Cholesterol metabolism is associated with soluble amyloid precursor protein production in Alzheimer's disease

Authors


Address correspondence and reprint requests to Julius Popp, MD,Department of Psychiatry, University Hospital of Lausanne, Route du Mont, 1008 Prilly-Lausanne, Switzerland.

E-mail: julius.popp@chuv.ch

Abstract

Disturbances of the cholesterol metabolism are associated with Alzheimer's disease (AD) risk and related cerebral pathology. Experimental studies found changing levels of cholesterol and its metabolites 24S-hydroxycholesterol (24S-OHC) and 27-hydroxycholesterol (27-OHC) to contribute to amyloidogenesis by increasing the production of soluble amyloid precursor protein (sAPP). The aim of this study was to evaluate the relationship between the CSF and circulating cholesterol 24S-OHC and 27-OHC, and the sAPP production as measured by CSF concentrations of sAPP forms in humans. The plasma and the CSF concentrations of cholesterol, 24S-OHC and 27-OHC, and the CSF concentrations of sAPPα, sAPPβ, and Aß1-42 were assessed in subjects with AD and controls with normal cognition. In multivariate regression tests including age, gender, albumin ratio, and apolipoprotein E (APOE)ε4 status CSF cholesterol, 24S-OHC, and 27-OHC independently predicted the concentrations of sAPPα and sAPPβ. The associations remained significant when analyses were separately performed in the AD group. Furthermore, plasma 27-OHC concentrations were associated with the CSF sAPP levels. The results suggest that high CSF concentrations of cholesterol, 24S-OHC, and 27-OHC are associated with increased production of both sAPP forms in AD.

Abbreviations used
ABCA1

ATP-binding cassette transporter A1 AD, Alzheimer's disease

APOE

apolipoprotein E

APP

amyloid precursor protein

amyloid beta

BACE1

β-site APP cleavage enzyme 1

OHC

hydroxycholesterol

VIF

variance inflation factor

The extracellular accumulation and deposition of amyloid beta (Aβ) in form of plaques is one of the core pathological hallmarks of Alzheimer's disease (AD). According to the amyloid hypothesis, accumulation of Aβ peptides triggers a cascade of pathophysiological changes leading to neurodegeneration (Hardy and Selkoe 2002). Aβ is a 38–43-amino acid peptide and a cleavage product of a much larger transmembrane peptide, the amyloid precursor protein (APP). Two major proteolytic processing pathways of APP have been identified, an amyloidogenic and a non-amyloidogenic one. The amyloidogenic processing of APP by β-site APP cleavage enzyme 1 (BACE1 or β-secretase) leads to the liberation of the soluble APPβ (sAPPβ) into the extracellular space. The cell-associated fragments are further cleaved by γ-secretase, generating a spectrum of Aβ peptides of varying length – including the most common Aβ40 and the less abundant, however, more fibrillogenic Aβ42. On the alternative non-amyloidogenic pathway, APP can be cleaved within the Aβ sequence by α-secretases resulting in the extracellular liberation of sAPPα.

Research of the last two decades revealed that cholesterol metabolism is involved in the AD pathogenesis, and in particular in amyloidogenesis. A strong genetic risk factor for late-onset AD is the presence of the ε4 allele of the apolipoprotein E (APOE) gene, which encodes a protein with crucial roles in cholesterol metabolism (Bu 2009). Carrying the APOEε4 allele may early, at pre-clinical disease stages contribute to pathological amyloid accumulation and deposition (Kok et al. 2009), and accelerate age-related changes of CSF Aß1-42 concentrations in middle-aged and older adults with normal cognition (Popp et al. 2010). A pathway analysis of the newly identified genes associated with AD also implicates cholesterol metabolism in the etiology of AD (Jones et al. 2010). Further evidence for the link between alterations of cholesterol metabolism and amyloid pathology comes from autopsy studies showing that middle-aged and elder subjects with hypercholesterolemia or dyslipidemia have significantly higher risk of cerebral amyloid plaque pathology (Launer et al. 2001; Pappolla et al. 2003; Matsuzaki et al. 2011).

