Amyloid precursor protein (APP) is a cell-surface protein that has been well studied for its role in Alzheimer’s disease (AD), a progressive neurodegenerative disorder that is the most common form of dementia in the elderly. In 1906, Alois Alzheimer described a case of presenile dementia in which amyloid plaques were present in the brain. Later, the primary component of these plaques was identified as a small protein containing 36–43 amino acid residues (Glenner and Wong 1984; Masters et al. 1985), which has since been named the β-amyloid protein (Aβ). Aβ is produced by proteolytic cleavage of APP (Kang et al. 1987). Mutations in APP that enhance the production of specific Aβ species have been shown to accelerate AD progression (Van Cauwenberghe et al. 2016). In addition, triploidy of chromosome 21 (Down syndrome), which results in an extra copy of the APP gene, leads to amyloid accumulation and AD-like pathology (Masters et al. 1985). Most recently, a mutation at codon 674 in the human APP gene has been identified that reduces the production of Aβ. This mutation has been shown to provide a protective effect against AD (Jonsson et al. 2012).
The role of APP in AD, and the regulation of APP cleavage to generate Aβ, has dominated AD research. The normal physiological function of APP and its mechanism of action are much less well understood. However, a number of studies indicate that APP has an important role in cell adhesion, growth, and cell signaling.
The APP Gene and Its Expression
Gene Structure and Alternative Splice Variants
The APP gene is located on the long arm of human chromosome 21 (21q21.3). The gene contains 20 exons that span approximately 290 kb and encodes an integral membrane protein of 770 amino acid residues with molecular mass of 87 kDa. The full-length APP mRNA is subject to alternative splicing and ten transcript variants have been identified. Three main splice variants expressed in the human brain encode the main isoforms: APP770, APP751, and APP695 (the nomenclature of the isoforms is based on the number of encoded amino acid residues). APP770, which contains all exons, and APP751, which lacks only exon 7, are expressed predominantly in nonneuronal cells within the central nervous system, whereas APP695, which lacks both exons 7 and 8, is expressed predominantly in neurons. L-APP, which lacks exon 15, is expressed by immune cells and astrocytes. Removal of exon 15 creates a consensus sequence for chondroiton sulfate attachment at the exon 14/16 boundary and can result in the expression of a proteoglycan form of APP known as appican.
APP is expressed by a variety of tissues, with high levels of expression in the brain. The APP gene is also transcribed at high levels in the human kidney, thyroid, and adrenal glands (Liu et al. 2008). The promoter sequence suggests that APP has a housekeeping function. It contains regulatory elements involved with the control of cell growth and maturation (Dawkins and Small 2014).
The APP Gene Family
The APP gene sequence has been conserved through evolution. Orthologous genes have been identified in mouse, frog, chick, fish, and other vertebrate species. In humans, there are also genes encoding two amyloid precursor-like proteins (APLP1 and APLP2). These genes are also present in some other vertebrates. APP therefore belongs to a small multigene family whose members have structural similarities, although neither APLP1 nor APLP2 contains the Aβ sequence. Phylogenetic analysis indicates that the APP gene family has arisen through two gene-duplication events during evolution. APP and APLP2 diverged most recently (Coulson et al. 2000). Consistent with the idea that these two genes may be related both structurally and functionally is the finding that a combined knockout of APP and APLP2 in mice is lethal around birth, whereas mice with a combined knockout of APLP1 and APP do not develop a phenotype distinguishable from that of a knockout of APP alone. Mice with a single deletion of APP, APLP1, or APLP2 alone are all viable and fertile with subtle phenotypes (Aydin et al. 2012).
Invertebrates Caenorhabditis elegans and Drosophila melanogaster also contain APP-like genes, termed apl-1 and Appl respectively. These genes are likely to be derived from a common ancestor of vertebrate and nonvertebrate families. Although apl-1 and Appl are more highly diverged from their mammalian orthologs than APP in lower vertebrates, the APP-like proteins in the worm and fly share the same basic motif structure of human APP and show a degree of functional conservation (Coulson et al. 2000). Yeast has no APP gene family members.
