The human β-amyloid precursor protein: biomolecular and epigenetic aspects

APP structure

The human APPgene is located on chromosome 21 (21q21.2-3), spans approximately 240 kb and contains at least 18 exons (5, 6) (Figure 1A). Structurally, APP has features of an integral type-I transmembrane glycoprotein. The size of the putative binding domain at the N-terminus suggests that APP may act as a cell-surface receptor for a ligand (1), act as a growth factor (14) or bind to an extracellular matrix component such as proteoglycan (15). The APP-promoter sequence indicates that the APP gene belongs to the class of housekeeping genes. The promoter lacks typical TATA and CAAT boxes but contains consensus sequences for the binding of a number of transcription factors, including SP-1, AP-1 and AP-4 sites, a heat-shock control element and two Alu-type repetitive sequences (16–18). The presence of SP-1, AP-1, and AP-4 sites in the APP promoter regulates, the expression of proteins associated with cell proliferation and mitosis as well as cell differentiation and suggests that APP has a function related to cell growth or maturation. Structurally, APP can be divided into three domains: APP extracellular domain (EC) (exons 1–17: amino acid residues 1–699), APP transmembrane domain (TM) (exons 17–18: amino acid residues 700–723) and APP intracellular domain (IC) (exons 17–18: amino acid residues 724–770). Of interest, the Aβ sequence (exons 16–17: amino acid residues 672–713) is not conserved and is unique to APP. The encoded protein contains a large N-terminal extracellular domain that includes a signal-peptide sequence (SP) (amino acid residues 1–18), a cysteine-rich globular domain (E1) (exons 1–5: amino acid residues 18–190), an acidic domain, a Kunitz-type protease inhibitor (KPI) domain, a helix-rich domain (E2) (exons 9–14: amino acid residues 366–568) and part of the Aβ sequence, which extends into the transmembrane domain. The relatively short intracellular domain contains the C-terminus, which has some phosphorylation sites and a YENPTY-sorting motif (Figure 1B).

Figure 1:

Structure of the amyloid precursor protein (APP): APPgene, APP₇₇₀ structure and major mRNA splice variants.

(A) APPgene structure. The APP gene is located on chromosome 21q21.2-3, spans approximately 240 kb and contains at least 18 exons. Exon 13a (highlighted in black) denotes a novel exon containing a 59-nucleotide coding region and a 393-nucleotide untranslated region, as published by de Sauvage and Octage [see Ref. (6)]. (B) Protein structure (APP₇₇₀). APP can be divided into three domains: APP extracellular domain (EC), APP transmembrane domain (TM) and APP intracellular domain (IC). APP has an N-terminal signal peptide (SP); E1 domain with a heparin-binding domain (HBD1), a copper-binding domain (CuBD); acidic region; APP₇₅₁ and APP₇₇₀ contain a Kunitz protease inhibitor (KPI) domain located in exon 7 and an Ox-2 antigen domain; E2 domain with a second heparin-binding domain (HBD2); Between the E2 and Aβ regions are two potential N-linked glycosylation sites (CHO). The amino-acid sequence of the Aβ region is shown along with the secretase cleavages sites: BACE cleaves APP after Met₆₇₁ (β) and Tyr₆₈₁ (β′), whereas ADAM10 [best candidate α-secretase, see Ref. (44)] processes APP within the Aβ peptide sequence after Lys₆₈₇ (α), thereby generating the p3 peptide. Gamma-secretase cleavage in the transmembrane region (TM) involves presenilin 1 (PS1) and generates Aβ peptides of mainly 40 and 42 amino acid residues long (γ₄₀) and (γ₄₂). PS-dependent proteolysis also occurs at other positions including the ε-site (Aβ₄₉) downstream of the γ-site proximal to the membrane-intracellular boundary (for details concerning the γ- ε- and ζ-sites, see text). Asterisks indicate the locations of mutations in APP causing early-onset of familial AD (FAD) (see AD Mutation Database, http://molgen-www.uia.ac.be/ADMutations). The intracellular (IC) C-terminal domain contains a YENPTY sorting motif. (C) Major mRNA splice variants. Differential mRNA splicing of exons 7, 8 (highlighted in black) can lead to the expression of 695, 751 and 770 amino acid isoforms.