Findings from cell culture and animal studies suggest that changing levels of cholesterol and its elimination products, the oxysterols 27-hydroxycholesterol (27-OHC) and the brain-specific 24S-hydroxycholesterol (24S-OHC), may contribute to amyloidogenesis by increasing sAPP production. Lowering cholesterol levels has been shown to reduce BACE1 activity and Aβ production (Simons et al. 1998; Fassbender et al. 2001; Ehehalt et al. 2003; Xiong et al. 2008). Increasing cholesterol levels in cultured cells and in transgenic mouse models of AD resulted in decreased sAPPα production (Bodovitz and Klein 1996; Refolo et al. 2000; Shie et al. 2002; Canepa et al. 2011), whereas lowering cellular cholesterol content increased the α-secretase activity (Kojro et al. 2001). Other experimental studies revealed an influence of the Aβ levels on cholesterol de-novo synthesis and concentrations (Grimm et al. 2005; Wirths et al. 2006; Canepa et al. 2011).

In addition to cholesterol, both 27-OHC and 24S-OHC may have effects on the APP and amyloid production. Experimental studies provided at least partially conflicting results, however. Both 27-OHC and 24S-OHC have been reported to inhibit APP secretion and amyloid production, with 24S-OHC having much more marked effects than 27-OHC (Brown et al. 2004). Others reported total APP and sAPPα to increase after treatment with 24S-OHC, but not with 27-OHC (Famer et al. 2007). Further studies have found 27-OHC to increase BACE1, APP, and Aβ levels (Sharma et al. 2008; Marwarha et al. 2010; Prasanthi et al. 2009, 2011; Dasari et al. 2010) as well as 24S-OH to induce the expression of the cholesterol efflux molecule ATP-binding cassette transporter A1 (ABCA1) with higher levels of ABCA1 to be associated with increased Aβ production (Fukumoto et al. 2002).

Although in vitro and animal studies provide substantial evidence that alterations of cholesterol metabolism may affect APP and Aβ production, and contribute to amyloid pathology, it is not known to date whether changes in cholesterol, 24S-OHC, and 27-OHC levels are related to amyloidogenesis in humans. The aim of this study was to evaluate the relationship between the cerebral and extracerebral cholesterol and its metabolites as reflected by their CSF and plasma concentrations, and the CSF concentrations of sAPP forms and of Aß1-42 in patients with AD and subjects with normal cognition. Preliminary findings of this study have been presented at the Alzheimer's Association International Conference 2011 and published as an abstract (Popp et al. 2011).

Subjects and methods

Ninety-seven participants with the diagnosis of probable AD (n = 53) or with normal cognition (n = 43) were included into this study. The participants with AD were patients investigated for their cognitive impairment at the Memory Clinic, at the Department of Psychiatry, University of Bonn. They met clinical diagnostic criteria for probable AD according to the National Institute of Neurological and Communicative Disorders and Stroke and Related Disorders Association (NINCDS-ADRDA) (McKhann et al. 1984), and DSM-IV criteria for dementia of the Alzheimer type. The diagnosis of AD was based on comprehensive neuropsychological and clinical evaluation and was made by a consensus conference of psychiatrists and neuropsychologists prior to the CSF analysis. The presence of relevant vascular cerebral lesions was excluded for all study participants by computed tomography or magnetic resonance tomography.

The participants in the control group were subjects with normal cognition referred to the Department of Neurology, University of Bonn for the evaluation or exclusion of neurological disorders. They had disorders not affecting the CNS, like tension headache (n = 8), peripheral neuropathy (n = 4), peripheral cranial nerve palsy (n = 4), and myopathy (n = 1), or affections potentially associated with pathological conditions of the CNS: spinal stenosis (n = 22), idiopathic intracranial hypertension (n = 1), seizures (n = 1), spinal cord infarction (n = 2). All participants underwent detailed clinical evaluation that consisted of medical history, physical and neurological examination. The cognitive function of all participants in this control group was assessed with the Mini-Mental State Examination (MMSE) (Folstein et al. 1975). The included subjects had MMSE scores ≥ 27 and no signs or symptoms suggesting cognitive decline. Structural imaging of the brain was performed by magnetic resonance tomography, or, in some cases, by computed tomography. Subjects with symptomatic cardiac disease, renal or hepatic dysfunction, insulin-dependent diabetes mellitus, untreated thyroidal dysfunction, inflammatory disease, and alcohol abuse or receiving treatment with known cholesterol-lowering medication were excluded from this study. In addition, subjects with CSF samples indicating a highly permeable blood–CSF barrier or inflammatory signs defined as a CSF:plasma albumin quotient (QAlb) > 9.0 and more than 5 leukocytes/mm3, or with more than 500 erythrocytes/uL were excluded. Patient characteristics are given in Table 1.

