Alzheimer’s disease: the lipid connection


Address correspondence and reprint requests to Tobias Hartmann, Universität des Saarlandes, Uniklinikum Homburg, Neurobiologie, Neurologie, Gebäude 90, 66421 Homburg/Saar, Germany.


Alzheimer’s disease is the most common neurodegenerative disorder. This short review summarizes the current knowledge about the role of lipids, especially cholesterol, sphingolipids, plasmalogens, and polyunsaturated fatty acids in Alzheimer’s disease etiology, pathogenesis, risk-factors, prevention, treatment, and the function of the amyloid precursor protein and the amyloid peptides in lipid homeostasis.

Abbreviations used

arachidonic acid


Alzheimer’s disease


disintegrin and metalloprotease


anterior pharynx defective 1 homologue


amyloid precursor-like proteins APLP


amyloid precursor protein

amyloid peptide


clinical dementia ratings




docosahexaenoic acid


familiar forms of AD


3-hydroxy-3-methylglutaryl-CoA reductase




preserilin enhancer 2 homologue


protein kinase C






polyunsaturated fatty acids


reactive oxygen species






tumor necrosis α-converting enzyme

Alzheimer’s disease

Alzheimer’s disease (AD) is the most common cause of dementia among neurodegenerative diseases in the elderly population affecting about 25 million people worldwide (2005). In industrialized nations, it is the fourth common cause of death and its incidence and costs are predicted to increase due to increasing life expectancy and average age (Brookmeyer et al., 1998; von Strauss et al., 1999). Hence it was estimated that the number of affected people will double every 20 years (Ferri et al., 2005), emphasizing the role of AD as a major health problem that will gain even more ethical and economical relevance in near future.

The characteristic histopathological changes of AD are extracellular neuritic plaques composed of a dense amyloid core of the β-amyloid peptide (Aβ) and intracellular neurofibrillary tangles that consist of an abnormally phosphorylated form of the protein tau. Clinical manifestations of this progressive and irreversible dementia that is eventually fatal are memory loss and cognitive decline (for reviews see Hardy and Higgins, 1992; Masters and Beyreuther, 1989; Masters et al., 2006; Selkoe, 2001; Sisodia and St George-Hyslop, 2002).

Amyloid precursor protein processing

The main component of the extracellular amyloid deposits, Aβ, is generated by sequential proteolytic cleavage of the amyloid precursor protein (APP). APP is a 695–770 amino acid type I transmembrane protein with a large hydrophilic N-terminal extracellular domain, a single hydrophobic transmembrane domain comprising 23 amino acids (aa) and a small C-terminal cytoplasmic domain (Kang et al., 1987; Younkin, 1991). It belongs to an evolutionary conserved protein family including the amyloid precursor-like proteins (APLP) 1 and 2 that are not only ubiquitously expressed in many mammalian tissues but homologues are also found in invertebrates. Proteolytic processing of APP occurs in two different pathways: a non-amyloidogenic and an amyloidogenic way (see Fig. 1).

Figure 1.

 Processing of the amyloid precursor protein (APP). The non-amyloidogenic pathway is initiated by α-secretase and results in the cleavage of APP within the amyloid peptide (Aβ) domain, thus precluding the generation of Aβ. In the amyloidogenic or pathological pathway, APP is cleaved by β-secretase whereby the N-terminus of the Aβ peptide is generated. Both, α- and β-secretase cleavage occur within the APP ectodomain at a short distance to the transmembrane domain, releasing soluble forms of APP, sAPPα and sAPPβ, respectively. The remaining membrane-bound C-terminal fragments of APP, C α-terminal fragment (CTF) and β-CTF, are substrates for γ-secretase, a ubiquitous multimeric protease. The γ-secretase cleavage of α-CTF results in the generation of a truncated non-amyloidogenic peptide (p3), whereas γ-secretase cleavage of β-CTF leads to the generation of the amyloidogenic Aβ. Independent of amyloidogenic or non-amyloidogenic processing of APP, the intracellular APP domain (AICD) is released into the cytosol and may migrate to the nucleus where it is supposed to activate gene transcription. As β-CTF can be cleaved at different sites by γ-secretase, Aβ varies in its length; the main variants are the 40 amino acid long peptide Aβ40 and the two amino acids longer peptide Aβ42. Because of its increased hydrophobicity, Aβ42 is able to form aggregates and is the predominant species of Aβ in senile plaques (Iwatsubo et al., 1994). The relevance of Aβ42 in Alzheimer’s disease (AD) is further supported by studies of the rare familiar forms of AD (FAD). Mutations in the genes encoding APP and presenilin, a component of the γ-secretase, increase the production of Aβ42. Moreover, the age of disease onset correlates directly with the Aβ40/42 ratio (Duering et al., 2005).

