APOE Genotype Differentially Modulates Plasma Lipids in Healthy Older Individuals,with Relevance to Brain Health

Abstract

Apolipoprotein E (APOE) genotype is an established genetic risk factor for sporadic Alzheimer’s disease (AD) but the extent to which APOE genotype influences the plasma lipidome is unknown, even though lipids are potential diagnostic or prognostic biomarkers for AD. We quantified plasma lipids using untargeted liquid chromatography coupled mass spectrometry in a total of 152 non-demented participants aged 65–100 years carrying at least one ɛ2 or ɛ4 allele (ɛ2/ɛ2 or ɛ2/ɛ3, n = 38: ɛ4/ɛ3 or ɛ4/ɛ4, n = 38), who were roughly matched to an ɛ3/ɛ3 control by age, sex, and lipid-lowering medication (n = 76). Low density lipoprotein cholesterol levels were genotype dependent (ɛ4/ɛ4> ɛ4/ɛ3> ɛ3/ɛ3> ɛ2/ɛ3> ɛ2/ɛ2). The greatest variation in lipids was related to the ɛ2 isoform, where various lysophosphatidylcholines and all phosphatidylethanolamine (PE) subclasses were elevated relative to ɛ3/ɛ3 and ɛ4 carriers. APOE ɛ4 carriers had reduced phosphatidylinositol relative to ɛ3/ɛ3 and ɛ2 carriers. Logistic regression revealed that ɛ2 carriers were at least 4 times higher odds of being in the highest tertile of PE lipid level relative to ɛ3/ɛ3. The elevation in PE and other phospholipids in ɛ2 carriers may indicate the protective effect of ɛ2 is linked to these phospholipids. Additionally, high baseline PE in cognitively normal participants predicted protection against cognitive decline six years later. Our data suggest substantial modulation of plasma lipids by APOE genotype and therefore indicates possible lipid targets and pathomechanisms involved in AD risk.

Keywords

Alzheimer’s disease apolipoproteins lipids lipidomics liquid chromatography-mass spectrometry plasma

INTRODUCTION

Apolipoprotein E (ApoE) is a 34 kDa, 299 amino acid glycoprotein encoded on chromosome 19 in humans which is required for clearance of plasma chylomicron and very low density lipoprotein (VLDL) remnants by interaction with the Low density lipoprotein (LDL) receptor and the LDL-related receptor. ApoE is also expressed in the central nervous system (CNS) by astrocytes and microglia to distribute cholesterol among neurons [1].

The function of ApoE is isoform dependent. Three alleles exist, ɛ2, ɛ3, and ɛ4, giving rise to three isoforms: ApoE2, ApoE3, and ApoE4, respectively, which combine to form six genotypes. ApoE3 is considered the parent isoform with highest prevalence (77%), followed by ApoE4 (15%) and ApoE2 (8%) [2]. ApoE2 differs from ApoE3 by the single amino acid substitution Arg158Cys and exhibits poorer binding capacity to the LDL receptor, impairing clearance of triglyceride-rich lipoprotein remnants [3], while ApoE4 has the amino acid substitution Cys112Arg, and the resulting conformational change induces lipolytic processing of VLDL in ApoE4 carriers [4].

In the periphery, ApoE4 increases risk of dyslipidemias and atherosclerosis by elevating LDL-C levels [5], while ApoE2 decreases LDL-C levels. In the CNS, the relationship between APOE and late-onset AD is well established, with an enrichment of APOE ɛ4 among persons with AD [2, 6] and carriers of APOE ɛ4 are at greater risk of developing AD compared to APOE ɛ3/ɛ3 or APOE ɛ2 carriers, while ɛ2 is considered protective [6]. The mechanisms by which ApoE isoforms modulate AD risk is still unclear, but possibly ApoE isoforms differentially affect amyloid-β processing and clearance [7]. Non-amyloid mechanisms have also been suggested, such as neuroinflammation, loss of synaptic function, and tau hyperphosphorylation [8, 9].

However, the extent to which APOE genotype modulates lipid levels in human plasma or the brain remains relatively unknown. The brain is a lipid rich organ, where synaptogenesis, neuronal, and glial maintenance depend on lipid trafficking. Thus, altered lipid metabolism and distribution related to ApoE is likely to be implicated in neurodegeneration independent of amyloid-β pathology. In this study, we aim to characterize the extent of variation in the human plasma lipidome by comparing non-demented participants carrying ɛ2, ɛ3, and ɛ4 alleles. To our knowledge, this is the first study to investigate the plasma lipidome variation based on APOE genotype in older aged participants.

MATERIALS AND METHODS

Participant cohorts

Blood and patient data were collected from participants aged 65–100 years (n = 152) enrolled in three Australian population-based aging studies, including the Sydney Memory and Ageing Study (MAS, n = 136), the Sydney Centenarian Study (SCS, n = 6), and the Hunter Community Study (HCS, n = 10). These protocols have been published [10 –12], along with methods for assessment of cognitive domains, and consensus diagnosis criteria for identifying and classifying mild cognitive impairment (MCI)/AD [12 –14]. Each study features several waves of data collection (baseline and follow-up each two years) for longitudinal analysis, with the initial (i.e., baseline) collection following recruitment called “Wave 1”. Wave 1 plasma was analyzed from each cohort, while additional consensus diagnosis data from MAS participants was taken during Wave 1 up to and including Wave 4 (third follow-up held six years after the first).

Ethics approval

The SCS and MAS were approved by the Ethics Committees of the University of New South Wales and the South Eastern Sydney and Illawarra Area Health Service (ethics approvals HC12313 and HC14327). The HCS was approved by the University of Newcastle and Hunter New England Human Research Ethics Committees (HREC 03/12/10/3.26). All work involving human participants was carried out in accordance with the principles of the Declaration of Helsinki of the World Medical Association.

Study design

We randomly selected equal numbers of non-demented participants carrying at least one ɛ2 allele, referred to as ɛ2 carriers (n = 38, of whom 7 were ɛ2/ɛ2), and participants carrying at least one ɛ4 allele, referred to as ɛ4 carriers (n = 38, of whom 21 were ɛ4/ɛ4). The numbers of subjects in each APOE genotype group based on study cohort is summarized in Supplementary Table 1. Exclusion criteria were mixed ɛ2/ɛ4 genotype, Mini-Mental State Examination (MMSE) scores <20 (indicating moderate to severe cognitive deficits), neuropsychiatric disorders or cardiovascular complications. Each ɛ2 or ɛ4 carrier was approximately matched by age (maximum difference of 1 year), sex, and lipid-lowering medication status to an ɛ3/ɛ3 participant (n = 76, Table 1). This was done to ensure similar distributions of participant characteristics in the pairs of genotype groups to be compared.

