Worldwide currently there are more than 46 million people suffering from dementia and the number of affected individuals is estimated to double every 20 years. Alzheimer’s disease (AD) is a devastating neurodegenerative disorder, which is the most common cause of dementia in the elderly population. Clinically AD is characterized by a progressive loss of cognitive brain functions leading to memory dysfunction, impaired judgment, disorientation and finally to a total loss of memory and personality (Plassman et al., 2007; World Alzheimer Report, 2015). AD-patients typically die in average within 3–10 years after diagnosis due to secondary disorders (Zanetti et al., 2009). The clinical symptoms of AD might be caused by an extensive loss of synapses and neurons leading to a strong hippocampal and cortical atrophy (Scheff and Price, 1993; Gómez-Isla et al., 1996; Mouton et al., 1998; Dickerson et al., 2001). The characteristic neuropathological hallmarks of the disease are intracellular neurofibrillary tangles (NFTs) and extracellular localized amyloid plaques. While the NFTs are composed of the microtubuli-associated protein tau in a hyperphosphorylated state (Grundke-Iqbal et al., 1986a,b), the amyloid plaques are mainly built up of amyloid-β (Aβ) peptides. Aβ-peptides are hydrophobic, 38–43 amino acid long products generated by the sequential proteolytic processing of the amyloid precursor protein (APP; Glenner and Wong, 1984; Masters et al., 1985; Kang et al., 1987). The significant cerebral accumulation of Aβ, starting several years prior to the first symptoms, is respected to trigger the disease process (Glenner and Wong, 1984; Glenner, 1989; Hardy and Higgins, 1992; Hardy and Selkoe, 2002). Especially the accumulation of Aβ42 (indicating 42 amino acids), which is the major Aβ species found in neuritic plaques, is considered to initiate AD progression (Iwatsubo et al., 1994; Tamaoka et al., 1995). Due to the additional hydrophobic amino acids isoleucine and alanine Aβ42 has a higher tendency to aggregate compared to the more prevalent Aβ40 (indicating 40 amino acids; Jarrett et al., 1993). Increasing evidence suggests small oligomers of Aβ to represent the most toxic form of the peptide (Lambert et al., 1998; Lesné et al., 2006; Shankar et al., 2008). Several mechanisms are discussed to contribute to Aβ neurotoxicity, among them the induction of inflammatory processes, a disruption of calcium homeostasis and membrane integrity, cholinergic and mitochondrial dysfunction and increased oxidative stress (Grimm and Hartmann, 2012).

There are two forms of AD, the more common sporadic AD with a disease onset after the age of 65 (late onset AD, LOAD) and the genetically based form (familial AD, FAD) with an earlier manifestation of symptoms. The two variants are basically distinguishable from each other in clinical and neuropathological terms. Less than 5% of all AD-cases belong to FAD which is caused by mutations in the genes encoding for APP and the presenilins (PS) 1 and 2, proteins involved in proteolytic APP-processing (Levy et al., 1990; Goedert et al., 1994; Levy-Lahad et al., 1995; Sherrington et al., 1995; Tanzi, 2012). Besides aging, hypercholesterolemia, hypertension, atherosclerosis, homocysteinemia, diabetes mellitus and obesity are discussed as non-genetic risk factors for LOAD (Barnes and Yaffe, 2011; Polidori et al., 2012). The ε4 allele of the apolipoprotein E (ApoE) has been identified as the most important genetic risk factor for the sporadic form of the disease (Corder et al., 1993; Strittmatter et al., 1993).

