Lipid metabolism and Alzheimer's disease: clinical evidence, mechanistic link and therapeutic promise

Abstract

Alzheimer’s disease (AD) is an age-associated neurodegenerative disorder with multifactorial etiology, intersecting genetic and environmental risk factors, and a lack of disease-modifying therapeutics. While the abnormal accumulation of lipids was described in the very first report of AD neuropathology, it was not until recent decades that lipid dyshomeostasis became a focus of AD research. Clinically, lipidomic and metabolomic studies have consistently shown alterations in the levels of various lipid classes emerging in early stages of AD brains. Mechanistically, decades of discovery research have revealed multifaceted interactions between lipid metabolism and key AD pathogenic mechanisms including amyloidogenesis, bioenergetic deficit, oxidative stress, neuroinflammation, and myelin degeneration. In the present review, converging evidence defining lipid dyshomeostasis in AD is summarized, followed by discussions on mechanisms by which lipid metabolism contributes to pathogenesis and modifies disease risk. Furthermore, lipid-targeting therapeutic strategies, and the modification of their efficacy by disease stage, ApoE status, and metabolic and vascular profiles, are reviewed.

Abbreviations

AA

arachidonic acid

ABCA1/7

ATP- binding cassette subfamily A member 1/7

ACAT1

acyl CoA:cholesterol acyltransferase 1

ACC

acetyl-CoA carboxylase

AD

alzheimer’s disease

ALCAR

acetyl-L-carnitine

ApoE

apolipoprotein E

APP

amyloid precursor protein

Aβ

β-amyloid

BBB

blood-brain barrier

CE

cholesteryl ester

CLU

clusterin

CNS

central nervous system

COX-1/2

cyclooxygenase-1/2

CPT

carnitine palmitoyltransferase

CSF

cerebrospinal fluid

DAG

diacylglycerol

DHA

docosahexaenoic acid

EPA

eicosapentaenoic acid

ER

endoplasmic reticulum

FFA

free fatty acid

fAD

familial AD

FAS

fatty acid synthase

HDL

high-density lipoprotein

LD

lipid droplet

LOAD

late-onset AD

LRP1

lipoprotein receptor-related protein 1

LXR

liver X receptor

MAG

monoacylglycerol

MAM

mitochondria-associated membrane

MCI

mild cognitive impairment

MUFA

monounsaturated fatty acid

NFT

neurofibrillary tangle

PGE2

prostaglandin E2

PICALM

phosphatidylinositol-binding clathrin assembly protein

PLA2

phospholipase A2

PPAR

peroxisome proliferator-activated receptor

PSEN1/2

presenilin-1/2

PUFA

polyunsaturated fatty acid

ROS

reactive oxygen species

RXR

Retinoid X receptor

SNP

single nucleotide polymorphism

SPM

specialized pro-resolution mediator

SREBP

sterol regulatory-element binding protein

T2D

type 2 diabetes

TAG

triacylglycerol

TREM2

triggering receptor expressed on myeloid cells 2

TZD

thiazolidinedione

VLCFA

very-long-chain fatty acid

Introduction

Alzheimer’s disease (AD) is the most prevalent form of dementia. Worldwide, approximately 50 million people suffer from AD, and this number is projected to reach 82 million in 2030 and 152 million in 2050 [[1]]. AD is clinically characterized by progressive memory deficits, cognitive impairment, and other behavioral changes accompanied by structural abnormalities in the brain [[2, 3]]. Neuropathologically, AD features extracellular aggregation of β-amyloid (Aβ) forming senile plaques, intraneuronal aggregation of hyperphosphorylated tau forming neurofibrillary tangles (NFT), and loss of neurons and synapses in the hippocampus and neocortex among other brain regions [[2, 3]]. Biomarkers of these three key features constitute the core of a biological framework—the amyloid, tau, neurodegeneration (AT(N)) system—to define disease onset and progression [[4]].

While diagnosed in different stages, AD develops along a continuum with progressive accumulation of biomarkers, neural damage, and cognitive decline [[4]]. Before disease onset, preclinical AD is characterized by subtle cognitive decline and accumulated Aβ deposits, and the subsequent transition to mild cognitive impairment (MCI) is associated with a high probability of progressing to AD dementia [[5]]. There are two types of AD—early-onset and late-onset. The late-onset form (LOAD), which accounts for more than 95% of total cases [[6, 7]], has a multifactorial nature and poorly understood molecular mechanisms. The early-onset, and autosomal dominant, form of AD (familial AD; fAD), caused by rare mutations in amyloid processing genes including APP (amyloid precursor protein), PSEN1 (presenilin-1), and PSEN2, accounts for ~ 1% of total cases [[8]].

In the past few decades, therapeutics targeting the Aβ cascade have predominantly failed at various stages of clinical trials [[9-12]], calling for refined and personized intervention strategies and/or alternative therapeutic targets [[13, 14]]. Despite the recent U.S. FDA approval of AD drug aducanumab, which raised debates on its ambiguous efficacy and safety profile [[15]], there is still a lack of disease-modifying or preventative strategy beyond symptom alleviation.

Beyond pathological hallmarks involving Aβ and tau, AD brains are characterized by microglia-mediated inflammation, as well as metabolic abnormalities encompassing glucose hypometabolism, mitochondrial dysfunction, oxidative stress, and lipid dyshomeostasis [[16-20]]. Hypometabolism of glucose is the best-known form of metabolic dysfunction in AD. It not only occurs as an early event in AD-vulnerable regions in the brain [[21, 22]] but its presence prior to amyloid or tau detection results in the fastest rate of disease progression [[23]]. In addition to glucose metabolism, abundant clinical investigations and epidemiological studies have strongly connected disrupted lipid metabolism with altered disease risk and the pathogenesis and progression of AD. In fact, lipid-related abnormalities are among the initial neuropathological findings identified by Alois Alzheimer in 1907 [[24]].

Lipids are a major class of nutrients that are vital for all organisms. They represent a diverse group of biomolecules that are structurally and functionally involved in a wide variety of cellular processes and tissue functions. As the body’s second most lipidated organ—only after adipose tissue–10% to 12% of the fresh weight, and more than 50% of the dry weight, of the brain is composed of lipids [[25, 26]]. Major lipid species in the brain can be generally categorized to phospholipids, sphingolipids, glycerolipids, fatty acids, and sterols, with phospholipids accounting for ~ 50% of total lipid content [[27]]. Except for cholesterol, all lipids with various backbones carry at least one fatty acid (acyl) chain. For cholesterol, it can also form a cholesteryl ester (CE) by attaching to a fatty acid chain (esterification). Functionally, these lipids are key components of cellular membranes including synapses and myelin sheath; they can serve to transduce signaling to regulate a range of biological processes; and in some circumstances, lipids can be used as bioenergetic fuels [[28]].

Given the significance of various lipids in brain physiology, it is not surprising that accumulating research has discovered complex and diverse mechanisms that connect lipid metabolism with AD-related pathophysiologies. Further, development of lipid-targeting therapeutics has exhibited various degrees of efficacy against AD pathologies. In this review, I will summarize clinical evidence that demonstrates altered lipid metabolism in AD, known mechanisms that connect lipid metabolism and AD etiology, and the potential of AD therapeutics that target lipid metabolism.

Correlating lipids with AD: clinical evidence

Changes in lipid levels in AD

Fatty acids

Fatty acids are building blocks for all lipid classes except cholesterol. They are hydrocarbon chains with varying lengths terminated with carboxylic acid groups (Fig. 1). Elevated levels of free fatty acids, and their metabolic intermediates acyl-carnitines and acyl-CoA, are neurotoxic and can induce mitochondrial uncoupling and bioenergetic dysfunction [[29, 30]]. Perturbations to brain fatty acid metabolism in AD are suggested by altered levels of free fatty acid levels. While the total free fatty acid levels are higher in the cerebrospinal fluid (CSF) of AD brains [[31]], each subclass of fatty acid show diverse shifts.

Fig. 1

Classification of fatty acids. Fatty acids can be classified to saturated- and unsaturated fatty acids, and the unsaturated ones can be further grouped to MUFA and PUFA. Chemical structures of palmitate (saturated fatty acid), oleate (MUFA) and DHA (PUFA) are shown as examples of each class. Saturated fatty acids are negative regulators of mitochondrial function. MUFA can be generated from saturated fatty acid by SCD-1. ω-3 PUFAs, such as DHA and EPA generated from α-linolenic acid (ALA) or from diet, are precursors of SPMs, which are involved in inflammation resolution. AA, a ω-6 PUFA generated from LA or obtained from diet, elicits pro-inflammatory effect via the production of prostaglandin E₂ (PGE₂) by cyclooxygenase 1/2 (COX1/2) and prostaglandin E synthase (PGES). ALA and LA are essential fatty acids that can only be obtained from diet for human. PGH₂, Prostaglandin H₂.

Fatty acid can be classified by their saturation status to saturated- and unsaturated fatty acid, and the latter can be further grouped to monounsaturated- (MUFA) and polyunsaturated fatty acid (PUFA) (Fig. 1). Changes in the abundance of unsaturated fatty acid are found associated with AD in multiple studies. Across these studies, both brain (cortex, middle frontal gyrus, and inferior temporal gyrus)- and plasma levels of unsaturated fatty acids, including ω-3 PUFAs and a MUFA (oleic acid), are found lower in AD individuals than cognitively normal controls, and these changes lead to an overall reduced unsaturation index [[32, 33]]. As the most abundant PUFA in the brain, docosahexaenoic acid (DHA) levels are also lower in AD brains, particularly in regions affected by AD such as the hippocampus [[34, 35]]. Importantly, DHA levels are positively correlated with cognitive performance. Corresponding to the declines in unsaturated fatty acids and the unsaturation index, even-chain saturated fatty acids are increased in the CSF of AD patients [[31]].