Enzymes involved in proteolytic processing of APP


The α-secretase cleavage does not only preclude the formation of the Aβ peptide but also causes the release of the large ectodomain of APP (sAPPα) which has neuroprotective and memory enhancing effects (Furukawa et al., 1996; Mattson et al., 1999; Meziane et al. 1998). The proteases a disintegrin and metalloprotease (ADAM) 9, 10, and 17 have emerged as putative α-secretase candidates. The ADAMs are type I integral membrane proteins with a multidomain structure including an N-terminal signal peptide, a pro-domain, a cleavage site for the prohormone convertases, a zinc-binding motif, a cysteine-rich/disintegrin-/EGF-like domain followed by a hydrophobic transmembrane domain, and a short cytoplasmatic tail. The action of the α-secretase is up-regulated by protein kinase C (PKC) (Buxbaum et al., 1993; LeBlanc et al., 1998) and regulated via other signaling mechanisms (Allinson et al., 2003; Hwang et al., 2006; Mills and Reiner, 1999). Tumor Necrosis α-Converting Enzyme (TACE; also called ADAM17) was initially identified as protease being responsible for the release of the inflammatory cytokine tumor necrosis factor α (Black et al., 1997; Moss et al., 1997). First evidence of TACE being involved in α-secretase cleavage of APP was reported by Buxbaum et al. (Buxbaum et al., 1998) who showed that fibroblasts from TACE-knockout (KO) mice are defective in their PKC-stimulated APP secretion. Inhibition of TACE in primary neurons leads to the suppression of regulated α-secretase activity but not of constitutive α-secretase activity (Blacker et al., 2002). In immunohistochemical analysis of human brains, TACE was found to localize to neurons and in addition colocalization of TACE with amyloid plaques in AD brains was described (Skovronsky et al., 2001). The over-expression of another α-secretase candidate, ADAM10, in human embryonic kidney cells 293 resulted in elevation of both constitutive and PKC-regulated release of sAPPα (Lammich et al., 1999). Like TACE, ADAM10 specifically cleaves synthetic peptides containing the α-secretase cleavage site. However, embryonic fibroblasts from ADAM10-deficient KO mice have no alteration in their production of sAPPα suggesting that a collective of proteases is able to perform α-secretase cleavage via compensatory mechanisms (Hartmann et al., 2002). ADAM9 was identified and cloned as disintegrin protein first from mouse lung (Weskamp et al., 1996). Coexpression of ADAM9 with APP695 in COS cells upon phorbol ester treatment revealed increased production of sAPPα (Koike et al., 1999). The deletion of the ADAM9 in mice resulted in no major abnormalities during development or adult life (Weskamp et al., 2002) again supporting the hypothesis that several members of the ADAM family are involved in APP processing and that there may be functional overlap between these proteases.


