Female APOE ɛ4 Carriers with Slow Rates of Biological Aging Have Better Memory Performances Compared to Female ɛ4 Carriers with Accelerated Aging

Abstract

Background:

Evidence suggests that APOE ɛ4 carriers have worse memory performances compared to APOE ɛ4 non-carriers and effects may vary by sex and age. Estimates of biological age, using DNA methylation may enhance understanding of the associations between sex and APOE ɛ4 on cognition.

Objective:

To investigate whether associations between APOE ɛ4 status and memory vary according to rates of biological aging, using a DNA methylation age biomarker, in older men and women without dementia.

Methods:

Data were obtained from 1,771 adults enrolled in the 2016 wave of the Health and Retirement Study. A series of ANCOVAs were used to test the interaction effects of APOE ɛ4 status and aging rates (defined as 1 standard deviation below (i.e., slow rate), or above (i.e., fast rate) their sex-specific mean rate of aging on a composite measure of verbal learning and memory.

Results:

APOE ɛ4 female carriers with slow rates of GrimAge had significantly better memory performances compared to fast and average aging APOE ɛ4 female carriers. There was no effect of aging group rate on memory in the female non-carriers and no significant differences in memory according to age rate in either male APOE ɛ4 carriers or non-carriers.

Conclusion:

Slower rates of aging in female APOE ɛ4 carriers may buffer against the negative effects of the ɛ4 allele on memory. However, longitudinal studies with larger sample sizes are needed to evaluate risk of dementia/memory impairment based on rates of aging in female APOE ɛ4 carriers.

Keywords

Aging epigenetic clocks memory

INTRODUCTION

Alzheimer’s disease (AD) is the most common cause of dementia and the 5th leading cause of death among older adults in the United States (US) [1]. Currently there is no cure or prevention for AD, rendering early diagnosis and detection pivotal to maximizing the efficacy of symptom modifying treatments [2]. To that end, risk stratification has become a common approach to optimizing clinical decision making and informing screening protocols [3, 4]. Frequently referred to as the triad of AD risk factors [5], older age, female sex, and APOE ɛ4 allele, have been extensively investigated risk factors. Advanced age, considered as the single greatest risk factor for late onset AD (LOAD), affects 5% of individuals aged 65–74, 13.1% of individuals aged 75–84 and 33.3% of individuals over 84 years [1]. Of all LOAD cases, two-thirds are women [6], a marked disparity that has largely been attributed to longer survival in women in US studies [7, 8], although evidence from other studies suggest that women are disproportionately affected by LOAD, even after accounting for survival bias [9, 10].

In terms of the genetic risk for AD, the ɛ4 allele of the APOE lipid transporter gene [11] has been identified as the strongest genetic risk for LOAD [12, 13]. The ɛ3 allele is the most prevalent (79%), followed by the ɛ4 allele (15–25%) and the ɛ2 allele (3–12%) [14]. Approximately 50–60% of individuals with LOAD have at least one APOE ɛ4 allele. Risk of LOAD is threefold for carriers of one ɛ4 allele, while two copies of the allele increase risk for LOAD by 9–15 fold [15]. In contrast, the presence of the ɛ2 allele is considered to be protective against LOAD [16, 17], while the ɛ3 allele is associated with minimal or neutral risk for AD [18].

In addition to increasing risk for LOAD, age, sex, and APOE ɛ4 have been associated with differences in cognition in healthy aging [19 –24]. For example, age-related declines in abstract reasoning, working memory, episodic memory, and processing speed decline are widely documented while other cognitive abilities such as vocabulary, general knowledge, and other skill-based knowledge remain stable or even improve with increased age [19, 25]. Regarding sex differences in cognitive abilities, the most consistent findings show that women tend to outperform men on measures of verbal memory while men outperform women on measures of visual-spatial ability [24 , 27]. Additionally, APOE ɛ4 carriers have lower performance on tests of episodic memory, as well as other cognitive abilities such as working memory and processing speed compared to non-carriers [28]. Furthermore, evidence that the cognitive effects of APOE ɛ4 may be exacerbated by increased age and female sex [28, 29], suggests that there are likely independent and interactive effects for each of these factors on cognitive function and risk for LOAD that warrant continued investigation.

Age, sex, and APOE ɛ4 status are typically considered as ‘non-modifiable’ and therefore not direct targets of intervention to slow AD progression; however, with respect to ‘age’, advancements in estimating ‘biological age’ have challenged this assumption [30]. Biological age broadly refers to physiological health of an individual and is distinct from chronological age which refers to the number of years lived [31]. Studies have shown that there is substantial variability in the biological ages of individuals with similar chronological ages such that some people are significantly older or younger than their chronological ages [31]. Furthermore, biological age may be a more accurate predictor of age-related health outcomes such as cancer, diabetes, and depression than chronological age [31, 32].

Although the underlying mechanisms driving differences in biological ages are not well understood, lifestyle factors such as educational attainment, physical activity, and diet have been shown to contribute to variability in biological age [33, 34] suggesting that biological age may be modifiable. However, there is a lack of consensus on a gold standard marker of biological age as numerous phenotypic (e.g., frailty, lipids) and molecular markers (e.g., telomere length, DNA methylation) have been proposed [35]. Despite this challenge, attempts to develop a reliable and accurate aging biomarker continue to gain momentum because of the enormous potential for developing interventions that may slow, arrest, or even reverse biological aging and attenuate the risk of age-related diseases [36, 37].

Currently, DNA methylation-based age estimates or so called ‘epigenetic clocks’ are considered the most reliable estimators of biological age [30]. DNA methylation (DNAm) refers to the addition or removal of methyl groups to cytosine-phosphate-guanine (CpG) dinucleotides sequences clustered in regions referred to as CpG islands [38]. During aging, select changes in these CpG islands occur proximal to genes involved in developmental processes and morphogenesis [38]. DNAm can be influenced by environmental factors and have been proposed to be one mechanism through which cumulative environmental exposures can alter health [39]. Although numerous DNAm clocks have been developed, the first and most commonly investigated clocks include the Horvath pan tissue [40] and Hannum blood [41] clocks, which are based on 353 CpGs and 71 CpGs related to age, respectively. The residuals from regressing these clocks onto chronological age are used as a marker of how fast or slow an individual is aging [30]. Faster aging, indexed from DNAm clocks, have been associated with increased risk of all-cause mortality, physical, psychiatric, and cognitive function as well as several other markers of brain health including magnetic resonance imaging (MRI) structural markers of gray matter volume and white matter hyperintensities [30 , 42–44].

Specific to cognitive aging, the GrimAge clock [45], developed in whole blood samples, has been shown to outperform other clocks in predicting cognitive function and brain health [45, 46], possibly due to the inclusion of other DNAm based markers of plasma proteins associated with health and lifespan. The GrimAge clock also includes a greater number of CpGs (i.e., 1030) relative to the Horvath (n = 353), Hannum (n = 71), or PhenoAge clock (n = 513). In our previous work [44], we showed that women had significantly slower rates of GrimAge compared to men, despite similar chronological age and that sex differences in memory, speed, and executive functioning were differentially mediated by rates of GrimAge. Specifically, women’s slow rates of aging fully accounted for sex differences in processing speed and executive functioning and partially accounted for differences in verbal memory performances. We interpreted the weaker effect of aging rates on women’s verbal memory performances as demonstrating a sex-specific verbal reserve, and parallels findings from other studies [47, 48]. While verbal memory in women may be relatively robust to the effects of aging and age-related disease, accelerated aging may decrease memory resiliency to other adverse factors without significant detriment to memory performances. As such, women with accelerated aging may be more vulnerable to the negative effects of APOE ɛ4 on memory compared to their female peers with slower rates of aging.

