ABCA7 expression is associated with Alzheimer’s disease polymorphism and disease status

Abstract

Genome-wide association studies (GWAS) have implicated a series of single nucleotide polymorphisms (SNPs) in Alzheimer’s disease (AD) risk. Elucidating the function of these SNPs is critical to identifying the underlying pathways and, potentially, novel therapeutic agents. SNPs within the gene ATP binding cassette A7 (ABCA7) reached significance in these studies, warranting investigation into their actions. Here, we analyzed ABCA7 expression in a set of human brain samples as a function of AD-associated SNPs and AD status. We report that the rs3764650T allele that decreases AD risk is associated with increased ABCA7 expression. However, ABCA7 expression is increased in AD individuals. We interpret our findings as suggesting a model wherein increased ABCA7 expression reduces AD risk and that the increased ABCA7 observed in AD reflects an inadequate compensatory change.

1. Introduction

AD is defined by short-term memory loss progressing into dementia and neuropathology including senile plaques and neurofibrillary tangles [12, 16, 35, 42]. AD risk increases with age with the disease affecting as many as one out of three people by 85 years of age [5, 9, 11, 15, 29]. Since AD is an impending health crisis, we seek to identify the underlying mechanisms and translate these mechanisms into preventative agents.

GWAS have identified genes in pathways that modulate AD risk, including genetic variants in ABCA7 (reviewed in [3, 36]). ABCA7 is part of a large family of transporters expressed in the brain on microglia and neurons and is capable of fluxing phospholipids as well as enhancing phagocytosis of apoptotic cells [17, 22-24, 43]. As both a risk factor in AD and potential drug target, ABCA7 merits further investigation.

Here we hypothesized that rs3764650, the primary AD-associated SNP, modulates ABCA7 expression. In evaluating this hypothesis, we genotyped several SNPs and quantified ABCA7 expression in 57 human brain samples. We interpret our results as suggesting a model wherein increased ABCA7 expression reduces AD risk, possibly through enhanced microglial activation and phagocytosis.

2. Materials and methods

2.1. Ethics statement

The work described here was conducted under the approval of the University of Kentucky Institutional Review Board.

2.2. Human autopsy tissue

Human anterior cingulate samples were supplied by the University of Kentucky AD Center Neuropathology Core and have been described previously [28, 46, 47]. Evaluation of AD status was conducted by the AD Center Neuropathology and Clinical Cores using guidelines set forth by the National Institute on Aging Reagan Institute (NIA-RI) that included evaluation of neurofibrillary tangles and neuritic senile plaques [32-34]. Age at death for the cognitively intact, i.e. non-AD donors, was 82.3 ± 1.6 (mean ± SE, n = 28) while AD donors were 81.6 ± 1.1 (mean ± SE, n = 29). The post-mortem interval for non-AD and AD donors was 2.8 ± 0.2 (n = 29) hours and 3.4 ± 0.1 (n = 27), respectively. By NIA-RI neuropathology criteria, the 28 non-AD individuals included 12 with no AD likelihood, 10 with low likelihood, and six with moderate AD likelihood. Neuropathology in the 29 AD individuals was uniformly robust, i.e., definite AD by NIH-RI criteria [33].

2.3. qPCR

ABCA7 expression levels were determined by quantitative polymerase chain reaction (qPCR) (Bio-Rad, Hercules, CA). Assays contained the sense primer corresponding to sequence in exon 44, i.e., 5’-TCCTTTGGAACAGCCTTTTG-3’ and antisense primer corresponding to sequence in exon 45, i.e., 5’-CTGCCCTTGAGATGTTGC-3’. Each 20 μL reaction mixture contained 20 ng of cDNA, 1 μM of each primer, and 2x SYBR Green (Quanta Biosciences, Gaithersburg, MD) and was subjected to a qPCR profile of 40 cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 20 s. A standard curve was amplified in parallel to determine the copy number of qPCR product per sample. ABCA7 copy numbers were normalized to the geometric mean of the reference genes eukaryotic translation initiation factor 4H (EIF4H) and ribosomal protein L32 (RPL32). Assays were repeated three times.

2.4. Genotyping

DNA samples were genotyped for rs3764650, rs3752246, rs4147914, rs3752229, rs7247084, rs757232, rs2279796, rs7251792, 12710103, and rs4147930 by using Taqman FAM and VIC dye-labeled probes (Life Technologies, Foster City, CA) by qPCR (Bio-Rad, Hercules, CA).

2.5. Statistical analysis

The extent that ABCA7 mRNA levels were associated with AD-associated SNPs and/or AD-status was determined by using linear regression analysis (SPSS v.19 (IBM, Somers, NY)). A dominant allelic model was assumed, i.e. individuals homozygous for the rs3764650 minor G allele were grouped with those heterozygous for the SNP and compared with individuals homozygous for the rs3764650 major T allele. The model also included age at death and sex as covariates.