Table 1

Patient characteristics and lipid profiles by APOE genotype

	ɛ2 carriers	ɛ3/ɛ3 matched to ɛ2 carriers	ɛ4 carriers	ɛ3/ɛ3 matched to ɛ4 carriers	χ² (2 df)^a
N	38	38	38	38	N/A
Age	78.0 (6.3)	77.2 (5.4)	75.6 (3.7)	76.1 (5.1)	2.046
Males N (%)	20 (52.6%)	20 (52.6%)	16 (42.1%)	16 (42.1%)	0.844
BMI	27.4 (4.0)	27.3 (3.0)	27.9 (5.8)	27.8 (4.0)	0.707
Lipid-lowering medication	16 (42.1%)	19 (50.0%)	23 (62%)	25 (61%)	3.225
Years of Education	11.1 (3.2)	11.2 (3.1)	12.1 (3.84	10.6 (3.4)	3.489
MMSE score	28.2 (1.5)	27.8 (3.2)	28.6 (1.4)	28.4 (1.5)	2.105
WHR	0.92 (0.10)	0.89 (0.07)	0.88 (0.07)	0.91 (0.10)	3.643
LDL-C (mmol/L)	2.60 (0.85)	2.80 (0.92)	2.98 (0.80)	2.78 (0.73)	3.477
HDL-C (mmol/L)	1.50 (0.50)	1.33 (0.41)	1.39 (0.37)	1.49 (0.40)	0.346
Total Cholesterol (mmol/L)	4.69 (0.96)	4.62 (1.04)	4.84 (0.95)	4.76 (0.81)	0.744
Triglycerides (mmol/L)	1.21 (0.75)	1.07 (0.42)	1.00 (0.42)	1.07 (0.58)	2.622

BMI, body mass index; MMSE, Mini-Mental State Examination; WHR, waist-hip ratio; LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol. Values represent mean and standard deviation (SD). ^aKruskall Wallis test was used for all variables except the use of Lipid-lowering medications, in which case the χ² test for equality of proportions was used. For these comparisons, the two ɛ3/ɛ3 groups were combined into a single group, and then compared against ɛ2 carriers and ɛ4 carriers. None of the variables were significantly different by APOE matched groups (all p > 0.05).

This study had two components. The first component involved lipidomic analysis using plasmas collected from subjects shortly after recruitment (Wave 1). The second component of the study was a longitudinal analysis where consensus diagnosis data was analyzed from MAS participants across time. Consensus diagnosis assessment was performed from Wave 1, and every two years following, with the latest assessment 6 years after Wave 1 designated “Wave 4”. Of MAS Wave 1 participants included in our study, n = 127 provided consensus diagnosis data at Wave 1. This dropped to 115 in Wave 2, 108 in Wave 3, and 103 in Wave 4. This consensus diagnosis data was used to assess whether lipidomic profiles from plasma in Wave 1 could be useful at predicting potential cognitive decline to MCI/dementia.

Plasma collection, handling, and storage

To minimize pre-analytical variability [15, 16], fasting EDTA plasma was separated from whole blood within 2–4 h of venepuncture and immediately stored at –80°C prior to bio-banking, and lipids were extracted within 15 min of freeze thawing.

Targeted assays of plasma lipids and ApoE protein

Plasma total cholesterol, LDL-C, High-density lipoprotein cholesterol (HDL-C), and triglycerides (TG) were measured by enzymatic assay at SEALS pathology (Prince of Wales Hospital) [17], using a Beckman LX20 Analyzer with a timed-endpoint method (Fullerton, CA). Plasma ApoE levels were measured as previously described [17].

APOE genotyping

DNA was extracted from samples using established procedures. Two APOE single nucleotide polymorphisms (SNPs rs7412, rs429358) were genotyped using Taqman genotyping assays (Applied Biosystems Inc., Foster City, CA) to determine the APOE haplotype, which has three alleles (ɛ2, ɛ3, ɛ4) using Taqman genotyping assays (Applied Biosystems Inc., Foster City, CA). Detailed information on our standard genotyping procedure has been previously published [14].

Plasma lipid extraction: Single phase 1-butanol/methanol

Plasma lipids were extracted as previously described [18]. We added 100 pmol/10μL of internal lipid standards mixture (Avanti, Splash Lipidomix, Alabaster, USA) to 10μL plasma [19] in 0.5 mL polypropylene tubes (Eppendorf, Sydney, Australia). A solution of 1-butanol/methanol (100μl, 1:1 v/v) containing 5 mM ammonium formate was then added and samples vortexed (10 s), and waterbath sonicated (1 h). Tubes were centrifuged at 13,000 g (10 min, ambient temperature). The supernatant was transferred to a liquid chromatography coupled mass spectrometry (LC-MS) vial with glass insert and stored at –80°C prior to analysis.

Liquid chromatography/mass spectrometry

Lipid analysis was performed by LC ESI-MS/MS using a Thermo QExactive Plus Orbitrap mass spectrometer (Bremen, Germany) [19]. A Waters ACQUITY UPLC CSH^TM C18 1.7μm, 2.1×100 mm column was used for liquid chromatography. Solvents A and B consisted of acetonitrile:MilliQ water (6:4 v/v) and isopropanol:acetonitrile (9:1 v/v) respectively, both containing 10 mM ammonium formate and 0.1% formic acid at a flow rate of 260μL/min, using the following gradient conditions: 32% solvent B to 100% over 25 min, a return to 32% B, and finally 32% B equilibration for 5 min prior to the next injection. Product ion scanning was performed in both positive and negative ion modes. Sampling order was randomized prior to analysis and pooled quality control samples run every 20 injections (additional information in the Supplementary Methods).

Alignment and peak detection

The raw data was aligned, chromatographic peaks selected, specific lipids identified and their peak areas integrated using LipidSearch software v4.1 (Thermo Fischer Scientific, Waltham MA), and data exported to Excel for manual processing (Supplementary Methods). The raw abundances (peak areas) were normalized by dividing each peak area by the raw abundance of the corresponding internal standard for that lipid class.

Data analysis

Data analysis and graphs were generated using IBM SPSS 24.0, R and Graphpad Prism 7. Multiple linear regression analyses were applied to assess differences in lipids by lipid-lowering medication status, correcting for age, sex, and body mass index (BMI), with dependent variables first log₁₀-transformed if non-normal. In these linear regression analyses, dummy variables representing the contrasts between medication and medication naïve groups were included in the equations. Likewise, group differences in lipids between ɛ2 and ɛ3/ɛ3 samples, and between ɛ4 and ɛ3/ɛ3 samples were assessed using multiple linear regression analyses, additionally correcting for LDL-C, HDL-C, and TG. In these linear regression analyses, dummy variables representing the contrasts between the above two pairs of APOE genotype groups were included in the equations. In each dummy variable, the relevant comparison ɛ3/ɛ3 group was treated as the reference group and coded “0”, while the ɛ2 and ɛ4 carrier groups were coded “1”. Binary logistic regression was used to examine the effect of APOE genotype on classification of lipid levels by tertiles, controlling for age, sex, BMI, and lipid-lowering medication, as well as to examine the effect of lipid tertiles on later conversion to MCI or AD. Measurements within the highest tertile of lipids were coded ‘1’, otherwise ‘0’, sex was coded ‘1’ for females and ‘0’ for males, and later conversion to MCI/AD was coded ‘1’, otherwise ‘0’. Cox regression was used to describe the effect of lipid levels on cognitive decline over time, which also utilized consensus diagnosis data from Wave 2 and Wave 3. For this analysis, Wave 1 was denoted as time 0, and Wave 2, 3, and 4 denoted as time 1, 2, and 3 respectively. The event was “decline” versus “non-decline”. A “decline” event was noted during the first wave at which subjects who were normal at Wave 1 progressed to MCI or dementia and did not improve back to normal by Wave 4, or the first wave at which subjects with MCI at Wave 1 progressed to dementia. Otherwise, a “non-decline” event was noted for the subject in question at the last time point at which the subject had available consensus diagnosis data. Partial correlations were used to identify associations between cognitive domains and lipid abundance of Wave 1 plasma, correcting for the effect of age, sex, years of education, APOE genotype, BMI, and lipid-lowering medication. For t-tests of lipid classes between APOE genotypes, we corrected levels of statistical significance for multiple testing using the Benjamini-Hochberg false discovery method [20] with an FDR of 0.05.

