BDNF expression mediates verbal learning and memory in women in a cohort enriched with risk for Alzheimer's disease

Abstract

INTRODUCTION

This study examined whether sex differences in verbal learning and memory (VLM) are mediated by plasma brain‐derived neurotrophic factor (BDNF) expression.

METHODS

In a sample of n = 201 participants (63.81 ± 6.04 years, 66.2% female, 65.7% family history of Alzheimer's disease [AD], 38% apolipoprotein E [APOE] ε4+) from the Wisconsin Registry for Alzheimer's Prevention, VLM was measured using trials 3 through 5 and delayed recall from the Rey Auditory Verbal Learning Test. Plasma BDNF was measured using a Human BDNF Quantikine Immunoassay. Mediation analysis used bootstrapping, and stratified mediation models tested the conditional dependence of APOE ε4 carriage.

RESULTS

BDNF partially mediated the sex–VLM relationship (β = −0.07; 95% confidence interval [CI]: −0.18, −0.01). Female APOE ε4 carriers had higher VLM scores (β = −0.53; p = 0.03), while female non‐carriers had both higher BDNF levels (β = −0.68; p < 0.01) and VLM scores (β = −1.06; p < 0.01); BDNF was again a significant mediator (β = −0.18; 95% CI: −0.37, −0.05).

DISCUSSION

This study found that circulating BDNF mediates higher verbal memory scores in females—particularly in APOE ε4 non‐carriers.

Highlights

• Sex differences in verbal learning and memory (VLM) were mediated by plasma brain‐derived neurotrophic factor (BDNF) levels.

• Women exhibited higher VLM scores and plasma BDNF levels compared to men.

• The protective effect of BDNF in women was attenuated by apolipoprotein E ε4 carriage.

• Findings suggest sex‐specific mechanisms against verbal memory decline in aging.

1. BACKGROUND

In Alzheimer's disease (AD) research, the “cognitive reserve” theory proposes that persons with favorable premorbid neurocognitive factors (e.g., higher education, IQ) maintain normal cognitive function in aging, despite the accumulation of AD‐related brain pathology. ¹ For those with higher cognitive reserve, the transition to amnestic mild cognitive impairment (aMCI) is delayed; however, these individuals then exhibit accelerated cognitive decline ² –likely due to advanced accumulated neuropathology. ³ Cognitive reserve can therefore be understood as a dynamic construct that emphasizes the summative effects of innate and acquired factors on brain structure and cognitive ability.

At the neuronal level, cognitive reserve is thought to protect neuroplasticity and synaptic function, therefore delaying the onset of MCI; however, persons with higher reserve typically exhibit accelerated cognitive decline upon MCI diagnosis, due in part to more advanced AD neuropathology. ⁴ Other factors thought to contribute to cognitive reserve include, but are not limited to, genetics, education level, occupational complexity, multilingualism, socioeconomic status, psychosocial support, and physical activity. ⁵ While the theory presents a key framework for exploring why some individuals retain normal cognitive function despite evidence of neuropathology, the underlying mechanisms and factors associated with cognitive reserve remain poorly characterized.

Evidence from recent meta‐analyses indicates that women slightly outperform men in verbal learning and memory (VLM) ability, ⁶ a neurocognitive domain sensitive to decline in AD. ⁷ Consistent with cognitive reserve theory, the female advantage in VLM appears to persist during the preclinical stages of AD ⁸ but quickly wanes upon transition to AD dementia. ⁹ This advantage may therefore reflect a sex‐specific form of cognitive reserve, which risks masking the cognitive effects of neuropathological changes in women—thereby delaying aMCI diagnosis during the early stages of AD.

