More Than Cholesterol Transporters: Lipoprotein Receptors in CNS Function and Neurodegeneration

Abstract

Members of the low-density lipoprotein (LDL) receptor gene family have a diverse set of biological functions that transcend lipid metabolism. Lipoprotein receptors have broad effects in both the developing and adult brain and participate in synapse development, cargo trafficking, and signal transduction. In addition, several family members play key roles in Alzheimer's disease pathogenesis and neurodegeneration. This review summarizes our current understanding of the role lipoprotein receptors play in CNS function and AD pathology, with a special emphasis on amyloid-independent roles in endocytosis and synaptic dysfunction.

Introduction

The low-density lipoprotein receptor (LDLR) family is a highly conserved receptor family with diverse functions in cellular physiology. The core members of the LDL receptor family include the LDL receptor (Yamamoto et al., 1984), LDLR-related protein 1 (LRP1) (Herz et al., 1988), the very-low-density lipoprotein receptor (VLDLR) (Takahashi et al., 1992), megalin (Saito et al., 1994), Apoer2 (LRP8) (Kim et al., 1996; Novak et al., 1996), LRP4 (Nakayama et al., 1998), and LRP1b (Liu et al., 2000). Orthologues of LDLR family members have been found in species throughout the animal kingdom, highlighting the fundamental role these receptors play in basic cellular processes.

The members of the LDLR family share a conserved structure. At the C-terminus, a short cytoplasmic tail contains one to three NPxY motifs, which facilitate signal transduction and endocytic trafficking through interactions with cytosolic adaptor proteins (Gotthardt et al., 2000). The extracellular domain has a varying number of complement-type repeats and epidermal growth factor precursor homology domains, which are required for ligand binding and pH-dependent ligand release, respectively (Davis et al., 1987; Esser et al., 1988). Additionally, the extracellular domain can interact with a variety of co-receptors.

The lipoprotein receptors bind a wide variety of ligands with differing specificity, helping to explain the diversity of their actions. These ligands include, but are not limited to developmental proteins (Sonic hedgehog (Christ et al., 2012), Wnt (Tamai et al., 2000), and Reelin (D'Arcangelo et al., 1999)), apolipoproteins (ApoE and ApoB), proteases and protease inhibitors (α2-macroglobin (Strickland et al., 1990)), carriers of plasma vitamins (Moestrup and Verroust, 2001), chaperones (RAP (Kounnas et al., 1992)), and inflammatory mediators (TGFβ (Muratoglu et al., 2011)).

Through interactions with these ligands, the lipoprotein receptors have a diverse set of functions in the developing and adult nervous system, from lipoprotein trafficking, to synaptic plasticity, to cell migration and development. Additionally, lipoprotein receptors affect neurodegenerative processes and are the key receptors for the main Alzheimer's disease risk factor, ApoE4. This review will touch on each of these aspects of lipoprotein receptor signaling in the nervous system.

The Lipoprotein Receptors and Cholesterol Metabolism

The lipoprotein receptors were initially identified for their role in peripheral cholesterol transport. Cholesterol is required for several processes in the body. It is a key component of the plasma membrane and a precursor for the production of steroid hormones. In the periphery, lipoprotein receptors bind triglyceride- and cholesterol-carrying lipoprotein particles such as chylomicrons, VLDL, and LDL for endocytosis via clathrin-coated pits into cells for further metabolism. (Go and Mani, 2012). Similar to their role in the periphery, lipoprotein receptors mediate cholesterol transport in the central nervous system (CNS). Plasma lipoproteins cannot cross the blood brain barrier, and thus cholesterol-carrying particles must be produced within the CNS (Bjorkhem et al., 1998). In the CNS, glia-derived cholesterol is essential for the formation and maintenance of mature synapses (Mauch et al., 2001).Astrocytes release ApoE-containing ‘HDL-like’ particles to transport cholesterol and phospholipids between glia and neurons (Boyles et al., 1985; Pitas et al., 1987). ApoE binds to a variety of cell surface receptors in the LDLR family for particle uptake, including LDLR, LRP1, VLDLR, and Apoer2 (Beffert et al., 2004). Lrp1 appears to have the highest transport capacity for ApoE, due to its rapid endoyctic recycling rates (Li et al., 2001). Moreover, selective deletion of Lrp1 in forebrain neurons leads to a global defect in brain lipid metabolism and neurodegeneration in mice (Liu et al., 2010). Other lipoprotein receptor family members may also participate in CNS cholesterol homeostasis.

The importance of cholesterol homeostasis for brain function is highlighted by Niemann-Pick type C (NPC) disease. In NPC, autosomal recessive mutations in NPC1 or NPC2 result in impaired cholesterol trafficking, which leads to a buildup of cholesterol and lipids in the lysosome and a deficiency of these molecules for use both in the membrane and as a precursor for steroid synthesis (Carstea et al., 1997; Naureckiene et al., 2000). NPC is characterized primarily by progressive neurodegeneration, which indicates the brain's high requirement for cholesterol relative to other organs.

Lipoprotein Receptors and the Developing Brain

Several lipoprotein receptors play a vital role in the developing embryo. Megalin, or Lrp2, is a 600-kDa lipoprotein receptor expressed predominantly in the developing embryo and the mammalian kidney (Saito et al., 1994). Lrp2 is required for normal organogenesis, and Lrp2 knockout mice have holoprosencephaly with fused hemispheres and no olfactory bulbs (Willnow et al., 1996). This phenotype results from increased BMP4 expression and a reduction of sonic hedgehog (Shh) expression, which leads to a loss of interneurons and oligodendroglial cell populations (Spoelgen et al., 2005). Individuals with autosomal recessive loss of function mutations in LRP2 develop Donnai-Barrow syndrome, which is characterized by facial dysmorphology, ocular and hearing anomalies, and agenesis of the corpus callosum (Kantarci et al., 2007).

The lipoprotein receptors Apoer2 and Vldlr also affect development as mediators of Reelin signaling. Reelin is a large secreted extracellular matrix protein, released early in development by Cajal-Retzius cells and later in life by a subset of GABAergic interneurons (Curran and D'Arcangelo, 1998). Reelin signaling is required for proper neuronal migration and maturation, and Reelin knockout (reeler) mice have inverted cortical layering, cerebellar hypoplasia, and immature synapses (Falconer, 1951; Niu et al., 2008). Reeler mice also have impaired learning and motor coordination as well as severe neurodevelopmental deficits that lead to strain background-dependent lethality by about 5 weeks of age (D'Arcangelo et al., 1995). Since Reelin uses both Apoer2 and Vldlr to mediate its signaling, Apoer2 and Vldlr double knockout mice also have a reeler phenotype (Trommsdorff et al., 1999). Patients with mutations in RELN or VLDLR have been identified that recapitulate certain aspects of the reeler phenotype (Hong et al., 2000; Ozcelik et al., 2008), which underscores the importance of these receptors in the development of the CNS.