As already mentioned, Aβ is generated by proteolytic processing of the precursor protein APP. APP is a ubiquitously expressed type I-transmembrane protein cycling between the plasma membrane and acidic intracellular compartments (Haass et al., 1992; Koo and Squazzo, 1994; Thinakaran and Koo, 2008). It consists of a large ectodomain, a single transmembrane domain and a short intracellular part. APP belongs to an evolutionary conserved protein family including the APP-like proteins 1 and 2 (APLP1, APLP2) in mammals. APP can be sequentially cleaved via two different pathways (Haass et al., 1992; Thinakaran and Koo, 2008; De Strooper, 2010; Figure 1). In the predominant non-amyloidogenic processing pathway the generation of Aβ is precluded. It is initiated by the α-secretase dependent cleavage of APP within the Aβ-domain shedding off the soluble ectodomain sAPPα and generating the membrane-anchored C-terminal fragment (CTF) C83 (indicating 83 amino acids). Members of the ADAM (a disintegrin and metalloprotease) protein family have been identified as catalytically active α-secretases with ADAM10 representing the physiologically relevant, constitutive α-secretase in neurons (Lammich et al., 1999; Kuhn et al., 2010). In contrast, the aspartyl protease β-site APP cleaving enzyme 1 (BACE1) initiates the amyloidogenic APP-processing pathway generating the membrane-spanning CTF C99 (indicating 99 amino acids) and releasing sAPPβ into the extracellular space (Vassar et al., 1999). The two alternative pathways differ in their subcellular localization: due to the acidic pH-optimum of BACE1 the amyloidogenic APP-processing is localized in acidic intracellular compartments, while non-amyloidogenic APP-processing mainly takes place at the cell surface (Parvathy et al., 1999; Grbovic et al., 2003; Carey et al., 2005). In both pathways the CTFs are subsequently processed by the γ-secretase complex, which consists of the proteins PS1 or PS2 as the catalytic core, Aph1 (anterior pharynx defective 1) a or b, PEN2 (presenilin enhancer 2) and nicastrin (Baulac et al., 2003; Edbauer et al., 2003; Kimberly et al., 2003). The γ-secretase possesses the unusual property to cleave its substrates within their transmembrane domains after shedding off the ectodomain, a process called regulated intramembrane proteolysis (RIP; Brown et al., 2000; Lichtenthaler et al., 2011). This catalytic activity leads to the generation of the non-toxic peptide p3 out of C83 and of Aβ out of C99 combined with the release of APP intracellular domain (AICD) into the cytosol in both processing pathways (Passer et al., 2000; Kakuda et al., 2006; Grimm and Hartmann, 2012). Due to multiple γ-secretase cleavage sites within the transmembrane domain of APP, the generated Aβ- and AICD-peptides can vary in length (Funamoto et al., 2004; Qi-Takahara et al., 2005; Kakuda et al., 2006). AICD is reported to translocate to the nucleus and to regulate the transcription of target genes, among them the genes encoding for APP, BACE1, the Aβ-degrading protease neprilysin (NEP) as well as several enzymes involved in lipid metabolism (Cao and Südhof, 2001; von Rotz et al., 2004; Grimm et al., 2011b,d, 2012c, 2013, 2015a).

A link between AD pathology and lipids was already observed more than a century ago by Alois Alzheimer, who described a higher occurrence of “adipose inclusions” or “lipoid granules” in post mortem AD-brain tissue as a third pathological hallmark of the disease (Foley, 2010). In the meantime the content of several lipid classes and fatty acids has been found to be altered in the brain of AD-patients. A physiological function of Aβ and AICD in the regulation of several lipid metabolism pathways has been reported, possibly explaining the altered cerebral content of some lipid species in AD-affected brain tissue. Inversely, APP-processing is strongly influenced by the surrounding lipid environment indicating a bidirectional link between APP-proteolysis and lipid metabolism (Grimm et al., 2012b; Mett et al., 2014).

The link between lipid homeostasis and AD-pathology is strengthened by the identification of the ApoEε4-allele as the most important genetic risk factor for LOAD. ApoE is a lipoprotein involved in the transport of cholesterol and other lipids in the central nervous system (CNS). In humans there are three different ApoEε alleles encoding for the isoforms ApoEε2, ApoEε3 and ApoEε4 (Weisgraber et al., 1981; Mahley et al., 1996; Holtzman et al., 2012). The ApoEε4-allele is associated with an increased AD-risk, earlier disease onset and enhanced cerebral plaque load (Corder et al., 1993; Kuusisto et al., 1994; Breitner et al., 1999; Tiraboschi et al., 2004). In contrast, ApoEε2-carriers have a reduced risk of developing AD (Corder et al., 1994). These associations might be explained by an isoform-dependent binding of ApoEε (ε2 > ε3 > ε4) to Aβ-peptides influencing the clearance and aggregation of the peptide (Ma et al., 1994; Deane et al., 2008; Castellano et al., 2011; Holtzman et al., 2012).

