**3. Brain lipid alterations in Alzheimer's disease**

#### **3.1. Cholesterol and sphingolipid homeostasis in AD**

parenchyma of aged wild-type mice, originating from axonal varicosities, further supporting this hypothesis. These data suggest that axonal dystrophy occurs first, leading thereafter to extracellular amyloid deposition in the early stages of the disease. In fact, it has been proposed that neurite dystrophy could reflect a conserved neuroprotective strategy to overcome the agerelated accumulation of misfolded proteins, which in turn may represent a molecular mechanism of Aβ plaque deposition that potentially underlies the shift from normal to pathological aging [24, 25]. Nevertheless, Aβ alone may promote axonal atrophy through its interactions with the p75 neurotrophin receptor (p75NTR) in axon membranes [26]. Together, the evidence suggests that dystrophy and extracellular Aβ deposition are involved in a positive feedback loop whereby axon dystrophy is a source of extracellular Aβ, and the latter

In terms of neuroinflammation, it is widely accepted that Aβ deposition alone might be sufficient to induce an inflammatory reaction that subsequently contributes to neuronal death and cognitive decline in AD [15]. However, this fact does not necessarily imply that Aβ plaque formation precedes microglial activation in AD. During normal aging, microglial activation aims to clear the misfolded proteins contained in fragmented neurites and aggregated into senile plaques. Interestingly, during AD-related pathological ageing, microglia cells recruited around plaques phagocytose Aβ and this could constitute part of the microglial mechanism to clear misfolded proteins, also during normal ageing [25]. Thus, in a scenario characterized by age-related chronic inflammation, microglia would be highly responsive to further activation which would drive their differentiation toward a classic phenotype characterized by pro-inflammatory cytokine secretion, in turn impairing axon trafficking, promoting Aβ accumulation and cell death [25, 27]. However, this putative role for AD-associated neuroin‐ flammation is not supported by evidence showing that the inflammatory response is not neurotoxic and, indeed, it is even neuroprotective in a transgenic mouse model of AD [28]. In fact, from early in the amyloid pathology, alternative neuroprotective microglia are activated around amyloid plaques supporting neuronal survival, and this alternative phenotype is also present during animal ageing. By contrast, the classic microglial phenotype that is character‐ ized by cytotoxic cytokine secretion only appears at advanced ages, associated with the presence of soluble Aβ oligomers and neuronal loss [27, 28]. Thus, these evidences show that alternative neuroprotective microglia may be present at advanced ages and coexist with classic microglial activation. In summary, although it is widely accepted that neuroinflammation promotes neuronal degeneration, it remains unclear how brain inflammation participates in

Hence, determining whether amyloid pathology is the first event in the pathway to ADassociated neuronal degeneration and dementia appears to be a particularly relevant issue, especially after the repeated fiascos in clinical trials of drugs targeting Aβ and related molec‐ ular entities. There is a close relationship among Aβ, inflammatory and neurite pathologies in AD because they all appear at early stages of the disease and all three are involved in neuronal death. In the present chapter, we will review how these neuropathological hallmarks are related to AD-associated membrane lipid alterations, as there can now be no shadow of doubt that brain lipids and the pathways they are involved in influence the pathophysiology of AD.

promotes axonal atrophy.

132 Update on Dementia

the shift from normal to pathological ageing.

The amyloid cascade hypothesis was postulated because FAD mutations cause Alzheimer's disease, and they induce abnormal APP processing that leads to the well-characterized amyloid pathology [9]. Since the pathological hallmarks are exactly the same for both FAD and SAD, the same cascade of neuropathological events is thought to occur in both these disease variants. However, in addition to the influence of FAD clinical mutations on APP metabolism, these mutations may also have additional effects on other signaling cascades. In fact, presenilins (PSs) are the catalytic center of the γ-secretase complex, which cleaves more than 60 type I membrane proteins (one type of single transmembrane spanning region in integral proteins) [29, 30]. More than 160 clinical mutations have been described for PS1 and most of those that were studied induce loss of function of γ-secretase activity [31, 32]. These mutations may exert additional effects on cellular signaling as a consequence of the altered processing of certain membrane proteins that could influence lipid cellular homeostasis. Interestingly, γ-secretase loss of function induced by the ablation of PSs or by transgenic expression of PS1 mutants provoked a severe imbalance in the cholesterol content of the plasma membrane and intracellular membranes [33, 34]. In this sense, PS ablation increased the overall levels of cholesterol and sphingomyelin (SM) in cells, whereas the local concentration of cholesterol at the plasma membrane was dramatically reduced, resulting in the intracellular accumulation of cholesterol and cholesterol-rich membrane domains, such as lipid rafts [33, 34]. These observations demonstrate the impact of γ-secretase loss of function on the cell membrane lipid composition.

