Abstract
Cognitive neuroscience of aging studies traditionally target participants age 65 and older. However, epidemiological surveys show that many women report increased forgetfulness earlier in the aging process, as they transition to menopause. In this population-based fMRI study, we stepped back by over a decade to characterize the changes in memory circuitry that occur in early midlife, as a function of sex and women's reproductive stage. Participants (N = 200; age range, 45–55) performed a verbal encoding task during fMRI scanning. Reproductive histories and serologic evaluations were used to determine menopausal status. Results revealed a pronounced impact of reproductive stage on task-evoked hippocampal responses, despite minimal difference in chronological age. Next, we examined the impact of sex and reproductive stage on functional connectivity across task-related brain regions. Postmenopausal women showed enhanced bilateral hippocampal connectivity relative to premenopausal and perimenopausal women. Across women, lower 17β-estradiol concentrations were related to more pronounced alterations in hippocampal connectivity and poorer performance on a subsequent memory retrieval task, strongly implicating sex steroids in the regulation of this circuitry. Finally, subgroup analyses revealed that high-performing postmenopausal women (relative to low and middle performers) exhibited a pattern of brain activity akin to premenopausal women. Together, these findings underscore the importance of considering reproductive stage, not simply chronological age, to identify neuronal and cognitive changes that unfold in the middle decades of life. In keeping with preclinical studies, these human findings suggest that the decline in ovarian estradiol production during menopause plays a significant role in shaping memory circuitry.
SIGNIFICANCE STATEMENT Maintaining intact memory function with age is one of the greatest public health challenges of our time, and women have an increased risk for memory disorders relative to men later in life. We studied adults early in the aging process, as women transition into menopause, to identify neuronal and cognitive changes that unfold in the middle decades of life. Results demonstrate regional and network-level differences in memory encoding-related activity as a function of women's reproductive stage, independent of chronological age. Analyzing data without regard to sex or menopausal status obscured group differences in circuit-level neural strategies associated with successful memory retrieval. These findings suggest that early changes in memory circuitry are evident decades before the age range traditionally targeted by cognitive neuroscience of aging studies.
Introduction
Our society is rapidly aging (Howden and Meyer, 2010), and three out of four older adults report problems with their memory (Koivisto et al., 1995). Thus, maintaining intact memory function with age may be one of the greatest public health challenges of our time. Intervening early with high-risk individuals is critical for the attenuation and prevention of disability, but early targets for treatment have not been identified. Given evidence that women are at a higher risk for memory disorders than men later in life (Gao et al., 1998; Mielke et al., 2014), applying a sex-dependent lens to the study of memory circuitry aging may help identify early antecedents of future memory decline. Moreover, studying adults early in the aging process, as women transition into menopause, may reveal sex-dependent characteristics underlying the aging of memory circuitry in the middle decades of life. Substantial evidence from animal studies indicates that sex steroid hormones, including 17β-estradiol, influence synaptic organization within the hippocampus and prefrontal cortex (PFC), regions that support episodic memory encoding (Bailey et al., 2011; Boulware et al., 2012). Despite clear implications for human health, less is known about the role of sex steroids in the aging of memory circuitry at a human cognitive neuroscience level.
Verbal memory encoding is thought to rely on the coordinated activity of neurons in a distributed set of brain regions, including the hippocampus and other medial temporal lobe structures, ventrolateral and dorsolateral PFC, and posterior parietal cortex (PPC; Buckner et al., 1999; Uncapher et al., 2006; Blumenfeld and Ranganath, 2007; Spaniol et al., 2009; Cabeza and Dennis, 2012). Age-related changes in neural activity during memory encoding are well established (Morrison and Baxter, 2012; Reuter-Lorenz and Park, 2010; Grady and Craik, 2000; Morcom et al., 2003; Rajah et al., 2015), with healthy older adults typically showing altered responses in PFC and hippocampus relative to younger adults. A distinct characteristic of these two regions is that they are densely populated with sex steroid hormone receptors, including estrogen receptors α and β (ER-α and ER-β).
Two decades of experimental evidence in rodents have established estradiol's role in hippocampal structural plasticity (Woolley and McEwen, 1994; McEwen 2002; Brinton 2009; Dimitriu et al., 2010; Liu et al., 2008). For example, dendritic spine density in hippocampal CA1 neurons varies over the course of the estrous cycle (Woolley et al., 1990; Woolley and McEwen, 1993), and ovariectomization leads to a 30% loss in spine density, which is reversed by estradiol replacement (Dimitriu et al., 2010). In nonhuman primates, natural menopause reduces the density of perforated synapse spines, which is correlated with worse recognition memory (Hara et al., 2012). Previous findings at the epigenetic level suggest that estradiol shapes hippocampal-dependent memory in part by inducing chromatin modifications that promote hippocampal plasticity (Fortress and Frick, 2014). In parallel, nonhuman primate studies have made substantial progress toward characterizing the role of estradiol in PFC synaptic plasticity and PFC-mediated cognitive functions (Hao et al., 2006; Morrison et al., 2006; Wang et al., 2010; Rapp et al., 2003). Mounting evidence from human neuroimaging studies further implicates sex steroids in the regulation of memory circuitry (Berman et al., 1997; Shaywitz et al., 1999; Duff and Hampson, 2000; Sherwin, 2003; Grigorova et al., 2006; Dumas et al., 2010; Jacobs and D'Esposito, 2011; Epperson et al., 2012; Hampson and Morley, 2013; Shanmugan and Epperson, 2014; Jacobs et al., 2016). This research builds on the pioneering work of Berman et al. (1997) and Shaywitz et al. (1999), who used pharmacological blockade and hormone replacement techniques to illustrate estradiol and progesterone's influence on regional activity in memory circuitry.