Structure of APP
E1 and E2 Domains
The cysteine-rich E1 domain close to the N-terminus can be subdivided into two distinct regions: a heparin-binding domain (HBD), and a copper/metal-binding domain (Rossjohn et al. 1999). The HBD may interact with heparan-sulfate proteoglycans, which are macromolecules of the cell surface and extracellular matrix involved with the control of neurite outgrowth, adhesion, and growth-factor presentation. Adjacent to the HBD is a hydrophobic pocket that could form a ligand-binding or dimerization site. Indeed, dimerization of the complete E1 domain can be induced by heparin (Dahms et al. 2010). The HBD is adjacent to a metal-binding domain that is able to bind and reduce a single copper ion. The physiological role of copper binding is uncertain, but APP reportedly dimerizes in a copper-dependent manner as well (Baumkotter et al. 2014). Adjacent to the E1 domain is an acidic region rich in glutamate and aspartate residues. This region has also been proposed to be involved in the regulation of APP dimerization (Hoefgen et al. 2014). The E2 domain is rich in α-helices and contains a second putative HBD and two putative metal-binding sites. Similar to the E1 domain, heparin binding to the E2 domain can readily induce dimerization, suggesting the involvement of this region in APP homodimer formation.
KP1 and Ox-2 Antigen Domains
Within the flexible linker that separates E1 and E2 domains, the APP751 and APP770 isoforms contain a Kunitz-type serine protease inhibitor (KPI) domain that is not found in APP695. APP isoforms that contain the KPI domain are highly expressed in platelets where they may influence wound repair by regulating blood clotting serine proteases. This domain is also present in APLP2, but not in the more evolutionarily distant APLP1. Adjacent to the KP1 domain is an Ox-2 antigen domain. The Ox-2 domain is a 19 amino acid residue insert that is similar to the Ox-2 antigen, a cell-surface glycoprotein with homology to immunoglobulin light chains that is found on lymphocytes and neurons. This region may be involved in cell-surface binding and recognition.
Aβ Domain and Transmembrane Region
The Aβ domain extends from the APP ectodomain, close to the membrane, into the transmembrane region. There is poor amino acid sequence conservation in this region within the APP gene family. Indeed, only APP contains the Aβ sequence. Residues within the transmembrane region bind to cholesterol, a structural component of the plasma membrane that is enriched in lipid rafts (Barrett et al. 2012).
The short cytoplasmic domain, known as the APP intracellular domain (AICD), contains 47 amino acid residues that do not adopt a well-ordered structure when unbound to ligands, but can adopt regions of stable structure when bound to a range of intracellular proteins. Interestingly this region is the most highly conserved domain suggesting that it plays an important role in APP function. The region contains a YENPTY protein interaction motif that is completely evolutionarily conserved across the APP family, from C. elegans to humans. This motif ensures correct trafficking of APP within the cell and may facilitate clathrin-mediated endocytosis. AICD may also play an important role in synaptogenesis and neural plasticity.
Processing of APP
Posttranslational Modifications, Trafficking, and Endocytosis
APP is posttranslationally modified in the Golgi apparatus. Posttranslational modifications include N- and O-linked glycosylation, tyrosine sulfation, phosphorylation, and palmitoylation. Some of the modified protein may then be trafficked to the cell surface, while some molecules may remain inside the cell. In neurons, vesicular APP undergoes kinesin-mediated fast anterograde transport to the synapse. Cell-surface APP can be rapidly internalized by clathrin-mediated endocytosis. The resulting APP-containing endosomes either recycle APP back to the cell surface, or fuse with lysosomes, after which APP is degraded. In many cells, APP is rapidly turned over with a half-life in the order of hours.
APP is proteolytically cleaved under normal physiological conditions. The two best-characterized proteolytic pathways involve an initial cleavage by either α- or β-secretase (Fig. 1b). These cleavages occur at well-defined positions close to the transmembrane domain and lead to the shedding of the majority of the ectodomain. Members of the disintegrin metallopeptidase (ADAM) family of proteases act as α-secretases. In neurons, the principal α-secretase is ADAM10. Cleavage of APP by α-secretase releases a secreted ectodomain, sAPPα, which is reported to possess neurotrophic properties. Alternatively, the cleavage of APP by β-secretase (β-site APP cleaving enzyme 1 or BACE1), a protease abundantly expressed in the brain, releases the ectodomain fragment, sAPPβ.
The initial cleavage of APP triggers a second cleavage catalyzed by a complex of membrane proteins referred to as the γ-secretase. This second cleavage occurs within the transmembrane domain and releases a small N-terminal fragment (either Aβ or p3 depending upon whether APP is cleaved by β- or α-secretase, respectively) and a C-terminal fragment, AICD. The α-secretase pathway precludes Aβ formation. In contrast, the β-secretase pathway results in Aβ formation. It is currently unclear how much either pathway contributes to APP turnover in vivo.