While ω-3 fatty acids are believed to be anti-inflammatory, arachidonic acid (AA) of the ω-6 fatty acid family is largely pro-inflammatory, and the levels of free AA increase in AD brains. Such an increase is associated with the decrease in its precursor linoleic acid (LA), and a pro-inflammatory- and pro-oxidative state in the CSF [[35]]. Corresponding to the increase in free AA, the levels of AA in phospholipids are reduced in the hippocampus of AD subjects [[36]]. AA can be produced from phospholipids by the hydrolysis at the sn-2 acyl bond by phospholipase A2 (PLA2) [[37]]. Consistent with the shift in AA pool from being phospholipid-bound towards the free form, total PLA2 activity, as well as the expression of the Ca²⁺-dependent cytosolic PLA2 (cPLA2) are higher in the cortex and CSF of AD brains [[38, 39]].

Lipid rafts are cholesterol- and sphingolipid-enriched membrane structure, where most of the proteins involved in synaptic transmission and the amyloidogenic secretases are located [[40, 41]]. Consistent with changes at tissue level, the fatty acid profiles of lipid rafts in AD brains are characterized by lower ω-3 PUFA and MUFA (primarily oleic acid), and a reduced unsaturation index for phospholipid acyl chains in the cortex [[42, 43]]. Of note, these changes also manifest in the earliest stages of AD in entorhinal cortex and frontal cortex [[44]]. These data suggest that lipid rafts are potentially in the center of altered lipid homeostasis in early stages of Alzheimer’s brains.

Glycerophospholipids and sphingolipids

The two most abundant lipid classes in the brain are glycerophospholipids and sphingolipids. They are both amphiphilic lipids with two hydrophobic fatty acyl chains and a hydrophilic head attached to an alcohol-based backbone. Glycerophospholipids represent the major form of phospholipids in cell membranes. For each glycerophospholipid molecule, two fatty acyl tails and one phosphate polar head are attached to a glycerol backbone. Across lipidomic studies with postmortem brains of MCI and AD subjects, decreased levels of glycerophospholipid contents including phosphatidylcholine (PC) [[45, 46]], phosphatidylinositol (PI) [[47, 48]], and phosphatidylethanolamines (PE) [[45, 46]] are detected in multiple brain regions, primarily those vulnerable to AD pathology such as the hippocampus and the cortex.

Sphingolipids, including ceramide, sphingomyelin, and glycosphingolipid, are a lipid class containing a long-chain sphingoid base backbone (primarily sphingosine that is de novo synthesized from serine and palmitoyl-CoA). Ceramide is the key molecule in the synthesis, recycling, and degradation of other sphingolipids. Sphingomyelin consist of a phosphatidylcholine group attached to a ceramide and can also be classified as phospholipid (sphingophospholipid). An elevation in ceramide levels occurs early in AD brains, including the frontal and temporal cortices [[49-51]], which is coupled with a reduction in sphingomyelin levels [[52, 53]]. It should be noted that sphingomyelin changes in AD are disease stage- and brain region-specific [[54, 55]]. Gene and protein expression studies further suggest that changes in sphingomyelin and ceramide are associated with upregulated levels of genes involved in ceramide de novo synthesis and sphingomyelin degradation and downregulated expression of genes for sphingomyelin synthesis [[52, 56]]. As increased ceramide levels promote lipid peroxidation, oxidative stress, mitochondrial dysfunction, and neuronal death [[57, 58]], these findings collectively suggest a critical role of dysregulated sphingolipid metabolism in early stages of AD.

Sulfatides represent another class of ceramide-derived lipids that are enriched in myelin. Similar to sphingomyelins, sulfatide levels in both the gray matter and the white matter decline at prodromal and early stages of AD [[51, 59]]. Galactosylceramide, another class of lipids found in myelin, is also depleted in the hippocampus of early-stage AD brains proceeding NFT pathology [[60]]. The early loss of sulfatide and galactosylceramide in AD can be metabolically traced to the loss of their common biosynthetic precursor, very long chain ceramide, and the reduced activity of the enzyme catalyzing its synthesis, ceramide synthase 2 [[60]]. Functionally, reductions in these myelin-specific lipids are associated with myelin degeneration and loss of white matter integrity in the AD brains (discussed in section “Myelin degeneration and white matter integrity” below).

Glycerolipids

Glycerolipids are mono-, di-, or tri-esterified glycerol, corresponding to mono-, di-, and tri-acylglycerol (MAG, DAG, and TAG), respectively. They share the same glycerol backbone with glycerophospholipids but have variable numbers of acyl groups and do not carry a phosphate group. While TAG levels are not changed in AD brains, both MAGs and DAGs are elevated in the frontal cortex of MCI and AD brains [[61-63]], suggesting disrupted glycerolipid metabolism as an early event in AD. Consistently, lipid droplets (LDs), the organelles that serve as a storage of neutral lipids [[64]], accumulate in AD [[65]]. The expression of perilipin-2 (Plin2, a specific surface marker of LD), but not other perilipin family members, is selectively upregulated in the frontal cortex [[66]]. Increased neutral lipids and LDs in post-mortem AD brains, AD patient-derived fibroblasts [[67]], and peripheral blood mononuclear cells [[68]] collectively signify disrupted neutral lipid metabolism in AD (see reviews [[69, 70]]).

Cholesterol and cholesteryl esters

With its 2% of total body weight, the brain contains 25% of the total cholesterol [[71]]. Because of the blood-brain barrier (BBB), the majority of brain cholesterol is de novo synthesized in the central nervous system (CNS) [[72]]. Most studies suggest that the levels of both circulating and brain cholesterol are elevated in AD patients than in control cases [[50, 73-75]], although unchanged brain cholesterol levels are observed in some other reports [[76, 77]], which could be related to different brain regions and disease stages across these studies. Notably, increased cholesterol is also seen in the cores of senile plaques in human brain [[78]], and brain cholesterol levels are positively correlated with the severity of AD [[50]]. Under physiological conditions, excess cellular cholesterol is esterified to CEs, a form of neutral lipids stored in LDs together with TAG. Cholesterol esterification is catalyzed by acyl CoA:cholesterol acyltransferase 1 (ACAT1, also known as sterol O-acyltransferase 1) [[79]]. In the entorhinal cortex of human AD brains, total CE levels are elevated by ~ 1.8 fold, which is resembled in AD mouse models [[62, 80]].

Collectively, across various lipid classes discussed above, it can bse summarized that alterations in lipid raft fatty acid compositions, ceramide, sphingomyelin, sulfatide, and glycerolipid (MAG and DAG) temporally emerge in early stages of AD and/or spatially locate to brain regions that are affected by pathology at first. These lines of evidence support an early, and potentially initiating, role of lipid metabolism in AD.

Lipid-related risk factors for AD

The critical, and potentially initiating, role of lipid metabolism in AD is further supported by the intimate relationships between key AD risk factors and various aspects of lipid metabolism (Fig. 2). Such factors include both genetic risk factors identified from genome-wide association studies (GWAS) and environmental or lifestyle risks discovered from epidemiological studies [[3]].

Fig. 2

Lipid-related risk factors in AD. A variety of AD genetic risk factors are involved in different aspects of lipid metabolism including lipid trafficking, lipid synthesis and lipid signaling. Peripheral lipid modifiers including metabolic and vascular risk factors also alter AD risks.

Genetic risk factors

Among the top risk genes for the sporadic LOAD, APOE, TREM2, APOJ, PICALM, ABCA1, and ABCA7 are all directly involved in lipid trafficking or metabolism [[81]] (Fig. 2). A key regulator of cholesterol metabolism, SREBP-2 is also genetically associated with altered AD risk [[82]]. While this section focuses on risk factors for the predominant sporadic AD (LOAD), causal genes for fAD (APP, PSEN1 and PSEN2) also modulate lipid metabolism, which will be discussed in detail in section “Amyloid metabolism” below.

Apolipoprotein E

The human APOE gene encodes the 34KD ApoE protein, which is best known as a lipid-binding protein responsible for transporting lipids across organs in the periphery, and between cells in the brain [[83]]. Compared with the most common ε3 isoform, the ε4 isoform of ApoE (ApoE4) is the strongest genetic risk factor for LOAD [[84, 85]] whereas the ε2 isoform (ApoE2) significantly reduces the risk [[86, 87]]. Each ApoE4 allele increases AD risk by 3 ~ 4-fold, while lowers the age at onset by ~ 8 years [[88-90]]. Beyond its role in regulating Aβ production and clearance, the pathological effect of ApoE4 is also driven by lipid-centric mechanisms [[91]]. Consistent with its loss-of-function as an intercellular lipid carrier relative to ApoE3, ApoE4 carriers have higher plasma levels of total cholesterol and TAG but reduced high-density lipoprotein (HDL)-cholesterol; in contrast, plasma levels of total cholesterol and HDL-cholesterol in ApoE2 carriers are lower and higher, respectively [[92, 93]]. Moreover, the link between ApoE4 and disrupted lipid metabolism is supported by studies showing that ApoE4 enhances the activity of the AA-producing cPLA2 [[94]], whereas ApoE4-related pathologies can be alleviated by DHA-rich diet but exacerbated by high-cholesterol diet [[95]].

