Markers of ovarian reserve are associated with reproductive age acceleration in granulosa cells from IVF patients

Abstract

STUDY QUESTION

Is reproductive aging in granulosa cells associated with markers of ovarian reserve?

SUMMARY ANSWER

Age acceleration was associated with anti-Mullerian hormone (AMH) levels, antral follicle count (AFC), oocyte yield and maturity, and the number of successfully fertilized embryos.

WHAT IS KNOWN ALREADY

The rate of reproductive aging varies among women of the same age. DNA methylation can be used to predict epigenetic age in a variety of tissues.

STUDY DESIGN, SIZE, DURATION

This was a cross-sectional study of 70 women at the time of oocyte retrieval.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The 70 participants were recruited for this study at an academic medical center and they provided follicular fluid samples at the time of oocyte retrieval. Granulosa cells were isolated and assessed on the MethylationEPIC array. Linear regression was used to evaluate the associations between DNA methylation-based age predictions from granulosa cells and chronological age. Age acceleration was calculated as the residual of regressing DNA methylation-based age on chronological age. Linear regressions were used to determine the associations between age acceleration and markers of ovarian reserve and IVF cycle outcomes.

MAIN RESULTS AND THE ROLE OF CHANCE

Participants were a mean of 36.7 ± 3.9 years old. In regards to race, 54% were white, 19% were African American and 27% were of another background. Age acceleration was normally distributed and not associated with chronological age. Age acceleration was negatively associated with AMH levels (t = −3.1, P = 0.003) and AFC (t = −4.0, P = 0.0001), such that women with a higher age acceleration had a lower ovarian reserve. Age acceleration was also negatively correlated with the total number of oocytes retrieved (t = −3.9, P = 0.0002), the number of mature oocytes (t = −3.8, P = 0.0003) and the number of fertilized oocytes or two-pronuclear oocytes (t = −2.8, P = 0.008) in the main analysis.

LIMITATIONS, REASONS FOR CAUTION

This study used pooled follicular fluid, which does not allow for the investigation of individual follicles. Infertility patients may also be different from the general population, but, as we used granulosa cells, the participants had to be from an IVF population.

WIDER IMPLICATIONS OF THE FINDINGS

This study demonstrated that epigenetic age and age acceleration can be calculated from granulosa cells collected at the time of oocyte retrieval. GrimAge most strongly predicted chronological age, and GrimAge acceleration was associated with baseline and cycle characteristics as well as cycle outcomes, which indicates its potential clinical relevance in evaluating both oocyte quantity and quality.

STUDY FUNDING/COMPETING INTEREST(S)

This study was supported by the National Institutes of Health (UL1TR002378) and the Building Interdisciplinary Research Careers in Women’s Health Program (K12HD085850) to A.K.K. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funding source had no role in any aspect of this study. J.B.S. serves as Vice Chair for the American Society for Reproductive Medicine Education Committee, is a Medical Committee Advisor for the Jewish Fertility Foundation and works with Jscreen. J.B.S. has received funding from Georgia Clinical Translational Research Alliance. H.S.H., J.B.S. and A.K.S. have received NIH funding for other projects. A.K.K., S.A.G., S.G., Q.S.K., L.J.M. and W.S. have no conflicts of interest.

TRIAL REGISTRATION NUMBER

N/A.

Introduction

Infertility is a major health problem in the USA with ∼2% of infants being born due to ART, including IVF (Sunderam et al., 2019). In the USA, only 40% of IVF cycles result in a live birth in women under 35 years of age, and there is a well-known decline with advancing reproductive age (Centers for Disease Control and Prevention et al., 2017). This decline is mainly due to a decrease in oocyte quantity (ovarian reserve) and quality as women age (te Velde and Pearson, 2002). This age-related decline occurs at different rates and has traditionally been estimated using serum levels of anti-Mullerian hormone (AMH), which is secreted by the granulosa cells that surround the oocyte in preantral and small antral follicles, consequently serving as a marker of the remaining follicular pool (Moolhuijsen and Visser, 2020). Ovarian reserve can also be assessed via an antral follicle count (AFC) from a transvaginal ultrasound and is associated with the response to ovarian stimulation and the number of retrieved oocytes (Fleming et al., 2015). However, neither measure is a consistent predictor of pregnancy and live birth (Fleming et al., 2015; Moolhuijsen and Visser, 2020).