Together, these findings provide converging evidence that functional changes in estrogen receptor-rich regions of memory circuitry are tied to ovarian status. Thus, the decline in ovarian estradiol production during the menopausal transition in women may impact specific neural circuits early in the aging process (Hogervorst et al., 2000; Adams et al., 2002; Morrison et al., 2006; Sherwin, 2006; Brinton 2009; Boulware et al., 2012; Maki and Henderson, 2012; Epperson, 2013; Jacobs et al., 2016). In this population-based functional MRI study of early midlife men and women (N = 200; age range, 45–55), we demonstrate that regional and network-level changes in memory circuitry during verbal encoding are evident early in the aging process as a function sex, women's reproductive stage, and sex steroid hormone concentrations.
Materials and Methods
Subjects.
Participants were selected from 17,741 pregnancies that constitute the New England Family Study (NEFS), a Boston–Providence subsidiary of the national Collaborative Perinatal Project (CPP). The NEFS is a prospective study initiated over 50 years ago to investigate prenatal and familial antecedents of pediatric, neurological, and psychological disorders of childhood (Niswander and Gordon, 1972). Pregnant women, recruited between 1959 and 1966 were representative of patients receiving prenatal care in the Boston–Providence area. In a series of studies over the last 20 years, we have followed a subset of NEFS offspring to investigate the fetal programming of adult psychiatric and general medical disorders and sex differences therein. We recently completed a study of the fetal programming of sex differences in memory circuitry aging (NIMH R01 MH090291). In this set of analyses, 200 offspring (100 males, 100 females) were recruited in early midlife, 45–55 years of age, and completed clinical, cognitive, and neuropsychological assessments in addition to functional and structural magnetic resonance and diffusion tensor imaging (fMRI, structural MRI, Diffusion Tensor Imaging). Exclusionary criteria included any history of neurologic disease, CNS damage, head injury with loss of consciousness, endocrine disorders, heart disease, alcohol-related diseases, current or history of psychosis, other medical illnesses that may significantly alter CNS function, or any MRI contraindication. Four subjects (all men) were excluded from fMRI analyses due to excessive motion (>15% of trials were flagged as motion outliers), and two subjects (both men) did not complete the functional runs, producing a sample of 194 (100 women). Human subjects' approval was granted by Partners Healthcare and Brown University. All volunteers gave written informed consent and were paid for their participation.
Study design and procedures.
Subjects were seen at Brigham and Women's Hospital Outpatient Clinical Research Center. Women who were still menstruating were scheduled in the early follicular phase (day 3–5) of their menstrual cycle, persuant to subject report. Based on common reference ranges (Beckman Coulter), progesterone levels for 75% of premenopausal women were indicative of follicular phase testing; however, eight women had progesterone levels suggestive of luteal phase testing. Subjects fasted for ≥8 h before a morning baseline blood draw. Subjects were offered a light standardized breakfast (excluding caffeine) followed by a 1 h MRI scanning session. Following the scan, subjects completed a structured clinical interview, a basic neuropsychological battery, two behavioral memory tasks, family medical history, and a reproductive/menstrual cycle history administered by an experienced clinical interviewer/clinician in a private testing room.
Neuropsychological assessments.
Participants completed a basic mood and neuropsychological battery that included the following: a digit span (Wechsler, 1997), a Controlled Oral Word Association Test for verbal fluency to the letters F, A, and S (FAS) and categories (Benton, 1968), the American National Adult Reading Test (Nelson, 1982), the State-Trait Anxiety Inventory, and the Profile of Moods Questionnaire (POMS). In addition, two measures of episodic memory, the 12-item Face Name Associative Memory Exam (Sperling et al., 2003; Rentz et al., 2011) and the 6-trial Selective Reminding Test (SRT; Grober et al., 2000; Lemos et al., 2014), were administered based on previous evidence of their increased sensitivity to detecting early cognitive decline (Hedden et al., 2012). Findings from these behavioral memory tasks were described in detail previously (Rentz et al., 2016).
Endocrine assessments.
Trained nurses inserted a saline-lock intravenous line in the nondominant forearm and acquired a fasting morning blood at ∼8:00 A.M. to evaluate hypothalamic–pituitary–gonadal axis hormones, including serum levels of sex steroids (estradiol, progesterone, and testosterone) and gonadotropins [follicle-stimulating hormone (FSH)]. Approximately 10 cc of blood were collected at Brigham and Women's Hospital Center for Clinical Investigation. Samples were allowed to clot for 30–45 min, after which blood was centrifuged (1500 × g for 10 min) and serum was aliquoted into 2 ml microtubes. Serum aliquots were stored at −20°C for neuroendocrine evaluations and archiving. 17β-Estradiol, progesterone, and testosterone concentrations were determined via liquid chromatography–mass spectrometry at the Brigham and Women's Hospital Research Assay Core. Assay sensitivities, dynamic range, and intra-assay coefficients of variation were as follows, respectively: estradiol, 1 pg/ml, 1–500 pg/ml, <5% RSD; progesterone, 0.05 ng/ml, 0.05–10 ng/ml, 5.75% relative standard deviation; testosterone, 1.0 ng/dl, 1–2000 ng/dl, <2% RSD. FSH levels were determined via chemiluminescent assay (Beckman Coulter). The assay sensitivity was 0.2 mIU/ml, the dynamic range was 0.2–200 mIU/ml, and the intra-assay coefficient of variation was 3.1–4.3%. 17β-Estradiol values were log-transformed to achieve a normal distribution.
Menopausal staging.
The timing of menopausal transition between the first clinical appearance of decreased ovarian function (i.e., shorter intermenstrual time periods) to menstrual irregularity and final amenorrhea is highly variable and can occur over several years. Women in this sample were between the ages of 45–55 years and were in various states of ovarian decline. Some women were already in menopause with permanent amenorrhea, low estradiol levels, and elevated gonadotropins; some exhibited signs of follicular failure (elevated FSH and oligoamenorrhea); and some showed normal cycling. Reproductive histories and hormonal evaluations were used to determine the reproductive stage of women in our sample following the Stages of Reproductive Aging Workshop (STRAW)-10 staging system (Harlow et al., 2012). Principal staging criteria were based on menstrual cycle characteristics, with supportive criteria provided by FSH levels. Women were categorized as late reproductive (“premenopause”; n = 32), menopausal transition (“perimenopause”; n = 29), or early postmenopausal (“postmenopause”; n = 31). The women in our sample ranged from Stage −3b, characterized by regular cycling and low FSH, to Stage +1c, characterized by amenorrhea and elevated FSH. An additional eight women reported current use of a hormone replacement regimen and were excluded from all analyses related to reproductive stage. Their data were included in the “supergroup” analyses of fMRI data used to generate generic, task-evoked functional regions of interest (ROIs).