A strong impact of the surrounding lipid bilayer on APP-processing is given by the fact that APP as well as all secretases are transmembrane proteins and that γ-secretase dependent APP cleavage even takes place in the hydrophobic membrane environment. For example, the exact position of γ-secretase cleavage and hence the length of the generated Aβ-peptides depends on membrane thickness (Grziwa et al., 2003; Winkler et al., 2012). In addition, the membrane fluidity influences APP-processing. Increased membrane fluidity seems to stimulate the non-amyloidogenic APP-processing by reducing APP internalization (Kojro et al., 2001). In this context it is important to note that APP-processing is also influenced by the subcompartmentalization of the membrane. Lipid raft microdomains are compact, dynamic assemblies of membrane proteins enriched in cholesterol, gangliosides and other sphingolipids. They are detergent-resistent and strongly differ in their lipid composition from the surrounding non-raft domains. Implications of lipid rafts in the intracellular protein trafficking, protein-lipid and protein-protein interactions as well as transmembrane signaling have been reported (Brown and Rose, 1992; Lingwood and Simons, 2010). The generation of Aβ has been shown to mainly take place in lipid rafts due to the co-localization of APP with BACE1 and the γ-secretase complex within these membrane microdomains (Lee et al., 1998; Riddell et al., 2001; Ehehalt et al., 2003; Vetrivel et al., 2004). In contrast, the non-amyloidogenic APP-proteolysis seems to occur predominantly in non-raft regions (Ehehalt et al., 2003; Harris et al., 2009). All these details indicate that a modulation of the membrane lipid composition might provide the opportunity of influencing Aβ-generation.

In the following sections of this article, the impact of several lipids and fatty acids on Aβ-associated AD-pathology is reviewed as well as the regulation of the corresponding metabolic pathways by APP-processing.

The brain is the most cholesterol-rich organ in the body (23 mg/g), it contains 23% of the total body sterol while only accounting for 2.1% of the total body weight (Dietschy and Turley, 2004). Within brain tissue cholesterol is mainly present in myelin sheaths and in the membranes of glial cells and neurons in its unesterified form. Due to the limited transport of cholesterol across the blood-brain barrier the cerebral cholesterol level is mainly dependent on de novo synthesis by oligodendrocytes, astrocytes and to a lesser extent by neurons. The conversion of 3-hydroxy-3-methylglutaryl-CoA to mevalonate catalyzed by the hydroxymethylglutaryl-CoA reductase (HMGCR), which is inhibited by statins, is the rate-controlling step in cholesterol biosynthesis (Martins et al., 2009; Di Paolo and Kim, 2011). The first evidence for a link between AD-pathogenesis and cholesterol metabolism was provided in 1994 by the observation that dietary cholesterol increases Aβ-production in rabbits (Sparks et al., 1994). Today there are many lines of evidence arguing for a connection between the pathology of AD and cholesterol homeostasis which are summarized below.

In several epidemiological studies elevated serum/plasma cholesterol contents have been identified as a risk factor for developing AD. Especially high serum cholesterol level in midlife are associated with a higher AD-risk (Pappolla et al., 2003; Solomon et al., 2009; Matsuzaki et al., 2011; Meng et al., 2014). Additionally, enhanced level of low-density lipoprotein (LDL) cholesterol and reduced level of high-density lipoprotein (HDL) cholesterol in serum correlate with the cerebral amyloid deposition in living human beings (Reed et al., 2014). In line, in human post mortem AD-brains cholesterol was found to be elevated and highly enriched in amyloid plaques (Cutler et al., 2004; Xiong et al., 2008; Panchal et al., 2010).

Most cell culture studies revealed that increasing cellular cholesterol level lead to an enhanced Aβ production whereas a depletion or reduction of cholesterol by e.g., cyclodextrin or statins shows the opposite effect (Simons et al., 1998; Fassbender et al., 2001; Maulik et al., 2013). The Aβ increasing property of cholesterol is based on a direct activation of β- and γ-secretase proteolytic activity (Kalvodova et al., 2005; Grimm et al., 2008; Osenkowski et al., 2008). Cholesterol is enriched in lipid raft membrane microdomains, in which amyloidogenic APP-processing mainly takes place. Thus modulating cellular cholesterol content inevitably affects membrane structure, membrane fluidity as well as cellular processes associated with lipid raft microdomains. Cholesterol depletion leads to the disruption of lipid rafts and therefore to a reduced association of APP, BACE1 and the components of the γ-secretase complex to lipid raft membrane microdomains, resulting in decreased amyloidogenic APP processing. Vice versa, an increase of cellular cholesterol leads to a higher lipid raft content of the membranes and hence to Aβ-overproduction (Simons et al., 1998; Hao et al., 2001; Hicks et al., 2012). High membrane cholesterol levels additionally promote APP endocytosis leading to enhanced Aβ-production in acidic intracellular compartments (Cossec et al., 2010). Conversely, APP is primarily localized at the cell surface in cholesterol-depleted cells leading to increased α-secretase-dependent non-amyloidogenic APP processing (Kojro et al., 2001). Beside the cholesterol-mediated effects on APP-proteolytic processing, cholesterol has been shown to promote Aβ-aggregation and -toxicity (Schneider et al., 2006; Ferrera et al., 2008; Abramov et al., 2011).