The β-secretase (also BACE for β-site APP-cleaving enzyme), the first prerequisite for generation of Aβ, is a type I integral membrane protein. It belongs to the pepsin family of aspartyl proteases but is not inhibited by the classical aspartyl protease inhibitor pepstatin (Vassar, 2001). The close homolog of BACE1, BACE2, exhibits an α-secretase-like activity that cleaves within the Aβ domain (Farzan et al., 2000) and does thus not contribute to the amyloidogenic processing of APP. The homozygous KO of the BACE1 gene leads to the abolishment of Aβ generation in embryonic cortical neurons (Cai et al., 2001) and in mice who display an normal phenotype (Luo et al., 2001; Roberds et al., 2001). Recently, it was reported that BACE1 is required for peripheral nerve myelination and correct bundling of axons by Schwann cells (Willem et al., 2006). Interestingly, BACE1 displays a higher affinity for APP bearing the familiar forms of AD (FAD) -associated Swedish mutation. For this reason, it was supposed that APP is not the main substrate for BACE1 and/or that α-secretase has a higher affinity for the wild-type APP. Indeed, other substrates for BACE1 such as the sialyl-transferase ST6Gal I and other sialyltransferases (Kitazume et al., 2004; Kitazume et al., 2006), P-selectin glycoprotein ligand-1 (Lichtenthaler et al., 2003), and Aβ itself (Fluhrer et al., 2003) have been identified. BACE1 is mainly expressed in neurons and the majority of BACE molecules are located in the Golgi apparatus and endosomes (Vassar et al., 1999). After passing through the secretory pathway, BACE1, like APP, clusters within lipid rafts where preferentially the APP cleavage by BACE1 takes place (Ehehalt et al., 2003). In addition, it was demonstrated that BACE1 is internalized from the cell surface to the endosomes and can recycle back to the cell surface similar to APP (Huse et al., 2000). For this reason and because of the acidic pH optimum of BACE1 it was suggested that β-secretase cleavage of APP occurs predominantly in acidic compartments like endosomes (Koo and Squazzo, 1994, Vassar, 2001#12).


The γ-secretase is responsible not only for APP-C-terminal fragment cleavage but also for the cleavage of other membrane-associated fragments derived from type I integral membrane proteins such as APLP1, APLP2, Notch 1-4, ErbB-4, E-cadherin, LRP, Nectin 1-α, Delta, Jagged, and CD44. The γ-secretase cleaves within the middle of the transmembrane domain, a process also known as regulated intramembrane proteolysis (Brown et al., 2000; Grziwa et al., 2003; Selkoe and Kopan, 2003). At least the four proteins presenilin 1 (PS1) or presenilin 2 (PS2), anterior pharynx defective 1 homologue (APH1), preserilin enhancer 2 homologue (PEN2), and nicastrin are required to form the active γ-secretase complex.

The PSs are assumed to have nine transmembrane domains (Laudon et al., 2005; Oh and Turner, 2005a) with the N-terminus on the cytosolic side (Doan et al., 1996) and luminal C-terminus (Oh and Turner, 2005b). Both proteins, the 46-kDa PS1 and 55-kDa PS2, share 67% amino acid identity. The protein is endoproteolyzed into two fragments, a N-terminal fragment and a C-terminal fragment that remain associated and display the mature, active form of PS (Ratovitski et al., 1997; Thinakaran et al., 1996). More than 100 autosomal dominant PS point mutations have now been identified, which all cause the aggressive early-onset form of AD. The KO of both PSs or the mutation of the aspartic residues D257 and D385 result in complete loss of γ-secretase activity and Aβ production (De Strooper, 2003; Grimm et al., 2002), identifying the PSs as active centre of the γ-secretase complex. Nicastrin is a highly glycosylated 120–140 kDa type I integral membrane protein, with almost the entire protein being lumenal/extracellular. APH1 (Francis et al., 2002; Goutte et al., 2002; Lee et al., 2002) and PEN2 (Francis et al., 2002; Steiner et al., 2002) encode proteins of seven and two predicted transmembrane domains, respectively. APH1 together with nicastrin are thought to serve as initial scaffolding proteins during the stepwise assembly of the complex. The association of the hairpin PEN2 protein to the complex triggers the endoproteolytic cleavage of the PS holoprotein (Kaether et al., 2006). All of these proteins form a single complex and are the essential components for γ-secretase activity (Fraering et al., 2004).