Verbal encoding fMRI paradigm and subsequent memory retrieval task.
Participants performed a verbal encoding task during fMRI scanning (Stone et al., 2005; Golby et al., 2001). The task consisted of two conditions, “Novel” and “Repeat.” In each condition, subjects were presented with a pair of common nouns on a black background presented centrally (4000 ms duration) with a variable interstimulus interval (600–1500 ms). Subjects were asked to silently generate a sentence using both words and were instructed to remember the stimuli for a later test. In the Repeat condition, subjects viewed the same noun pair repeated throughout each block of a run and were instructed to generate the same sentence each time they saw the word pair. In the Novel condition, subjects viewed novel word pairs and generated a new sentence in response to each pair. Subjects were instructed to respond to every word pair with a single button press (pointer finger) to indicate that they had successfully formed a sentence in their mind. Subjects performed two experimental runs of the task. Each run contained three Repeat blocks and three Novel blocks, for a total of six blocks per condition.
A subsequent recognition memory task was administered immediately following the encoding task using the same response box and while the subject remained in the scanner. Subjects viewed single nouns, including 24 previously presented nouns and 24 foils. Each stimulus was presented for 4000 ms with a variable interstimulus interval. Subjects were instructed to indicate, using one of two buttons, whether they had seen the item on the screen in the previous task (yes) or not (no). Response times (RT) and accuracy (d′) were recorded. Response time values <100 ms were considered null and not included in the computation of subjects' average RT. The sensitivity index d′ was calculated (Wickens, 2001) as d′ = z[probability(hits)] − z[probability(false alarms)]. In accordance with signal detection theory, a higher d′ represents a greater distinction between signal and noise (i.e., better signal detection).
fMRI data acquisition.
MRI data were acquired with a Siemens 3T Tim Trio scanner equipped with a 12-channel head coil. Functional data were obtained using a T2* weighted echoplanar imaging sequence sensitive to blood oxygenation level-dependent (BOLD) contrast (repetition time, 2000 ms; echo time, 30 ms; field of view, 200 mm; flip angle, 90°; voxel size, 3.1 × 3.1 × 3.0). Each functional volume consisted of 33 3 mm oblique axial slices. A T1-weighted image was collected using a high resolution 3D multiecho MPRAGE sagittal sequence with an isotropic resolution of 1 mm3. Following acquisition, MRI data were converted to Nifti format and preprocessed in SPM8 (Wellcome Department of Cognitive Neurology, London, United Kingdom). Preprocessing included realignment and geometric unwarping of echoplanar imaging images using magnetic field maps, correction for head motion, nonlinear volume-based spatial normalization (Montreal Neurological Institute template MNI-152), and spatial smoothing with a Gaussian filter (6 mm full-width at half-maximum). Additional software (http://web.mit.edu/swg/software.htm) was used to identify and exclude outliers in the global mean image time series (threshold, 3.0 SD from the mean) and movement (threshold, 1.0 mm; measured as scan-to-scan movement, separately for translation and rotation) parameters. Statistical parametric maps of BOLD activation were calculated in SPM8 using the general linear model approach (Worsley and Friston, 1995).
fMRI data analyses.
Hemodynamic responses were modeled using a gamma function and convolved with onset times of Novel and Repeat blocks to form the general linear model (GLM) at the single subject level. Outlier time points and the six rigid-body movement parameters were included in the GLM as covariates of no interest. To test a priori hypotheses targeting ROIs within memory encoding circuitry (Golby et al., 2001; Blumenfeld and Ranganath 2007; Spaniol et al., 2009; Uncapher and Wagner, 2009), we conducted ROI analyses on functionally defined masks of left dorsal/posterior VLPFC (BA44/BA45/BA9), mid-VLPFC (BA45), ventral/anterior VLPFC (BA47; MNI coordinates, −51, 20, 25; −51, 30, 10; and −39, 26, −2, respectively), left posterior parietal cortex (BA7; −24, −61, 49; 10 mm spheres around peak loci), and anatomically defined masks of the hippocampus. Functional ROIs were defined from whole-brain analyses in the larger sample (N = 194; see Fig. 2B) based on peak task-evoked activity generated at p < 10−16, T = 8.99, and df = 193. The hippocampal ROI was anatomically defined using a manually segmented MNI-152 brain (based on methods previously published by the Center for Morphometric Analysis at Massachusetts General Hospital and Harvard Medical School; Makris et al., 2013). ROIs were implemented as overlays on the SPM8 canonical brain using the Wake Forest University PickAtlas ROI toolbox for SPM (Maldjian et al., 2003). Mean β weights from the ROIs were extracted for each participant as a function of encoding load (Novel > Repeat) using the REX toolbox (Whitfield-Gabrieli, 2009). For each participant and ROI, β estimates were entered into an ANOVA with reproductive status (premenopausal, perimenopausal, postmenopausal) as a between-subjects factor, and age as a covariate. Parallel analyses were run with sex (male vs female) as the between-subjects factor.