In the human brain, cholesterol is mainly transported in lipoprotein particles that predomi‐ nantly contain ApoE. Interestingly, ApoE has been identified as a risk factor for SAD suggest‐ ing that altered cholesterol transport might also be related to the pathogenesis of late-onset AD [35]. The human ApoE protein is comprised of 299 amino acids and it has three isoforms, namely ApoE2, ApoE3, and ApoE4. The differences between these three isoforms lie in the amino acid residues at positions 112 and 158: ApoE2 (Cys112, Cys158), ApoE3 (Cys112, Arg158), and ApoE4 (Arg112, Arg158). In particular, subjects carrying the ApoE4 allele have a 3- to 4-fold higher risk of developing AD than those who do not carry this allele. Furthermore, ApoE4 was observed to exhibit a gene dose–effect, such that individuals who carry two copies of this allele have an even higher risk of suffering AD and an earlier age of onset. The effects of the ApoE4 isoform on AD risk are maximal between the ages of 60 and 70 years old, ApoE4 allele being present in more than 50% of all AD cases. Conversely, ApoE2 carriers appear to be somewhat protected from AD compared with ApoE3 carriers [36]. In this context, the ApoE4 isoform is less efficient in promoting cholesterol flux in neurons and astrocytes, and it also compromises cell uptake of cholesterol-containing lipoproteins compared with the other ApoE isoforms [37]. Furthermore, individuals carrying the ApoE4 allele accumulate less ApoE lipoprotein in the brain than non-ApoE4 carriers [38]. Hence, the expression of ApoE4 in SAD cases appears to alter cholesterol homeostasis in neurons in a similar way as that induced by γ-secretase loss-of-function in PS1-deficient cells and transgenic models of AD harboring clinical PS1 mutations [33, 34]. In such AD models, the loss of γ-secretase activity leads to impaired uptake of lipoproteins from the extracellular media due to the poor internalization of ApoE receptors like the LDLR (low-density lipoprotein receptor) [34]. In AD patients with the ApoE4 allele, cholesterol uptake would be impaired due to the lower affinity of ApoE4 to bind neuronal lipoprotein receptors, and to the lower concentration of circulating ApoE than in individuals carrying the ApoE2 or ApoE3 alleles [38, 39]. In any case, poorer membrane incorporation of neuronal cholesterol leads to increased *de novo* cholesterol synthesis and an altered neuronal distribution. Thus, altered cholesterol homeostasis is a key aspect of AD pathogenesis and alterations to cholesterol may represent a meeting point in the pathogenesis of FAD and SAD, driving the same neuropathological events in both disease variants, such as increased amyloidogenic APP processing.

The central nervous system (CNS) contains around 25% of the cholesterol in the body and evidence is accumulating that cholesterol homeostasis is indeed associated with AD patho‐ genesis. High cholesterol and high-density lipoprotein (HDL) in blood plasma are correlated with Aβ load in the brains of patients with AD [40, 41] and that increased cholesterol levels are associated with the incidence of AD [42, 43]. Furthermore, high or low cholesterol levels have often been related to enhanced or diminished Aβ production, respectively, in cell and animal models of AD, although these results are a little controversial [42, 44, 45]. What is more, lipidomic studies have shown that levels of cholesterol, certain cholesterol esters, and certain SM species are upregulated in the brain of patients with AD. This correlation is particularly strong in the case of patients with AD harboring the ApoE4 allele, although some contradictory results have also been reported in this respect [46–49]. Finally, altered cholesterol distribution and transport have been causally linked to neurodegenerative diseases in addition to AD, such as Huntington's and Niemann–Pick Type C diseases [44].

Cholesterol is an essential structural component of cell membranes and one of the major components of the functional membrane microdomains known as lipid rafts, together with sphingolipids such as SM and gangliosides. These microdomains are highly ordered mem‐ brane structures that serve as platforms for cell signaling, ligand-receptor binding, protein sorting, and other activities in the cell. Interestingly, amyloidogenic APP processing and Aβ aggregation have been proposed to take place in lipid rafts [50]. In fact, the activities of both BACE-1 and γ-secretase are enhanced in this type of membrane microdomains [51, 52]. In this context, compelling evidence supports the involvement of cholesterol and sphingolipids in the amyloidogenic processing of APP. On the one hand, membrane enrichment of these lipids could alter the biophysical properties of the lipid bilayer, affecting secretase activity in a manner that leads to the production of the longer pathogenic Aβ peptides instead of the shorter p3 peptide [53] (see **Figure 1**). On the other hand, cholesterol and SM storage disorders impair intracellular trafficking of APP, resulting in the accumulation of APP, APP-CTFs, and Aβ in autophagic vesicles of the endolysosomal pathway [54, 55]. Accordingly, impaired distribution of cholesterol and SM is accompanied by the downregulation of proteins involved in endoso‐ mal redistribution and fusion to the plasma membrane (SNAREs and RABs) in PS1-deficient cells [33]. These evidences suggest that dysfunctional vesicular trafficking between the plasma membrane and intracellular compartments may be caused by membrane lipid alterations that lead to the neuritic pathology and altered APP processing in FAD transgenic models [33, 56]. Additional studies have also linked shingolipid lysosomal accumulation to autophagic dysfunction and dystrophic neurite formation in AD [55, 57]. Such results indicate that cellular accumulation of sphingolipids could induce key cytopathological changes characteristic of AD, such as alterations to the autophagic/lysosomal system, increased generation of Aβ and accumulation of APP-CTFs in autophagic vesicles at dystrophic neurites, as occurs in an agedependent manner in transgenic mouse models of AD [58]. Interestingly, a cholesterolenriched diet in healthy mice also leads to insulin-like growth factor 1 (IGF1) impairment and insulin-mediated pro-survival signaling, which in turn promotes tau hyperphosphorylation in neurons [59]. Together, this evidence suggests that altered cholesterol/sphingolipid homeostasis may promote the neurite pathology, tau hyperphosphorylation, and amyloido‐ genic APP processing in AD.