Lipoprotein Receptors, Reelin, and Synaptic Plasticity

Several of the lipoprotein receptors are expressed in the adult CNS. Lrp1 is expressed in astrocytes, microglia, and neurons in the post-synaptic density (May et al., 2004). VLDLR is expressed by many cell types, and in the brain it is found in glia, neuroblasts, matrix cells, and pyramidal neurons on cell membranes outside of lipid rafts (Christie et al., 1996; Duit et al., 2010). Similarly, LDLR is expressed in both astrocytes and neurons (Fan et al., 2001). Conversely, Apoer2 is largely restricted to the testes and brain, where it is expressed in neurons throughout the CNS and traffics into the post-synaptic density (Beffert et al., 2005; Clatworthy et al., 1999). Due to their localization and expression in neurons, lipoprotein receptors are well-placed to affect synaptic function and synaptic plasticity.

The lipoprotein receptors Apoer2 and Vldlr play a role in synaptic plasticity through their interaction with the glycoprotein Reelin. At the adult synapse, Reelin has a key effect of enhancing long-term potentiation (LTP). Reelin binding clusters Apoer2 and Vldlr, leading to receptor dimerization and tyrosine phosphorylation of the cytosolic protein Dab1 (Benhayon et al., 2003; D'Arcangelo et al., 1999; Herz and Chen, 2006; Hiesberger et al., 1999; Howell et al., 1997; Strasser et al., 2004). Dab1 then activates Src tyrosine kinases (SFKs), including Fyn, which phosphorylate NMDA receptors, enhancing their Ca²⁺ influx (Chen et al., 2005). This increased Ca²⁺influx allows for a stronger NMDA current at similar levels of glutamate release, resulting in greater AMPA insertion into the membrane and enhanced LTP (Weeber et al., 2002). Consistent with its ability to enhance LTP, acute intra-ventricular Reelin injections improve memory performance of wild type mice and can rescue learning and memory deficits in heterozygous reeler mice (Rogers et al., 2011; Rogers et al., 2012).

The robust effect of Reelin signaling has led to a search for other Reelin receptors. In addition to Apoer2 and Vldlr, Reelin also binds EphB2, a member of the EphB receptor family (Bouche et al., 2013). EphB receptors are required for proper CNS development and synaptic function (Klein, 2009; Xu and Henkemeyer, 2012). EphB2 activation by its ligand Ephrin-B1 results in the clustering of EphB2 and NMDA receptors and promotes excitatory synaptic development (Dalva et al., 2000). In mature synapses, EphB2 promotes synaptic targeting of NR2B subunits, which allows greater calcium influx upon NMDA activation and may shift the synaptic plasticity balance towards LTP (Nolt et al., 2011). In vitro, Reelin can at least partially activate EphB2 signaling; however, the effects of Reelin-mediated EphB2 signaling on NMDA receptor targeting and adult synaptic function have yet to be determined (Bouche et al., 2013). Intriguingly, Apoer2 binds NR2 to promote its phosphorylation, while EphB2 binds NR1 (Beffert et al., 2005; Dalva et al., 2000; Hoe et al., 2006). Reelin binds to each of these receptors at a distinct and separate site, suggesting that Apoer2, NMDAR, and EphB2 may form a macromolecular complex at the synapse (Bouche et al., 2013). (Figure 1C.) In addition to EphB receptors, it has been reported that LDLR may interact with Reelin (D'Arcangelo et al., 1999), albeit much more weakly than Apoer2 and Vldlr, Functionally, these are the main Reelin receptors, since Apoer2/Vldlr double knockout mice are phenotypically identical phenotype to reeler mice (Trommsdorff et al., 1999).

Figure 1. ApoE, Reelin, and Synaptic Plasticity.

A.) Reelin binds to its receptor, Apoer2, activating a signaling cascade through Dab1 and SFKs to phosphorylate NMDA receptors, resulting in greater Ca²⁺influx upon glutamate signaling (Benhayon et al., 2003; Chen et al., 2005). Dab1 also acts to counteract Aβ and PP2B's effects on Akt, GSK3β, and tau phosphorylation (Beffert et al., 2002; Durakoglugil et al., 2009). B.) Reelin signaling results in the endocytosis of Apoer2, AMPA, and NMDA receptors. ApoE isoforms differentially affect the recycling of these endosomes (Chen et al., 2010). ApoE3-containing endosomes readily recycle back to the surface, while those containing ApoE4 remain trapped (trapped vesicles are displayed at reduced size). C.) Apoer2 and EphB2 bind the NR2 and NR1 subunits of NMDAR, respectively, regulating receptor function and endocytosis (Beffert et al., 2005; Dalva et al., 2000). Reelin binds to EphB2 and Apoer2 through independent, distinct sites (Bouche et al., 2013). Akt, protein kinase B; AMPA, alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid; ApoE, apolipoprotein E; Apoer2, apolipoprotein E receptor 2; Dab1, disabled-1; glu, glutamate; GSK3β, glycogen synthase kinase 3 beta; LTP, long term potentiation; NMDA, N-methyl-D-aspartate; PI3K, phosphoinositide 3-kinase; PP2B, protein phosphatase 2B, calcineurin; SFKs, Src-family kinases.

Most studies of the synaptic role of Reelin have focused on its post-synaptic effects; however, new research suggests that Reelin also affects the release of VAMP7-containing synaptic vesicles (Bal et al., 2013). Previously, Apoer2 and Vldlr had been thought to be predominantly postsynaptic proteins. Intriguingly, reeler mice have more vesicles at the presynaptic bouton and a reduction in SNAP25, an element of the vesicle fusion machinery, suggesting impairment in vesicle fusion and release. However, this effect is not dependent on either Apoer2 or Vldlr, but rather the integrin pathway (Hellwig et al., 2011). The precise details by which Reelin signals at the pre-synapse and the role of Reelin receptor signaling remain to be clarified.

The Role of Lipoprotein Receptors in the Peripheral Nervous System

The neural function of lipoprotein receptors is not limited to the brain; in fact, they play a large role in the development of the neuromuscular junction (NMJ). The vertebrate NMJ is a specialized cholinergic synapse in which presynaptic motor neurons connect with post-synaptic muscle fibers at the motor endplate. The lipoprotein receptor Lrp4 plays several roles in NMJ formation. First, Lrp4 forms a complex with MuSK to pre-pattern the muscle (Kim and Burden, 2008). Activation of MuSK results in the recruitment of several downstream pathways that result in nicotinic acetylcholine receptor (AChR) clustering. When the motor neuron arrives, it releases agrin, which stimulates MuSK to promote maturation of the synapse. Lrp4 is required for the relay of the agrin signaling to MuSK and in the absence of Lrp4, agrin's ability to induce MuSK phosphorylation is minimal (Kim et al., 2008; Zhang et al., 2008). Finally, independent of agrin or MuSK, Lrp4 acts as a retrograde messenger to promote pre-synaptic vesicle clustering (Yumoto et al., 2012). Surprisingly, Lrp4 only seems to be required for NMJ development in some mammalian species. While Lrp4 knockout mice exhibit neonatal lethality due to failure to form the NMJ (Weatherbee et al., 2006), homozygous mutations in LRP4 that would be predicted to abolish its functionality cause Cenani-Lenz syndrome in humans and mulefoot disease in cattle (Johnson et al., 2006; Karner et al., 2010; Li et al., 2010), with apparently functional NMJs. These findings could point to a different mechanism in the organization of the NMJ in cattle and humans or a redundancy that may compensate for functional Lrp4 loss. This redundancy could potentially be mediated by Lrp10, a LDLR family member that is virtually identical to Lrp4; however, it remains to be determined if LRP10 indeed arose from a recent gene duplication.