To investigate encoding-dependent alterations in functional connectivity by reproductive status and sex, we performed psychophysiological (PPI) analyses with seeds placed in two targeted regions: the hippocampus and dorsal VLPFC (BA44/BA45/BA9). The hippocampus was a clear candidate for examining the impact of reproductive status on functional connectivity given the strong a priori evidence implicating sex steroids in the regulation of this region and the results of the univariate analyses (see Results, Regional BOLD response in VE circuitry). Second, the dorsal VLPFC region showed the highest magnitude of task-evoked activity, making it a compelling candidate for examining functional interactions with PFC. Given the left-hemisphere dominance in verbal encoding tasks (Blumenfeld and Ranganath 2007; Spaniol et al., 2009), time courses from the left VLPFC seed and left hippocampus were extracted for PPI analyses. For each participant, subject-level GLMs were constructed as described earlier, with the addition of the seed time course as a regressor and two additional PPI regressors (the interaction of the seed time course with the regressors for Novel and Repeat conditions). These interaction regressors are orthogonal to the task and seed regressors and describe the contribution of the interaction above and beyond the main effects of the task and seed time course (McLaren et al., 2012). VLPFC and hippocampal connectivity was measured at the single subject level by estimating the difference between the interaction of the seed time course with the regressor for Novel versus Repeat blocks. Single subject activation maps were entered into second-level random effects analysis to probe group differences in memory encoding-dependent VLPFC or hippocampal connectivity. Given extensive evidence for the involvement of posterior parietal cortices in verbal memory tasks (Uncapher and Wagner, 2009) and sex differences in PFC–PPC structural covariance (Abbs et al., 2011), we focused our analyses to examine connectivity between seed regions and PPC by applying small volume correction to a mask of the parietal lobe. The small volume correction approach in SPM8 limits voxelwise analyses to voxels within an a priori volume, defined here as any encoding-sensitive region within parietal cortex. The mask was generated by taking the conjunction of voxels that fell within an anatomical parietal mask and those that demonstrated encoding-related activity from the supergroup (N = 194) activity map. Parallel analyses were run with sex (male vs female) as the between-subjects factor. Connectivity between the left hippocampal seed and right hippocampus was examined by applying a small volume correction across the anatomically defined right hippocampus.
Behavioral data were analyzed with SPSS version 20.0 using ANOVAs for demographic, neuropsychological, and clinical characteristics with continuous variables where equal variance was assumed. ANCOVAs were used to asses group differences in regional BOLD responses after partialling out variance attributable to chronological age. Our strongest a priori hypothesis was that advanced reproductive stage and, in particular, the decline in 17β-estradiol concentrations would alter task-evoked activity in the hippocampus and functional connectivity from the hippocampal seed. We conducted parallel analyses in VLPFC and posterior parietal cortex, two additional key nodes in memory encoding circuitry. Spearman rank-order correlations were conducted to determine the relationship between BOLD β values, behavioral performance, and sex steroid hormone concentrations. Correlations were bootstrapped (1000 iterations), and 95% confidence intervals were computed. A p < 0.05 was designated for statistical significance (any p < 0.08 is noted for completeness).
Results
Demographic, neuropsychological, and clinical characteristics
The community-based sample was 92% Caucasian and 8% African American. Participants were in early midlife (mean age, 49.9; SD, 2.1), with an average of 2 years of college and an average verbal IQ of 116.8 (SD, 10.5). Table 1 reports demographic characteristics of the sample in men and by reproductive stage in women. Groups were comparable on body mass index, education, parental socioeconomic status, estimated verbal IQ, and ethnicity. Although age ranges and mean age were similar across groups (differing, on average, by <18 months), a group difference in age was significant (F(3,182) = 2.8, p = 0.04, η2 = 0.04), showing that premenopausal women were younger than postmenopausal women (t(61) = −2.96, p = 0.009) and men (t(124) = −2.5, p = 0.013). Thus, all analyses included chronological age as a covariate. Perimenopausal and postmenopausal women and men did not differ significantly from one another (all p values >0.2).
Demographic characteristics of the sample in women by menopausal stage and men
Neuropsychological and mood assessments indicated that groups did not differ on verbal fluency (composite of FAS and categories, F(3,181) = 0.40, p = 0.75), digit span (backward, F(3,181) = 0.97, p = 0.41), state anxiety (F(3,181) = 0.09, p = 0.96), trait anxiety (F(3,181) = 0.17, p = 0.92), or mood scores from the POMS (subscores for tension–anxiety, depression–dejection, vigor–activity, fatigue–inertia, confusion–bewilderment, anger–hostility; all F < 1.7, p > 0.2). Performance on two episodic memory tests, the Face Name Associative Memory Exam (Rentz et al., 2011; Papp et al., 2014; Face-Name) and six-trial SRT (Masur et al., 1989), differed by sex and reproductive stage. These findings were reported in detail previously (Rentz et al., 2016). To summarize, in aggregate, women outperformed men on both the Face Name test (free recall, t(184) = −3.26, p < 0.001, two-tailed, r = 0.23) and SRT (delayed recall, t(183) = −4.15, p < 0.0001, two-tailed, r = 0.28). However, subsequent analysis revealed that the Face Name test performance differed by reproductive stage (adjusted for age; F(3,179) = 5.85, p = 0.001, partial η2 = 0.09), with postmenopausal women (mean, 9.34; SE, 0.73) performing worse than premenopausal women (mean, 11.75; SE, 0.72; p = 0.021) and perimenopausal women (mean, 11.67; SE, 0.76; p = 0.028), but not significantly different from men (mean, 8.93; SE, 0.42; p = 0.62). A similar pattern was observed with the SRT task (Rentz et al., 2016). Medication use was determined by structured clinical interview and medical history information. Psychotropic medication use was uncommon and similarly distributed across groups: benzodiazapines (3 premenopausal, 2 perimenopausal, 3 postmenopausal women, 9 men) and antidepressants (selective serotonin reuptake inhibitors; 7 premenopausal, 3 perimenopausal, 4 postmenopausal women, 5 men). Nonpsychotropic medications consisted of those used as a preventative measure for cardiac-related symptoms, e.g., statins (9 premenopausal, 10 perimenopausal, 10 postmenopausal women, 43 men).