Link between lipids and Alzheimer’s disease

First principal indications that lipids may play an important role in APP processing and Aβ production are given by the common feature that all proteins involved in processing are integral membrane proteins. Moreover, the Aβ producing γ-secretase cleavage takes place in the middle of the membrane suggesting that the lipid environment of the cleavage enzymes influences Aβ production and hence AD pathogenesis (Grziwa et al., 2003).

Cholesterol and Alzheimer’s disease

The brain is the most cholesterol-rich organ in the body and contains about 25% of total body cholesterol. Cholesterol is an essential factor of cell membranes and plays a crucial role in the development and maintenance of neuronal plasticity and function (Pfrieger, 2003). The homeostasis of cholesterol including synthesis, removal, storage, or transport within the brain is strictly regulated. One of the key enzymes in the biosynthetic pathway of cholesterol is the 3-hydroxy-3-methylglutaryl-CoA reductase (HMGR) which catalyzes the formation of mevalonate (Michal, 1999) (see Fig. 2).

Figure 2.

 Pathway of the cholesterol de novo synthesis, highlighted are the interaction points of the cholesterol lowering drugs statins and BM15.766. Note that statins act at the very beginning and BM15.766 very end of this pathway.

As this enzyme catalyzes the rate limiting and committed step of the cholesterol biosynthesis, HMGR is both a target for the physiological regulation of the cholesterol homeostasis and for pharmacological inhibitors (e.g. statins) (Steinberg, 2006).

In mature neurons, endogenous cholesterol synthesis is reduced and surplus cholesterol required for synaptic plasticity is believed to be supplied by glial cells, primarily astrocytes (Poirier et al., 1993). The major part of cholesterol present in the brain is synthesized in the CNS. Transport of cholesterol from peripheral circulation to the brain is prevented by an intact blood–brain barrier (only the oxidized cholesterol products 27-hydroxycholesterol and 24S- hydroxycholesterol (Papassotiropoulos et al., 2000) can pass the blood–brain barrier). The exchange rate between peripheral and brain cholesterol is less than 1% per day. Therefore, it has to be assumed that cholesterol homeostasis of the brain is predominantly, but not entirely independent of blood cholesterol level (Dietschy and Turley, 2004).

Hypercholesterolemia is an early risk factor for the development of amyloid pathology and longitudinal, population-based studies demonstrated that cholesterol is associated with AD in later lifespan (Kivipelto et al., 2001).

In addition, Corder et al. reported that the risk for AD increases and mean age of onset decreases from 84 to 68 years with increasing number of APOEε4 alleles in families with late onset AD (Corder et al., 1993). Homozygozity for APOEε4 increases the relative risk by 15- to 18-fold. This makes, beside age, APOE the most relevant risk factor for AD. In respect to the AD-lipid link, it is important to notice that APOE is a lipoprotein involved in lipid trafficking between neurons and astrocytes, although the molecular mechanism how APOE alters the AD age of onset is unclear.

Molecular and mechanistic evidence points toward a direct role of cholesterol in AD pathogenesis, e.g. cholesterol enhances γ-secretase-mediated Aβ production (Fassbender et al., 2001; Simons et al., 1998; Wahrle et al., 2002; Zha et al., 2004), whereas cholesteryl esters stimulate non-amyloidogenic APP degradation (Puglielli et al., 2001).

The important role of cholesterol in APP processing is further substantiated by experiments in which cholesterol de novo synthesis is inhibited by pharmacological drugs. In these experiments, the HMGR inhibitors simvastatin or lovastatin were able to reduce intracellular and extracellular levels of Aβ42 and Aβ40 peptides in primary cultures of hippocampal neurons and mixed cortical neurons. In vivo, guinea pigs treated with high doses of simvastatin showed a strong and reversible reduction of cerebral Aβ42 and Aβ40 levels in the cerebrospinal fluid and brain homogenate (Fassbender et al., 2001). BM15.766, which reduces cholesterol de novo synthesis by inhibition of 7-dehydro-cholesterol-reductase, one of the final steps of cholesterol biosynthesis, decreases amyloid load as well, emphasizing that the amyloid reducing effect is not limited to statins but involves sterol depletion (Refolo et al., 2001).