E1 domain and acidic region

The crystal structure of the E1 domain shows similarities to known cysteine-rich growth factors, and so it is termed ‘growth factor-like domain’ (GFLD) (14). The E1 domain is divided into two distinct regions, the heparin-binding domain (HBD1) and the copper/metal-binding domain (CuBD) (Figure 1B). HBD1 is composed of a single α-helix and an anti-parallel β-sheet, with a loop rich in basic residues (96–110) that bind to heparin (14, 15). Immediately adjacent to the HBD1 is a hydrophobic pocket, which could form either a protein-binding site or a dimerization site (14). Adjacent to the HBD1 is the copper/metal-binding domain, which contains a single α-helix and a short β-sheet. This region can bind several metal ions (19). The role of this domain is unclear, but it has been suggested that copper (II)-binding and reduction may be a principal function (20). On the C-terminal side of the E1 domain is an acidic region of unknown significance that is rich in glutamic acid and aspartic acid residues (Figure 1B).

KPI and Ox-2 antigen domains

Longer isoforms of APP (APP₇₇₀ and APP₇₅₁) may contain a Kunitz-type protease inhibitor (KPI) domain located in exon 7 and an Ox-2 antigen domain (Figure 1B). APP isoforms containing the KPI domain are more commonly expressed in non-neuronal cells (21). KPI-containing APP isoforms are highly expressed in platelets, where they can influence wound repair by regulating blood clotting serine proteases (22). As serine proteases are also implicated in neuronal cell growth (23), it is possible that KPI-containing APP isoforms regulate cell growth by inhibiting one or more of these proteases. Concerning the Ox-2 domain, this is an insert of 19 amino acid residues that is similar to a region of the Ox-2 antigen. The Ox-2 antigen is a lymphoid and neuronal cell-surface glycoprotein that has homology to Thy-1 and immunoglobulin light chains (24). As immunoglobulin loop domain are commonly found in cell-surface receptors and involved in cell-surface binding and recognition, it seems likely that the Ox-2 domain in APP has a similar function.

E2 domain

The E2 domain is an α-helix rich region that can readily dimerize (25) and may therefore be involved in APP self-association. Specially, X-ray analysis has revealed that the E2 domain of APP could form parallel or antiparallel dimers (26), and the latter structure implies that there is a potential to function in trans-cellular adhesion. Indeed, cell-culture studies support the homo-or hetero-dimers formation of APP family members, and trans-dimerization has been shown to promote cell-cell adhesion (27). It has been further demonstrated that heparin binding to the E1 or E2 region would induce the formation of APP dimerization (28). Downstream of the E1 and E2 regions, a ‘RHDS’ motif in the extracellular domain of APP within the Aβ sequence also appears to promote cell adhesion (Figure 1B). It is believed that this region acts in an integrin-like manner by homology to ‘RGD’ sequences (29). In this regard, it is interesting that APP co-localizes with integrins on the surface of axons and at sites of adhesion (30, 31). The E2 domain has also a heparin-binding site (HBD2) (32, 33) (Figure 1B) as well as a number of putative metal-binding sites that may hold the E2 domain in a rigid conformation (34). Between the E2 domain and the Aβ region are found two potential N-linked glycosylation sites (CHO) at residues 542 and 571 (Figure 1B).

Aβ, transmembrane domain and intracellular domain

The Aβ region on the C-terminal side of the E2 domain lies partly within the extracellular domain and partly within the transmembrane domain (Figure 1B). The Aβ peptides vary in size from 39 to 42 amino acids, and Aβ_1–42 aggregates more readily than the other molecules (35). A GxxxG sequence motif within the transmembrane domain has been implicated in homodimerization (36) and in cholesterol-binding (37) (Figure 1B). The intracellular domain contains a YENPTY sorting motif located between residues 757 and 762 of the APP₇₇₀ isoform (Figure 1B). This YENPTY sorting motif interacts with several proteins containing phosphotyrosine binding-domains (PTB-domain), including Mint/X11 adaptor proteins. This motif is involved in the facilitation of clathrin-mediated endocytosis and is present in many tyrosine receptor kinases, non-receptor tyrosine kinases, low-density lipoprotein-receptor-related family proteins and integrins (38, 39). Consistent with this role, many studies have demonstrated that the YENPTY motif in APP is involved in the regulation of its trafficking and endocytosis (40–42).