Hormonal evaluations
Analysis of sex steroid hormones and gonadotropins confirmed that serum estradiol (F(2,89) = 35.35, p < 0.001, r = 0.67) and progesterone (F(2,89) = 12.04, p < 0.001, r = 0.22) levels declined, while FSH levels rose (F(2,90) = 47.31, p < 0.001, r = 0.52) over the menopausal transition (Fig. 1).
Sex steroid and gonadotropin hormone concentrations by menopausal stage. Serum estradiol and progesterone levels declined and FSH levels rose as a function of advancing reproductive age in women. Pre, Premenopause (n = 32); Peri, perimenopause (n = 29); Post, postmenopause (n = 31). Error bars represent ±1 SEM. ***p < 0.001.
Verbal retrieval behavioral performance
Behavioral performance on the verbal retrieval task was comparable across groups, with no significant differences observed for measures of accuracy (percentage correct, F(3,177) = 0.21, p = 0.89; d′ sensitivity index, F(3,177) = 0.29, p = 0.93) or response time (F(3,177) = 0.41, p = 0.75; Fig. 2A).
Verbal encoding circuitry and subsequent memory retrieval performance. A, Memory retrieval response time (RT) and accuracy (discriminability index, d′) by group. B, Task-evoked BOLD responses throughout verbal encoding circuitry, Novel > Repeat, displayed at p < 10−16 (N = 194). Error bars represent ±1 SEM. Pre, Premenopause (n = 32); Peri, perimenopause (n = 29); Post, postmenopause (n = 31); men (n = 94).
Regional BOLD response in VE circuitry
The verbal encoding paradigm evoked robust responses within memory circuitry regions (Fig. 2B). Region of interest analyses examined group differences in task-evoked activity within VLPFC subdivisions (along a dorsal/posterior to ventral/anterior gradient), hippocampus, and posterior parietal cortex (Fig. 3A). We began by analyzing the data by sex, regardless of women's reproductive status. Adjusting for age and performance, no significant differences between men and women were observed for task-evoked responses within VLPFC or hippocampus (all F(1,173) < 0.7, all p > 0.4). The only exception was PPC, which men activated more strongly than women (F(1,173) = 4.78, p = 0.03, r = 0.16; Fig. 3C).
Modulation of verbal encoding-related fMRI BOLD activity in midlife by menopausal status and sex. A, Surface location of functionally defined masks of VLPFC subdivisions, BA44/BA45/BA9 (turquoise), BA45 (dark blue), BA47 (green), and posterior parietal (BA7, orange) on a rendered brain. Functional ROIs were generated from a supergroup activity map (Novel−Repeat; N = 194). The white box reveals a cutout of the anatomically defined left hippocampal mask (yellow). B, Task-evoked activity in left hippocampus decreased over the transition to menopause. Premenopausal (n = 32) and perimenopausal (n = 29) women exhibited greater activity relative to postmenopausal women (n = 31). C, Men (n = 94) showed greater activity in left posterior parietal cortex relative to women (n = 92), regardless of menopausal stage. Parameter estimates are adjusted for chronological age. Error bars represent ±1 SEM. *p < 0.05.
However, group differences emerged after taking into account the reproductive status of women. Significant changes in hippocampal activity were observed as a function of women's reproductive stage, independent of chronological age. Task-evoked activity in left hippocampus decreased over the menopausal transition (F(2,85) = 3.5, p = 0.035, r = 0.28), with premenopausal (mean, 0.28; SE, 0.06; p = 0.016) and perimenopausal (mean, 0.23; SE, 0.06; p = 0.043) women exhibiting greater activity relative to postmenopausal women (mean, 0.03; SE, 0.06; Fig. 3B). Linear regression analyses indicated that, controlling for age, as endogenous estradiol concentrations declined, the magnitude of left hippocampal activity decreased (β = 0.15, t = 2.0, p = 0.05). Univariate activity in VLPFC and PPC did not differ by women's reproductive status (all F < 1.3, p > 0.3).
Functional connectivity during verbal encoding
Generalized psychophysiological interaction analyses examined group differences in verbal encoding-dependent connectivity from two seed regions, dorsal VLPFC and hippocampus. Analyses by sex revealed stronger intra-VLPFC connectivity in men. Clusters in ventral/anterior VLPFC displayed greater functional connectivity with the dorsal VLPFC seed in men relative to women (coordinates of peak voxel, −33, 29, −8; peak-level pFWE corrected = 0.008; cluster-level pFWE corrected = 0.018; Fig. 4A). Men also showed evidence of heightened connectivity between VLPFC and bilateral inferior parietal lobule (BA40) relative to women. Two clusters (left PPC, −51, −43, 49; right PPC, 54, −49, 52) showed trend-level significance at pFWE corrected = 0.08. At a threshold of pFWE corrected < 0.05, no clusters showed stronger connectivity with the VLPFC seed in women (as a whole) compared to men.
Functional connectivity within encoding circuitry by sex. Left, Relative to women (n = 92), men (n = 94) showed greater functional connectivity between dorsal VLPFC (turquoise seed) and ventral VLPFC (BA47, coordinate of peak voxel, −33, 29, −8; cluster significant at p < 0.05, FWE corrected). Center, Right, Men also exhibited greater connectivity between the VLPFC seed and two clusters in bilateral posterior parietal cortex. Clusters in left BA40 (−51, −43, 49) and right BA40 (54, −49, 52) reached trend significance at p = 0.08, FWE corrected. To illustrate the spatial extent of the pattern of connectivity, clusters are displayed at p < 0.005, uncorrected.