Nevertheless, it cannot be ruled out that AD-related membrane lipid alterations can also potentiate the neurotoxicity of the Aβ oligomers in AD patient's brains. In fact, lipid rafts may serve as a platform for the cellular interactions with soluble Aβ oligomers, in turn promoting tau hyperphosphorylation and inhibiting synaptic plasticity by hindering LTP (long-term potentiation) in the brain [60, 61]. Moreover, raft-associated lipids such as cholesterol, SM, and the GM1 ganglioside revert the fibrillar Aβ into soluble oligomers, such that altered cellular lipid homeostasis may actually potentiate the severity of the amyloid pathology in AD [62].

#### **3.2. Polyunsaturated fatty acids in AD**

clinical PS1 mutations [33, 34]. In such AD models, the loss of γ-secretase activity leads to impaired uptake of lipoproteins from the extracellular media due to the poor internalization of ApoE receptors like the LDLR (low-density lipoprotein receptor) [34]. In AD patients with the ApoE4 allele, cholesterol uptake would be impaired due to the lower affinity of ApoE4 to bind neuronal lipoprotein receptors, and to the lower concentration of circulating ApoE than in individuals carrying the ApoE2 or ApoE3 alleles [38, 39]. In any case, poorer membrane incorporation of neuronal cholesterol leads to increased *de novo* cholesterol synthesis and an altered neuronal distribution. Thus, altered cholesterol homeostasis is a key aspect of AD pathogenesis and alterations to cholesterol may represent a meeting point in the pathogenesis of FAD and SAD, driving the same neuropathological events in both disease variants, such as

The central nervous system (CNS) contains around 25% of the cholesterol in the body and evidence is accumulating that cholesterol homeostasis is indeed associated with AD patho‐ genesis. High cholesterol and high-density lipoprotein (HDL) in blood plasma are correlated with Aβ load in the brains of patients with AD [40, 41] and that increased cholesterol levels are associated with the incidence of AD [42, 43]. Furthermore, high or low cholesterol levels have often been related to enhanced or diminished Aβ production, respectively, in cell and animal models of AD, although these results are a little controversial [42, 44, 45]. What is more, lipidomic studies have shown that levels of cholesterol, certain cholesterol esters, and certain SM species are upregulated in the brain of patients with AD. This correlation is particularly strong in the case of patients with AD harboring the ApoE4 allele, although some contradictory results have also been reported in this respect [46–49]. Finally, altered cholesterol distribution and transport have been causally linked to neurodegenerative diseases in addition to AD, such

Cholesterol is an essential structural component of cell membranes and one of the major components of the functional membrane microdomains known as lipid rafts, together with sphingolipids such as SM and gangliosides. These microdomains are highly ordered mem‐ brane structures that serve as platforms for cell signaling, ligand-receptor binding, protein sorting, and other activities in the cell. Interestingly, amyloidogenic APP processing and Aβ aggregation have been proposed to take place in lipid rafts [50]. In fact, the activities of both BACE-1 and γ-secretase are enhanced in this type of membrane microdomains [51, 52]. In this context, compelling evidence supports the involvement of cholesterol and sphingolipids in the amyloidogenic processing of APP. On the one hand, membrane enrichment of these lipids could alter the biophysical properties of the lipid bilayer, affecting secretase activity in a manner that leads to the production of the longer pathogenic Aβ peptides instead of the shorter p3 peptide [53] (see **Figure 1**). On the other hand, cholesterol and SM storage disorders impair intracellular trafficking of APP, resulting in the accumulation of APP, APP-CTFs, and Aβ in autophagic vesicles of the endolysosomal pathway [54, 55]. Accordingly, impaired distribution of cholesterol and SM is accompanied by the downregulation of proteins involved in endoso‐ mal redistribution and fusion to the plasma membrane (SNAREs and RABs) in PS1-deficient cells [33]. These evidences suggest that dysfunctional vesicular trafficking between the plasma membrane and intracellular compartments may be caused by membrane lipid alterations that

increased amyloidogenic APP processing.