A strong correlation between hypercholesterolemia and enhanced Aβ level has also been observed in several animal models (Sparks et al., 1994; Refolo et al., 2000; Maulik et al., 2013). Inversely, a reduction of accumulated Aβ-peptides along with improved behavioral memory was achieved in animal models after administration of cholesterol-lowering drugs including statins (Fassbender et al., 2001; Refolo et al., 2001; Kurata et al., 2012). It should be noted, that there are also a few studies in which statins had no or oppositional effects on the cerebral Aβ-content in vivo (Park et al., 2003; Cibickova et al., 2009).

The impact of statins on AD has also been analyzed in observational studies and randomized controlled trials leading to inhomogeneous results. Statin intake is associated with a reduced incidence of AD or dementia in general in most, but not all of these studies (Wolozin et al., 2000, 2007; Rea et al., 2005; Arvanitakis et al., 2008; Haag et al., 2009). Especially the reduction of serum cholesterol level by the intake of statins in midlife might have a preventive effect towards the development of AD (Kivipelto et al., 2002; Pappolla et al., 2003; Shinohara et al., 2014). In strong contrast, most clinical trials failed to observe any benefit of statins in individuals already suffering from AD (Feldman et al., 2010; Sano et al., 2011; McGuinness et al., 2014), indicating cholesterol-lowering drugs to have rather a protective than a therapeutic potential in respect to AD.

Beside the described influence of cholesterol on APP-proteolysis, there is also an impact of APP-processing on cholesterol homeostasis. APP/APLP2- and PS1/PS2-deficient fibroblasts have a significantly increased cellular cholesterol content, which can be reversed by the supplementation of Aβ40-peptides. In line with this, enhanced cerebral cholesterol concentrations were found in APP- and PS-deficient mice (Grimm et al., 2005; Umeda et al., 2010). Analysis of the underlying mechanisms revealed that Aβ40 reduces cholesterol de novo synthesis by inhibiting HMGCR activity (Grimm et al., 2005).

Docosahexaenoic acid (DHA, 22:6) is a polyunsaturated fatty acid (PUFA) naturally occurring in high amounts in marine food, especially in fish oil (Mann et al., 2010). It accounts for 30%–40% of all esterified fatty acids in neuronal plasma membrane phospholipids and for 8% of the brain dry weight, thus belonging together with α-linolenic acid (ALA, 18:3) and eicosapentaenoic acid (EPA, 20:5) to the most important ω3-fatty acids in the CNS (Lauritzen et al., 2001; Muskiet et al., 2006). As endogenous DHA-biosynthesis is highly limited in humans, the main part of this fatty acid is provided by dietary intake (Pawlosky et al., 2001). DHA is efficiently transported across the blood brain barrier (Ouellet et al., 2009; Nguyen et al., 2014) and rapidly incorporates into phospholipids of cellular membranes leading to increased membrane fluidity (Horrocks and Farooqui, 2004; Yang et al., 2011).

The DHA content is reported to be reduced in the serum/plasma of AD-patients as well as in certain regions of post mortem AD-brains (Söderberg et al., 1991; Conquer et al., 2000; Tully et al., 2003). Because of its six double-bonds DHA is very susceptible to lipid-peroxidation resulting in oxidative stress known to be involved in AD pathogenesis (Smith et al., 1994; Yatin et al., 1998; Fam et al., 2002; Cai et al., 2011). Indeed, the levels of PUFA oxidation products are elevated in AD-affected brains, indicating the reduced DHA content in these tissues to be caused by increased oxidative damage (Sayre et al., 1997; Markesbery and Lovell, 1998; Montine and Morrow, 2005; Grimm et al., 2016a).