Age remains the most informative predictor of IVF success, but there is considerable variability around the age at which women reach the end of their reproductive potential and experience menopause (Broekmans et al., 2009; Perheentupa and Huhtaniemi, 2009; Cedars, 2013). This variability may be due to differences between a woman’s biological age at a cellular level and their chronological age. These differences can arise based on a woman’s environment, exposures and health behaviors. One way to quantify these differences in biological age is through the use of epigenetic clocks.

Epigenetic clocks are based on DNA methylation, which is known to change reproducibly with age (Horvath, 2013). DNA methylation is also responsive to the environment and helps to regulate gene expression (Bonasio et al., 2010; Joseph et al., 2018). DNA methylation at cytosine-guanine dinucleotide (CpG) islands typically results in gene silencing, while DNA methylation in gene bodies is associated with increased gene expression (Moore et al., 2013). These properties make DNA methylation a good candidate for characterizing biological age. The first epigenetic clock was based on a weighted average of 353 CpG sites that, together, were able to predict chronological age within ∼3 years in a range of tissues and cell types (Horvath, 2013). Subsequent clocks were built in a single tissue, such as blood, as DNA methylation is tissue-specific (Hannum et al., 2013; Olsen et al., 2020). One such clock has also been developed for granulosa cells (Olsen et al., 2020).

Further research has yielded second-generation epigenetic clocks that incorporate clinical parameters during their development, allowing them to better assess biological age in health-related outcomes (Levine et al., 2018; Lu et al., 2019; Crimmins et al., 2021). This assessment is based on calculations of age acceleration, which is the difference (residual) between chronological and predicted age. Participants with a higher predicted age than their chronological age are experiencing age acceleration, while participants with a lower predicted age than their chronological age are experiencing age deceleration. Age acceleration is typically associated with negative health outcomes, such as cardiovascular disease and cancer, while age deceleration is associated with healthy lifestyle choices, such as exercise (Declerck and Vanden Berghe, 2018).

The relationship between age acceleration and reproductive aging is just beginning to be discovered. In a small pilot study, Monseur and colleagues tested whether age acceleration in blood was associated with indicators of low ovarian reserve prior to menopause. They found that age acceleration was associated with lower AMH values and lower oocyte yield and suggests that age acceleration may be a useful clinical test of ovarian reserve (Monseur et al., 2020). These findings strongly suggest that age acceleration is related to reproductive aging and IVF success.

Since DNA methylation is tissue-specific, assessing reproductive cells may provide more information about underlying reproductive age. Granulosa cells are specialized to promote and support oocyte growth, making them an ideal tissue for studies on reproductive aging (Perheentupa and Huhtaniemi, 2009). Using Horvath’s Pan-tissue clock, one study found that DNA methylation-based age was not associated with age in cumulus cells, and cumulus cells were predicted to be younger than matched white blood cells (Horvath, 2013; Morin et al., 2018). A subsequent study confirmed that there was no association between age and the Pan-tissue clock in granulosa cells, but there was an association with another clock (termed Skin & Blood), which was further adapted to create a ‘Granulosa Cell clock’ that was associated with chronological age (Olsen et al., 2020).

The goal of this study was to build on previous work examining granulosa cell aging by testing its association with markers of ovarian reserve and IVF performance. We hypothesized that epigenetic age acceleration will be associated with markers of lower ovarian reserve and cycle characteristics including poor response to stimulation, lower oocyte yield, lower oocyte maturity rate and lower fertilization rate.

Materials and methods

Participants

Women undergoing an oocyte retrieval at the Emory Reproductive Center were approached to participate in this study prior to the procedure. Both IVF and fertility preservation cycles were included in the study. To elucidate how ovarian aging impacts oocyte quantity and quality, 70 participants were recruited from May 2018 to August 2020 and had DNA methylation data available. Women were included in this study if they were between 18 and 50 years of age, could understand written and spoken English and did not have a history of chemotherapy. Missing data are disclosed in Table I.

Table I.

Participant demographics and IVF cycle characteristics (n = 70).

Characteristic	Mean (SD)
BMI (kg/m2)	26.8 (6.8)
AMH (ng/ml)	3.2 (3.0)
Baseline AFC	16.9 (9.8)
Endometrial thickness at trigger (mm)*	10.1 (3.0)
Total gonadotropin dose (IU)	4098 (2220)
Oocytes retrieved	14.4 (10.0)
Mature (MII) oocytes	11.1 (8.36.4)
Number of successfully fertilized 2PNs#	7.7 (6.1)
	N (%)
Age (years)
<30	1 (1.4)
30–35	23 (32.9)
35–40	33 (47.1)
>40	13 (18.6)
Race
White	38 (54)
Black	13 (19)
Other	19 (27)

^*Two missing values.