TREM2

Triggering receptor expressed on myeloid cells 2 (TREM2), primarily expressed on microglial surfaces, mediates phagocytosis and inflammatory response by these resident myeloid cells in the brain [[96, 97]]. TREM2 loss-of-function mutations cause an early-onset dementia called Nasu-Hakola disease, which is characterized by severe myelin loss and neurodegeneration [[98, 99]]. A rare variant of TREM2 (R47H) is associated with a significant increase in AD risk [[100, 101]]. Beyond its interaction with Aβ, TREM2 has been identified as a sensor of phospholipids, lipoproteins and apolipoproteins [[102-104]].

Clusterin

Clusterin (ApoJ, gene symbol CLU) is identified as an AD risk gene based on the association between multiple single nucleotide polymorphisms (SNPs) at the CLU locus and altered disease risk [[105, 106]]. Beyond its role in the folding of secreted proteins as a molecular chaperone and its effect on Aβ aggregation, clusterin is also involved in lipid transport and metabolism in both the brain and the periphery [[107]].

PICALM

As another major AD risk gene [[105]], phosphatidylinositol binding clathrin assembly protein (PICALM) is involved in clathrin-mediated endocytosis, which is essential in the internalization and transport of not only proteins but also lipids in lipoprotein particles [[108]]. Interestingly, the Aβ clearance role of PICALM is mediated by its binding to the low-density lipoprotein receptor related protein 1 (LRP1) [[109]].

ABCA1 and ABCA7

Both ABCA1 and ABCA7 (ATP-binding cassette subfamily A member 1 and 7) are members of the ABC transporter family, sharing 54% sequence identity [[110]]. ABCA1 initiates the efflux of lipids such as cholesterol and phospholipids by loading them to lipid-free lipoproteins, with ApoE being the primary substrate for such lipidation in the brain [[111, 112]]. A loss-of-function mutation in ABCA1 is associated with low plasma levels of ApoE and higher AD risk [[111]]. ABCA7 is also involved in the transport of cholesterol and phospholipids [[113]]. Human-based genetic and epigenetic studies have identified multiple SNPs, variants, alternative splicings, and methylations of ABCA7 gene that contribute to its loss-of-function, altered lipid- and Aβ metabolism and increased AD risks [[110, 114]].

SREBP-2

Sterol regulatory-element binding proteins (SREBPs) are a family of transcriptional factors that regulate the biosynthesis of lipids [[115]]. As the master regulator of cholesterol synthesis, SREBP-2 is found connected with AD risk and progression. SREBF2 (gene encoding SREBP-2) polymorphism (rs2269657) is associated with LOAD biomarkers and gene expression, and brain SREBF2 mRNA levels are negatively correlated with the age at death [[82]]. SREBP-2 could thus be a potential risk factor for AD (also see section “Fatty acid and cholesterol synthesis” below).

Metabolic and vascular factors

The essential role of lipid homeostasis in the etiology of AD is also supported by the modification of AD risks by multiple lipid-related factors—either positively or negatively.

It has been established epidemiologically that metabolic disorders such as obesity and type 2 diabetes (T2D) are associated with increased risk for AD (Fig. 2), and diets enriched in saturated fatty acids are linked to learning and memory deficits [[116-121]]. Midlife, but not late-life overweight or obesity is associated with increased AD risk and an earlier onset [[122, 123]]. Multiple studies show that higher systemic insulin resistance in AD patients is positively correlated with Aβ deposition in brain [[124, 125]]. Interestingly, in a study conducted in non-dementia subjects, midlife, but not late-life systemic insulin resistance is associated with greater brain Aβ burdens 15 years later [[126]]. Notably, recent neuropathological studies suggest that the association between T2D and AD pathology (Aβ deposits and NFTs) is not significant [[127, 128]]. Despite the lack of evidence supporting a direct effect of T2D on AD pathogenesis, it has been suggested that metabolic mechanisms that are commonly affected in both disorders, such as dyslipidemia, oxidative stress, and cerebrovascular function, mediate their epidemiological connection [[127, 128]].

Acylcarnitines are intermediates of fatty acid metabolism during fatty acid transport into the mitochondria by the carnitine-carrier system. Two pilot studies suggest that serum levels of acetyl-l-carnitine and other acyl-carnitines decrease along the continuum from cognitively normal to MCI, and to AD [[129, 130]]. These studies suggest a decreased systemic fatty acid catabolism in AD that may indicate altered fatty acid degradation in the brain. Interestingly, the decline in circulating acyl-carnitines is also associated with lower levels of ketone bodies [[130]], which are the major alternative substrate that the brain can utilize upon glucose insufficiency or hypometabolism. These findings connecting midlife alterations in systemic lipid metabolism with late life AD risk collectively underline an early role of abnormalities in lipid metabolism in modifying AD susceptibility.

While the cholesterol pool of the CNS is separate from its systemic counterpart, high circulating levels of cholesterol, familial hypercholesterolemia, and low circulating HDL-cholesterol are known risk factors for AD [[131-133]] (Fig. 2). High midlife cholesterol levels increase the risk of AD by 2.8-fold [[134]]. Persistent midlife hypertension is another significant risk factor for late life incidence of both all-cause dementia and AD [[135, 136]] independent of cardiovascular diseases [[137]]. In contrast, people with a normal cardiovascular profile have a lower 10-year risk of all-cause dementia and of AD [[138]]. In a separate cohort, a comprehensive cardiovascular health score considering both metabolic and vascular factors in midlife (diet, body mass index, physical activity, fasting glucose, blood pressure, and blood cholesterol) is associated with a lower risk of dementia later in life [[139]], although this study did not separate AD from other dementia etiologies (such as vascular dementia).

These findings, along with the alterations in lipid species occurring early in the brain, consistently suggest that both central- and peripheral lipid dysregulation play an important role in the initial stages of AD.

Deciphering lipid metabolism in AD: mechanistic links

Lipid homeostasis in AD - the dynamics of lipid synthesis, transport and degradation

Because of their essential roles, the levels of lipid species are finely regulated for their homeostasis and functionality in the brain. In this section, key players and mechanisms involved in the regulation of lipid homeostasis will be discussed in the context of AD. Given that fatty acids are the essential building blocks for nearly all lipid species and excess fatty acids are associated with lipotoxicity to both neuronal and non-neuronal cells via multiple mechanisms [[140-144]] (reviewed in [[145]]), the discussion will be centered on the anabolic and catabolic metabolisms and the transport of fatty acids, with connections to the overall lipid profile in the brain (Fig. 3).

Fig. 3

Fatty acids in the center stage of the dynamic regulation of brain lipid homeostasis. Free fatty acids can be generated from: 1) lipolysis of phospholipids by PLA and TAG by adipose triglyceride lipase (ATGL) or hormone-sensitive lipase (HSL); 2) de novo synthesis from acetyl-CoA involving key enzymes ACC and fatty acid synthase; or 3) autophagic degradation of lipophagy. After being converted to acyl-CoA by various forms of ACS, fatty acids can be: 1) used to synthesize phospholipid, ceramide and sphingomyelin (SM); 2) transported into the mitochondria by carnitine palmitoyltransferase 1/2 (CPT1/2) for β-oxidation, which further generates acetyl-CoA for ATP production or ketogenesis (primarily occurs in astrocytes); or 3) sequestered in LDs in the forms of TAG or cholesterol ester (CE; acyl chain attached to cholesterol by ACAT). DAG, diacylglycerol; SMS, sphingomyelin synthase; SMase, sphingomyelinase; KB, ketone body; ACAT, acyl CoA:cholesterol acyltransferase.

Fatty acid and cholesterol biosynthesis

While short- and medium-chain fatty acids are transported from the periphery, long- and very-long-chain fatty acids are primarily de novo synthesized in the brain from acetyl-CoA [[146]]. Additionally, brain fatty acids can be generated by phospholipase A2 (PLA2)-catalyzed phospholipid hydrolysis [[147]], and potentially by lipophagy [[148]] (Fig. 3). In the brain of a fAD mouse model overexpressing mutant APP and PSEN1 (APP/PS1), increased palmitic acid (C16) levels are accompanied by upregulated protein levels of fatty acid synthase (FAS) [[149]]. In human AD cortex, FAS protein levels are also elevated, especially at areas around the plaques [[149, 150]]. Consistently, another key enzyme controlling fatty acid synthesis, acetyl-CoA carboxylase (ACC), is activated in fAD mouse brain as suggested by a decrease in its inhibitory phosphorylation [[151]].

Uncontrolled cholesterol synthesis in brain is also associated with worsened AD-related phenotypes. Increased cholesterol synthesis by overexpressing SREBP-2 in APP/PS1 mice accelerates oxidative damage, amyloid accumulation, neuroinflammation, and cognitive decline, and remarkably, it induces tau hyperphosphorylation and NFT formation in the absence of tau transgene or mutation [[152]]. Conversely, astrocyte-specific (astrocytes are the major cholesterol-synthesizing cells in the brain) SREBF2 knockout dramatically reduces both Aβ- and tau pathologies [[153]]. A recent study suggests that tau, in the opposite direction, alters the distribution and signaling of SREBPs in AD [[154]]. Overall, these findings validate the detrimental effect of SREBP-2 on AD mortality in humans [[82]].

Fatty acid oxidative degradation

The oxidative degradation of free fatty acids via β-oxidation is primarily in the domain of mitochondria, and secondarily, peroxisomes. While the role of fatty acids, as fuel substrates, in meeting the energetic demand of the brain remains unclear [[145, 155]], a timely degradation of free fatty acids, especially those peroxidized ones, is critical for brain health because of their negative effect on cellular function [[29]]. Compared with the oxidative degradation of glucose, fatty acid β-oxidation by mitochondria is associated with higher production of reactive oxygen species (ROS). Correspondingly, neurons, which have high metabolic rate but limited ROS-detoxification capacity [[156]], possess low β-oxidation activity.