134 Update on Dementia

as Huntington's and Niemann–Pick Type C diseases [44].

Polyunsaturated fatty acids (PUFAs) are those fatty acids that contain more than one double bond in their backbone. They are abundant in cell membranes, and they are mainly incorpo‐ rated into membrane phospholipids. The carbon next to the carboxyl group is known as the α carbon, the next one is the β carbon, and so forth, until the final carbon called the ω carbon. Thus, ω-3 fatty acids have the first double bond between the third and fourth C atoms from the ω carbon. For instance, 22:6 ω-3 or 22:6 n-3 (docosahexaenoic acid, DHA) indicates a 22 carbon chain with six double bonds and with the first double bond between the third and fourth carbons from the CH3 end. The physiological properties of unsaturated fatty acids largely depend on the position of the first unsaturation relative to the end position. The essential fatty acids α-linolenic acid (ALA, 18:3 ω-3) and linoleic acid (LA, 18:2 ω-6) must be incorporated through the diet, and they are the starting point for the synthesis of longer and more unsaturated PUFAs such as arachidonic acid (ARA, 20:4 ω-6), eicosapentaenoic acid (EPA, 20:5 ω-3), and DHA (22:6 ω-3). However, conversion of ALA to longer PUFAs in humans is very inefficient and therefore, these long PUFAs are normally incorporated through the diet, particularly through fish intake [63].

The membranes of the cells in the brain are rich in ω-3 PUFAs such as DHA and EPA. Since AD is a cognitive disorder and DHA is involved in normal cognitive development, the DHA levels in the AD brain have been analyzed extensively. As a result, it is widely accepted that in the human brain AD courses with diminished DHA levels, although a number of discrep‐ ancies in this respect have also been observed [64]. These discrepancies may reflect the brain region studied as the neurodegeneration associated with AD does not affect all brain areas homogeneously. In the hippocampus, one of the regions primarily affected in AD, decreased DHA levels are associated with reduced levels of PE (phosphatidylethanolamine) or PE plasmalogens [65–69], supporting a relationship between lower DHA levels and cognitive decline in AD. Moreover, there is significant experimental evidence in animal models that hippocampal DHA deficiency or enrichment is associated with reduced or increased learning memory abilities, respectively [70]. At the cellular level, exposure to ω-3 PUFAs enhances synaptic plasticity by increasing LTP and synaptic protein expression, in turn leading to increased dendritic spine density and hippocampal neurogenesis. In addition, ω-3 PUFAs have antioxidant, anti-inflammatory, and anti-apoptotic effects, thereby promoting neuronal survival during normal ageing and in AD. On the other hand, PUFA deficits are related to enhanced amyloidogenic APP processing and cell susceptibility to Aβ neurotoxicity, particu‐ larly as ω-3 PUFA deficiency downregulates neuroprotective signaling (e.g., ERK signaling). Therefore, PUFA deficits may enhance neuron degeneration and cognitive impairment in AD [71].

It still remains largely unclear how ω-3 PUFAs exert their cellular functions and consequently, what signaling cascades are impaired in the brain due to their deficiency. Such ω-3 PUFAs maintain the structural functionality of neural cell membranes. Indeed, in consonance with the reduced levels of DHA in the human AD brain, lipid rafts obtained from AD brain cortex also exhibited significantly less DHA than age-matched controls [72]. Interestingly, the biophysical and structural properties of PE and DHA in membranes are opposed to those of cholesterol and SM. Thus, these abnormalities in lipid raft composition may provoke strong modifications to the membrane structure of neurons such as alteration of membrane viscosity, rigidity and thickness, lateral lipid packing, lipid order, and other parameters, which could in turn be relevant to secretase activity and the production of Aβ [73]. Accordingly, decreased PUFA levels in lipid rafts would be coupled to enriched cholesterol and sphingolipids, thereby promoting the detrimental effects on neurons including the neurite dystrophy, tau hyper‐ phosphorylation, and amyloidogenic APP processing that drives neuronal degeneration (see Section 3.1.).

