Next Article in Journal
Investigating the Impact of a Curse: Diseases, Population Isolation, Evolution and the Mother’s Curse
Next Article in Special Issue
The Genetic Side of the Mood: A Scientometric Review of the Genetic Basis of Mood Disorders
Previous Article in Journal
Plasmid Genomes Reveal the Distribution, Abundance, and Organization of Mercury-Related Genes and Their Co-Distribution with Antibiotic Resistant Genes in Gammaproteobacteria
Previous Article in Special Issue
Recent Developments in Autism Genetic Research: A Scientometric Review from 2018 to 2022
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Genes
Volume 13
Issue 11
10.3390/genes13112150
genes-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

SNP-by-CpG Site Interactions in ABCA7 Are Associated with Cognition in Older African Americans

by
Dima L. Chaar
1,
Kim Nguyen
1,
Yi-Zhe Wang
1,
Scott M. Ratliff
1,
Thomas H. Mosley
2,
Sharon L. R. Kardia
1,
Jennifer A. Smith
1,3,† and
Wei Zhao
1,3,*,†
1
Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, MI 48109, USA
2
Memory Impairment and Neurodegenerative Dementia (MIND) Center, University of Mississippi Medical Center, Jackson, MI 39216, USA
3
Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI 48104, USA
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
Genes 2022, 13(11), 2150; https://doi.org/10.3390/genes13112150
Submission received: 12 September 2022 / Revised: 21 October 2022 / Accepted: 10 November 2022 / Published: 18 November 2022
(This article belongs to the Special Issue Genetic and Epigenetic Factors and Their Interactions in the Susceptibility to Multifactorial Diseases)
Browse Figures
Review Reports Versions Notes

Abstract

:
SNPs in ABCA7 confer the largest genetic risk for Alzheimer’s Disease (AD) in African Americans (AA) after APOE ε4. However, the relationship between ABCA7 and cognitive function has not been thoroughly examined. We investigated the effects of five known AD risk SNPs and 72 CpGs in ABCA7, as well as their interactions, on general cognitive function (cognition) in 634 older AA without dementia from Genetic Epidemiology Network of Arteriopathy (GENOA). Using linear mixed models, no SNP or CpG was associated with cognition after multiple testing correction, but five CpGs were nominally associated (p < 0.05). Four SNP-by-CpG interactions were associated with cognition (FDR q < 0.1). Contrast tests show that methylation is associated with cognition in some genotype groups (p < 0.05): a 1% increase at cg00135882 and cg22271697 is associated with a 0.68 SD decrease and 0.14 SD increase in cognition for those with the rs3764647 GG/AG (p = 0.004) and AA (p = 2 × 10−4) genotypes, respectively. In addition, a 1% increase at cg06169110 and cg17316918 is associated with a 0.37 SD decrease (p = 2 × 10−4) and 0.33 SD increase (p = 0.004), respectively, in cognition for those with the rs115550680 GG/AG genotype. While AD risk SNPs in ABCA7 were not associated with cognition in this sample, some have interactions with proximal methylation on cognition.

1. Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disease characterized by the dysregulation of the amyloid-β (Aβ) pathway, leading to Aβ plaques [1] and the aggregation of tau tangles [2]. AD accounts for 60–80% of dementia cases in the elderly [3,4,5]. Approximately 6.2 million Americans age 65 and older are living with AD, and this estimate is projected to rise to 13.8 million by 2060 [3]. AD risk differs by race, with African Americans (AA) twice as likely to develop AD compared to European Americans (EA) [6]. Because this health disparity places a greater burden of personal and medical care on AA, it is crucial to better understand AD and its development in this population.
AD is a multifactorial disease that is likely influenced by interactions between genetic, environmental, and epigenetic factors, along with age-related neurodegeneration [7]. In addition to age, genetic variants in the apolipoprotein E (APOE) gene are the largest risk factor for AD in AA [8], with one copy of the APOE  ε 4 allele increasing AD risk 3–5-fold [9,10,11]. ABCA7 is the second largest genetic risk factor for AD in AA, with genetic variants increasing AD risk by 70–80% [8]. The ABCA7 gene encodes the ATP-binding cassette (ABC) transporter A7, which regulates homeostasis of phospholipids and cholesterol in the central nervous system and peripheral tissues [12,13,14]. This gene is mostly expressed in the brain, spleen, lungs, and adrenal gland [15]. Studies have suggested that mutations in ABCA7 are associated with AD susceptibility through the dysregulation of lipid metabolism, which facilitates A β clearance [13,14].
Though ABCA7 is a risk locus for AD in both EA and AA, the specific risk variants differ across groups [16]. In EA, three ABCA7 SNPs, rs3764650, rs3752246, and rs4147929, are associated with AD. They represent two independent signals, as rs3752246 and rs4147929 are in nearly complete linkage disequilibrium (LD) in EA. Although rs3764650 shows the strongest association with AD in EA, it is only nominally associated in AA [16,17]. In AA, two additional ABCA7 SNPs, rs3764647 and rs3752239, have stronger associations with AD [17], with rs3764647 being in the same LD block as rs3764650 in AA. Interestingly, another independent SNP in ABCA7, rs115550680, which is monomorphic in EA, is strongly associated with AD in AA. In particular, the G allele of rs115550680 confers an AD risk comparable to APOE ε4 (OR = 1.79) in AA [8].
Epigenetic modifications, such as DNA methylation, are potential molecular mechanisms that can modulate the effect of genetic risk factors [18]. When methylation sites (CpGs) are clustered together as a CpG island (CGI), it often serves as a hub for gene expression regulation. CGIs in the promoter region usually suppress transcription, whereas CGIs in the intragenic region can interact with multiple regulatory elements and have a variety of impacts on gene expression (e.g., influencing mRNA isoforms or promoting enhancer function) [19]. Given the regulatory role of DNA methylation on gene expression, there has been a growing interest in understanding the extent to which DNA methylation contributes to AD risk [20,21,22,23,24]. In particular, recent studies on post-mortem brain tissue found evidence of association between DNA methylation in ABCA7 and both AD and AD-related pathologies, including Aβ load and tau tangle density [21,22]. This evidence suggests that methylation in ABCA7 has a non-trivial functional role which is worthy of further investigation.
Although the relationships between AD and ABCA7 SNPs are well-characterized, there are limited studies on the association between genetic variation in ABCA7 and measures of cognitive function and/or cognitive decline prior to the development of dementia. An imaging study showed that ABCA7 SNPs were associated with amyloidosis among cognitively healthy individuals and those with mild cognitive impairment, but not among those with AD, suggesting an early effect of ABCA7 on cognition and cognitive decline [25]. A few studies in EA found inconsistent results regarding the effect of ABCA7 SNPs on cognition, with associations varying by sex, APOE status, and disease progression [26]. For example, in healthy older adults, a longitudinal study found an association between rs3764650 and cognitive decline, but only in females [27]. Additionally, interactions between the APOE ε4 allele and SNPs rs3764650 and rs3752246 were associated with three cognitive factor scores related to verbal learning and memory, working memory, and intermediate memory, in a genotype-dependent manner: in the absence of ABCA7 minor alleles, each additional ε4 allele was associated with lower memory scores; and conversely, in the presence of ABCA7 minor alleles, each additional ε4 allele was associated with better memory scores [28]. Lastly, rs3764650 was significantly associated with increased rates of memory decline among individuals with mild cognitive impairment or AD [29].
To our knowledge, no study has investigated the relationship between ABCA7 genetic variation and cognition in cognitively healthy AA. Further, few studies have examined the relationship between DNA methylation in ABCA7 and/or its interaction with genetic variants on general cognitive function. In this study, we investigate whether previously identified risk SNPs (referred to as sentinel SNPs) in ABCA7, DNA methylation in ABCA7, and their interactions are associated with general cognitive function in older AA without dementia. In order to better understand the functional consequences of these risk factors at the molecular level, we also evaluated whether identified epigenetic or genetic risk factors are associated with transcript level ABCA7 gene expression in transformed B-lymphocytes from the same cohort. A thorough investigation of the relationship between these multi-omic layers and later-life cognition can help characterize the underlying genetic architecture of cognition in older adulthood, prior to dementia onset. This may allow the identification of targets for intervention and treatment, especially in populations that are most at risk [30].

2. Materials and Methods

2.1. Sample

The Genetic Epidemiology Network of Arteriopathy (GENOA) study is a community-based longitudinal study aimed at examining the genetic effects of hypertension and related target organ damage [31]. European American (EA) and African American (AA) hypertensive sibships were recruited if at least two siblings were clinically diagnosed with hypertension before age 60. All other siblings were invited to participate, regardless of hypertension status. Exclusion criteria included secondary hypertension, alcoholism or drug abuse, pregnancy, insulin-dependent diabetes mellitus, active malignancy, or serum creatinine levels > 2.5mg/dL. In Phase I (1996–2001), 1854 AA participants (Jackson, MS) and 1583 EA participants (Rochester, MN) were recruited [31]. In Phase II (2000–2004), 1482 AA participants and 1239 EA participants were successfully followed up, and their potential target organ damage from hypertension was measured. Demographics, medical history, clinical characteristics, information on medication use, and blood samples were collected in each phase. Methylation levels were measured only in AA participants using blood samples collected in Phases I and II. In an ancillary study (2001–2006), 1010 AA and 967 EA GENOA participants underwent a battery of established neurocognitive tests to assess several measures of cognitive function, including learning, memory, attention, concentration, and language. Written informed consent was obtained from all participants, and approval was granted by participating institutional review boards (University of Michigan, University of Mississippi Medical Center, and Mayo Clinic).
A total of 850 AA participants had non-missing genetic and demographic data. Since participants with a history of stroke or dementia may have had changes in general cognitive function that differed from non-pathological cognitive aging, we excluded those who had a history of stroke (n = 43) and/or preliminary evidence of dementia as indicated by a score of <24 on the Mini-Mental State Examination (MMSE) (n = 76) [32]. We also excluded participants younger than age 45 (n = 16). A total of 634, 494, and 429 participants were available for SNP, methylation, and gene expression analyses, respectively (Figure S1).