3. Results

The ABCA7 SNP rs3764650 [T/G] is associated with AD risk [3]. To assess the hypothesis that ABCA7 expression is modulated by this SNP, we quantified ABCA7 copy number and then normalized these values to the geometric mean of two housekeeping genes [7, 27]. As depicted in Figure 1, ABCA7 expression appears reduced in individuals with the AD-risky G allele and increased in those with AD. To evaluate this apparent trend, we used linear modeling that included rs3764650, AD status, age, and sex. Table 1 shows that ABCA7 expression was significantly associated with rs3764650 genotype and AD status. Overall, we interpret these results as suggesting that ABCA7 expression is reduced with the minor rs3764650G allele and increased with AD.

Figure 1. ABCA7 expression as a function of rs3764650 genotype and AD status.

ABCA7 expression is decreased in individuals carrying rs3764650G and increased with. Expression was evaluated as a function of the AD-associated SNP rs3764650 in a dominant model, i.e., rs3764650 GT and GG individuals were compared against TT individuals.

Table 1.

Rs3764650 and AD status are associated with ABCA7 expression. Linear regression statistical analysis included AD status, rs3764650 genotype, age at death and sex. The adjusted R² for the model was 0.285. Age and sex were not significantly associated with ABCA7 expression (p=.44 and p=.26, respectively).

	Standard Beta
Parameter	Coefficient	P-Value
rs3764650	−0.375	3.5×10−3
AD Status	0.502	1.1×10−5

Several ABCA7 SNPs have been associated with ABCA7 expression in brain or monocytes by others [2, 45]. To evaluate whether these SNPs are associated with ABCA7 expression in our brain samples, we genotyped the samples for these SNPs and performed linear regression analysis with each individual SNP. Overall, the SNPs showed variable association with ABCA7 expression with rs3764650 having the lowest p value (Table 2). We hypothesized that if these other SNPs were associated with expression only through rs3764650, the linkage disequilibrium between each SNP and rs3764650 would correlate with the strength of the SNP’s association with ABCA7 expression. We found a robust correlation in this analysis (Figure 2). We interpret this positive relationship as supporting the concept that these other SNPs are associated with ABCA7 expression only through rs3764650.

Table 2.

Relationship between SNP linkage disequilibrium with rs3764650 and association with ABCA7 expression. The linkage disequilibrium between rs3764650 (r²) and each of the other SNPs was determined by using SNAP (http://www.broadinstitute.org/mpg/snap/ldsearch.php). The association between each individual SNP and ABCA7 expression was determined by linear regression analyses that included AD status, age, sex and the SNP in a dominant model. The rationale for including each SNP is noted in the comment column.

SNP	r2 with rs3764650	Standard Beta Coefficient	P-Value	Comment (reference)
rs3764650	1.000	−0.402	0.001	AD GWAS [15, 31]
rs4147914	0.584	−0.296	0.021	Exonic SNP
rs3752229	0.541	−0.296	0.028	Associates with ABCA7 expression in brain [2]
rs7251792	0.245	0.161	0.209	Associates with ABCA7 expression in monocytes [43]
rs3752246	0.230	−0.231	0.066	Non-synonymous exonic SNP [31]
rs757232	0.173	−0.170	0.174	Associates with ABCA7 expression in brain [2]
rs4147930	0.166	−0.190	0.138	Exonic SNP
rs2279796	0.036	0.123	0.354	Associates with ABCA7 expression in monocytes [43]
rs7247087	0.007	0.169	0.171	Associates with ABCA7 expression in brain [2]
rs12710103	0.006	0.037	0.772	Associates with ABCA7 expression in monocytes [43]

Figure 2. SNP linkage disequilibrium with rs3764650 correlates with SNP association with ABCA7 expression.

The r² for the relationship between SNP linkage disequilibrium and association with ABCA7 expression was 0.93.

4. Discussion

Our primary findings are that ABCA7 expression is decreased in individuals that carry the AD-risky rs3764650G allele. Also, ABCA7 expression is increased with AD. The most parsimonious interpretation of our results is that the AD-risky rs3764650G allele decreases ABCA7 prior to AD onset, i.e., increased ABCA7 expression would reduce AD risk. From these results we propose that increased ABCA7 expression during AD is insufficient to block disease progression. The mechanistic underpinnings of ABCA7 action are unclear but likely involve lipid regulation or phagocytosis.

Our finding that ABCA7 expression is increased in AD is consistent with a prior report from Karch et al that ABCA7 expression was increased in individuals with increased cognitive decline [20]. Our finding that ABCA7 expression is reduced with the AD-risky rs3764650G allele has not been reported previously. Instead, Allen et al identified several other SNPs that were associated with ABCA7 expression [2]. We evaluated these SNPs and found that their association with ABCA7 expression appears to reflect their linkage disequilibrium with rs3764650. However, Karch et al examined rs3764650 directly and did not detect an association between this SNP and ABCA7 expression. Several technical and analytical differences between Karch et al and this study may underlay the differences in our findings. Technical differences include that we used SYBR-green based qPCR and normalized to the geometric mean of RPL32 and EIF4H while Karch et al used TaqMan-based qPCR and normalized relative to GAPDH. Additionally, we used a dominant model for rs3764650 while Karch et al used an additive model; we re-analyzed our data with an additive model and found a less robust significance for SNP association with ABCA7 expression (p=0.011 for additive model as opposed to p=0.0035 for dominant model). Karch et al examined 112 brain samples while our study included 57 samples; a post-hoc power analysis of the SNP association with expression here was 85.3%. A greater number of samples from non-AD individuals with the rs3764650G allele would provide greater insight on rs3764650G effects on ABCA7 expression in the absence of pathology. In summary, the normalization genes may be the most likely variable underlying our differing results, although additional unknown variables are also possible.