APP expression

During transcription, alternative splicing generates APP mRNAs encoding several isoforms that range from 365 to 770 amino acid residues. The major expressed isoforms of APP have 695, 751 or 770 amino acid residues (Figure 1C). APP₇₅₁ and APP₇₇₀ contain a domain homologous to the Kunitz-type serine protease inhibitors (KPI) located in exon 7 of the extracellular sequences (Figure 1B and C), and these isoforms are commonly expressed in non-neuronal cells (21). APP₆₉₅ isoform lacks the KPI domain (Figure 1C) and is predominantly expressed in neurons and accounts for the primary source of APP in brain (43). The raison and functional significance for this apparent tissue-specific alternative splicing is poorly understood.

APP processing

APP is processed in the constitutive secretory pathway and is post-translationally modified by N- and O-glycosylation, phosphorylation and tyrosine sulfation (44). Full-length APP is sequentially processed by at least three proteinases, termed α-, β- and γ-secretases (Figure 2). Cleavage by α-secretase or β-secretase within the luminal/extracellular domain results in the shedding of nearly the entire ectodomain to yield large soluble APP derivatives (called APPsα and APPsβ, respectively) and generation of membrane-tethered α- or β-carboxyl-terminal fragments (APP-CTFα and APP-CTFβ). The APP-CTFs are subsequently cleaved by γ-secretase to generate either a 3-kDa product (p3, from APP-CTFα) or Aβ (from APP-CTFβ), and the APP intracellular domain (AICD) (Figures 1B and 2). The major neuronal β-secretase is a transmembrane aspartyl protease, termed BACE1 (β-site APP cleaving enzyme 1 or β-amyloid cleaving enzyme 1; also called Asp-2 and memapsin-2) (45–48). There is an alternative BACE (β′) cleavage site following Glu at position+11 of the Aβ peptide (49) (Figure 1B). In addition, there is a BACE2 homolog (β-site APP cleaving enzyme 2) that is expressed widely but does not appear to play a role in Aβ generation as it appears to cleave near the α-secretase site (50, 51). Whereas cleavage at the β-site is specific to BACE1, it was initially believed that a number of proteases, specifically members of the ADAM (a disintegrin and metalloprotease) family of proteases including ADAM9, ADAM10 and ADAM17, are candidates for the α-secretase (52). However, it should be noted that cleavage of APP by α-secretase processing only precludes the formation of an intact full length Aβ peptide. Although this latter event is commonly called the non-amyloidogenic pathway, that is unfortunately a bit of a misnomer because truncated Aβ (p3 peptide) from 17–42 (Figures 1B and 2) is also deposited in the brain of AD and Down syndrome patients (53, 54), indicating that shorter Aβ peptides starting at the α-secretase site may contribute to some aspects of AD-associated amyloid pathology (55, 56). Regarding γ-secretase cleavage that releases Aβ from the membrane, this activity is executed by a high molecular weight complex consisting of presenilin PS1 or PS2, nicastrin, anterior pharynx defective phenotype (APH1) and presenilin enhancer 2 (PEN2) (57, 58). Although these four proteins form the mature γ-secretase complex, it appears that the core γ-secretase activity resides within presenilin itself, functioning as an aspartyl protease (59–61). Gamma-secretase cleavage is a type of regulated intramembrane proteolysis (RIP), as cleavage occurs in the middle of the transmembrane domain (62). In addition to generating Aβ peptides of different lengths, RIP of APP is thought to occur as a series of cleavages, starting from the C-terminal end of the substrate and moving toward the N-terminal region of the transmembrane domain. These cleavage sites have been termed the γ- ε- and ζ-sites (62) (Figure 1B). In addition to the γ-cleavage that yields Aβ 40 and 42, PS-dependent proteolysis also occurs at other positions including the ε-site (Aβ 49) downstream of the γ-site proximal to the membrane-intracellular boundary (63–65). Recent data provide support for a sequential cleavage model in which ε-cleavage serves as the initial cutting site followed by ζ- and γ-processing within the membrane (65–67). Thus, the ε-cleavage of APP may represent the primary PS-dependent processing event. This is important because this cleavage releases AICD and more than 20 proteins have been reported to interact with AICD, including some that are necessary for AICD-dependent functions in signal transduction, apoptosis or modulation of cytoskeletal dynamics (68). It is also important to note that initially it was believed that the cleavage of APP is physiologically carried out by α-secretase, and it prevents amyloidogenesis in AD. In reality, this is an oversimplification, because only a fraction of the total pool of APP is cleaved by α-secretase in most cell types, leaving most of the APP protein intact. Furthermore, β- and γ-secretase processing of APP (at the N- and C-terminals of the Aβ sequence) also occur under physiological conditions; this indicates that all fragments of APP, including the Aβ peptide, are part of normal physiology (44). It is also worth mentioning that none of the secretases have unique substrate specificity toward APP, and the cleavage of transmembrane proteins by an ADAM or BACE (ectodomain shedding) is commonly involved in the activation of a number of functional pathways. Ectodomain shedding by ADAMs is essential for the release of many cytokines and growth factor ligands, such as epidermal growth factor (EGF) (69). Additionally, ADAMs are involved in the ectodomain shedding of growth-factor receptors, such as human epidermal growth factor receptor 2 (70) and Notch (71). Ectodomain shedding by BACE is also likely to be required for the proper function of a number of proteins (72). For example, neuregulin-1 (NRG1) is cleaved by BACE1 and ADAM17 to release an ectodomain fragment that acts in a paracrine manner to stimulate myelination (73). Therefore, cleavage by ADAMs or BACE can potentially facilitate cellular signaling in a variety of ways, either by the release of growth factors or by the ligand-dependent activation of cellular receptors. RIP by γ-secretase is also a process involved in the normal function of many proteins. RIP can serve two general functions. First, it can remove the membrane-associated fragment that is produced by ectodomain shedding. Second, it can catalyze the production of intracellular signaling domains (62). Gamma-secretase has been reported to cleave more than 50 type-I membrane proteins in addition to APP (74). Apart from APP, the most well-known γ-secretase substrate is the developmental protein Notch, which is activated by γ-secretase cleavage (75, 76). Therefore, it is possible that γ-secretase cleavage may also be involved in the function of APP.