2.2. Measures

2.2.1. General Cognitive Function

General cognitive function was calculated using five neurocognitive measures evaluated at Phase II [32,33]:
  • The Weschler Adult Intelligence Scale-Revised: Digit Symbol Substitution Test (DSST) measured complex visual attention, sustained and focused concentration, response speed, and visuomotor coordination. The DSST is related to the executive function of working memory in cognition [34]. In this test, participants matched symbols to numbers according to a key located at the top of the page. The DSST score comprised the number of symbols correctly matched within 90 s.
  • The Controlled Oral Word Association Test (COWA-FAS) tested for verbal fluency (phonetic association) and language. This required participants to generate as many words as possible that start with F, A, and S in 1 min. The score consisted of the total number of admissible words generated.
  • The Rey Auditory Verbal Learning Test (RAVLT) measured delayed recall, relating to the cognitive functions of new learning, immediate memory span, and vulnerability to learning interference, and recognition memory. Scores were determined by the number of words recalled after a 30-min delay. Scores ranged from 0 to 15.
  • The Stroop Color–Word Test (SCWT) assessed concentration effectiveness by requiring participants to state the color of a word, rather than the word written. The score sums the number of color words that were correctly stated in 45 s. Specifically, the ability to shift perceptual sets in response to novel stimuli was tested.
  • The Trail Making Test A (TMTA) evaluated visual conceptual tracking, as participants are required to connect a set of 25 circles quickly and accurately. TMTA provided information on the cognitive functions of visual search, scanning, processing speed, and executive functions. The TMTA score was measured as the amount of time (seconds) the participants took to complete the task. The maximum time allowed was 240 s. Prior to analysis, TMTA scores were natural log-transformed and recoded so that higher scores indicated better cognitive function.
General cognitive function, a measure of overall cognitive performance, can be quantified as a summary measure of cognitive tests in multiple cognitive domains [35]. In this study, general cognitive function was calculated as the first unrotated principal component (FUPC) from a principal component analysis (PCA) of the five neurocognitive measures in the full sample (n = 634). The FUPC accounted for 53% of the total variance in the neurocognitive measures, and loading values of the five measures ranged from 0.52 to 0.87.

2.2.2. Demographic Data

Age was assessed at the time of cognitive testing. Educational attainment, measured at Phase II, was categorized into a three-level variable: (1) less than high school degree (reference group), (2) high school degree or GED, and (3) at least some college. Smoking has been shown to have a substantial impact on the epigenome [36], so we used smoking data concurrent with DNA methylation measures (Phase I). Participants were categorized as current, former, or never smokers (reference group).

2.2.3. Genetic Data

Blood samples were genotyped using the Affymetrix® Genome-Wide Human SNP Array 6.0 or the Illumina 1M Duo. Samples and SNPs with a call rate <95%, samples with mismatched sex, and duplicate samples were removed. Genotypes were imputed using the 1000 Genomes Project phase I integrated variant set (v.3) (Hg19, released in March 2012). Of the six SNPs of interest identified from the existing literature (rs3764647, rs3764650, rs115550680, rs3752246, rs3752239, and rs4147929), five had high imputation quality (r2 > 0.7), and one (rs3752239) was excluded due to low imputation quality (r2 = 0.49). SNPs were coded as the dosage of the corresponding AD risk allele, as specified in the previous literature. Genetic principal components were calculated from genotyped SNPs and included in regression models to control for population stratification. In order to evaluate confounding and/or effect modification by APOE isoforms known to influence dementia risk, we measured rs7412 (to capture the APOE ε2 allele) and rs429359 (to capture the APOE ε4 allele) using a TaqMan assay and ABI Prism© Sequence Detection (Applied Biosystems, Foster City, CA, USA) in 1544 participants. Participants were classified as having 0, 1, or 2 copies of ε2 (represented by the rs7412 T allele) and/or ε4 (represented by the rs429359 C allele).

2.2.4. DNA Methylation Data

Genomic data was extracted from stored peripheral blood leukocytes from 1106 AA participants from Phase I and 304 AA participants from Phase II using the AutoGen FlexStar (AutoGen, Holliston, MA, USA). Bisulfite conversion was performed with the EZ DNA Methylation Kit (Zymo Research, Irvine, CA, USA), and methylation was measured using the Illumina HumanMethylationEPIC BeadChip. The raw intensity data were visualized using the shinyMethyl R package [37] to identify sex mismatches and outliers, which were removed. Samples with incomplete bisulfite conversion were identified using Qcinfo in the Enmix R package [38], and then removed. Background correction and dye-bias normalization were performed using Noob in the Minfi R package [39,40]. We also checked sample identity using the 59 SNP probes on the EPIC chip, and mismatched samples were removed. Probe-type bias was adjusted using the Regression on Correlated Probes (RCP) method [41]. Probes with detection p-value < 10−16 were considered successfully detected, and probes and samples with a detection rate of <10% were removed [42]. After quality control, a total of 1396 samples (n = 1100 from Phase I and n = 294 from Phase II) and 857,121 CpG sites were available for analyses. For this analysis, all methylation data were collected from Phase I samples.
We selected all CpG sites within 5kb of the ABCA7 gene (a total of 72 CpG sites within the ABCA7 region: chr19, 1040102–1065570, and hg19). We used Illumina annotation [43] to characterize each CpG site as being in a promoter region and/or CGI, CGI shore, or CGI shelf. White blood cell proportions for CD8+ T lymphocytes, CD4+ T lymphocytes, natural killer cells, B cells, monocytes, and granulocytes were estimated using the Houseman method [44]. For each CpG site prior to analysis, the methylation beta-value [45,46] was multiplied by 100 to approximate the percent of methylation at that site. Methylation beta-values were pre-adjusted for batch effects (sample plate, row, and column) and white blood cell proportions using linear mixed modeling, and the resulting residuals were added to the mean values.

2.2.5. Gene Expression Data

Gene expression levels in transformed B-lymphocyte cell lines from blood samples taken primarily at GENOA Phase II were measured using the Affymetrix Human Transcriptome Array 2.0. The Affymetrix Expression Console was used for quality control, and all array images passed visual inspection. Affymetrix Power Tool software was used to process raw intensity data [47]. We normalized Affymetrix CEL files using the Robust Multichip Average (RMA) algorithm, including background correction, quantile normalization, log2-transformation, and probe set summarization [48]. Linearity was also maintained using GC correction (GCCN), signal space transformation (SST), and gain lock (value = 0.75). We used the Brainarray custom CDF [49] version 19 to map the probes to genes, specifically removing probes with non-unique matching cDNA/EST sequences that could be assigned to more than one gene cluster. As a result, the gene expression data processed through the custom CDF are expected to be free of mappability issues; however, alignment bias may still exist due to genetic variation, errors in reference genome, and other complications [50]. After mapping, Combat was used to remove batch effects [51].

2.3. Statistical Analysis

2.3.1. Genetic Analysis

We first calculated Pearson correlations between sentinel SNPs. Next, the association between ABCA7 sentinel SNPs and general cognitive function was analyzed using linear mixed models with random effects to adjust for relatedness. Model 1 adjusted for age at cognitive testing, sex, and the first four genetic principal components (PC1-4), with family as a random effect to account for sibships. Model 2 additionally adjusted for educational attainment. Model 3 further adjusted for APOE ε2 and ε4. For any SNPs that were significantly associated with general cognitive function, we further examined the association between those SNPs and each of the five neurocognitive measures to identify the domain(s) that most strongly drive the association. Since prior studies have suggested that the effect of ABCA7 SNPs may vary by sex, education, and/or APOE status, we also assessed the interaction between the sentinel SNPs and sex, education, or APOE (ε2 and ε4) on cognitive outcomes.

2.3.2. Epigenetic Analysis

Pearson correlations were calculated for all 72 CpG sites. Next, linear mixed models were used to test the associations between each of the 72 CpG sites and general cognitive function. Model 1 adjusted for age at cognitive testing, sex, four genetic principal components, age difference between methylation and cognition measurements, smoking status, and family as a random effect to account for sibships. Model 2 additionally adjusted for educational attainment, and Model 3 further adjusted for APOE ε2 and ε4. The coMET package in R was used to create a regional plot to visualize association p-values, correlations, and Ensembl genes [52]. BioRender was used to annotate and format the figure [53]. For any CpGs that were significantly associated with general cognitive function, we further examined the association between those CpGs and each of the five neurocognitive measures in order to identify the domain(s) that most strongly drive the association.

2.3.3. Genetic-Epigenetic Interaction Analysis

Next, we examined the interaction between each CpG site and sentinel ABCA7 SNPs in association with general cognitive function. In this analysis, we adjusted for age at cognitive testing, sex, four genetic principal components, age difference between methylation and cognition measurements, smoking status, and APOE ε2 and ε4, with family as a random effect to account for sibships (Model 4). Models 1–4, which were used to assess genetic, epigenetic, and genetic–epigenetic interaction associations with general cognitive function are shown in Figure S2. To improve interpretability, we mean-centered methylation so that the β estimates from the regression models reflect the effect sizes for those with average methylation in the population. For any identified significant interaction, we stratified the genotypes by number of risk alleles (0, 1, or 2 risk alleles) and conducted contrast tests using the Emtrends function in the Emmeans package in R [54] to obtain the effect size of the CpG associated with general cognitive function in each genotype group. Minor homozygote genotype groups that made up <5% of the sample size were grouped with heterozygous genotype groups to increase power as appropriate. Plots of SNP-by-CpG interactions on general cognitive function were generated using the effects [55] and ggplot2 [56] packages in R. Any identified SNP-by-CpG interactions significantly associated with general cognitive function were also tested for association with each of the five neurocognitive measures.
As a sensitivity analysis for significant interactions (FDR q < 0.1), we tested the association after excluding outlying CpG values that were more than four standard deviations from the mean (Model 4). We then assessed whether the SNP-by-CpG interactions (FDR q < 0.1) were driven by potential SNP-CpG correlations by testing the association between each SNP and its corresponding CpG, adjusting for age at methylation measurement, sex, and the first four genetic principal components, with family as a random effect. If the SNP and CpG were associated at p < 0.05, we adjusted out the effect of the SNP from the CpG site and re-tested the interaction (Model 4).

