and menopause.

• Epigenetic clocks reflect some of the hallmarks of aging, including nutrient sensing, mitochondrial dysfunction and stem cell exhaustion.

Future perspective

• Determining causality between changes in DNA methylation and reproductive aging will be vital to better understand underlying mechanisms.

• Future studies should include larger, more diverse populations and should explore the potential clinical utility of using changes in DNA methylation to better understand or predict reproductive aging.