Figure 2:

Schematic diagram of APP processing pathways (not drawn to scale).

The Aβ domain is highlighted in black. For simplicity, only one cleavage site is shown for each enzyme. EC, extracellular domain; TM, transmembrane domain; IC, intracellular domain.

Putative functions of APP

Despite the large number of published studies on APP, there is still no clear consensus on the protein’s function. There are many excellent reviews that summarize this area in detail but are otherwise beyond the scope of this article. Briefly, an analogy of the secondary structures and proteolytic processing profiles between Notch and APP predicts that APP could function as a cell-surface receptor similar to Notch (77). The E1 and E2 regions in the extracellular domain of APP have been shown to interact with extracellular matrix proteins and heparin sulfate proteoglycans (78), supporting its role in cell-substratum adhesion, cell-cell adhesion, dimerization, ligand-binding and metal-binding. A number of publications have pointed to an important role of the APP extracellular domain in neurite outgrowth and synaptogenesis, both as a full-length protein and as a secreted molecule (APPs) following ectodomain-shredding. Furthermore, a role of APP in cell signaling and apoptosis via AICD has been also documented (68). Thus, APP may exert these activities in both autocrine and paracrine fashions. Because levels of APPs have been reported to be reduced in individuals with AD (79), the possibility has been raised that the loss of the trophic activity of APPs, in concert with a reduction of growth factors in the brain, may contribute to neurodegeneration in AD. One of the mechanisms that regulate APP function is likely to be protein phosphorylation. APP can be phosphorylated at multiple sites in both extracellular and intracellular domains (80). In particular, protein phosphorylation by protein kinase C (PKC) has been shown to play a pivotal role in the control of APP metabolism and amyloid formation (80). Among different sites of phosphorylation, the phosphorylation at the threonine₆₆₈ residue (Thr₆₆₈, numbering for the APP₆₉₅ isoform, i.e., Thr₇₄₃ for the APP₇₇₀ isoform) in the APP intracellular domain has received the most attention to date. This phosphorylation has been implicated in regulating APP localization to the growth cones and neurites (81, 82). Significantly, the Thr₆₆₈ phosphorylated APP is shown to be preferentially transported to the nerve terminals (83), and the Thr₆₆₈ phosphorylated APP fragments are increased in AD, but not in control subjects (84), raising the possibility that this phosphorylation event may contribute to AD pathogenesis by regulating Aβ generation in neurons. A recent work provided a possible link between Thr₆₆₈ phosphorylation and APP processing through the prolyl isomerase Pin1. The authors showed that Pin1 binds to the phosphorylated Thr₆₆₈-Pro motif and promotes the isomerization of the proline residue (85). This leads to a conformational change of the APP intracellular domain and alteration of APP processing and Aβ production. Furthermore and interestingly, the APPThr₆₈₈ phosphorylation and adaptor protein interaction may be functionally coupled. Biochemical and structural studies indicate that Thr₆₈₈ phosphorylation results in a conformational change by which it negatively regulates APP binding to Fe65 and reduces the stability of the APP intracellular domain (86). Overall, these findings lend support for an important role of Thr₆₆₈ phosphorylation and Fe65 in regulating APP dynamics. Concerning the two other human proteins that have a high degree of homology to APP, but do not contain the Aβ sequence, APLP1 and APLP2, and in regard to the pathology of AD, although APLP1 and APLP2 do not produce the toxic Aβ peptide, their roles in functioning separately from, but in support of, APP suggest that they may play a role in the development of the disease. Indeed, it was reported that both APLP1 and APLP2, like APP, formed transcriptionally active triple protein complexes with Mint3 and transcriptional co-activators Taz and Yap to activate transcription of target genes, and the complex formation was regulated by the γ-secretase cleavage of APLP1 and APLP2. The presence of Mint1 instead of Mint3 in the complex prevented its translocation to the nucleus. APLP1 displayed much lower transactivation levels compared to APP and APLP2. These results indicate that all these three human APP family members are capable of activating gene transcription via Mint3-Taz and Mint3-Yap (87).

