Potential role of APOE ɛ4 allele as a modifier for the association of BDNF Val66Met polymorphisms and cognitive impairment in community-dwelling older adults

Objective To determine whether the brain-derived neurotrophic factor (BDNF) Val66Met polymorphism is associated with cognitive impairment (CI) in community-dwelling Chinese older adults, and to investigate whether this relationship is modified by the Apolipoprotein E (APOE) ɛ4 allele. Methods The study is a secondary analysis of 703 participants aged ≥60 years randomly enrolled from the Beijing Longitudinal Study of Aging II prospective cohort. The education-adjusted Mini-Mental State Examination and the Clinical Dementia Rating Scale were used to measure the cognitive performance of the subjects. The main effects and interactions (additive and multiplicative) of the BDNF Met and the APOE ε4 alleles on CI were estimated by logistic regression models. Results In total, 84 out of 703 older adults aged ≥60 years old had CI. No significant difference was observed in the risk of CI between participants with the BDNF Met allele and that of subjects without the BDNF Met allele (p = 0.213; p = 0.164). Individuals carrying both the BDNF Met and APOE ε4 alleles had an almost 1.5-fold increased odds of CI compared with carriers of the BDNF Met allele but without the APOE ε4 allele. The additive association indicated a positive interaction of both BDNF Met and APOE ε4 alleles with wide CIs (p = 0.021; p = 0.018). Conclusion The results suggest that the APOE ε4 allele may be a potential modifier for the association of the BDNF Val66Met polymorphism with CI in community-dwelling older adults.


Introduction
Aging is usually described as a multifaceted deterioration within cognitive disorders and other physical dysfunctions that are known to induce the loss of functional capacity and decreased quality of life in older adults.Aging is often accompanied by a decline in some cognitive domains including memory, learning, concentration, execution, and calculation.However, cognitive dysfunction ascribed to senescence is not equally observed in adults, and a substantial proportion of individuals still maintain healthy cognitive function even into old age.It has been reported that some genetic predisposition to different cellular and molecular neurobiological factors affecting long-term cognitive ability may explain the heterogeneity seen in cognitive performance in older people (Alzheimer's, 2012).Comprehending the genetic sources of heterogeneity in cognitive aging could provide crucial contributions to future endeavors aimed at screening, treatment, or prevention of neurodegenerative diseases such as Alzheimer's disease (AD).
It has been reported that some cognitive disorders observed in the elderly may be associated with disruptions in the neurotrophic systems (Mattson et al., 2004).Genetic studies have indicated that variations of alleles within the neuroplasticity-related gene encoding for neurotrophins may be potential sources of some individual variations in cognitive aging (Barha et al., 2019).Brain-derived neurotrophic factor (BDNF) is characterized to partake in neuronal growth and differentiation and synaptic plasticity, which performs an essential role in memory storage and learning (Song et al., 2015).The human BDNF gene located on chromosome 11p14.1, is composed of 11 exons and 9 functional promoters.A guanine to adenine single nucleotide polymorphism (SNP) within the pro-domain region of the BDNF gene at position 196 of exon 2 results in an amino acid substitution from valine (Val) to methionine (Met) at codon 66, which is associated with reduced levels of the activity-dependent neuronal secretion of the mature form of BDNF (Egan et al., 2003;Chen et al., 2008).Although numerous studies have shown a relationship between Val66Met polymorphism in the BDNF gene and cognitive performance, some have failed to establish such an association (He et al., 2007;Bicalho et al., 2018;Zhao et al., 2018); the influence of the Met allele, in particular, has not been determined (Toh et al., 2018).Even the relationship between BDNF Val66Met polymorphisms and cognitive function remains controversial (Brown et al., 2020).Previous studies have confirmed that older adults with the Met allele have a higher risk of cognitive impairment (CI) compared to Val homozygotes carriers (de Azeredo et al., 2017;Lim et al., 2018;Tan et al., 2018), and some have found that the Val allele may increase this risk (Ventriglia et al., 2002;Matsushita et al., 2005;Rezaei et al., 2017).These contradictory findings may be partly due to the omission of some confounding factors such as age (Brown et al., 2020), sex (Li et al., 2017;Barha et al., 2019) and other neurodegenerative pathologies.
The apolipoprotein E epsilon 4 (APOE ε4) allele is a major genetic risk factor for sporadic AD (Reiman et al., 2007).In recent years, BDNF Val66Met polymorphism has been reported to interact with APOE ɛ4 on working memory, verbal and visual episodic memory, and the progression of mild cognitive impairment (MCI) (Forlenza et al., 2010;Ward et al., 2014;Lim et al., 2015;Gomar et al., 2016).While some studies did not indicate the influence of BDNF/APOE gene-gene interaction on cognitive performance, incidents of AD or MCI (Forlenza et al., 2010;Richter-Schmidinger et al., 2011;Zhao et al., 2018).There is limited information and uncertain conclusions regarding the interactive effect of the BDNF Val66Met gene polymorphism and the APOE ε4 allele on the risk of CI in communitydwelling elderly.The purpose of this study is to investigate whether the BDNF Val66Met polymorphism may confer CI risk through the modifying effect of APOE ε4, based on data from the Beijing Longitudinal Study of Aging II (BLSA II) prospective cohort.