Similar results were achieved by physical extraction of cholesterol with cyclodextrin, which depletes cholesterol out of the membrane (Fassbender et al., 2001). In vivo experiments with rabbits fed a cholesterol enriched diet showed an increasingly mild-to-moderate-to-severe accumulation of intracellular immunolabeled Aβ (Sparks et al., 2005). Moreover, cholesterol-enriched diet causes increased cerebral amyloid load in APP transgenic mice (Refolo et al., 2001; Shie et al., 2002).

In line with these results, statin treatment is associated with a reduced AD risk in some epidemiological studies (Jick et al., 2000; Rockwood et al., 2002; Wolozin et al., 2000). Pilot prospective trials came out with mixed results. Treatment with high statin dosage (80 mg simvastatin/day) was able to reduce cerebral Aβ levels and induced a small cognitive benefit in AD patients (Simons et al., 2002). A second high dosage study (Sparks et al., 2005), using 80 mg/day atorvastatin, reported a similar benefit for mild and moderate AD, possibly due to the extended treatment duration. In contrast, low statin dosage treatment for up to 12 month did not show influence on Aβ production (Hoglund et al., 2004).

The finding that Aβ aggregates preferentially bind cholesterol (Avdulov et al., 1997) and the increase in brain cholesterol levels during early disease progression (Cutler et al., 2004; Wood et al. 2002) further support the importance of cholesterol and lipids in AD.

Sphingolipids and APP processing

Sphingolipids represent another major lipid class of eucaryotic cell membranes. The main fraction of sphingolipids is made up by sphingomyelin (SM) whose head group is represented by phosphorylcholine esterified to the 1-hydroxyl group of ceramide (Cer). Cer is the most important branching point in the sphingolipid synthesis resulting in different metabolites summarized in Fig. 3.

Figure 3.

 Important metabolites and enzymes in the sphingolipid pathway. Ceramide is the central metabolite of several molecular pathways. For several of these links to Alzheimer’s disease have been reported.

In contrast to cholesterol SM decreases Aβ production by inhibition of the γ-secretase (Grimm et al., 2005) which was confirmed by sphingomyelinase (SMase) inhibition experiments. SMases are enzymes that break down SM to Cer. Inhibition of these degrading enzymes lead to elevated SM levels and reduced Aβ-level. Interestingly, SMase activity is increased in PS-FAD mutations, which demonstrate the involvement of SMases in AD (Grimm et al., 2005).

Apart from SM, other sphingolipids are linked to AD as well. Sphingosine is assumed to affect the PKC (Smith et al., 2000) and therewith to influence α-secretase. Further studies pointed out that in AD brain, the total ganglioside pattern is significantly altered, which emphasizes the importance of glycosphingolipids in AD (Kracun et al., 1990). For instance, elevated GM1 (monoasialoganglioside) and increased sialidase activity have been observed in AD patients (Blennow et al., 1991; Svennerholm et al., 1994). Micropathological analysis of amyloid plaques revealed that GM1-like cholesterol binds to Aβ and it was suggested that the GM1/Aβ complex might serve as a seed for amyloid fibril formation. Zha et al. demonstrated that GM1 is the most effective natural compound increasing Aβ production. In presence of GM1 increased γ-secretase and decreased α-secretase activity resulted in up to a 10-fold increase of Aβ levels (Zha et al., 2004).

Inhibition of one of the rate-limiting enzyme in the ganglioside synthesis, the glucosylceramidsynthase, decreased Aβ production. In addition, sAPP levels and APP maturation also decreased suggesting that transport of APP is influenced by an altered ganglioside level. Similar data were obtained in mouse models and in cell lines with defective ganglioside biosynthesis (Tamboli et al., 2005). In agreement with these findings glycosphingolipids have been shown to influence the forward transport at distinct steps at the secretory pathway (Muniz et al., 2001). However, injection of GM1 into βAPP transgenic mice decreased amyloid burden and the level of soluble Aβ. It was assumed that GM1 in peripheral blood binds to Aβ, which is then degraded (Matsuoka et al., 2003).