To understand the in vivo function of APP and its processing products, the in vivo gain-and loss-of-function phenotypes associated with the APP family of proteins in model systems (C. elegans, Drosophila and mice) were also investigated (88). In C. elegans, although the cloning of the C. elegans APP homolog APL-1 was published in 1993, to date there has been no report describing the APL-1 null mutant. It was reported that, by using an RNA-interference (RNAi) method, worms treated with APL-1 RNAi exhibited a defect in pharyngeal pumping, and this phenotype was shared with Fe65 homolog-1 (Feh-1)-treated worms, suggesting that these two proteins act in the same pathway, possibly through direct physical interactions. Mutation in Feh-1 resulted in a reduction of acetylcholinesterase gene expression. These findings taken together support a conserved role of the APP family of proteins in the cholinergic pathway, possibly in a Fe65-dependent manner. For Drosophila, it was reported that mutants lacking APPL were viable and fertile but exhibit subtle behavioral defects that can be partially rescued by human APP, demonstrating functional conservation. Subsequent analysis revealed that these mutant flies show reduced synaptic bouton numbers at the neuromuscular junction, and that this activity appears to require the formation of a complex with the cell adhesion molecule fasciclin and Drosophila Mint/X11. Consistent with a role of APPL in synapse development, ectopic over-expression of APPL leads to satellite bouton formation at the Drosophila neuromuscular junction, and this activity requires the APP YENPTY domain where adaptor protein interaction takes place. In mice, homozygous APP-deficient mice are viable and fertile. However, the mutant animals were smaller (15–20% less body weight) than age-matched controls and exhibited decreased locomotor activity and forelimb grip strength, indicating compromised neuronal or muscular function. In addition, the majority of the mice show reactive gliosis, suggesting undefined neuronal damage in brain activity. The relatively subtle phenotype of the APP-deficient mice indicated that the presence of other APP family members may compensate for the loss of APP. Indeed, whereas mice with individual deletion of APP, APLP1 and APLP2 are viable and fertile, APP/APLP2 and APLP1/APLP2 double-knockout mice showed early postnatal lethality. Intriguingly, the APP/APLP1 double-null mice are viable, revealing a property of APLP2 that is uniquely required when APP or APLP1 is absent. Although the neuromuscular junction of APP or APLP2 single-null mice did not show overt structural abnormalities, the APP/APLP2 double-knockout animals exhibited poorly formed neuromuscular junctions with reduced apposition of pre- and postsynaptic elements of the junctional synapses. The number of synaptic vesicles at the presynaptic terminals was also reduced. Mice deficient in all three APP family members (APP/APLP1/APLP2 triple-knockout mice) were lethal in the early postnatal period. Analysis of these APP/APLP1/APLP2 triple-knockout mice revealed that the majority of the animals showed cortical dysplasia suggestive of neuronal migration abnormalities and partial loss of cortical Cajal Retzius cells. In sum, the loss-of-function studies present a convincing picture that members of the APP gene family play essential roles in the development of the peripheral and central nervous systems relating to synapse structure and function, as well as in neuronal migration or adhesion.