Study design and participants
This is a secondary analysis of data from the BLSA II prospective cohort project, which randomly enrolled 10,039 community-dwelling residents aged ≥55 years old in Beijing in 2009 (Ji et al., 2020).All eligible residents provided their informed consent.The research ethics committee of Xuanwu Hospital at Capital Medical University provided approval for the protocol of this study.
DNA samples were available for 8,405 subjects of the BLSA II prospective cohort, among which 730 older adults aged ≥60 years were randomly enrolled in the present study and completed both BDNF Val66Met and APOE genotype testing.For the current analysis, 27 participants without completed cognitive assessments (n = 21) or education data (n = 6) were excluded.

Cognitive measures and CI definition
The global cognitive function of the subjects was measured using the Mini-Mental State Examination (MMSE) and the Clinical Dementia Rating Scale (CDR).As described elsewhere (Cui et al., 2011), CI was defined by the CDR score ≤ 1, and the MMSE score ≤ 17 (illiterate) ≤ 20 (primary school) or ≤ 24 (secondary school or above).

Assessment of covariates
These covariates were selected a priori as potential confounders based on the literature (Stuck et al., 1999;Kalaria et al., 2008).A standardized self-administered questionnaire comprising information on demographics, smoking and drinking history, comorbidities (i.e., cerebrovascular disease, hypertension, heart disease, diabetes, cataracts, hearing impairment) and depression was administered in a face-to-face interview by trained physicians and nurses.Smoking or drinking status was categorized as "never" or "past and current." Depression was assessed by the 15-item Geriatric Depression Scale (GDS) with a score of ≥5 (Almeida and Almeida, 1999).

Statistical analysis
Allele and genotype frequencies were determined by counting and calculating sample proportions, and the Hardy-Weinberg equilibrium was estimated by the Chi-squared (χ 2 ) test.Continuous variables were described as mean ± standard deviation (SD) and analyzed by the Mann-Whitney U test based on distributional properties.Categorical variables were presented as percentages and frequencies and analyzed by the χ 2 tests.
Participants were defined as BDNF Met allele carriers who had at least one allele of Met (Val/Met and Met/Met), and the others were defined as BDNF Met allele (Val/Val) negative.The association of the BDNF Val66Met polymorphism with CI was assessed by binary logistic regression analyses, which were performed with initial adjustment for age, sex, and years of education (model 1), and further adjustment for the variables in model 1 plus other potential confounding factors (model 2).
All individuals were divided into four groups to estimate possible joint effects of the BDNF Met and the APOE ε4 alleles on CI: participants carrying neither the BDNF Met allele nor the APOE ε4 allele (group A, reference group), participants carrying the BDNF Met allele but not the APOE ε4 allele (group B), participants without the BDNF Met allele but with the APOE ε4 allele (group C) and participants with both the BDNF Met and APOE ε4 alleles (group D).The risk of CI in all groups was tested by binary logistic regression models (model 1 and Model 2).Possible additive effects of the BDNF Met and APOE ε4 alleles were examined by comparing the odds ratios (ORs) of group D with the summed ORs of groups B and C, with a calculation of the relative excess risk due to interaction (RERI) (RERI score>0: positive additive interaction; RERI score<0: negative interaction) (Hosmer and Lemeshow, 1992;Muller-Gerards et al., 2019).
The effect of the APOE ε4 allele on the relationship between the BDNF Val66Met polymorphism and CI was calculated with two logistic regression models (model 1 and model 2).The ORs for incident CI in APOE ε4 individuals carrying the BDNF Met allele were compared with the ORs in the subjects without the BDNF Met allele.The measure of interaction between the BDNF Met and APOE ε4 alleles on a multiplicative scale was calculated based on the following logistic regression model (de Mutsert et al., 2009):