^#Twenty-two missing values due to oocyte cryopreservation only.

BMI, body mass index; AMH, anti-Mullerian hormone; AFC, antral follicle count; MII, metaphase II; 2PNs, two-pronuclear oocytes.

Medical record data were abstracted by a qualified physician (S.A. and S.A.G.). Each participant underwent an infertility evaluation prior to the beginning of treatment which included markers of ovarian reserve. AMH was abstracted from the medical record. The Emory Reproductive Center uses Mayo Medical Laboratories, which uses the Ansh laboratory ELISA assay (CV = 3.6–4.6%). In the case of multiple measurements, the AMH closest in time to the IVF cycle which yielded granulosa cells was used (mean = 7.0 months, SD = 6.4 months). The stimulation protocol was based on physician and patient preference. AFC was assessed via transvaginal ultrasound at the baseline visit, which occurs on stimulation Day 1, and it measured all antral follicles between 2 and 10 mm. Follicular development was then monitored by ultrasound and serum estradiol levels throughout the stimulation cycle. The gonadotropin dose was adjusted accordingly, and final oocyte maturation was triggered using human chorionic gonadotropin or a GnRH agonist at the physician’s discretion. The number of follicles >14 mm and endometrial thickness was ascertained via ultrasound on the day of the trigger injection. Ultrasound-guided oocyte retrieval was performed 35 h later. Cycle outcomes were determined by the embryology lab and included the total number of oocytes retrieved, maturity rate, fertilization rate and blastocyst progression rate. Oocyte maturity was determined based on the presence of a polar body in the first 24 h after retrieval. The oocyte maturity rate was calculated as the number of mature oocytes divided by the total number of oocytes retrieved. Both mature oocyte cryopreservation cycles and IVF cycles were included for this measurement. For those subjects undergoing IVF, both ICSI and conventional insemination were included. Successful fertilization was indicated by the presence of two pronuclei. Blastocysts were graded by the Gardner Schoolcraft criteria (Gardner and Schoolcraft, 1999). The blastocyst progression rate was calculated as the number of blastocysts divided by total number of fertilized oocytes.

Biological sample collection and processing

At the time of oocyte retrieval, 70–100 ml of fluid was pooled from multiple follicles from the same participant. Generally, this volume represents fluid from all follicles, but in some cases in which the patient had a very large number of follicles, volumes >100 ml were discarded. Follicular fluid was stored at 4°C until processing. Follicular fluid was processed via centrifugation for 15 min at 1200g, followed by a PBS wash and subsequent centrifugation for 8 min at 300g. The supernatant was discarded and granulosa cells were pelleted, resuspended and stored at −80°C. Granulosa cell samples may include a small proportion of epithelial or thecal cells, as we did not perform a purification step for granulosa cells. Granulosa cell DNA samples were extracted using the QIAmp DNA Mini or DNeasy Blood and Tissue kit (Qiagen, Germany) according to the manufacturer’s instructions and were quantified with the Quant-it Pico Green kit (Thermo Fisher Scientific, USA). A minimum of 500 ng of DNA were plated and then assayed for DNA methylation on the Illumina MethylationEPIC BeadChip (Illumina, USA).

DNA methylation and statistical analysis

DNA methylation data were background and dye bias corrected using the R package ‘minfi’ (Aryee et al., 2014). Initial quality control was performed using the R package ‘CpGassoc’, which removed probes and samples with low signal and missing data (Barfield et al., 2012). Cross-reactive probes identified by McCartney et al. (2016) were removed. The first epigenetic clock, the Pan-tissue clock, was developed using samples from 82 datasets, representing 51 cell and tissue types, with the goal of accurately predicting chronological age. Using an elastic net regression, 353 CpG sites were selected for inclusion into this clock, which predicts age within ∼3 years in a range of cell and tissue types (Horvath, 2013). The Pan-tissue DNA methylation age was calculated using the methods described by Horvath (2013). The next generation of epigenetic clocks was built to predict health span-related outcomes, such as cardiovascular disease and all-cause mortality. The PhenoAge clock was developed using clinical markers, resulting in the selection of 513 CpG sites (Levine et al., 2018). PhenoAge was calculated using the methods described by Levine and colleagues (Horvath, 2013; Levine et al., 2018). Finally, GrimAge was developed with the goal of predicting lifespan. This clock, consisting of 1030 CpG sites, was built using a two-step method that took plasma protein levels, smoking, sex and chronological age into account (Lu et al., 2019). GrimAge was calculated using methods described by Lu and colleagues (Horvath, 2013; Lu et al., 2019). DNA methylation age was also calculated according to a granulosa cell-specific clock developed by Olsen et al. (2020). This clock is based on 391 CpG sites identified from pooled mural granulosa cells, skin and blood samples (Olsen et al., 2020). Linear regression was used to determine the association between each measure of DNA methylation age and chronological age. An overview of epigenetic clocks is presented in Supplementary Table SI. Age acceleration was calculated for all clocks by taking the residual of DNA methylation age on chronological age. Linear regression was used to evaluate the associations between both Granulosa Cell Clock age acceleration and GrimAge acceleration and cycle characteristics and outcomes. Two sensitivity analyses were performed: one controlled for chronological age and the second controlled for AMH.