The amyloid precursor protein, APP, which is discussed in depth later in the next section for its role in Alzheimer's disease, also takes part in NMJ development. APP, or its paralogue APLP2, is required both pre- and post-synaptically for NMJ development (Akaaboune et al., 2000; Wang et al., 2005). One hypothesis is that APP acts trans-synaptically to promote cross-talk between the pre- and post-synaptic compartments. While this has not been conclusively shown, it is known that APP can bind both to itself and to APLP2 in trans to promote cell adhesion in vitro (Soba et al., 2005). At the presynaptic side, APP interactions promote choline transporter (CHT) activity and expression, and they promote L-type calcium channel function, both of which are required for proper NMJ function (Wang et al., 2007; Yang et al., 2007). Finally, APP acts post-synaptically on the motor endplate to promote Lrp4 and MuSK interaction (Choi et al., 2013). These findings at peripheral synapses may also apply at central synapses and point the way for future studies on the role of lipoprotein receptors in central synapse development.

Lipoprotein Receptors and Neurodegenerative Disease

Alzheimer's disease (AD) is a devastating neurodegenerative disorder characterized by progressive memory loss and a pathological accumulation of amyloid plaques and neurofibrillary tangles. Considerable research into the disease has focused on the genetic, cognitive, and pathological changes found in individuals with early onset familial disease, who represent approximately 1% of the patient population. These individuals have mutations in either the amyloid precursor protein (APP) or its cleavage proteins PS1 and PS2, resulting in the early accumulation of plaque-forming amyloid β (Aβ) (Weggen and Beher, 2012). However, the majority of AD patients have late-onset disease (LOAD), with no mutations in the Aβ generating machinery. The major genetic risk factor for LOAD is the ε4 isoform of apolipoprotein E – a cholesterol transport protein.

In humans, ApoE exists in three isoforms that differ at two residues: ε2 (ApoE2 with cys112 and cys158), ε3 (ApoE3 with cys112 and arg158), and ε4 (ApoE4 with arg112 and arg158). These variations confer significant differences in disease risk. ApoE3 is the most common isoform and thus is considered the gold standard from the population genetic risk perspective. Relative to ApoE3, ApoE4 is present in approximately 15 - 20% of the population and is associated with greatly increased risk and earlier onset of AD (Corder et al., 1993; Schmechel et al., 1993). Meanwhile, ApoE2 isoform is protective against AD relative to ApoE3 and ApoE4 (Corder et al., 1994). The structural differences between these isoforms contribute to differences in protein conformation, receptor binding, and endocytic trafficking (reviewed in Kanekiyo et al., 2014). Interestingly, genetic association studies of polymorphisms in lipoprotein receptors with AD risk have only revealed potential minor effects of LRP1 (Beffert et al., 1999; Jin et al., 2013). This may be due to the fact that several of the lipoprotein receptors have key developmental roles, which could preclude major effects on neurodegeneration. By contrast, a significant association of SORL1 with AD has been reported (Rogaeva et al., 2007), which has been corroborated mechanistically on the molecular level (Schmidt et al., 2012).

What links lipoprotein receptors to AD? The difficulty in finding the answer stems from the fact that late-onset AD is a multifactorial disease involving endocytic dysfunction, lipoprotein signaling, and synaptic dysregulation – essentially three different disciplines. The remainder of this review focuses on the variety of roles played by lipoprotein receptors in neurodegenerative processes.

APP Processing and the Endosome

The pathological hallmark of Alzheimer's disease is progressive accumulation of amyloid plaques. These plaques are formed of aggregated amyloid β (Aβ), the end product of sequential cleavage of amyloid precursor protein (APP) (reviewed in (Zhang et al., 2012)). Briefly, APP is cleaved first by β-secretase (BACE-1) to sAPPβ and β C-terminal fragment (βCTF) (Cai et al., 2001). βCTF is then cleaved by γ-secretase: a complex of multiple proteins that includes presenilin 1 or 2 (PS1 or PS2), APH, nicastrin, and PEN2 (St George-Hyslop and Fraser, 2012), resulting in the final cleavage product, Aβ.

The APP metabolites resulting from this sequential cleavage play their own roles in neuronal function, reviewed in depth in (Chow et al., 2010). sAPPβ, generated by β-cleavage, appears to be involved in synaptic pruning and apoptosis (Nikolaev et al., 2009). In contrast to amyloidogenic β-cleavage, APP can be cleaved by α-secretase within the Aβ segment to generate the non-amyloidogenic sAPPα (Vingtdeux and Marambaud, 2012). sAPPα is generally considered to be neuroprotective, enhancing LTP and spatial learning (Taylor et al., 2008). The other cleavage products created by α- and β-cleavage, CTF83 and CTF99, respectively, do not have well defined roles. However, γ-secretase cleavage of the CTFs results in the release of the APP Intracellular Domain (AICD), which binds Fe65 and TIP60 for potential transcriptional activation of downstream targets (Cao and Sudhof, 2001), ultimately promoting cell death and impairing neurogenesis (Ghosal et al., 2010; Wang et al., 2014). The final and most-studied products of APP cleavage are the Aβ species, predominantly Aβ₄₀ and some Aβ₄₂. Of these two species, Aβ₄₂ is considered to be more amyloidogenic and thus more toxic (Aoki et al., 2008; Kim et al., 2007; McGowan et al., 2005). Mutations causing early onset Alzheimer's disease directly affect this process – mutations in App facilitate cleavage by BACE1 to amyloidogenic species, while mutations in PS1 and PS2 enhance their γ-cleavage ability and tend to produce a higher Aβ₄₂:Aβ₄₀ ratio (Weggen and Beher, 2012). By contrast, in LOAD, there is no inherent dysfunction of the key enzymes of Aβ generation, yet Aβ still accumulates.

Whether APP is α- or β- cleaved is determined by the localization of APP processing proteins. α-secretase is active at the membrane surface, whereas BACE1 is most active at the lower pH found in endosomes (Cole and Vassar, 2007). As a result, conditions that result in increased trafficking of APP to the endosome are pro-amyloidogenic, whereas those that maintain APP at the surface are anti-amyloidogenic. Moreover, in healthy brains APP and BACE1 are largely localized to different compartments, while AD brains show increased colocalization of the two proteins within acidic microdomains (Das et al., 2013). Additionally, overexpression of ApoE4 promotes colocalization of APP and BACE1 in vitro (Rhinn et al., 2013). However, localization to the membrane surface is not enough to prevent APP β-cleavage. Lipid rafts in the membrane also have increased BACE1 activity and decreased α-secretase activity (Ehehalt et al., 2003). It is likely that β-cleavage in lipid rafts contributes to Aβ generation; however, the bulk of Aβ generation, approximately 70%, is dependent on endocytic recycling (Cirrito et al., 2008).