Table 1. Subject characteristics and biochemical data by clinical diagnosis
Subject characteristics Controls (= 43) AD patients (= 53) 
  1. a

    Pearson's χ2 test;

  2. b

    t-test;

  3. c

    Adjusted for age and gender.

  4. SD, Standard-Deviation; MMSE, Mini-Mental State Examination; CH, Cholesterol; 27-OHC, 27-hydroxy-cholesterol; 24-OHC, 24S-hydroxycholesterol; Qalb, albumin ratio.

Gender (f/m): n 22/21 33/20 χ2=1.196; = 0.305a
APOEε4 (non-carrier/carrier): n 33/10 22/31 χ2=12.045; = 0.001a
Age (years): mean (SD) 67.33 (9.04) 71.23 (8.29) = 0.016; = 0.032b
MMSE: mean (SD) 28.71 (1.54) 20.37 (5.56) = 36.469; < 0.001b
Biochemical data anovac
CH (mg/dL): mean (SD) 225.33 (39.67) 235.76 (46.61) = 0.388; = 0.535
27-OHC (ng/mL): mean (SD) 199.58 (40.71) 219.63 (79.43) = 2.263; = 0.137
24S-OHC (ng/mL): mean (SD) 62.28 (16.07) 78.82 (30.52) = 6.082; = 0.016
24S-OHC/27-OHC: mean (SD)0.319 (0.086)0.371 (0.111)= 2.716; = 0.103
CHcsf (mg/dL): mean (SD) 0.58 (0.19) 0.51 (0.19) = 4.443; = 0.038
27-OHCcsf (ng/mL): mean (SD) 2.00 (0.89) 1.80 (0.52) = 1.129; = 0.291
24S-OHCcsf (ng/mL): mean (SD) 2.68 (1.07) 2.99 (1.39) = 0.653; = 0.421
24S-OHCcsf/27-OHCcsf: mean (SD)1.422 (0.432)1.812 (0.436)= 10.868; = 0.002
sAPPα (ng/mL): mean (SD) 111.82 (40.17) 125.93 (45.59) = 1.933; = 0.168
sAPPβ (ng/mL): mean (SD) 142.79 (52.04)159.52 (57.25)= 1.072; = 0.303
Aβ1-42 (pg/mL): mean (SD) 797.27 (216.44) 428.33 (191.71) = 75.648; < 0.001
Qalb: mean (SD) 6.55 (2.96) 5.90 (2.68) = 0.092; = 0.763