Astrocytes are a highly heterogeneous population of neural cells and represent the primary brain cell type degrading fatty acids [[141, 157-159]] (Fig. 3). Relative to neurons, the metabolic role of astrocytes in AD and neurodegeneration has just started to be revealed. In the context of AD, a recent study by our group suggests that compromised astrocytic degradation of fatty acids could constitute another potential mechanism underlying lipid dysregulation in brain [[160]]. In astrocytes, human ApoE4 knockin induces a metabolic shift towards enhanced glucose metabolism and reduced fatty acid β-oxidation, which subsequently elicits lipid accumulation in astrocytes. Consistently, this metabolic shift in ApoE4 astrocytes is accompanied by an increase in fragmented mitochondria [[160]], which are less capable of distributing and oxidizing fatty acid than fused ones [[161]]. ApoE4-induced astrocytic lipid accumulation is also seen in immortalized cells [[162]] and human iPSC-derived cells [[163]]. Further, our group developed an ex vivo brain fatty acid β-oxidation assay with acute brain slices that enables metabolic assessment of brain lipid catabolism [[164]]. Using this method, a reduced fatty acid degradation and increased TAG accumulation in the hippocampus of young ApoE4 knockin mice is observed, confirming the unique role of astrocytes in brain fatty acid degradation. These findings are consistent with a previous report in Drosophila that loss of a mitochondrial fatty acid importer (carnitine palmitoyltransferase 2; CPT2) leads to LD accumulation in the brain [[165]]. In line with reduced fatty acid oxidation capacity associated with ApoE4 and potentially AD, the mRNA levels of peroxisome proliferator-activated receptor α (PPARα), the nuclear receptor promoting fatty acid degradation, is significantly reduced in AD brains [[166]].

Beside astrocytes, microglia also express enzymes for β-oxidation [[167]], and a growing body of literature suggest fatty acid metabolism as a determinant of microglial phenotype. Evidence in macrophages suggest that their inflammatory response and polarization are modulated by fatty acid oxidation [[168]]. Inhibition of macrophage β-oxidation exacerbates palmitate-induced inflammatory activation, whereas enhancement of fatty acid oxidation reduces TAG accumulation and lipid-induced inflammation [[169, 170]]. In microglia, mRNA levels of fatty acid oxidation genes are repressed by pro-inflammatory activation by lipopolysaccharides (LPS) or interferon-γ (IFN-γ) [[171]]. Consistently, interventions that activate mitochondrial function and facilitate microglial metabolic reprogramming alleviate neuroinflammation and promote Aβ clearance [[172-175]].

The peroxisome is another organelle that performs β-oxidation, specifically for very-long-chain fatty acids (VLCFAs; >= 22 carbons) [[176]]. Evidence for the impairment of peroxisomal function in advanced AD stages include the accumulation of VLCFA in AD brains with Braak stages V-VI, the increased peroxisomal volume in the soma of neurons, and a loss of peroxisomes in neuronal processes with hyperphosphorylated tau [[177]]. Intriguingly, this study suggests that peroxisomal dysfunction has a stronger association with NFT than with neurite plaques. Studies also suggest the role of glial peroxisomes in AD-relevant brain phenotypes such as demyelination and neuroinflammation. In oligodendrocytes, a loss of functional peroxisomes causes demyelination, axonal degeneration, and neuroinflammation, which is coupled with an accumulation of VLCFAs, and these neurotoxic VLCFAs further lead to neurodegenerative phenotypes relevant to AD [[176]]. Likewise, peroxisomal dysfunction of β-oxidation in microglia caused by the loss of multi-functional protein-2 (MFP2) elicits a profound pro-inflammatory response [[178]].

Fatty acid and cholesterol sequestration in lipid droplets (LDs)

LDs are one of the first layers of protection against free fatty acid - or cholesterol-related lipotoxicity by sequestering them as the neutral TAGs and CEs [[64]] (Fig. 3). LDs are detected in multiple cell types in brain including astrocyte, microglia, ependymal cell and neuron across aging-, AD-, and other neurodegenerative brains (reviewed in [[69]]). While the exact role of LD in AD pathogenesis remains to be determined, elevated LD accumulation in AD is consistent with the function of these organelles as sequesters of peroxidized fatty acids as well as unmodified PUFAs that are prone to peroxidation [[69]]. In a fAD mouse model overexpressing mutant APP, PSEN1, and MAPT (3xTG), LD accumulation in ependymal cells of the subventricular zone proceeds amyloid and tau pathology and suppresses neural stem cell proliferation [[179]]. Further, LD-accumulating microglia have pro-inflammatory and defective phagocytosis phenotype and possess transcriptional signatures overlapping with that of AD microglia [[180]]. In cultured neurons, ApoE4 induces suppressed fatty acid sequestering in LDs, which leads to elevated free fatty acid levels, disrupted bioenergetics, and synaptic dysfunction [[160]]. Moreover, CEs are upstream regulators of both Aβ and tau pathologies during early AD development [[181]].

It should be noted that while LDs are increased in aging and neurodegenerative diseases, they serve as a protective and adaptive mechanism to control lipid-mediated neurotoxicity by buffering excess free fatty acid and cholesterol, particularly during the initial stages of these lipid-implicated abnormalities. This notion is supported by the finding that blockage of glial LD accumulation causes more severe neurodegeneration [[182, 183]].

Intercellular lipid transport

Fatty acid transport

Because of their low fatty acid degradation capacity, neurons rely on glia, primarily astrocytes, to clear and degrade excessive free fatty acids and neutral lipids in LDs [[141, 160, 182, 183]]. Upon oxidative stress or neuronal stimulation, peroxidized fatty acids are transported from neurons to astrocyte, via an ApoE-dependent mechanism [[141, 182]] (Fig. 4). Our recent study suggests that neuron-to-astrocyte transport of lipids is impaired by ApoE4 [[160]]. Lipids within neuronal LDs are eliminated upon astrocyte exposure and transferred to astrocytes, with ApoE4 in either neurons or astrocytes diminishing the transport efficiency. Further, we demonstrate that the clearance of neuronal LDs is integral to the metabolic and neurotrophic support provided by astrocytes: when the export of neuronal LD is inhibited, the metabolic enhancement of neurons by astrocyte is abrogated [[160]]. Notably, this study also suggests a differential role of supplemental ApoE3 versus ApoE4 in regulating the clearance of neuronal lipids. Supplemental recombinant ApoE3, but not recombinant ApoE4, promotes astrocyte-induced clearance of neuronal LDs, which suggests ApoE3 as an enhancer, while ApoE4 as a negative regulator, of neuronal lipid clearance [[160]].

Fig. 4

Intercellular lipid transport in the brain. Cholesterol synthesized in astrocyte by HMG-CoA reductase (HMG-CoAR) is packaged and exported in ApoE-containing lipoproteins (ApoE-particles) by ATP-binding cassette transporter A1 (ABCA1) and subsequently transported to neuron for neurite growth and synaptogenesis or for conversion to 24-hydroxycholesterol (24-OHC) by CYP46 (cholesterol 24-hydroxylase). Neurons, upon stimulation or oxidative stress, release fatty acids in ApoE-particles to astrocytes for degradation by mitochondrial β-oxidation, or for storage in LDs. fatty acids and other lipids required for oligodendrocyte-mediated myelination or remyelination are synthesized by both oligodendrocyte and astrocyte, and ApoE particles are potentially involved in astrocyte-to-oligodendrocyte lipid transport. Excessive astrocyte-derived saturated fatty acids induce oligodendrocyte death via lipoapoptosis pathway. Lipids from myelin debris can bind to and activate TREM2 signaling to induce the expression of lipid metabolism genes to facilitate the lipolysis of CE-rich LDs and efflux of lipids. Syk, spleen tyrosine kinase.

In postmortem human brains, increased LDs in the choroid plexus are significantly correlated with worsened cognitive function and AD neuropathology including neuroinflammation [[184]]. Further, such a correlation resembles the phenotypes seen in ApoE knockout mice or hyperlipidemic ApoE4 mice (on high-fat diet). This study further supports the connection between ApoE regulation of brain LD metabolism and AD pathogenesis. In addition to ApoE, other AD risk factor genes such as ABCA1, ABCA7, and PICALM are also involved in the formation of glial LDs [[185]]. Collectively, these studies suggest that impairment in lipid clearance and transport could underlie lipid dyshomeostasis and reduced neurotrophic function observed in aging and AD brains.

Cholesterol transport

Cholesterol in brain is essential for neuronal function due to their roles in membrane fluidity, vesicle formation, and synaptic transmission [[186]]. Different from fatty acid transport, intercellular transport of cholesterol is primarily from glia to neurons (Fig. 4). This mechanism is important for synaptic function and sterol homeostasis because: (a) the cholesterol required for synaptogenesis in neurons is transported from astrocytes because of the suppression of cholesterol biosynthetic pathway in neuron [[187]] and (b) cholesterol needs to be hydroxylated to 24-hydroxycholesterol in neurons before excretion from the brain [[188, 189]]. When cholesterol synthesis in astrocytes is suppressed by conditional SREBP-2 knockout, synaptic, behavioral and motor functions are impaired [[190]]

ApoE is the primary cholesterol transporter in the brain by facilitating its efflux from astrocytes and its uptake by neurons [[191, 192]]. Upon ApoE knockout, cholesterol biosynthesis and hence its levels are reduced in the brain [[193]], and these mice develop age-associated cognitive deficits [[194]]. Further, the depletion of LRP1, a major neuronal receptor for ApoE-containing lipoproteins, induces lipid dysregulation, neuroinflammation and synapse loss [[195]]. In addition to cholesterol, astrocyte-to-neuron ApoE particles are found to carry miRNAs to inhibit neuronal cholesterol biosynthesis, which subsequently elevates neuronal levels of acetyl-CoA (substrate for cholesterol synthesis) and epigenetically regulates the expression of neuronal genes for memory consolidation [[196]]. With regard to the effect of ApoE genotype on these processes, ApoE4 is consistently shown to be associated with reduced capacity in binding, secreting, and transporting cholesterol and other lipids [[197, 198]]. Together, these studies suggest that disrupted intercellular exchanges of fatty acids and cholesterol contribute to lipid dyshomeostasis in AD, and furthermore, they unanimously highlight the detrimental effect of ApoE4 in these mechanisms.