Building upon our prior work, we investigated whether there was a differential influence of APOE ɛ4 on verbal learning and memory according to rates of aging. We hypothesized that the effect of APOE ɛ4 on memory would be strongest in women classified a priori as ‘fast agers’ relative to ‘average’ or ‘slower’ agers. While we expected this effect to be greater in women given their overall higher performances (thus more variability to detect differences), we also examined these associations in men to aid interpretations of findings.

METHODS

Participants

Participant data for the current study were obtained from the 2016 Venous Blood Study (VBS) [6] and the Harmonized Cognitive Assessment Protocol (HCAP) [49], sub studies of the 2016 wave of the Health and Retirement Study (HRS). The HRS is an ongoing biennial longitudinal panel study of 20,000 adults aged 50 years or older and living in the United States (US). The HRS is sponsored by the National Institute on Aging (NIA U01AG009740) and is conducted by the University of Michigan. Full details of the study procedures have been previously published [50]. Participants who completed the VBS and HCAP and who had undergone APOE genotyping from salivary DNA collection in either 2006, 2008, 2010, and 2012 were selected. We excluded participant data from our analyses if there was an informant-report (available from the HCAP study) that the participant had a prior history of stroke (n = 159), Parkinson’s disease (n = 23), or AD (n = 36). The final sample included 1,771 participants (mean age = 75, 57% female, 76% White, non-Hispanic, 13% African American, 9% Hispanic, and 2% other race/ethnicity). Among this sample, 415 participants were APOE ɛ4 carriers (40 homozygotes, 375 heterozygotes) and 1,356 non-carriers.

Measures

Biological age

The GrimAge clock [45] is a second-generation clock that incorporates age, sex, a surrogate DNAm marker of smoking pack years and DNAm biomarkers of seven plasma proteins: adrenomedullin, beta-2 microglobulin, cystatin C, leptin, plasminogen activation inhibitor 1, tissue inhibitor metalloproteinase, and growth differentiation factor 15 (GDF 15) to capture physiology risk and stress factors associated with mortality and morbidity [51, 52]. This composite biomarker was developed using elastic net regression models to select CpGs (identified from the Illumina Infinium 450K array) most predictive of lifespan. Following the methods outline by Lu et al. (2019), the GrimAge clock was constructed from whole blood samples collected during the VBS by HRS staff at the University of Michigan and made publicly available via written request [53]. The measure of rate of aging was derived by regressing GrimAge on chronological age to derive the residual. Men and women were defined as ‘slow’, ‘average’, and ‘fast’ agers based on 1 standard deviation (SD) above the sex-specific mean age rate which resulted in 6 groups (3 male: slow n = 92, average n = 543, fast n = 121; and 3 female: slow n = 143, average n = 701, fast n = 171), a classification method for defining thresholds/cut points based on sample distribution for continuous variables [54, 55]. Although this approach to defining cut offs is arbitrary, categorizing individuals in this way would enable us to demarcate aberrant agers from relatively “normal” agers.

APOE ɛ4 carrier status

Saliva samples were collected during home visits by HRS staff during 2006, 2008, 2010, and 2012. DNA extraction and genotyping were performed by researchers at the National Institute of Health Center for Inherited Disease research and archived by the national center for biotechnology information. Full details of genotyping are published elsewhere [56]. In brief, genotyping was performed using predesigned TaqMan allelic discrimination SNP assays rs429358 and rs7412. Recommended quality metrics for using APOE ɛ4 outlined by the HRS recommends excluding participants with imputed data with a posterior probability of <0.8 for either SNP. However, data used in the current study was unaffected by this recommendation as no participant included in the current study analyses had a posterior probability of <0.8. For the current study APOE genotypes were dichotomized to represent carriers with at least 1 ɛ4 allele (coded as ‘1’) and ɛ4 non-carriers (coded as ‘0’). Although some studies have excluded individuals with the ɛ2/ɛ4 genotype based on evidence of a protective effect of the ɛ2 allele against AD [57, 58], other studies have not [59, 60], thus, we choose to include these pairs in our analyses and code as ‘1’ or ‘carriers’ to maximize power in our main analyses. A total of 29 participants had a ɛ2/ɛ4 genotype.

Neuropsychological test measures

Several measures of verbal learning and memory were administered as part of the HCAP. The total number of correct words recalled over three trials from the 10-item word list subtest of the Consortium to Establish a Registry for Alzheimer’s disease (CERAD) [61], the total score from the immediate recall trials of the Logical Memory Wechsler Memory Scale [62], and Brave Man Story from the East Boston Memory Test [63] were used as the measures of verbal learning. The delayed recall scores from each of these three tests as well as the total recognition scores (of correct) from the CERAD wordlist and Logical Memory story were used as the measures of verbal memory. There was not a recognition trial for the Brave Man story. To maximize reliability, we ran two separate unrotated principal components analysis (PCAs) to extract the shared variance among the individual measures of verbal learning and memory. All raw scores were z-transformed prior to the PCAs. The learning PC explained 63 % of the variance across the learning trials, with individual test loading ranging from 0.77 to 0.83. The memory PC explained 67% of the variance across the five measures of memory with loadings ranging from 0.79 to 0.84. Higher PC values indicated higher performances.

Covariates

Age, sex, race/ethnicity, and education (years) were obtained from self-report. To account for acute blood infections that could influence DNA methylation patterns, we adjusted analyses using the GrimAge rate variable for white blood cell (WBC) count as recommended by others [64, 65]. Given associations between depressive symptoms on memory in older adults [66], we also controlled for current self-reported symptoms of depression using the total score form the 11-item Center for Epidemiologic Studies Depression Scale (CESD-11), that was administered as part of the HCAP. Higher scores on the CESD-11 indicated more depressive symptoms.

Analytical approach

All analyses were conducted using the Statistical Package for Social Sciences (SPSS; version 28) for Windows (IBM ®, Armonk, NY, 2021). Descriptive statistics were used to generate means and SDs for all continuous variables or percentages for categorical variables. Pearson’s bivariate correlational analyses was used to examine associations between all variables. We examined differences according to sex and APOE ɛ4 carrier status in chronological age and GrimAge (continuous) using one-way ANOVAs. To supplement the descriptive analyses of the data set, we also compared men and women on verbal learning and memory using ANCOVAs controlling for APOE ɛ4 status, chronological age, CESD scores, and education without the inclusion of GrimAge rates.

To aid interpretation of results from our main analysis, we first examined whether ɛ4 carriers differed according to sex using a chi-square (χ²) test of independence. We then stratified our sample by sex and examined whether APOE ɛ4 carrier status differed by age rate (i.e., slow, average, and fast) using a chi-square test of independence. In a prior study using this sample, we showed that women outperformed men on verbal learning and memory, independent of chronological age and GrimAge rates, suggesting that women may maintain a sex-specific memory advantage/reserve regardless of GrimAge rates. Thus, sex comparisons of the effects of APOE ɛ4 on learning and memory may be confounded by a female advantage in verbal learning/memory so we stratified our sample by sex and examined the effect of APOE ɛ4 status on verbal learning and memory using an ANCOVA. Covariates included in these models included CESD scores, chronological age, and years of education. To assess whether the effect of APOE ɛ4 on learning and memory was greater in men or women, we also tested the APOE ɛ4 x sex interaction in separate models, controlling for covariates.