Venous and lumbar punctures were performed after overnight fasting, at the Departments of Psychiatry or Neurology, University of Bonn. For lumbar puncture, a standardized technique with a 22G ‘atraumatical’ spinal needle was applied. The CSF samples were kept on ice for a maximum of 1 h until being centrifuged for 10 min at 2000 g at 4°C. Samples were aliquoted to 0.25 mL and were stored in polypropylene tubes at −80°C until assay procedures. Quantitative assessment of serum and CSF albumin was performed with nephelometry as a part of the routine CSF/serum analysis. CSF Aβ1-42 concentrations were measured by ELISA using commercially available assays (Innogenetics, Gent, Belgium). Leukocyte genomic DNA was isolated from 10 mL EDTA blood with the Qiagen blood isolation kit (Qiagen, Hilden, Germany) and the APOE genotype was determined as described before (Hixson and Vernier 1990). Soluble APPα and APPβ CSF concentrations were measured with a multiplexing assay of Meso Scale Discovery (Gaithersburg, MD, USA) as described earlier (Lewczuk et al. 2010, 2012). Plasma and CSF sterols and oxysterols were extracted after alkaline hydrolysis using cylohexane and separated as their trimethylsilylethers on cross-linked methyl silicone DB-XLB 122-1232 fused silica capillary columns (J&W, Folsom, CA, USA) (30 m × 0.25 mm i.e. × 0.25 μm film thickness) as described previously (Kolsch et al. 2010). Briefly, plasma concentrations of cholesterol were measured using gas chromatography – flame ionization detection on an HP 6890 series II plus GC (Agilent Technologies, Böblingen, Germany) using 5α-cholestane as an internal standard. CSF cholesterol was measured by gas chromatography-mass spectrometry with electron impact ionization (70 eV) using selected ion-monitoring (GC-MS-SIM) at m/z 458 and epicoprostanol (m/z 370) as internal standard. Plasma and CSF concentrations of 24S- and 27-hydroxycholesterol (m/z 413 and m/z 456, respectively) were measured by GC-MS-SIM using deuterium-labeled 24R,S and 27-hydroxycholesterol as internal standards (isotope dilution methodology) (m/z 416 and m/z 461, respectively). All GC-MS-SIM measurements were performed on an HP GC-MSD system (HP5890 series II GC combined with a 5971 mass-selective detector; Agilent Technologies). An aliquot of 1μL was injected in a splitlessmode at 280°C by an automated sampler and injector (HP 7683). Helium was used as carrier gas with an inlet pressure of 9.9 psi, resulting in a total gas-flow of 1.1 mL/min. The temperature program was as follows: 150°C for 1 min, followed by 20°C/min up to 260°C, and 10°C/min up to 280°C (for 15 min). Identity of all sterols was proven by comparison with the full-scan mass spectra of authentic compounds (range, m/z 50–500). Additional qualifier (characteristic fragment) ions were used for structural identification. The intra- and interday coefficients of variation for all sterols was below 3%. Accuracy of the method was established by recovery experiments, day-to-day variation (below 3%), limit of detection, and limit of quantification below the present concentrations for each sterol. All measurements were performed in duplicate.

All participants of the study gave informed written consent. The study was approved by the Human Ethics Committee of the Faculty of Medicine of the University Clinics of Bonn, Germany.

Statistical analysis

Univariate analysis of variance (anova) including age and gender as further factors, followed by post hoc Bonferroni's multiple comparison test was used to compare the biochemical values between AD patients and controls. Correlations between measured values were analyzed with Pearson's correlation coefficient. Multivariate regression models with the levels of the sAPP forms or of Aβ1-42 as dependent variables, and including age, gender, albumin ratio, APOEε4 status, and diagnosis were used to investigate associations of plasma and CSF cholesterol, 24S-OHC, and 27-OHC with the amyloid-related molecules and considering further potentially relevant factors. In addition, analyses were separately performed in the groups of controls with normal cognition and of patients with the AD to further explore group-specific aspects. As the 24S-OHC/ 27-OHC ratio has been proposed to be of particular importance for the generation of amyloid in the brain (Famer et al. 2007; Prasanthi et al. 2009), we further performed separate analyses of its association with the CSF concentrations of the sAPP forms and Aβ1-42. As it has been hypothesized that hypercholesterolemia-induced Aβ accumulation in the brain may be mediated by 27-OHC entering the CNS from the circulation (Bjorkhem 2006; Sharma et al. 2008), we used multivariate regression models to explore association between plasma and CSF 27-OHC concentrations. To rule out multicollinearity between the included parameters model tolerance and the variance inflation factor (VIF) were assessed using multiple linear regression with multicollinearity testing. Statistical analyses were performed using the statistical analysis software package pasw Statistics 18.0 for Windows (SPSS Inc., Chicago, IL, USA).