Ethical approval

All subjects provided written informed consent. This study was approved by the Emory University Institutional Review Board.

Results

Demographic and cycle characteristics/outcomes are presented in Table I. Participants had an average age of 37 years and were racially diverse. AMH levels were associated with response to stimulation (t = 8.1, P = 1.6e−11; Supplementary Fig. S1). The most common reasons for seeking infertility treatment were male factor (47.1%), tubal factor (12.9%) and unexplained infertility (20%). There were 14 participants (20%) on a minimal stimulation protocol (clomiphene citrate and low-dose gonadotropin) for poor responders, while 3 participants (4.3%) were on a high-dose Lupron flare protocol, and 53 (75.7%) participants were on an antagonist protocol with a variable gonadotropin dose depending on ovarian reserve protocol. For 61 participants (87%), hCG was used as the ovulation trigger and for 9 participants (13%), leuprolide acetate was used. Of the 70 women, 48 (69%) had at least one 2-pronuclear oocyte (2PN) and 38 women (54%) had a viable blastocyst at the end of their cycle.

GrimAge acceleration/deceleration was independent of chronological age (P = 1). It was not associated with race (P > 0.45) or with tobacco (t = −0.10, P = 0.92) or alcohol use (t = −0.21, P = 0.83).

To determine which epigenetic clock should be used for this analysis, we considered the Pan-tissue clock, the Granulosa Cell Clock, the PhenoAge clock and the GrimAge clock (Supplementary Table SI). Predicted age was not correlated with chronological age for either the Pan-tissue clock (N = 70, r = 0.02, P = 0.90) or the PhenoAge (N = 70, r = 0.07, P = 0.58) clock. However, correlations with chronological age were observed with the Granulosa Cell Clock age (N = 70, r = 0.44, P = 0.0002) and the GrimAge clocks (N = 70, r = 0.75, P = 9.6e−14). Because GrimAge was most highly correlated with chronological age (Fig. 1), we prioritized it for the main analysis of age acceleration and deceleration and ovarian reserve. Results for all of the clocks are presented in Supplementary Table SII.

Figure 1.

GrimAge is associated with chronological age (t = 9.3, P = 9.6e−14). Points above the line (red) represent women with a chronological age that is older than their GrimAge indicating age acceleration. Points below the line (blue) represent women with a chronological age higher than their GrimAge, indicating age deceleration.

We examined markers of ovarian reserve prior to IVF start. Lower AMH (N = 70, t = −3.1, P = 0.003) was associated with GrimAge acceleration (Fig. 2) even when controlling for age (N = 70, t = −3.2, P = 0.002). At the beginning of ovarian stimulation, AFC was also associated with GrimAge acceleration (N = 70, t = −3.9, P = 0.0002), even when controlling for age (N = 70, t = −4.23, P = 7.2e−5) and AMH (N = 70, t = −3.2, P = 0.002). Next, we assessed endometrial lining thickness and cycle characteristics measured at the end of the stimulation that may be predictive of the number of mature oocytes. While correlated with AFC (N = 69, r = 0.76, P = 4.5e−14), lower numbers of follicles measuring >14 mm on the day of the trigger shot were also associated with GrimAge acceleration (N = 69, t = −3.6, P = 0.0006; Fig. 3), indicating that age acceleration may be associated with having fewer mature oocytes. This association was independent of age (N = 69, t = −3.88, P = 0.0002) and AMH (N = 69, t = −2.0 P = 0.047). The endometrial thickness on the day of the trigger (N = 67, t = −1.40, P = 0.17) was not associated with GrimAge acceleration in granulosa cells. GrimAge acceleration was associated with lower estrogen levels on the day of the trigger (t = −3.5, P = 0.0007). GrimAge acceleration was higher in participants on a clomiphene citrate protocol compared to women on other protocols (t = −2.5, P = 0.02). The trigger type (leuprolide acetate or hCG) was not associated with GrimAge acceleration (t = −1.6, P = 0.13).