Results
The full sample of this study consisted of 703 participants which included 84 (11.95%) individuals with CI and 619 subjects with normal cognitive function.There were 210 subjects carrying (29.87%)Val/Val homozygotes, 350 (49.79%) with Val/Met heterozygotes, and 143 with (20.34%)Met/Met homozygotes, with no deviation from the Hardy-Weinberg equilibrium (p = 0.99).A total of 264 individuals carried at least one APOE ε4 allele.Table 1 shows the demographic and clinical characteristics.
Results of binary logistic regression analyses for the independent influence of the APOE ε4 or BDNF Met alleles on CI are shown in Table 2. Participants carrying the APOE ε4 allele had a higher risk of CI compared to those without the APOE ε4 allele (p < 0.001 for model 1 and model 2).There was no significant variation in the distribution of CI status between participants with the BDNF Met allele and subjects without the BDNF Met allele (p = 0.213 for model 1; p = 0.164 for model 2).
Tables 3, 4 show the results of logistic regression analyses for the interactive effect of the APOE ε4 and BDNF Met alleles on CI, with the highest OR for CI in group D (p = 0.006 for model 1; p = 0.003 for model 2).Among APOE ε4 allele carriers, individuals with the BDNF Met allele showed an almost 1.5-fold higher OR for CI compared with subjects without the BDNF Met allele (p = 0.041 for model 1; p = 0.036 for model 2), which was not observed in participants without the APOE ε4 allele (right column in Tables 3, 4).RERI scores indicated a positive additive interactive effect of both the BDNF Met allele and the APOE ε4 allele on the risk of CI (p = 0.021 for model 1; p = 0.018 for model 2).A possible multiplicative positive interaction was detected for the two adjusted models, but neither of the two multiplicative scales reached statistical significance (p = 0.081 for model 1; p = 0.089 for model 2).

