Apolipoprotein E, amyloid-ß clearance and therapeutic opportunities in Alzheimer's disease

Alzheimer's disease (AD) is a progressive neurodegenerative disease characterised by extracellular amyloid-ß (Aß) and intraneuronal tau protein brain pathologies. The most significant risk factor for non-familial AD is the presence of the E4 isoform of the cholesterol transporter apolipoprotein E (apoE). Despite extensive basic research, the exact role of apoE in disease aetiology remains unclear. Correspondingly, therapeutic targeting of apoE in AD is at an early preclinical stage. In this review, I discuss the key interactions of apoE and Aß pathology, the current progress of preclinical animal models and the caveats of existing therapeutic approaches targeting apoE. Finally, novel Alzheimer's genetics and Aß-independent disease mechanisms are highlighted.


Introduction
Alzheimer's disease (AD) is the most common cause of dementia in aged populations, being characterized by cerebrovascular and neuronal dysfunctions that induce a progressive decline in cognitive functions [1]. Th e occurrence of AD in individuals aged over 65 years is defi ned as late-onset AD (LOAD) -representing the majority of AD suff erers. Patients with early-onset AD (EOAD) represent approximately 1% of the overall population [2].
Symptomatic AD is diagnosed clinically using a battery of cognitive tests, with signifi cant eff orts ongoing to move diagnosis to earlier disease stages using the additional tools of genetic testing, blood and cerebrospinal fl uid biomarkers and neuroimaging [3]. Previous to these advances, however, AD could only be defi nitively diagnosed as the cause of dementia by post-mortem detection of two major neuropathologies. Th ese comprise senile plaques of aggregated Aβ peptide, and neurofi brillary tangles of hyperphosphorylated, aggregated tau protein.

Amyloid-β
Aβ peptides are produced through sequential proteolysis of the amyloid precursor protein (APP) by β-secretase/ BACE and the γ-secretase complex (partly comprising the presenilin PS1 or PS2). Aβ peptides vary in length from 39 to 43 amino acids with the predominant species being Aβ40 and Aβ42 [4]. Disease-modifying AD drug discovery research has focused on strategies targeting production or clearance of the Aβ peptide. Th is 'amyloid hypothesis' has been driven by the fact that familial EOAD with autosomal dominant inheritance is caused by mutations in the APP, PS1 or PS2 genes. In simple terms, the net eff ect of these mutations is to increase either bulk Aβ levels or the ratio of Aβ42:Aβ40 production [5]. An increase in brain Aβ42 levels, whether absolute or ratiometric, is hence critical to the aetiology of familial EOAD.
In agreement with the amyloid hypothesis, studies in transgenic mouse models of AD imply a cascade of events in which abnormal forms of tau act as downstream mediators of Aβ toxicity [6,7]. Contrary to this proposed cascade, however, whilst neuronal loss and neurofi brillary tangle counts strongly predict cognitive status in LOAD patients, total Aβ plaque load correlates weakly with cognitive impairment [8]. Th e prevalent explanation for this disparity is that it is diff usible Aβ oligomers, rather than Aβ plaques, that represent the actual toxic species. Th e E693Δ APP mutation, for example, causes Alzheimer's-type dementia through the toxicity of nonfi brillar, intracellular Aβ oligomers [9]. Conversely, the ' Arctic' APP mutation (E693G) induces formation of large Aβ oligomers known as protofi brils [10]. Experimental disagreement over the physicochemical nature of toxic oligomers in LOAD has hampered delineation of their exact role in disease [11].

Apolipoprotein E
Apolipoprotein E (apoE) is the primary transporter of cholesterol in the central nervous system (CNS), being synthesised within the blood brain barrier (BBB) pre dominantly by astrocytes [12]. Th ree apoE polymorphic alleles (APOE2, APOE3, and APOE4) encode three protein isoforms (apoE2, apoE3 and apoE4) that diff er by cysteine/arginine polymorphisms at position 112 or 158. Th e APOE4 allele, found in 15% of the population, remains the most signifi cant genetic risk factor for LOAD [13].
In support of the amyloid hypothesis, APOE4 carrier status is associated with greater Aβ plaque load in both AD patients and cognitively normal individuals [14,15]. Th e APOE4 allele also correlates with increased cerebrovascular Aβ deposition [16] and, correspondingly, is a risk factor for cerebral amyloid angiopathy [17]. As a consequence, research into the mechanistic connection between apoE4 and LOAD has focused on delineating the interaction of apoE with Aβ pathology (Figure 1). Experimental data now support a clear and necessary role for apoE in Aβ toxicity.