Results

Subjects' characteristics and biochemical data by clinical diagnosis group are given in Table 1. Significant correlations were observed between the 24S-OHC CSF concentrations and the sAPPα (r = 0.398, p < 0.001) and sAPPβ (r = 0.326, p = 0.001) concentrations in the pooled sample. Furthermore, 24S-OHC CSF concentrations in both the control and the AD group correlated with sAPPα (r = 0.370, p = 0.015, and r = 0.495; p = 0.003, respectively) and sAPPβ (r = 0.351, p = 0.021, and r = 0.292; p = 0.032) CSF levels (Fig. 1). There were no correlations between the CSF cholesterol and 27-OHC concentrations or the plasma cholesterol, 24S-OHC, and 27-OHC concentrations, and the measured sAPP forms. Furthermore, neither the plasma nor the CSF 24S-OHC/27-OHC ratios were correlated with the concentrations of the sAPP forms (data not shown). Plasma cholesterol was correlated with plasma 27-OHC (r = 0.588; p < 0.001). In multivariate regression analyses including age, gender, Qalb, APOEε4 status (carrier of at least one ε4 allele vs. non-carrier), and diagnosis, the CSF cholesterol, 24S-OHC, and 27-OHC concentrations independently predicted the concentrations of sAPPα (cholesterol: F = 7.726, p = 0.001; ß = 0.607, p = 0.001; 24S-OHC: F = 13.771, p < 0.001; ß = 0.565, p < 0.001; and 27-OHC: F = 8.769, p = 0.001; ß = 0.568, p = 0.001) and sAPPβ (cholesterol: F = 6.593, p = 0.003; ß = 0.494, p = 0.007; 24S-OHC: F = 10.729, p < 0.001; ß = 0.473, p < 0.001; and 27-OHC: F = 8.287, p = 0.001; ß = 0.486, p = 0.004). When separately performed in the control and the AD group, the analyses revealed significant associations in the AD group for sAPPα and CSF cholesterol (F = 5.880, p = 0.005; ß = 0.673, p = 0.002), 24S-OHC (F = 10.614, p < 0.001; ß = 0.626, p < 0.001), and 27-OHC (F = 4.902, p = 0.007; ß = 0.733, p = 0.001) as well as for sAPPβ and CSF cholesterol (F = 4.902, p = 0.007; ß = 0.496, p = 0.025), 24S-OHC (F = 7.148, p = 0.002; ß = 0.506, p = 0.001), and 27-OHC (F = 7.676, p = 0.002; ß = 0.673, p = 0.004), whereas no association was found between the sAPP forms and CSF cholesterol, 24S-OHC, and 27-OHC in the control group (data not shown). Qalb significantly contributed to all models in both the analyses in the pooled sample of controls and patients with AD and in the AD subgroup. The tolerance of all entered variables was > 0.5 and the VIF was < 2.0 for all variables except for CSF cholesterol concentrations and Qalb (tolerance = 0.328; VIF = 3.048). Carrying the APOEε4 allele did not contribute to these models. There was no association of the CSF 24S-OHC/27-OHC ratio with the CSF concentrations of the sAPP forms (data not shown).

Figure 1.

Correlations between the soluble amyloid precursor protein (sAPP)α (a) and sAPPβ (b) and 24S-hydroxy-cholesterol (24S-OHC) CSF concentrations in the patients with AD and controls with normal cognition.

No association was found between plasma concentrations of cholesterol and 24-OHC or the plasma 24S-OHC/27-OHC ratio, and the sAPP forms in the pooled sample or in the control and the AD group (not shown). Plasma 27-OHC concentrations were associated with the CSF sAPPα and sAPPβ concentrations in the pooled sample (F = 4.723, p = 0.013; ß = 0.317, p = 0.011, and F = 4.890, p = 0.004; ß = 0.310, p = 0.011) whereas a weak association with sAPPα (F = 4.860, p = 0.033; ß = 0.312, p = 0.033) and trend for sAPPβ (F = 3.492, p = 0.068; ß = 0.268, p = 0.068) were observed in the AD group. No association was found between the sAPP forms and plasma 27-OHC in the control group. In a model including plasma 27-OHC, Qalb, gender, and diagnosis (F = 19.255, p < 0.001), plasma 27-OHC independently predicted CSF 27-OHC concentrations (ß = 0.322, p = 0.007).

No association was observed between CSF Aβ1-42 and the CSF or plasma concentrations of cholesterol, 24S-OHC, and 27-OHC (data not shown).