Histological evidence suggests that dysfunction of endocytic recycling is one of the earliest pathological changes in AD. In postmortem studies of AD patients, early endosomes were enlarged up to 32 fold the volume of normal endosomes (Cataldo et al., 1997). In theory, the enlarged endosomes give APP more time to interact with β- and γ-secretases and thus generate more Aβ. Remarkably, endosomal enlargement begins prior to the appearance of clinical disease in APOE ε4 carriers, and this finding was reproduced in induced pluripotent stem cells derived from AD patient fibroblasts (Cataldo et al., 2000; Israel et al., 2012). Partly because of these findings, recent AD research has focused on changes in endosome recycling associated with the members of the LDL receptor family.

Several members of the lipoprotein receptor family affect APP endocytosis. For example, LRP1 has an endocytosis rate that is much faster than APP. As a result, when the LRP1 extracellular domain interacts with APP, APP endocytosis is accelerated, leading to increased processing to Aβ (Cam et al., 2005). Overexpression of a functional LRP1 mini-receptor lead to increased soluble Aβ levels in the PDAPP mouse model (Zerbinatti et al., 2006; Zerbinatti et al., 2004). However, LRP1 also plays a role in promoting neuronal Aβ clearance and neuronally restricted disruption of Lrp1 in an APP/PS1 model mice accelerated Aβ accumulation (Kanekiyo et al., 2013). Additionally, the scaffolding protein RanBP9 increases APP localization to lipid rafts and enhances the association of LRP1, APP, and BACE1, to increase Aβ generation (Lakshmana et al., 2009). RanBP9 is increased in AD brains and in APP mice, suggesting that it may have a role in disease pathogenesis (Lakshmana et al., 2010; Woo et al., 2012). A newly identified member of the LDLR family, low-density-lipoprotein receptor class A domain containing 3 (LRAD3), also binds to APP and promotes amyloidogenic processing (Ranganathan et al., 2011). In contrast to LRP1 and LRAD3, the related receptor LRP1B has a slower rate of endocytosis and retains APP at the cell surface, which leads to decreased Aβ generation (Cam et al., 2004). Similarly, LRP10, another novel LDLR family member binds to APP and promotes its trafficking to the Golgi complex, which also prevents amyloidogenic processing (Brodeur et al., 2012).

The role of the LDL receptor family member Apoer2 in the endocytic processing of APP is less clear. Apoer2 alone, in the absence of ligand, can cause APP localization to lipid rafts, which leads to increased Aβ generation in the absence of endocytosis (Fuentealba et al., 2007). In the presence of F-spondin, which interacts with Apoer2 and APP, Apoer2's slower endocytosis rate inhibits APP endocytosis, maintaining APP at the surface and reducing Aβ generation (Hoe et al., 2005). However, ApoE binding to Apoer2 causes endocytosis of APP, Apoer2 and BACE1 (He et al., 2007). Together, these data suggest that the effect of Apoer2 on APP endocytosis depends on the ligands present. Intriguingly, VLDLR and APP can increase each other's surface expression through mutual interactions with Fe65 (Dumanis et al., 2012).

In addition to the LDL receptor gene family of ApoE receptors, members of the structurally related Vps10p domain containing receptor family are also involved in APP trafficking. Sorting protein–related receptor with A-type repeats (SorLA, alternatively SORL1 or LR11) is a neuronal Vps10p containing receptor that interacts with APP and generally reduces its cleavage by either α- or β-secretases (Schmidt et al., 2007). Human studies have shown that SorLA is decreased in late-onset AD, and certain genetic variants of SorLA are associated with AD, indicating a possible pathogenic role for SorLA (Dodson et al., 2006; Rogaeva et al., 2007; Scherzer et al., 2004). Recent data support a novel inhibitory mechanism for SorLA, where it binds to APP monomers and prevents them from forming dimers, which are the preferred ligands for both α- and β-secretases (Schmidt et al., 2012). Overexpression of SorLA causes the redistribution of APP to the Golgi complex, where all secretases are less active, leading to a decrease in Aβ levels (Andersen et al., 2005). Recently, another member of the Vps10p family, sortilin, was identified as being involved with APP endocytosis in vitro. Sortilin promotes the endocytosis of APP and the generation of sAPPα (Gustafsen et al., 2013). However, an in vivo study in sortilin knockout mice did not reveal a change in sAPPα or sAPPβ levels (Carlo et al., 2013). (Figure 2).

Figure 2. APP Processing and Endocytosis.

In the pro-amyloidogenic pathway, APP is cleaved sequentially by BACE1 to sAPPβ and C99 (Cai et al., 2001), and then sAPP β is cleaved by γ-secretase to Aβ and AICD (St George-Hyslop and Fraser, 2012). In the non-amyloidogenic pathway, α-secretase cleaves APP through the Aβ site to generate C83 and sAPPα, which is cleaved to P3 and AICD by γ-secretase (Vingtdeux and Marambaud, 2012). β-cleavage is favored in the endosome and in lipid rafts, while α-cleavage is favored at the cell surface outside of lipid rafts (Ehehalt et al., 2003). Several members of the low-density lipoprotein receptor family affect APP localization. LRP1 increases APP trafficking to the endosome (Cam et al., 2005), while LRP1B decreases trafficking (Cam et al., 2004). Apoer2 can direct APP to lipid rafts or to α-secretase in the presence of F-spondin (Hoe et al., 2005). SORLA can prevents α- and β-cleavage and promotes APP transport to the Golgi (Andersen et al., 2005; Schmidt et al., 2007). Aβ, amyloid beta; AICD, APP intracellular domain; Apoer2, apolipoprotein E receptor-2; APP, amyloid precursor protein; Arc, activity-regulated cytoskeleton-associated protein; BACE1, beta-site APP cleaving enzyme 1; C83, APP C-terminal C83 fragment; C99, APP C-terminal C99 fragment; LRP1, low density lipoprotein receptor-related protein 1; LRP1B, low density lipoprotein receptor-related protein 1B; sAPPα, soluble APP α; sAPPβ, soluble APP β; SORLA, sortilin-related receptor.

Alterations in endocytosis of APP clearly play a role in the pathology of AD, based on evidence from patient tissue and mouse models of disease. However, it is still not entirely clear what role ApoE isoforms play physiologically in APP endocytosis and subsequent Aβ-generation. In one study, the ApoE4 isoform caused increased endocytosis of APP, Apoer2, and BACE1 in neuroblastoma cells (He et al., 2007). Another study showed that ApoE4 delayed recycling of endosomes in hepatocytes (Heeren et al., 2004) and neuronal cells (Rellin et al., 2008). However, in vivo studies using microdialysis of the interstitial fluid (ISF) in ApoE transgenic mice showed that ApoE isoform has no effect on Aβ synthesis, only its clearance (Castellano et al., 2011).