The reduced activity of α-secretase by gangliosides could also be mechanistically explained by inhibition of PKC, which enhances α-cleavage (Cedazo-Minguez et al., 2001). Other explanations were given by the stiffening effect of GM1 on membranes. In this model, it is assumed that gangliosides inhibit lateral movement and reduce the required contact between substrate and enzyme, which results in a decreased α-cleavage.

The effects of (glyco-)sphingolipids on Aβ production underline that not only cholesterol but also various other lipids are influencing APP processing. Thus, pharmaceutical drugs mediating the homeostasis of other lipids might have beneficial effects for AD treatment as well. DL-PDMP, a potent inhibitor of the glycosylceramid synthase or GW4869, an inhibitor of the nSMase (Grimm et al., 2005), could already demonstrate to have such effects on APP processing (Tamboli et al., 2005).

Other lipid changes in AD brains

Various other lipid alterations occur during the pathogenesis of AD besides the above mentioned modifications described in (Han, 2005; Prasad et al., 1998; Wells et al., 1995) and others.

A significant and selective decrease of ethanolamine plasmalogens (PLs) relative to phosphatidyl ethanolamine in autopsy brain samples from AD patients that does neither occur in Huntington’s nor in Parkinson’s disease was identified in several studies (Ginsberg et al., 1995; Guan et al., 1999; Wells et al., 1995). PLs are characterized by a vinyl-ether bond at the sn-1 position of their glycerol backbone. Ethanolamine PLs composes up to 90 mol% of total ethanolamine glycerophospholipids content or about 30 mol% of total phospholipids in neuronal cell membranes hence being a major membrane component (Han et al., 2001). Besides AD also other neuropathological conditions have been linked with decreased PL levels such as ischemia, spinal cord injury, Down syndrome, Zellweger syndrome, adrenoleuko-dystrophy, Niemann–Pick disease, multiple sclerosis, malnutrition, and fetal alcohol syndrome (Farooqui and Horrocks, 2001). Han et al. (Han et al., 2001) correlated the PL deficiency in AD with different clinical dementia ratings (CDR) of AD patients. A dramatic decrease up to 40 mol% in PL content in white matter was detected already at a very early stage of AD (CDR 0.5). In gray matter, PL levels are reduced to 10 mol% in early stages (CDR 0.5) and to 30 mol% at severe dementia stage (CDR 3).

Nevertheless, the role of PL reduction in AD is poorly understood. It may be related with synapse loss and neurodegeneration because the PL reduction may lead to membrane instability (Ginsberg et al., 1998). One explanation for this depletion might be the increased oxidative damage in the AD brain due to several factors including age-related decrements in energy availability, mitochondrial dysfunction, glutamateric neurotransmission, and accumulation of Aβ (Albers and Beal, 2000; Bassett and Montine, 2003). Aβ itself has been described as being an oxidant species (Davis, 1996; Markesbery and Carney, 1999). Cultured embryonic rat brain oligodendrocytes treated with Aβ up to 24 h showed a significant decrease in ethanolamine PL content (Cheng et al., 2003). Moreover, PLs are affected by reactive oxygen species (ROS) due to their vinyl-ether bond that makes them susceptible to oxidation. ROS are generated under oxidative stress in AD and thus PL depletion may occur due to ROS-mediated degradation (Han, 2005). Indeed, elevated degradation products generated by oxidation with lipid peroxylradicals have been found in the aging human brain (Weisser et al., 1997). On the other hand, it was observed that in AD brains phospholipase A2 levels are elevated compared with age-matched controls (Farooqui et al., 2003). This is accompanied with increased levels of lipid peroxides, lysophosphatidylcholine, 4-hydroxynonenal [product of arachidonic acid (AA) peroxidation], phosphomonoesters, and phosohodiesters (Mulder et al., 2003; Pettegrew et al., 1995; Selley et al., 2002).