2.3.4. Gene Expression Analysis

Among the 494 participants with methylation and genetic data, 429 participants also had gene expression data. Figure S3 presents a graphical depiction of ABCA7 transcripts observed in the Genotype Tissue Expression (GTEx) project [57], which assesses gene expression levels in a variety of cell types. A total of 17 transcripts, along with a measure of overall ABCA7 gene expression, were available for analysis in our study. For SNPs, CpGs, or interactions that were significantly associated with general cognitive function, we assessed their association with ABCA7 gene-level and transcript-level expression (Model 5) using linear mixed models. Model 5 adjusted for age at which gene expression data was generated (age at blood draw), sex, first four genetic principal components, and family as a random effect. For models that included CpG sites, Model 5 also included the age difference between methylation and gene expression measurements. Similarly, for any significant interaction effects, contrast tests were conducted to obtain the effect size in each genotype group. Minor homozygote genotype groups (<5% sample size) were grouped with heterozygous genotype groups to increase power as appropriate.
We next evaluated whether the identified CpG sites within the ABCA7 region correlate with gene expression of ABCA7 and/or nearby genes in an external public database with multiple cells/tissues. For this, we used cis- expression quantitative trait methylation (cis-eQTM) results from peripheral blood mononuclear cells (PBMCs) and three specific white blood cell types (CD4 + T lymphocytes, monocytes, and neutrophils) in the iMETHYL database [58,59], which integrated genotype, methylation, and gene expression data from 102 individuals. We also examined gene expression levels of ABCA7 in different cell types available from the Genotype Tissue Expression (GTEx) project [57].

2.3.5. Multiple Testing Correction

All statistical analyses were conducted using R (Version 3.6) [60]. For genetic analysis, the Bonferroni corrected p-value cut-off (p < 0.05/5) was used to claim significance. For all other analyses, false discovery rate (FDR) correction was applied to each model, and FDR q < 0.1 was considered significant. Since the SNPs, CpG sites, and transcripts in ABCA7 were all correlated, applying stringent multiple testing corrections might have been too conservative; thus, any nominal associations were also noted.

3. Results

3.1. Sample Characteristics

The sample included 634 AA without dementia (Table 1). Overall, participant age ranged from 45 to 85 years (mean = 63.3 years), and the mean age difference between Phase I methylation and cognitive measurements was 6.0 years (SD = 1.3). More than half of participants (74.9%) were female, and 47.3% had at least some college education. General cognitive function was normally distributed. Mean RAVLT score was 7.1 (SD = 3.3) words recalled, mean DSST score was 34.4 (SD = 12.6) symbols, mean COWA-FAS score was 29.7 (SD = 11.6) words, mean SCWT score was 22.5 (SD = 9.8) items, and mean TMTA score was 61.6 (SD = 32.0) seconds to completion.

3.2. Correlation among Six Cognitive Outcomes

Pearson correlations (r) among the six cognitive outcomes (general cognitive function and the five individual neurocognitive measures) are shown in Table S1. The five neurocognitive measures were moderately correlated (Pearson r ranged from 0.24 to 0.66), with the highest correlation between DSST and TMTA (r = 0.66, p < 0.001).

3.3. Correlation among ABCA7 SNPs

Pearson correlations among the five sentinel ABCA7 SNPs are shown in Table S2. Rs3764647 was strongly correlated with rs3764650 (r = 0.84, p < 0.001), and rs3752246 was highly correlated with rs4147929 (r = 0.96, p < 0.001). The other sentinel SNP pairs were only weakly correlated or uncorrelated.

3.4. Genetic Associations

In Models 1 and 2, there were no ABCA7 SNPs that met the nominal significance threshold (p < 0.05, Table S3). Although APOE was not part of the primary analysis, APOE ε2 and ε4 were analyzed separately as exposures in Models 1 and 2. APOE ε4 was associated with general cognitive function in both models in the expected direction (higher dosage of ε4 was associated with lower cognitive function). After adjusting for educational attainment and APOE ε2 and ε4 in Model 3, sentinel SNPs remained unassociated with general cognitive function. There were no observed nominal or significant interactions between SNPs and sex, APOE isoforms, or educational attainment on general cognitive function.

3.5. Epigenetic Associations

Among the 72 CpG sites examined, six were nominally associated with general cognitive function in at least one of the three Models (Table S4). After adjusting for educational attainment and APOE ε2 and ε4 (Model 3), five CpGs (cg22271697, cg00874873, cg11714200, cg26264438 and cg12082025) in the ABCA7 region were nominally associated with general cognitive function. Figure 1 illustrates the regional plot of association p-values of the 72 CpGs in the ABCA7 region with general cognitive function according to the chromosomal positions of CpG sites, as well as the correlations between the CpGs (Model 3).

3.6. Genetic-Epigenetic Interactions

Since rs3764647 and rs3764650, as well as rs4147929 and rs3752246, are highly correlated with each other (Table S2), we removed one SNP from each pair. We selected rs3764647 because it had stronger evidence of association with AD in AA than rs3764650 [8]. We selected rs3752246 because it was a missense variant and more likely to have a functional effect than rs4147929, which is intronic [61]. Thus, we analyzed three independent risk SNPs (|r| < 0.60) in the interaction analysis. We assessed the interaction between each of the three independent sentinel SNPs (rs3764647, rs115550680, and rs3752246) and 72 CpG sites on general cognitive function, and identified four significant SNP-by-CpG interactions (FDR q < 0.1) that were associated with general cognitive function (Table 2): rs3764647*cg00135882 (p = 1.46 × 10−4), rs3764647*cg22271697 (p = 5.77 × 10−4), rs115550680*cg06169110 (p = 2.18 × 10−4), and rs115550680*cg17316918 (p = 4.84 × 10−4). The two SNPs and four CpGs that were involved in the four significant SNP-by-CpG interactions are shown in Figure 1 to highlight their positions with respect to neighboring genes, regulatory elements, and CGIs in the ABCA7 region. All interactions with at least nominal significance are shown in Table S5. Notably, an additional seven CpG sites had nominally significant interactions with rs115550680, and one additional site had a nominally significant interaction with rs3764647. In Table S6, we present Pearson correlations among the ABCA7 CpG sites that were nominally associated with general cognitive function (Table S4) and/or were involved in an FDR-significant SNP-by-CpG interaction (Table 2). The majority of these CpGs were weakly correlated or uncorrelated.
For interactions with FDR q < 0.1, we performed contrast tests to estimate the effect size of the specific CpG site per genotype group. In all four cases, the minor homozygote genotype group had a small frequency (<5% of the sample size); thus, we combined them with the corresponding heterozygote genotype group. Contrast tests showed that methylation is associated with general cognitive function in some genotype groups, but not others (p < 0.05; Table 3 and Figure 2).
Rs3764647 had significant interactions with two CpGs (cg00135882 and cg22271697). For those with the risk genotype (GG/AG), a 1% increase at cg00135882 was associated with a 0.68 SD decrease in general cognitive function (p = 0.004, Figure 2A); whereas for those with the AA genotype, a 1% increase at cg22271697 was associated with a 0.14 SD increase in general cognitive function (p = 2.00 × 10−4, Figure 2B). Similarly, rs115550680 had interactions with two CpGs (cg06169110 and cg17316918). For those with the risk genotype (GG/AG), a 1% increase at cg06169110 was associated with a 0.37 SD decrease in general cognitive function (p = 2.00 × 10−4, Figure 2C), and a 1% increase at cg17316918 was associated with a 0.33 SD increase in general cognitive function (p = 0.004, Figure 2D).
We performed a sensitivity analysis by excluding outlying CpG values beyond four standard deviations of mean methylation, and our results remained consistent (Table S7). To test whether the interaction was driven by potential SNP-CpG correlation, we assessed the association between each SNP-CpG pair. We observed nominal associations between rs3764647 and cg22271697, as well as between rs115550680 and cg06169110. For these two SNP-CpG pairs, we regressed out the SNP effect from the corresponding CpGs and re-tested the interactions. The results remained consistent with those reported in Table 3 (Table S8). We also tested the association between all four significant interactions with each of the five neurocognitive domains. Similar interactions were observed for multiple neurocognitive measures, especially DSST and SCWT, in which all four interactions were significantly associated (Table S9).

3.7. Gene Expression Associations

To understand the functional effects of identified SNP-by-CpG interactions, we examined their interaction effects (Tables S10 and S11) as well as marginal effects (Tables S12 and S13) on ABCA7 gene and transcript expression. At the gene level, none of the identified SNP-by-CpG interactions were associated with gene expression in our sample. However, we found a negative association between one of the SNPs, rs115550680, and gene level expression of ABCA7 (ENSG00000064687): for each additional rs115550680 G allele, there was a 0.05 decrease in gene expression (p = 0.027).
At the transcript level, two SNP-by-CpG interactions (rs115550680*cg17316918 and rs3764647*cg22271697) were nominally associated with two different transcripts (ENST00000525939 and ENST00000531467) (Table S10). ENST00000531467 (Chromosome 19: 1,062,261–1,063,945 forward strand) is a protein coding transcript with four coding exons (Figure S3). ENST00000525939 (Chromosome 19: 1,062,261–1,063,945 forward strand) is a retained intron, found primarily in the spleen, pituitary, whole blood, and brain (cerebellum and cerebellar hemisphere) (Figure S3). Although the interactions were only nominally significant, we performed contrast tests to estimate the effect size of the CpG site in each genotype group on each identified transcript. Contrast tests showed that methylation at cg17316918 trended toward a positive association with ENST00000525939 among those with the rs115550680 risk genotype (GG/AG) but did not reach nominal significance (Table S11). We also assessed the marginal associations of the two SNPs and two CpGs involved in the interactions on each of the ABCA7 transcripts (Tables S12 and S13). We found that rs115550680 was negatively associated with 11 ABCA7 transcripts, including ENST00000531467, at FDR q < 0.1 (Table S12). Rs3764647 was nominally associated with only ENST00000530703 (p = 0.037; Table S12). Among the CpGs involved in the interactions, cg06169110 was nominally associated with two transcripts (Table S13).
The iMETHYL [62] cis-eQTM results for PBMCs and the three white blood cell types showed that there were CpGs within the ABCA7 region, including within the promoter region, that regulate expression of both ABCA7 and nearby genes. However, the CpGs identified in the significant SNP-by-CpG interactions in our study were not associated with gene expression of ABCA7 or nearby genes at FDR q < 0.05.