Lipid shorthand notation

Lipids are named according to LIPID MAPS convention [21] (Supplementary Table 2), with slight modification to denote summation of lipids of a particular class/subclass: Cer(d18:1/X) refers to the sum of all ceramides (Cer) with an 18:1 fatty acid in the sn-1 position, while CE(18:X) refers to the sum of cholesterol esters (CE) with an 18 carbon chain length.

RESULTS

Participant demographics

Participant demographics are summarized in Table 1, showing no significant differences on any of these measures, nor on additional lipid measures. There were some sex differences (higher years of education and waist-hip ratio in males, higher HDL-C and total cholesterol in females, p < 0.05, Supplementary Table 3).

Effect of lipid-lowering medication on lipids

Participants on lipid-lowering medication had significantly lower total cholesterol and LDL-C (t = –7.266 and –6.47, respectively; p < 0.05, Supplementary Figure 1A), and trended toward lower HDL and a higher HDL-C/LDL-C ratio (t = –1.922 and 1.766, respectively; p < 0.10, Supplementary Figure 1A). There were no differences in log₁₀-transformed abundances of most lipid classes by lipid-lowering medications after adjustment for age, sex, and BMI, except in diacylglycerol (DG) and lysophosphatidylcholines (LPC) levels (Supplementary Figure 1B, p < 0.05).

Associations of LDL-C, HDL-C, and ApoE with APOE genotype

Among all subjects, there were no differences in LDL-C levels between genotypes (p > 0.05, Table 1). In lipid-lowering medication-naïve participants (n = 71), median LDL-C levels increased in a genotype dependent manner (Fig. 1A), with ɛ2/ɛ2 having the lowest levels, followed by ɛ2/ɛ3, ɛ3/ɛ3, then ɛ3/ɛ4 and ɛ4/ɛ4 having the highest LDL-C levels. Pairwise comparisons showed ɛ2 carriers had significantly lower LDL-C levels (Fig. 1A) compared to matched ɛ3/ɛ3 individuals (p = 0.04) and ɛ4 carriers (p = 0.013, paired Wilcoxon tests). This relationship was not conserved in participants on lipid-lowering medications, where LDL-C and HDL-C levels, as well as their ratio, did not differ by APOE genotype (p > 0.05).

Fig.1

Differences in LDL-C (mmol/L) in individuals who are not on lipid lowering medication, and ApoE protein levels (μg/mL) by APOE genotype. Boxplots show (A) LDL-C and (B) ApoE protein levels across APOE genotype from ɛ2/ɛ2 to ɛ4/ɛ4. LDL-C and ApoE levels are significantly different across APOE genotypes (Kruskall Wallis test and ANCOVA respectively, p < 0.05). Circles represent modest outliers (defined as points 1.5–3.0×interquartile range above or below the highest and lowest quartiles respectively), while asterisks represent extreme outliers more than 3.0×interquartile ranges above or below the highest and lowest quartiles, respectively.

Among all subjects, ApoE plasma levels were significantly associated with APOE genotype (Fig. 1B, F = 10.26, p < 0.001), with highest levels in ɛ2 carriers (especially ɛ2/ɛ2) and lowest levels in ɛ4 carriers (especially ɛ4/ɛ4).

Associations of lipid classes with APOE genotype and ApoE levels

Grouped differences between APOE ɛ2 carriers and ɛ3/ɛ3 matched participants, and APOE ɛ4 carriers and ɛ3/ɛ3 matched participants were determined by multiple linear regression, adjusting for age, sex, BMI, lipid-lowering medication, and plasma LDL-C, HDL-C, and TG levels in the model (Fig. 2), with APOE groups expressed as dummy variables. LPC(22:X), phosphatidylcholine (PC)(18:0/X), and all phosphatidylethanolamine (PE) subclasses were higher among ɛ2 carriers relative to ɛ3/ɛ3 participants, though LPC(22:X), PC(18:0/X) and PE(20:0/X) did not survive multiple testing correction. There were fewer differences in lipid classes between ɛ4 carriers and ɛ3/ɛ3 participants, where only Cer(d16:1/X) and phosphatidylinositol (PI) were increased among ɛ4 carriers, but PI was not significant after multiple testing correction.

Fig.2

Lipid profiles by APOE genotype. Grouped boxplots are shown with comparisons of ɛ2 carriers versus matched ɛ3/ɛ3 groups, ɛ4 carriers versus matched ɛ3/ɛ3 groups, and ɛ2 versus ɛ4 carriers. Significant differences were determined between APOE genotype groups by using APOE genotype as a dummy variable in a multiple linear regression model with age, sex, BMI, lipid-lowering medication usage, LDL-C, HDL-C, and TG levels as covariates. A red line indicates significance of the t-test between the groups overlapped, while a bracket indicates significance of groups between ɛ2 carriers and ɛ4 carriers. Threshold for significance is set at ^*p = 0.012. ^†0.012 < p<0.05.

Comparing grouped differences between ɛ2 carriers and ɛ4 carriers, LPC(22:X), total PC, total PE, and all PC and PE subclasses, as well as PI were significantly higher in ɛ2 carriers compared to ɛ4 carriers, though total PC and PC(18:0/X) were not significant after correction for multiple testing (p < 0.012). PE lipids were highly inter-correlated, especially among ɛ2 carriers (Supplementary Figure 2A and B), indicating the relationships for PE subclasses are also reflected ubiquitously at the lipid species level. Furthermore, subjects with the highest tertile of total PE tended to have the highest tertile of most PE species, compared to subjects in the lower two tertiles. We found that of 49 individual PE species analyzed, a median 83.6% of these species were also of the highest tertile if the subject had the highest tertile of total PE. By contrast, subjects who did not have the highest tertile of total PE also tended not to have the highest tertile of individual PE species, where among these subjects, a median of only 4.1% of individual PE species were among the highest tertile. This shows that total PE is a good indicator of levels of most individual PE species.

Pearson partial correlations correcting for the effect of age, sex, BMI, APOE genotype, and lipid lowering medication revealed significant correlations of LDL-C with CE (r = 0.42, p < 0.001), Cer (r = 0.364, p < 0.001), PC (r = 0.30, p = 0.002), sphingomyelin (SM) (r = 0.43, p < 0.001), and TG (r = 0.25, p = 0.01). ApoE protein levels were only significantly associated with total TG and DG (r = 0.43 and 0.49 respectively, p < 0.001, controlling for age, sex, BMI, and lipid lowering medications).

Associations of lipid classes with cognitive domains

Partial correlations of cognitive domains [14] measured during Wave 1 and lipid abundance, including LDL-C and ApoE (correcting for the effect of age, sex, years of education, APOE genotype, BMI, and lipid-lowering medication) were not significant (Supplementary Table 4), although there was a minor and positive partial correlation of total SM and LDL-C with the visuospatial domain (r = 0.20, p = 0.03, and r = 0.21, p = 0.04, respectively). These did not survive correction for multiple testing.