Interactions of ApoE with Aβ pathology
In vitro studies have demonstrated that apoE4 more than apoE3 interacts directly with Aβ [18], enhancing Aβ fi brillisation [19]. Interpretation of such data is complicated by the diffi culties of replicating in vivo Aβ conformation and apoE lipidation status. However, early Aβ amyloidosis mouse model data also support a clear role for apoE in Aβ pathology [20]. As a consequence of these fi ndings, apoE/Aβ interaction inhibitors are being developed as AD therapeutics. Small Aβ-mimetic peptides intially demonstrated reductions in apoEstimulated formation of neurotoxic Aβ aggregates in vitro [21], with these data being subsequently confi rmed in vivo using a mouse model of Aβ brain amyloidosis [22].
ApoE proteins comprise an amino-terminal receptorinteracting domain and carboxy-terminal lipid-binding domain. Fluorescence lifetime imaging-fl uorescence resonance energy transfer (FLIM-FRET) studies on human post-mortem tissue sections indicate that Aβ is preferentially associated with the carboxyl terminus of apoE4 compared to that of apoE3, and that apoE4 undergoes greater amino-terminal degradation, prolonging Aβ interaction [23]. Th is prolonged interaction may enhance forma t ion and stabilisation of toxic Aβ oligomers [24]. Analyses of AD brain samples have demonstrated a higher burden of oligomeric Aβ in APOE4 carriers with increased amyloid plaque-associated synaptic loss. ApoE4 co localises with oligomeric Aβ at the synapse, indicating a key role as a co-factor in Aβ toxicity [25].
Th e greater susceptibility of apoE4 to proteolytic cleavage, and the subsequent prolongation of Aβ interactions, is thought to be a consequence of diff erential domain interaction. Th e C112R polymorphism in apoE4 results in a salt bridge between R61 and E255, which is lacking in apoE3 [26]. Th is brings the amino-and carboxy-terminal domains into closer proximity and exposes the hinge region of apoE4 to proteolysis [23]. Consequently, the development of small-molecule 'structure correctors' that shift apoE4 to an apoE3-like conformation has also been proposed as a therapeutic strategy for AD [27].
Th e main challenge for small molecule approaches aiming to disrupt apoE intradomain or apoE/Aβ proteinprotein interactions is to achieve a compound with suffi cient potency, specifi city and BBB permeability to be suitable for clinical trials.

ApoE mouse models of Aβ amyloidosis
Multiple mouse models of Aβ brain amyloidosis have been generated, predominantly comprising familial, EOAD APP and PS1/2 mutations either alone or in combination [28]. To varying degrees, these mice recapitulate brain parenchymal and cerebrovascular Aβ deposition with cognitive behavioural disorder; however, neuronal loss is relatively lacking in most models. When considering the impact of apoE on Aβ pathology in these mice it is important to consider that endogenous murine apoE is non-polymorphic and does not display domain interaction [29]. Consequently, mouse apoE behaves most similarly to human apoE3. In order to determine the eff ects of human apoE isoforms, Aβ amyloidosis trans genics have now been combined with a variety of human apoE mouse models. Th ese crosses display delayed onset of Aβ pathology relative to their murine equivalents, emphasising the importance of interspecies diff erences [30].
Mice expressing mutant V717F APP in conjunction with human apoE isoform knock-ins (PDAPP/TRE mice) show isoform-dependent Aβ deposition, with apoE4 showing the strongest eff ect followed by apoE3 and then apoE2 [31].
Gene dosage is critically important, with haploinsuffi ciency of both human apoE3 and apoE4 k nock-in isoforms causing marked reductions in Aβ deposition in APP/PS1 mutant mice [32,33]. Th is is a key point, as there is an ongoing debate regarding the potential therapeutic benefi ts of raising versus lowering apoE expression levels. Whilst the transgenic data indicate that reducing apoE levels would be more benefi cial, small-molecule upregulation of apoE levels, particularly through agonism of the lipid X receptor (LXR) [34] or retinoid X receptor (RXR) [35], has been reported as a promising therapeutic approach. In vivo studies of such agonists, whilst successfully demonstrating reductions in Aβ pathology, were carried out against a background of endogenous murine apoE. It remains a possibility, therefore, that increasing expression of human apoE4 may actually be deleterious to disease. It should also be noted that LXR/ RXR agonism has side eff ects, such as hyper tri gly ceridaemia, and the relatively hydrophobic nature of ligands makes complicating interactions with the γ-secretase multispan membrane complex a possibility [36].

ApoE and Aβ production
Th ere is limited evidence for modulation of Aβ production by apoE with in vitro studies using cultured cells co-overexpressing apoE and APP -a relatively unphysiological paradigm [37]. ApoE4-induced increases in Aβ production could be mediated by a novel, apoEinteracting protein, TMC22, proposed to facilitate an interaction between APP and the γ-secretase complex [38].

LXR RXR
Neuron  [40]. It is possible that apoE acts to stabilise oligomeric Aβ, causing enhanced toxicity and seeding deposition of larger aggregates [24].