4. Discussion

While previous studies have implied that ABCA7 is a causal gene for AD [63,64,65,66], there is a dearth of studies examining the relationship between ABCA7 and cognitive function. AD is a gradual neurodegenerative disease, characterized by noticeable cognitive impairment in areas of episodic memory, semantic memory, and executive function, with pathophysiology preceding the illness decades prior [67,68]. Studying the relationship between SNPs and CpGs in ABCA7 and cognition may enhance our understanding of cognitive health and further elucidate the role of ABCA7 in cognitive aging preceding AD. To our knowledge, this study is the first assessment of the association and interaction between DNA methylation and genetic risk factors in ABCA7 on cognition in AA without dementia.
In this study, we found no association between known AD-associated SNPs and cognitive measures. This is, perhaps, not surprising, as previous studies have been inconsistent regarding the association between ABCA7 SNPs and cognition. Most of the studies, however, have been conducted in primarily European ancestry populations [27,28,29,69]. For example, the Three-City Dijon study found no association between ABCA7 common variants and global cognition, nor other cognitive outcomes [69]. Other studies in EA have shown that SNPs may be associated with cognition in subgroups stratified on gender [27], APOE status [28], or disease progression [29]. In light of this, we also assessed whether ABCA7 SNP associations are modified by sex, APOE major isoforms, and/or education status. Unlike prior studies [27,28], we did not find any evidence of interaction. Lack of association with cognition for the sentinel SNP-by-sex and SNP-by-APOE interactions may be due to differences in ancestry or to small sample size, as those studies have sample sizes ranging from 1153 to 3267 [39,40]. Our study also did not find SNP-by-education associations interactions on cognition. This is consistent with another study, which observed no interaction between education and ABCA7 variants on memory performance in either EA or AA; however, a weak signal was observed for memory decline in AA, which is a cognitive measure more closely related to AD and dementia than general cognitive function [70].
Other lines of evidence also suggest that the ABCA7 risk variants may not be highly relevant to the neurological pathways underlying normal cognitive function and/or cognitive reserve. For example, previous GWAS for general cognitive function and AD have shown few overlapping loci [35,71]. Further, studies of cognitively “resilient” individuals who live to an older age with intact cognitive function, despite the presence of AD neuropathology, have found the genetic architecture of cognitive resilience to be distinct from that of AD [72]. At this point, relatively little is known about the pathways involving genetic variants and cognitive aging in those without dementia. Thus, studying variants that affect general cognitive function before development of dementia may identify novel pathways for therapeutic targets.
Only one epigenome-wide association study (EWAS) has examined the association between all CpG sites across the genome, including CpGs in ABCA7 gene, and general cognitive function in participants from multi-ethnic backgrounds [73]. This study did not identify any significant associations between ABCA7 and general cognitive function. However, due to the large numbers of CpG sites tested, the EWAS could have missed signals with smaller effect sizes. Moreover, the EWAS sample was mostly composed of EA. Our study, which focused on CpG sites in ABCA7 in an AA cohort, would give us more power to detect an association in this region among AA. Nevertheless, we also failed to detect any associations between CpGs and general cognitive function after multiple testing correction, although six CpGs were associated at a nominal level. Importantly, we examined methylation levels in whole blood leukocytes, which is not the most relevant tissue for brain function. A study in post-mortem brain tissue found associations between CpGs in ABCA7 and AD, as well as an increased burden of pathologies (e.g., Aβ load and tau tangle density), whereas another study failed to demonstrate differential methylation in peripheral blood between AD patients and controls [21]. Although methylation patterns differ between blood and brain tissues [23,74], blood cells touch every cell bed that affects the brain, and are related to chronic inflammation and oxidative stress, which are linked to cognitive performance [75,76]. Studying methylation in blood also allows us to study epigenetic associations with cognition in living participants in an inexpensive and non-invasive manner.
Although ABCA7 sentinel SNPs and CpG sites were not associated with general cognitive function, we did find evidence of SNP-by-CpG interactions. Four interactions reached FDR significance (rs3764647*cg00135882, rs3764647*cg22271697, rs115550680*cg06169110, and rs115550680*cg17316918). Further, a total of nine CpG sites had at least nominally significant interactions with rs115550680 on cognition function. For participants who are homozygous for the rs115550680 major allele (AA), local methylation does not seem to have an effect on cognitive function. However, for participants who carry the risk allele (GG/AG), methylation at local CpGs may play an important role in cognition. This might be related to the different ABCA7 transcripts that are involved in each case. Rs115550680 is located in an LD block that spans several introns and exons [8]. A prior study suggested that there is a 44-base pair exonic deletion (rs142076058, p.Arg578 fs) among rs115550680 G carriers, which could cause a frameshift in the ABCA7-coding sequence, resulting in the formation of a premature termination codon [77]. Indeed, our gene expression analysis found that the risk allele (G) at rs115550680 was strongly associated with decreased expression of 11 ABCA7 transcripts. Taken together, these data suggest that this SNP might influence the major isoforms that are expressed, and the expressed alternative transcripts may influence cognitive function. Furthermore, alternative transcripts that are expressed in those carrying the risk allele may be further modulated by methylation level at local CpG sites, which may lead to differences in cognitive function in this group. Consistent with this hypothesis, methylation at cg17316918 was associated with transcript ENST00000525939 in rs115550680 risk allele carriers (GG/AG) only. Interestingly, this transcript is largely expressed in the brain. However, there is no prior evidence to show an association between this transcript and AD and/or cognition. Nonetheless, alternative splicing of ABCA7 is likely to play a similar important role in cognition, as has been demonstrated in AD [78,79].
The other SNP that had significant interactions with ABCA7 CpG sites, rs3764647, is a missense mutation, where the risk allele (G) leads to the amino acid change p.His395Arg in the first extracellular loop of the ABCA7 protein [16]. One CpG site (cg00135882) is associated with cognitive function in participants who carry the risk allele (GG/AG), and another CpG site (cg22271697) is associated with cognitive function in those who do not carry the risk allele (AA). This differential pattern may be due to different functions of the two transcripts instead of alternative splicing. Consistently, we did not observe a direct association between this SNP or CpG with expression of ABCA7 transcripts. Notably, three of the CpGs (cg00135882, cg22271697, and cg06169110) in the significant SNP-by-CpG interactions were either flanking or within CGIs. Active intragenic CGIs may change the major isoforms that are expressed by interfering with splicing and/or polyadenylation. Alternatively, they may promote enhancer function or act directly as an enhancer to regulate gene expression [19]. Consistent with this hypothesis, all four CpGs are located in regions that contain at least one important regulatory element (i.e., promoters, enhancers, and/or CTCF binding sites). Taken together, these results suggest that SNPs and CpG sites in ABCA7 may interact to modulate the expression and/or function of ABCA7 transcripts, and that some of the affected transcripts may influence cognitive function in older AA.
Indeed, recent literature suggests that SNP-by-CpG interactions might be an important mechanism underlying human complex diseases [80,81,82]. Similar SNP-by-CpG interactions have been identified in association with complex human disorders, such as breast cancer [83], type 2 diabetes [84], alcohol dependence [85], and suicide attempts in schizophrenia [86]. One factor to note, however, is that SNPs could have a cis-regulatory effect on local CpGs, which could cause a spurious interaction. However, our sensitivity analysis demonstrated that the interactions which we observed were not solely due to SNP-CpG correlations. In summary, we have demonstrated that a complicated interplay between genetic and epigenetic risk factors in the ABCA7 region may play an important role in cognitive function. Future studies are needed to disentangle this complicated relationship.
Our study is not without limitations. First, our gene expression measures were taken from transformed B-lymphocytes from immortalized cell lines. While transformed B-lymphocytes are a convenient source of DNA, the transformation process causes epigenetic changes to the immortalized cells that are not fully understood [87]. However, they provide a unique and efficient way to examine the functional effects of genetic and epigenetic variation on gene expression, since the environmental conditions of the cells are the same across individuals. In addition, previous cis-eQTM studies in white blood cells have shown that at least some CpGs within the ABCA7 region promote or repress gene expression of ABCA7 and nearby genes, but we did not observe eQTM relationships with those same CpGs in our study. One reason for this may be that our methylation was measured in blood and included a mix of white blood cells, while our gene expression was measured in transformed B-lymphocytes. Additional work is needed to understand how ABCA7 CpGs and their interactions with SNPs influence proximal gene expression in a variety of white blood cell types, which would further shed light on the complicated biological mechanisms that contribute to cognitive function. We also acknowledge that our findings need to be replicated in a larger sample of AA. Further studies in animal and cellular models are also warranted to confirm our findings and to reveal how SNPs and methylation jointly contribute to cognitive function. Finally, due to the cross-sectional nature of our study, we cannot infer causality of our findings or quantify how the SNP-by-CpG interactions alone impact cognition. To that end, longitudinal studies are necessary to investigate how SNPs and/or CpGs affect cognitive changes over time.
Our study also has notable strengths. To our knowledge, our study is the first to take a multi-omic approach to investigate the relations between the ABCA7 gene region and cognitive function in a population-based cohort of older adults without diagnosed dementia. Our study was also conducted on AA, an understudied population with a higher prevalence of AD [3,5] and higher conferred risk of AD from ABCA7, compared to EA [8]. Additionally, with comprehensive cognition measures, we were able to assess associations with multiple neurocognitive domains, as well as general cognitive function.

5. Conclusions

In the present study, we evaluated the association between ABCA7 genetic, epigenetic, and transcriptomic markers and cognitive function in 634 AA participants without preliminary evidence of dementia. We found that DNA methylation levels at local CpG sites modify the relationship between genetic variants and general cognitive function. Specifically, two SNPs in the ABCA7 gene region (rs3764647 and rs115550680) may regulate the effects of methylation on cognition. Differential gene expression analysis further highlighted the potentially causal transcripts. In conclusion, our findings suggest that a complicated interplay between genetic and epigenetic factors in ABCA7 may influence cognition in older AA without dementia.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/genes13112150/s1, Figure S1: Flow diagram illustrating sample sizes for genetic (n = 634), epigenetic (n = 494), and transcriptomic (n = 429) analyses in GENOA AA; Figure S2: Models used to assess genetic, epigenetic and genetic-epigenetic interaction associations with general cognitive function; Figure S3: Transcript expression of ABCA7: ENSG00000064687 (12 ATP binding cassette subfamily A member 7 [Source: HGNC Symbol; Acc:HGNC:37]); Table S1: Pearson’s correlations among the six cognitive measures (n = 634); Table S2: Pearson’s correlations among the five sentinel ABCA7 SNPs (n = 634); Table S3. Association between ABCA7 sentinel SNPs and general cognitive function (n = 634); Table S4. Association of CpGs in the ABCA7 region and general cognitive function (p < 0.05; n = 494); Table S5. Interaction between ABCA7 sentinel SNPs and CpG sites on general cognitive function (p < 0.05; n = 494); Table S6. Pearson’s correlations among ABCA7 CpG sites (n = 494); Table S7. Estimated effect of CpG site on general cognitive function for given ABCA7 SNP genotype group, after excluding outlying values for CpG sites; Table S8. Estimated effect of CpG site on general cognitive function for given ABCA7 SNP genotype group, after adjusting for SNP effect; Table S9: Interaction between ABCA7 sentinel SNPs and CpG sites on neurocognitive measurements (n = 494); Table S10. Interaction between ABCA7 sentinel SNPs and CpG sites on transcripts in the ABCA7 gene region (p < 0.05; n = 429); Table S11. Estimated effect of CpG site on ABCA7 transcripts for given ABCA7 SNP genotype group (n = 429); Table S12. Association of SNPs on transcripts in the ABCA7 gene region (p < 0.05; n = 429); Table S13. Association of CpG sites on transcripts in the ABCA7 region (p < 0.05; n = 429).