Alternatively, DHA may be released from phospholipids due to the activity of PLA2 (phos‐ pholipase A2), acting as a signaling molecule, and DHA can be hydroxylated to produce several secondary bioactive lipids such as resolvins (RVs) and protectins. DHA hydroxylation is mediated through lipoxygenase-15 (LOX-15) or acetylated cyclooxygenase-2 (COX-2) [63]. Compounds derived from DHA are classified as D-series RVs or protectins, while those formed from EPA are designated as E-series RVs. DHA can be hydroxylated on carbon 17 by 15-LOX or acetylated COX-2, leading to stereoselective formation of 17S- or 17R-hydroxy-DHA (17- HDHA), respectively. These derivatives may be further hydroxylated to give rise to trihydroxy derivatives such as the D1, D2, D3, and D4 17-(S/R)-RVs (D-series RVs), and the dihydroxy 17- (R)- and 17(S)-protectin, the latter also known as neuroprotectin D1 (NPD1). EPA can be stereoselectively hydroxylated to 18-(S/R)-hydroxy-EPA (18-HEPA) by cytochrome P450 or acetylated COX-2, which is further processed to form E1, E2 and E3 18-(S/R)-RVs (E-series RVs: **Figure 3**). Both, 17-HDHA and 18-HEPA serve as markers for RVs and protectins, and remarkably, their presence in blood is directly related to the intake of ω-3 PUFAs in animal models [74]. In addition, these PUFA derivatives are thought to exert their biological function by mechanisms that go beyond the simple regulation of lipid membrane composition and structure. In fact, non-esterified DHA, RVs and protectins may bind to different fatty acid (FA) receptors such as the retinoid X receptor (RXR), G protein-coupled receptors (GPCRs), peroxisome proliferator-activated receptors (PPARs), and fatty-acid binding proteins (FABPs). Although the exact signaling cascade mediated by many of these proteins has not been identified, the mechanism of action of DHA or HDHA derivatives like NPD1 has been proposed to involve PPARγ activation. Indeed, NPD1 is known to promote PPARγ activation more intensely than DHA and as such, the neuroprotective effects of DHA may be mediated by NPD1 and/or other DHA-derived hydroxylated bioactive derivatives in the brain [75, 76].

homogeneously. In the hippocampus, one of the regions primarily affected in AD, decreased DHA levels are associated with reduced levels of PE (phosphatidylethanolamine) or PE plasmalogens [65–69], supporting a relationship between lower DHA levels and cognitive decline in AD. Moreover, there is significant experimental evidence in animal models that hippocampal DHA deficiency or enrichment is associated with reduced or increased learning memory abilities, respectively [70]. At the cellular level, exposure to ω-3 PUFAs enhances synaptic plasticity by increasing LTP and synaptic protein expression, in turn leading to increased dendritic spine density and hippocampal neurogenesis. In addition, ω-3 PUFAs have antioxidant, anti-inflammatory, and anti-apoptotic effects, thereby promoting neuronal survival during normal ageing and in AD. On the other hand, PUFA deficits are related to enhanced amyloidogenic APP processing and cell susceptibility to Aβ neurotoxicity, particu‐ larly as ω-3 PUFA deficiency downregulates neuroprotective signaling (e.g., ERK signaling). Therefore, PUFA deficits may enhance neuron degeneration and cognitive impairment in AD

It still remains largely unclear how ω-3 PUFAs exert their cellular functions and consequently, what signaling cascades are impaired in the brain due to their deficiency. Such ω-3 PUFAs maintain the structural functionality of neural cell membranes. Indeed, in consonance with the reduced levels of DHA in the human AD brain, lipid rafts obtained from AD brain cortex also exhibited significantly less DHA than age-matched controls [72]. Interestingly, the biophysical and structural properties of PE and DHA in membranes are opposed to those of cholesterol and SM. Thus, these abnormalities in lipid raft composition may provoke strong modifications to the membrane structure of neurons such as alteration of membrane viscosity, rigidity and thickness, lateral lipid packing, lipid order, and other parameters, which could in turn be relevant to secretase activity and the production of Aβ [73]. Accordingly, decreased PUFA levels in lipid rafts would be coupled to enriched cholesterol and sphingolipids, thereby promoting the detrimental effects on neurons including the neurite dystrophy, tau hyper‐ phosphorylation, and amyloidogenic APP processing that drives neuronal degeneration (see

Alternatively, DHA may be released from phospholipids due to the activity of PLA2 (phos‐ pholipase A2), acting as a signaling molecule, and DHA can be hydroxylated to produce several secondary bioactive lipids such as resolvins (RVs) and protectins. DHA hydroxylation is mediated through lipoxygenase-15 (LOX-15) or acetylated cyclooxygenase-2 (COX-2) [63]. Compounds derived from DHA are classified as D-series RVs or protectins, while those formed from EPA are designated as E-series RVs. DHA can be hydroxylated on carbon 17 by 15-LOX or acetylated COX-2, leading to stereoselective formation of 17S- or 17R-hydroxy-DHA (17- HDHA), respectively. These derivatives may be further hydroxylated to give rise to trihydroxy derivatives such as the D1, D2, D3, and D4 17-(S/R)-RVs (D-series RVs), and the dihydroxy 17- (R)- and 17(S)-protectin, the latter also known as neuroprotectin D1 (NPD1). EPA can be stereoselectively hydroxylated to 18-(S/R)-hydroxy-EPA (18-HEPA) by cytochrome P450 or acetylated COX-2, which is further processed to form E1, E2 and E3 18-(S/R)-RVs (E-series RVs: **Figure 3**). Both, 17-HDHA and 18-HEPA serve as markers for RVs and protectins, and remarkably, their presence in blood is directly related to the intake of ω-3 PUFAs in animal

[71].