Several epidemiological trials found the dietary intake of DHA or higher DHA serum/plasma levels to be associated with a reduced risk of developing AD indicating a potential of DHA in AD-prevention (Kalmijn et al., 1997; Barberger-Gateau et al., 2002; Morris et al., 2003b). However, other studies failed to find an association between PUFAs and AD-risk (Engelhart et al., 2002; Kröger et al., 2009; Jicha and Markesbery, 2010; Mett et al., 2014).

We and others analyzed the impact of DHA on APP-processing revealing the fatty acid to reduce Aβ-levels via pleiotropic mechanisms. DHA reduces β- and γ-secretase activity and stimulates α-secretase-dependent APP-cleavage. In addition to direct effects, the activities of γ- and β-secretase are reduced by DHA due to a PS1-displacement out of lipid rafts and a reduced BACE1 internalization. The stimulated α-secretase activity in presence of DHA is based on the enhanced gene expression and protein stability of ADAM17. Altogether these effects lead to a shift from amyloidogenic to non-amyloidogenic APP-processing and thus to reduced total Aβ-level. DHA additionally has cholesterol-lowering effects further inhibiting Aβ-production. It reduces cholesterol de novo synthesis via inhibition of HMGCR and disturbs lipid raft integrity by shifting cholesterol out of these membrane microdomains (Hashimoto et al., 2005a; Stillwell et al., 2005; Grimm et al., 2011c). Beside the described effects on APP-processing an impact of DHA on Aβ-degradation and -aggregation has also been reported. We recently observed a highly enhanced insulin-degrading enzyme (IDE)-dependent Aβ-degradation in neuroblastoma cells after the supplementation of DHA- and EPA-containing phosphatidylcholine (PC; Grimm et al., 2016b). Others reported an increased microglial phagocytosis of Aβ as well as a reduction of Aβ-fibrillation and Aβ-induced toxicity in the presence of DHA (Hossain et al., 2009; Hjorth et al., 2013).

A protective effect of dietary DHA with regard to cerebral Aβ-level and amyloid plaque load could be further confirmed in vivo in several animal models (Lim et al., 2005; Green et al., 2007; Perez et al., 2010). In line with this, higher cognitive performances were observed in AD-animal models after DHA supplementation (Hashimoto et al., 2002, 2005b; Calon et al., 2004). However, others failed to find any beneficial effect of DHA in AD transgenic mice (Arendash et al., 2007).

A possible therapeutic use of DHA regarding AD has been investigated in several clinical trials showing inconsistent results. Some studies revealed a beneficial effect of daily DHA treatment in patients with very mild cognitive dysfunctions (Freund-Levi et al., 2006; Kotani et al., 2006; Chiu et al., 2008). Others did not observe any influence of DHA on AD-biomarkers and cognitive decline in AD patients (Freund-Levi et al., 2009; Quinn et al., 2010). It should be mentioned, that oxidized DHA species and the lipid-peroxidation products of PUFAs are able to increase amyloidogenic APP-processing and hence Aβ-generation. In a recent study we demonstrated, that only 1% oxidized DHA reverts the positives effects of DHA on Aβ-production indicating that PUFAs have to be prevented from oxidation in nutritional approaches (Grimm et al., 2016a). In such approaches DHA often is combined with E-vitamins due to their high antioxidative properties acting as scavengers of radicals and peroxides (Kamal-Eldin and Appelqvist, 1996). However, we demonstrated in two recent studies that several tocopherol and tocotrienol species have beside their protective antioxidative properties the undesirable effect of increasing amyloidogenic APP processing and reducing the enzymatic degradation of Aβ-peptides (Grimm et al., 2015b, 2016c).

Trans fatty acids (TFAs) are unsaturated fatty acids, which are characterized by having at least one double-bond in trans-configuration. This means that the two hydrogen atoms are, in contrast to cis-configuration, localized on opposite sides of the double-bond. Because of their straighter shape compared to the cis-counterparts, TFAs have higher melting points and lead to a decreased fluidity of biological membranes (Roach et al., 2004; Ibrahim et al., 2005). TFAs in our diet arise from industrial procedures and to a lesser extent from biological processes in the digestive tract of ruminant animals. The key source of TFAs is commercially prepared food due to hydrogenation or thermal treatment of oils (Bhardwaj et al., 2011). Accumulation of these fatty acids in the body as well as incorporation in brain tissue has been reported indicating an impact of TFAs on cerebral biochemistry (Laryea et al., 1990; Teixeira et al., 2012).