Figure 2.

Baseline characteristics associated with GrimAge acceleration (Grim AA), calculated as the residual between GrimAge and chronological age. (A) Anti-Mullerian hormone (AMH), (B) Baseline antral follicle count (AFC), N = 70.

Figure 3.

Cycle outcomes are associated with GrimAge acceleration (Grim AA). (A) The number of follicles >14 mm on the day of the ovulation trigger (N = 69), (B) the total number of oocytes retrieved (N = 70), (C) the number of mature (metaphase II, M2) oocytes retrieved (N = 70) and (D) the number of fertilized oocytes (2PNs, N = 48).

Finally, we assessed indicators of oocyte yield and quality following retrieval. GrimAge acceleration associated with having fewer total oocytes retrieved (N = 70, t = −3.9, P = 0.0002) and fewer mature oocytes (N = 70, t = −3.8, P = 0.0003). Furthermore, GrimAge acceleration was associated with the total number of fertilized oocytes, i.e. the presence of 2PNs (N = 48, t = −2.7, P = 0.008; Fig. 3). These associations also remained significant when controlling for age and AMH (N = 48, 0.02<P < 7.5e−5), except for the association with the number of 2PNs in a sensitivity analysis with AMH (N = 48, t = −1.9, P=.06). These outcomes, though somewhat related, indicate that women with age acceleration have a somewhat poorer prognosis than women with similar markers of the ovarian reserve without age acceleration. However, we did not observe an association between age acceleration and the oocyte maturity rate (N = 70, t = −0.97, P = 0.34).

The total number of embryos progressing to the blastocyst stage was associated with age acceleration (N = 38, t = −2.4, P = 0.02), but the blastocyst progression rate (number of blastocysts divided by number of 2PNs) was not (N = 38, t = −0.42, P = 0.67). The association with the number of blastocysts remained significant in an analysis controlling for age (N = 38, t = −2.4, P=.02) but not in an analysis controlling for AMH (N = 38, t = −1.8, P = 0.07). Thus, age acceleration may be related to the total number of oocytes retrieved, which correlates with the total number of blastocysts produced, but not the blastocyst progression rate. All analyses were also performed in a subset of participants who were over 35 years of age (Supplementary Table SII).

After examining associations with the GrimAge clock described above, we also repeated all analyses using the Granulosa Cell clock, as Granulosa Cell Clock Age was also associated with chronological age in these samples. However, the only association between age acceleration and deceleration measured by the Granulosa Cell clock (Olsen et al., 2020) and any of our variables related to ovarian reserve, cycle characteristics during the stimulation or indicators of cycle success following retrieval, was with levels of AMH (t = 2.0, P = 0.048).

Discussion

These data show that GrimAge acceleration, which is independent of chronological age, is associated with the lower markers of ovarian reserve, AMH and AFC. In addition, we identified associations between GrimAge acceleration and having fewer follicles greater than 14 mm on the day of the trigger shot, lower numbers of total oocytes and mature oocytes retrieved, and the lower numbers of successfully fertilized oocytes and blastocysts. Acknowledging that cycle performance is highly associated with ovarian reserve, we performed sensitivity analyses controlling for AMH. AFC, the number of follicles >14 mm on the day of the ovulation trigger, and the number and maturity of oocytes retrieved remained significant. We also assessed the association of GrimAge acceleration with cycle characteristics and identified associations between GrimAge acceleration and being on a clomiphene citrate protocol. We believe this is due to the lower ovarian reserve in patients on a minimal stimulation protocol, placing them in a higher-risk group for age acceleration. Lower estrogen levels on the day of the ovulation trigger were also associated with age acceleration, which is expected as participants with a larger number of oocytes will produce more estrogen. The identified associations with oocyte quantity and quality suggest that reproductive age acceleration is associated with response to stimulation, and oocyte and blastocyst yield. There was no association between age acceleration/deceleration and the oocyte maturity rate or the blastocyst progression rate. However, we are unable to distinguish whether these results are indicative of oocyte quality versus the quality of granulosa cells.