Discussion

In this study, we identified the CSF cholesterol, 24S-OHC, and 27-OHC concentrations as predictors of both sAPPα and sAPPβ CSF levels in the group of patients with AD, but not in the control group. In addition, plasma 27-OHC concentrations were associated with the CSF concentrations of sAPPα and sAPPβ in the pooled sample of controls and patients with AD whereas a weak association with sAPPα and trend for sAPPβ were observed in the AD group.

Cerebral cholesterol levels may directly influence amyloidogenesis by modulating the activity of the enzymes involved in APP processing. The amyloidogenic processing of APP is believed to occur in close proximity to lipid rafts, cholesterol-rich membrane microdomains containing both β- and γ-secretases complexes (Cordy et al. 2003; Ehehalt et al. 2003). In cell culture and animal studies, lowering cholesterol was shown to reduce BACE1 activity and Aβ production (Simons et al. 1998; Fassbender et al. 2001; Ehehalt et al. 2003; Xiong et al. 2008). Adding cholesterol to cultured cells and cholesterol-rich diet in transgenic APP mice resulted in decreased sAPPα production (Bodovitz and Klein 1996; Refolo et al. 2000; Shie et al. 2002; Canepa et al. 2011) while lowering cellular cholesterol content was reported to increase the α-secretase activity (Kojro et al. 2001). Accordingly, it has been proposed that a cholesterol-rich environment contributes to the amyloidogenesis, whereas lowering cholesterol levels may favor the non-amyloidogenic APP cleavage, and by this reduce the amyloid production. Other experimental studies revealed an influence of the Aβ40 and Aβ42 as well as of a further APP cleavage product, the amyloid precursor protein intracellular domain, on cholesterol concentrations or cholesterol de-novo synthesis, secretion, and transport suggesting more complex interrelations between cholesterol levels and Aβ production, however (Canepa et al. 2011; Grimm et al. 2012).

It is not known to date whether the relationships described from cell culture and animal studies may be of relevance in humans. In a clinical study in patients with AD, cholesterol-lowering treatment with simvastatin was shown to lower both CSF sAPPα and sAPPβ, but the plasma and CSF levels of Aß1-42 were unchanged (Sjogren et al. 2003). In line with these findings, we found higher cholesterol levels in the CSF to be associated with increased sAPPα and sAPPβ CSF concentrations, whereas there was no association of cholesterol with Aß1-42 levels. Together, these results suggest that cholesterol levels are related to the production of the sAPP forms in AD, and that lowering cholesterol may decrease both sAPPα and sAPPβ concentrations in the CNS.

The main elimination pathway of cerebral cholesterol is the secretion of its side-chain oxidized brain-specific metabolite 24S-OHC. While more than 90% of the circulating 24S-OHC originate from the brain (Lutjohann 2006), 27-OHC is the most common extracerebral oxysterol, and 27-OHC in the CNS is thought to derive mainly from the circulation (Babiker et al. 1997). Circulating and CSF concentrations of 24S-OHC may be increased in AD patients during early clinical stages of the disease compared with controls with normal cognition (Lutjohann et al. 2000; Papassotiropoulos et al. 2002). Furthermore, increased 27-OHC levels in the brain and in the CSF have been found in patients with AD compared with controls (Heverin et al. 2004; Leoni et al. 2004). Both 27-OHC and 24S-OHC may have effects on APP and amyloid production, however, experimental studies provided at least partially conflicting results. Both 27-OHC and 24S-OHC have been reported to inhibit APP secretion and amyloid production in neurons, with 24S-OHC having much more marked effects than 27-OHC (Brown et al. 2004). Others showed total APP and sAPPα to increase after treatment with 24S-OHC, but not with 27-OHC in neuroblastoma cells (Famer et al. 2007). Further in vitro and ex vivo studies found 27-OHC to increase APP, BACE1, and Aβ levels (Sharma et al. 2008; Marwarha et al. 2010; Prasanthi et al. 2009, 2011; Dasari et al. 2010) as well as 24S-OH to induce ABCA1 expression, with higher levels of ABCA1 to be associated with increased Aβ production (Fukumoto et al. 2002). As 24S-OHC was found to increase sAPPα levels, and 27-OHC to increase BACE1 and Aβ levels, some authors proposed the 24S-OHC/27-OHC ratio to be of particular importance for the generation of amyloid in the brain (Famer et al. 2007; Prasanthi et al. 2009). These relationships were not addressed in humans so far. In our study, both 24S-OHC and 27-OHC CSF concentrations were strong and independent predictors of both sAPP forms in the AD, but not in the control group. The results suggest that both oxysterols are associated with increased sAPP production, and that these associations are related to the pathological changes in AD. Neither the plasma nor the CSF 24S-OHC/27-OHC ratios were associated with the concentrations of the sAPP forms or of Aß1-42 in the CSF, not supporting the hypothesis of the particular role of these ratios for the generation of sAPP and Aß1-42 in humans. We observed no associations of plasma and CSF concentrations of cholesterol or its metabolites with the Aß1-42 CSF levels. This finding may be explained at least in part by the contribution of multiple factors other than the Aß production to the cerebral amyloid accumulation, and by the fact that Aß accumulation in AD results in decreased CSF Aß1-42 levels.