Studies analyzing the relationship between TFAs and AD-risk or the progression of cognitive decline came to inconsistent results. A positive correlation between dietary TFA intake and AD-risk was found in one study while others reported the AD-risk not to be influenced by TFAs (Engelhart et al., 2002; Morris et al., 2003a). Similarly, some authors observed the TFA intake to result in a higher rate of cognitive decline in women with type 2 diabetes, in persons with high copper consumption and in older persons in general while others failed to find a relationship between TFA intake and cognitive decline in women (Morris et al., 2004, 2006; Devore et al., 2009; Naqvi et al., 2011; Okereke et al., 2012).

We investigated the effects of TFAs on APP-processing and Aβ-generation in neuroblastoma cells compared to their cis-counterparts. In presence of TFAs, we found a shift from non-amyloidogenic to amyloidogenic APP-processing accompanied by a significant increase in Aβ-production. TFA supplementation increases the activity of β- and γ-secretase due to direct effects and an enhanced gene expression of BACE1 and the γ-secretase complex components (Grimm et al., 2012a). The direct effect on γ-secretase activity was confirmed by others demonstrating the activity of purified γ-secretase to be stimulated by an increased trans/cis-ratio of supplemented fatty acids (Holmes et al., 2012). In contrast, non-amyloidogenic APP-processing is reduced in TFA-treated cells because of enhanced APP-internalization and a reduction in ADAM10 gene expression. Additionally, we found TFAs to stimulate Aβ-aggregation in vitro (Grimm et al., 2012a).

The impact of TFAs on cerebral Aβ-levels and cognition has also been investigated in vivo with less clear results. In a study by Phivilay et al. (2009), Aβ- and tau-pathology was unaltered in the brain tissue of an AD-mouse model after dietary supplementation of TFAs. Another study reported a declined spatial learning performance of mice fed with a TFA- and monosodium glutamate-rich diet (Collison et al., 2010).

As TFAs are reported to be linked to cholesterol and DHA homeostasis they might also affect APP-processing and Aβ-generation via indirect mechanisms. The dietary intake of TFA leads to an inauspicious enhanced ratio of LDL/HDL plasma cholesterol (Mensink and Katan, 1990; Judd et al., 1994), which might be associated with a higher AD-risk as described above. Furthermore, high TFA consumption was shown to modify the fatty acid profile of murine brain tissue with a reduction in DHA content. Nevertheless, in this study the cerebral Aβ-levels were unaltered as already mentioned (Phivilay et al., 2009).

Plasmalogens (PL) are commonly occurring phospholipids accounting for 22% of the total phospholipid mass in human brain tissue. They are characterized by an enol ether double-bond at the sn1-position, which links an alkenyl chain to the glycerol backbone. At the sn2-position they are enriched in PUFAs including DHA and arachidonic acid (AA, 20:4). Phosphatidylethanolamine (PE) and PC are the most common polar head groups of PLs, which have a high susceptibility to oxidative stress due to their enol ether double-bond (Broniec et al., 2011; Braverman and Moser, 2012). PL-biosynthesis takes place in peroxisomes and the endoplasmic reticulum. The initial committed step reaction of PL de novo synthesis is catalyzed by the peroxisomal enzyme alkyl-dihydroxyacetonephosphate-synthase (AGPS; De Vet et al., 1999). PL level in the human body are mainly modulated by PL metabolism, but to a lesser extent also by the dietary consumption of PL-rich meat and fish (Blank et al., 1992).

While one study by Pettegrew et al. (2001) did not detect an AD-dependent alteration in the cerebral PL-content, we and others found a reduction of PE-PLs and PC-PLs in human post mortem AD-brains (Ginsberg et al., 1995; Han et al., 2001; Grimm et al., 2011a; Igarashi et al., 2011; Rothhaar et al., 2012). In line with this, a reduced PE-PL content was also observed in the serum and in erythrocyte membranes of AD-patients (Goodenowe et al., 2007; Oma et al., 2012).

The reduction of PL content in AD-affected brain tissue might be explained by enhanced PL degradation due to increased oxidative stress and a stimulated activity of phospholipases in presence of Aβ-peptides (Sanchez-Mejia et al., 2008). Additionally, we demonstrated PL biosynthesis to be regulated by APP-processing. Under physiological conditions AGPS gene expression and hence PL biosynthesis is upregulated by AICD. In contrast, under pathological conditions the Aβ-induced reactive oxidative species impair AGPS protein stability leading to a decreased PL de novo synthesis (Grimm et al., 2011d).