136 Update on Dementia

Section 3.1.).

**Figure 3.** Chemical structure of specialized pro-resolving mediators derived from DHA and EPA ω-3 fatty acids. DHA (docosahexaenoic acid) and EPA (eicosapentaenoic acid) may be released from phospholipids through PLA2 (phos‐ pholipase A2) activity and converted into bioactive hydroxylated fatty acids with potent anti-inflammatory properties, known as resolvins and protectins. This conversion may be mediated by several enzymes, including lipoxygenase-15 (LOX-15), acetylated cyclooxigenase-2 (COX-2), and cytochrome P450. Compounds derived from DHA are classified as D-series resolvins (RVs: left panel), while those formed from EPA are designated as E-series resolvins (right panel). Although resolvins normally includes trihydroxy fatty acids, DHA can be also transformed into dihydroxylated com‐ pounds denominated as protectins. Within this group, neuroprotectin D1 (NPD1: see left panel) is the best studied DHA-derived hydroxylated compound in terms of AD therapy, and it displays anti-inflammatory, anti-apoptotic, and anti-amyloidogenic properties.

The balance between ω-6 to ω-3 intake has a strong impact on brain health. In Western diets, this ratio is about 10–20:1, while in other cultures and also historically, this ratio has been as low as 1–2:1. Total fat intake as well as the ω-6 to ω-3 ratio in Western diets has increased significantly since the Industrial Revolution, indicating that Western diets are deficient in ω-3 PUFAs [77]. Epidemiological studies, including correlational studies and migration studies, suggest a protective effect against AD of ω-3 PUFAs and fish oil (an important source of ω-3 PUFAs), such that the role of nutrition in preventing AD arouses increasing hope, particularly with reference to ω-3 PUFA dietary intake. One recent meta-analysis reviewed a total of six cohort studies performed in the USA and Europe to address how dietary intake of long-chain ω-3 PUFAs or fish correlates with the incidence of dementia and AD [78]. This meta-analysis found a significant lower risk of AD associated with high fish intake. Such an association was most pronounced when the follow-up period was at least five years and fish intake was 500 g or more per week, such that fish consumption is inversely correlated with AD incidence in a temporal and quantitative manner. A dose–response meta-analysis also showed that for every 100 g per week dietary fish intake the risk of AD falls 11%. This neuroprotective effect of fish intake was mainly attributed to its high long-chain ω-3 PUFA content, particularly DHA [79]. Interestingly, the same meta-analysis also revealed that dietary intake of ω-3 PUFAs alone (not linked to fish consumption) did not lower the risk of dementia or AD. Moreover, an earlier randomized trial reached the same conclusions in patients with mild-to-moderate AD who were administered DHA [80]. Nevertheless, most of the individual studies evaluating the relationships between ω-3 PUFA intake and AD risk suggest there is a potential protective effect of these long ω-3 PUFAs on the incidence of AD, although no significant statistical differences were reached in the pooled analysis.

The discrepancies between fish and ω-3 PUFA consumption in relation to AD incidence may be explained by different factors in terms of the dietary composition or socioeconomic status of the individual. In this context, dietary intake of long-chain ω-3 PUFAs may also be accom‐ panied by the intake of other saturated fats, which would attenuate the neuroprotective effect of ω-3 PUFAs. Alternatively, fish is also a good source of vitamins, essential amino acids and other nutrients, which could in turn be responsible for the beneficial effect attributed to fish in AD prevention. The fact that DHA is converted into bioactive derivatives that mediate its beneficial effects in CNS cannot be overlooked. In this context, the neuroprotective effect of fish intake could be also attributed to PUFA derivatives present in fish, such as hydroxylated forms of PUFAs or PUFA forms easily transformable into bioactive derivatives similar to NPD1 [81]. In fact, fish oil consumption has recently been related to increased levels of total DHA and NPD1-like derivatives in the mouse brain, without any modification of free (unesterified) DHA levels [82]. Hence, fish oil intake promotes elevated levels of NPD1 without affecting basal levels of free DHA in the brain. These data bring to light a central role for ω-3 PUFA hydroxylated bioactive derivatives in the prevention and treatment of AD (see Section 5.2.).