Synaptic Activity Increases Aβ Generation

The physiological role of Aβ has long been debated. A few studies have demonstrated that blocking Aβ in young mice leads to reduced performance on memory tasks, suggesting that at low levels Aβ plays a role in synaptic function (Abramov et al., 2009; Morley et al., 2010; Puzzo et al., 2011; Puzzo et al., 2008). Remarkably, synaptic activity increases levels of extracellular Aβ and decreases levels of intra-neuronal Aβ (Cirrito et al., 2005). This is achieved by two separate mechanisms. First, the activity-dependent generation of extracellular Aβ is dependent on APP endocytosis and processing by BACE1 and γ-secretase (Cirrito et al., 2008; Kamenetz et al., 2003). This is in part mediated by Arc, an intermediate early gene that is expressed shortly following synaptic activity (Soule et al., 2008). Arc increases the association of γ-secretase with APP in endosomes (Wu et al., 2011). Conversely, synaptic activity decreases intra-neuronal Aβ42 through increased processing by neprilysin, the main Aβ protease (Tampellini et al., 2009). Intriguingly, these results suggest a physiological role for Aβ. In response to synaptic activity, Aβ is increased extracellularly, which dampens LTP at dendrites in the area through a number of candidate receptors including NMDA receptors, mGluR5, and PrPc (Lauren et al., 2009; Li et al., 2011a; Renner et al., 2010). This could indicate that Aβ is involved in homeostatic scaling, a process where highly active synapses begin to dampen their responses, and less active ones become more sensitive (reviewed in (Sheng et al., 2012)).

Reelin, Aβ, and Tau at the Synapse

Aβ oligomers have been shown to inhibit LTP induction (Nomura et al., 2012; Shankar et al., 2008; Walsh et al., 2002). Aβ activates caspase-3 through a mechanism that is currently unknown but probably involves mitochondrial dysfunction. Caspase-3 then activates PP2B (calcineurin), which directly de-phosphorylates NMDAR at the Fyn phosphorylation site (D'Amelio et al., 2011; Kim et al., 2009; Snyder et al., 2005). Caspase-3 also cleaves Akt, rendering Akt inactive and unable to phosphorylate and inhibit GSK3β (Jo et al., 2011). When left uninhibited GSK3β blocks LTP induction by an unknown mechanism that may involve tau (see below). Reelin has the opposite effect from Aβ on GSK3β activity. Dab1 activates PI3K, which activates Akt, leading to decreased GSK3β activation (Beffert et al., 2002). Convincingly, Aβ inhibition of LTP in hippocampal slices is prevented by co-application of Reelin in the ACSF (Durakoglugil et al., 2009). Thus, Reelin signaling through ApoE receptors intersects with Aβ at the level of tyrosine phosphorylation of NMDA receptors to affect LTP induction, learning and memory (Figure 1A).

As previously mentioned, the two main Reelin receptors, Apoer2 and Vldlr, also bind ApoE. Different ApoE isoforms differentially affect Reelin's ability to signal through the ApoE receptors by interfering with receptor recycling to and from the synapse. Once Reelin activates its receptors, the receptors are endocytosed, along with Reelin and ApoE. While ApoE2 and ApoE3 do not interfere with endosomal recycling, ApoE4 impairs it. This impaired recycling decreases the available surface pool of Apoer2 and thus Reelin signaling (Figure 1B). As a result, hippocampal slices from ApoE4 transgenic mice do not enhance LTP upon Reelin addition to the perfusate (Chen et al., 2010). Importantly, because Reelin is unable to efficiently signal in ApoE4 knockin mice, Reelin is unable to protect against the Aβ-induced synaptic suppression. Consequently, ApoE4 promotes increased susceptibility to Aβ damage at the synapse (Chen et al., 2010).

The importance of Reelin in AD pathology is implicated by several studies. Reelin levels are reduced in the entorhinal cortex in postmortem AD brain and hAPP mouse models (Chin et al., 2007; Herring et al., 2012). Interestingly, another group found that Reelin levels were elevated in frontal cortex of AD patients, suggesting potential compensation in that region for the elevated Aβ levels (Botella-Lopez et al., 2006; Botella-Lopez et al., 2010). Additionally, cognitively normal individuals with a high level of plaques and tangles possess certain SNPs in RELN that may make them more prone to elevating Reelin levels in response to high Aβ load (Kramer et al., 2011). For review, see (Krstic et al., 2013). Finally, Reelin is not the only ligand that binds to Apoer2 and Vldlr to activate Dab1 signaling. Clusterin, a.k.a. ApoJ, also binds to Apoer2 and Vldlr and can activate Reelin signaling (Andersen et al., 2003; Leeb et al., 2013). Clusterin SNPs are associated with AD, and clusterin levels are elevated in AD (Harold et al., 2009; Lambert et al., 2009; May et al., 1990), further highlighting the role of ApoE receptors in AD pathogenesis.

As described earlier, activation of the Reelin pathway culminates in GSK3β inhibition. In many disease states, including Alzheimer's disease, GSK3β is unregulated, leading to its over-activation (Jope et al., 2007). One proposed mechanism for GSK3β's effects on synaptic plasticity is tau hyperphosphorylation. Aggregated, hyperphosphorylated tau comprises the second pathological hallmark of AD, neurofibrillary tangles (NFTs). In a diseased brain, tau becomes phosphorylated at several sites by a number of kinases, leading to its detachment from the microtubule (Gotz et al., 2010). Hyperphosphorylated tau migrates from the axonal compartment to the somatodendritic compartment, where it destabilizes the synapse (Hoover et al., 2010; Li et al., 2011b). Tau aggregates to form NFTs, ultimately resulting in neuronal death. Remarkably, activated GSK3β is one of the kinases that phosphorylates tau, indicating a potential mechanism for its inhibition of LTP (Mulot et al., 1994). Indeed, hippocampal slices from tau knockout mice are resistant to Aβ-induced LTP reduction (Shipton et al., 2011), as well as learning and memory deficits (Roberson et al., 2007).

The effect of Reelin pathway dysfunction on tau phosphorylation appears to be strain-specific. While some mouse strains have high levels of tau hyperphosphorylation in the absence of Dab1, other strains are relatively unaffected. A quantitative trait locus analysis comparing Dab1-deficient mice with high or low levels of tau phosphorylation revealed a strong hit on chromosome 16 at a locus that includes the genes for APP and SOD, as well as a suggestive hit at the locus that includes Psen1 (Brich et al., 2003). The genetic interplay between these key proteins highlights the integrative effects of the Aβ, Reelin, and tau pathways in disease pathogenesis.