ApoE and Aβ clearance
Aβ is cleared from the brain by proteolytic degradation [41], bulk fl ow along the perivascular interstitial fl uid drainage pathway [42], or by receptor-mediated clearance across the BBB [43]. In addition, the 'peripheral sink' hypothesis postulates that clearance of Aβ from the brain is accelerated by removal of Aβ from the plasma via the liver and kidneys [44]. APOE4 carriers may display clearance defi cits in both compartments as Aβ removal from both the CNS and the plasma is reduced in human apoE4 knock-in mice [31,45].
ApoE isoform status may infl uence CNS Aβ degradation through indirect mechanisms such as regulation of cellular cholesterol -enhancing endocytosis and lysosomal degradation of Aβ [46]. Th e major impact of apoE is, however, likely to be through interaction of Aβ with cell-surface apoE receptors, including LDL receptorrelated protein 1 (LRP1), the LDL receptor (LDLR) and the VLDL receptor (VLDLR) [47]. Receptor binding of Aβ, alone or in complex with apoE, either delivers Aβ to the lysosome or leads to transcytosis into the plasma via the BBB. LRP1 is perhaps the best characterised transporter acting in the latter instance [48]. ApoE isoforms (apoE4 > apoE3 > apoE2) may disrupt rapid, LRP1mediated clearance of unbound Aβ by diverting it to the VLDLR, which has a slower rate of endocytosis [49].
From a therapeutic perspective, peripheral administration of soluble fragments of LRP1 has been shown to reduce brain Aβ load in K670N/M671L APP mice thr ough plasma Aβ binding -theoretically exploiting the peripheral sink hypothesis [50]. However, the primary investigation of this type of approach has been through enhancement of peripheral Aβ clearance through anti-Aβ immunisation strategies. Th ese remain, despite early setbacks, one of the most promising current therapeutic avenues. Passive immunisation with the humanised anti-Aβ antibody bapinuezumab demonstrated lower effi cacy in APOE4 carriers with a corresponding increase in vasogenic oedema, suggestive of transient increases in vascular permeability [51,52]. If phase III trials are positive, determination of APOE status is likely to become an important aspect of treatment.
In addition to LRP1, LDLR has also been implicated in Aβ removal from the CNS. LDLR over-expression decreased Aβ deposition and enhanced clearance in the K670N/M671L APP, ΔE9 PS1 amyloidosis mouse model [53]. LDLR knockout data are inconsistent, however, as whilst two studies reported increased Aβ load [54,55] a further analysis failed to show any eff ect [56]. Although LDLR-upregulating compounds have been reported [57], clinical usage of such drugs would be challenging due to specifi city and toxicity concerns.

Aβ-independent disease mechanisms
Collaborative large-scale genome-wide association studies have identifi ed, in addition to apoE, novel LOAD risk genes. Th ese include CLU (encoding apolipoprotein J), PICALM, CR1 and BIN1 [58]. Conversely, variants of APP and PS1/2, which increase Aβ42 production in familial EOAD, were not hits in these studies. Th e genetic drivers of LOAD and EOAD are hence likely to be diff erent. Whilst the novel LOAD risk genes may function in either Aβ clearance [43,59] or toxicity [60], there remains a possibility that key implicated pathways, such as lipid homeostasis and innate immunity, play Aβindependent roles in the aetiology of LOAD. ApoE is linked to autoimmune infl ammation, diabetes and coronary heart disease -environmental risk factors for LOAD magnifi ed by the APOE4 genotype [61]. Th e clinical failures of non-steroidal anti-infl ammatories [62], a peroxi some proliferator-activated receptor (PPAR)γ agonist [63] and HMG-CoA reductase inhibitors [64] suggest, however, that targeting mid-life risk factors for LOAD in late stage disease is unlikely to be therapeutically successful. Such treatments, including apoE-based thera peutics, may need to be given earlier in the disease process. Th is places additional importance on early diagnosis of AD and/or preventative treatment in individuals at high risk of developing LOAD.
ApoE, and related cell signalling, is also purported to modulate synaptic plasticity, tau phosphorylation, and neuroinfl ammation [47]. Th e extent to which apoE drives the aetiology of LOAD through these mechanisms is unclear; however, apoE mimetic peptides designed to mediate putative, benefi cial eff ects of apoE demonstrated both behavioural and pathological benefi ts in mutant APP mice [65]. Th e main challenge with such an approach will be to achieve a candidate molecule with appropriate physicochemical properties for clinical use.

Conclusions
Understanding of the interplay between APOE genotype and Aβ pathology has progressed signifi cantly in recent years, particularly with respect to human apoE knock-in animal models of Aβ amyloidosis. Th ese demonstrate an isoform-specifi c role for apoE4 in retarding Aβ clearance from the CNS. By virtue of the nature of the target, however, apoE therapeutics are still at an early preclinical stage, with appreciable chemistry challenges facing small-molecule approaches. Th e most immediate impact of apoE on AD therapeutics will likely be the profi ling of patients for APOE4 status to help determine dosing of anti-Aβ immunotherapy treatments. ApoE has multiple systemic functions, some of which relate to novel LOAD risk genes, which may also aff ect the aetiology of AD independently of Aβ. Th e understanding, and modelling, of these functions remain goals for future research.

Competing interests
Adam Kline was in the past 5 years an employee of Eisai Limited and received a fi xed salary. Adam Kline was not an Eisai employee at the time of publication.