In contrast, women (as a whole) showed greater encoding-related connectivity between the left hippocampal seed and right hippocampus (coordinates of peak voxel, 33, −25, −14; peak-level pFWE corrected = 0.045; cluster-level pFWE corrected = 0.04). However, analyzing data with respect to women's reproductive stage revealed that bilateral hippocampal activity was driven by postmenopausal women, who displayed heightened connectivity relative to premenopausal women, perimenopausal women, and men (all peak-level and cluster-level pFWE corrected < 0.05; Fig. 5A,B). Relative to premenopausal women, postmenopausal women showed heightened connectivity between left hippocampus and three right hippocampal clusters: 21, −10, −26 (peak-level pFWE corrected = 0.007; cluster-level pFWE corrected = 0.027); 27, −40, −2 (peak-level pFWE corrected = 0.033; cluster-level pFWE corrected = 0.042); 21, −10, −26 (peak-level pFWE corrected = 0.041; cluster-level pFWE corrected = 0.042). Relative to perimenopausal women, postmenopausal women showed greater connectivity with one right hippocampal cluster: 21, −10, −26 (peak-level pFWE corrected = 0.029; cluster-level pFWE corrected = 0.042). Finally, relative to men, postmenopausal women showed greater connectivity with two clusters in right hippocampus: 21, −10, −26 (peak-level pFWE corrected = 0.011; cluster-level pFWE corrected = 0.033) and 30, −25, −17 (peak-level pFWE corrected = 0.013; cluster-level pFWE corrected = 0.022).
Hippocampal functional connectivity by menopausal status. A, Left, A slice through right hippocampus (white box) is visible on a rendered brain. Right, Postmenopausal women (n = 31) showed greater connectivity between the left hippocampal seed and multiple clusters in right hippocampus relative to premenopausal (n = 32) and perimenopausal (n = 29) women (all significant at p < 0.05, FWE corrected; for coordinates, see Results). Clusters are displayed at p < 0.005, uncorrected. B, Beta values extracted from the entire right hippocampal mask (542 voxels) represent the average magnitude of connectivity between left (L) and right (R) hippocampus during verbal encoding. Bilateral hippocampal connectivity increased over the menopausal transition. Men shown for comparison. C, Scatter plot displays the correlation between endogenous estradiol concentrations (log transformed) and the magnitude of bilateral hippocampal connectivity in women. As estradiol concentrations decline, the magnitude of bilateral hippocampal connectivity increases. Pre, Premenopause; Peri, perimenopause; Post, postmenopause. Gray dotted lines represent 95% CI. Error bars represent ±1 SEM. +p < 0.06; **p < 0.01; ***p < 0.005.
To plot the magnitude of bilateral hippocampal connectivity across groups and control for chronological age, mean β values were extracted from the entire right hippocampal mask (542 voxels) to avoid biasing the results by selecting clusters known to differ across groups. The magnitude of connectivity between the left hippocampal seed region and right hippocampal ROI (averaged across all voxels) showed a similar impact of menopausal group (F(3,184) = 3.58, p = 0.015, r = 0.24), partialling out variance attributable to age. Post hoc comparisons confirmed that postmenopausal women exhibited greater bilateral hippocampal connectivity compared to premenopausal women (mean, −0.21; SD, 0.54; p = 0.002), perimenopausal women (mean, −0.04; SD, 0.61; p = 0.053), and men (mean, −0.07; SD, 0.52; p = 0.007; Fig. 5B). Finally, to investigate the role of gonadal hormones in shaping bliateral hippocampal connectivity in women, mean PPI β values from the right hippocampal mask (representing the magnitude of functional connectivity with the left hippocampal seed) were related to estradiol concentrations. Linear regression analyses indicated that, controlling for age, as estradiol levels declined, the magnitude of left–right hippocampal connectivity substantially increased (F(2,86) = 8.89, p < 0.001, r = 0.42; estradiol, β = −0.57, t = −3.76, p < 0.001; Fig. 5C). This association was not observed for progesterone (F(2,86) = 0.78, p = 0.65).
In sum, during verbal memory encoding, men showed heightened intra-VLPFC functional connectivity (from more dorsal/posterior to ventral/anterior subdivisions) and some evidence of greater PFC–PPC connectivity relative to women. In contrast, women showed heightened bilateral hippocampal connectivity as a function of advanced menopausal status and declining estradiol levels, independent of chronological age.
Finally, to further investigate the altered pattern of hippocampal functional connectivity in postmenopausal women, participants were subdivided into tertiles of “low” (n = 10), “middle” (n = 11), and “high” (n = 10) performers based on a composite index of performance on the Face Name Associative Memory task (Fig. 6A), a task shown previously to be sensitive to early changes in memory function (Rentz et al., 2011; Hedden et al., 2012). Within postmenopausal women, the highest performers showed the least recruitment of contralateral right HIPP (i.e., smallest magnitude of left–right hippocampal connectivity; F(2,30) = 4.3, p = 0.024, r = 0.49; Fig. 6B). Post hoc comparisons revealed that high performers (mean, −0.28; SD, 0.57) differed significantly from low (mean, 0.67; SD, 1.0; p = 0.007) and middle (mean, 0.38; SD, 0.52; p = 0.045) performers, while low and middle tertiles did not differ from one another (p = 0.39). In addition, postmenopausal women with preserved memory function recruited VLPFC (BA45) more strongly than poorer performing subjects (F(2,30) = 3.36, p = 0.050; high vs low, p = 0.056; high vs middle, p = 0.023; Fig. 6C).
High-performing postmenopausal women show the least bilateral hippocampal connectivity and enhanced VLPFC activity relative to low performers. A, Postmenopausal women were categorized into tertiles of low (n = 10), middle (n = 11), and high (n = 10) performers based on a composite summary score of the Face Name Associative Memory Task (see Materials and Methods). B, C, High-performing women showed the least functional connectivity between left (L) and right (R) hippocampus (B) and the greatest recruitment of VLPFC relative to the two lower performing tertiles (C). Error bars represent ±1 SEM. +p < 0.06; *p < 0.05; **p < 0.01.