There are several potential applications for the use of GrimAge and GrimAge acceleration in both clinical and research settings. One of the most pressing questions in the field of ART is how to determine the quality of oocytes. Oocyte quality declines with age and could be due to a number of different factors, including the breakdown of meiotic process and mitochondrial function, telomere attrition or free radicals (Keefe et al., 2015). Poor oocyte quality results in higher rates of aneuploidy, leading to higher rates of miscarriage (Broekmans et al., 2009; MacLennan et al., 2015). Recent studies have proposed biomarkers of oocyte and embryo quality such as oocyte diameter, gene expression and cytokine expression (McKenzie et al., 2004; Devjak et al., 2016; Singh et al., 2016; Bassil et al., 2021). However, they have yet to be implemented in clinical practice and require further research. GrimAge acceleration may offer another unique perspective on oocyte quality and could provide additional information for counseling patients regarding IVF performance if granulosa cells from single follicles were available. We were unable to evaluate individual oocyte quality in this study as only pooled follicular fluid samples were available from each participant. Additional studies are needed to assess the association between GrimAge and risk of miscarriage or live birth rates.

Another potential application of GrimAge acceleration is to use it to quantify the impact on IVF success of potential treatment modifications, including dietary supplements, changes in diet, weight changes and tobacco cessation. GrimAge acceleration could serve as a tool to evaluate different modifications in either a cross-sectional study or clinical trial to provide support for novel treatment recommendations. Clinically, this could give study investigators a quantifiable reflection of how these modifications may be lowering age acceleration and improving IVF performance. Future longitudinal studies of multiple cycles are needed for further evaluation.

In addition to associations between GrimAge acceleration and our variables of interest, we also tested associations with age acceleration from the PhenoAge, the Pan-tissue clock and the Granulosa Cell clock (Supplementary Table SII). However, we caution that associations with age acceleration for clocks not associated with chronological age may not be interpretable. The Granulosa Cell clock was associated with chronological age, although the association was less strong than with the GrimAge clock. This is likely due to these two clocks being composed of different sets of CpG sites, which may capture different aspects of the aging process. In addition, the Granulosa Cell clock may not have performed as well as GrimAge due to the small number of granulosa cell samples used to train the clock being added to the pre-existing Skin & Blood clock training datasets, instead of being built on a larger set of granulosa cell samples alone (Olsen et al., 2020). The Granulosa Cell clock was also trained only on samples from patients with AMH levels between the 25th and 75th percentiles in their dataset, which may limit generalizability to samples from patients with diminished ovarian reserve or polycystic ovary syndrome, who generally have low and high AMH levels, respectively (Olsen et al., 2020).

That study also showed that Horvath’s Pan-tissue clock predicted a younger epigenetic age in granulosa cells but was not associated with chronological age (Horvath, 2013; Olsen et al., 2020). Another study by Morin et al. (2018) also evaluated the Horvath Pan-tissue clock and noted that predicted age was lower in cumulus cells compared to white blood cells. This study also showed no associations between age acceleration in white blood cells and ovarian response to stimulation, and cumulus-cell predicted age was not different between high and low responders in the same age group (Morin et al., 2018). The lack of association between white blood cells and ovarian response may have been due to a small sample size (N = 77, divided into four groups).

There are a few limitations to this study, including having a relatively small sample size and thus a lack of power to examine clinical pregnancy rates, pregnancy complications and live birth rates. We also used pooled follicular fluid to limit the impact on patient care, which does not allow us to examine follicular fluid linked to individual oocytes. However, this study examines summary measures of ovarian reserve and response to stimulation, so a pooled approach is appropriate. Cells from pooled follicular fluid may contain small amounts of thecal or epithelial cells, and there is the possibility that changes in follicular fluid cell methylation may be related to the cellular composition of follicular fluid. In addition, infertility patients may have an altered GrimAge acceleration compared to the general population and thus the results might not be able to be extrapolated to the general population. Finally, this study relied on GrimAge from granulosa cells, which limits its application for fertility patients not pursuing IVF; future studies should evaluate the degree to which GrimAge acceleration in peripheral tissues, such as blood, may be predictive of fertility status.

In conclusion, this study demonstrated that epigenetic age and age acceleration can be calculated from granulosa cells collected at the time of oocyte retrieval. GrimAge most strongly predicted chronological age, and GrimAge acceleration was associated with baseline and cycle characteristics as well as cycle outcomes, which indicates its potential usefulness in evaluating both oocyte quantity and quality. Future directions will include the evaluation of associations with long-term pregnancy outcomes and the use of granulosa cells derived from single follicles to further interrogate oocyte quality.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.