Because of the altered PL content in AD-brain tissue, we analyzed the impact of PLs on APP-processing. Our results demonstrate that PLs reduce γ-secretase activity in living cells as well as in purified membranes derived from neuroblastoma cells and murine brain tissue. Compared to the corresponding phospholipids lacking the enol ether, all tested PC-PL- and PE-PL-species independent of the bound fatty acid directly inhibited γ-secretase activity. In contrast, the activities of α- and β-secretase remained unchanged after PL-supplementation (Rothhaar et al., 2012). The direct inhibitory effect of PE-PLs on the γ-secretase complex has been recently confirmed by others (Onodera et al., 2015). Interestingly, in our study the addition of PLs to cellular membranes derived from human AD-brains also resulted in a decreased γ-secretase activity. This indicates the rebuilding of a normal PL level to have a positive impact in the pathologic situation of AD (Rothhaar et al., 2012). However, such ex vivo experiments have their clear limitations and further studies are necessary to analyze the in vivo relevance of PLs on APP-processing. In addition to Aβ-production, there is also an impact of PLs on the aggregation of Aβ-peptides. PE-PL has been reported to eliminate the neurotoxicity-associated Aβ-oligomerization phase while allowing fibril formation (Lee et al., 2011).

Sphingolipids are an inhomogeneous group of lipids characterized by a backbone consisting of the amino alcohol sphingosine. Sphingolipid biosynthesis is initiated by the serine palmitoyl-CoA transferase (SPT) catalyzing the condensation of palmitoyl-CoA and L-serine to 3-ketosphinganin, which is further metabolized to ceramide. Ceramide is the most important branching point within the sphingolipid metabolism pathways serving as precursor for the generation of sphingosine, sphingomyelin (SM) and more complex glycosphingolipids.

All sphingolipids are anchored in the membrane bilayer via their ceramide moiety, besides cholesterol they represent major components of lipid raft membrane microdomains (Posse de Chaves and Sipione, 2010). The first evidence for a role of sphingolipids in neurodegeneration came from the observation of lysosomal storage diseases, inherited disorders characterized by the lysosomal accumulation of different sphingolipids. These diseases are associated with early dementia and the development of AD-related Aβ- and tau-pathology (Tarasiuk et al., 2012). The link between sphingolipid metabolism and AD-pathogenesis is further strengthened by alterations of several sphingolipids in post mortem AD-brain tissue and their potential to modulate APP-processing and Aβ-aggregation summarized below. Additionally, SPT gene expression and hence total sphingolipid biosynthesis is downregulated by the APP-processing product AICD (Grimm et al., 2011b).

Regarding therapeutic interventions for AD an early diagnosis of the disease and hence the identification of biomarkers, which can be used for the in vivo diagnosis prior to the first symptoms, is important. So, the identification of early AD-biomarkers with a high specificity and reliability is a central topic in AD research (Fiandaca et al., 2014). The lipid alterations connected to AD, which are partially detectable at the very early disease stages as described above, might have the potential to be used as biomarkers for early AD diagnosis by lipidomic approaches. For example, Mapstone et al. (2014) discovered a set of eight PC species and two acylcarnitines in the peripheral blood that predicts the development of mild cognitive impairment or AD within 2–3 years with an accuracy of more than 90%. However, further studies are needed to identify combinations of lipidomics-based biomarkers which can be used for the detection of preclinical AD with the required sensitivity and specificity.

In conclusion all these findings demonstrate a close link of APP, APP processing and AD to lipid homeostasis. It could be demonstrated that APP processing and especially AICD has a physiological function in in the regulation of several lipid metabolic pathways. Inversely, APP-processing is strongly dependent on the lipid microenvironment indicating a bidirectional link between APP-proteolysis and lipid metabolism. This results in tightly connected complex regulatory cycles (Figure 2). Under pathological situations such as AD, this balanced regulation might be disrupted leading to pathological alterations in lipid homeostasis and Aβ peptide overproduction, resulting in increased neurodegeneration.

MOWG, JM, HSG and TH wrote the manuscript.

According to the author guidelines, funding for the research leading to these results were received from: the EU FP7 project LipiDiDiet, Grant Agreement No. 211696. Moreover funding for MOWG and TH was provided by Fundació la Maratò de TV3 and by JPND MindAD 1ED1508.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.