To test our primary hypotheses that the negative effects of APOE ɛ4 on learning and memory may be greater in women with faster relative to slower rates of aging, we tested the interaction effect of GrimAge rates and APOE ɛ4 status on learning and memory performances using ANCOVAs. For models with the GrimAge rate group, WBC was included as an additional covariate. Post-hoc comparisons were Bonferroni corrected for multiple tests. Statistical significance was set at 95% confidence intervals (CI). The partial eta-squared (η_p²) was used as the measure of effect size. All variables were converted to z-scores prior to running the ANCOVAs to facilitate the interpretation of results.

Sensitivity analyses

In response to the recommendation of a reviewer, we performed several additional analyses to aid interpretation of results. Firstly, given evidence that the ɛ2 allele may be protective against memory decline, we reran our main analyses excluding ɛ2/ɛ4 allele pairs (n = 29). We also examined whether including immune cell types (basophils, eosinophils, lymphocytes, monocyte counts, neutrophils) in addition to white blood cell types as covariates influenced the overall pattern of results, to account for the possibility that immune functioning may be driving results as suggested by prior studies. Finally, we reran the main analyses using PhenoAge clock [32] in place of GrimAge to compare results with this second-generation clock. Although in a prior study, PhenoAge was not associated with cognition using this sample [44], the possibility that there would be a significant interaction with other variables such as APOE ɛ4 carrier status was considered.

RESULTS

Descriptive statistics for the whole and sex-stratified sample are summarized in Table 1 and bivariate correlations are shown in Supplementary Table 1.

Table 1

Sample characteristics (n = 1771)

Variable	Mean or (n)	SD or %	Min.	Max.	APOEɛ4	APOEɛ4
					non-carriers	carriers
					Mean (SD/N)	Mean (SD)
Chronological age	74.84	7.04	64	98	75.03 (7.1)	74.3 (6.7)
Men	75.06	7.10	64	98	75.16 (7.1)	74.59 (7.1)
Women	74.67	6.99	64	96	74.87 (7.2)	74 (6.5)
Education (y)	12.95	3.06	0	17	12.85 (3.1)	13.15 (3.1)
Men	13.26	2.93	0	17	13.12 (3.04)	13.75 (2.5)
Women	12.71	3.14	0	17	12.7 (3.1)	12.76 (3.8)
Race/ethnicity
White		75.8%
Men	(593)	78.4%			(461)	(132)
Women	(750)	73.9%			(583)	(167)
African American		12.8%
Men	(86)	11.4%			(62)	(24)
Women	(140)	13.8%			(93)	(47)
Hispanic		9.5%
Men	(65)	8.6%			(52)	(13)
Women	(103)	10.1%			(78)	(25)
Other race/ethnicity^a		1.9%			(25)	(11)
Men	(12)	1.6%			(5)	(7)
Women	(22)	2.2%			(19)	(3)
DNAm GrimAge	72.10	7.05	55.96	97.05	72.3 (7.1)	71.55 (6.8)
Men	74.10	7.10	57.03	97.05	74.26 (7.0)	73.4 (7.26)
Women	70.61	6.64	55.96	92.53	70.76 (6.8)	70.12 (6.1)
DNAm GrimAge_rate (continuous)	– 0.01	0.99	– 2.62	4.37	– 0.00 (1.0)	– 0.05 (0.95)
Men	0.38	0.98	– 2.62	4.37	0.41 (1.0)	0.31 (0.96)
Women	– 0.31	0.90	– 2.47	3.04	– 0.31 (1.0)	– 0.30 (0.9)
WBC	6.63	1.97	1.70	21.8	6.7 (1.9)	6.4 (1.9)
Men	6.59	1.90	2.30	19.4	6.7 (1.9)	6.3 (1.85)
Women	6.66	2.03	1.70	21.8	6.7 (2.0)	6.5 (2.1)
MMSE	27.41	2.67	9.00	30	27.46 (2.6)	27.06 (3.03)
Men	27.08	2.83	9.00	30	27.17 (2.7)	26.85 (3.2)
Women	27.66	2.51	12.00	30	27.75 (2.4)	27.37 (2.7)
Memory PC	– 0.01	1.00	– 3.30	2.12	0.05 (0.97)	– 0.20 (1.1)
Men	– 0.22	1.00	– 3.30	2.12	– 0.16 (0.95)	– 0.38 (1.1)
Women	0.15	0.97	– 3.29	2.12	0.22 (0.94)	– 0.07 (1.05)
Learning PC	0.01	0.98	– 3.15	2.73	0.04 (0.98)	– 0.14 (1.1)
Men	– 0.20	0.98	– 3.15	2.55	– 0.18 (0.96)	– 0.32 (1.07)
Women	0.17	0.95	– 3.15	2.73	0.19 (0.95)	– 0.01 (1.0)
CESD	1.75	2.25	0	11	1.75 (2.26)	1.73 (2.2)
Men	1.58	2.09	0	11	1.6 (2.1)	1.55 (2.0)
Women	1.87	2.35	0	10	1.88 (2.3)	1.92 (2.4)

WBC, white blood cell count; MMSE, Mini-Mental Status Exam; CESD, Center for Epidemiological Scale for Depression, ^aother race/ethnicity includes Pacific Islander, Native Hawaiian, and Asian.

Results from one-way ANOVAs comparing ɛ4 carriers to ɛ4 non-carriers on chronological age and the continuous measure of GrimAge showed no significant differences for either chronological age, p = 0.064 or GrimAge, p = 0.071. When stratified by sex, results remained unchanged; no significant differences by carrier status in chronological age or GrimAge. Associations between chronological age and GrimAge by ɛ4 carrier status and sex are illustrated by scatter plots in Supplementary Figures 1 and 2.

Comparisons of men and women on chronological age also did not reveal significant differences, p = 0.272; however there was a significant sex difference in GrimAge, F(1, 1790) = 111.378, p < 0.001, η_p² = 0.06, such that women had younger GrimAges (Mean = 70.68, SD = 6.65) compared to men (Mean = 74.14, SD = 7.13) and slower rates of GrimAge acceleration compared to men (standardized mean difference = 0.70, SE = 0.045, p < 0.001, η_p² = 0.12).

Results from ANCOVAS comparing performances of men and women on verbal learning and memory, adjusting for APOE ɛ4 carrier status, chronological age, education, and CESD scores, revealed that women outperformed men on verbal learning p < 0.001 (η_p² = 0.06) and memory p < 0.001 (η_p² = 0.073). There was a significant main effect of APOE ɛ4 status such that carriers had lower performances on learning, p < 0.001 η_p² = 0.131) and memory p < 0.001, η_p² = 0.025) compared to non-carriers. The results of these analyses are displayed in the Supplementary Table 3.

Sex-stratified APOE genotype ɛ4 carrier status is displayed in Table 2. There was not a significant association between sex and APOE ɛ4 carriers and non-carriers, p = 0.516. Similarly, there was no association between sex/gender and APOE genotype, p = 0.694. Rates of aging (slow, average, and fast) were not associated with APOE ɛ4 carrier status or APOE genotype in men (APOE ɛ4 carrier status: p = 0.544; APOE genotype: p = 0.470), or women (APOE ɛ4 carrier status: p = 0.475, APOE genotype: p = 0.673).