In conclusion, it is clear that APP undergoes complex regulation and is important for neuronal and synaptic function in both central and peripheral nervous systems. This may involve the APP extracellular domain, the APP intracellular domain, the Aβ sequence or, indeed, cross communication among these motifs. It is therefore reasonable to speculate that the misregulation of APP could contribute to neuronal and synaptic impairment occurring in AD in an Aβ-dependent or Aβ-independent manner. As such, a more complete understanding of AD pathogenesis will likely require greater insights into the physiological function of APP.

Regulation of alternative APP pre-mRNA splicing

As mentioned in the Introduction, human APP is the best known as the precursor molecule whose proteolysis generates β-amyloid (Aβ), a 39-42 amino acid peptide whose amyloid fibrillar form is the primary component of amyloid plaques found in the brains of individuals with Alzheimer’s disease (AD) and Down syndrome (7, 9). Current therapeutics designed to treat AD have not been successful in effectively treating the progressive nature of cognitive decline observed in patients. Cholinesterase inhibitors donepezil, rivastigmine and galantamine, huperzine-A, and N-methyl-D-aspartate antagonist memantine are approved drugs available for AD; however, these drugs fail to treat the underlying cause of neurodegeneration and only provide modest short-term symptomatic relief (89). One of the underlying reasons for the failure of treatment measures is the lack of knowledge pertaining to the epigenetic, environmental, and mechanistic drivers of neurodegeneration. Many studies have focused on determining the genetic contributions to AD. In FAD, autosomal dominant mutations in APP, PS1, and PS2 [all genes involved in the production of Aβ, see Ref. (11)] are mainly responsible for the early-onset form of the disease. Recently, the triggering receptor expressed on myeloid cells 2 (TREM2) was designed as another rare monogenetic candidate for early onset of disease risk (90). However, these mutations in FAD represent a very small percentage of cases (∼1%), and ∼99% of cases are SAD (11). Multiple studies conducted to determine disease-causative loci have revealed that AD is highly complex and heterogeneous in nature, susceptibility loci vary according to gene penetrance, ethnicity, geography, environment and sample size of various studies; therefore, identifying genes responsible for SAD is highly challenging (91–93). Linkage and genome-wide association studies (GWAS) have mapped regions in chromosomes 9, 10 and 12 that are associated with AD risk; several genes, such as apolipoprotein E (APOE), BACE1 and BACE2, for example, are thought to be prime candidates to confer risk because of their role in pathways associated with Aβ biosynthesis and deposition (94). However, approximately 300 genes have been implicated without strong causative evidence to increase risk for AD (92). Therefore, nongenetic factors, such as epigenetic modifications, may also be causative and currently the subject of intense research. Epigenetics is the study of changes in gene expression and/or chromatin structure and cell function caused by mechanisms other than changes in the underlying DNA sequence (95). This field is an important area of investigation because epigenetic modifications may explain differential regulation of AD risk genes and genomic regions without changes to their DNA sequence and, therefore, undetected in genetic studies. These modifications can occur on DNA molecules (mainly on cytosine bases at cytosine-guanine: CpG sites) or on histone proteins, which make up the fundamental structure of chromatin (96, 97). Modifications to histones include methylation, acetylation, phosphorylation, ubiquitination, sumoylation, etc., and more than 100 different modification residues have been characterized (98). Variations in promoter sequences can alter gene expression directly by altering a transcription-binding site or indirectly by changing the organization of chromatin. Promoter variants with effects on the transcriptional activity of certain human genes and in the regulation of alternative pre-mRNA splicing have been identified, and genetic association studies have suggested that some of these variants may be disease risk factors (99, 100). Recently, noncoding RNAs (ncRNAs), such as long noncoding RNAs (LncRNAs) and microRNAs (miRNAs), among other regulatory RNA molecules, have been shown to play an epigenetic role in the regulation of