Association of lipid tertiles with APOE genotype and future cognitive decline

Binary logistic regression revealed significant associations of APOE genotype with the highest tertile of various lipid classes. Relative to ɛ3/ɛ3, ɛ2 carriers (Fig. 3A) had significantly higher odds of containing the highest tertile of total PE (OR = 5.60, 95% CI = 2.23, 13.5) and its subclasses, as well as the highest tertile of Cer(d16:1/X), lysophosphatidylethanolamine (LPE)(20:X) and LPE(22:X) (OR = 4.22, 2.81, and 4.80, respectively; 95% CI = 1.77, 10.1; 1.17, 6.76; and 2.01, 11.52, respectively). Relative to ɛ3/ɛ3 participants, ɛ4 carriers (Fig. 3B) had significantly higher odds of having the highest tertile of total SM (OR = 2.93, 95% CI = 1.24, 6.90) and its subclasses, and significantly reduced odds of having the highest tertile of total PI (OR = 0.33, 95% CI = 0.12, 0.92).

Fig.3

Odds ratios and confidence intervals for the highest tertile of lipids of ɛ2 and ɛ4 carriers relative to ɛ3/ɛ3 after controlling for age, sex, BMI and lipid-lowering medication usage. Error bars indicate the 95% confidence interval of the odds ratio of belonging to the highest tertile of each lipid class/subclass, relative to ɛ3/ɛ3, for (A) ɛ2 allele carriers, and (B) ɛ4 allele carriers (but excluding ɛ2/ɛ4). Odds for lipids where the confidence interval bars are completely left, or completely right of the OR = 1 line (i.e., not touching the line) are considered significant at p < 0.05, and are also marked with an asterisk.

For participants with consensus diagnosis data in both Wave 1 and Wave 4 of MAS (n = 97), we analyzed whether baseline tertiles of lipids could predict future cognitive decline (Supplementary Table 5). We defined converters (n = 22) as cognitively normal participants at baseline who progressed to MCI or AD, or MCI participants who progressed to AD, at the end of six years (Wave 4). Of 26 subjects with highest tertile of measured PE(18:1/X), only 2 of these subjects were converters (7.7%), compared against the lowest tertile where 20/71 subjects were converters (28.2%). This means 20/22 converters (90.9%) had the lower two tertiles of PE(18:1/X). Participants with the highest tertile of PE(18:1/X) had significantly reduced odds of converting, relative to participants in the lowest two tertiles of these lipid classes (OR = 0.17, 95% C.I. = [0.03, 0.92], p = 0.04), corresponding to roughly 5.9-fold reduced odds of conversion among subjects with the highest tertile of PE(18:1/X) lipids. Other PE subclasses examined were just below the threshold of significance (p = 0.07). No other lipid class tertiles, nor LDL-C or ApoE tertiles, were significantly associated with conversion to MCI or AD.

Cox regression analysis of progression to cognitive decline over time

Cox regression was applied to assess progression to decline across the study waves. A “decline” event was noted during the first wave at which subjects who were normal at Wave 1 progressed to MCI or dementia and did not improve back to normal by Wave 4, or the first wave at which subjects with MCI at Wave 1 progressed to dementia. Otherwise, a “non-decline” event was noted for the subject in question at the last timepoint at which the subject had available consensus diagnosis data.

The survival curves (time to decline) for the highest versus lowest tertile of PE(18:1/X) lipids are shown in Fig. 4. The survival curve for the lower two tertiles relative to the highest tertile of PE(18:1/X) was steeper but not significantly different (p = 0.23, HR = 2.025, 95% C.I. = [0.639, 6.418]. After including age, sex, APOE genotype, BMI, and lipid-lowering medication, the Cox regression model trended toward significance (p = 0.064, χ² = 16.19). Other lipid classes examined by Cox regression and were not significant by lipid tertile (p > 0.05).

Fig.4

Cox regression survival curves showing differences in rate of decline between subjects with the highest tertile of PE(18:1/X) at Wave 1 (green line), compared to the lower two tertiles (blue line). Survival refers to a successful non-decline event, i.e., subject normal at Wave 1 who did not decline to MCI or dementia, or an MCI subject at Wave 1 who improved to normal, or did not progress to dementia in subsequent waves.

DISCUSSION

APOE genotype is considered the strongest genetic risk factor for sporadic AD and has associations with cognitive change [14, 17]. In the present study, we attempt to understand the relationship between APOE genotype and the plasma lipidome in older aged participants taken from three independent population aging studies [10 –12]. Previously, we explored associations of various apolipoproteins, including ApoE, as well as the plasma lipidome with cognition and aging in elderly adults [14, 19]. Any differences in lipid profile related to APOE genotype could potentially explain how AD risk is modulated. A major finding of our study was that the ɛ2 allele is associated with higher levels of various phospholipid classes, particularly PE, and baseline PE levels may potentially differentiate between subjects who remain cognitively healthy and those who go on to develop MCI/AD.

Associations of APOE genotype with cholesterol and ApoE levels

Preliminary analyses did not identify statistically significant differences in LDL-C, HDL-C, total cholesterol, or TG between APOE genotypes. However, participants on lipid lowering medication, primarily statins, had reduced LDL-C, HDL-C, and total cholesterol independent of APOE genotype. Thus lipid-lowering medication is a possible confounder. Repeating the analysis with participants naïve for lipid-lowering medications uncovered a statistically significant dose dependent effect of APOE genotype on LDL-C, which has previously been described [22], and explains the relative high risk of dyslipidemias and coronary events in ɛ4 carriers relative to non-carriers [22], while ɛ2 carriers may be protected against hypercholesterolemia. The positive correlation of LDL-C with Cer and SM levels has previously been reported in our laboratory [19], though in the present study, we also found minor relationships of LDL-C with TG and PC, which was not seen in the previous study. The difference in associations for these lipid classes may be due to the fact the previous study focused purely on ɛ3/ɛ3 individuals with a smaller proportion of subjects on lipid-lowering medications. We did not find any relationship of LDL-C levels with PE lipids in either study, which could imply PE lipids are transported more abundantly in other lipoproteins relative to LDL [23].

Apart from LDL-C, plasma protein levels of ApoE have previously been associated with genotype, with ApoE highest among ɛ2 carriers relative to ɛ3/ɛ3 and ɛ4 carriers [1 , 17], which we confirmed in the present study. However, plasma ApoE levels were only associated with plasma TG and DG, and not with any other lipid classes, so APOE genotype could account for some additional variance in plasma lipids independent of ApoE levels.