#### **3.3. Specific lipid alterations as potential biomarkers in AD**

Modern lipidomic analysis allows a comprehensive atlas to be built up of all the lipid altera‐ tions existing in the AD brain. Current laboratory techniques, such as ultra/high pressure liquid chromatography (U/HPLC) and gas chromatography (GC) coupled to mass spectrom‐ etry (MS) allow the vast majority of lipids in cells and animal tissues to be studied. Since the brain is the most lipid-enriched organ in the human body, after adipose tissue, alterations in lipid composition might be involved in many neurological disorders, including AD [44]. An in-depth lipidomic analysis performed in the postmortem brain of patients with AD showed heterogeneous changes in lipid metabolism in AD-affected patients [47]. As expected, the cerebellum lipid profile was largely unaffected whereas significant lipid changes were observed in the prefrontal and entorhinal cortex of AD brains when compared with agematched controls. These changes demonstrate that lipid alterations are restricted to ADaffected brain regions (principally the cortex and hippocampus) and that they are not present in unaffected regions like the cerebellum. Interestingly, the prefrontal cortex displays more severe lipid alterations, with a decrease in PE, LPC (lyso-phosphatidylcholine), and sulfatides, together with elevated levels of ceramides (including glucosyl- and galactosyl-ceramides, Cer) and DAG (diacylglycerol). By contrast, in the entorhinal cortex, significant increases are only evident in LBPA (lysobiphosphatidic acid), SM, ganglioside GM3, and cholesterol esters (ChoE). In addition, polyunsaturated PE 40:6, 38:6, and 38:4 species were markedly downre‐ gulated in the prefrontal cortex, whereas there was a general decrease in long-chain fatty acids (≥40C) and a corresponding increase in short-chain fatty acids (≤34C) that is compatible with the lower levels of PE carrying DHA in the brain of patients with AD. Unexpectedly, the entorhinal cortex displays more species of the polyunsaturated lipid pools in PC (phosphati‐ dylcholine) and PE. The different lipid alterations between these two brain regions may reflect different aspects or stages of AD pathophysiology, since the entorhinal cortex is known to be affected earlier and more severely than neocortical areas [83].

with reference to ω-3 PUFA dietary intake. One recent meta-analysis reviewed a total of six cohort studies performed in the USA and Europe to address how dietary intake of long-chain ω-3 PUFAs or fish correlates with the incidence of dementia and AD [78]. This meta-analysis found a significant lower risk of AD associated with high fish intake. Such an association was most pronounced when the follow-up period was at least five years and fish intake was 500 g or more per week, such that fish consumption is inversely correlated with AD incidence in a temporal and quantitative manner. A dose–response meta-analysis also showed that for every 100 g per week dietary fish intake the risk of AD falls 11%. This neuroprotective effect of fish intake was mainly attributed to its high long-chain ω-3 PUFA content, particularly DHA [79]. Interestingly, the same meta-analysis also revealed that dietary intake of ω-3 PUFAs alone (not linked to fish consumption) did not lower the risk of dementia or AD. Moreover, an earlier randomized trial reached the same conclusions in patients with mild-to-moderate AD who were administered DHA [80]. Nevertheless, most of the individual studies evaluating the relationships between ω-3 PUFA intake and AD risk suggest there is a potential protective effect of these long ω-3 PUFAs on the incidence of AD, although no significant statistical

The discrepancies between fish and ω-3 PUFA consumption in relation to AD incidence may be explained by different factors in terms of the dietary composition or socioeconomic status of the individual. In this context, dietary intake of long-chain ω-3 PUFAs may also be accom‐ panied by the intake of other saturated fats, which would attenuate the neuroprotective effect of ω-3 PUFAs. Alternatively, fish is also a good source of vitamins, essential amino acids and other nutrients, which could in turn be responsible for the beneficial effect attributed to fish in AD prevention. The fact that DHA is converted into bioactive derivatives that mediate its beneficial effects in CNS cannot be overlooked. In this context, the neuroprotective effect of fish intake could be also attributed to PUFA derivatives present in fish, such as hydroxylated forms of PUFAs or PUFA forms easily transformable into bioactive derivatives similar to NPD1 [81]. In fact, fish oil consumption has recently been related to increased levels of total DHA and NPD1-like derivatives in the mouse brain, without any modification of free (unesterified) DHA levels [82]. Hence, fish oil intake promotes elevated levels of NPD1 without affecting basal levels of free DHA in the brain. These data bring to light a central role for ω-3 PUFA hydroxylated bioactive derivatives in the prevention and treatment of AD (see Section 5.2.).