genes by various mechanisms, including recruiting DNA methyltransferases and chromatin modifiers to their targets, inhibiting translation of mRNA, and in the degradation or stability of mRNA by sequence complementarity with their targets (101). Epigenetic modifications are therefore central facilitators of the nexus between genes and the environment. The role of epigenetic modifications has been studied in various fields of biology, especially in developmental and cancer biology; erroneous regulation of epigenetic modifications has been linked to many developmental deficiencies, including neurogenesis, and in the formation and progression of cancer (102, 103). The role of epigenetic modifications in neurological disease has also been studied; erroneous regulation of epigenetic modifications results in a number of neurological disorders such as Rett syndrome, mental retardation, autism and AD (104–107). Recently, I performed a study (108) searching for a link between APP and Lesch-Nyhan syndrome (LNS), a neurogenetic disorder of purine metabolism in which the enzyme hypoxanthine-guanine phosphoribosyltransferase (HGprt, EC. 2.4.2.8; MIM 308000) is defective. Complete or severe deficiency of HGprt enzyme activity leads to LNS (MIM 300322). Classical features of LNS include hyperuricemia and its sequelae (gout, nephrolithiasis and tophi), intellectual impairment and self-injurious behavior (109). Partial deficiency of HGprt enzyme activity (MIM 300323) leads to its attenuated variants (LNVs). LNV is characterized by the consequences of overproduction of uric acid and a variable spectrum of neurological manifestations, without the manifestations of self-injurious behavior (110, 111). The etiology of LNS and LNV involves a mutation of the HPRT1 gene, which is on the long arm of the X chromosome (Xq26.1), and it contains nine exons and eight introns (110, 111). Because the HPRT1 gene is on the X chromosome, males are affected and females in the families are at risk for being carriers of the mutation. A major unsolved question is how the loss of HGprt enzyme function affects the brain to cause the neurobehavioral syndrome in LNS and its attenuated variants (LNVs). To address this issue, a link between LNS and the aberrant basal ganglia function, including the dysfunction of dopaminergic pathways, was reported (112, 113). However, the mechanism by which features of LNS result from impaired purine metabolism is still not well understood. It was also documented that adhesion of HGprt-deficient neuroblastoma as well as fibroblasts from patients with LNS exhibited dramatically enhanced adhesion compared to control cells (114) and could have consequences for the maturation of the central nervous system, as seen in the smaller brain size of LNS and LNVs children (115–117). It was also reported that HGprt-deficiency was accompanied by aberrations in a variety of pathways known to regulate neurogenesis or to be implicated in neurodegenerative disease, including the canonical Wnt/β-catenin and the AD/presenilin signaling pathways (118). Furthermore, it was also documented that the HPRT1 gene regulated multiple developmental and metabolic pathways of murine embryonic stem cell neuronal differentiation, and the neural aberrations of HGprt-deficiency could result from the combinatorial effects of these multi-system metabolic errors (119). As some of these aberrations are also found in forms of AD and Huntington’s disease, these authors predicted that some of the systems defects play similar neuropathogenic roles in diverse neurodevelopmental and neurodegenerative diseases (119). A role for APP related to cell-cell or cell-substrate adhesion and important for brain morphology and highly coordinated brain functions such as memory and learning has been suggested (12, 88). Hence, the APP pathway is possibly implicated in the development of LNS and LNVs. In an attempt to search for a link between APP and LNS (108), I have examined the APP-mRNA profile and the genomic APP-DNA in fibroblasts from normal subjects and HGprt-deficient LNS and LNVs patients. Interestingly, during this study (108), I identified for the first time, in fibroblasts from normal subjects as well as from LNS and LNVs patients, several APP-mRNA isoforms encoding divers APP protein isoforms ranging from 120 to 770 amino acids (with or without mutations and/or deletions) (Table 1, see Supplementary Material for additional details about the experimental procedures). There was no mutation found in the 18 exons and flanking intronic