Sulfatides are depleted up to 93% in gray matter and up to 58% in white matter, whereas ceramides were elevated more than threefold in brains of AD patient at a stage of very mild dementia (Han et al., 2002). Sulfatides (galactosyl-3′-sulfate ceramides) represent acidic glycosphingolipids containing sulfate esters on their oligosaccharide chains (see Fig. 3). In mammals, sulfatides are present in various tissues such as brain, kidney, testis, erythrocytes, platelets, and granulocytes (Ishizuka, 1997). They are involved in many biological processes including protein trafficking, cell growth, signal transduction, cell–cell recognition, neuronal plasticity, and cell morphogenesis (Vos et al., 1994, Ishizuka, 1997#77).

Ceramides are composed of a long-chain or sphingoid base linked to a fatty acid via an amide bond. Although they show a very low abundance in most tissues, ceramides exert important biological functions as second messenger mediating diverse cellular processes. The reasons for the Cer and sulfatide alterations are still unresolved, but the elevation of Cer might result from the degradation of sulfatide with the loss of a galactosylsulfate, increased de novo synthesis, or the increased hydrolysis of SM or cerebrosides (Han et al., 2002). As Cer can modulate a variety of cellular effects ranging from proliferation and differentiation to growth arrest and apoptosis (Venkataraman and Futerman, 2000). Cer-induced apoptosis might play an important role under pathological conditions such as in AD (Thevissen et al., 2006).

Involvement of n-3 and n-6 fatty acids in AD

The nervous system is highly enriched in polyunsaturated fatty acids (PUFAs), especially in docosahexaenoic acid (DHA, 22:6ω-3) that alone constitutes more than 17% by weight of the total fatty acids in the brain of adult rats (Hashimoto et al., 2002). PUFA oxidation, either enzyme-catalyzed or free radical-mediated, is thought to contribute to the molecular pathogenesis of AD. Studies demonstrated that AD is associated with depletions of DHA in the brain, specifically in hippocampus, and also AA (20:4ω-6) was found to be reduced (Soderberg et al., 1991; Tully et al., 2003). Therefore, recent investigations have focused mainly on these two PUFAs that differ in their distribution within the brain: DHA is highly enriched in neuronal membranes, whereas AA shows an even distribution in gray and white matter and in different cell types (Montine and Morrow, 2005). The released AA and DHA are metabolized to eicosanoids and docosaniods, respectively, which are involved in signal transduction as second messengers (Di Marzo, 1995; Farooqui et al., 1995). Numerous epidemiological studies indicate an inverse correlation between DHA intake and AD incidence (Barberger-Gateau et al., 2002; Kalmijn et al., 1997; Morris et al., 2003). The daily administration of ω-3 fatty acids (DHA and eicosapentaenoic acid) to patients at a very mild stage of AD resulted in positive effects, but in patients who already had mild-to-moderate AD no beneficial effect was observed (Freund-Levi et al., 2006). Furthermore, dietary DHA reduces the production and accumulation of Aβ and decreases Aβ42 levels in aged Alzheimer mouse models (Lim et al., 2005). In rats, administration of DHA had positive effects on the learning ability and suppressed the increase in lipid peroxide and ROS levels in the cerebral cortex and hippocampus, suggesting an elevated antioxidative defense (Hashimoto et al., 2002). The intake of DHA or fish oil (contains DHA and eicosapentaenoic acid) rich diets by APP/PS1 transgenic mice resulted in decreased hippocampal Aβ levels (Oksman et al., 2006). In the same study, cells treated with fish oil revealed a decrease in Aβ production and neuroblastoma cells incubated with DHA showed reduced Aβ levels in a dose-dependent manner suggesting that DHA is involved in down-regulation of the amyloidogenic pathway.

Physiological function of Aβ in lipid homeostasis

Beside the pathological aspect of Aβ in AD very little is known about a physiological function of this small peptide. In the last few years, cumulating evidences suggest that APP processing and especially Aβ is tightly connected to lipid homeostasis. In the following complex regulatory cycles between Aβ and different lipids are elucidated. As Aβ has an ambiguous role both as a physiological regulator and as a toxin believed to be responsible for AD, understanding the physiological function of Aβ might help to elucidate the early events in AD pathogenesis.