Brain–hormone–behavior relationships
The magnitude of encoding-related activity within memory circuitry was strongly related to subsequent memory retrieval. In men, regional activity in mid-VLPFC (BA45; ρ = 0.27; 95% CI, 0.04–0.47; p = 0.008, two-tailed; n = 94) and ventral/anterior VLPFC (BA47; ρ = 0.23; 95% CI, 0.01–0.42; p = 0.03, two-tailed) was related to more successful memory retrieval performance. For women, better retrieval was related to activity in dorsal/posterior VLPFC (BA44/BA45/BA9; ρ = 0.32; 95% CI, 0.11–0.52; p = 0.003, two-tailed; n = 91; Fig. 7A,C). With respect to gonadal hormones, in women, successful retrieval was related to higher endogenous estradiol levels (ρ = 0.29; 95% CI, 0.06–0.49; p = 0.011, two-tailed; Fig. 7B). In men, a similar pattern emerged with higher testosterone levels related to better memory retrieval, but the findings were not statistically significant, in part due to low variability of testosterone among the men (ρ = 0.13; 95% CI, −0.04–0.35; p = 0.13).
Relationship between VLPFC activity and memory retrieval in women and men. A, C, Scatter plots display the correlation between the magnitude of encoding-related activity (β value) in VLPFC subdivisions and performance on a subsequent memory retrieval task. A, In women (n = 92), the magnitude of activity in dorsal VLPFC (BA44/BA45/BA9) predicted better memory retrieval. B, Higher estradiol levels were also associated with better retrieval in women. C, In men (n = 94), activity in mid (BA45) and ventral (BA47) VLPFC subdivisions was associated with more successful retrieval. Gray dotted lines represent 95% CIs.
Discussion
Summary of findings
Cognitive aging studies traditionally target participants aged 65 and older, yet epidemiological surveys show that many women report increased forgetfulness and “brain fog” earlier in the aging process, as they transition through menopause (Greendale et al., 2011). In this population-based fMRI study, we stepped back by over a decade to characterize the changes in memory circuitry that occur in early midlife (age ∼45–55) as a function of sex and women's menopausal status. Women's reproductive stage shaped task-evoked hippocampal activity during verbal memory encoding, despite minimal variance in chronological age. Premenopausal and perimenopausal women recruited left hippocampus more strongly than postmenopausal women. Next, using generalized psychophysiological interaction analysis, we explored the impact of reproductive stage on integrated activity across task-related brain regions. Analyses within women revealed a reorganization of functional networks across the menopausal transition. Postmenopausal women showed heightened bilateral hippocampal connectivity relative to premenopausal and perimenopausal women. Furthermore, the magnitude of regional activity and functional connectivity during encoding was related to verbal memory retrieval.
While the influence of menopausal status was greatest in the hippocampus, a number of sex differences were observed in prefrontal and parietal cortices, which were independent of women's reproductive stage. For example, men showed greater task-evoked activity in left PPC and heightened VLPFC–PPC functional connectivity relative to women. Critically, analyzing these data without regard to sex obscured group differences in the circuit-level neural strategies associated with successful memory performance.
At the neuroendocrine level, lower 17β-estradiol was related to more pronounced alterations in hippocampal activity, hippocampal connectivity, and poorer performance on a subsequent memory task, strongly implicating sex steroids in the regulation of this circuitry. Subgroup analyses revealed that high-performing postmenopausal women (relative to low and middle performers) exhibited a neuronal response pattern similar to that of premenopausal women. Together, these findings underscore the importance of considering reproductive stage and sex steroid hormones, not simply chronological age, to identify neuronal and cognitive changes that unfold in the middle decades of life. More broadly, these findings suggest that early changes in memory circuitry are evident decades before the age range typically targeted by cognitive neuroscience of aging studies.
Study design and limitations
Given that chronological age differed marginally between groups (<18 months, on average), the observed differences in task-evoked BOLD responses are unlikely to be attributable to age-related changes in cerebral vasculature (D'Esposito et al., 2003). Furthermore, the impact of reproductive status on BOLD was region-specific and thus not likely driven by global changes in blood oxygenation. One limitation of this study is its cross-sectional design, which precluded our ability to directly assess incremental changes over the transition to menopause. However, by using a midlife cohort for whom chronological age was similar but menopausal status varied, we were better able to tease apart the influence of reproductive aging from chronological aging. Second, medical histories indicated that some participants in our midlife cohort reported use of a prescription drug. To limit confounding effects of medication status, we ensured that psychotropic medication use, while uncommon, was evenly distributed across groups. Furthermore, plotting ROI β estimates by medication status and group produced highly comparable values across every region of interest. This ensured us that medication status was unlikely to be driving the observed findings.
Changes in memory circuitry are evident in midlife
Despite the tradition in the aging literature of studying adults 65 and older, more attention is being paid to the neural and cognitive changes that happen in the preceding decades, as adults enter midlife. Previous findings suggest that changes in memory encoding and retrieval ability and related neural activity are evident by midlife (Park et al., 2013; Cansino et al., 2015; Kwon et al., 2015). During source encoding, Cansino et al. (2015) observed underrecruitment of prefrontal regions in middle aged relative to younger adults. During spatial and temporal context memory, Kwon et al. (2015) observed no difference in encoding-related brain activity, but heightened PFC activation in middle aged adults during retrieval. Across the small handful of functional MRI studies to investigate memory function in midlife, none reported changes in middle temporal lobe regions (when performance is matched across groups), and changes in PFC function are less consistent across studies (Kennedy et al., 2012; Grady et al., 2006; Kwon et al., 2015). Although these studies represent a critical step toward characterizing early changes in memory circuitry, none of the studies examined their findings with respect to sex or women's reproductive stage. Given that this time period marks one of the most significant periods of hormonal change for women, and given the role of sex steroid hormones in regions central to memory encoding and retrieval processes, considering women's menopausal status may be critical for fully understanding the changes in memory circuitry that take place in midlife. In fact, one of the most consistently observed cognitive changes in women transitioning through menopause is in the domain of verbal learning and memory (Berent-Spillson et al., 2012; Epperson et al., 2013). Advancing our understanding of the hormonal regulation of memory circuitry may provide insights as to why women are at higher risk for memory disorders later in life (Mielke et al., 2014).