Discussion
In the present study, no independent effect of the BDNF Val66Met polymorphism on CI was observed in communitydwelling older adults.Our results showed a higher risk of incident CI in individuals with the BDNF Met allele compared with Val/Val homozygotes in APOE ε4 allele carriers, and a positive interactive effect of carrying the BDNF Met allele and the APOE ε4 allele on CI, which suggested that BDNF Val66Met polymorphism might confer the risk of CI via the interaction of the APOE ε4 allele in communitydwelling older adults.
Many studies have attempted to explore the correlation between Val66Met polymorphisms in the BDNF gene and CI with ambiguous findings.While some studies indicated that the BDNF Met allele was associated with cognitive dysfunction (Ventriglia et al., 2002;Matsushita et al., 2005;Lin et al., 2014), similar to our findings, other works failed to find this significant relationship (Desai et al., 2005;He et al., 2007;Kim et al., 2011).Our results provided a possible explanation that the BDNF Met allele might increase the risk of CI among community-dwelling elderly residents, potentially interacting with the APOE ε4 allele.A number of possible reasons could explain the modifying effect of the APOE ε4 allele.
First, many lines evidence suggest that BDNF Val66Met is a downstream mediator of amyloid beta (Aβ) toxicity on hippocampal function (Lim et al., 2013).Lim et al. found that subjects with the BDNF Met allele showed significant cognitive decline as compared to Val/Val homozygotes, in healthy individuals with high levels of Aβ accumulation.BDNF Val66Met was not observed to be associated with cognitive deficits in subjects with low Aβ.These findings show that carrying the BDNF Met allele could hasten the onset of clinically significant cognitive dysfunction related to the presence of a high brain Aβ load (Lim   Aβ deposition is well known to be associated with APOE genotype, particularly with the number of APOE ε4 alleles (Verghese et al., 2011;Villemagne et al., 2011;Liu et al., 2013).It has been reported that among APOE ε4 carriers, carrying the BDNF Met allele correlates to a greater Aβ load compared to Val homozygotes, particularly in the precuneus, orbitofrontal cortex, gyrus rectus, and lateral prefrontal cortex (Adamczuk et al., 2013).Also, BDNF Met within APOE ε4 carriers had significantly more amyloid deposition in regions typically affected by AD (Stonnington et al., 2020).This supports our findings that APOE ε4 may influence the association of BDNF Val66Met polymorphism and CI through Aβ deposition.
Second, the interactive effect of APOE/BDNF on the integrity of brain functional connectivity may provide another explanation for our findings.Recent work showed slightly decreased functional connectivity within the Dorsal Attention Network (DAN) in BDNF Met carriers/APOE ε3 homozygotes compared to BDNF Met/APOEε4 carriers in healthy older adults (Pietzuch et al., 2021).
However, the mechanism underlying the correlation between APOE/BDNF interactions and CI remains elusive.It's reported that the Met allele alters the trafficking and packaging of intracellular pro-BDNF, thereby regulating the secretion of mature BDNF (mBDNF) (Chen et al., 2004).Although the BDNF gene Val66Met polymorphism does not appear to affect the secretion of pro-BDNF and mBDNF (Mo et al., 2021), APOE ε4 blocks the secretion of mature-BDNF (Rainey-Smith et al., 2014;Sen et al., 2017).Pro-BDNF and mBDNF have opposing biological processes (Lu et al., 2005).There may be a mechanistic link between the BDNF Val66Met polymorphism and APOE isoforms in regulating mature-BDNF secretion and conversion of pro-BDNF to mBDNF.
Although there have been many suggestions that BDNF Val66Met polymorphism is involved in cognitive dysfunction (de Azeredo et al., 2017;Lim et al., 2018;Tan et al., 2018), no clear evidence of the association between BDNF Val66Met and CI has been explored in previous studies.The individuals in the current study were randomly selected from community-dwelling elderly subjects in the BLSA II prospective cohort.This increases the authenticity and reliability of the data and reduces recruitment bias.Our study adds cross-sectional evidence to a growing body of literature on the correlation of BDNF Val66Met polymorphism with CI.Our findings demonstrate the association of BDNF Val66Met with CI, and show the necessity of APOE ε4 for this relationship.The strength of these results provides more evidence to re-evaluate the effect of the BDNF Met and the APOE ε4 alleles on cognitive dysfunction in the elderly.
Our research has several limitations.First, the number of participants was limited, although individuals in the current study were randomly selected from the BLSA II cohort.Second, the MMSE and CDR scores with evidence of a marked ceiling effect were used to define CI, and may have missed important levels of CI.Additionally, the physical activity of older adults, which could potentially affect the association of BDNF Val66Met polymorphisms with cognitive function was not included in our analysis.Further longitudinal studies are required to explore the association of BDNF Val66Met polymorphisms with CI or other forms of cognitive decline of various etiologies to confirm our findings.

Conclusion
In conclusion, the results demonstrate the role of the APOE ε4 allele in modifying the association between BDNF Val66Met and CI in community-dwelling older adults.This finding may provide further evidence to evaluate the influence of BDNF Val66Met polymorphism and the APOE ε4 allele on cognitive dysfunction in the elderly and contribute to the screening, treatment, or prevention of CI in the future.

TABLE 2
Odds ratios for CI by the occurrence of the APOE ε4 allele and the BDNF Met allele.

TABLE 3
Risk of CI for BDNF Met allele and APOE ε4 genotype groups (model 1).
ORs are adjusted for age, sex, education.

TABLE 1
Study participant characteristics by cognitive function status.