As reviewed above, lipid environment affects Aβ production. First indications that not only lipids influence APP processing but inversely APP processing also influences lipid metabolism were observed when fibroblasts deficient in PS and hence unable to produce Aβ were studied. In absence of γ-secretase activity membrane fluidity is decreased (Grimm et al., 2006). As cholesterol is an important regulator of membrane fluidity, this suggests a potential link between cholesterol homeostasis and γ-secretase activity. Indeed in absence of PS, cholesterol levels are increased. Incubation with γ-secretase inhibitors in wild-type cells revealed that γ-secretase activity is responsible for this effect. Similar results were obtained with APP KO cells, linking APP processing with lipid homeostasis. In line with these results Aβ rescued increased cholesterol levels in APP or PS KO cells by inhibiting the enzyme catalyzing the committed step reaction of the cholesterol synthesis, the HMGR. Similar altered cholesterol levels are found in mouse brains deficient of APP or PS. FAD mutations with increased Aβ42 and decreased Aβ40 levels, show an increased cholesterol level, too (Grimm et al., 2005). Taken together with the finding that cholesterol affects Aβ production and Aβ inhibits cholesterol synthesis a negative feed back cycle can be postulated (see Fig. 4).

Figure 4.

 Model of regulatory cycles between sphingolipids, cholesterol, and amyloid precursor protein (APP) processing. Complex regulatory cycles connect APP processing with cholesterol and sphingolipid homeostasis. Central to this is the lipid modulated release of amyloid peptide (Aβ) from APP by γ-secretase and the Aβ-mediated regulation of sphingomyelinase (SMase) and 3-hydroxy-3-methylglutaryl-CoA reductase (HMGR) activity, which alters cholesterol and sphingomyelin levels. Only this central part of these regulatory cycles is shown here. Further cycles include e.g. the other APP-secretases and additional cholesterol and sphingomyelin regulatory mechanisms.

It is unclear if the regulation of HMGR by Aβ40 is direct or mediated via another metabolite. As the HMGR is localized in the membrane of the endoplasmic reticulum (Friesen and Rodwell, 2004), both options seem to be possible, but due to the low Aβ40 level in the endoplasmic reticulum (Hartmann et al., 1997) a signal transmission via another molecule might be preferred.

This regulatory cycle shows high analogies to the regulatory system described by Brown and Goldstein. Both regulatory cycles show a lipid-controlled regulated intramembrane proteolysis mechanism. In the Brown and Goldstein system, high cholesterol inhibits site-1 protease cleavage of sterol regulatory element-binding protein and therefore decreases the transcription of the HMGR (Brown et al., 2002), whereas cholesterol increases γ-secretase cleavage of APP, resulting in high Aβ levels, which inactivate HMGR enzyme activity.

Beside cholesterol homeostasis, APP processing and in particular Aβ is highly involved in sphingolipid homeostasis. It could be shown that both APP and PS deficient or γ-secretase inhibitor treated cells have an increased SM level. The enzymes, affected by PS and APP are SMases, which degrade SM to Cer (Grimm et al., 2005). Physiological concentrations of Aβ42 directly activate SMase, whereas Aβ40 has only a minor effect on SMase. This results in a further regulatory cycle summarized in Fig. 4. As before Aβ regulated cholesterol levels, this aspect of lipid homeostasis is altered by PS-FAD mutations which cause increased SMase activity and therefore a correspondingly decreased SM level.


There is currently no non-symptomatic treatment for AD. Some small pilot clinical trials have thus far produced a delay in disease progression at best, whereas all major clinical trials failed to provide sufficient evidence for any kind of effective non-symptomatic treatment.

There are few lines of research which show as much intrinsic linkage to AD than the study of lipids covering all aspects from most basic mechanism to epidemiological risk factors and therapeutic approaches. Such a broad connection is usually a good basis to seek for therapeutic approaches. Targeting lipid homeostasis carries one additional advantage, and it can often be designed to minimize potential adverse side effects. This might prove especially important for the design of preventive treatments and treatment of the earliest stages of AD.