Sex differences in midlife
Although no differences were observed in VLPFC as a function of reproductive stage, we observed an overall sex difference in this region in our midlife cohort, with men showing stronger functional connectivity within VLPFC (from dorsal to ventral subdivisions) and between VLPFC and PPC relative to women. We also found that subdivisions of VLPFC were differentially associated with verbal retrieval performance in men and women, with women reliant on dorsal subregions (bordering BA44/BA45/BA9) and men reliant on middle and ventral subregions (BA45 and BA47). Findings from the resting-state literature suggest that these VLPFC subdivisions are part of different functional networks, with the dorsal VLPFC region falling within the dorsal frontoparietal attention network and the more anterior/ventral VLPFC regions falling within the ventral controlled retrieval network (which includes the hippocampus; Nyhus and Badre, 2015). Although speculative, this hints that men and women may be relying on different neural strategies (represented here as different functional networks) to successfully perform the task.
Reproductive stage and estradiol impact hippocampal function
Previous cognitive aging studies found age-related underrecruitment of left VLPFC during episodic encoding (Spaniol and Grady, 2012). Age-related changes in bilateral parietal cortex have also been observed, with greater encoding-related activity in older versus younger adults (unmatched for performance; Sperling et al., 2003; Grady et al., 2006; Murty et al., 2006). In this study, neither VLPFC nor PPC activity differed as a function of reproductive age in women, despite differences in hippocampal activity over this same period. This suggests that the hippocampus may be more sensitive to the impact of reproductive aging than VLPFC or PPC. From a neuroendocrine perspective, this makes sense given that the hippocampus contains some of the highest concentrations of ERα and ERβ in the brain (second only to hypothalamic nuclei and central medial nucleus of the amygdala; Österlund et al., 1999, 2000). In fact, we found that 17β-estradiol concentrations were related to alterations in hippocampal connectivity and poorer performance on a subsequent memory task across women, strongly implicating sex steroids in the modulation of hippocampal function. We did not find evidence of a relationship between progesterone and hippocampal activity, but our data do not rule out the possibility that additional gonadal hormones, including progesterone and its neurosteroid metabolites, contribute to the observed effects. Future studies should examine reproductive age-related changes using more challenging episodic memory tasks, such as tasks that rely on context memory as opposed to item recognition. Doing so may reveal functional differences across a broader network of regions, including PFC (Kwon et al., 2015).
High-performing postmenopausal women show premenopausal-like pattern of neural activity
The purpose of this study was to characterize the impact of menopausal status on memory circuitry function. However, an important exploratory question is whether some women progress through menopause but retain a pattern of brain activity akin to premenopausal women and show elevated cognitive function in domains known to decline as a function of reproductive stage (Epperson et al., 2013; Rentz et al., 2016). To address this, we used an individual-differences driven approach to test whether memory circuitry function differed between low- and high-performing postmenopausal women. Using a strategy similar to previous aging studies (Gazzaley et al., 2005), we defined tertiles of low, middle, and high performers based on the 12-item Face Name Associative Memory Exam. This task was chosen because of its increased sensitivity for detecting early changes in memory function (Rentz et al., 2011, 2016). On average, postmenopausal women showed heightened bilateral hippocampal connectivity during encoding relative to premenopausal and perimenopausal women. However, subgroup analysis revealed that this heightened connectivity was characteristic of low and middle performers and uncharacteristic of postmenopausal women with preserved memory function.
Ample preclinical evidence demonstrates that the decline in ovarian estradiol production during menopause leads to neuronal changes in the hippocampus. For example, Hara et al., 2012 showed that natural menopause in female macaques results in a lower density of perforated synapse spines in the hippocampus and worse recognition memory. Our data support and extend this literature by demonstrating alterations in hippocampal function in women as a function of reproductive stage. Our data are consistent with estradiol's general role in modulating hippocampal function. However, an outstanding question is how the subgroup of high-performing postmenopausal women compensated even in the face of low estradiol. Note that in our sample, low, middle, and high performers did not differ in demographic characteristics (e.g., age, body mass index, socioeconomic status), estradiol and progesterone concentrations, or the number of years since their final menstrual period. One possibility is that as ovarian sources of estradiol decline, secondary estrogenic support from other peripheral sources may play a role in maintaining hippocampal function and hippocampal-dependent memory performance. For high-performing women, the loss of ovarian estradiol may be compensated for through other endocrine pathways, a possibility we are currently pursuing.
Moving forward, understanding the cellular, synaptic, and circuit-level mechanisms for maintaining memory function in the face of reduced ovarian function is a critical challenge for future research, given the potential for identifying therapeutic targets (Frick, 2012). Identifying the distinguishing characteristics between low and high performers will be an important step toward understanding divergent trajectories of cognitive aging as they unfold in the middle years of life.
Conclusions
These results contribute to our broader understanding of the impact of sex and reproductive status on the aging of memory circuitry. In keeping with preclinical studies, our findings suggest that the loss of ovarian estradiol during menopause plays a significant role in shaping memory circuitry.
Footnotes
This work was supported by NIMH Grant MH090291 (J.M.G.); the Office of Research on Women's Health and NICHD Building Interdisciplinary Research Careers in Women's Health program, Grant K12 HD051959 (E.G.J.); and the Harvard Clinical and Translational Science Center (NIH Grant UL1 RR025758). We thank Harlyn Aizley for administering structured clinical interviews, Anne Remington for project administration, Katie Lancaster and Julia Longenecker for help with data collection, and Dr. Shalendar Bhasin for overseeing mass spectrometry.
The authors declare no competing financial interests.
- Correspondence should be addressed to either Dr. Jill M. Goldstein or Dr. Emily G. Jacobs, Brigham and Women's Hospital, Harvard Medical School, 1 Brigham Circle, Boston, MA 02120. jill_goldstein{at}hms.harvard.edu or emily.jacobs{at}psych.ucsb.edu