Table 2

Sex-stratified APOE genotype and ɛ4 carrier status

2/2		0.5%
Male	5	0.70%
Female	4	4.0%
2/3
Male	101	13.4%
Female	118	11.6%
2/4
Male	12	1.6%
Female	17	1.7%
3/3
Male	473	62.6%
Female	655	64.5%
3/4
Male	151	20.0%
Female	195	19.2%
4/4
Male	14	1.9%
Female	26	2.6%
APOE carrier status
ɛ4 non-carriers	1356	76.6%
Male	579	76.6%
Female	777	76.6%
ɛ4 carriers	415	23.4%
Male	177	23.4%
Female	238	23.4%

Sex-stratified analyses of APOE ɛ4 status on verbal learning and memory

Verbal learning. There was a significant main effect of APOE ɛ4 status on verbal learning among men, F(4,761) = 11.283, p < 0.001, η_p² = 0.015, and women, F(4,1026) = 15.471, p < 0.001, η_p² = 0.015, such that ɛ4 carriers had significantly lower memory performance compared to ɛ4 non-carriers: men (mean difference = 1.251, SE = 0.075, p < 0.001, 95% CI [–0.40, –0.10]); women (mean difference = –0.240, SE = 0.061, p < 0.001, 95% CI [–0.36, –0.12]).

When verbal memory was used as the outcome, a similar pattern of findings were revealed for men, F(4, 756) = 16.362, p < 0.001, η_p² = 0.02; mean difference = –0.307, SE = 0.076, p < 0.001, 95% CI [–0.46, –0.16] and women, F(4,1015) = 27.811, p < 0.001, η_p² = 0.027; mean difference = –0.33, SE = 0.062, p < 0.001, 95% CI [–0.45; –0.21].

Analyses of APOE ɛ4 X sex interaction

There was no significant APOE ɛ4 status X sex interaction effect on learning, p = 0.939, or memory, p = 0.798.

Sex-stratified analyses of APOE ɛ4 status and age rates on learning and memory

Learning. There was not a significant interaction between APOE ɛ4 status and rates of aging among women on learning, F (9, 1014) = 1.663, p = 0.190 so this term was excluded from the final model. The main effect of age-rate group was not significant, p = 0.076; however, the effect of APOE ɛ4 status remained significant in this model, F (6,1026) = 15.391, p < 0.001, η_p² = 0.015.

Memory

There was a significant interaction effect of APOE ɛ4 status X age-rate on verbal memory in women, F (9,1015) = 3.851, p = 0.022, η_p² = 0.01. Post-hoc comparisons revealed that the interaction between APOE ɛ4 status and group age-rate was driven by differences in ɛ4 carriers only (Fig. 1). Specifically, slow aging female ɛ4 carriers outperformed average aging female ɛ4 carriers, mean difference = 0.44, SE = 0.16, p = 0.017, 95% CI [0.06, 0.85], η_p² = 0.007 and fast aging female ɛ4 carriers, mean difference = 0.61 SE = 0.20, p = 0.006, 95% CI [0.14, 1.10], η_p² = 0.024. There was not a significant difference between average and fast aging ɛ4 carriers, p = 0.751. In female ɛ4 non-carriers, there was no significant difference in memory performances between any of the age-rate group comparisons. Similarly, the effect of APOE ɛ4 status on memory performances was only statistically significant in average and fast female aging groups, such that ɛ4 carriers had lower memory performances compared to ɛ4 non-carriers (average aging group: mean difference = –0.36, p < 0.001, 95% CI [0.21, 51], η_p² = 0.03, fast aging group: mean difference = –0.51, p < 0.001, 95% CI [0.22, 0.81], η_p² = 0.07. In contrast, there was no significant difference in memory performance between ɛ4 carriers and ɛ4 non-carriers in the slow aging female group, mean difference = –0.08, p = 0.632, 95% CI [–0.40, 0.25], η_p² = 0.031 = 0.002 (Fig. 2).

Fig. 1

Comparisons of memory performances in women according to ɛ4 status and age rate. *Indicates significant differences between groups at p < 0.05 (i.e., slow agers versus fast: p = 0.006, slow versus average: p = 0.017). Number of individuals in each group as follows: ɛ4 non-carriers: slow agers = 109, average agers = 537, fast agers = 131; ɛ4 carriers: slow n = 34, average n = 164, fast agers = 40.

Fig. 2

Comparisons of memory performances in women according to ɛ4 status and age rate. *Indicates significant differences at p < 0.001. Number of individuals in each group as follows: ɛ4 non-carriers: slow agers = 109, average agers = 537, fast agers = 131; ɛ4 carriers: slow n = 34, average n = 164, fast agers = 40.

Male group comparisons

There was a significant main effect of APOE ɛ4 status on learning F(7, 761) = 10.445, p < 0.001, η_p² = 0.014. There was not a significant main effect of age rate group, p = 0.255 or the interaction term, p = 0.921 on learning. Results were similar when the Memory PC was used as the outcome such that there was a significant main effect of APOE ɛ4 but a non-significant main effect of age rate, p = 0.053 and a non-significant interaction effect of APOE ɛ4 X age-rate, p = 0.672.

Sensitivity analyses

Results from rerunning the main analyses without the ɛ2/ɛ4 genotype did not alter results; all significant association remained, and all non-significant associations remained (Supplementary Tables 1 and 2). Results from including immune cell types as covariates in our main analyses did not change the results. Additionally, none of the immune cell types had a significant main effect on the outcome measures in any of the models (Supplementary Table 4). Finally, when we replaced the GrimAge rate group variable with the PhenoAge rate group variable and reran the analyses, there was not a significant interaction effect between APOE ɛ4 and PhenoAge on verbal learning or memory among men or women (Supplementary Table 5).

DISCUSSION

Findings from the current study showed that carriers of at least one ɛ4 allele had lower verbal learning and memory performances compared to ɛ4 non-carriers, consistent with other studies [13, 29]. Although some studies have found that effect of APOE ɛ4 on memory is greater in women compared to men [5, 12], we did not find a significant differences in effect size in our sex-stratified analysis or a significant sex X APOE ɛ4 on memory. However, we found that the negative effect of APOE ɛ4 on memory was strongest in women classified as ‘average’ and ‘fast’ agers compared to their ‘slow’ aging females ɛ4 carriers. Additionally, slow aging female APOE ɛ4 carriers had similar memory performances compared to slow aging female ɛ4 non-carriers. In males, rates of aging did not appear to influence the effect of APOE ɛ4 status on learning and memory, despite lower performances on these measures in ɛ4 carriers compared to ɛ4 non-carriers. Together these findings suggest that slow rates of aging, indexed by GrimAge may buffer against the negative effects of APOE ɛ4 on memory performances but only in women.

This is the first known published study to examine the effects of DNAm aging rates on the association between APOE ɛ4 status and verbal memory using a sex-stratified approach. While not directly comparable, one study examined associations between sex, APOE ɛ4 status, and their interaction on rates of aging, indexed by several epigenetic clocks, including GrimAge (continuous variable) and found no significant differences [67], consistent with our finding that accelerated aging and APOE ɛ4 carrier status may not be directly linked. However, the authors did not examine whether the associations between sex, age acceleration, and cognitive impairment varied by APOE ɛ4.

Findings from a recent systematic review suggested that there is currently insufficient evidence to support the role of epigenetic aging markers as predictors of cognitive decline or dementia [68]; however, the authors noted that only 5 out of the 30 studies identified in their review conducted a sex-stratified analysis. For studies that did, at least one measure of DNAm clock was associated with cognitive decline across studies in either just men [69 –71], women [72] or both [43], which further highlights the need for sex-stratified approaches in evaluating these biomarkers in cognitive aging research. Moreover, adjusting for sex may not be sufficient given evidence that women tend to have younger biological ages and slower rates of aging compared to men [73]. Furthermore, interpretation of sex differences when verbal memory is used as an outcome may be confounded by a female verbal memory advantage, such that a lack of a significant sex difference may not negate the possibility of a sex-specific effect. For example, in the current study female ɛ4 non-carriers outperformed male ɛ4 non-carriers (consistent with findings using the whole sample). In contrast, female ɛ4 carriers also outperformed male ɛ4 carriers and there was not a statistically significant difference between female ɛ4 carriers and male ɛ4 non-carriers. Together these findings suggest that the difference in verbal memory between ɛ4 female and ɛ4 male carriers in the present study may reflect a preexisting sex difference in verbal memory rather than a sex-specific effect of APOE ɛ4 as female ɛ4 non-carriers also outperformed male ɛ4 non-carriers (an example of false positive sex difference of APOE ɛ4). Similarly, a non-statistically significant difference between ɛ4 female carriers and male ɛ4 non-carriers in memory performance in the current study suggests that the presence of the ɛ4 allele attenuated the female verbal memory advantage to a level comparable to male ɛ4 non-carriers. These findings demonstrate the difficulty in identifying and differentiating sex differences in memory from differences driven by APOE ɛ4 without knowledge of ɛ4 carrier status and may lead to false negative sex differences in memory [74].