The process by which dendritic tau causes synaptic dysfunction is largely unknown. One hypothesis is that tau acts as a carrier for Fyn kinase to the dendrites. At the synapse, Fyn kinase phosphorylates the NR2 subunit of NMDA receptors, which facilitates its interaction with PSD-95, and may lead to increased glutamate excitotoxicity. In the absence of tau, or with a truncated version of tau incapable of entering the dendrite (Δtau), Fyn remains localized to the soma. As a result, APP23 mice with no tau or Δtau have fewer memory deficits despite having similar levels of APP and Aβ as those with full-length tau (Ittner et al., 2010). Fyn's role in tau pathology is supported by the fact that overexpression of Fyn leads to greater impairment in spatial learning in APP mice (Chin et al., 2005). However, the tau-Fyn hypothesis, reviewed in depth in (Ittner and Gotz, 2011), is seemingly at odds with the physiological requirement for Fyn at the synapse and with the proposed mechanism of Reelin enhancement of LTP (Grant et al., 1992). As described above, Reelin activates Fyn kinase, but this enhances LTP, rather than causing excitotoxicity. The answer to this conundrum may lie in a different compartmentalization of these effectors: for example, the effect of Aβ and tau on extrasynaptic versus synaptic NMDA receptors, and the role of NR2B versus NR2A subunits (Hardingham et al., 2002; Tackenberg et al., 2013; Talantova et al., 2013) (see Parsons and Raymond, 2014 for a review). Clearly, more work is required to better understand the interactions of Aβ, tau, and Reelin at the synapse.

ApoE and Aβ Clearance

Once Aβ is generated by secretase cleavage, it is released to the ISF where it is cleared by a variety of mechanisms: efflux across the blood brain barrier (BBB), uptake by cells for lysosomal degradation, and cleavage by Aβ-specific proteases. Dysfunctional clearance of Aβ plays a key role in amyloid accumulation, since late-onset AD patients do not have increased Aβ production, but rather insufficient Aβ clearance (Mawuenyega et al., 2010). Further, Aβ42 is generally cleared more slowly than Aβ40, which leads to a preferential accumulation of the more toxic species (Bell et al., 2007). Aβ is cleared from the ISF in an ApoE-isoform dependent manner: ApoE2 allows for the more rapid clearance of Aβ in vivo, while ApoE4 slows clearance. Thus, the presence of ApoE4 promotes the buildup of Aβ in the ISF (Castellano et al., 2011). The ApoE receptors LRP1, LDLR, and VLDLR all appear to play a key role in the clearance of Aβ. As these Aβ-clearing effects of ApoE were recently reviewed in depth (Kanekiyo et al., 2014), we will only touch on them briefly here.

One method of Aβ clearance is efflux across the BBB to the periphery, where Aβ is rapidly degraded in the plasma. ApoE receptors directly mediate this process, primarily LRP1, which uses its ligands, α2-macroglobulin, RAP, or ApoE, to promote Aβ efflux (Bell et al., 2007; Hughes et al., 1998; Kanekiyo and Bu, 2009; Narita et al., 1997). ApoE2 and ApoE3 use both LRP1 and Vldlr to cross the BBB, leading to faster Aβ efflux. ApoE4 shifts Aβ transport from LRP1 to Vldlr, resulting in a slower rate of clearance and build-up of Aβ in the ISF (Deane et al., 2008). The LDLR, was also recently found to promote Aβ efflux across the BBB (Castellano et al., 2012). While it is clear that lipoprotein receptors have a large effect on Aβ efflux at the BBB, further work is required to elucidate their exact mechanisms of action.

Normally, transport across the BBB is highly regulated by transporters such as the lipoprotein receptors; however, if the BBB is damaged, this regulation is lost. Perhaps unsurprisingly, the BBB is found to be defective in some, but not all, postmortem AD samples (Clifford et al., 2007). If the BBB has been damaged, Aβ species may leak back into the CNS and accumulate in neurons, contributing to amyloid pathology (Clifford et al., 2007). In vitro experiments disagree on the role of ApoE and its isoforms on BBB maintenance (Hafezi-Moghadam et al., 2007; Nishitsuji et al., 2011). The importance of ApoE4 in BBB stability in vivo needs to be clarified, since only some AD patients have weakened BBB, and the deterioration might be a result of disease pathology, rather than the cause.

In addition to efflux across the BBB, Aβ can be cleared centrally by endocytic trafficking to lysosomes or degradation by Aβ proteases. ApoE3 promotes Aβ uptake and degradation by lysosomes in vitro, while ApoE4 has a weaker effect (Li et al., 2012). In addition to lysosomal proteases, there are two Aβ-specific proteases. Intracellularly, the degradation of Aβ is mediated by neprilysin. While all ApoE isoforms enhance Aβ degradation by neprilysin, ApoE4 appears to have the weakest effect in vitro (Jiang et al., 2008). Extracellularly, the degradation of Aβ is mediated by insulin degrading enzyme (IDE), a protease that is released by microglia and astrocytes into the extracellular milieu and whose levels are reduced by ApoE4 (Jiang et al., 2008; Qiu et al., 1998).

The interaction between ApoE and Aβ is affected by the lipidation state of ApoE. It has been shown in vitro that lipidated ApoE is more effective at promoting both Aβ efflux across the BBB and Aβ degradation by proteases (Jiang et al., 2008). Lipidated ApoE is generated in the brain by ATP binding cassette transporter A1 (ABCA1), which transfers cholesterol from the plasma membrane to lipid-poor ApoE (Lawn et al., 1999). Overexpression of ABCA1 increases ApoE lipidation and reduces amyloid deposition in a mouse model of AD (Wahrle et al., 2008); conversely, decreasing ABCA1 expression has the opposite effect (Wahrle et al., 2005). One explanation for this effect is that ApoE binds Aβ more efficiently when lipidated (Tokuda et al., 2000); however, ApoE shows minimal interaction with Aβ in vivo and may instead promote Aβ clearance indirectly (Verghese et al., 2013). (Figure 3).

Figure 3. Removing Aβ from the CNS.

Aβ can be degraded by IDE in the ISF or taken up into microglia for degradation by neprilysin (nep) (Jiang et al., 2008; Qiu et al., 1998). Aβ can also be cleared across the BBB to the plasma, where it is degraded by plasma proteases. This process is accelerated by Aβ binding to ApoE particles. While ApoE3 uses both Vldlr and LRP1 to cross the BBB, ApoE4 primarily uses Vldlr, which results in slower efflux (Deane et al., 2008). Lipidated ApoE is more effective at transporting Aβ across the BBB than alipidated forms (Jiang et al., 2008). ABCA1 is a membrane protein which lipidates ApoE particles (Lawn et al., 1999) and whose expression is upregulated by LXR and RXR (Beaven and Tontonoz, 2006). Finally, when the tight junctions of the BBB are weakened, Aβ can leak back into the ISF (Clifford et al., 2007). Aβ, amyloid beta; ABCA1, ATP-binding cassette transporter; AD, Alzheimer's disease; ApoE, apopolipoprotein E; APP, amyloid precursor protein BBB, blood brain barrier; IDE, insulin degrading enzyme; ISF, interstitial fluid; LRP1, low density lipoprotein receptor-related protein 1; LXR, liver X receptor; RXR, retinoid X receptor; Vldlr, very low density lipoprotein receptor.