Author Contributions

Conceptualization, D.L.C., K.N., Y.-Z.W., S.M.R., T.H.M., S.L.R.K., J.A.S. and W.Z.; methodology, D.L.C., K.N., J.A.S. and W.Z.; software, formal analysis, and investigation, D.L.C., K.N., S.M.R. and W.Z.; resources, T.H.M. and S.L.R.K.; writing—original draft preparation, D.L.C., K.N., J.A.S. and W.Z.; writing—review and editing, D.L.C., J.A.S. and W.Z.; visualization, D.L.C.; supervision, J.A.S. and W.Z.; funding acquisition, S.L.R.K. and J.A.S. All authors have read and agreed to the published version of the manuscript.

Funding

Support for the Genetic Epidemiology Network of Arteriopathy (GENOA) was provided by the National Heart, Lung and Blood Institute (NHLBI, U01HL054457, RC1HL100185, R01HL087660, R01HL119443, R01HL133221) and the National Institute of Neurological Disorders and Stroke (NINDS, R01NS041558) of the NIH.

Informed Consent Statement

Written informed consent was obtained from all subjects involved in the study.

Data Availability Statement

The phenotype data and APOE genotypes used in the current study are available upon reasonable request to J.A.S. and S.L.R.K., and with a completed data use agreement (DUA). All other genotype data are available from the Database of Genotypes and Phenotypes (dbGaP): phs001401.v2.p1. Methylation and gene expression data are available from the Gene Expression Omnibus (GEO): GSE210256 and GSE138914. Due to IRB restriction, mapping of the sample IDs between genotype data (dbGaP) and methylation data (GEO) cannot be provided publicly, but is available upon written request to J.A.S. and S.L.R.K.