We found that ɛ4 female carriers with slow rates of aging had significantly better verbal memory performances when compared to ɛ4 female carriers with average and fast rates of aging is conceptually consistent with findings from other studies reporting stronger associations between APOE ɛ4 and cognition with chronologically older relative to younger individuals; an effect that varied according to sex and cognitive domain measured [75 –77]. However, whether these findings represent differences driven by indirect effects of APOE ɛ4 (via greater accumulation of prodromal AD pathology with increased age) versus a direct effect of APOE ɛ4 genotype on memory functioning that is exacerbated by increased age is less clear, given evidence to support both interpretations [13]. While we excluded individuals with an informant report of a prior dementia diagnosis as well as other neurological disorders, it is likely that the current sample includes a subset of individuals with prodromal AD or dementia. This is a limitation in so far as generalizing these effects to truly normal cognitive aging samples. Future studies with access to AD biomarkers and/or neuroimaging data could exclude individuals with evidence of AD to evaluate this further.

In support of a direct effect of APOE ɛ4 on memory, evidence of a protective effect of ɛ4 on cognition relative to non-carriers in younger individuals has been documented [78, 79], although other studies have not found this benefit [80]. For studies that have found a memory benefit of the ɛ4 allele in younger samples, these findings have been interpreted as a reflection of an antagonistic pleiotropy function of the ɛ4 allele, such that in older ages the positive effect of ɛ4 on cognition transmutes to a negative effect with advancing age possibly due to age comprised recruitment of cognitive compensatory networks observed only in ɛ4 carriers [81].

Our findings did not reveal a significant interaction between age-rate and ɛ4 carrier status on learning or memory in men. However, given the small group sizes, smaller effects may have been missed in the male sample, especially since overall memory performances were lower in men compared to women and therefore any effect of age-rate beyond education may have been much more limited. However, future research would benefit from examining this in a larger sample.

Other limitations concerning the sample size include not having adequate power to examine these associations according to APOE ɛ4 genotypes. Specifically, we were unable to examine whether carriers of two copies of the ɛ4 allele had worse memory performances compared to carriers of just one ɛ4 allele or whether having an ɛ2 allele in the ɛ4 carrier groups related to slower aging and/or better memory performances relative to ɛ3/ɛ4 or ɛ4/ɛ4 carriers. Additionally, larger ethnically diverse samples could be further stratified to examine whether these effects generalized across different race/ethnic groups, given the mixed findings on associations between APOE ɛ4 status and cognitive function/AD risk in relation to race/ethnicity. For example, evidence suggests that APOE ɛ4 association with risk of AD and memory is observed in Caucasian samples and does not appear to represent a substantial AD risk for African American or Asian groups [82, 83]. However, findings from other studies suggest that lower levels of baseline education, a factor strongly associated with cognitive function [84], may account for weaker associations between APOE ɛ4 and memory in older African Americans. From this perspective, the negative effect of APOE ɛ4 on cognition in African Americans and other racial/ethnic minorities may be obscured by an overall greater vulnerability to AD due to having less room to ‘fall’ cognitively relative to Caucasians who tend to have higher levels/better quality education [85, 86]. In another study, no significant racial/ethnic disparity in AD risk among APOE ɛ4 carriers was observed [10]. The groups included in this study were African American (n = 172), Japanese Americans (n = 81), Latino (n = 42), Native Hawaiian (n = 56), and White (n = 69). The authors reported that in ɛ4 non-carriers significant racial/ethnic disparities were observed with the highest risk in African Americans and Native Hawaiians and lowest in Asian Americans relative to Whites. Additionally, a 17% age-related increased risk for AD was noted in women relative to men, even after accounting for longevity. As noted by the authors, race/ethnicity and sex disparities in AD risk are likely mediated by competing causes of premature death with advancing age. As such, differences in sample rates of biological aging may account for some of these mixed findings on race/ethnic disparities and APOE ɛ4 in AD research. This suggestion is supported by prior findings which show that, on average, African Americans had faster rates of GrimAge compared to Whites or Hispanics [6]. Speculatively, the effect of APOE ɛ4 carrier status on memory may not be as pronounced in the setting of accelerated aging which may reduce overall resiliency and perhaps account for the higher prevalence of AD in African Americans relative to other ethnicities. However, future research is needed to investigate this possibility further.

The cross-sectional design is also a limitation with regard to understanding the significance of these associations on risk for future cognitive decline and dementia. However, future studies could examine longitudinal outcomes using the current framework for identifying subgroups at risk based on APOE ɛ4 carrier status and accelerated DNAmGrimAge rates. Additionally, whether these results translate to other domains of cognition (e.g., executive functioning, speed) were not explored in the current study given that our focus was on deepening understanding of female advantage in verbal learning/memory; however future research would benefit from exploring these associations in relation to non-memory domains.

Refining measurement approaches in the development of epigenetic clocks is an additional avenue for strengthening future research on this topic given inconsistencies between the associations using the GrimAge clock and PhenoAge clock. Although these clocks are both considered second generation clocks, the number of CpGs included in their construction differ (1030 CpGs in GrimAge and 513 CpGs in PhenoAge), which may drive differences in associations with cognition. The greater number of CpGs represented by GrimAge may capture greater shared variance between biological and cognitive aging. Other studies have also reported stronger associations between GrimAge and cognitive outcomes compared to other clocks including PhenoAge [43, 46]. GrimAge and PhenoAge also differ, however, with respect to CpGs included based on clinical markers in addition to age, decomposing these composite markers in relation to cognition may enhance understanding of specific CpGs linked to cognition. Finally, although we used an arbitrary cut off of 1 SD below and above the mean to identify subgroups of individuals with extreme aging rates, future research using other methods to define ‘abnormal’ rates of aging would be enhance understanding of aberrant rates of aging.

Conclusion

Female APOE ɛ4 carriers with slower than average DNAm aging rates may be buffered from the adverse effects of the APOE ɛ4 allele on memory. These findings are important from a mechanistic, prognostic and treatment perspective, especially since women are disproportionately affected by AD. Although longitudinal research is needed to better understand the significance of these differences, if female APOE ɛ4 carriers with slow rates of aging are at a decreased risk of AD and/or cognitive decline relative to the other ɛ4 female age groups, such findings would strengthen support for using epigenetic markers of aging as targets for intervention in female ɛ4 carriers. With emerging findings showing that age-related DNA methylation markers can be modified through lifestyle changes [87, 88] holds promise for future clinical studies aimed at modifying the single greatest risk factor for AD and a path forward in minimizing sex/gender disparities in this disease.

Footnotes

ACKNOWLEDGMENTS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

FUNDING

The Health and Retirement Study is sponsored by NIA (U01AG009740) and conducted by the University of Michigan. The HRS was approved by the University of Michigan Institutional Review Board.