Lipid interaction with AD pathogenic mechanisms

Amyloid metabolism

Aβ peptides are generated from the sequential cleavage of APP by β- and γ-secretases [[199]]. Such a process, also known as the amyloidogenic pathway, is believed to occur within membrane microdomains (lipid rafts) [[200, 201]], which are enriched in cholesterol and sphingolipids [[202]]. It is thus not surprising that cholesterol and sphingolipids affect amyloidogenesis (Fig. 5). Further, multiple surface receptors of the low-density lipoprotein receptor family, such as LRP1, LRP1B, SORL1, and ApoER2 have been connected with APP metabolism, mostly involving their role in endocytic trafficking (reviewed in [[203]]).

Fig. 5

Interactions between lipid metabolism and AD pathogenic mechanisms. Lipid-related factors could contribute to AD pathogenesis via interactions with key mechanisms involved in AD etiology. Inner circle, key AD mechanisms; outer circle, lipid factors affecting each mechanism. Boxes in cyan are lipid species, and boxes in gold represent lipid regulators.

Cholesterol

A large volume of studies suggests that amyloidogenic pathway is regulated by the levels of cholesterol and CE [[17, 204]]. The activity of both β- and γ-secretases is enhanced by high cholesterol and inhibited by low cholesterol [[203]]. The mechanism of cholesterol on Aβ production is recently furthered by superresolution imaging [[153]]. It is shown that suppressed cholesterol synthesis in astrocytes substantially reduces amyloid and tau pathology in a mouse model of AD. This effect is mediated by decreased cholesterol in neurons, which facilitate APP to traffic out of lipid rafts, where β- and γ-secretases are localized. Outside of the lipid rafts, APP is instead converted to the neuroprotective [[205]] and nonamyloidogenic soluble APP-α (sAPP-α) by α-secretase [[153, 206]].

Sphingolipid

In the case of sphingolipid, ceramide is proposed as a key contributor to AD pathophysiology by affecting the production of Aβ and hyperphosphorylation of tau. When sphingomyelinase, the enzyme hydrolyzing sphingomyelin to ceramide, is inhibited, γ-secretase activity and Aβ secretion is reduced [[207]]. This finding is likely related to the effect of ceramide promoting amyloid biogenesis in lipid rafts by stabilizing β-secretase [[208]].

The intimate connection between Aβ generation and cholesterol- and sphingolipid metabolism is further manifested by the regulation in the reverse direction, where Aβ₄₂ and Aβ₄₀ reduces sphingomyelin and cholesterol levels by activating neutral sphingomyelinases and by inhibiting hydroxymethylglutaryl-CoA reductase activity, respectively [[207]]. Likewise, fibrillar Aβ induces sphingomyelinase activity and the subsequent ceramide production via ROS-dependent mechanisms [[207, 209, 210]] (also see the Oxidative Stress section below). These mechanisms are consistent with the reduction in sphingomyelin levels in human AD brains [[52]]. Thus, the bidirectional effect between Aβ production and ceramide levels could form a vicious cycle that exacerbate initial ceramide accumulation in AD brains [[211]]. Upstream of Aβ peptides, APP and γ-secretase are also important regulators of lipid metabolism via modulating the synthesis and turnover of cholesterol, the accumulation of acylglycerols, and the homeostasis of lipoproteins [[212-215]].

Mitochondria-associated membranes

Mitochondria are the powerhouses of the cell and play fundamental roles in processes such as redox homeostasis, apoptosis, steroid biosynthesis, calcium homeostasis, and cell fate regulation [[216, 217]]. Mitochondria perform these functions by forming a dynamic network with close interactions with other components in the cell including the endoplasmic reticulum (ER) [[218-220]]. Intraneuronally, lipid metabolism and Aβ metabolism converge at the mitochondria-associated membranes (MAMs), a unique intracellular interface between the mitochondria and the ER [[221]]. MAMs are enriched in cholesterol, sphingomyelin, and proteins involved in phospholipid exchange, calcium signaling, and redox regulation. Structurally, MAMs are similar to lipid rafts [[221]]. Accumulating evidence has established the essential role of MAMs in lipid synthesis and trafficking [[41, 222]]. MAM regulation of Aβ production is suggested by the enriched β- and γ-secretases and elevated β-secretase cleaved product of APP (C99) localized to MAMs [[223, 224]]. Intriguingly, this mechanism is connected to reduced mitochondrial function, via a cascade initiated from increased C99 to enhanced sphingomyelin hydrolysis by sphingomyelinases and to increased ceramides [[223, 225]]. Besides C99, Aβ peptides are also known to directly modulate the sphingomyelinase-ceramide pathway [[210, 226]].

Conversely, fatty acids could directly affect axonal Aβ generation by regulating palmitoylation and enhancing MAM activity [[227]]. Palmitoylated APP is enriched in lipid rafts including MAMs, where it serves as a preferred substrate towards β-secretase-mediated cleavage and amyloidogenesis; upon inhibition of ACAT, an enzyme that transfers fatty acyl groups, the levels of both palmitoylated APP and Aβ are substantially reduced [[228]]. These findings are consistent with the upregulated MAM activity, disrupted phospholipid synthesis, and accumulated neutral lipids in AD- and ApoE40-positive human cells [[229, 230]].

Gangliosides

Gangliosides are a group of complex glycosphingolipids with one or more sialic acids connected to the sugar chain and are abundant in the rafts of plasma and organelle membranes [[231]]. GM1, a ganglioside that contains one sialic acid residue is known to act as a seed for Aβ binding (by forming GM1‑bound Aβ) and aggregation, leading to the formation of amyloid fibrils [[232]]. In this process, GM1 clusters are formed because of disrupted endosomal function or lipid metabolism, which interact with Aβ and lead to the sequential formation of preamyloid β-sheet-rich oligomers and tape-like amyloid fibrils [[233]].

ApoE

ApoE is involved in almost all aspects of Aβ pathology, from its production, aggregation, and deposition, to its degradation and clearance (reviewed in [[234]]). Functionally, ApoE isoforms differentially affect the size, composition, and extent of lipidation of ApoE particles. Compared with ApoE3 group, ApoE in ApoE4 carriers is less lipidated while ApoE in ApoE2 carriers is more lipidated [[235, 236]]. Such isoform-dependent lipidation status also pertains to ApoE particles in the CSF: those in E2/E3 subjects have larger sizes, whereas those in E3/E4 subjects have smaller sizes, relative to E3/E3 controls [[237]]. Further, the smaller size and lower lipidation rate of ApoE4 particles are associated with reduced receptor binding ability and resultantly, compromised lipid carrying and trafficking capacity [[160, 198, 236]].

Compared with cognitively normal individuals, those with MCI have higher CSF levels of lipid-depleted (non-lipidated) Aβ, which is less likely to be cleared by enzymatic degradation, phagocytosis, or BBB transport out of the brain [[234, 238]]. Moreover, the lipidation status of CSF Aβ is modified by ApoE genotype, with ApoE4 carriers exhibiting reduced lipidation of both Aβ and ApoE itself, which could be relevant to the differential stability of ApoE/Aβ complexes [[239]]. Further, the lipidation status of ApoE and Aβ in CSF is reduced after a 4-week high saturated fat and high glycemic dietary intervention, whereas the same parameters are increased by diet low in saturated fat and glycemic index [[238]]. This study corroborates the close relationships among AD pathology, ApoE status and lipid profile. Consistent with findings in human CSF, a higher deposition of Aβ in ApoE4 mouse brains is associated with a longer half-life of Aβ [[240]], highlighting the central role of ApoE in Aβ clearance.

Two competitive bindings among ApoE, lipids, and Aβ have been proposed [[234]]: (a) Aβ may compete with lipids for ApoE binding, and (b) ApoE may compete with Aβ for their common cell surface receptors. Therefore, it can be speculated that disruptions to lipid metabolism could affect Aβ metabolism, and vice versa. In the case of ApoE4, its reduced lipidation and transport efficiency could affect both Aβ clearance and intercellular lipid exchange and subsequent degradation. Further, the potential competition between accumulated Aβ and uncleared lipids for lipid carriers could further exacerbate the lagged clearance of both species.

Tau

Relative to the intimate and bidirectional relationship between Aβ and lipids, the interaction between tau and lipids is less understood, and existing studies suggest their connections are be primarily indirect – in many instances through Aβ [[17]]. This is likely related to the fact that tau is a cytosolic microtubule-associated protein, and their phosphorylation and aggregation do not occur at lipid-rich membrane structures. Moreover, as tau pathology is primarily intraneuronal, glia-centered mechanisms of lipid synthesis and catabolism are unlikely to be directly affected by tau. Nevertheless, several studies provide initial evidence that brain cholesterol accumulation and high dietary cholesterol are associated with increased tau phosphorylation and aggregation [[241, 242]] (Fig. 5). Intraneuronally, the lipid-rich ER-mitochondria interaction site (MAMs) is also regulated by tau [[243]]. Additionally, a significant association between T2D and CSF tau levels is detected in ApoE4 carriers but not in noncarriers [[244]].