Acknowledgments

The authors wish to thank the staff and participants of the GENOA study.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Hampel, H.; Hardy, J.; Blennow, K.; Chen, C.; Perry, G.; Kim, S.H.; Villemagne, V.L.; Aisen, P.; Vendruscolo, M.; Iwatsubo, T.; et al. The Amyloid-β Pathway in Alzheimer’s Disease. Mol. Psychiatry 2021, 26, 5481–5503. [Google Scholar] [CrossRef] [PubMed]
  2. Grundke-Iqbal, I.; Iqbal, K.; Tung, Y.C.; Quinlan, M.; Wisniewski, H.M.; Binder, L.I. Abnormal Phosphorylation of the Microtubule-Associated Protein Tau (Tau) in Alzheimer Cytoskeletal Pathology. Proc. Natl. Acad. Sci. USA 1986, 83, 4913–4917. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  3. 2021 Alzheimer’s Disease Facts and Figures. Alzheimers Dement. 2021, 17, 327–406. [CrossRef]
  4. Barnes, L.L.; Bennett, D.A. Alzheimer’s Disease in African Americans: Risk Factors and Challenges for the Future. Health Aff. 2014, 33, 580–586. [Google Scholar] [CrossRef] [PubMed]
  5. Tang, M.X.; Cross, P.; Andrews, H.; Jacobs, D.M.; Small, S.; Bell, K.; Merchant, C.; Lantigua, R.; Costa, R.; Stern, Y.; et al. Incidence of AD in African-Americans, Caribbean Hispanics, and Caucasians in Northern Manhattan. Neurology 2001, 56, 49–56. [Google Scholar] [CrossRef] [Green Version]
  6. Ridge, P.G.; Mukherjee, S.; Crane, P.K.; Kauwe, J.S.K.; Consortium, A.D.G. Alzheimer’s Disease: Analyzing the Missing Heritability. PLoS ONE 2013, 8, e79771. [Google Scholar] [CrossRef] [Green Version]
  7. Migliore, L.; Coppedè, F. Genetics, Environmental Factors and the Emerging Role of Epigenetics in Neurodegenerative Diseases. Mutat. Res. 2009, 667, 82–97. [Google Scholar] [CrossRef]
  8. Reitz, C.; Jun, G.; Naj, A.; Rajbhandary, R.; Vardarajan, B.N.; Wang, L.-S.; Valladares, O.; Lin, C.-F.; Larson, E.B.; Graff-Radford, N.R.; et al. Variants in the ATP-Binding Cassette Transporter (ABCA7), Apolipoprotein E Ε4,and the Risk of Late-Onset Alzheimer Disease in African Americans. JAMA 2013, 309, 1483–1492. [Google Scholar] [CrossRef]
  9. Strittmatter, W.J.; Saunders, A.M.; Schmechel, D.; Pericak-Vance, M.; Enghild, J.; Salvesen, G.S.; Roses, A.D. Apolipoprotein E: High-Avidity Binding to Beta-Amyloid and Increased Frequency of Type 4 Allele in Late-Onset Familial Alzheimer Disease. Proc. Natl. Acad. Sci. USA 1993, 90, 1977–1981. [Google Scholar] [CrossRef] [Green Version]
  10. Strittmatter, W.J.; Roses, A.D. Apolipoprotein E and Alzheimer Disease. Proc. Natl. Acad. Sci. USA 1995, 92, 4725–4727. [Google Scholar] [CrossRef]
  11. Corder, E.H.; Saunders, A.M.; Strittmatter, W.J.; Schmechel, D.E.; Gaskell, P.C.; Small, G.W.; Roses, A.D.; Haines, J.L.; Pericak-Vance, M.A. Gene Dose of Apolipoprotein E Type 4 Allele and the Risk of Alzheimer’s Disease in Late Onset Families. Science 1993, 261, 921–923. [Google Scholar] [CrossRef]
  12. Lamartinière, Y.; Boucau, M.-C.; Dehouck, L.; Krohn, M.; Pahnke, J.; Candela, P.; Gosselet, F.; Fenart, L. ABCA7 Downregulation Modifies Cellular Cholesterol Homeostasis and Decreases Amyloid-β Peptide Efflux in an In Vitro Model of the Blood-Brain Barrier. J. Alzheimers Dis. 2018, 64, 1195–1211. [Google Scholar] [CrossRef]
  13. Aikawa, T.; Holm, M.-L.; Kanekiyo, T. ABCA7 and Pathogenic Pathways of Alzheimer’s Disease. Brain Sci. 2018, 8, 27. [Google Scholar] [CrossRef] [Green Version]
  14. Zhao, Q.-F.; Yu, J.-T.; Tan, M.-S.; Tan, L. ABCA7 in Alzheimer’s Disease. Mol. Neurobiol. 2015, 51, 1008–1016. [Google Scholar] [CrossRef]
  15. Wang, N.; Lan, D.; Gerbod-Giannone, M.; Linsel-Nitschke, P.; Jehle, A.W.; Chen, W.; Martinez, L.O.; Tall, A.R. ATP-Binding Cassette Transporter A7 (ABCA7) Binds Apolipoprotein A-I and Mediates Cellular Phospholipid but Not Cholesterol Efflux. J. Biol. Chem. 2003, 278, 42906–42912. [Google Scholar] [CrossRef] [Green Version]
  16. Logue, M.W.; Lancour, D.; Farrell, J.; Simkina, I.; Fallin, M.D.; Lunetta, K.L.; Farrer, L.A. Targeted Sequencing of Alzheimer Disease Genes in African Americans Implicates Novel Risk Variants. Front. Neurosci. 2018, 12, 592. [Google Scholar] [CrossRef]
  17. N’Songo, A.; Carrasquillo, M.M.; Wang, X.; Burgess, J.D.; Nguyen, T.; Asmann, Y.W.; Serie, D.J.; Younkin, S.G.; Allen, M.; Pedraza, O.; et al. African American Exome Sequencing Identifies Potential Risk Variants at Alzheimer Disease Loci. Neurol. Genet. 2017, 3, e141. [Google Scholar] [CrossRef] [Green Version]
  18. Jaenisch, R.; Bird, A. Epigenetic Regulation of Gene Expression: How the Genome Integrates Intrinsic and Environmental Signals. Nat. Genet. 2003, 33, 245–254. [Google Scholar] [CrossRef] [PubMed]
  19. Cain, J.A.; Montibus, B.; Oakey, R.J. Intragenic CpG Islands and Their Impact on Gene Regulation. Front. Cell Dev. Biol. 2022, 10, 832348. [Google Scholar] [CrossRef] [PubMed]
  20. Li, Q.S.; Sun, Y.; Wang, T. Epigenome-Wide Association Study of Alzheimer’s Disease Replicates 22 Differentially Methylated Positions and 30 Differentially Methylated Regions. Clin. Epigenet. 2020, 12, 149. [Google Scholar] [CrossRef]
  21. Yu, L.; Chibnik, L.B.; Srivastava, G.P.; Pochet, N.; Yang, J.; Xu, J.; Kozubek, J.; Obholzer, N.; Leurgans, S.E.; Schneider, J.A.; et al. Association of Brain DNA Methylation in SORL1, ABCA7, HLA-DRB5, SLC24A4, and BIN1 with Pathological Diagnosis of Alzheimer Disease. JAMA Neurol. 2015, 72, 15–24. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  22. De Jager, P.L.; Srivastava, G.; Lunnon, K.; Burgess, J.; Schalkwyk, L.C.; Yu, L.; Eaton, M.L.; Keenan, B.T.; Ernst, J.; McCabe, C.; et al. Alzheimer’s Disease: Early Alterations in Brain DNA Methylation at ANK1, BIN1, RHBDF2 and Other Loci. Nat. Neurosci. 2014, 17, 1156–1163. [Google Scholar] [CrossRef] [Green Version]
  23. Lunnon, K.; Smith, R.; Hannon, E.; De Jager, P.L.; Srivastava, G.; Volta, M.; Troakes, C.; Al-Sarraj, S.; Burrage, J.; Macdonald, R.; et al. Methylomic Profiling Implicates Cortical Deregulation of ANK1 in Alzheimer’s Disease. Nat. Neurosci. 2014, 17, 1164–1170. [Google Scholar] [CrossRef] [Green Version]
  24. Yamazaki, K.; Yoshino, Y.; Mori, T.; Yoshida, T.; Ozaki, Y.; Sao, T.; Mori, Y.; Ochi, S.; Iga, J.; Ueno, S. Gene Expression and Methylation Analysis of ABCA7 in Patients with Alzheimer’s Disease. J. Alzheimers Dis. 2017, 57, 171–181. [Google Scholar] [CrossRef]
  25. Apostolova, L.G.; Risacher, S.L.; Duran, T.; Stage, E.C.; Goukasian, N.; West, J.D.; Do, T.M.; Grotts, J.; Wilhalme, H.; Nho, K.; et al. Associations of the Top 20 Alzheimer Disease Risk Variants with Brain Amyloidosis. JAMA Neurol. 2018, 75, 328–341. [Google Scholar] [CrossRef]
  26. De Roeck, A.; Van Broeckhoven, C.; Sleegers, K. The Role of ABCA7 in Alzheimer’s Disease: Evidence from Genomics, Transcriptomics and Methylomics. Acta Neuropathol. 2019, 138, 201–220. [Google Scholar] [CrossRef] [Green Version]
  27. Nettiksimmons, J.; Tranah, G.; Evans, D.S.; Yokoyama, J.S.; Yaffe, K. Gene-Based Aggregate SNP Associations between Candidate AD Genes and Cognitive Decline. Age 2016, 38, 41. [Google Scholar] [CrossRef] [Green Version]
  28. Engelman, C.D.; Koscik, R.L.; Jonaitis, E.M.; Okonkwo, O.C.; Hermann, B.P.; La Rue, A.; Sager, M.A. Interaction between Two Cholesterol Metabolism Genes Influences Memory: Findings from the Wisconsin Registry for Alzheimer’s Prevention. J. Alzheimers Dis. 2013, 36, 749–757. [Google Scholar] [CrossRef]
  29. Carrasquillo, M.M.; Crook, J.E.; Pedraza, O.; Thomas, C.S.; Pankratz, V.S.; Allen, M.; Nguyen, T.; Malphrus, K.G.; Ma, L.; Bisceglio, G.D.; et al. Late-Onset Alzheimer’s Risk Variants in Memory Decline, Incident Mild Cognitive Impairment, and Alzheimer’s Disease. Neurobiol. Aging 2015, 36, 60–67. [Google Scholar] [CrossRef] [Green Version]
  30. Cacabelos, R.; Torrellas, C. Epigenetics of Aging and Alzheimer’s Disease: Implications for Pharmacogenomics and Drug Response. Int. J. Mol. Sci. 2015, 16, 30483–30543. [Google Scholar] [CrossRef]
  31. Daniels, P.R.; Kardia, S.L.R.; Hanis, C.L.; Brown, C.A.; Hutchinson, R.; Boerwinkle, E.; Turner, S.T.; Genetic Epidemiology Network of Arteriopathy study. Familial Aggregation of Hypertension Treatment and Control in the Genetic Epidemiology Network of Arteriopathy (GENOA) Study. Am. J. Med. 2004, 116, 676–681. [Google Scholar] [CrossRef] [PubMed]
  32. Lezak, M.D.; Howieson, D.B.; Loring, D.W.; Hannay, H.; Fischer, J. Neuropsychological Assessment; Oxford University Press: New York, NY, USA, 2004; ISBN 978-0-19-511121-7. [Google Scholar]
  33. Smith, J.A.; Mosley, T.H., Jr.; Turner, S.T.; Kardia, S.L. Shared Genetic Effects among Measures of Cognitive Function and Leukoaraiosis. In Brain Injury–Pathogenesis, Monitoring, Recovery and Management; Agrawal, A., Ed.; Intechopen: London, UK, 2012; p. 39. [Google Scholar]
  34. Jaeger, J. Digit Symbol Substitution Test. J. Clin. Psychopharmacol. 2018, 38, 513–519. [Google Scholar] [CrossRef] [PubMed]
  35. Davies, G.; Armstrong, N.; Bis, J.C.; Bressler, J.; Chouraki, V.; Giddaluru, S.; Hofer, E.; Ibrahim-Verbaas, C.A.; Kirin, M.; Lahti, J.; et al. Genetic Contributions to Variation in General Cognitive Function: A Meta-Analysis of Genome-Wide Association Studies in the CHARGE Consortium (N = 53 949). Mol. Psychiatry 2015, 20, 183–192. [Google Scholar] [CrossRef] [Green Version]
  36. Gao, X.; Jia, M.; Zhang, Y.; Breitling, L.P.; Brenner, H. DNA Methylation Changes of Whole Blood Cells in Response to Active Smoking Exposure in Adults: A Systematic Review of DNA Methylation Studies. Clin. Epigenet. 2015, 7, 113. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  37. Fortin, J.-P.; Fertig, E.; Hansen, K. ShinyMethyl: Interactive Quality Control of Illumina 450k DNA Methylation Arrays in R. F1000 Res. 2014, 3, 175. [Google Scholar] [CrossRef]
  38. Xu, Z.; Niu, L.; Li, L.; Taylor, J.A. ENmix: A Novel Background Correction Method for Illumina HumanMethylation450 BeadChip. Nucleic Acids Res. 2016, 44, e20. [Google Scholar] [CrossRef]
  39. Fortin, J.-P.; Triche, T.J., Jr.; Hansen, K.D. Preprocessing, Normalization and Integration of the Illumina HumanMethylationEPIC Array with Minfi. Bioinformatics 2017, 33, 558–560. [Google Scholar] [CrossRef] [Green Version]