This study was supported by grants to JEG from the National Institute on Aging (R01AG071514, R01AG071514S1, R01NS101483, and R01NS101483S1), the Harry T. Mangurian Foundation, and the Leo and Anne Albert Charitable Trust. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

CONFLICT OF INTEREST

JG is an Editorial Board Member of this journal but was not involved in the peer-review process nor had access to any information regarding its peer-review.

DATA AVAILABILITY

HRS data are available at:

The supplementary material is available in the electronic version of this article: .

References

(2022) 2022 Alzheimer’s disease facts and figures. Alzheimers Dement 18, 700–789.

De Roeck

, Engelborghs

, Dierckx

(2016) Next generation brain health depends on early Alzheimer disease diagnosis: From a timely diagnosis to future population screening. J Am Med Dir Assoc 17, 452–453.

Giang

, Nguyen

, Tran

(2020) Stratifying patients using fast multiple kernel learning framework: Case studies of Alzheimer’s disease and cancers. BMC Med Inform Decis Mak 20, 108.

Hampel

, Vergallo

, Giorgi

, Kim

, Depypere

, Graziani

, Saidi

, Nisticò

, Lista

(2018) Precision medicine and drug development in Alzheimer’s disease: The importance of sexual dimorphism and patient stratification. Front Neuroendocrinol 50, 31–51.

Zhao

, Ren

, Yamazaki

, Qiao

, Li

, Felton

, Mahmoudiandehkordi

, Kueider-Paisley

, Sonoustoun

, Arnold

, Shue

, Zheng

, Attrebi

, Martens

, Li

, Bastea

, Meneses

, Chen

, Thompson

, St John-Williams

, Tachibana

, Aikawa

, Oue

, Job

, Yamazaki

, Liu

, Storz

, Asmann

, Ertekin-Taner

, Kanekiyo

, Kaddurah-Daouk

, Bu

(2020) Alzheimer’s risk factors age, APOE genotype, and sex drive distinct molecular pathways. Neuron 106, 727–742.e726.

Crimmins

, Thyagarajan

, Levine

, Weir

, Faul

(2021) Associations of age, sex, race/ethnicity, and education with 13 epigenetic clocks in a nationally representative U.S. sample: The Health and Retirement Study. J Gerontol A Biol Sci Med Sci 76, 1117–1123.

Kawas

, Gray

, Brookmeyer

, Fozard

, Zonderman

(2000) Age-specific incidence rates of Alzheimer’s disease: The Baltimore Longitudinal Study of Aging. Neurology 54, 2072–2077.

Tom

, Hubbard

, Crane

, Haneuse

, Bowen

, McCormick

, McCurry

, Larson

(2015) Characterization of dementia and Alzheimer’s disease in an older population: Updated incidence and life expectancy with and without dementia. Am J Public Health 105, 408–413.

Letenneur

, Gilleron

, Commenges

, Helmer

, Orgogozo

, Dartigues

(1999) Are sex and educational level independent predictors of dementia and Alzheimer’s disease? Incidence data from the PAQUID project. J Neurol Neurosurg Psychiatry 66, 177–183.

10.

Lim

, Wang

, Park

, Bogumil

, Wu

, Cheng

, Haiman

, Le Marchand

, Wilkens

, White

, Setiawan

(2022) Risk of Alzheimer’s disease and related dementia by sex and race/ethnicity: The Multiethnic Cohort Study. Alzheimers Dement 18, 1625–1634.

11.

Mahley

, Rall

Jr. (2000) Apolipoprotein E: Far more than a lipid transport protein. Annu Rev Genomics Hum Genet 1, 507–537.

12.

Altmann

, Tian

, Henderson

, Greicius

, Alzheimer’s Disease Neuroimaging Initiative Investigators (2014) Sex modifies the APOE-related risk of developing Alzheimer disease. Ann Neurol 75, 563–573.

13.

O’Donoghue

, Murphy

, Zamboni

, Nobre

, Mackay

(2018) APOE genotype and cognition in healthy individuals at risk of Alzheimer’s disease: A review. Cortex 104, 103–123.

14.

Eisenberg

, Kuzawa

, Hayes

(2010) Worldwide allele frequencies of the human apolipoprotein E gene: Climate, local adaptations, and evolutionary history. Am J Phys Anthropol 143, 100–111.

15.

Yamazaki

, Zhao

, Caulfield

, Liu

, Bu

(2019) Apolipoprotein E and Alzheimer disease: Pathobiology and targeting strategies. Nat Rev Neurol 15, 501–518.

16.

Nagy

, Esiri

, Jobst

, Johnston

, Litchfield

, Sim

, Smith

(1995) Influence of the apolipoprotein E genotype on amyloid deposition and neurofibrillary tangle formation in Alzheimer’s disease. Neuroscience 69, 757–761.

17.

Salvadó

, Grothe

, Groot

, Moscoso

, Schöll

, Gispert

, Ossenkoppele

(2021) Differential associations of APOE-ɛ2 and APOE-ɛ4 alleles with PET-measured amyloid-β and tau deposition in older individuals without dementia. Eur J Nucl Med Mol Imaging 48, 2212–2224.

18.

Fernández-Calle

, Konings

, Frontiñán-Rubio

, García-Revilla

, Camprubí-Ferrer

, Svensson

, Martinson

, Boza-Serrano

, Venero

, Nielsen

, Gouras

, Deierborg

(2022) APOE in the bullseye of neurodegenerative diseases: Impact ofthe APOE genotype in Alzheimer’s disease pathology and braindiseases. Mol Neurodegener 17, 62.

19.

Salthouse

(2019) Trajectories of normal cognitive aging. Psychol Aging 34, 17–24.

20.

Daseking

, Petermann

, Waldmann

H-C

(2017) Sex differences in cognitive abilities: Analyses for the German WAIS-IV. Pers Individ Dif 114, 145–150.

21.

Haller

, Montandon

, Rodriguez

, Garibotto

, Herrmann

, Giannakopoulos

(2019) Hippocampal volume loss, brain amyloid accumulation, and APOE status in cognitively intact elderly subjects. Neurodegener Dis 19, 139–147.

22.

Caselli

, Dueck

, Osborne

, Sabbagh

, Connor

, Ahern

, Baxter

, Rapcsak

, Shi

, Woodruff

, Locke

, Snyder

, Alexander

, Rademakers

, Reiman

(2009) Longitudinal modeling of age-related memory decline and the APOE epsilon4 effect. N Engl J Med 361, 255–263.

23.

Barel

, Tzischinsky

(2018) Age and sex differences in verbal and visuospatial abilities. Adv Cogn Psychol 2, 51–61.

24.

Munro

, Winicki

, Schretlen

, Gower

, Turano

, Muñoz

, Keay

, Bandeen-Roche

, West

(2012) Sex differences in cognition in healthy elderly individuals. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 19, 759–768.

25.

Lindenberger

(2001) Lifespan theories of cognitive development. In International Encyclopedia of the Social and Behavioral Sciences, Smelser NJ, Baltes PB, eds. Elsevier Science, Oxford, pp. 8848-8854.

26.

Grossi

, Matarese

, Orsini

(1980) Sex differences in adults’ spatial and verbal memory span. Cortex 16, 339–340.

27.

Brunet

, Caldwell

JZK

, Brandt

, Miller

(2020) Influence of sex differences in interpreting learning and memory within a clinical sample of older adults. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 27, 18–39.

28.

Wisdom

, Callahan

, Hawkins

(2011) The effects of apolipoprotein E on non-impaired cognitive functioning: A meta-analysis. Neurobiol Aging 32, 63–74.

29.