Neuroinflammation

Neuroinflammation is a key element in the pathogenesis and progression of AD, signatured by increased pro-inflammatory cytokines, decreased anti-inflammatory responses, and amplified reactivity of glial cells including microglia and astrocytes [[245]]. The immunomodulatory properties of lipids, in particular fatty acids and their derivatives, were discovered decades ago. Depending on the class of the lipids and the cell type in which they are located, both pro- and anti-inflammatory responses can be induced [[246]].

Specialized pro-resolution mediators

Lipid metabolism can be involved in modulating neuroinflammation via a class of lipid-derived signaling molecules called specialized pro-resolution mediators (SPMs). SPMs, including resolvins, lipoxins, protectins, and maresins, are produced from PUFA (primarily ω-3 PUFAs such as DHA and eicosapentaenoic acid (EPA)) and orchestrate the resolution of acute responses by activating cellular events that repress inflammatory processes and restore tissue homeostasis [[247]] (Fig. 1). This mechanism is used by both the periphery and the CNS to avoid persistent, chronic inflammation. In AD, mounting evidence reveals an impairment of the resolution of inflammation, which contributes to chronic neuroinflammation and exacerbated AD pathology [[248]]. In human AD brains, multiple SPMs are decreased in AD-affected brain regions compared with controls, whereas the protein expression of SPM receptors is upregulated [[249, 250]]. Importantly, a positive correlation exists between CSF levels of SPMs (LXA4 and RvD1) and cognitive function across control- and AD cases [[251]]. In AD mouse models, SPM administration not only restores the decline in hippocampal SPM levels but also reduces pro-inflammatory cytokines and glial reactivity and alleviates Aβ and tau pathology [[252, 253]]. Lipid metabolism is also linked with neuroinflammation via the expression of anti-inflammatory genes induced by lipid-sensing nuclear receptors such as liver X receptors (LXRs) and PPARs [[254, 255]].

Prostaglandins

In contrast to the pro-resolution SPMs, PUFAs (primarily AA generated from PLA2-catalyzed phospholipid hydrolysis) are also substrates of pro-inflammatory prostaglandins, particularly prostaglandin E2 (PGE₂), through sequential reactions catalyzed by cyclooxygenase-1 (COX-1) and COX-2, and prostaglandin E synthases [[256]] (Fig. 1). CSF levels of PGE₂ is elevated in early-, but not late-, stage of AD [[257, 258]]. In fAD mouse models, deletion of PGE₂ receptors alleviates neuroinflammation, oxidative stress, Aβ deposition, and memory deficit via microglia-centered mechanisms [[259-261]]. PGE₂, via its EP2 receptor, promotes microglial glucose sequestration and mitochondrial bioenergetic deficits. By blocking microglial EP2 signaling, cellular energy metabolism, brain inflammatory state, and synaptic function are restored [[262]]. These findings are consistent with retrospective studies supporting the effect of long-term use of non-steroidal anti-inflammatory agents (NSAIDs, inhibitors of cyclooxygenases) on reducing AD risks [[263-266]], although such a beneficial effect is not observed in randomized trials [[267, 268]]. Notably, pro-inflammatory fatty acids generated from phospholipid hydrolysis are also involved in Aβ-mediated neurotoxicity. cPLA2 activity and the production of AA can be dose-dependently activated by Aβ in neuronal cultures, whereas cPLA2 inhibition diminishes Aβ-induced neurotoxicity and protects fAD mice from memory and learning deficits [[269]]. Spatially, lipid rafts may play a critical role in initiating inflammatory response in glial cells involving cholesterol- and sphingolipid metabolism, where high cholesterol levels are associated with clustering of inflammatory proteins and elevated inflammation [[270]].

Collectively, lipid metabolism, via the production of both pro-inflammatory and pro-resolving lipid-derived messengers, contributes to the chronic and unresolved neuroinflammation in AD (Fig. 5).

Oxidative stress

Oxidative stress refers to oxidative damage to cell constituents caused by accumulation of ROS and reactive nitrogen species (RNS) that overwhelm cellular antioxidative capacity [[217, 271]]. Increased generation of ROS and oxidative stress is a hallmark of AD and contributes to the neurotoxic environment with disease progression [[16]]. Compared with glial cells, neurons are particularly vulnerable to oxidized redox environment becausse of their limited antioxidative capacity.

Oxidative stress in brain is largely manifested by lipid peroxidation, owning to its high oxidation-prone lipid content. In AD brains, amyloid plaques and their immediate surroundings are characterized by the presence of oxidized lipids, whereas in cognitively normal subjects, the levels of oxidized lipids are lower even in the presence of plaques [[272]]. The interactions between lipids and oxidative stress in AD are three-fold. First, some lipid species are known as ROS inducers. In addition to the direct effect on Aβ, ceramides promote ROS generation in AD [[273]]. Second, selective lipids, primarily PUFAs are prone to lipid peroxidation to generate highly reactive electrophilic aldehydes including 4-hydroxy-2-nonenal (4-HNE), acrolein, and malondialdehyde [[274]]. Accordingly, markers for lipid peroxidation are increased early in AD brains along with reduced PUFA levels [[275-277]], and ApoE4 is associated with enhanced lipid peroxidation [[278]]. Last, ROS such as H₂O₂ can directly [[279]] or indirectly (via cPLA2 and AA) [[280]] activate sphingomyelinases, which catalyze the hydrolysis of sphingomyelin to generate ceramide. Consistently, protein levels of acid sphingomyelinase are increased in the cortex of AD patients and fAD mice [[52, 281, 282]]. In fAD mouse models, Aβ deposition and memory impairment are improved after genetic- or pharmacological reduction of acid- or neutral sphingomyelinases [[282, 283]]. These three mechanisms constitute the interplays between oxidative stress and lipid dyshomeostasis implicated in AD pathogenesis (Fig. 5).

Myelin degeneration and white matter integrity

Around 40% of the human brain contains white matter that consists of densely packed fibers, and myelin is the major component of these structures [[284]]. In the CNS, myelin sheath is a plasma membrane extension of oligodendrocyte that wraps around neuronal axons to facilitate rapid nerve conduction and efficient propagation of action potential [[285]]. About 80% of the dry weight of myelin is lipid [[286]]. Accumulation of fatty acids and glycosphingolipids are coupled with myelin dysfunction in multiple inherited lysosomal and peroxisomal diseases [[287]]. In addition to their role as key inducer of multiple sclerosis, white matter loss and demyelination is also observed in AD brains [[288]]. White matter hyperintensities are correlated with CSF levels of Aβ and tau in individuals with MCI or preclinical AD [[289, 290]].

Glycolipids and phospholipids together comprise the majority of myelin membrane lipids, and both classes use fatty acids as structural blocks for their synthesis. Altered fatty acid metabolism could thus contribute to AD-related structural and functional abnormalities of white matter. It is recently reported that fatty acid synthesis in oligodendrocytes is essential for myelination with correct lipid composition [[291]] (Fig. 5). Further, lipid biosynthesis by astrocytes is also required for myelination via astrocyte-to-oligodendrocyte transport [[292]], suggesting complementary roles of these two types of glia in supplying lipid building blocks for myelination (Fig. 4). While astrocyte-derived fatty acids are important for myelination, excess saturated fatty acids released from reactive astrocytes also mediate the toxic effect that induces oligodendrocyte death [[140]].

Besides the interaction with Aβ [[293]], TREM2 can also be bound by lipid species and lipoproteins [[102-104]]. Remarkably, the AD-related R47H mutation impairs its lipid-detecting function [[102]]. More recently, there has been convincing evidence on the direct role of TREM2 in orchestrating myelin debris clearance and remyelination by regulating cholesterol esterification and LD metabolism in microglia (Fig. 4), where TREM2 deficiency elicits dysregulatied lipid metabolism and transport [[294, 295]]. These findings suggest cholesterol metabolism and lipid signaling as key players in TREM2 polymorphism-associated myelin degeneration and increased AD risk.

Blood-brain barrier (BBB) integrity

The BBB plays an essential role in both separating the brain from, and connecting the brain to, the peripheral circulation by forming an interface for transcytotic exchange of substances including lipids [[296]]. BBB dysfunction is manifested in early stages of AD irrespective of changes in Aβ or tau biomarkers [[297, 298]]. Aβ endothelial transcytosis and clearance across the BBB is mediated by lipid related genes PICALM and LRP1 [[109]] (Fig. 5). While it remains unclear how lipid metabolism contributes to disrupted BBB integrity in AD, recent studies suggest that lipid profile and metabolic status in both the brain and the periphery modulate BBB function. Mfsd2a, a phospholipid flippase and transporter expressed by BBB endothelial cells, suppresses transcytosis and ensures BBB integrity [[299]]. It is discovered that, as a major DHA transporter across BBB, Mfsd2a increases DHA-containing phospholipids in CNS endothelial cells, which inhibit the caveolae-mediated transcytosis and maintains BBB integrity [[300]]. This is consistent with (a) the increased BBB permeability in AD, (b) lowered DHA levels in AD brains, and (c) DHA-free diet accelerates cognitive decline in AD mice [[301]]. In another study focusing on the effect of peripheral lipid profile, obesity-associated high circulating saturated fatty acids lead to elevated CSF levels of palmitate and increased BBB permeability, and high brain levels of palmitate further elicit neuroinflammation and impair synaptic and cognitive functions [[302-304]].