The position of the final cleavage by γ-secretase is not restricted to a single site. Cleavage appears to be stepwise and not site-specific, with cleavages occurring at sites termed, γ, ζ, and ε, which are separated by approximately three amino acid residues. The Aβ protein typically ranges in size from 37 to 43 amino acid residues.
Functions of APP
APP has a putative role in cell-cell and cell-substrate adhesion. In particular, APP can bind to heparan-sulfate proteoglycans found both on the surface of cells and in the extracellular matrix. Thus binding occurs through sites in the E1 and E2 domains. These domains can also bind collagen type I and laminin, and can reportedly direct neurite outgrowth through binding to β-1 integrin and neural cell adhesion molecule 140 (NCAM-140). Modification of carbohydrate residues at the chondroitin sulfate attachment motif in the L-appican isoform is also believed to influence cell-adhesion (Dawkins and Small 2014).
Cell migration is closely linked with cell adhesion, and deletion of APP impairs cell motility, both in neurons and peripheral cells, such as macrophages. APP KO mice are viable and fertile, albeit with a 15–20% decrease in body weight and 10% reduction in brain weight compared with wild-type animals. This subtle phenotype reflects a high degree of functional redundancy within the APP family. A triple knockout of APP, APLP1, and APLP2 in mice that deletes E1 and E2 domains results in perinatal lethality and lissencephaly, potentially as a result of abnormal neuron migratory behavior during colonization of the developing embryonic cortex (Aydin et al. 2012). It has been suggested that the interaction of APP with pancortins is responsible for modulating migration (Rice et al. 2012).
APP has been implicated in the regulation of neurite outgrowth, which is a process that involves the formation of membranous protrusions that subsequently elongate under the regulation of complex signaling events. The subsequent formation of connections between neurons (synaptogenesis) is carefully regulated. APP expression during development of the embryonic nervous system correlates with periods of intense neurite outgrowth and synaptogenesis, and it is now generally accepted that APP plays a role in both processes within the brain. For example, strong APP expression is observed in neurites during their extension into the olfactory bulb and subsequent synapse formation (Clarris et al. 1995). APP is highly expressed in olfactory receptor neurons and in the hippocampus, which are sites of continued neurogenesis in the adult. This expression probably reflects a demand for growth, cell proliferation, and synaptogenesis in these regions (Wang et al. 2009).
Increased APP expression in cultured neurons stimulates neurite outgrowth and enhances synaptogenesis. Neurons cultured from APP KO mice display abnormal neurite outgrowth, characterized by hyperextension of the longest neurite and abnormalities in others. APP knockout mice have a lower synaptic density than wild-type mice, suggesting a requirement for APP in synapse formation. The knockout mice also have locomotor and grip-strength defects that may result from deficits at the neuromuscular junction, as well as memory impairments associated with decreased long-term potentiation (LTP), and an increased sensitivity to kainic acid-induced seizures. Within the dendritic spine, APP may be needed for synaptic vesicle release during neurotransmission, and for activation of N-methyl-D-aspartate (NMDA) excitatory receptors (Priller et al. 2006). Transgenic animals that overexpress human mutant APP consistently demonstrate synaptic impairment. However this synaptic impairment could conceivably be related to the production of a toxic form of Aβ.
APP may act as a synaptic adhesion molecule in neurons. The involvement of APP with synaptic adhesion mechanisms is thought to be mediated through regulation of actin dynamics and extracellular binding (Müller and Zheng 2012). APP is also implicated in the process of neurogenesis. Strong evidence indicates that APP stimulates proliferation of neuronal stem/progenitor cells (NSPCs), and APP may also regulate their subsequent differentiation into neurons. APP-stimulated proliferation requires production of the cysteine protease, cystatin C, whereas the differentiation effect requires the transcription factor neurogenin 2 (Dawkins and Small 2014).
Proposed Mechanism of APP Action
It is unclear whether the function of APP in vivo is mediated by the full-length molecule or via one or more of its cleavage products that include the secreted ectodomain and the short AICD. The structure of APP suggests that it could be a receptor for extracellular ligands (Fig. 1a) (Deyts et al. 2016). Indeed, APP bears similarities to the developmentally expressed receptor, notch-1, which is also proteolytically processed by γ-secretase in response to ligand binding. There is evidence that APP could be a receptor for soluble cues such as netrin or F-spondin (Ho and Sudhof 2004; Lourenco et al. 2009). However, no ligands have been definitively confirmed as being important for APP function in vivo.