Hypercholesterolemia in middle-aged and elder subjects was linked to significantly higher amounts of cerebral amyloid pathology in later life (Launer et al. 2001; Pappolla et al. 2003; Matsuzaki et al. 2011). Circulating cholesterol cannot cross the blood–brain barrier, however (Bjorkhem 2006). Significant influx of the major peripheral side-chain oxidized cholesterol metabolite 27-OHC from circulation into the brain has been observed in mice (Heverin et al. 2004) and humans (Heverin et al. 2005). As there is good correlation between plasma concentrations of cholesterol and 27-OHC, it has been hypothesized that hypercholesterolemia-induced Aβ accumulation in the brain may be mediated by 27-OHC (Bjorkhem 2006; Sharma et al. 2008). In our study, we observed both a strong correlation between plasma levels of cholesterol and of 27-OHC as well as an association between plasma and CSF 27-OHC concentrations. Furthermore, we found weak but significant associations of plasma 27-OHC concentrations with the CSF levels of sAPPα and sAPPβ, supporting the hypothesis that 27-OHC may link increased circulating cholesterol levels to enhanced cerebral sAPP and amyloid production.

In contrast to the findings in the AD group, there were no associations of the plasma and CSF concentrations of cholesterol, 24S-OHC, and 27-OHC with the CSF levels of the sAPP forms in the control group. This indicates that these associations may be related to specific pathological changes occurring in AD. In addition to the levels of cholesterol, 24S-OHC, and 27-OHC, the contribution of other factors possibly influencing both the production and the clearance of the sAPP forms may differ between the AD and the control group. One such important factor seems to be the blood–CSF barrier permeability. Qalb as a marker of the blood–CSF barrier permeability was previously found to be associated with CSF concentrations of the sAPP forms and of cholesterol (Lewczuk et al. 2012; Vanmierlo et al. 2011). In the present study, Qalb significantly contributed to all regression models showing associations between the CSF levels of cholesterol, 24S-OHC and 27-OHC with the CSF levels of the sAPP forms. These findings suggest that the blood–CSF barrier permeability may be a determinant of the relationship between cholesterol metabolism and sAPP production.

In summary, our results suggest that high CSF concentrations of cholesterol and of its metabolites 24S-OHC and 27-OHC are associated with increased production of both sAPPα and sAPPβ. Influx of 27-OHC from the circulation into the CNS may link hypercholesterolemia to increased cerebral amyloid production and accumulation, and may explain the association of hypercholesterolemia with a higher risk of AD. The underlying mechanisms and the pathophysiological relevance of the observed relationships between cholesterol metabolism and sAPP levels for amyloidogenesis remain to be clarified.

Acknowledgements

This study was supported by the German Ministry of Education and Research (BMBF): Competence Network Degenerative Dementias (KNDD): 01GI0711. P. Lewczuk is a consultant of Innogenetics, Ghent, Belgium. There are no other potential conflicts of interest.

Ancillary