Targeting lipid metabolism against AD: therapeutic potential

Activating lipid-sensitive nuclear receptors

Retinoid X receptors (RXRs), LXRs and PPARs belong to the nuclear receptor superfamily, and they are the master regulators of both peripheral and central lipid homeostasis by transactivating genes involved in lipid metabolism [[305]]. Multiple therapeutic strategies for AD have been investigated by targeting these receptors.

LXR/RXR

LXRs, including LXRα and LXRβ, form heterodimers with RXRs, and their primary ligands are oxysterols, products of cholesterol hydroxylation [[305]]. Natural RXR ligands include retinoic acid, linoleic acid, linolenic acid, and DHA. Besides LXR, RXR can form dimers with itself, or with PPARs and retinoic acid receptors (RARs) [[306]]. LXR functions as a sensor of intracellular cholesterol levels and promotes the transcription of genes involved in cholesterol efflux, including ApoE [[307]]. LXRβ is the major LXR form expressed in brain, 3- to 5-fold higher than LXRα [[308]]. LXRβ knockout mice exhibit neuronal loss and impaired motor coordination, accompanied by lipid accumulation, astrogliosis, and axonal atrophy [[309]].

The LXR/RXR dimer can be activated by either LXR or RXR ligands. LXR agonists (e.g., GW3965 and T0901317) promote Aβ clearance, attenuate neuroinflammation, and mitigate cognitive impairments in fAD mice, and these effects are primarily mediated by ABCA1 and ApoE [[310, 311]]. Bexarotene, an RXR agonist and a drug for cutaneous T-cell lymphoma, also stimulates the clearance of Aβ peptides and improves cognitive deficit [[312]] by upregulating ABCA1 activity and reversing the hypolipidation of ApoE4 [[313]] in AD mouse models. However, clinical studies of Bexarotene suggest no effect on Aβ metabolism or cognitive function [[314, 315]]. Interestingly, animal studies reported a sexual dimorphism of Bexarotene in modulating synaptic function in a fAD mouse model carrying five mutations to APP and PSEN1 (5xFAD), with male mice more responsive than female [[316]].

PPARγ

PPARs are initially discovered for their role in promoting peroxisomal proliferation before they are regarded as major regulators of lipid- and carbohydrate metabolism [[317]]. PPARγ is the most investigated and targeted PPAR in AD research. In peripheral tissues, PPARγ regulates lipogenesis and insulin sensitization [[318]]. Earlier studies have connected the effect of PPARγ agonists against AD-like pathologies with anti-inflammatory mechanisms including the suppression of pro-inflammatory transcription factors such as NFκB and STATs [[246, 319]], further underlining the intimate connection between lipid metabolism and inflammatory response.

In animal models, PPARγ agonists attenuate Aβ pathology and improve cognitive function by modulating both the production and clearance of Aβ [[320-326]]. In human studies, despite initial favorable outcomes from trials with PPARγ agonist pioglitazone [[327, 328]], recent report on a large-scale phase III trial in people at risk of AD shows that pioglitazone fails to delay the onset of MCI [[329]]. A similar scenario applies to the clinical evaluation of another PPARγ agonist of the thiazolidinedione (TZD) class, rosiglitazone, for which predominantly negative results are obtained from larger trials, whether rosiglitazone is used alone [[330-332]] or as adjunctive therapy [[333]]. Notably, a phase II trial in patients with mild to moderate AD finds that rosiglitazone, at a higher dose, can selectively improve cognition in ApoE4 noncarriers but not ApoE4 carriers [[334]]. This interesting interaction between ApoE4 and a lipid modulator further highlight the critical role of lipid metabolism in AD pathogenesis and susceptibility.

Since existing PPARγ agonists have a low to moderate ability to cross the BBB, the question arises as to whether next generation PPARγ agonists with improved brain penetration can elicit improved efficacy. It is also plausible that many of the beneficial effects seen in animal models and early phase trials are relevant to the insulin sensitizing and mitochondria promoting properties of TZD signaling [[335]], whereas the activation of lipogenic pathway downstream of PPARγ is not an effective therapeutic target. This speculation is consistent with the increased expression of PPARγ mRNA in the frontal cortex of human AD brains [[166]]. Given the accumulation of many toxic lipid species and reduced lipid catabolism in AD brains, the activation of PPARα, which promotes lipid degradation, may represent a better strategy to intervene lipid dysregulation in AD (see section “Reducing lipid accumulations” below).

Reducing lipid accumulations

PPARα controls the expression of genes encoding key enzymes of fatty acid oxidation, such as carnitine palmitoyl transferases (CPTs) and acyl-CoA oxidase [[336]]. PPARα is highly expressed in tissues with higher fatty acid oxidation capacity such as liver, heart, kidney, and skeletal muscle. In the brain, PPARα is predominantly expressed in astrocytes [[167, 337]], which is consistent with the unique capability of astrocytes in performing fatty acid oxidation [[29]]. Beyond their role in astrocytes, PPARα in neurons maintains spatial learning and memory by regulating the CREB pathway [[338]].

In the context of AD pathology, PPARα modulates amyloid metabolism by activating the non-amyloidogenic α-secretase while inhibiting the amyloidogenic β-secretase [[339, 340]]. While the pathophysiological role of PPARα in neurodegeneration remains elusive, initial interventional studies in rodent models suggest that endogenous ligands or exogenous agonists of PPARα promote synaptic function of hippocampal neurons and protect them from oxidative damage and Aβ toxicity [[341, 342]]. In vivo, PPARα agonist Wy-14643 alleviates AD-like pathologies and memory deficits in the APP/PS1 mice [[343, 344]]. Further, PPARα agonists elicit anti-inflammatory effects and promote survival in a mouse model of amyotrophic lateral sclerosis (ALS) [[345]], suggesting a common mechanistic role of PPARα across neurodegenerative diseases.

Acetyl-l-carnitine (ALCAR) is an ester of the trimethylated amino acid l-carnitine, and it can be metabolized to acetyl-CoA and carnitine, which can serve as tricarboxylic acid (TCA) cycle fuels, and facilitators of fatty acid transport into mitochondria, respectively [[346]]. Serum levels of ALCAR and CSF levels of l-carnitine are decreased in AD [[129, 347]]. In primary neurons, ALCAR ameliorates Aβ-oligomer-evoked changes in mitochondrial respiration, fragmentation, and movement [[348]]. Animal studies show that ALCAR improves age-associated mitochondrial dysfunction and cognitive deficit [[349, 350]]. However, meta-analyses of AD clinical trials with ALCAR yield conflicting and inconclusive results regarding the effect of ALCAR on cognitive measures [[351, 352]].

Besides promoting lipid degradation using PPARα agonists, another strategy to prevent the accumulation of toxic fatty acids, amyloidogenic ceramides, and pro-inflammatory prostaglandins is to inhibit the de novo synthesis of fatty acid in the brain. It is found recently that an inhibition of FAS in fAD mice alleviates lipid peroxidation, neuroinflammation, and cognitive loss [[149]]. In line with this report, another compound that reduces brain inflammation and cognitive deficit in fAD mice, CAD-31, elicits its protective effect by simultaneously promoting fatty acid oxidation and inhibiting fatty acid synthesis [[151]].

Boosting lipid transporters

ApoE and ABCA1 are the major regulators of lipid transport in the brain. Multiple in vitro studies support the potential of using ApoE4 structure correctors to mitigate ApoE4-induced pathological effects [[353, 354]]. In human iPSC-derived neurons, the compound PH002 is able to decrease ApoE4 degradative fragments, Aβ production, and tau hyperphosphorylation in a dose-dependent manner [[353]]. Future in vivo studies are needed to determine the efficacy of this ApoE4 structure-correcting approach in rescuing ApoE4-involving AD pathologies including the disrupted lipid transport.

Given the gain-of-toxic function of ApoE4 in AD, strategies aiming to reduce its expression are also being evaluated. At the RNA level, knockdown of ApoE4 in the APP/PS1 mice by antisense oligonucleotides (ASO) decreases neurite dystrophy independent of Aβ pathology, and this strategy is only effective if initiated during the seeding stage before fibril formation [[355]]. At the protein level, neutralizing antibodies that target nonlipidated and aggregated ApoE reduce Aβ deposition in the APP/PS1/ApoE4 mice [[356]]. A more direct approach to facilitate ApoE lipidation and lipid transport is to increase ABCA1 activity (see review [[236]]). In mice, intraperitoneal injection of an ABCA1 agonist, CS-6253, increases ApoE4 lipidation and decreases ApoE4-driven amyloid and tau pathologies [[357]]. While the primary outcome for these ApoE- or ABCA1-targeting studies is Aβ pathology, it would be interesting to investigate whether lipid metabolism and trafficking is involved.

Supplementing PUFAs

In distinction to approaches that intervene lipid metabolism by manipulating lipid enzymes or sensors, lipids that decline in AD and are prone to peroxidation such as PUFAs, have been assessed as AD treatment- or prevention strategies. Epidemiologically, increased intake of ω-3 PUFA, especially DHA and EPA, reduces AD risk, whereas lower ω-3 PUFA intake increases the risk [[58, 358]].