APP homodimerization may be involved in cell signaling. Dimerization-induced receptor activation is a process employed by many cell-surface receptors, and cis-dimerization of APP could promote clustering. In contrast, transdimerization of full-length APP could mediate both cell-cell and transsynaptic adhesion. There is currently limited evidence to suggest that APP normally acts as a receptor in either a monomeric or homodimeric form in vivo. Rather, it has been proposed that APP-mediated signaling occurs through the binding of coreceptors, of which a number have been identified as candidates. These putative coreceptors include members of the lipoprotein receptor protein family, LRP1 and LRP2 (megalin), the related type 1 transmembrane protein alcadein, the low-density lipoprotein receptor-related, sortilin-related receptor, Nogo-66 receptor, and notch2 (Deyts et al. 2016). The significance of these interactions with APP is still unclear. The large range of proposed binding partners suggests that APP could act as a modulator of many functions, with the activation of specific signaling cascades dependent on the suite of coreceptors bound to APP.
The secreted ectodomain of APP that results from α-secretase cleavage (sAPPα) may act like a trophic factor to facilitate neuronal development, survival, and maintenance of connections with neighboring neurons. Addition of sAPPα to cultured cells or its expression in vivo is reported to rescue phenotypic effects arising from deletion of APP, suggesting that the secreted APP ectodomain has a biological function.
The YENPTY motif in the AICD is an important site for protein interaction, as it is a motif found in many receptor tyrosine kinases, nonreceptor tyrosine kinases, low-density lipoprotein receptor related proteins, and integrins (Dawkins and Small 2014). More than 20 proteins are reported to bind the YENPTY motif, notably those proteins having domains that bind phosphotyrosine. A protein that is particularly well studied is the brain enriched adaptor protein, Fe65, which binds and stabilizes the relatively labile AICD, and can alter APP processing (McLoughlin and Miller 2008). Other proteins that can bind AICD include the adaptor proteins Dab1, JIP, Numb, X11/mint family members, several kinases, and Grb2 (Deyts et al. 2016).
Different models for AICD’s putative role in APP-mediated signaling can be proposed. One model would involve a conformational change in the AICD that releases a bound cytosolic adaptor protein. This adaptor protein could translocate to the nucleus and modify transcription. In this scenario, AICD would remain embedded in the membrane. Interaction with binding partners could affect the conformation of the C-terminus and change the phosphorylation status of the AICD. Phosphorylation has been shown to induce structural changes that alter the AICD’s affinity for adaptor proteins. In another model, AICD would be released from the membrane and translocate to the nucleus, either alone, or bound to an adaptor protein, where it could form a transcription regulatory complex on the DNA. Indeed, AICD is analogous to the notch-1 intracellular domain (NICD), which is released from notch-1 upon γ-secretase cleavage and which is translocated to the nucleus to activate transcription. Currently, evidence for the presence of AICD in the nucleus in vivo is limited, and the biological relevance of AICD to endogenous signaling is unclear. Putative targets of APP-mediated transcriptional modification include KAI, GSK3β, neprilysin, EGFR, p53, LRP, APP itself, and various genes involved in cytoskeletal dynamics and calcium regulation (Müller and Zheng 2012).
It is unclear whether Aβ has a function. A variety of putative receptors for Aβ have been reported, including the receptor for advanced glycosylation end products (RAGE) on microglia. However, as Aβ binds hydrophobically to many proteins in a nonspecific manner, it is unclear which, if any, of these molecules may be a receptor for Aβ in vivo.
APP is a type-1 transmembrane protein expressed on the plasma membrane that can be processed by regulated intramembrane proteolysis. Proteolytic processing of APP can give rise to a secreted ectodomain sAPPα with putative trophic properties, and a short intracellular domain AICD implicated in gene regulation. APP is likely to have an important function in neurite outgrowth, synaptogenesis, neurogenesis, cell adhesion, and the control of gene transcription, and is implicated in processes that regulate these functions, such as cell membrane motility.
A considerable amount of research is ongoing to determine the function of APP and that of its proteolytic fragments. A more complete understanding of APP function may be important for the development of AD therapies aimed at lowering Aβ production, as these approaches have the potential to perturb the normal physiological role of APP.