Finally, Aβ binding to lipidated ApoE particles can also be enhanced by the acquisition of sulfatides, sulfated derivatives of galactocerebrosides that are primarily produced by oligodendrocytes and form the major components of the myelin sheath surrounding the axon (Eckhardt, 2008). Enhanced Aβ binding in the presence of sulfatides promotes Aβ clearance through ApoE-mediated endocytosis. In AD patients, the sulfatide content in the CNS is reduced, which is mirrored by a reduction of sulfatides in the CSF (Han et al., 2002; Han et al., 2003). This sulfatide loss is specific to AD, as it is not found in other neurodegenerative diseases such as Parkinson's and dementia with Lewy bodies (Cheng et al., 2003). Lipoprotein receptors play a specific role in this process, as reduction of Lrp1 led to sulfatide loss in a mouse model (Liu et al., 2010).

As described earlier, lipoprotein receptors affect both Aβ production and clearance. A few key mouse and human studies strongly suggest that clearance is the process primarily affected in AD and by ApoE isoforms (Castellano et al., 2011; Mawuenyega et al., 2010). However, it is also important to note that ApoE4 carriers and pre-clinical AD patients have hyperactivity in several brain regions and that increased synaptic activity leads to overproduction of Aβ (Cirrito et al., 2005; Filippini et al., 2009; Quiroz et al., 2010). For review, see (Palop and Mucke, 2010). Whether Aβ buildup causes hyperactivity or vice versa in the early phases of disease is still debated.

Vascular Aβ and Lipoprotein Receptors

Cerebral amyloid angiopathy (CAA) is the deposition of amyloid in brain capillaries, leptomeningal vessels, and arterioles. Most commonly, CAA is the result of deposition of Aβ in the smooth muscle cell layer, which gradually replaces the smooth muscle wall and leaves the vessel more susceptible to hemorrhage (Vinters, 1987). The majority of patients with AD also have CAA, which can independently trigger intracranial hemorrhage and progressive dementia (Ellis et al., 1996; Rensink et al., 2003). Similar to AD, possession of the ApoE4 allele causes patients to be more susceptible to CAA and hemorrhage, a finding which is mirrored in mouse models of AD (Fryer et al., 2005; Greenberg et al., 1995). Interestingly, ApoE2, which is protective for AD, is associated with an increased risk of hemorrhage when CAA is present (Nicoll et al., 1996).

Aβ drains out of the CNS through ISF drainage pathways in the smooth muscle cell layer of arterioles, leptomeningal arteries, and capillaries, which is primarily where Aβ builds up in CAA. Lrp1 is required for clearance of Aβ through smooth muscle cells, since smooth muscle cell-specific knockout of Lrp1 leads to the buildup of both parenchymal amyloid plaques and CAA in a mouse model of AD (Kanekiyo et al., 2012). Smooth muscle cell levels of Lrp1 are reduced in AD (Bell et al., 2009). It is not yet clear if Lrp1 is the only lipoprotein receptor involved in CAA, or if it is the mechanism by which ApoE4 increases CAA risk. For an in-depth review of neurovascular dysfunction in AD, see (Zlokovic, 2011).

ApoE Fragmentation

Full-length ApoE may not be the only ApoE form that is involved in AD pathology. In postmortem AD brains, there is an accumulation of C-terminal-truncated ApoE (Huang et al., 2001). While this could simply be the result of increased degradation of ApoE, it has been shown that the addition of ApoE fragments to neuronal cultures leads to the accumulation of NFTs (Brecht et al., 2004). Cell stress in neurons in vitro results in ApoE fragmentation and the generation of toxic C-terminal-truncated fragments, which can escape the secretory pathway and cause NFT formation (Brecht et al., 2004). ApoE4 is particularly susceptible to proteolysis due to a domain interaction between Arg-61 and Glu-255, which is facilitated by the Arg-112 specific to apoE4 (Dong and Weisgraber, 1996; Dong et al., 1994), and expression of truncated ApoE4 in transgenic mice leads to AD-like neurodegeneration and death (Harris et al., 2003). These studies, reviewed in (Mahley and Huang, 2012), indicate a role for ApoE fragments as toxic particles that may cause τ aggregation (Brecht et al., 2004), rather than just as markers of ApoE degradation. It is important to point out that ApoE expression is normally absent in neurons but occurs under stress conditions (Xu et al., 2008), and the fragmentation hypothesis requires ApoE to be present in the neuron at time of injury in order for proteolysis to occur.

The Common Thread: ApoE4 and Endocytic Recycling

The seemingly disparate effects of ApoE4 – altered APP processing, reduced Aβ clearance, and impaired synaptic plasticity – share a common mechanism: alterations in endocyctic recycling. To understand this finding, one must remember that the role of lipoprotein receptors and apolipoproteins in the periphery is to transport cholesterol- and triglyceride-rich lipoproteins. In hepatocytes, ApoE-containing particles are taken up primarily through LRP1 and LDLR (Beisiegel et al., 1989; Rohlmann et al., 1998). While cholesterol and ApoB are transported to the lysosome, ApoE recycles back to the surface (Rensen et al., 2000). However, ApoE4 recycles poorly in comparison to the other isoforms, leading to its intracellular accumulation (Heeren et al., 2004). The reason for this may be that at the lower pH that prevails in endosomes, ApoE4 is more likely to form a molten globule (Morrow et al., 2002). The altered structure of ApoE4 at low pH may make it more prone to aggregation, leading to disrupted retroendocytosis.

As described in the previous sections, the disruption of endocytic recycling by ApoE4 plays a key role in AD. First, ApoE4 has been reported to promote colocalization of APP and BACE1 in early endosomes, which may lead to increased processing of APP (Rhinn et al., 2013). Second, ApoE4 promotes intra-neuronal buildup of Aβ by affecting its trafficking (Zhao et al., 2013). Finally, ApoE4 causes reduced surface expression of Reelin receptors, blocking Reelin's ability to protect the synapse against Aβ toxicity (Chen et al., 2010). While the relative contributions of these three mechanisms to AD pathogenesis remain to be determined, these data nevertheless clearly show that trafficking presents an appealing target for AD prevention. Furthermore, a recent genomic study showed that gene expression changes in endocytic trafficking molecules were common to both ApoE4 carriers and LOAD patients (Rhinn et al., 2013). One of the identified trafficking proteins was SV2A, a synaptic vesicle protein. Treatment of human fibroblasts with Levetiracetam, an inhibitor of SV2A, reversed the endosomal trafficking dysfunction and upregulated APP processing caused by ApoE4, but had no effect in the presence of ApoE3 (Rhinn et al., 2013). Levetiracetam also reversed behavioral and electrophysiological deficits in the hAPP mouse model of AD, and it reversed the abnormal hippocampal hyperactivity found in mild cognitive impairment (MCI) patients (Bakker et al., 2012; Sanchez et al., 2012). Levetiracetam is currently used clinically for the treatment of epilepsy, where it has several significant side effects, including mood disorders and excessive sleepiness (Bootsma et al., 2007). It is currently unknown if these are on- or off-target effects, and while inhibiting SV2A with Levetiracetam may not be the ideal therapeutic strategy, taken together with other findings, these results strongly indicate that restoring endocytic recycling is an attractive strategy for preventing AD.