Wang

, Zhou

, Ye

, Lin

, Zhang

(2019) Sex difference in the association of APOE4 with memory decline in mild cognitive impairment. J Alzheimers Dis 69, 1161–1169.

30.

Horvath

, Raj

(2018) DNA methylation-based biomarkers and the epigenetic clock theory of ageing. Nat Rev Genet 19, 371–384.

31.

Jylhävä

, Pedersen

, Hägg

(2017) Biological age predictors. EBioMedicine 21, 29–36.

32.

Brown

, Wall

, Chen

, Levine

, Yaffe

, Roose

, Rutherford

(2018) Biological age, not chronological age, is associated with late-life depression. J Gerontol A Biol Sci Med Sci 73, 1370–1376.

33.

Ryan

, Wrigglesworth

, Loong

, Fransquet

, Woods

(2020) A systematic review and meta-analysis of environmental, lifestyle, and health factors associated with DNA methylation age. J Gerontol A Biol Sci Med Sci 75, 481–494.

34.

Bae

, Piao

, Kim

, Im

, Kim

, Choi

, Cho

(2021) Biological age and lifestyle in the diagnosis of metabolic syndrome: The NHIS health screening data, 2014-2015. Sci Rep 11, 444.

35.

Hartmann

, Hartmann

, Secci

, Hermann

, Fuellen

, Walter

(2021) Ranking biomarkers of aging by citation profiling and effort scoring. Front Genet 12, 686320.

36.

Justice

, Ferrucci

, Newman

, Aroda

, Bahnson

, Divers

, Espeland

, Marcovina

, Pollak

, Kritchevsky

, Barzilai

, Kuchel

(2018) A framework for selection of blood-based biomarkers for geroscience-guided clinical trials: Report from the TAME Biomarkers Workgroup. Geroscience 40, 419–436.

37.

Kudryashova

, Burka

, Kulaga

, Vorobyeva

, Kennedy

(2020) Aging biomarkers: From functional tests to multi-omics approaches. Proteomics 20, e1900408.

38.

Gardiner-Garden

, Frommer

(1987) CpG islands in vertebrate genomes. J Mol Biol 196, 261–282.

39.

Quach

, Levine

, Tanaka

, Lu

, Chen

, Ferrucci

, Ritz

, Bandinelli

, Neuhouser

, Beasley

(2017) Epigenetic clock analysis of diet, exercise, education, and lifestyle factors. Aging (Albany NY) 9, 419.

40.

Horvath

(2013) DNA methylation age of human tissues and cell types. Genome Biol 14, R115.

41.

Hannum

, Guinney

, Zhao

, Zhang

, Hughes

, Sadda

, Klotzle

, Bibikova

, Fan

, Gao

, Deconde

, Chen

, Rajapakse

, Friend

, Ideker

, Zhang

(2013) Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 49, 359–367.

42.

Levine

, Lu

, Bennett

, Horvath

(2015) Epigenetic age of the pre-frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer’s disease related cognitive functioning. Aging (Albany NY) 7, 1198.

43.

Hillary

, Stevenson

, Cox

, McCartney

, Harris

, Seeboth

, Higham

, Sproul

, Taylor

, Redmond

, Corley

, Pattie

, Hernández

, Muñoz-Maniega

, Bastin

, Wardlaw

, Horvath

, Ritchie

, Spires-Jones

, McIntosh

, Evans

, Deary

, Marioni

(2021) An epigenetic predictor of death capturesmulti-modal measures of brain health. Mol Psychiatry 26, 3806–3816.

44.

O’Shea

, Maynard

, Tremont

(2022) DNA methylation “GrimAge” acceleration mediates sex/gender differences in verbal memory and processing speed: Findings from the Health and Retirement Study. J Gerontol A Biol Sci Med Sci 77, 2402–2412.

45.

, Quach

, Wilson

, Reiner

, Aviv

, Raj

, Hou

, Baccarelli

, Li

, Stewart

(2019) DNA methylation Grim Age strongly predicts lifespan and healthspan. Aging (Albany NY) 11, 303.

46.

McCrory

, Fiorito

, Hernandez

, Polidoro

, O’Halloran

, Hever

, Ni Cheallaigh

, Lu

, Horvath

, Vineis

, Kenny

(2021) GrimAge outperforms other epigenetic clocks in the prediction of age-related clinical phenotypes and all-cause mortality. J Gerontol A Biol Sci Med Sci 76, 741–749.

47.

Sundermann

, Biegon

, Rubin

, Lipton

, Landau

, Maki

(2017) Does the female advantage in verbal memory contribute to underestimating Alzheimer’s disease pathology in women versus men? J Alzheimers Dis 56, 947–957.

48.

Subramaniapillai

, Almey

, Natasha Rajah

, Einstein

(2021) Sex and gender differences in cognitive and brain reserve: Implications for Alzheimer’s disease in women. Front Neuroendocrinol 60, 100879.

49.

Langa

, Ryan

, McCammon

, Jones

, Manly

, Levine

, Sonnega

, Farron

, Weir

(2020) The Health and Retirement Study Harmonized Cognitive Assessment Protocol Project: Study design and methods. Neuroepidemiology 54, 64–74.

50.

Fisher

, Ryan

(2018) Overview of the Health and Retirement Study and introduction to the special issue. Work Aging Retire 4, 1–9.

51.

Ignjatovic

, Lai

, Summerhayes

, Mathesius

, Tawfilis

, Perugini

, Monagle

(2011) Age-related differences in plasma proteins: How plasma proteins change from neonates to adults. PloS One 6, e17213.

52.

Ridker

, Buring

, Cook

, Rifai

(2003) C-reactive protein, the metabolic syndrome, and risk of incident cardiovascular events: An 8-year follow-up of 14 719 initially healthy American women. Circulation 107, 391–397.

53.

Crimmins

, Kim

, Fisher

, Faul

(2020) HRS Epigenetic Clocks–Release 1. Survey Research Center, Univeristy of Michigan.

54.

Heaton

(2004) Revised comprehensive norms for an expanded Halstead-Reitan Battery: Demographically adjusted neuropsychological norms for African American and Caucasian adults, professional manual, Psychological Assessment Resources.

55.

Casaletto

, Elahi

, Staffaroni

, Walters

, Contreras

, Wolf

, Dubal

, Miller

, Yaffe

, Kramer

(2019) Cognitive aging is not created equally: Differentiating unique cognitive phenotypes in “normal” adults. Neurobiol Aging 77, 13–19.

56.

Faul

, Collins

, Smith

, Zhao

, Weir

(2021) APOE and serotonin transporter alleles–early release. Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI.

57.

Salvadó

, Ferreira

, Operto

, Cumplido-Mayoral

, Arenaza-Urquijo

, Cacciaglia

, Falcon

, Vilor-Tejedor

, Minguillon

, Groot

, van der Flier

, Barkhof

, Scheltens

, Ossenkoppele

, Kern

, Zettergren

, Skoog

, Hort

, Stomrud

, van Westen

, Hansson

, Molinuevo

, Wahlund

, Westman

, Gispert

(2022) The protective gene dose effect of the APOEɛ2 allele on gray matter volume in cognitively unimpaired individuals. Alzheimers Dement 18, 1383–1395.

58.

Evans

, Dowell

, Prowse

, Tabet

, King

, Rusted

(2020) Mid ageɛ4 carriers show memory-related functional differences and disrupted structure-function relationships in hippocampal regions. Sci Rep 10, 3110.

59.

MacAulay

, Halpin

, Cohen

, Calamia

, Boeve

, Zhang

, Brouillette

, Foil

, Bruce-Keller

, Keller

(2020) Predictors of heterogeneity in cognitive function: APOE-e4, sex, education, depression, and vascular risk. Arch Clin Neuropsychol 35, 660–670.