APOE genotype-dependent lipidomic differences

Our study found relatively few changes in lipids among ɛ4 carriers compared to ɛ3/ɛ3 individuals. Levels of Cer(d16:1/X) were elevated in ɛ4 carriers relative to ɛ3/ɛ3 individuals. ɛ4 carriers also had lower total PI levels (though not significant) and odds of attaining the lowest tertile to PI relative to ɛ2 carriers. Significantly reduced PI among ɛ4 carriers relative to ɛ3/ɛ3 individuals could be important since PI-related metabolites have previously been implicated in clinical AD progression [24], modulated by the APOE ɛ4 allele [25]. ɛ4 carriers also had significantly higher odds of having the highest tertile of SM relative to ɛ3/ɛ3 which has been linked with increased AD risk among males over the age of 55 [26]. A small number of previous studies have also investigated whether the APOE allele is associated with plasma or brain lipids. APOE4 knock-in mice models failed to show major perturbations in brain lipid levels, compared to the ɛ2 and ɛ3 alleles [27], except in oxysterols [28], while another study in postmortem cerebrospinal fluid (CSF) samples among AD patients found no relationship between ɛ4 allele and lipids [28].

While there were minimal differences between the lipidome among ɛ4 carriers relative to ɛ3/ɛ3 participants, we identified substantial differences in the lipidome of ɛ2 allele carriers relative to ɛ3/ɛ3 and ɛ4 carriers, suggesting that the ɛ2 allele could be a strong determinant of plasma lipid levels in non-demented individuals [28]. In particular, PE and LPC lipid subclasses were elevated among ɛ2 carriers relative to ɛ3/ɛ3 and ɛ4 carriers, while PC levels were also higher among ɛ2 carriers relative to ɛ4 carriers.

The increase of plasma PE and LPC lipids occurs even though LDL-C levels are decreased among ɛ2 allele carriers, and implies that other lipoprotein particles such as HDL or VLDL may be involved [23]. ApoE is more abundant in HDL, and VLDL receptors bind equally well to the ɛ2 isoform as it does to ɛ3 [29] even though the LDL receptor binds with much lower affinity to ɛ2 compared to ɛ3. Among ɛ2/ɛ2 participants with the lowest LDL-C levels, LDL particles could have altered composition resembling more closely intermediate density lipoprotein (IDL), which includes increased phospholipid density [30]. Altered lipoprotein composition, as well as an increase in plasma ApoE concentrations among ɛ2 carriers could account for the elevation in phospholipids.

Could these lipidomic changes related to ɛ2 be protective?

Since ɛ2 is considered protective against atherosclerosis and AD, it is interesting to speculate whether the mechanism might involve ɛ2-associated elevation in phospholipids and lysophospholipids. Levels of PC and LPC, and the LPC to PC ratio are decreased in the CSF of AD participants [31]. Published work on blood lipidomics has consistently found differences in PC and LPC levels or their ratios in MCI and AD participants relative to healthy age-matched controls [32, 33], and recently, a 10 biomarker panel including several PCs was shown to predict phenoconversion from cognitive healthy to MCI and AD [34]. Since the PCs listed in this biomarker panel were all decreased in AD converters of the mentioned study relative to cognitively healthy non-converters, the higher levels of total PC and LPC found in our ɛ2 carriers could imply a lipidomic profile that is protective against cognitive decline and AD.

Previous postmortem lipidomic analysis has revealed reduced levels of PE and other lipids in AD brains, indicative of abnormal neuronal membrane repair, or degeneration [35, 36]. The established body of work in various tissues including brain, CSF, and blood implicates reduced phospholipid levels in AD and MCI. Therefore, the elevated levels of phospholipids, especially PE and LPCs, in our ɛ2 carriers supports the possibility that ɛ2 associated phospholipid alterations have a protective function against AD and possibly age-related cognitive decline. While we were unable to identify relationships between plasma lipid classes and cognitive domains, likely due to very gradual nature of cognitive changes, these lipids may still have an important role in maintaining cognitive health. Indeed, participants with the highest tertile of PE(18:1/X) at baseline had a nearly six-fold reduced odds of deteriorating in cognitive status to MCI or AD six years later. Unfortunately, the Cox regression analysis did not indicate differences between survival (time to decline) curves between the different tertile groups. This result may be due to relative small samples across waves available for this type of analysis. Nevertheless, the hazard ratio does suggest a reduced risk of decline among those with the highest PE(18:1/X) and supports the “endpoint” analysis between Wave 1 and Wave 4 performed using logistic regression.

Any protective mechanism of phospholipids is likely related to their role in maintaining brain membrane integrity and synaptic function [37, 38], and phospholipids are markedly reduced with ageing and dementia [35, 39]. Elevated baseline levels of phospholipids in plasma of ɛ2 carriers could confer protection against neurodegeneration and oxidative stress relative to ɛ3/ɛ3 and ɛ4 carriers [40, 41]. PE (especially plasmalogens) has recently been shown to combat neuro-inflammation and apoptosis and attenuate cognitive decline in an AD rat model [42], and regulate autophagy and enhance longevity in yeast and Drosophila [43], while the asymmetric distribution of PE and PC in plasma membranes is vital to maintaining membrane integrity [44], and PE may even modulate membrane disruption by amyloid proteins [45]. In humans, the protective capacity of phospholipids has recently been assessed in early clinical trials of a nutrient food supplement called Souvenaid (Fortasyn) [46, 47]. This supplement contains a combination of rate-limiting factors involved in the Kennedy pathway, required for PE and PC synthesis [48]. The rationale for using Souvenaid in early AD is that the nutrient combination would support synaptic formation and function [48], especially since many of these phospholipid forming substrates are reduced in both CSF and plasma of MCI and AD participants relative to healthy controls [49]. Souvenaid has been associated with enhanced cerebral blood flow, and also strengthens synapse formation and brain functional connectivity in both wild-type and APOE ɛ4 mice [46], as well as in humans [50]. Unfortunately, clinical trials have failed to meet primary endpoints with no notable improvement in global measures of cognition among subjects with moderate AD [51], and a more recent clinical trial found no significant improvements in its primary endpoint (neuropsychological test battery scores) conferred by Souvenaid [52]. Nevertheless, this study did meet two of its secondary endpoints in measures of cognition and brain atrophy, and the authors suspected that cognition decline was inherently lower among the subjects chosen, which could render the investigation underpowered. While the results of these trials should be treated with caution, the evidence does suggest these cofactors integrate well into the CNS and could be beneficial to subjects during the earliest stages of AD, as opposed to more moderate and severe forms of AD [51, 52], where neurodegeneration may be more difficult to control and reverse. Thus, establishing a suitable biomarker for detecting prodromal AD as early as possible is an important research priority. In this endeavor, the potential relationship between PE with predicted cognitive decline as described in this study represents a positive starting point for further research into how lipid profiles may be disrupted, contributing to AD [15, 53].

Limitations and future directions

Lipid lowering medications do affect levels LDL-C, confirming expected trends and could be a possible confounder. Nevertheless, the associations of lipids and their tertiles with APOE genotype were conserved after adjusting for lipid lowering medications, and their use conveyed no significant changes to the lipidomic profile. Association of lipids with future cognitive decline should be interpreted with caution due to the relatively small sample size of converters (identified after initial subject selection), though our results suggest the influence of PE, particularly reduced PE(18:1/X) lipids with increased risk of conversion, which should be explored in a larger set of subjects.