Mechanistically, ω-3 PUFA affects multiple aspects of AD pathologies. DHA and EPA decrease Aβ production while increase Aβ degradation by reducing β- and γ-secretase activity and stimulating the activity of insulin-degrading enzyme (IDE), respectively [[359, 360]]. ω-3 PUFA modulates brain inflammatory response to Aβ [[361]]. In addition to the direct effect of ω-3 PUFA on Aβ metabolism, some of the beneficial effects of ω-3 PUFA are contributed by its role as an activator of RXR and PPARs [[362-364]]. Clinically, small scale studies suggest potential beneficial effects of ω-3 PUFA on pathological outcomes in MCI or AD patients [[365, 366]]. However, it should be noted that in many instances of these studies, ω-3 PUFA is supplemented in combination with other compounds such as antioxidants. It thus remains to be determined whether the observed effects are induced by ω-3 PUFA, by other ingredients, or by multiple ingredients combined. The last possibility pointing to a synergistic effect of ω-3 PUFA and antioxidants is supported by a report in which oxidized DHA, even at low levels, abrogates the protective effect of DHA on Aβ production [[367]].

In two larger scale, randomized clinical trials, DHA fails to delay the cognitive decline in mild to moderate AD subjects [[368, 369]]. Nevertheless, subgroup analysis in these two trials detected positive effect of DHA in very mild AD cases [[368]] and in ApoE4 noncarriers [[369]], respectively. These interesting findings corroborate the critical role of disrupted lipid metabolism in early AD stages, as well as its close interaction with ApoE functionality. Notably, ω-3 fatty acid supplementation to MCI patients increases the levels of a SPM (RvD1) in their macrophage coupled with increased Aβ phagocytosis and decreased cytokine expression [[365]], suggesting some of the beneficial effect of ω-3 fatty acids might be mediated by the resolution of neuroinflammation via SPMs [[365]] (see section “Neuroinflammation” above).

Replenishing ketogenic fuels

Ketone bodies, such as acetoacetate and β-hydroxybutyrate, are mainly derived from fatty acid oxidation in liver upon starvation and represent the major non-glucose fuel source to the brain [[370]]. An adaptive metabolic switch to ketones in the brains of normal chronological aging, female reproductive aging, and AD models have been observed [[371-375]]. Since glucose hypometabolism is an early and persistent hallmark of AD, it is proposed that increased ketone body availability to the brain may promote its resistance to bioenergetic deficits [[376]]. Further, β-hydroxybutyrate exhibits antioxidative and anti-inflammatory effects in the brain [[377]]. How ketone body interventions affect brain lipid profiles and whether lipid metabolism is involved in the CNS effect of ketone bodies, warrant further investigation.

In both aging and AD animal models, ketogenic diets (high saturated fat and low carbohydrate) markedly attenuate Aβ pathology and improve brain metabolic and cognitive function [[378, 379]]. However, despite initial indicators of efficacy for ketogenic diet/formula in human trials [[380-382]], daily administration of a ketogenic medium-chain TAG (AC-1204) fails to alter cognitive function in a phase III trial with mild-to-moderate AD subjects. Interestingly, subgroup analyses of two trials of ketogenic intervention suggest that its cognition-improving effect is more notable in early stage AD patients with Mini-Mental State Examination (MMSE) score ≥ 15 [[383]], or ApoE4 noncarriers [[384, 385]]. These modifications of responsiveness by ApoE4 and disease stage resemble the findings in DHA interventions [[368, 369]].

Taken together, while ketogenic interventions are effective in animal models and early phase clinical trials, its long-term effect in a broader AD population warrants further investigation. Moreover, the effectiveness of ketogenic strategies against AD is dependent on the individual metabolic status and lipid profile, which are markedly modified by ApoE genotype and disease stage as discussed in previous sections.

Restricting cholesterol levels

While retro/prospective, observational analyses suggest a protection of the use of cholesterol-lowering statins in reducing AD risk [[386, 387]], randomized clinical trials find no benefit of statin treatment in AD patients [[388-391]]. In several reports, statin-associated short-term and reversible cognitive impairment are observed [[392-394]]. Schultz and colleagues ascribed such discrepancy to the independent effect of cholesterol in the central- and peripheral systems [[395]]. It is proposed that the risk-lowering effect of statins primarily comes from the reduced circulating cholesterol, which benefits cardiovascular and cerebrovascular systems and reduces vascular inflammation [[395]]. The negative impact of high-dose statins on cognition may stem from the decreased local cholesterol synthesis within the CNS (as seen in ApoE knockout mice [[193]]), where cholesterol participates in essential processes including synaptogenesis, myelination, and neurotransmission [[395]]. It is also plausible that the impact of statins on AD risk is in part contributed by its effect on vascular dementia, which is often misdiagnosed as AD [[395]]. Recent analyses of the UK Biobank database suggest that potential beneficial effect of statins is selective for ApoE4 carriers [[386, 396]], which is in agreement with ApoE4-associated central- and systemic cholesterol dyshomeostasis. In addition to ApoE4 status, age [[387]] and sex [[396]] are also factors that modify AD risk profiles in statin users. These findings support the necessity of a precise and personized approach when evaluating the efficacy of cholesterol-lowering medicine for AD.

Besides the idea of reducing cholesterol biosynthesis by statins, studies in AD animal models indicate potential effect by modulating the levels of cholesterol derivatives. For example, boosted conversion of cholesterol to the brain-exportable 24S-hydroxycholesterol by expressing exogenous cholesterol 24-hydroxylase reduces Aβ pathology and improves spatial memory [[397]]. A pharmacological approach using Efavirenz, an FDA-approved anti-retroviral drug, to activate cholesterol 24-hydroxylase results in similar effect in the 5xFAD mice [[398]]. Alternatively, suppression of CE production by genetic knockdown or pharmacological inhibition of ACAT1 attenuates Aβ deposition and restores cognition in fAD mice [[399-402]].

Conclusions and perspectives

Large volumes of clinical evidence have clearly demonstrated the alterations in lipid species and lipid metabolism in the pathogenesis and progression of AD. It is noteworthy that many lipid-related changes emerge in early stages of the disease. This suggests that lipid dyshomeostasis may be an initiation mechanism of the disease, likely via interactions with amyloidogenesis, synaptogenesis and hypometabolism, thus highlighting the importance of early intervention for lipid-centric strategies towards improved efficacy.

Further, many of these lipid-related changes persist or exacerbate through later disease stages. These changes could be relevant to the critical interactions between lipid species and mechanisms such as bioenergetics, neuroinflammation, and oxidative stress, which potentially form a vicious cycle promoting neuronal loss and cognitive dysfunction. These lines of evidence indicate that in neurodegeneration, lipid metabolism could serve as a bidirectional link between bioenergetic decline and chronic neuroinflammation [[403]], both of which are hallmarks of AD [[16]]. Moreover, given the intimate and strong mechanistic connections among Aβ, ApoE and lipid trafficking, lipid metabolism could be the hub where Aβ-dependent and -independent mechanisms of ApoE-associated pathologies converge.

Several lessons are learned from the successful preclinical studies and failed human clinical trials of lipid-targeting AD therapies.

First, the efficacy of lipid-related therapeutics is substantially affected by genetic modifiers of AD risk, with ApoE status being the most significant one. It appears that fatty acid-centric therapeutics are more, and in some cases only, effective in ApoE4 noncarriers, such as PPAR agonists, DHA supplement and ketogenic intervention, whereas cholesterol targeting strategies such as statins are more selective for ApoE4 carriers. It is plausible that for ApoE4 carriers, fatty acid-related lipid dyshomeostasis is too severe to be corrected. In agreement with this speculation, diabetes worsens cognitive decline in ApoE3 and ApoE2 carriers, but not ApoE4 carriers [[404]]. Such an ApoE genotype-specific effect is in part mediated by vascular impairment, for which ApoE4 carriers have reached a plateau regardless of diabetic status [[404]]. A strong interaction with ApoE genotype also pertains to lifestyle-focusing strategies to reduce AD risk. It is estimated that improvements in modifiable risk factors can prevent or delay 30% to 40% of dementia cases [[133, 405]]. However, a recent study suggests that the responsiveness to improvements to modifiable-risk factors (smoking, depression, diabetes, physical activity, social isolation, and healthy diet) is detectable in ApoE4 noncarriers only [[406]].

Second, emerging evidence suggests that many of the neuroprotective effect of lipid-modulating approaches are relevant to their indirect effect on peripheral metabolism. In addition to the ApoE status, metabolic- and vascular profile both modulate brain lipid metabolism and modify the responsiveness to lipid-modulating interventions. For example, PPARγ agonists are more effective in the AD population with metabolic disorders, whereas statins are more beneficial for those with hyperlipidemia, vascular abnormalities, and increased stroke risks. A personalized approach incorporating these interacting factors should be considered when intervening lipid metabolism against AD.

Third, developing therapeutics by targeting lipid metabolism should consider brain selectivity and peripheral side effects. For example, nuclear receptor (LXR/RXR/PPAR) agonists are known to elicit unwanted disruption to lipid synthesis and homeostasis in metabolic organs such as the liver and adipose tissue, which may complicate their direct effect on the brain.

Collectively, findings from large-scale populational studies and responsiveness analyses to investigational therapies in the past two decades emphasize the necessity to incorporate individual risk profile, especially genetic- and non-genetic factors that affect metabolic and vascular functions, when intervening lipid metabolism against AD. Further, successful lipid-targeting therapeutics should consider the dynamic changes and adaptations of lipid metabolism throughout the prodromal, early, and late phases of the disease.

Acknowledgements

This work has been supported by the National Institute on Aging (NIA) grants RF1AG068175 to FY, P01AG026572 (Project 1 and Analytic Core to FY), Arizona Alzheimer’s Consortium Pilot Project grants to FY, and the Packer-Wenz research endowment to FY. Figures were created with BioRender.com.

Conflict of interest

The author declares no conflict of interest.

Author contribution

FY conceptualized the scope, wrote the manuscript, and prepared the figures.

Open Research

Data availability statement

Data sharing is not applicable to this review article as no new data were created or analyzed.