Insights and Developments in Diagnostics

The abundance of molecular data on the role of ApoE receptors and Aβ in AD is now setting the stage for increasingly rational and sophisticated diagnostic and therapeutic approaches. In our clinics today, the prediction, diagnosis, and monitoring of Alzheimer's disease remains inexact. Drug testing and development occurs either in patients with advanced disease, for whom interventions may be too late, or in a cohort of as yet unaffected individuals, which necessarily then must be very large to achieve statistical significance. Similarly, drug efficacy is primarily determined by cognition scores, which may take months or years to change significantly compared to the placebo control group. It is thus important to develop new ways to monitor disease, and the clinical evaluation of AD is rapidly being transformed by the development of new imaging techniques and biomarkers.

Historically, the diagnosis of AD could only be confirmed after death via detection of amyloid plaques and neurofibrillary tangles. Through recent advances in non-invasive imaging technology, amyloid deposition can now be observed in living patients using positron-electron tomography (PET). Patients are injected with Pittsburgh B (PIB) or similar compounds, which are PET tracers that bind to amyloid in the brain (Klunk et al., 2004). Studies of PIB imaging have shown that the PIB signal roughly follows what is already known about amyloid deposition from postmortem tissue. In cognitively normal individuals, no PIB signal is present until around 50 years of age, at which point it starts to appear in a percentage of the population. Once individuals begin to be PIB positive, the amount of signal increases over time (Morris et al., 2009). As the PIB signal crosses a vague threshold, cognitive decline starts to occur, with greater PIB signal predicting greater cognitive loss. Since PIB signal closely shadows amyloid deposition, one can examine risk factors, modifiers, and treatments for AD in living patients. For example, ApoE4 carriers have higher PIB signals at younger ages, while ApoE2 carriers have lower signals (Reiman et al., 2009). The PIB signal also increases at a faster rate in ApoE4 carriers (Grimmer et al., 2010).

While PIB PET reveals the static deposition of amyloid, FDG-PET and fMRI are able to detect functional loss in the affected areas of brain. FDG-PET measures glucose metabolism in the brain, while fMRI can be used to look at blood flow, oxygen consumption and network connectivity. These tools, reviewed in (Perrin et al., 2009), have shown that changes in connectivity and brain function occur long before plaque deposition and cognitive decline in ApoE4 carriers. Cognitively normal ApoE4 carriers already have decreased regional volume by MRI in those brain areas that are the first to be affected in AD, as well as decreased glucose metabolism by FDG-PET (Chen et al., 2012). Similarly, young, PIB negative, ApoE4 carriers already have some fMRI connectivity deficits that are similar to those found in AD patients (Sheline et al., 2010). These functional alterations may begin very early in development, since ApoE4-carrying infants as young as 2 months of age exhibited reduced white matter myelin water fraction (MWF) and gray matter volume (GMV) in several areas affected in AD (Dean et al., 2014). Additionally, ApoE4 carriers show greater hippocampal activation during memory encoding tasks and greater activation of the “default mode” network at rest (Filippini et al., 2009). Since neuronal activity leads to increased levels of Aβ, increased hippocampal activity may be both, driver and first indicator of amyloid accumulation in these patients.

The availability of these new imaging tools has led to important prospective studies in a large familial AD cohort in Colombia. Individuals in this cohort harbor the E280A mutation of PS1, which shifts APP processing towards generation of longer Aβ species, particularly Aβ42. These individuals follow a temporally fixed path of disease development (Van Vickle et al., 2008). Disease onset occurs within a narrow window around 45 years of age, presenting initially with memory loss and then progressing to include personality changes, loss of language, and finally gait abnormalities (Acosta-Baena et al., 2011; Lopera et al., 1997). Since disease progression is so well characterized in the Colombian cohort, we can study neurological changes that occur before the onset of clinical disease. For example, 15 years prior to disease onset, carriers of the PS1 mutation have increased hippocampal activation despite similar performance during memory tasks relative to healthy controls (Quiroz et al., 2010). The Colombian cohort has already been used to update AD diagnostic criteria in the clinic. Studies in pre-dementia E280A carriers identified new cognition tests that could indicate pre-clinical dementia, including the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) memory test, the naming of famous faces, and semantic variation (Arango-Lasprilla et al., 2007; Cuetos et al., 2007; Tirado et al., 2008). This year, clinical trials using anti-a1yloid therapeutics will begin in pre-dementia individuals of the Colombian cohort.

The distinction between “early-onset” AD, found in the Colombian cohort, and “late-onset” AD, found in the majority of patients, has a well-understood mechanistic basis. Both diseases involve elevated levels of Aβ, which through mechanisms that are not yet entirely clear triggers tau pathology, cognitive impairment and ultimately results in neurodegeneration. However, both forms of the same disease, i.e. AD, differ mechanistically in that familial AD involves APP mutations that increase Aβ production, whereas ApoE4 impairs clearance of Aβ. Moreover, ApoE4 appears to alter brain function already early in life (Dean et al., 2014), which may leave certain brain regions more susceptible to Aβ toxicity (Figure 4).

Figure 4. ApoE4 and AD Progression.

ApoE4 carriers have alterations in brain structure and function starting early in life, most notably in AD related areas. As early as infancy, MRI reveals that there is reduced gray matter volume in poster/middle cingulate, lateral temporal, and medial occipitotemporal regions in ApoE4 carriers, a trend that continues throughout life (Dean et al., 2014). These findings are mirrored in middle age by reduced glucose metabolism in the same areas in ApoE4 carriers, prior to the onset of cognitive dysfunction (Protas et al., 2013). With age, ApoE4 carriers have earlier deposition of Aβ plaques compared to ApoE3 carriers, which is then followed by the development tau tangles and frank atrophy. The altered baseline brain function and accelerated Aβ deposition combine to cause ApoE4 carriers to have both an earlier age of disease onset and a more rapid decline in cognitive function (Caselli et al., 2009). Aβ, amyloid beta; AD, Alzheimer's disease; ApoE, apolipoprotein E; MRI, magnetic resonance imaging.

Alzheimer's disease is a complex disorder whose underlying molecular and network dysfunctions continue to emerge. It is clear that ApoE4 and lipoprotein receptors have pivotal roles in AD pathogenesis by altering cellular trafficking and Aβ clearance on one hand, as well as affecting synaptic plasticity on the other. With so many interacting proteins and higher order network systems that are affected as the disease progresses, it is vital to take an interdisciplinary approach to reduce this unwieldy complexity to its root causes. We are at a thrilling time in Alzheimer's research as new molecular insights into the disease mechanisms reveal novel neurobiological principles, and promising therapeutic approaches feed the desperately needed drug pipeline.