  40. Aryee, M.J.; Jaffe, A.E.; Corrada-Bravo, H.; Ladd-Acosta, C.; Feinberg, A.P.; Hansen, K.D.; Irizarry, R.A. Minfi: A Flexible and Comprehensive Bioconductor Package for the Analysis of Infinium DNA Methylation Microarrays. Bioinformatics 2014, 30, 1363–1369. [Google Scholar] [CrossRef] [Green Version]
  41. Niu, L.; Xu, Z.; Taylor, J.A. RCP: A Novel Probe Design Bias Correction Method for Illumina Methylation BeadChip. Bioinformatics 2016, 32, 2659–2663. [Google Scholar] [CrossRef] [Green Version]
  42. Lehne, B.; Drong, A.W.; Loh, M.; Zhang, W.; Scott, W.R.; Tan, S.-T.; Afzal, U.; Scott, J.; Jarvelin, M.-R.; Elliott, P.; et al. A Coherent Approach for Analysis of the Illumina HumanMethylation450 BeadChip Improves Data Quality and Performance in Epigenome-Wide Association Studies. Genome Biol. 2015, 16, 37. [Google Scholar] [CrossRef]
  43. Hansen, K. IlluminaHumanMethylationEPICanno.ilm10b2.hg19: Annotation for Illumina’s EPIC Methylation Arrays, R package version 0.6.0; BioConductor: Boston, MA, USA, 2016. [Google Scholar]
  44. Houseman, E.A.; Accomando, W.P.; Koestler, D.C.; Christensen, B.C.; Marsit, C.J.; Nelson, H.H.; Wiencke, J.K.; Kelsey, K.T. DNA Methylation Arrays as Surrogate Measures of Cell Mixture Distribution. BMC Bioinform. 2012, 13, 86. [Google Scholar] [CrossRef] [Green Version]
  45. Du, P.; Zhang, X.; Huang, C.-C.; Jafari, N.; Kibbe, W.A.; Hou, L.; Lin, S.M. Comparison of Beta-Value and M-Value Methods for Quantifying Methylation Levels by Microarray Analysis. BMC Bioinform. 2010, 11, 587. [Google Scholar] [CrossRef] [Green Version]
  46. Weisenberger, C.D.J.; Van den Berg, D.; Pan, F.; Berman, B.P.; Laird, P.W. Comprehensive DNA Methylation Analysis on the Illumina® Infinium® Assay Platform; Illumina: San Diego, CA, USA, 2008; p. 4. [Google Scholar]
  47. Lockstone, H.E. Exon Array Data Analysis Using Affymetrix Power Tools and R Statistical Software. Brief. Bioinform. 2011, 12, 634–644. [Google Scholar] [CrossRef] [Green Version]
  48. Irizarry, R.A.; Bolstad, B.M.; Collin, F.; Cope, L.M.; Hobbs, B.; Speed, T.P. Summaries of Affymetrix GeneChip Probe Level Data. Nucleic Acids Res. 2003, 31, e15. [Google Scholar] [CrossRef]
  49. Dai, M.; Wang, P.; Boyd, A.D.; Kostov, G.; Athey, B.; Jones, E.G.; Bunney, W.E.; Myers, R.M.; Speed, T.P.; Akil, H.; et al. Evolving Gene/Transcript Definitions Significantly Alter the Interpretation of GeneChip Data. Nucleic Acids Res. 2005, 33, e175. [Google Scholar] [CrossRef] [Green Version]
  50. Saha, A.; Battle, A. False Positives in Trans-EQTL and Co-Expression Analyses Arising from RNA-Sequencing Alignment Errors. F1000Research 2018, 7, 1860. [Google Scholar] [CrossRef] [Green Version]
  51. Johnson, W.E.; Li, C.; Rabinovic, A. Adjusting Batch Effects in Microarray Expression Data Using Empirical Bayes Methods. Biostatistics 2007, 8, 118–127. [Google Scholar] [CrossRef]
  52. Martin, T.C.; Hardiman, T.; Yet, I.; Tsai, P.-C.; Bell, J.T. CoMET: Visualisation of Regional Epigenome-Wide Association Scan (EWAS) Results and DNA Co-Methylation Patterns. BMC Bioinform. 2015, 16, 131. [Google Scholar] [CrossRef] [Green Version]
  53. BioRender. Available online: https://app.biorender.com/illustrations/634485ef72e6ef474cc65999 (accessed on 10 October 2022).
  54. Lenth, R.V.; Buerkner, P.; Herve, M.; Love, J.; Riebl, H.; Singmann, H. Emmeans: Estimated Marginal Means, Aka Least-Squares Means, R package version: 1.8.2; CRAN: Stamford, CT, USA, 2022. [Google Scholar]
  55. Fox, J.; Weisberg, S.; Price, B.; Friendly, M.; Hong, J.; Andersen, R.; Firth, D.; Taylor, S. ; R Core Team. Effects: Effect Displays for Linear, Generalized Linear, and Other Models, R package version4.2-2; CRAN: Stamford, CT, USA, 2022. [Google Scholar]
  56. Ggplot2 Package—RDocumentation. Available online: https://www.rdocumentation.org/packages/ggplot2/versions/3.3.5 (accessed on 17 December 2021).
  57. The Gtex Consortium. The GTEx Consortium Atlas of Genetic Regulatory Effects across Human Tissues. Science 2020, 369, 1318–1330. [Google Scholar] [CrossRef]
  58. Gene: ABCA7 (ENSG00000064687)-Summary–Homo_sapiens-Ensembl Genome Browser 107. Available online: https://useast.ensembl.org/Homo_sapiens/Gene/Summary?db=core;g=ENSG00000064687;r=19:1039997-1065572 (accessed on 14 October 2022).
  59. Komaki, S.; Shiwa, Y.; Furukawa, R.; Hachiya, T.; Ohmomo, H.; Otomo, R.; Satoh, M.; Hitomi, J.; Sobue, K.; Sasaki, M.; et al. IMETHYL: An Integrative Database of Human DNA Methylation, Gene Expression, and Genomic Variation. Hum. Genome Var. 2018, 5, 18008. [Google Scholar] [CrossRef]
  60. Hachiya, T.; Furukawa, R.; Shiwa, Y.; Ohmomo, H.; Ono, K.; Katsuoka, F.; Nagasaki, M.; Yasuda, J.; Fuse, N.; Kinoshita, K.; et al. Genome-Wide Identification of Inter-Individually Variable DNA Methylation Sites Improves the Efficacy of Epigenetic Association Studies. NPJ Genom. Med. 2017, 2, 11. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  61. The R Project for Statistical Computing. Available online: https://www.r-project.org/ (accessed on 14 June 2022).
  62. Naj, A.C.; Jun, G.; Beecham, G.W.; Wang, L.-S.; Vardarajan, B.N.; Buros, J.; Gallins, P.J.; Buxbaum, J.D.; Jarvik, G.P.; Crane, P.K.; et al. Common Variants in MS4A4/MS4A6E, CD2AP, CD33 and EPHA1 Are Associated with Late-Onset Alzheimer’s Disease. Nat. Genet. 2011, 43, 436–441. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  63. Vasquez, J.B.; Fardo, D.W.; Estus, S. ABCA7 Expression Is Associated with Alzheimer’s Disease Polymorphism and Disease Status. Neurosci. Lett. 2013, 556, 58–62. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  64. Steinberg, S.; Gene, D.; Stefansson, H.; Jonsson, T.; Johannsdottir, H.; Ingason, A.; Helgason, H.; Sulem, P.; Magnusson, O.T.; Gudjonsson, S.A.; et al. Loss-of-Function Variants in ABCA7 Confer Risk of Alzheimer’s Disease. Nat. Genet. 2015, 47, 445–447. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  65. De Roeck, A.; Duchateau, L.; Van Dongen, J.; Cacace, R.; Bjerke, M.; Van den Bossche, T.; Cras, P.; Vandenberghe, R.; De Deyn, P.P.; Engelborghs, S.; et al. An Intronic VNTR Affects Splicing of ABCA7 and Increases Risk of Alzheimer’s Disease. Acta Neuropathol. 2018, 135, 827–837. [Google Scholar] [CrossRef] [Green Version]
  66. De Roeck, A.; Van den Bossche, T.; van der Zee, J.; Verheijen, J.; De Coster, W.; Van Dongen, J.; Dillen, L.; Baradaran-Heravi, Y.; Heeman, B.; Sanchez-Valle, R.; et al. Deleterious ABCA7 Mutations and Transcript Rescue Mechanisms in Early Onset Alzheimer’s Disease. Acta Neuropathol. 2017, 134, 475–487. [Google Scholar] [CrossRef] [Green Version]
  67. Hof, P.R.; Glannakopoulos, P.; Bouras, C. The Neuropathological Changes Associated with Normal Brain Aging. Histol. Histopathol. 1996, 11, 1075–1088. [Google Scholar]
  68. Perl, D.P. Neuropathology of Alzheimer’s Disease. Mt. Sinai J. Med. 2010, 77, 32–42. [Google Scholar] [CrossRef]
  69. Vivot, A.; Glymour, M.M.; Tzourio, C.; Amouyel, P.; Chêne, G.; Dufouil, C. Association of Alzheimer’s Related Genotypes with Cognitive Decline in Multiple Domains: Results from the Three-City Dijon Study. Mol. Psychiatry 2015, 20, 1173–1178. [Google Scholar] [CrossRef] [Green Version]
  70. Smith, J.A.; Kho, M.; Zhao, W.; Yu, M.; Mitchell, C.; Faul, J.D. Genetic Effects and Gene-by-Education Interactions on Episodic Memory Performance and Decline in an Aging Population. Soc. Sci. Med. 2021, 271, 112039. [Google Scholar] [CrossRef]
  71. Davies, G.; Lam, M.; Harris, S.E.; Trampush, J.W.; Luciano, M.; Hill, W.D.; Hagenaars, S.P.; Ritchie, S.J.; Marioni, R.E.; Fawns-Ritchie, C.; et al. Study of 300,486 Individuals Identifies 148 Independent Genetic Loci Influencing General Cognitive Function. Nat. Commun. 2018, 9, 2098. [Google Scholar] [CrossRef]
  72. Dumitrescu, L.; Mahoney, E.R.; Mukherjee, S.; Lee, M.L.; Bush, W.S.; Engelman, C.D.; Lu, Q.; Fardo, D.W.; Trittschuh, E.H.; Mez, J.; et al. Genetic Variants and Functional Pathways Associated with Resilience to Alzheimer’s Disease. Brain 2020, 143, 2561–2575. [Google Scholar] [CrossRef]
  73. Marioni, R.E.; McRae, A.F.; Bressler, J.; Colicino, E.; Hannon, E.; Li, S.; Prada, D.; Smith, J.A.; Trevisi, L.; Tsai, P.-C.; et al. Meta-Analysis of Epigenome-Wide Association Studies of Cognitive Abilities. Mol. Psychiatry 2018, 23, 2133–2144. [Google Scholar] [CrossRef]
  74. Yu, L.; Chibnik, L.B.; Yang, J.; McCabe, C.; Xu, J.; Schneider, J.A.; De Jager, P.L.; Bennett, D.A. Methylation Profiles in Peripheral Blood CD4+ Lymphocytes versus Brain: The Relation to Alzheimer’s Disease Pathology. Alzheimers Dement. 2016, 12, 942–951. [Google Scholar] [CrossRef] [Green Version]
  75. Ashraf-ganjouei, A.; Moradi, K.; Bagheri, S.; Aarabi, M.H. The Association between Systemic Inflammation and Cognitive Performance in Healthy Adults. J. Neuroimmunol. 2020, 345, 577272. [Google Scholar] [CrossRef]
  76. Weinstein, G.; Lutski, M.; Goldbourt, U.; Tanne, D. C-Reactive Protein Is Related to Future Cognitive Impairment and Decline in Elderly Individuals with Cardiovascular Disease. Arch. Gerontol. Geriatr. 2017, 69, 31–37. [Google Scholar] [CrossRef]
  77. Cukier, H.N.; Kunkle, B.W.; Vardarajan, B.N.; Rolati, S.; Hamilton-Nelson, K.L.; Kohli, M.A.; Whitehead, P.L.; Dombroski, B.A.; Van Booven, D.; Lang, R.; et al. ABCA7 Frameshift Deletion Associated with Alzheimer Disease in African Americans. Neurol. Genet. 2016, 2, e79. [Google Scholar] [CrossRef] [Green Version]
  78. Raj, T.; Li, Y.I.; Wong, G.; Humphrey, J.; Wang, M.; Ramdhani, S.; Wang, Y.-C.; Ng, B.; Gupta, I.; Haroutunian, V.; et al. Integrative Transcriptome Analyses of the Aging Brain Implicate Altered Splicing in Alzheimer’s Disease Susceptibility. Nat. Genet. 2018, 50, 1584–1592. [Google Scholar] [CrossRef]
  79. Humphries, C.; Kohli, M.A.; Whitehead, P.; Mash, D.C.; Pericak-Vance, M.A.; Gilbert, J. Alzheimer Disease (AD) Specific Transcription, DNA Methylation and Splicing in Twenty AD Associated Loci. Mol. Cell. Neurosci. 2015, 67, 37–45. [Google Scholar] [CrossRef] [Green Version]
  80. Tsuboi, K.; Nagatomo, T.; Gohno, T.; Higuchi, T.; Sasaki, S.; Fujiki, N.; Kurosumi, M.; Takei, H.; Yamaguchi, Y.; Niwa, T.; et al. Single CpG Site Methylation Controls Estrogen Receptor Gene Transcription and Correlates with Hormone Therapy Resistance. J. Steroid Biochem. Mol. Biol. 2017, 171, 209–217. [Google Scholar] [CrossRef]