Lipids and apolipoproteins measured in plasma are not necessarily concordant with levels measured in the CSF or brain tissue [54], and our study did not measure lipids associating with ApoE directly, but suggests that there may be effects on the broad plasma lipidome. We also note from the literature that peripheral and central ApoE are considered distinct and that minimal, if any, peripheral ApoE bypasses the blood-brain barrier [55]. Therefore, it is difficult to conclude to what extent APOE genotype affects lipids in the CNS. Peripheral ApoE is generally expressed in the liver and macrophages, while the CNS synthesizes ApoE locally via astrocytes, and to a smaller extent in neurons. Nevertheless, CSF does exchange metabolites with the vascular system and this exchange may be elevated in case of increased blood-brain barrier permeability in disease, and many lipids and other metabolites in plasma have been shown to correlate with CSF levels [56] and affect CNS lipid metabolism [46]. Peripheral ApoE has also been shown to affect cognition independent of CNS ApoE levels [57], and is therefore important in cognitive health. Although distinct pools of ApoE regulate central and peripheral compartments, CNS ApoE is expressed at a high level, second only to the liver [58], suggesting a highly important role in brain lipid metabolism that could mirror their roles in peripheral metabolism [59]. The relationship between ApoE protein concentrations and APOE genotype in plasma is also observed centrally, with CSF ApoE levels decreasing in the order ɛ2/ɛ2> ɛ3/ɛ3> ɛ4/ɛ4 [60].

One known role of central ApoE, like its peripheral counterpart, is to remodel membranes by recruiting cholesterol and phospholipids (though without triglycerides, which are not present in the CNS) [1]. Further, CNS ApoE tends to be elevated in response to stress or injury [61], most likely to facilitate neuronal repair. It is interesting to note that ApoE particles in the CNS contain a much higher composition of PE lipids compared to their peripheral counterparts, though the reason for this is unclear [1, 59]. We hypothesize that this could be linked to the intimate role PE plays in promoting highly flexible neuronal membranes of the CNS. Taken together, these studies identify ApoE as an important regulator of cholesterol and phospholipid transport and distribution in both the periphery and the CNS, supporting not only cholesterol transport, but also membrane remodeling and repair. Altered concentrations of ApoE due to APOE genotype leads to differential functioning of ApoE in lipid transport, and given that PE is a large component of CNS apolipoproteins, could explain in part the differential risk of AD conferred by APOE genotype. Our results indicate that ɛ2 carriers have increased plasma phospholipids, and additional work is needed to confirm if this is also the case in CSF, where ApoE serves as the most abundant apolipoprotein involved in lipid transport relative to the periphery and is expressed in situ [54].

Conclusion

We investigated whether APOE genotype influences plasma lipid levels among participants without AD aged over 65 years. ɛ2 carriers had lower levels of LDL-C, followed by ɛ3/ɛ3 while ɛ4 carriers had the highest levels of LDL-C. There were few associations of ApoE4 and plasma lipids, except for an association with lower PI, while ɛ2 carriers had significantly increased levels of most phospholipids and LPC relative to ɛ3/ɛ3 and ɛ4 carriers. Logistic regression analyses showed that ɛ2 carriers had over four times the odds of having the highest tertile of PE lipids relative to ɛ3/ɛ3. Participants with the highest tertile of baseline PE had lower odds of converting from control to MCI or AD, or from MCI to AD, six years later. The pattern of lipid changes observed complements published data on lipid changes in AD. Collectively, our data suggest that a mechanism by which the APOE ɛ2 allele is protective against MCI and AD could be by its association with elevated levels of specific classes of phospholipids, particularly PE.

Footnotes

ACKNOWLEDGMENTS

The authors would like to acknowledge the contributions of Dr. Kristan Kang and Mahboobeh Hosseini, who provided data collection and biobanking for plasma samples, as well as Ms. Angela Russell and Dr. Sophia Dean, who provided administrative support. The authors also thank the Mark Wainwright Analytical Centre in conjunction with the Bioanalytical Mass Spectrometry Facility for providing use of the QExactive mass spectrometer, analytical grade reagents and data processing software. The authors would also like to thank CHeBA and SEALs Pathology (Prince of Wales Hospital) for handling and preparing MAS and SCS samples, and Prof. Peter Schofield for HCS samples, as well as the MAS, HCS. and SCS research teams and participants for making this project possible.

This work was supported by a National Health & Medical Research Council of Australia Program Grant (APP1054544). M.W.W. is the recipient of an Australian Postgraduate Award (APA) from the Australian Commonwealth Government. N.B. is the recipient of the Australian Research Council Postdoctoral Research Fellowship. The authors thank the Rebecca Cooper Medical Research Foundation for their ongoing financial support.

Authors’ disclosures available online ().

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

References

Mahley

(2016) Central nervous system lipoproteins: ApoE and regulation of cholesterol metabolism. Arterioscler Thromb Vasc Biol 36, 1305–1315.

Heffernan

, Chidgey

, Peng

, Masters

, Roberts

(2016) The neurobiology and age-related prevalence of the ɛ4 allele of apolipoprotein E in Alzheimer’s disease cohorts. J Mol Neurosci 60, 316–324.

Hernández-Nazará

, Ruiz-Madrigal

, Martínez-López

, Roman

, Panduro

(2008) Association of the epsilon 2 allele of APOE gene to hypertriglyceridemia and to early-onset alcoholic cirrhosis. Alcohol Clin Exp Res 32, 559–566.

Phillips

(2014) Apolipoprotein E isoforms and lipoprotein metabolism. IUBMB Life 66, 616–623.

Larifla

, Armand

, Bangou

, Blanchet-Deverly

, Numeric

, Fonteau

, Michel

C-T

, Ferdinand

, Bourrhis

, Vélayoudom-Céphise

F-L

(2017) Association of APOE gene polymorphism with lipid profile and coronary artery disease in Afro-Caribbeans. PLoS One 12, e0181620.

Corder

, Saunders

, Strittmatter

, Schmechel

, Gaskell

, Small

, Roses

, Haines

, Pericak-Vance

(1993) Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science 261, 921–923.

Castellano

, Kim

, Stewart

, Jiang

, DeMattos

, Patterson

, Fagan

, Morris

, Mawuenyega

, Cruchaga

, Goate

, Bales

, Paul

, Bateman

, Holtzman

(2011) Human apoE isoforms differentially regulate brain amyloid-beta peptide clearance. Sci Transl Med 3, 89ra57.

Liao

, Yoon

, Kim

(2017) Apolipoprotein E metabolism and functions in brain and its role in Alzheimer’s disease. Curr Opin Lipidol 28, 60–67.

Kim

, Basak

, Holtzman

(2009) The role of apolipoprotein E in Alzheimer’s disease. Neuron 63, 287–303.

10.

Sachdev

, Levitan

, Crawford

, Sidhu

, Slavin

, Richmond

, Kochan

, Brodaty

, Wen

, Kang

, Mather

(2013) The Sydney Centenarian Study: Methodology and profile of centenarians and near-centenarians. Int Psychogeriatr 25, 993–1005.

11.

McEvoy

, Smith

, D’Este

, Duke

, Peel

, Schofield

, Scott

, Byles

, Henry

, Ewald

, Hancock

, Smith

, Attia

(2010) Cohort profile: The Hunter Community Study. Int J Epidemiol 39, 1452–1463.

12.

Sachdev

, Brodaty

, Reppermund

, Kochan

, Trollor

, Draper

, Slavin

, Crawford

, Kang

, Broe

(2010) The Sydney Memory and Ageing Study (MAS): Methodology and baseline medical and neuropsychiatric characteristics of an elderly epidemiological non-demented cohort of Australians aged 70–90 years. Int Psychogeriatr 22, 1248–1264.

13.

Lipnicki

, Crawford

, Kochan

, Trollor

, Draper

, Reppermund

, Maston

, Mather

, Brodaty

, Sachdev

(2017) Risk factors for mild cognitive impairment, dementia and mortality: The Sydney Memory and Ageing Study. J Am Med Dir Assoc 18, 388–395.

14.

Muenchhoff

, Song

, Poljak

, Crawford

, Mather

, Kochan

, Yang

, Trollor

, Reppermund

, Maston

, Theobald

, Kirchner-Adelhardt

, Kwok

, Richmond

, McEvoy

, Attia

, Schofield

, Brodaty

, Sachdev

(2017) Plasma apolipoproteins and physical and cognitive health in very old individuals. Neurobiol Aging 55, 49–60.

15.

Wong

, Braidy

, Poljak

, Pickford

, Thambisetty

, Sachdev

(2017) Dysregulation of lipids in Alzheimer’s disease and their role as potential biomarkers. Alzheimers Dement 13, 810–827.

16.

Burla

, Arita

, Bendt

, Cazenave-Gassiot

, Dennis

, Ekroos

, Han

, Ikeda

, Liebisch

, Lin

, Loh

, Meikle

, Oresic

, Quehenberger

, Shevchenko

, Torta

, Wakelam

MJO

, Wheelock

, Wenk

(2018) MS-based lipidomics of human blood plasma: A community-initiated position paper to develop accepted guidelines. J Lipid Res 59, 2001–2017.

17.

Song

, Poljak

, Crawford

, Kochan

, Wen

, Cameron

, Lux

, Brodaty

, Mather

, Smythe

, Sachdev

(2012) Plasma apolipoprotein levels are associated with cognitive status and decline in a community cohort of older individuals. PLoS One 7, e34078.

18.

Alshehry

, Barlow

, Weir

, Zhou

, McConville

, Meikle

(2015) An efficient single phase method for the extraction of plasma lipids. Metabolites 5, 389–403.

19.

Wong

MWK

, Braidy

, Pickford

, Vafaee

, Crawford

, Muenchhoff

, Schofield

, Attia

, Brodaty

, Sachdev

, Poljak

(2019) Plasma lipidome variation during the second half of the human lifespan is associated with age and sex but minimally with BMI. PLoS One 14, e0214141.

20.

Benjamini

, Hochberg

(1995) Controlling the False discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 57, 289–300.

21.

Fahy

, Subramaniam

, Murphy

, Nishijima

, Raetz

, Shimizu

, Spener

, van Meer

, Wakelam

, Dennis

(2009) Update of the LIPID MAPS comprehensive classification system for lipids. J Lipid Res 50 Suppl, S9–14.

22.

Bennet

, Di Angelantonio

, Ye

, Wensley

, Dahlin

, Ahlbom

, Keavney

, Collins

, Wiman

, de Faire

, Danesh

(2007) Association of apolipoprotein E genotypes with lipid levels and coronary risk. JAMA 298, 1300–1311.

23.

Dashti

, Kulik

, Hoek

, Veerman

, Peppelenbosch

, Rezaee

(2011) A phospholipidomic analysis of all defined human plasma lipoproteins. Sci Rep 1, 139.

24.

Bennett

SAL

, Valenzuela

, Xu

, Franko

, Fai

, Figeys

(2013) Using neurolipidomics to identify phospholipid mediators of synaptic (dys)function in Alzheimer’s disease. Front Physiol 4, 168.

25.

Zhu

, Zhong

, Elder

, Sano

, Holtzman

, Gandy

, Cardozo

, Haroutunian

, Robakis

, Cai

(2015) Phospholipid dysregulation contributes to ApoE4-associated cognitive deficits in Alzheimer’s disease pathogenesis. Proc Natl Acad Sci U S A 112, 11965–11970.

26.

Mielke

, Haughey

, Han

, An

, Bandaru

VVR

, Lyketsos

, Ferrucci

, Resnick

(2017) The association between plasma ceramides and sphingomyelins and risk of Alzheimer’s disease differs by sex and APOE in the Baltimore Longitudinal Study of Aging. J Alzheimers Dis 60, 819–828.

27.

Sharman

, Shui

, Fernandis

, Lim

, Berger

, Hone

, Taddei

, Martins

, Ghiso

, Buxbaum

, Gandy

, Wenk

, Martins

(2010) Profiling brain and plasma lipids in human APOE epsilon2, epsilon3, and epsilon4 knock-in mice using electrospray ionization mass spectrometry. J Alzheimers Dis 20, 105–111.

28.

Jenner

, Lim

WLF

, Ng

MPE

, Wenk

, Shui

, Sharman

, Gandy

, Martins

(2010) The effect of APOE genotype on brain levels of oxysterols in young and old human APOE ɛ2, ɛ3 and ɛ4 knock-in mice. Neuroscience 169, 109–115.

29.

Takahashi

, Oida

, Ookubo

, Suzuki

, Kohno

, Murase

, Yamamoto

, Nakai

(1996) Very low density lipoprotein receptor binds apolipoprotein E2/2 as well as apolipoprotein E3/3. FEBS Lett 386, 197–200.

30.

Murdoch

, Boright

, Paterson

, Zinman

, Steffes

, Cleary

, Edwards

, Marcovina

, Purnell

, Brunzell

(2007) LDL composition in E2/2 subjects and LDL distribution by Apo E genotype in type 1 diabetes. Atherosclerosis 192, 138–147.

31.

Mulder

, Wahlund

, Teerlink

, Blomberg

, Veerhuis

, van Kamp

, Scheltens

, Scheffer

(2003) Decreased lysophosphatidylcholine/phosphatidylcholine ratio in cerebrospinal fluid in Alzheimer’s disease. J Neural Transm (Vienna) 110, 949–955.

32.

Whiley

, Sen

, Heaton

, Proitsi

, Garcia-Gomez

, Leung

, Smith

, Thambisetty

, Kloszewska

, Mecocci

, Soininen

, Tsolaki

, Vellas

, Lovestone

, Legido-Quigley

(2014) Evidence of altered phosphatidylcholine metabolism in Alzheimer’s disease. Neurobiol Aging 35, 271–278.

33.

Klavins

, Koal

, Dallmann

, Marksteiner

, Kemmler

, Humpel

(2015) The ratio of phosphatidylcholines to lysophosphatidylcholines in plasma differentiates healthy controls from patients with Alzheimer’s disease and mild cognitive impairment. Alzheimers Dement (Amst) 1, 295–302.

34.

Mapstone

, Cheema

, Fiandaca

, Zhong

, Mhyre

, MacArthur

, Hall

, Fisher

, Peterson

, Haley

, Nazar

, Rich

, Berlau

, Peltz

, Tan

, Kawas

, Federoff

(2014) Plasma phospholipids identify antecedent memory impairment in older adults. Nat Med 20, 415–418.

35.

Kosicek

, Hecimovic

(2013) Phospholipids and Alzheimer’s disease: Alterations, mechanisms and potential biomarkers. Int J Mol Sci 14, 1310–1322.

36.

Gonzalez-Dominguez

, Garcia-Barrera

, Gomez-Ariza

(2014) Combination of metabolomic and phospholipid-profiling approaches for the study of Alzheimer’s disease. J Proteomics 104, 37–47.

37.

Chirackal Manavalan

, Kober

, Metso

, Lang

, Becker

, Hasslitzer

, Zandl

, Fanaee-Danesh

, Pippal

, Sachdev

, Kratky

, Stefulj

, Jauhiainen

, Panzenboeck

(2014) Phospholipid transfer protein is expressed in cerebrovascular endothelial cells and involved in high density lipoprotein biogenesis and remodeling at the blood-brain barrier. J Biol Chem 289, 4683–4698.

38.

Nguyen

, Ma

, Shui

, Wong

, Cazenave-Gassiot

, Zhang

, Wenk

, Goh

, Silver

(2014) Mfsd2a is a transporter for the essential omega-3 fatty acid docosahexaenoic acid. Nature 509, 503–506.

39.

Prasad

, Lovell

, Yatin

, Dhillon

, Markesbery

(1998) Regional membrane phospholipid alterations in Alzheimer’s disease. Neurochem Res 23, 81–88.

40.

Miyata

, Smith

(1996) Apolipoprotein E allele-specific antioxidant activity and effects on cytotoxicity by oxidative insults and beta-amyloid peptides. Nat Genet 14, 55–61.

41.

Rasmiena

, Barlow

, Stefanovic

, Huynh

, Tan

, Sharma

, Tull

, de Haan

, Meikle

(2015) Plasmalogen modulation attenuates atherosclerosis in ApoE- and ApoE/GPx1-deficient mice. Atherosclerosis 243, 598–608.

42.

Che

, Li

, Zhang

, Ding

, Zhang

, Shi

, Yanagita

, Xue

, Chang

, Wang

(2018) A comparative study of EPA-enriched ethanolamine plasmalogen and EPA-enriched phosphatidylethanolamine on Aβ42 induced cognitive deficiency in a rat model of Alzheimer’s disease. Food Funct 9, 3008–3017.

43.

Rockenfeller

, Koska

, Pietrocola

, Minois

, Knittelfelder

, Sica

, Franz

, Carmona-Gutierrez

, Kroemer

, Madeo

(2015) Phosphatidylethanolamine positively regulates autophagy and longevity. Cell Death Differ 22, 499.

44.

, Agellon

, Allen

, Umeda

, Jewell

, Mason

, Vance

(2006) The ratio of phosphatidylcholine to phosphatidylethanolamine influences membrane integrity and steatohepatitis. Cell Metab 3, 321–331.

45.

Sciacca

MFM

, Brender

, Lee

D-K

, Ramamoorthy

(2012) Phosphatidylethanolamine enhances amyloid fiber-dependent membrane fragmentation. Biochemistry 51, 7676–7684.

46.

Wiesmann

, Zerbi

, Jansen

, Haast

, Lutjohann

, Broersen

, Heerschap

, Kiliaan

(2016) A dietary treatment improves cerebral blood flow and brain connectivity in aging apoE4 mice. Neural Plast 2016, 6846721.

47.

Rijpma

, van der Graaf

, Lansbergen

, Meulenbroek

, Cetinyurek-Yavuz

, Sijben

, Heerschap

, Olde Rikkert

MGM

(2017) The medical food Souvenaid affects brain phospholipid metabolism in mild Alzheimer’s disease: Results from a randomized controlled trial. Alzheimers Res Ther 9, 51.

48.

van Wijk

, Broersen

, de Wilde

, Hageman

, Groenendijk

, Sijben

, Kamphuis

(2014) Targeting synaptic dysfunction in Alzheimer’s disease by administering a specific nutrient combination. J Alzheimers Dis 38, 459–479.

49.

van Wijk

, Slot

RER

, Duits

, Strik

, Biesheuvel

, Sijben

JWC

, Blankenstein

, Bierau

, van der Flier

, Scheltens

, Teunissen

(2017) Nutrients required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer’s disease dementia. Alzheimers Dement (Amst) 8, 139–146.

50.

de Waal

, Stam

, Lansbergen

, Wieggers

, Kamphuis

PJGH

, Scheltens

, Maestú

, van Straaten

ECW

(2014) The effect of Souvenaid on functional brain network organisation in patients with mild Alzheimer’s disease: A randomised controlled study. PLoS One 9, e86558.

51.

Onakpoya

, Heneghan

(2015) The efficacy of supplementation with the novel medical food, Souvenaid, in patients with Alzheimer’s disease: A systematic review and meta-analysis of randomized clinical trials. Nutr Neurosci 20, 219–227.

52.

Soininen

, Solomon

, Visser

, Hendrix

, Blennow

, Kivipelto

, Hartmann

(2017) 24-month intervention with a specific multinutrient in people with prodromal Alzheimer’s disease (LipiDiDiet): A randomised, double-blind, controlled trial. Lancet Neurol 16, 965–975.

53.

Wong

, Braidy

, Poljak

, Sachdev

(2017) The application of lipidomics to biomarker research and pathomechanisms in Alzheimer’s disease. Curr Opin Psychiatry 30, 136–144.

54.

Koch

, Furtado

, Falk

, Leypoldt

, Mukamal

, Jensen

(2017) Apolipoproteins and their subspecies in human cerebrospinal fluid and plasma. Alzheimers Dement (Amst) 6, 182–187.

55.

Linton

, Gish

, Hubl

, Butler

, Esquivel

, Bry

, Boyles

, Wardell

, Young

(1991) Phenotypes of apolipoprotein B and apolipoprotein E after liver transplantation. J Clin Invest 88, 270–281.

56.

Trushina

, Dutta

, Persson

X-MT

, Mielke

, Petersen

(2013) Identification of altered metabolic pathways in plasma and CSF in mild cognitive impairment and Alzheimer’s disease using metabolomics. PLoS One 8, e63644.

57.

Lane-Donovan

, Wong

, Durakoglugil

, Wasser

, Jiang

, Xian

, Herz

(2016) Genetic restoration of plasma ApoE improves cognition and partially restores synaptic defects in ApoE-deficient mice. J Neurosci 36, 10141–10150.

58.

Elshourbagy

, Liao

, Mahley

, Taylor

(1985) Apolipoprotein E mRNA is abundant in the brain and adrenals, as well as in the liver, and is present in other peripheral tissues of rats and marmosets. Proc Natl Acad Sci U S A 82, 203.

59.

Pitas

, Boyles

, Lee

, Hui

, Weisgraber

(1987) Lipoproteins and their receptors in the central nervous system. Characterization of the lipoproteins in cerebrospinal fluid and identification of apolipoprotein B,E(LDL) receptors in the brain. J Biol Chem 262, 14352–14360.

60.

Riddell

, Zhou

, Atchison

, Warwick

, Atkinson

, Jefferson

, Xu

, Aschmies

, Kirksey

, Hu

, Wagner

, Parratt

, Xu

, Li

, Zaleska

, Jacobsen

, Pangalos

, Reinhart

(2008) Impact of apolipoprotein E (ApoE) polymorphism on brain ApoE levels. J Neurosci 28, 11445.

61.

, Bernardo

, Walker

, Kanegawa

, Mahley

, Huang

(2006) Profile and Regulation of Apolipoprotein E (ApoE) expression in the CNS in mice with targeting of green fluorescent protein gene to the ApoE locus. J Neurosci 26, 4985.