Modern lipidomic analysis allows a comprehensive atlas to be built up of all the lipid altera‐ tions existing in the AD brain. Current laboratory techniques, such as ultra/high pressure liquid chromatography (U/HPLC) and gas chromatography (GC) coupled to mass spectrom‐ etry (MS) allow the vast majority of lipids in cells and animal tissues to be studied. Since the brain is the most lipid-enriched organ in the human body, after adipose tissue, alterations in lipid composition might be involved in many neurological disorders, including AD [44]. An in-depth lipidomic analysis performed in the postmortem brain of patients with AD showed heterogeneous changes in lipid metabolism in AD-affected patients [47]. As expected, the cerebellum lipid profile was largely unaffected whereas significant lipid changes were

differences were reached in the pooled analysis.

138 Update on Dementia

**3.3. Specific lipid alterations as potential biomarkers in AD**

AD progresses from a pre-symptomatic stage to mild cognitive impairment (MCI), mild AD and to severe AD with a gradual deterioration in cognitive abilities. Unfortunately, the clinical manifestation of the disease is preceded by a long prodromal phase, during which neuropa‐ thological lesions arise, including neuron death. For this reason, clinical diagnosis of AD is unreliable, particularly at early disease stages. Hence, there is a strong need to find peripheral biomarkers to reliably diagnose AD early, thereby enabling early treatment and better therapeutic efficacy. Most approaches to fluid-based biomarker discovery have focused on Aβ42, total tau and phosphorylated tau in cerebrospinal fluid (CSF). Although these are useful to distinguish symptomatic patients from normal controls or other dementias, these CSF biomarkers lack predictive value in preclinical patients, and they are only useful to confirm the clinical diagnosis [84]. Thus, given the brain lipid alterations in AD, lipidomic analysis of lipid derivatives in biological fluids may represent a reliable way to identify non-invasive biomarkers for early AD diagnosis [85].

Of the lipid changes reported in the CSF, plasma, and serum of patients with AD, many do not necessarily correlate with those described previously in the CNS [6]. For instance, free cholesterol and ChoE were reported to be downregulated in the CSF although they are increased in the brain of patients with AD [86] (see Section 3.1.). However, six different longchain ChoE species in plasma allowed patients with AD to be accurately discriminated from healthy controls (ChoE 32:0, 34:0, 34:6, 32:4, 33:6, and 40:4). These metabolites accumulated more strongly in healthy controls than in MCI, and in MCI than in AD, such that they were proposed as potential biomarkers for early AD diagnosis [87]. Total PC levels and specific PC species have also been proposed as reliable biomarkers, with diminished PC levels in the CSF of patients with AD accompanied by lowered LPC and increased PC hydrolytic products such as glycerophosphocholine and phosphocholine, suggesting that PC breakdown might be enhanced in AD pathogenesis [88]. Notably, a set of 10 PC metabolites was specifically depleted in the plasma of healthy individuals who later suffered phenoconversion towards MCI/AD. These subjects were diagnosed as AD during a 5-year follow-up even though they displayed no cognitive impairment at entry. The PC species identified were diacyl PC 36:6, 38:0, 38:6, 40:1, 40:2, 40:6, PC acyl-alkyl 40:6, and LPC 18:2, as well as the acylcarnitines (ACs) propionyl AC (C3), and C16:1-OH [89]. It is noteworthy that control subjects (not previously diagnosed with AD) did not display any of these modifications, while already diagnosed patients with AD also showed decreased levels of these PC species. Moreover, downregulation of this panel of lipids predicted phenoconversion from healthy to MCI/AD within a 2–3 year time frame with 90% accuracy [89]. These data were supported by independent studies showing decreased levels of PC 38:4, 38:6, and 40:6 in the plasma or serum of AD subjects [86, 90]. In addition, a variety of peripheral lipid changes were also reported that might potentially be useful for early AD diagnosis, such as lower levels of SM and increased levels of Cers in the plasma or serum of patients with AD. In particular, there were significantly fewer SM species containing long chains (e.g., 22 and 24 carbon atom acyl chains) in AD subjects [86, 91]. In parallel, increased Cer levels were reported in the plasma of patients with AD [91, 92]. SM can be metabolized into Cers, second messengers that regulate cellular differentiation, proliferation and apoptosis. Upregulated levels of Cers were concomitant with significant reductions in SM in the plasma of patients with AD. A correlation between the decrease in SM and the increase in Cers was particularly robust in the ratios of SM and Cer species with identical fatty acyl chains. Cer alterations were particularly evident in mild-to-moderate stages of AD [91]. Moreover, it is noteworthy that upregulated Cer levels were significantly correlated with the onset of memory impairment, supporting the role of Cers as potential AD biomarkers [92].

In conclusion, a wide range of peripheral fluid changes have been described that could be used as biomarkers for early AD diagnosis. However, many of the clinical studies involved are crosssectional in nature and some of them do not reveal reliable biomarkers to test disease pro‐ gression. Nevertheless, longitudinal studies with several years of follow-up do identify promising biomarkers for early AD diagnosis that reliably predict cognitive impairment and the onset of AD.