  81. Qiu, C.; Shen, H.; Fu, X.; Xu, C.; Deng, H. Meta-Analysis of Genome-Wide Association Studies Identifies Novel Functional CpG-SNPs Associated with Bone Mineral Density at Lumbar Spine. Int. J. Genom. 2018, 2018, 6407257. [Google Scholar] [CrossRef] [PubMed]
  82. Gertz, J.; Varley, K.E.; Reddy, T.E.; Bowling, K.M.; Pauli, F.; Parker, S.L.; Kucera, K.S.; Willard, H.F.; Myers, R.M. Analysis of DNA Methylation in a Three-Generation Family Reveals Widespread Genetic Influence on Epigenetic Regulation. PLoS Genet. 2011, 7, e1002228. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  83. Harlid, S.; Ivarsson, M.I.L.; Butt, S.; Hussain, S.; Grzybowska, E.; Eyfjörd, J.E.; Lenner, P.; Försti, A.; Hemminki, K.; Manjer, J.; et al. A Candidate CpG SNP Approach Identifies a Breast Cancer Associated ESR1-SNP. Int. J. Cancer 2011, 129, 1689–1698. [Google Scholar] [CrossRef] [PubMed]
  84. Dayeh, T.A.; Olsson, A.H.; Volkov, P.; Almgren, P.; Rönn, T.; Ling, C. Identification of CpG-SNPs Associated with Type 2 Diabetes and Differential DNA Methylation in Human Pancreatic Islets. Diabetologia 2013, 56, 1036–1046. [Google Scholar] [CrossRef] [Green Version]
  85. Taqi, M.M.; Bazov, I.; Watanabe, H.; Sheedy, D.; Harper, C.; Alkass, K.; Druid, H.; Wentzel, P.; Nyberg, F.; Yakovleva, T.; et al. Prodynorphin CpG-SNPs Associated with Alcohol Dependence: Elevated Methylation in the Brain of Human Alcoholics. Addict. Biol. 2011, 16, 499–509. [Google Scholar] [CrossRef] [Green Version]
  86. Bani-Fatemi, A.; Gonçalves, V.F.; Zai, C.; de Souza, R.; Le Foll, B.; Kennedy, J.L.; Wong, A.H.; De Luca, V. Analysis of CpG SNPs in 34 Genes: Association Test with Suicide Attempt in Schizophrenia. Schizophr. Res. 2013, 147, 262–268. [Google Scholar] [CrossRef]
  87. Fridman, A.L.; Tainsky, M.A. Critical Pathways in Cellular Senescence and Immortalization Revealed by Gene Expression Profiling. Oncogene 2008, 27, 5975–5987. [Google Scholar] [CrossRef]
Genes 13 02150 g001 550
Figure 1. Regional plot of the association between DNA methylation in the ABCA7 region and general cognitive function. The top panel shows −log10 (p value) for the association between methylation and general cognitive function, adjusting for age, sex, age difference between methylation and cognition measurements, educational attainment, APOE ε2, APOE ε4, smoking status, PC1-4, and familial relatedness (random effects; Model 3), according to chromosomal positions. Nominally significant (p < 0.05) associations are above the dashed line. The middle panels show Ensembl genes, regulatory elements, and CpG islands (UCSC Genome Browser) in the ABCA7 region. The lower panel shows the correlations in the DNA methylation levels among the 72 CpG sites in this region. The five CpGs that have a nominal association with general cognitive function are marked by asterisks. The four CpGs and two SNPs that were identified in the SNP-by-CpG interactions associated with general cognitive function are marked by diamond symbols (CpGs) and arrows (SNPs).
Figure 1. Regional plot of the association between DNA methylation in the ABCA7 region and general cognitive function. The top panel shows −log10 (p value) for the association between methylation and general cognitive function, adjusting for age, sex, age difference between methylation and cognition measurements, educational attainment, APOE ε2, APOE ε4, smoking status, PC1-4, and familial relatedness (random effects; Model 3), according to chromosomal positions. Nominally significant (p < 0.05) associations are above the dashed line. The middle panels show Ensembl genes, regulatory elements, and CpG islands (UCSC Genome Browser) in the ABCA7 region. The lower panel shows the correlations in the DNA methylation levels among the 72 CpG sites in this region. The five CpGs that have a nominal association with general cognitive function are marked by asterisks. The four CpGs and two SNPs that were identified in the SNP-by-CpG interactions associated with general cognitive function are marked by diamond symbols (CpGs) and arrows (SNPs).
Genes 13 02150 g001
Genes 13 02150 g002 550
Figure 2. Linear prediction of CpG sites (% methylated) on general cognitive function for a given SNP genotype group in the ABCA7 region: (A) rs3764647*cg00135882, (B) rs3764647*cg22271697, (C) rs115550680*cg06169110, and (D) rs115550680*cg17316918. Models were adjusted for age, sex, age difference between methylation measurement and cognition measurement, educational attainment, APOE ε2, APOE ε4, smoking status, PC1-4, and familial relatedness as a random effect (Model 4). Regression lines are shown with standard error bands. For rs3764647, GG (n = 17) and AG (n = 156) groups were combined in the GG/AG group (n = 173). For rs115550680, GG (n = 5) and AG (n = 54) groups were combined in the GG/AG group (n = 59).
Figure 2. Linear prediction of CpG sites (% methylated) on general cognitive function for a given SNP genotype group in the ABCA7 region: (A) rs3764647*cg00135882, (B) rs3764647*cg22271697, (C) rs115550680*cg06169110, and (D) rs115550680*cg17316918. Models were adjusted for age, sex, age difference between methylation measurement and cognition measurement, educational attainment, APOE ε2, APOE ε4, smoking status, PC1-4, and familial relatedness as a random effect (Model 4). Regression lines are shown with standard error bands. For rs3764647, GG (n = 17) and AG (n = 156) groups were combined in the GG/AG group (n = 173). For rs115550680, GG (n = 5) and AG (n = 54) groups were combined in the GG/AG group (n = 59).
Genes 13 02150 g002
Table
Table 1. Sample characteristics of Genetic Epidemiology Network of Arteriopathy (GENOA) African Americans (n = 634).
Table 1. Sample characteristics of Genetic Epidemiology Network of Arteriopathy (GENOA) African Americans (n = 634).
Mean (SD) or n%
Age at cognition measurement (years)63.31 (8.08)
Age difference between methylation and cognition measurements (years) a6.03 (1.29)
Sex
    Female475 (74.90%)
    Male159 (25.10%)
Educational attainment
    At least some college300 (47.32%)
    High school degree/GED169 (26.66%)
    Less than high School degree/GED165 (26.03%)
Smoking status
    Current smoker105 (16.56%)
    Former smoker146 (23.03%)
    Never smoker383 (60.41%)
General cognitive function0.00 (1.00)
Delayed recall (RAVLT, number of words recalled)7.05 (3.34)
Processing speed (DSST, number of symbols)34.44 (12.62)
Word fluency (COWA-FAS, number of words)29.73 (11.61)
Concentration effectiveness (SCWT, number of items)22.53 (9.83)
Visual conceptual tracking (TMTA, seconds to test completion)61.63 (31.96)
Abbreviations: HS, high school; RAVLT, Rey Auditory Verbal Learning Test; DSST, Digit Symbol Substitution Test; COWA-FAS, Controlled Oral Word Association Test; SCWT, Stroop Color–Word Test; TMTA, Trail Making Test A. a. The subset sample (n = 494) consists of subjects with available genetic and methylation data.
Table
Table 2. Interaction between ABCA7 sentinel SNPs and CpG sites on general cognitive function (FDR q < 0.1; n = 494).
Table 2. Interaction between ABCA7 sentinel SNPs and CpG sites on general cognitive function (FDR q < 0.1; n = 494).
SNP AnnotationCpG Site AnnotationMain EffectsInteraction
SNP*CpG Site InteractionSNPPositionRisk AlleleRAFCpg SitePositionSite TypeRelation to CGIβSNPp-ValueβCpGp-Valueβinteractionp-Value
rs3764647*cg00135882rs37646471044712G0.20cg001358821065783PromoterNorth Shore−0.010.8750.240.086−0.801.46 × 10−4 **
rs3764647*cg22271697rs37646471044712G0.20cg222716971042537PromoterNorth Shelf−0.070.3190.167.23 × 10−6 *−0.185.77 × 10−4 **
rs115550680*cg06169110rs1155506801050420G0.06cg061691101046615Gene BodyCG Island−0.230.045 *0.060.143−0.382.18 × 10−4 **
rs115550680*cg17316918rs1155506801050420G0.06cg173169181056930Gene BodyOpen Sea−0.050.661−0.060.1640.414.84 × 10−4 **
Abbreviations: AA, African American; EA, European American; RAF, risk allele frequency in GENOA. Model 4: General cognitive function ~ SNP + CpG + SNP*CpG + age at cognitive testing + age difference between methylation and cognition measurements + sex + educational attainment + APOE ε2 + APOE ε4 + smoking status + PC1-4 + familial relatedness (random effect). * p < 0.05, ** FDR q < 0.1.
Table
Table 3. Estimated effect of CpG site on general cognitive function for the given ABCA7 SNP genotype group (n = 494).
Table 3. Estimated effect of CpG site on general cognitive function for the given ABCA7 SNP genotype group (n = 494).
SNPCpg SiteGenotypeβCpGp-Value
rs3764647 acg00135882AA0.090.566
GG/AG−0.680.004 *
rs3764647 acg22271697AA0.14 2.00 × 10−4 *
GG/AG−0.02 0.719
rs115550680 bcg06169110AA0.05 0.221
GG/AG−0.37 2.00 × 10−4 *
rs115550680 bcg17316918AA−0.06 0.202
GG/AG0.330.004 *
a: GG (n = 17) and AG (n = 156) groups were combined in the GG/AG group (n = 173). b: GG (n = 5) and AG (n = 54) groups were combined in the GG/AG group (n = 59). Model 4: General cognitive function~ SNP + CpG + SNP*CpG + age at cognition measurement + age difference between methylation and cognition measurements + sex + educational status + APOE ε2 + APOE ε4 + smoking status + PC1-4 + familial relatedness (random effect). * p < 0.05.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Chaar, D.L.; Nguyen, K.; Wang, Y.-Z.; Ratliff, S.M.; Mosley, T.H.; Kardia, S.L.R.; Smith, J.A.; Zhao, W. SNP-by-CpG Site Interactions in ABCA7 Are Associated with Cognition in Older African Americans. Genes 2022, 13, 2150. https://doi.org/10.3390/genes13112150

AMA Style

Chaar DL, Nguyen K, Wang Y-Z, Ratliff SM, Mosley TH, Kardia SLR, Smith JA, Zhao W. SNP-by-CpG Site Interactions in ABCA7 Are Associated with Cognition in Older African Americans. Genes. 2022; 13(11):2150. https://doi.org/10.3390/genes13112150

Chicago/Turabian Style

Chaar, Dima L., Kim Nguyen, Yi-Zhe Wang, Scott M. Ratliff, Thomas H. Mosley, Sharon L. R. Kardia, Jennifer A. Smith, and Wei Zhao. 2022. "SNP-by-CpG Site Interactions in ABCA7 Are Associated with Cognition in Older African Americans" Genes 13, no. 11: 2150. https://doi.org/10.3390/genes13112150

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop