Age-Dependent Changes to Sphingolipid Balance in the Human Hippocampus are Gender-Specific and May Sensitize to Neurodegeneration

Abstract

The greatest risk factor for developing Alzheimer’s disease (AD) is aging. The major genetic risk factor for AD is the ɛ4 allele of the APOE gene, encoding the brain’s major lipid transport protein, apolipoprotein E (ApoE). The research community is yet to decipher why the ApoE4 variant pre-disposes to AD, and how aging causes the disease. Studies have shown deregulated levels of sphingolipids, including decreased levels of the neuroprotective signaling lipid sphingosine 1-phosphate (S1P), and increased ceramide content, in brain tissue and serum of people with pre-clinical or very early AD. In this study we investigated whether sphingolipid levels are affected as a function of age or APOE genotype, in the hippocampus of neurologically normal subjects over the age of 65. Lipids were quantified in 80 postmortem tissue samples using liquid chromatography tandem mass spectrometry (LC-MS/MS). Sphingolipid levels were not significantly affected by the presence of one ɛ4 or ɛ2 allele. However, ceramide, sphingomyelin, and sulfatide content was very significantly correlated with age in the hippocampus of males. On the other hand, S1P, normalized to its non-phosphorylated precursor sphingosine, was inversely correlated with age in females. Our results therefore establish gender-specific differences in sphingolipid metabolism in the aging human brain. Ceramide is a pro-apoptotic lipid, and heavily implicated as a driver of insulin resistance in metabolic tissues. S1P is a neuroprotective lipid that supports glutamatergic neurotransmission. Increasing ceramide and decreasing S1P levels may contribute significantly to a pro-neurodegenerative phenotype in the aging brain.

Keywords

Aging Alzheimer’s disease APOE ApoE ceramide sphingolipid sphingomyelin sphingosine 1-phosphate

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia, accounting for over half of all dementia cases [1]. Age is the greatest risk factor for developing the disease, and accordingly, the incidence of AD is expected to double every 20 years as life expectancy increases and populations age [2]. The most prevalent genetic risk factor for late-onset AD, which accounts for >99% of AD cases, is inheritance of the ɛ4 allele of the APOE gene [1 , 4]. There are three major alleles of the APOE gene, ɛ2, ɛ3, and ɛ4, of which ɛ3 is most common [1, 5]. Homozygous carriers of the ɛ4 allele are at ∼12-fold increased risk of developing AD compared with non-carriers, while for heterozygous carriers the risk is ∼3-fold higher than the general population [3, 6]. Conversely, the risk of developing AD is lower in carriers of the ɛ2 allele [5, 7]. A number of mechanisms have been proposed to explain why APOE genotype modulates AD risk, including a proposed role in the clearance of neurotoxic amyloid-β peptides [3, 4]. However, the primary function for ApoE is as a component of lipoprotein particles that transport lipids in the circulation and central nervous system. ApoE is the major brain lipoprotein, with a particular role as a ligand for lipoprotein receptors [8]. Allelic variants of other genes involved in lipid metabolism and transport, notably CLU, encoding the lipoprotein ApoJ, and the membrane transporter ABCA7, are also major modulators of AD risk [4], further highlighting the likelihood that altered lipid metabolism is a significant driving influence for AD.

The sphingolipids are a major family of lipids defined by their long chain base backbone, generally 18-carbon sphingosine or dihydrosphingosine in mammals [9]. Ceramide is the central metabolite in the sphingolipid pathway (Fig. 1), formed by addition of a variable length fatty acyl chain (C14 – C26) to the amine group of sphingosine or dihydrosphingosine. This reaction is catalyzed by the ceramide synthases, of which there are six isoforms in mammals (CerS1-6) [9, 10]. Ceramide is a pro-apoptotic lipid capable of promoting a neurodegenerative phenotype [10, 11], and is also heavily implicated in the pathogenesis of insulin resistance [10, 12].

Fig.1

Overview of the sphingolipid metabolic pathway.

The diversity in sphingolipid structures results from the different headgroups that can be added to the primary hydroxyl group of ceramide (Fig. 1), as well as the diversity of fatty acids that can be incorporated into the core ceramide molecule. Addition of a galactose headgroup to ceramide by galactosylceramide synthase forms galactosylceramide (GalCer), which can be further sulfated to yield sulfatide (ST); while the addition of a phosphocholine group to ceramide forms the most abundant of the sphingolipids, sphingomyelin (SM). ST, GalCer, and SM are major constituents of myelin [13], and essential for normal neurological function [14, 15].

Several studies have demonstrated significant changes to brain sphingolipids in the prodromal stages of AD, including depletion of myelin enriched sphingolipids [16, 17] and elevated levels of ceramides, SM, and sphingosine [16 , 18–22]. Longitudinal studies have demonstrated increased levels of specific ceramide and SM species in serum preceding memory impairment [23], and a higher risk of developing AD in women with high serum ceramide levels [24]. Ceramide, SM, and hexosylceramide (HexCer) levels also increase with age in the mouse brain [25, 26]. We have previously demonstrated loss of the neuroprotective sphingolipid S1P in the hippocampus CA1 region and inferior temporal grey matter of subjects with Braak stage III/IV pathology, corresponding to an early stage of AD pathogenesis [27]. Levels of S1P, normalized to its non-phosphorylated precursor sphingosine (i.e., the S1P/sphingosine ratio), were lower in the hippocampus of APOE ɛ4 allele carriers, suggesting that APOE genotype may directly affect S1P levels, however our sample size was small. The hippocampus is vulnerable to age-related changes in dendritic spine density and plasticity, and is one of the first brain regions to exhibit the pathological hallmarks of AD (amyloid-β plaques and tau hyperphosphorylation), together with reduced metabolic activity and neuronal loss [28 –31]. This makes the hippocampus a good region to determine whether perturbations in sphingolipid metabolism could serve as precursors to AD. In the present study, we investigated whether sphingolipid levels are altered as a function of age and APOE genotype in the hippocampus of cognitively normal donors. APOE genotype was not found to have a significant influence on sphingolipid levels, but highly significant correlations between age and the levels of multiple sphingolipid species were observed in elderly male subjects, while an age-dependent decline in S1P:sphingosine balance was observed in females.

MATERIALS AND METHODS

Human brain tissue samples

Human brain tissue samples were obtained from the New South Wales Brain Tissue Resource Centre (NSW BTRC) and Queensland Brain Bank (n = 40 from each). Frozen tissue samples were taken from the CA1 region of the hippocampus. Ethics approval was from the University of New South Wales Human Research Ethics Committee (HREC13120). Table 1 summarizes brain donor demographics, postmortem interval (PMI), and Braak stage. Braak staging of donor brains was performed according to published criteria [32]. Brain tissue was stored at –80°C until required for analysis.

Table 1

Donor demographics. Mean age and PMI were compared among the three major genotypes, ɛ3/ɛ3, ɛ2/ɛ3, and ɛ3/ɛ4, using ANOVA with Tukey’s post-test, as described in the results text

APOE genotype	ɛ3/ɛ3	ɛ2/ɛ3	ɛ3/ɛ4	ɛ2/ɛ4	ɛ2/ɛ2	Total
Number (%)	50 (63%)	10 (13%)	14 (18%)	4 (5%)	2 (3%)	80
Age (mean ± SD)	78 ± 6.8	82.6 ± 12.8	74.8 ± 7.2	78.8 ± 11.9	83.5 ± 12.0	78.2 ± 8.3
PMI (mean ± SD)	25.0 ± 17.8	41.7 ± 28.0	26.8 ± 17.8	23.3 ± 6.2	49.5 ± 48.8	27.9 ± 20.3
Gender (%)
male	62	50	57	75	0	60
female	38	50	43	25	100	40
Braak stage (n)
Not reported	2	0	4	1	2	9
0	11	4	3	1	0	19
I	25	2	5	2	0	34
II	11	4	2	0	0	17
III	1	0	0	0	0	1

APOE genotyping

Genomic DNA was extracted from 50–100 mg brain tissue using phenol/chloroform extraction. In brief, brain tissues were homogenized in lysis buffer (10 mM Tris, pH 7.8, 1 mM EDTA, 100 mM NaCl, and 1% w/v SDS), containing 0.5 mg/ml Proteinase K (Roche, North Ryde, Australia), and incubated overnight at 55°C. Samples were centrifuged (13,000 rpm, 5 min) and supernatant transferred to Eppendorf tubes containing 450μL of phenol: chloroform: isoamyl alcohol (25:24:1). Samples were mixed and centrifuged (13,000 rpm, 5 min) to produce a biphasic emulsion. The upper, aqueous layer was collected and transferred to an RNAse-free 1.5 mL tube. The phenol: chloroform: isoamyl alcohol step was repeated. Ice cold ethanol (900μL) and sodium acetate (45μL) were added to the supernatant, and samples placed at –80°C for 1 h. Samples were centrifuged (13,000 rpm, 30 min), supernatant removed, and the remaining DNA pellet re-suspended in 80μL Tris-EDTA buffer (100 mM Tris, 10 mM disodium EDTA).

Genomic DNA was genotyped for the presence of the three APOE variants (ɛ2, ɛ3, and ɛ4) based on TaqMan SNP genotyping assays for rs7412 (C 904973) and rs429358 (C 3084793) according to the manufacturer’s instructions (AB Applied Biosystems by Life Technologies, Victoria, Australia). Five percent of the samples were genotyped in duplicate, with 100% inter- and intra-assay concordance [33].

Lipid standards and solvents

Synthetic lipid standards were purchased from Avanti Polar Lipids, Inc. (Alabaster, USA). Mass spectrometry grade ammonium formate, formic acid, ethyl acetate, and 2-propanol were from Sigma-Aldrich (Castle Hill, Australia). Analytical grade methanol was purchased from Merck (Bayswater, Australia).

Sphingolipid extraction and quantification by liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Sphingolipids were extracted from approximately 20 mg brain tissue using a single phase extraction protocol with ethyl acetate, isopropanol, and water [34]. Brain tissue was pulverized over dry ice, and an internal standard cocktail comprising 250 pmoles GalCer (d18:1/12:0), 250 pmoles ST (d18:1/12:0), 1250 pmoles SM (d18:1/12:0), and 50 pmoles each of ceramide (d18:1/17:0), S1P (d17:1) and sphingosine (d17:1) was added at the start of the extraction process [34]. Samples were resolved on a 3×150 mm Agilent XDB-C8 HPLC column and quantified using multiple reaction monitoring (MRM) on a Thermo Fisher Scientific Quantum Access, as described previously [34, 35]. Run time per sample was 30 min and LC-MS/MS analysis was divided into two periods. Period 1 (0–7.5 min) scanned for sphingosine and S1P (d17:1, d18:0, d18:1), while the ceramide, GalCer, ST and SM (d18:1/12:0 –d18:1/26:0, d18:1/16:1 – d18:1/26:1) scan events were monitored in period 2 (7.5–30 min). Transitions monitored for specific sphingolipid species are listed in Supplementary Table 1.

Data processing was performed using our Metabolite Mass Spectrometry Analysis Tool [35] and the detected sphingolipid species were verified as conforming to a previously identified quadratic elution profile [34]. Individual lipid peak areas were normalized to the appropriate class-specific internal standard, and converted to a pmole amount using a single external standard curve for each lipid class. Lipid amounts in each sample were then expressed per mg of tissue extracted.

Statistical analysis

Lipid levels and gene expression values were log-transformed (natural log) to create normal distributions, as the untransformed data was not normally distributed. Correlations between age and lipids, or age and relative gene expression levels, were analyzed by Pearson analysis, using GraphPad PRISM software. All p values were adjusted for multiple comparisons using the false discovery rate method of Benjamini and Hochberg, with Q = 5%. For the false discovery rate correction, all p values in Table 2 were assessed together for consistency. Mean lipid levels in different APOE genotype sample groups were compared using one-way ANOVA, adjusting for age as a co-variate. Tukey’s post-test was applied to compare means of each genotype group, using SPSS software. The effect of Braak stage (0, I, or II) on sphingolipid levels was investigated by linear regression analysis, using each lipid as the response variable, and genotype, gender, age, PMI, and Braak stage as co-variates. Linear modelling was performed with the LIMMA package using R statistical software. p values were corrected for false discovery rate using the Benjamini-Hochberg method.

Table 2

Correlations between sphingolipids and age

Lipid	Male		Female		Both male and female
	r	p value	r	p value	r	p value
Total Ceramide	0.43	0.0022 ^*	–0.24	0.19	0.19	0.09
C16:0 Cer	0.43	0.0026 ^*	0.08	0.65	0.33	0.0028 ^*
C18:0 Cer	0.20	0.17	0.07	0.71	0.17	0.13
C20:0 Cer	–0.08	0.59	0.28	0.12	0.03	0.79
C22:0 Cer	0.13	0.40	–0.10	0.59	0.0005	1.00
C22:1 Cer	0.44	0.0018 ^*	–0.35	0.05	0.12	0.28
C24:0 Cer	0.28	0.05	–0.36	0.05	0.04	0.72
C24:1 Cer	0.47	0.0007 ^*	–0.37	0.04	0.14	0.20
Total GalCer	–0.02	0.90	–0.05	0.79	–0.04	0.75
C16:0 GalCer	0.16	0.27	–0.20	0.27	0.04	0.75
C18:0 GalCer	0.26	0.08	–0.25	0.17	0.07	0.56
C22:0 GalCer	–0.13	0.37	0.01	0.94	–0.10	0.39
C22:1 GalCer	0.14	0.35	–0.17	0.37	0.02	0.88
C24:0 GalCer	–0.05	0.76	–0.01	0.96	–0.02	0.85
C24:1 GalCer	0.07	0.66	–0.08	0.66	0.02	0.85
C24:0-OH GalCer	–0.32	0.03	0.16	0.39	–0.17	0.13
C24:1-OH GalCer	–0.08	0.61	0.08	0.67	–0.02	0.88
Total SM	0.40	0.0045 ^*	–0.12	0.53	0.19	0.08
C16:0 SM	0.48	0.0005 ^*	–0.11	0.53	0.26	0.02
C16:1 SM	0.40	0.0048 ^*	–0.36	0.04	0.13	0.26
C18:0 SM	0.16	0.28	0.05	0.79	0.13	0.24
C18:1 SM	0.36	0.01	–0.31	0.09	0.09	0.45
C20:0 SM	–0.08	0.61	0.35	0.05	0.10	0.37
C22:0 SM	0.22	0.14	0.19	0.29	0.22	0.04
C22:1 SM	0.51	0.0002 ^*	–0.07	0.72	0.32	0.0043 ^*
C24:0 SM	0.27	0.06	–0.24	0.19	0.07	0.53
C24:1 SM	0.32	0.03	–0.14	0.43	0.16	0.14
Total ST	0.51	0.0002 ^*	–0.31	0.08	0.18	0.11
C16:0 ST	–0.02	0.92	–0.05	0.77	–0.03	0.83
C18:0 ST	0.11	0.45	–0.21	0.25	–0.02	0.86
C22:0 ST	0.44	0.0019 ^*	–0.27	0.13	0.12	0.27
C22:1 ST	0.47	0.0010 ^*	–0.33	0.06	0.11	0.33
C24:0 ST	0.43	0.0026 ^*	–0.37	0.04	0.05	0.68
C24:1 ST	0.52	0.0001 ^*	–0.28	0.12	0.21	0.06
C24:0-OH ST	0.27	0.07	–0.26	0.15	0.04	0.75
C24:1-OH ST	0.49	0.0004 ^*	–0.25	0.18	0.20	0.07
Sphingosine	0.27	0.07	0.03	0.87	0.20	0.07
S1P	0.25	0.09	–0.44	0.01	0.01	0.91
S1P/Sphingosine	0.07	0.61	–0.52	0.0020 ^*	–0.14	0.22

Pearson correlation analysis was used to determine associations between age and lipid levels (log-transformed). Coefficients of correlation (r) and p value are shown. Lipids significant after adjusting p values for multiple comparisons (Benjamini-Hochberg correction) are marked with an asterisk.

Permutation testing: To investigate the effects of random sampling, permutation testing (100,000 permutations) of the Pearson’s product-moment correlation test was performed for the 15 lipid measures that were significantly correlated with age in males. Testing was performed on male and female patients independently to investigate the effects of random sampling. Permutation testing was performed using the perm.cor.test function of the RVAideMemoire package using R statistical software.

RESULTS

Donor demographics

The present study included 80 post-mortem hippocampal tissue samples from cognitively normal donors who showed no signs of neurological disease at the time of death. The mean age at death was 78 ± 8 years, and the cohort was comprised 60% of males (mean age = 77 ± 8) and 40% of females (mean age = 81 ± 8). Braak staging confirmed that neurofibrillary tangles (NFTs) were either absent or restricted to the entorhinal region (Braak I/II) in all but one case (Braak III), ruling out any influence of AD pathology in the hippocampus. Braak staging was not available for nine cases. Subjects were divided into five groups based on APOE genotype. Baseline characteristics for each genotype group are presented in Table 1. The distribution of APOE genotypes was similar to previous reports [3, 5], with the majority being ɛ3/ɛ3 (63%), followed by ɛ3/ɛ4 (18%), then ɛ2/ɛ3 (13%). Mean age at death was not significantly different between the three major APOE genotypes (F = 2.85, p = 0.064 by one-way ANOVA). PMI was also not significantly affected as a function of APOE gentoype (F = 3.1; p = 0.051 by one-way ANOVA), although mean PMI was higher in the ɛ2/ɛ3 compared to the ɛ3/ɛ3 group (p = 0.04 by Tukey’s post-test). There was no association between PMI and age at death (r = –0.157; p = 0.169).

We first tested for associations between PMI and lipid levels to determine if PMI must be accounted for in subsequent analyses (Supplementary Table 1). At a significance level of p < 0.01, only the low abundance signaling lipid S1P (inverse correlation, r = –0.30; p = 0.007) was affected by PMI. This association was not significant after adjusting for multiple comparisons. We previously showed that the S1P:sphingosine ratio, which is indicative of the balance between sphingosine kinase and phosphatase activity, is significantly reduced with increasing postmortem AD pathology and may be influenced by APOE genotype [27]. There was a clear effect of PMI on S1P:sphingosine ratio (r = –0.41; p = 0.0002), which was significant after adjusting for multiple comparisons.

Gender-specific effects of aging on hippocampal sphingolipids

As age at death differed significantly among the APOE genotype groups, we tested the relationship between age and hippocampal sphingolipids in univariate analyses (Table 2). Considering both males and females together, there were no significant correlations between age at death and total levels of GalCer, ST, and S1P. Total ceramide, SM and sphingosine levels showed a trend increase with age, but this was not significant (p = 0.09, 0.09, and 0.07, respectively). Gender is an important factor in the risk for developing dementia and other diseases, and has a profound influence on biochemical and physiological processes [36]. When analyzed separately, total ceramide, SM, and ST were very significantly correlated with age in males, but not females (Fig. 2 and Table 2).

Fig.2

Hippocampal sphingolipids correlate with age at death in males but not females. Total levels of ceramide (A and E), SM (B and F), and ST (C and G), and S1P:sphingosine ratio (D and H), in hippocampal tissue from male (A-D; n = 48) and female (E-H, n = 32) subjects, plotted against age at death. Natural log-transformed values are shown for lipids. A line of best fit (linear regression) is shown. Correlations were determined by Pearson analysis, using log-transformed lipid levels. Coefficient of correlation and p value are shown. (I) Results of permutation testing on individual lipids identified as significantly associated with age are reported in Table 2.

Ceramide is comprised of many different forms, dictated primarily by the length of the N-acyl chain. These different forms of ceramide are then converted to SM, GalCer, ST, and gangliosides (Fig. 1). Given the association of total ceramide, SM, and ST with age in males, we tested the association of the major forms of ceramide, SM, GalCer, and ST, with age. To account for multiple comparisons (117 p values calculated in Table 2), the Benjamini-Hochberg false-discovery rate correction (Q = 5%) was applied to determine which p values could be considered statistically significant (Table 2). C16:0 ceramide, C16:0 SM, C16:1 SM, C22:1 ceramide, C22:1 SM, C22:1 ST, C22:0 ST, C24:1 ceramide, C24:1 ST, C24:0 ST, and C24:1-OH ST were all positively correlated with age in males (Table 2). There were no significant correlations between age and levels of these lipids in females; however, C16:0 ceramide (r = 0.33, p = 0.0028) and C22:1 SM (r = 0.32, P = 0.0043) were positively associated with age in the combined male + female cohort. Permutation testing to control for effects of random sampling further confirmed the significant correlations between age and sphingolipid levels reported in Table 2 (Fig. 2I).

In contrast to the positive correlations between major sphingolipids and age in males, S1P:sphingosine ratio was inversely correlated with age at death in hippocampus tissue from females (Fig. 2 and Table 2). This association remained significant after adjusting for PMI (r = –0.52; p = 0.002).

APOE genotype does not influence sphingolipid levels in cognitively normal elderly subjects

Having determined that sphingolipid levels are significantly affected by age at death, we analyzed sphingolipid levels as a function of APOE genotype, adjusting for age at death as a co-variate. The ɛ2/ɛ4 (n = 4) and ɛ2/ɛ2 (n = 2) genotype groups were not included due to insufficient numbers. In univariate analysis, using a threshold p value of 0.01, total levels of SM, ceramide, GalCer, ST, sphingosine, and S1P did not vary with APOE genotype (Table 3). Of the individual lipid species, only C16:1 SM levels varied significantly as a function of APOE genotype, with lowest levels in the ɛ2/ɛ3 group and highest levels in the ɛ3/ɛ4 group (Table 3). This association was not significant after adjusting for multiple comparisons. In general, associations were stronger after adjusting for age, and no lipid was significantly affected by APOE genotype in univariate ANOVA without adjusting for age (data not shown).

Table 3

Sphingolipid levels as a function of APOE genotype

	Mean ± SD (pmoles lipid/mg tissue)			ANOVA adjusted for age at death
	ɛ2/ɛ3	ɛ3/ɛ3	ɛ3/ɛ4	F value	p value
Total Ceramide	2042.3 ± 2544. 8	2025.4 ± 1433.4	1521.7 ± 923.5	1.14	0.33
C16:0 Cer	108.1 ± 203.7	48.99 ± 44.58	47.40 ± 30.16	0.22	0.81
C18:0 Cer	826.9 ± 902.3	626.68 ± 395.67	488.6 ± 316.6	0.99	0.38
C20:0 Cer	25.42 ± 12.50	33.08 ± 25.14	33.23 ± 32.37	0.34	0.71
C22:0 Cer	9.67 ± 6.32	14.73 ± 15.65	13.98 ± 17.69	0.78	0.47
C22:1 Cer	^#11.18 ± 15.94	14.54 ± 10.68	11.58 ± 9.45	3.19	0.047
C24:0 Cer	26.34 ± 37.68	39.57 ± 79.60	26.16 ± 31.60	1.40	0.25
C24:1 Cer	902.7 ± 1244.3	1079.8 ± 916.0	785.9 ± 657.0	1.50	0.23
Total GalCer	309.2 ± 190.4	398.7 ± 916.03	387.8 ± 271.8	0.53	0.59
C16:0 GalCer	0.52 ± 0.45	0.73 ± 0.60	0.84 ± 0.88	1.29	0.28
C18:0 GalCer	21.00 ± 23.75	24.38 ± 25.05	22.91 ± 27.16	0.43	0.65
C22:0 GalCer	2.05 ± 1.40	3.23 ± 3.14	3.05 ± 2.98	0.81	0.45
C22:1 GalCer	0.87 ± 0.77	1.19 ± 0.84	0.99 ± 0.59	1.22	0.30
C24:0 GalCer	16.22 ± 11.60	18.31 ± 22.34	19.90 ± 27.75	0.10	0.91
C24:1 GalCer	135.1 ± 93.4	172.7 ± 108.4	173.0 ± 105.1	0.64	0.53
C24:0 GalCer-OH	12.79 ± 10.45	22.41 ± 21.66	20.27 ± 25.34	0.38	0.69
C24:1 GalCer-OH	32.64 ± 17.61	40.97 ± 26.55	39.51 ± 24.44	0.18	0.84
Total SM	3215.9 ± 1651.8	3173.6 ± 1105.2	3235.4 ± 934.2	0.36	0.70
C16:0 SM	588.3 ± 415.4	563.1 ± 177.3	571.9 ± 196.5	0.68	0.51
C16:1 SM	^#,*35.37 ± 21.61	45.65 ± 19.32	55.15 ± 26.79	5.04	0.009
C18:0 SM	1005.8 ± 442.0	942.0 ± 366.6	926.7 ± 247.1	0.04	0.96
C18:1 SM	464.1 ± 233.4	499.5 ± 164.6	508.4 ± 152.8	0.79	0.46
C20:0 SM	150.4 ± 81.80	116.58 ± 81.04	133.9 ± 96.0	0.62	0.54
C22:0 SM	26.64 ± 10.97	24.71 ± 11.00	31.16 ± 21.57	1.07	0.35
C22:1 SM	25.7 ± 20.42	27.52 ± 13.08	28.96 ± 17.55	1.76	0.18
C24:0 SM	31.59 ± 25.71	34.29 ± 29.57	36.94 ± 44.56	0.14	0.87
C24:1 SM	611.3 ± 353.3	658.5 ± 347.2	669.5 ± 308.1	0.42	0.66
Total ST	1954.2 ± 2480.2	1787.8 ± 2012.8	1760.1 ± 2324.2	0.67	0.58
C16:0 ST	1.97 ± 1.12	2.44 ± 1.84	3.22 ± 2.81	0.53	0.59
C18:0 ST	24.07 ± 18.52	45.20 ± 44.19	45.10 ± 45.78	1.24	0.30
C22:0 ST	3.95 ± 5.68	3.96 ± 5.41	2.79 ± 4.15	0.62	0.54
C22:1 ST	3.19 ± 5.91	2.46 ± 2.78	2.38 ± 3.65	1.42	0.25
C24:0 ST	109.3 ± 144.5	111.4 ± 126.6	112.3 ± 146.9	0.49	0.62
C24:1 ST	991.4 ± 1212.1	904.2 ± 1076.3	889.1 ± 1314.7	0.09	0.92
C24:0 ST-OH	26.54 ± 26.00	37.40 ± 49.00	45.84 ± 81.94	0.49	0.62
C24:1 ST-OH	78.63 ± 95.81	66.62 ± 90.85	61.22 ± 87.05	0.00	1.00
Sphingosine	12.22 ± 7.55	12.69 ± 11.18	15.54 ± 17.34	0.75	0.48
S1P	0.50 ± 0.88	0.53 ± 0.46	1.14 ± 2.25	3.07	0.053
S1P/Sphingosine	0.038 ± 0.048	0.050 ± 0.036	0.061 ± 0.044	2.04	0.14

Table shows the mean and standard deviation (SD) for each lipid as a function of APOE genotype, and results of one-way ANOVA after adjusting for age as a co-variate. Lipids significant without adjusting for multiple comparisons (p < 0.01) are in Italics; # indicates significantly different means ɛ2/ɛ3 compared to ɛ3/ɛ3, at p < 0.05; ^*indicates significantly different means ɛ2/ɛ3 compared to ɛ3/ɛ3.

While our previous data suggested that APOE genotype influences the hippocampal S1P:sphingosine ratio in a mixed cohort of subjects with and without AD pathology [27], in this study the S1P:sphingosine ratio was not associated with APOE genotype in univariate ANOVA, nor after adjusting for age, PMI, or both.

As the samples were pathologically staged (Braak staging), linear regression models were used to test for any effect of Braak stage on sphingolipid levels (Supplementary Table 2). A general trend for increased GalCer levels from Braak stage 0 to stage I/II was observed, and considered in isolation, C24:0 GalCer was significantly affected as a function of Braak stage (p = 0.0017). However, this was not significant after adjusting for multiple comparisons (p = 0.067). Braak stage was not found to have any significant effect on other sphingolipids.

DISCUSSION

The most important overall risk factor for developing AD is aging, while the predominant genetic risk factor is inheritance of the APOE ɛ4 allele. The research community is currently unable to explain why AD is a disease of aging, nor how APOE ɛ4 modulates AD risk. We herein sought to determine if sphingolipid levels in the CA1 hippocampal region of cognitively normal, elderly human donors were significantly affected by the presence of an APOE ɛ4 allele. APOE genotype did not have a significant influence on hippocampal sphingolipid levels, including S1P/sphingosine ratio. However, we did observe an increase in total ceramide, SM and ST with increasing age above 65 years, in males. This overall increase was comprised of increases in multiple different forms of these lipids, most notably those bearing C16, C22, and C24 N-acyl chains. These lipids did not increase with age in females; instead we observed a pronounced decline in the S1P:sphingosine ratio.

To the best of our knowledge, this is the first report describing a significant effect of aging on sphingolipid levels in human hippocampus, and therefore awaits confirmation in an independent cohort and/or in different brain regions. We are aware of only one prior study reporting on changes to lipid composition of the human brain with aging, in which a notable decline in myelin lipid content was observed in females aged over 90 years [37]. This study pre-dated the use of mass spectrometry for lipid quantification and the authors were therefore unable to differentiate different variants of cerebroside (i.e., GalCer) and sulfatide. The overall increase in hippocampal sphingolipid content with age in males, reported here, is supported by a recent study in mice [26], with the exception that in our study, increased lipid levels were seen only in males, whereas Vozella et al. reported significantly increased sphingolipid levels in both male and female mice [26]. Interestingly, the authors showed most robust increases in C16:0 and C24:1 SM and HexCer, and C24:1 ceramide, in further agreement with our data. Increased ceramide levels, particularly C24:0, were also reported in the aging mouse cerebral cortex [25]. Given that ST is derived by sulfation of GalCer, it was somewhat surprising that the other major myelin sphingolipid, GalCer, did not show the same association with age in males. In addition to GalCer, glucosylceramide (GluCer) is an endogenous sphingolipid found in all human tissues. GalCer and GluCer are stereoisomers that cannot be distinguished with the LC-MS/MS conditions used herein, and are therefore commonly referred to as hexosylceramide (HexCer). However, HexCer in human brain tissue is >99% GalCer [38, 39], so it is highly unlikely that GluCer is a significant contributor to the total HexCer content measured.

Ceramide has traditionally been considered a pro-apoptotic, pro-inflammatory lipid, and is associated with neurodegenerative phenotypes [11 , 25]. The pro-apoptotic functions of ceramide are, however, somewhat of a generalization, as different forms of ceramide have important physiological functions in different organ systems. In the brain, C18 ceramide synthesis is essential for neuronal viability, as loss of the CerS1 enzyme produces a neurodegenerative phenotype [40, 41]. CerS2 synthesizes C24 ceramides that are incorporated into the essential myelin lipids GalCer and ST, and is therefore also essential for CNS health [14 , 43], as well as overall metabolic health [44, 45]. C16 ceramide, synthesized by CerS5 and/or CerS6 has been associated with mitochondrial dysfunction, oxidative stress [44, 46], and systemic insulin resistance [44, 47]. Diabetes is a known risk-factor for AD and vascular dementia [48], and de la Monte and co-workers have proposed a major role for ceramides and insulin resistance as related causal factors in AD pathogenesis [49]. The early stages of AD are accompanied by a marked reduction in cerebral glucose utilization, potentially caused by deficits of insulin receptors at synaptic membranes of the cerebral cortex and hippocampus [50], as well as reductions in insulin signaling that worsen with AD progression [51]. C16 ceramide is known to antagonize insulin signaling [12 , 47], so increasing levels of C16:0 ceramide and SM in the aging male brain could promote an insulin-resistant phenotype that sensitizes to AD pathogenesis.

We did not observe any positive associations between sphingolipid levels and age in the female hippocampus samples. In fact, a trend toward an inverse association between age and levels of the signaling lipid S1P was observed (r = –0.44, p = 0.013). S1P is a neuroprotective signaling molecule [52, 53] created following the phosphorylation of sphingosine by sphingosine kinases 1 and 2 (SphK1 and 2). Protein phosphorylation is generally expressed relative to total levels of the relevant protein, and when S1P levels are normalized to sphingosine, a stronger inverse association with age was observed in hippocampal tissue from females (r = –0.53; p = 0.002). This internal sample normalization approach reduces the inherent variability associated with absolute quantification of metabolites across different human subjects. In contrast to S1P, sphingosine is pro-apoptotic [54], so declining S1P/sphingosine with age may sensitize to neurodegeneration. S1P was shown to protect neurons against amyloid-β in vitro, and is lost from vulnerable brain regions during the course of AD pathogenesis [27]. S1P up-regulates the expression of neurotrophic genes including brain derived neurotrophic factor and cholinergic differentiation factor [55], and potentiates glutamate release from pre-synaptic termini, thereby supporting long-term potentiation in hippocampal neurons [56, 57]. Thus declining S1P/sphingosine balance with age may be a marker of declining synaptic activity. Alternatively, loss of sphingosine kinase - S1P signaling at synaptic termini could of itself be a significant driver of declining synaptic activity.

It is important to note that our study was set up to examine associations of lipid levels with APOE genotype in people over the age of 65, and was therefore not optimally designed to examine changes to lipid and gene expression levels with age. Despite this, the gender-specific associations between lipid levels and age in the hippocampus were very strong and remained significant after adjusting for multiple comparisons, as well as following permutation testing. This study did not investigate the underlying biochemical basis for the observed effect of age on lipid levels. The increased levels of specific SM and ST species with age in males (e.g., C16:0 SM and C24:1 ST) could be traced back to elevated levels of their ceramide precursors (Table 2). Ceramide:SM ratios were not significantly correlated with age (data not shown), reflecting the fact that both ceramide and SM levels increase with age in males, and indicating that sphingomyelinase activity is not significantly modulated as a function of age. C16 and very long chain (C22–C26) ceramides are synthesized by different CerS isoforms. Changes to activity or expression of these isoforms may explain the associated changes in sphingolipid levels. Alternatively, the general increase in sphingolipid levels in males could be suggestive of an increase in activity and/or expression of the serine palmitoyl transferase complex (SPT), the rate-limiting initial step in the biosynthesis of sphingolipids [9]. A third possibility is decreased degradation of sphingolipids with aging, through decreased expression of degradative enzymes such as S1P lyase (SGPL1 gene), or reduced lysosomal function. Loss of S1P during AD pathogenesis has been attributed to loss of SphK1 and 2, and gain of S1P lyase (SGPL1) expression [27, 58]. It is likely that declining S1P:sphingosine levels with age in the female hippocampus results from de-regulation of these enzymes, or the S1P phosphatases Sgpp1 and 2.

The presence of one APOE ɛ2 or ɛ4 allele, in comparison to the most common ɛ3/ɛ3 genotype, did not have a significant effect on sphingolipid levels in the present study. Unfortunately, we did not have any ɛ4/ɛ4 genotypes, and only two ɛ2/ɛ2 genotypes, in the current cohort. Thus, we were unable to test the influence of individual ApoE isoforms, particularly ɛ4, in the absence of the ɛ3 allele. Homozygous carriers of the ɛ4 allele are highly likely to develop dementia and are generally underrepresented in the cognitively normal elderly population, as used in this study [3 , 6]. We were also unable to analyze the effect of APOE genotype separately in male and female brains, as there are insufficient numbers of ɛ2/ɛ3 and ɛ3/ɛ4 genotypes in each gender group. In the present study, we found no significant association between S1P:sphingosine ratio and APOE genotype, suggesting that the possible influence of APOE gentoype on S1P:sphingosine balance observed in our previous study [27] is primarily attributed to the influence of AD pathology. As demonstrated in the present study, hippocampal sphingolipid levels are not significantly affected by Braak stage I or II pathology, in which NFT pathology is restricted to the entorhinal and perirhinal cortical regions [28]. Our previous work included subjects with intermediate and advanced Braak stages (III-VI), in which the hippocampus is affected by NFT pathology. This concurs with other reports demonstrating that APOE genotype does not affect sphingolipid levels in cortical tissue [59] or plasma [60] of cognitively normal subjects, but does have an effect in the presence of AD pathology. Bandaru et al. proposed that pre-existing AD pathology is required to unmask the effect of APOE genotype on sphingolipid levels [59]. The present study did not include subjects with AD pathology, and we are therefore unable to verify this possibility. Similarly, transgenic expression of human ApoE4 versus ApoE3 in ApoE knockout mice has been shown to have little overall influence on brain lipid levels [61], although Han et al. reported that brain ST levels are higher in ApoE knockout mice compared to WT mice, and lower in the ApoE4 transgenics compared to the ApoE3 transgenics [62].

In summary, we show here that the presence of an ɛ2 or ɛ4 allele of the APOE gene did not affect hippocampal sphingolipid levels in cognitively normal adults aged 65 and older, however sphingolipid balance is significantly altered as a function of age. This is the first report describing such effects of age on sphingolipid levels in the human brain. Increased ceramides with aging in males and decreased S1P with aging in females are both pro-neurodegenerative influences that likely sensitize the aging brain to AD and other dementias.

Footnotes

ACKNOWLEDGMENTS

This work was supported by NHMRC-ARC dementia research fellowship APP1110400 (T.A.C.), Australian Postgraduate Awards to N.K and C.T., an Alzheimer’s Australia postgraduate top-up scholarship (N.K.), and NHMRC project grant APP1100626 (A.S.D.). Tissue samples were sourced from the Queensland Brain Bank, School of Chemistry and Molecular Biosciences, University of Queensland; and the NSW BTRC at the University of Sydney, which is supported by the Schizophrenia Research Institute and National Institute of Alcohol Abuse and Alcoholism (NIH (NIAAA) R28AA012725). We gratefully acknowledge subsidized access to the Bioanalytical Mass Spectrometry Facility at UNSW.

Authors’ disclosures available online ().

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

References

Thies

, Bleiler

, Association

Association

(2013) 2013 Alzheimer’s disease facts and figures. Alzheimers Dement 9, 208–245.

Prince

, Bryce

, Albanese

, Wimo

, Ribeiro

, Ferri

(2013) The global prevalence of dementia: A systematic review and metaanalysis. Alzheimers Dement 9, 62.

Liu

, Kanekiyo

, Xu

, Bu

(2013) Apolipoprotein E and Alzheimer disease: Risk, mechanisms and therapy. Nat Rev Neurol 9, 106–118.

Holtzman

, Morris

, Goate

(2011) Alzheimer’s disease: The challenge of the second century. Sci Transl Med 3, 77sr71.

Farrer

, Cupples

, Haines

, Hyman

, Kukull

, Mayeux

, Myers

, Pericak-Vance

, Risch

, van

Duijn CM

(1997) Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. JAMA 278, 1349–1356.

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.

Corder

, Saunders

, Risch

, Strittmatter

, Schmechel

, Gaskell Jr PC , Rimmler

, Locke

, Conneally

, Schmader

, et

al.

(1994) Protective effect of apolipoprotein E type 2 allele for late onset Alzheimer disease. Nat Genet 7, 180–184.

Vance

, Hayashi

(2010) Formation and function of apolipoprotein E-containing lipoproteins in the nervous system. Biochim Biophys Acta 1801, 806–818.

Merrill Jr AH (2011) Sphingolipid and glycosphingolipid metabolic pathways in the era of sphingolipidomics. Chem Rev 111, 6387–6422.

10.

Yuyama

, Mitsutake

, Igarashi

(2014) Pathological roles of ceramide and its metabolites in metabolic syndrome and Alzheimer’s disease. Biochim Biophys Acta 1841, 793–798.

11.

Wang

, Dinkins

, He

, Zhu

, Poirier

, Campbell

, Mayer-Proschel

, Bieberich

(2012) Astrocytes secrete exosomes enriched with proapoptotic ceramide and prostate apoptosis response 4 (PAR-4): Potential mechanism of apoptosis induction in Alzheimer disease (AD). J Biol Chem 287, 21384–21395.

12.

Chaurasia

, Summers

(2015) Ceramides - lipotoxic inducers of metabolic disorders. Trends Endocrinol Metab 26, 538–550.

13.

O’Brien

, Sampson

(1965)J Lipid Res 6, 537–544 Lipid composition of the normal human bra, Gray matter, white matter, and myelin. .

14.

Coetzee

, Fujita

, Dupree

, Shi

, Blight

, Suzuki

, Popko

(1996) Myelination in the absence of galactocerebroside and sulfatide: Normal structure with abnormal function and regional instability. Cell 86, 209–219.

15.

Bosio

, Binczek

, Stoffel

(1996) Functional breakdown of the lipid bilayer of the myelin membrane in central and peripheral nervous system by disrupted galactocerebroside synthesis. Proc Natl Acad Sci U S A 93, 13280–13285.

16.

Han

, Holtzman

, McKeel

, Kelley

, Morris

(2002) Substantial sulfatide deficiency and ceramide elevation in very early Alzheimer’s disease: Potential role in disease pathogenesis. J Neurochem 82, 809–818.

17.

Couttas

, Kain

, Suchowerska

, Quek

, Turner

, Fath

, Garner

, Don

(2016) Loss of ceramide synthase 2 activity, necessary for myelin biosynthesis, precedes tau pathology in the cortical pathogenesis of Alzheimer’s disease. Neurobiol Aging 43, 89–100.

18.

, Huang

, Li

, Gong

C-X

, Schuchman

(2010) Deregulation of sphingolipid metabolism in Alzheimer’s disease. Neurobiol Aging 31, 398–408.

19.

Kosicek

, Zetterberg

, Andreasen

, Peter-Katalinic

, Hecimovic

(2012) Elevated cerebrospinal fluid sphingomyelin levels in prodromal Alzheimer’s disease. Neurosci Lett 516, 302–305.

20.

Satoi

, Tomimoto

, Ohtani

, Kitano

, Kondo

, Watanabe

, Oka

, Akiguchi

, Furuya

, Hirabayashi

, Okazaki

(2005) Astroglial expression of ceramide in Alzheimer’s disease brains: A role during neuronal apoptosis. Neuroscience 130, 657–666.

21.

Filippov

, Song

, Zhang

, Vinters

, Tung

, Kirsch

, Yang

, Duerksen-Hughes

(2012) Increased ceramide in brains with Alzheimer’s and other neurodegenerative diseases. J Alzheimers Dis 29, 537–547.

22.

Chan

, Oliveira

, Cortes

, Honig

, Duff

, Small

, Wenk

, Shui

, Di

Paolo G

(2012) Comparative lipidomic analysis of mouse and human brain with Alzheimer disease. J Biol Chem 287, 2678–2688.

23.

Mielke

, Bandaru

VVR

, Haughey

, Rabins

, Lyketsos

, Carlson

(2010) Serum sphingomyelins and ceramides are early predictors of memory impairment. Neurobiol Aging 31, 17–24.

24.

Mielke

, Bandaru

, Haughey

, Xia

, Fried

, Yasar

, Albert

, Varma

, Harris

, Schneider

, Rabins

, Bandeen-Roche

, Lyketsos

, Carlson

(2012) Serum ceramides increase the risk of Alzheimer disease: The Women’s Health and Aging Study II. Neurology 79, 633–641.

25.

Cutler

, Kelly

, Storie

, Pedersen

, Tammara

, Hatanpaa

, Troncoso

, Mattson

(2004) Involvement of oxidative stress-induced abnormalities in ceramide and cholesterol metabolism in brain aging and Alzheimer’s disease. Proc Natl Acad Sci U S A 101, 2070–2075.

26.

Vozella

, Basit

, Misto

, Piomelli

(2017) Age-dependent changes in nervonic acid-containing sphingolipids in mouse hippocampus. Biochim Biophys Acta 1862, 1502–1511.

27.

Couttas

, Kain

, Daniels

, Lim

, Shepherd

, Kril

, Pickford

, Li

, Garner

, Don

(2014) Loss of the neuroprotective factor Sphingosine 1-phosphate early in Alzheimer’s disease pathogenesis. Acta Neuropathol Commun 2, 9.

28.

Braak

, Braak

(1995) Staging of Alzheimer’s disease-related neurofibrillary changes. Neurobiol Aging 16, 271–278.

29.

Price

, Ko

, Wade

, Tsou

, McKeel

, Morris

(2001) Neuron number in the entorhinal cortex and ca1 in preclinical alzheimer disease. Arch Neurol 58, 1395–1402.

30.

West

, Coleman

, Flood

, Troncoso

(1994) Differences inthe pattern of hippocampal neuronal loss in normal ageing andAlzheimer’s disease. Lancet 344, 769–772.

31.

Villeda

, Plambeck

, Middeldorp

, Castellano

, Mosher

, Luo

, Smith

, Bieri

, Lin

, Berdnik

, Wabl

, Udeochu

, Wheatley

, Zou

, Simmons

, Xie

, Longo

, Wyss-Coray

(2014) Young blood reverses age-related impairments in cognitive function and synaptic plasticity in mice. Nat Med 20, 659–663.

32.

Hyman

, Phelps

, Beach

, Bigio

, Cairns

, Carrillo

, Dickson

, Duyckaerts

, Frosch

, Masliah

, Mirra

, Nelson

, Schneider

, Thal

, Thies

, Trojanowski

, Vinters

, Montine

(2012) National Institute on Aging-Alzheimer’s Association guidelines for the neuropathologic assessment of Alzheimer’s disease. Alzheimers Dement 8, 1–13.

33.

Oldmeadow

, Holliday

, McEvoy

, Scott

, Kwok

, Mather

, Sachdev

, Schofield

, Attia

(2014) Concordance between directand imputed APOE genotypes using 1000 Genomes data. JAlzheimers Dis 42, 391–393.

34.

Hejazi

, Wong

, Cheng

, Proschogo

, Ebrahimi

, Garner

, Don

(2011) Mass and relative elution time profiling: Two-dimensional analysis of sphingolipids in Alzheimer’s disease brains. Biochem J 438, 165–175.

35.

Wong

, Abuhusain

, McDonald

, Don

(2012) MMSAT: Automated quantification of metabolites in selected reaction monitoring experiments. Anal Chem 84, 470–474.

36.

Mielke

, Vemuri

, Rocca

(2014) Clinical epidemiology of Alzheimer’s disease: Assessing sex and gender differences. Clin Epidemiol 6, 37–48.

37.

Svennerholm

, Bostrom

, Helander

, Jungbjer

(1991) Membrane lipids in the aging human brain. J Neurochem 56, 2051–2059.

38.

Vanier

, Svennerholm

(1975) Chemical pathology of Krabbe’s disease. III. Ceramide-hexosides and gangliosides of brain. Acta Paediatr Scand 64, 641–648.

39.

Svennerholm

, Vanier

, Mansson

(1980) Krabbe disease: Agalactosylsphingosine (psychosine) lipidosis.. J Lipid Res 21, 53–64.

40.

Zhao

, Spassieva

, Jucius

, Shultz

, Shick

, Macklin

, Hannun

, Obeid

, Ackerman

(2011) A deficiency of ceramide biosynthesis causes cerebellar purkinje cell neurodegeneration and lipofuscin accumulation. PLoS Genet 7, e1002063.

41.

Ginkel

, Hartmann

, vom

Dorp K

, Zlomuzica

, Farwanah

, Eckhardt

, Sandhoff

, Degen

, Rabionet

, Dere

, Dormann

, Sandhoff

, Willecke

(2012) Ablation of neuronal ceramide synthase 1 in mice decreases ganglioside levels and expression of myelin-associated glycoprotein in oligodendrocytes. J Biol Chem 287, 41888–41902.

42.

Imgrund

, Hartmann

, Farwanah

, Eckhardt

, Sandhoff

, Degen

, Gieselmann

, Sandhoff

, Willecke

(2009) Adult ceramide synthase 2 (CERS2)-deficient mice exhibit myelin sheath defects, cerebellar degeneration, and hepatocarcinomas. J Biol Chem 284, 33549–33560.

43.

Ben-David

, Pewzner-Jung

, Brenner

, Laviad

, Kogot-Levin

, Weissberg

, Biton

, Pienik

, Wang

, Kelly

, Alroy

, Raas-Rothschild

, Friedman

, Brugger

, Merrill Jr AH , Futerman

(2011) Encephalopathy caused by ablation of very long acyl chain ceramide synthesis may be largely due to reduced galactosylceramide levels. J Biol Chem 286, 30022–30033.

44.

Raichur

, Wang

, Chan

, Li

, Ching

, Chaurasia

, Dogra

, Öhman

, Takeda

, Sugii

, Pewzner-Jung

, Futerman

, Summers

(2014) CerS2 haploinsufficiency inhibitsβ-oxidation and confers susceptibility to diet-inducedsteatohepatitis and insulin resistance.. Cell Metab 20, 687–695.

45.

Park

, Park

, Kuperman

, Boura-Halfon

, Pewzner-Jung

, Futerman

(2013) Ablation of very long acyl chain sphingolipids causes hepatic insulin resistance in mice due to altered detergent-resistant membranes. Hepatology 57, 525–532.

46.

Zigdon

, Kogot-Levin

, Park

, Goldschmidt

, Kelly

, Merrill Jr AH , Scherz

, Pewzner-Jung

, Saada

, Futerman

(2013) Ablation of ceramide synthase 2 causes chronic oxidative stress due to disruption of the mitochondrial respiratory chain. J Biol Chem 288, 4947–4956.

47.

Turpin

Sarah M

, Nicholls

Hayley T

, Willmes

Diana M

, Mourier

, Brodesser

, Wunderlich

Claudia M

, Mauer

, Xu

, Hammerschmidt

, Brönneke

Hella S

, Trifunovic

, LoSasso

, Wunderlich

, Kornfeld

J-W

, Blüher

, Krönke

, Brüning

Jens C

(2014) Obesity-induced CerS6-dependent C16:0 ceramide production promotes weight gain and glucose intolerance. Cell Metab 20, 678–686.

48.

, Qiu

, Gatz

, Pedersen

, Johansson

, Fratiglioni

(2009) Mid- and late-life diabetes in relation to the risk of dementia: A population-based twin study. Diabetes 58, 71–77.

49.

la Monte SM

(2012) Triangulated mal-signaling in Alzheimer’s disease: Roles of neurotoxic ceramides, ER stress, and insulin resistance reviewed. (Suppl 2), S. J Alzheimers Dis 30, 231–249.

50.

Rivera

, Goldin

, Fulmer

, Tavares

, Wands

, de

la Monte SM

(2005) Insulin and insulin-like growth factor expression and function deteriorate with progression of Alzheimer’s disease: Link to brain reductions in acetylcholine. J Alzheimers Dis 8, 247–268.

51.

Steen

, Terry

, Rivera

, Cannon

, Neely

, Tavares

, Xu

, Wands

, de

la Monte SM

(2005) Impaired insulin and insulin-like growth factor expression and signaling mechanisms in Alzheimer’s disease–is this type 3 diabetes? . J Alzheimers Dis 7, 63–80.

52.

Pfeilschifter

, Czech-Zechmeister

, Sujak

, Mirceska

, Koch

, Rami

, Steinmetz

, Foerch

, Huwiler

, Pfeilschifter

(2011) Activation of sphingosine kinase 2 is an endogenous protective mechanism in cerebral ischemia. Biochem Biophys Res Commun 413, 212–217.

53.

Malaplate-Armand

, Florent-Bechard

, Youssef

, Koziel

, Sponne

, Kriem

, Leininger-Muller

, Olivier

, Oster

, Pillot

(2006) Soluble oligomers of amyloid-beta peptide induce neuronal apoptosis by activating a cPLA2-dependent sphingomyelinase-ceramide pathway. Neurobiol Dis 23, 178–189.

54.

Woodcock

(2006) Sphingosine and ceramide signalling in apoptosis. IUBMB Life 58, 462–466.

55.

Hoffmann

, Hofereiter

, Rubsamen

, Melms

, Schwarz

, Faber

, Weber

, Putz

, Loleit

, Weber

, Hohlfeld

, Meinl

, Krumbholz

(2015) Fingolimod induces neuroprotective factors in human astrocytes. J Neuroinflammation 12, 184.

56.

Kajimoto

, Okada

, Yu

, Goparaju

, Jahangeer

, Nakamura

(2007) Involvement of sphingosine-1-phosphate in glutamate secretion in hippocampal neurons. Mol Cell Biol 27, 3429–3440.

57.

Kanno

, Nishizaki

, Proia

, Kajimoto

, Jahangeer

, Okada

, Nakamura

(2010) Regulation of synaptic strength by sphingosine 1-phosphate in the hippocampus. Neuroscience 171, 973–980.

58.

Ceccom

, Loukh

, Lauwers-Cances

, Touriol

, Nicaise

, Gentil

, Uro-Coste

, Pitson

, Maurage

, Duyckaerts

, Cuvillier

, Delisle

(2014) Reduced sphingosine kinase-1 and enhanced sphingosine 1-phosphate lyase expression demonstrate deregulated sphingosine 1-phosphate signaling in Alzheimer’s disease. Acta Neuropathol Commun 2, 12.

59.

Bandaru

VVR

, Troncoso

, Wheeler

, Pletnikova

, Wang

, Conant

, Haughey

(2009) ApoE4 disrupts sterol and sphingolipid metabolism in Alzheimer’s but not normal brain. Neurobiol Aging 30, 591–599.

60.

Han

, Rozen

, Boyle

, Hellegers

, Cheng

, Burke

, Welsh-Bohmer

, Doraiswamy

, Kaddurah-Daouk

(2011) Metabolomics in early Alzheimer’s disease: Identification of altered plasma sphingolipidome using shotgun lipidomics. PLoS One 6, e21643.

61.

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.

62.

Han

, Cheng

, Fryer

, Fagan

, Holtzman

(2003) Novel role for apolipoprotein E in the central nervous system. Modulation of sulfatide content. J Biol Chem 278, 8043–8051.