60.

Memel

, Staffaroni

, Cobigo

, Casaletto

, Fonseca

, Bettcher

, Yassa

, Elahi

, Wolf

, Rosen

(2021) APOE moderates the effect of hippocampal blood flow on memory pattern separation in clinically normal older adults. Hippocampus 31, 845–857.

61.

Morris

, Heyman

, Mohs

, Hughes

, van Belle

, Fillenbaum

, Mellits

, Clark

(1989) The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 39, 1159–1165.

62.

Wechsler

(1987) Wechsler Memory Scale-Revised manual, Psychological Corporation, San Antonio.

63.

Scherr

, Albert

, Funkenstein

, Cook

, Hennekens

, Branch

, White

, Taylor

, Evans

(1988) Correlates of cognitive function in an elderly community population. Am J Epidemiol 128, 1084–1101.

64.

Horvath

, Levine

(2015) HIV-1 infection accelerates age according to the epigenetic clock. J Infect Dis 212, 1563–1573.

65.

Kananen

, Nevalainen

, Jylhava

, Marttila

, Hervonen

, Jylha

, Hurme

(2015) Cytomegalovirus infection accelerates epigenetic aging. Exp Gerontol 72, 227–229.

66.

O’Shea

, Fieo

, Hamilton

, Zahodne

, Manly

, Stern

(2015) Examining the association between late-life depressive symptoms, cognitive function, and brain volumes in the context of cognitive reserve. Int J Geriatr Psychiatry 30, 614–622.

67.

Inkster

, Duarte-Guterman

, Albert

, Barha

, Galea

LAM

, Robinson

, Alzheimer’s Disease Neuroimaging Initiative (2022) Are sex differences in cognitive impairment reflected in epigenetic age acceleration metrics? Neurobiol Aging 109, 192–194.

68.

Zhou

, Wu

, Phyo

AZZ

, Torres

, Vishwanath

, Ryan

(2022) Epigenetic aging as a biomarker of dementia and related outcomes: A systematic review. Epigenomics 14, 1125–1138.

69.

Paul

, Binder

, Horvath

, Kusters

, Yan

, Rosario

, Yu

, Bronstein

, Ritz

(2021) Accelerated hematopoietic mitotic aging measured by DNA methylation, blood cell lineage, and Parkinson’s disease. BMC Genomics 22, 696.

70.

Fransquet

, Lacaze

, Saffery

, Shah

, Vryer

, Murray

, Woods

, Ryan

(2021) Accelerated epigenetic aging in peripheral blood does not predict dementia risk. Curr Alzheimer Res 18, 443–451.

71.

Beydoun

, Shaked

, Tajuddin

, Weiss

, Evans

, Zonderman

(2020) Accelerated epigenetic age and cognitive decline among urban-dwelling adults. Neurology 94, e613–e625.

72.

Zheng

, Habes

, Gonzales

, Pomponio

, Nasrallah

, Khan

, Vaughan

, Davatzikos

, Seshadri

, Launer

, Sorond

, Sedaghat

, Wainwright

, Baccarelli

, Sidney

, Bryan

, Greenland

, Lloyd-Jones

, Yaffe

, Hou

(2022) Mid-life epigenetic age, neuroimaging brain age, and cognitive function: Coronary artery risk development in young adults (CARDIA) study. Aging (Albany NY) 14, 1691–1712.

73.

Kankaanpää

, Tolvanen

, Saikkonen

, Heikkinen

, Laakkonen

, Kaprio

, Ollikainen

, Sillanpää

(2022) Do epigenetic clocks provide explanations for sex differences in life span? A cross-sectional twin study. J Gerontol A Biol Sci Med Sci 77, 1898–1906.

74.

Hirnstein

, Stuebs

, Moe

, Hausmann

(2023) Sex/gender differences in verbal fluency and verbal-episodic memory: A meta-analysis. Perspect Psychol Sci 18, 67–90.

75.

Hyman

, Gomez-Isla

, Briggs

, Chung

, Nichols

, Kohout

, Wallace

(1996) Apolipoprotein E and cognitive change in an elderly population. Ann Neurol 40, 55–66.

76.

Swan

, Lessov-Schlaggar

, Carmelli

, Schellenberg

, La Rue

(2005) Apolipoprotein E epsilon4 and change in cognitive functioning in community-dwelling older adults. J Geriatr Psychiatry Neurol 18, 196–201.

77.

Cacciaglia

, Molinuevo

, Falcón

, Sánchez-Benavides

, Gramunt

, Brugulat-Serrat

, Esteller

, Morán

, Fauria

, Gispert

(2019) APOE-ɛ4 risk variant for Alzheimer’sdisease modifies the association between cognitive performance andcerebral morphology in healthy middle-aged individuals. Neuroimage Clin 23, 101818.

78.

, Lin

, Chen

, Hong

, Tsai

(2000) Intelligence and event-related potentials for young female human volunteer apolipoprotein E epsilon4 and non-epsilon4 carriers. Neurosci Lett 294, 179–181.

79.

Zink

, Bensmann

, Arning

, Beste

, Stock

(2019) Apolipoprotein ɛ4 is associated with better cognitive control allocation in healthy young adults. Neuroimage 185, 274–285.

80.

Ihle

, Bunce

, Kliegel

(2012) APOE epsilon4 and cognitive function in early life: A meta-analysis. Neuropsychology 26, 267–277.

81.

Tuminello

, Han

(2011) The apolipoprotein e antagonistic pleiotropy hypothesis: Review and recommendations. Int J Alzheimers Dis 2011, 726197.

82.

Beydoun

, Weiss

, Beydoun

, Hossain

, Maldonado

, Shen

, Evans

, Zonderman

(2021) Race, APOE genotypes, and cognitive decline among middle-aged urban adults. Alzheimers Res Ther 13, 120.

83.

Evans

, Bennett

, Wilson

, Bienias

, Morris

, Scherr

, Hebert

, Aggarwal

, Beckett

, Joglekar

, Berry-Kravis

, Schneider

(2003) Incidence of Alzheimer disease in a biracial urban community: Relation to apolipoprotein E allele status. Arch Neurol 60, 185–189.

84.

Lövdén

, Fratiglioni

, Glymour

, Lindenberger

, Tucker-Drob

(2020) Education and cognitive functioning across thelife span. Psychol Sci Public Interest 21, 6–41.

85.

Rajan

, McAninch

, Wilson

, Weuve

, Barnes

, Evans

(2019) Race, APOEɛ4, and long-term cognitive trajectories in a biracial population sample. J Alzheimers Dis 72, 45–53.

86.

Sisco

, Gross

, Shih

, Sachs

, Glymour

, Bangen

, Benitez

, Skinner

, Schneider

, Manly

(2015) The role of early-life educational quality and literacy in explaining racial disparities in cognition in late life. J Gerontol B Psychol Sci Soc Sci 70, 557–567.

87.

Fitzgerald

, Hodges

, Hanes

, Stack

, Cheishvili

, Szyf

, Henkel

, Twedt

, Giannopoulou

, Herdell

(2021) Potential reversal of epigenetic age using a diet and lifestyle intervention: A pilot randomized clinical trial. Aging (Albany NY) 13, 9419.

88.

Fiorito

, Caini

, Palli

, Bendinelli

, Saieva

, Ermini

, Valentini

, Assedi

, Rizzolo

, Ambrogetti

, Ottini

, Masala

(2021) DNA methylation-based biomarkers of aging were slowed down in a two-year diet and physical activity intervention trial: The DAMA study. Aging Cell 20, e13439.