Two primary mechanisms may underlie the actions of ABCA7 in AD. First, ABCA7 may modulate AD risk through its ability to transfer phospholipids to apolipoproteins such as apoE, apoJ and apoA-1 [1, 43]. This is suggestive that ABCA7 may act similarly to ABCA1, which has been shown to transfer cholesterol to apoE and modulate amyloid-beta (Aβ) levels [6, 8, 19, 23, 25, 44]. For example, ABCA1 deficient mice have decreased levels of lipidated apoE and increased Aβ deposits [26]. However, a key difference between ABCA1 and ABCA7 is that apoE levels are not altered in ABCA7 deficient mice [23]. Additionally, ABCA1 deficiency causes Tangiers disease in humans and a robust phenotype in mice but ABCA7 polymorphisms do not associate with lipid homeostasis and ABCA7 deficiency has only a modest effect on lipid homeostasis in mice [10, 20, 21, 23, 31]. Overall, the actions of ABCA7 on lipid homeostasis may be modest relative to ABCA1. A second mechanism whereby ABCA7 may modulate AD risk is through microglial phagocytosis [17, 23]. Indeed, ABCA7 is a homolog of ced-7, the C. elegans gene critical for apoptotic cell engulfment [17]. Here, we note that ectopic ABCA7 expression increases microglial phagocytosis of multiple substrates, including apoptotic cells, Aβ, and synthetic substrates in vitro [17, 23, 39-41]. Consistent with this possibility, ABCA7 deficient mice have increased Aβ deposition in vivo and rs3764650G in humans has been associated with increased neuritic plaque burden [23, 37, 38]. In summary, (i) ABCA7 increases phagocytosis activity in vitro, (ii) ABCA7 deficiency increases Aβ in a murine model, and (iii) the rs3764650G allele that correlates with reduced ABCA7 expression is associated with increased neuritic plaques in human. Hence, we propose that the most likely interpretation of current results is that ABCA7 modulates AD risk through phagocytosis.

This interpretation is consistent with two recent genetic findings that have implicated microglial function in AD risk. First, TREM2 increases microglial activation; a rare, apparently inhibitory, TREM2 allele is associated with robustly increased AD risk [14, 18]. Second, CD33 reduces microglial activation; a CD33 allele that reduces functional CD33, apparently by modulating exon 2 splicing, is associated with decreased AD risk [4, 13, 30]. These results overall suggest that increased microglial activation reduces AD risk. Hence, the observation that ABCA7 expression enhances microglial activation coupled with our finding that an AD-protective ABCA7 allele is associated with increased ABCA7 expression generates a consistent overall model wherein increased microglial activation, especially phagocytosis, is AD-protective (Figure 3). TREM2 and CD33 modulate phagocytosis by signaling through Syk and a Syk phosphatase, SHP1, respectively. The signaling pathway whereby ABCA7 modulates phagocytosis is unclear but may involve the ERK signaling [17].

Figure 3. ABCA7, TREM2 and CD33 modulate microglial function to alter AD risk.

This model proposes that enhanced microglial phagocytosis due to increased ABCA7 or TREM2 function, or reduced CD33 function, leads to decreased AD risk.

In summary, our primary findings are that ABCA7 expression is decreased with the AD-risky rs3764650G allele and increased with AD. The increase in ABCA7 expression in AD could reflect that ABCA7 contributes to AD risk. However, a parsimonious model is that the ABCA7 induction in AD is in response to the disease because ABCA7 expression was also increased with the AD-protective rs3764650T allele. Hence, we interpret our data overall as suggesting a model wherein a SNP-driven, presumably long-term, increase in ABCA7 expression reduces AD risk; increased ABCA7 expression during AD is insufficient to block disease progression.

Highlights.

• GWAS have identified an AD-associated SNP within ABCA7.

• We investigated the extent that this AD-associated SNP modulates ABCA7 expression.

• The rs3764650G allele, which increases AD risk, was associated with decreased ABCA7 expression.

• ABCA7 expression was increased with AD status.

• We propose a model wherein increased ABCA7 expression reduces AD risk with the increase observed in AD reflecting a poor compensatory response.

Abbreviations

GWAS

Genome-wide association studies

SNP

Single nucleotide polymorphism

AD

Alzheimer’s disease

ABCA7

ATP binding cassette A7

qPCR

quantitative polymerase chain reaction

Aβ

Amyloid-beta

EIF4H

eukaryotic translation initiation factor 4H

RPL32

ribosomal protein L32