In this regard, the role of brain‐derived neurotrophic factor (BDNF) may be of particular importance. ¹⁰ , ¹¹ At the neuronal level, BDNF is synthesized and secreted from several components of the synapse, including astrocytes, microglia, and post‐synaptic dendrites. ¹² This localized synaptic BDNF production is necessary for long‐term potentiation, which has a direct influence on memory consolidation; however, circulating BDNF measured from blood plasma has likewise been associated with improved cognitive function in aging. ¹¹ , ¹³

Within the brain, the highest levels of BDNF are found in the medial temporal lobe and hippocampus—regions of the cortex critical for receiving, consolidating, and retrieving memories. ¹⁴ These regions are likewise sensitive to decline in AD dementia, and there is emerging evidence that BDNF production and function may be influenced by apolipoprotein E (APOE) ε4 allele carriage ¹⁵ —a canonical risk factor for AD. ¹⁶ However, it remains unclear whether BDNF expression is associated with VLM differences between male and female APOE ε4 carriers. This study examines the mediating role of circulating BDNF on the relationship between sex and VLM in a sample enriched with risk for AD. Additionally, we explore whether this mediating relationship is conditionally dependent upon APOE ε4 carriage. Figure 1 shows a graphical summary of this study framework.

FIGURE 1.

Graphical summary of the stratified mediation framework for this study. a, b, c and c' are path coefficients representing standardized regression effects. Models were subsequently stratified by APOE ε4 carriage (0: no allele, 1: at least one allele). The c path coefficient represents the total effect of sex on VLM. APOE, apolipoprotein E; BDNF, brain‐derived neurotrophic factor; VLM, verbal learning and memory.

2. METHODS

2.1. Participants

Study participants were enrolled in the Wisconsin Registry for Alzheimer's Prevention (WRAP)—a longitudinal study of ≈ 1500 late‐middle‐aged adults who were cognitively healthy and between 40 and 65 years of age at study entry. ¹⁷ , ¹⁸ For this work, a sample of 201 participants from a WRAP‐linked ancillary study were selected on the basis of having complete data on all variables of interest. Sample enrichment with AD risk was evidenced by 66% percent of participants having at least one parent with AD and 38% were APOE ε4 positive (for participant demographic data, see Table 1) (Tables 2, 3, 4).

TABLE 1.

Participant data stratified by sex and APOE ε4 carriage.

Variable:	Females	Males	APOE ε4+	APOE ε4–
n (%)	133 (66)	68 (34)	77 (38)	124 (62)
Age (years), mean (SD)	63.3 (6.0)	64.8 (5.8)	63.0 (5.7)	64.3 (6.2)
Female, n (%)	—/—	—/—	51 (66%)	82 (66%)
Race, n (%) Black or African American American Indian or Alaska Native Asian White Other	3 (2%) 2 (2%) —/— 128 (96%) —/—	—/— 1 (1%) 1 (1%) 66 (97.1%) —/—	1 (1%) 2 (3%) —/— 74 (96%) —/—	2 (2%) 1 (1%) 1 (1%) 119 (96%) 1 (1%)
Education (years), mean (SD)	16.0 (2.0)*	17.4 (3.0)*	16.3 (2.1)	16.5 (2.9)
Hippocampal volume, mean (SD)	5.39 (0.48)*	5.06 (0.51)*	5.36 (0.52)	5.23 (0.51)
Family history of AD, n (%)	91 (68%)	41 (61%)	68 (88%)*	64 (52%)*
APOE ε4 carriage, n (%)	51 (38%)	26 (38%)	—/—	—/—
BDNFVal66met carriage, n (%)	41 (31%)	19 (28%)	26 (34%)*	34 (28%)*
BDNF (pg/mL), mean (SD)	6722.5 (6081.1)	5257.8 (4727.6)	5730.5 (4147.7)	6535.3 (6462.5)
VLM (z‐scores), mean (SD)	0.35 (0.82)*	−0.37 (1.13)*	0.06 (0.11)	0.14 (0.12)

Abbreviations: AD, Alzheimer's disease; APOE, apolipoprotein E; BDNF, brain‐derived neurotrophic factor; SD, standard deviation; VLM, verbal learning and memory.

*: $p<0.05.$

TABLE 2.

Model findings from analyses testing the mediating role of plasma BDNF expression on the relationship between sex and VLM scores.

95% CI
Type	Effect	β	Lower	Upper	p value
Indirect	Sex ⇒ BDNF ⇒ VLM	−0.07	−0.18	−0.01	0.01
Component	Sex ⇒ BDNF	−0.41	−0.89	−0.20	0.01
	BDNF ⇒ VLM	0.18	0.02	0.35	0.01
Total	Sex ⇒ VLM	−0.67	−0.97	−0.38	<0.01
Direct	Sex ⇒ VLM	−0.61	−0.90	−0.39	<0.01

Note: Covariates: All models were adjusted for age, time between blood sample collection and cognitive testing, education level, apolipoprotein E ε4+, BDNF_Val66met, family AD history, and hippocampal volume.

Abbreviations: AD, Alzheimer's disease; BDNF, brain‐derived neurotropic factor; CI, confidence interval; VLM, verbal learning and memory.

TABLE 3.

Model findings from analyses testing the mediating role of plasma BDNF expression in APOE ε4+ individuals (n = 77).

95% CI
Type	Effect	Β	Lower	Upper	p value
Indirect	Sex ⇒ BDNF ⇒ VLM	0.01	−0.05	0.08	0.57
Component	Sex ⇒ BDNF	−0.14	−2.81	6.38	0.60
	BDNF ⇒ VLM	−0.07	−0.99	0.81	0.53
Total	Sex ⇒ VLM	−0.53	−0.99	−0.07	0.03
Direct	Sex ⇒ VLM	−0.54	−1.00	−0.07	0.02

Note: Covariates: All models were adjusted for age, time between blood sample collection and cognitive testing, education level, BDNF_Val66met, family AD history, and hippocampal volume.

Abbreviations: AD, Alzheimer's disease; APOE, apolipoprotein E; BDNF, brain derived‐neurotrophic factor; CI, confidence interval; VLM, verbal learning and memory.

TABLE 4.

Model findings from analyses testing the mediating role of plasma BDNF expression in APOE ε4– individuals (n = 124).

95% CI
Type	Effect	Β	Lower	Upper	p value
Indirect	Sex ⇒ BDNF ⇒ VLM	−0.18	−0.37	−0.05	<0.01
Component	Sex ⇒ BDNF	−0.68	−1.06	−0.33	<0.01
	BDNF ⇒ VLM	0.25	0.10	1.01	<0.01
Total	Sex ⇒ VLM	−1.06	−1.46	−0.66	<0.01
Direct	Sex ⇒ VLM	−0.88	−1.28	−0.48	<0.01

Note: Covariates: All models were adjusted for age, time between blood sample collection and cognitive testing, education level, BDNF_Val66met, family AD history, and hippocampal volume.

Abbreviations: AD, Alzheimer's disease; APOE, apolipoprotein E; BDNF, brain‐derived neurotrophic factor; CI, confidence interval; VLM, verbal learning and memory.

2.2. Plasma BDNF collection and analyses

After a 12 hour fast from food, caffeine, and alcohol, blood samples were collected for plasma analyses into a 10 mL ethylenediaminetetraacetic acid (EDTA) vacutainer, which was then gently inverted 10 to 12 times. Within 1 hour of collection, these EDTA tubes were centrifuged at 3000 rpm for 15 minutes at room temperature. Cell‐free plasma was then aliquoted into cryovials and stored at –80°C until assayed for BDNF using total BDNF enzyme‐linked immunosorbent assay kits (R&D Systems) in accordance with manufacturer specifications. Frozen plasma samples were thawed at room temperature (25°C) and held on ice before being loaded in duplicate on a microplate and incubated, in accordance with the manufacturer protocols. Each assay included a recombinant protein as a positive control, and plates were read using SpectraMax Plus 384 Microplate Reader (Molecular Devices).

2.3. Cognitive evaluation

Participants in WRAP undergo a comprehensive battery of neuropsychological tests spanning traditional cognitive domains such as memory, attention, executive function, language, and visuospatial ability. ¹⁹ Prior factor analytic studies of these measures within the larger WRAP cohort indicated that they map on to six cognitive factors. ¹⁷ , ¹⁸ For this study, we sought to specifically interrogate the domain of VLM; the VLM factor score in WRAP is a reported index composed of the Rey Auditory Verbal Learning Test (RAVLT) learning trials 3 through 5, and delayed recall. ²⁰ Due to non‐normal distribution of the VLM index, we applied a z score transformation to standardize values for all analyses.

2.4. Genotyping

DNA was extracted from whole‐blood samples using the PUREGENE DNA Isolation Kit (Gentra Systems, Inc.). DNA concentrations were quantified using ultraviolet spectrophotometry (DU 530 Spectrophotometer; Beckman Coulter). Single nucleotide polymorphisms (SNPs) for APOE (rs429358 and rs7412) and BDNF_Val66met (rs6265) were genotyped by LGC Genomics (Beverly, MA) using competitive allele‐specific polymerase chain reaction (PCR)–based Kompetitive allele specific PCR (KASP) genotyping assays. Quality control (e.g., duplicate sample concordance rates and Hardy–Weinberg equilibrium) has been previously described ²¹ and was found to be satisfactory.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature using traditional electronic databases (e.g., PubMed). While brain‐derived neurotrophic factor (BDNF) is an important neurotrophin for protecting synaptic health in aging, literature is mixed regarding whether BDNF expression is associated with differences in verbal learning and memory (VLM) ability between male and female apolipoprotein E (APOE) ε4 carriers. Evidence from relevant systematic review literature was prioritized for this work and is appropriately cited herein.

2. Interpretation: Our findings indicate that higher circulating BDNF levels in women may be protective against verbal memory decline—but this relationship is attenuated by the deleterious effects of APOE ε4 allele carriage.

3. Future directions: This study motivates several potential avenues for continued investigation; examples include, (1) investigating the role of menopause and its influence on the sex–BDNF–VLM association; (2) examining the influence of physical activity on BDNF secretion and cognitive function; (3) examining commensurate effects on Alzheimer's disease biomarkers; and (4) exploring the influence of the BDNF_Val66met polymorphism.

2.5. Statistical analyses

Welch two‐sample t tests were performed to evaluate whether participant data differed by sex or APOE ε4 carriage. Next, linear regression models were assembled in a mediation analysis framework to test whether plasma BDNF expression mediated the association between sex (coded 1: female; 2: male) and VLM scores. Models were adjusted for time between blood sample collection and cognitive testing, as well as participant age, years of education, hippocampal volume, family history of AD, and APOE ε4 carriage. Bootstrapping of model coefficients was performed to estimate the indirect effect of plasma BDNF on the sex–VLM relationship. Finally, the dataset was stratified by APOE ε4 carriage (0: no allele, 1: at least one allele) and mediation analyses were repeated within each strata. All analyses were completed using IBM SPSS version 27.0 with the PROCESS Macro v3.4. ²² Findings with p < 0.05 (two‐tailed) were considered significant.

3. RESULTS

3.1. Participant data

Table 1 presents participant data stratified by sex and APOE ε4 allele carriage. The sample included n = 201 participants (66% female), with a mean ± standard deviation (SD) age of 63.8 ± 6.0 years. The group was highly educated (mean ± SD education 16.4 ± 2.7 years) and primarily White (94%, self‐reported). APOE ε4 carriage was present in 38% of the overall sample while BDNF_Val66met carriage was present in 31%, and 66% participants reported family history of AD.

APOE ε4 carriers had a significantly lower proportion of BDNF_Val66met carriage and a higher proportion with a family history of AD compared to non‐carriers (p < 0.05). Female participants had significantly higher hippocampal volumes than males, while males reported more years of education (p < 0.05). BDNF levels did not significantly differ between groups, although mean VLM z scores were higher in females than in male participants.

3.2. Influence of BDNF on the relationship between sex and VLM

As shown in Table 2, covariate‐adjusted regression models from the full study sample indicated that females had significantly higher VLM scores (β = −0.67; p < 0.01) and plasma BDNF levels (β = −0.41; p = 0.01) than males. Likewise, BDNF expression was positively correlated with VLM (β = 0.18; p = 0.01). Bootstrapping indicated that BDNF was a significant partial mediator of the association between sex and VLM (β = −0.07; 95% confidence interval [CI]: −0.18, −0.01).

3.3. Conditional dependence of BDNF mediation on APOE ε4

Models stratified by APOE ε4 carriage showed no significant association between sex and BDNF in APOE ε4+ individuals (β = −0.14; p = 0.60). However, females in this group again had significantly higher VLM scores than males (β = −0.53; p = 0.03). Table 3 presents these results.

By comparison, in the APOE ε4– group (Table 4), females had significantly higher BDNF levels (β = −0.68; p < 0.01) and VLM scores (β = −1.06; p < 0.01), and bootstrapping again showed BDNF to be a significant mediator, with an effect size more than double that which was observed in the full sample (β = −0.18; 95% CI: −0.37, −0.05).

4. DISCUSSION

The major findings of this study were that (1) BDNF partially mediates the relationship between sex and VLM scores in a population at risk for AD, such that females had greater circulating BDNF levels commensurate with higher VLM scores, and (2) that this association was only found in individuals without the APOE ε4 allele. As APOE ε4 carriage is a hallmark genetic risk factor for AD, this study principally describes an association between sex and verbal memory—likewise, that APOE ε4 allele carriage may attenuate the association between BDNF and VLM.

Altogether, these findings suggests that VLM in aging may be protected by a form of sex‐specific cognitive reserve; indeed, studies consistently indicate that female sex is associated with greater VLM ability across the human lifespan ²³ , ²⁴ but diminishes at a faster rate in females with dementia. ²⁵ However, the relationship between sex and VLM in the context of AD pathophysiology remains unclear. Neurocognitive tests such as the RAVLT are widely used to estimate cognitive decline in AD; at the cortical level, the medial temporal lobe has been shown to mediate verbal memory ²⁶ and is one of the first cortical regions to exhibit AD‐related neuropathology. ²⁷ Emerging evidence highlights sex‐based differences on the association between verbal memory and temporal lobe glucose metabolic rates (TLGluMR), an in vivo measure of neuronal integrity, in patients with MCI and AD dementia. ²⁸

Our study echoes these findings by investigating circulating BDNF as another mediator of cognitive function, and we further extend prior reports by stratifying individuals by APOE ε4 carriage. BDNF is known to play a crucial role in mediating memory formation and consolidation through supporting synaptic health and promoting long‐term potentiation. ²⁹ The highest levels of BDNF in the brain are found in the hippocampus, a region vital for learning and memory. Each of the three different isoforms of the APOE gene (ε2, ε3, and, ε4) have been found to differentially regulate BDNF production; a recent review by Miranda et al. ³⁰ proposed that APOE ε2 and APOE ε3 alleles may offer neuroprotection through synergistic effects on BDNF regulation, whereas APOE ε4 carriage—one of the strongest known risk factors for AD—may instead impact neurodegeneration and synaptic health through the inhibition of BDNF production. APOE ε4 has been shown to epigenetically suppress BDNF mRNA expression by inducing the nuclear translocation of histone deacetylases ³¹ which not only attenuates BDNF synthesis, but also blocks its secretion. ³² Our findings potentially reflect this suppressive effect, as we found that APOE ε4– females had almost 27% higher plasma BDNF than APOE ε4+ females.

BDNF transcription and secretion may also be influenced by hormones such as estrogens; ³³ indeed, estrogen may be a key explanatory mechanism behind our observations. Estrogen‐mediated synaptogenesis and neuroprotection may be regulated by central BDNF secretion. ³⁴ While the mechanisms here are still unclear, the BDNF gene contains a motif analogous to that of the canonical estrogen receptor element (ERE), which mediates the biological response of the hormone ³⁵ and may therefore explain estrogen's relationship with BDNF transcription. Gene transcription via ERE binding has been shown to increase downstream BDNF production, which could likewise explain estrogen's neuroprotective function. ³⁶ Furthermore, preclinical findings from estrogen replacement therapy found increased hippocampal BDNF expression and improved working and verbal memory in AD patients. ³⁷

As estrogen may play an important role in linking BDNF to the AD pathophysiological process, investigating the role of menopause by comparing pre‐ and post‐menopausal females may be an important direction for future investigation. The onset of menopause is usually associated with a significant decrease in estrogen production; a recent study by Konishi et al. ³⁸ found that lower BDNF in post‐menopausal female subjects was associated with worse working memory function—a relationship not observed in men or pre‐menopausal females—indicating a menopause‐specific effect on memory due to BDNF. Other studies have reported menopause‐dependent differences in amyloid beta (Aβ) and hyperphosphorylated tau burden ³⁹ hallmark biomarkers indicative of AD. Exercise may serve as a powerful treatment option for post‐menopausal females and AD risk as it has been shown to increase release of brain‐based BDNF into circulation. ⁴⁰ To best understand the effects of exercise and estrogen on BDNF and memory function, longitudinal exercise intervention studies are warranted.

There are limitations to note for this work. Our sample was composed almost entirely of White individuals with high levels of education; as education and hippocampal volume have been associated with VLM scores in healthy older adults, ⁴¹ both were included as covariates in all models. These findings therefore may not be generalizable to underrepresented communities or individuals with lower educational attainment. In general, the external validity of the present work was limited to a VLM construct derived from the RAVLT; as such, it is unclear whether these findings generalize to individuals from non‐English (and non–Latin‐based language) populations.

Likewise, sex differences in other dimensions of cognitive decline and risk of dementia may be attributable to differences in social determinants of health or lifestyle exposures. Exploring gene–lifestyle interactions—while disentangling sex from gender as a social construct—holds promise in elucidating sex‐specific differences in cognitive aging and the onset of AD. This could help to shed light on the distinctiveness of cognitive reserve in females, a concept currently challenging to disentangle from a potential female advantage in verbal tasks. Moreover, the opportunistic selection of males and females in the study, based solely on data completeness rather than demographic or health‐related variables, further complicates the interpretation of these findings as evidence for sexual dimorphism in cognitive reserve.

Finally, nearly two thirds of the present sample were female, which may have limited the statistical power of our models for male individuals. While our study design allowed us to explore potential differences in cognitive outcomes between APOE ε4 carriers and non‐carriers, the unequal distribution of participants across these groups challenges the achievement of optimal statistical power—particularly in the context of mediation analyses. Consequently, the interpretation of our findings should be tempered by the recognition of potential limitations in statistical power, which may have limited the detection of significant interactions or main effects.

Findings from this study highlight the need for continued investigation in several areas. First, testing the influence of menopausal status and estrogen levels on circulating BDNF and their commensurate effects on sex differences in cognitive function could clarify whether these findings are over‐generalized to all female individuals. There is also an opportunity to further explore additional factors that may regulate BDNF transcription and secretion; for example, the Val66Met polymorphism in the BDNF gene has been shown to modulate the secretion, translation, and intracellular trafficking of dendritic BDNF Mrna. ⁴² Testing the influence of cardiorespiratory fitness may be additionally insightful, as BDNF production likely interacts with other cardiometabolic factors, such as cardiovascular function, insulin regulation, and stress‐influenced inflammatory processes. However, the longitudinal investigation of these underlying mechanisms is lacking, and replication studies will be needed to further disambiguate observed sex differences in these processes.

CONFLICT OF INTEREST STATEMENT

All authors have contributed to this work and agree with the presented findings; this manuscript has not been published before, nor is it being considered for publication in another journal. The authors have no conflicts of interest to declare. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All study activities were approved by the University of Wisconsin–Madison Institutional Review Board. All human subjects provided written informed consent.

Supporting information

Supporting Information

Edmunds KJ, Pandos AA, Hoang I, et al. BDNF expression mediates verbal learning and memory in women in a cohort enriched with risk for Alzheimer's disease. Alzheimer's Dement. 2025;17:e270062. 10.1002/dad2.70062

DATA AVAILABILITY STATEMENT

Request for the data used in this study may be directed to the corresponding authors.