sequences of genomic APP-DNA from all fibroblasts. Divers APP-mRNA isoforms with or without mutations and/or deletions have been found in fibroblasts of all subjects (normal control as well as LNS and LNVs patients), and it is therefore conceivable that changes in epigenetic regulation caused by genetic and environmental factors, as well as life events and aging, could cause alterations in the regulation of alternative APP pre-mRNA splicing and result in an imbalance between different APP-mRNA isoforms. Some of these isoforms may be disease risk factors. It is the first time that the real profile of APP-mRNA isoforms accounting for epigenetic mechanisms in the regulation of alternative APP pre-mRNA splicing has been shown. Here, it is important to note that epigenetic modifications, due to gene-gene interactions (epistasis) (120), could affect the regulation of alternative APP pre-mRNA splicing in favor of APP-mRNA isoforms responsible for the disease. Epistasis is important, ubiquitous and has become a hot topic in complex disease genetics, such as AD, schizophrenia, autism, type-2 diabetes, sporadic breast cancer, sickle-cell anemia, etc., in recent years and even common for determining phenotypes for a number of rare Mendelian diseases such as cystic fibrosis, Hirschsprung disease, etc. (121). However, the data supporting epistasis in complex human diseases are emerging slowly. This is due to different difficulties that we face in detecting and characterizing epistasis, such as challenges of modeling non-linear interactions, and in the interpretation of results (120, 121). My present results accounted for epigenetic mechanisms in the regulation of alternative APP pre-mRNA splicing and may provide, therefore, a unique integrative framework for the pathologic diversity and complexity of neurodevelopmental and neurodegenerative disorders in which the APP gene is involved in the pathogenesis of the diseases such as autism, AD and, especially SAD. Although there is no experimental evidence at present proving the direct link between LNS/LNVs and APP, based on the results summarized in Table 1, it is conceivable that in LNS, epigenetic modifications due to epistasis between mutated HPRT1 and APP genes could affect the regulation of alternative APP pre-mRNA splicing in favor of APP-mRNA isoforms responsible for the modifications of brain anatomy and the neurobehavioral syndrome observed in LNS and LNVs. The severity of the affection would depend on how mutations ultimately alter the interactions between mutated HPRT1 and APP genes. The type of mutation and its location in the HPRT1 gene is therefore an important factor for provoking disease (LNS or LNV), not only through its effect on residual HGprt enzyme activity but also through its effect on interactions between mutated HPRT 1 and APP genes. For the same type of mutation in the HPRT1gene, the response to epigenetic modifications due to epistasis may be different from one patient to another and this could explain the manifestation of different clinical phenotypes from different patients and also from different affected family members (111, 122, 123).

Control of genomic rearrangements of APP gene

In the past 15 years, it has become evident that higher-order genomic architectural features can lead to a susceptibility to DNA rearrangements that are frequent causes of disease (124). Genomic rearrangements describe mutational changes in the genome such as duplication, deletion, insertion, inversion and translocation that are different from the traditional Watson-Crick base pair alterations (125). Genomic rearrangements can represent polymorphisms that are neutral in function, or they can convey phenotypes via diverse mechanisms, including changing the copy number (that is, copy number variation or CNV) of dosage-sensitive genes, disrupting genes, creating fusion genes or other mechanisms (125). The pathological conditions caused by genomic rearrangements are collectively defined as genomic disorders (125–127). CNV is a major source of genetic variation among humans. CNV can convey clinical phenotypes, including genomic disorders, sporadic diseases and complex human traits such as autism (128), schizophrenia (129) and mental retardation (130). Thus, uncovering the mechanisms underlying CNV formation has tremendous implications for the diagnostics of CNV-associated diseases. Typically, the term ‘genomic rearrangements’ is only used to describe gross DNA changes ranging from thousands to sometimes millions of base pairs that can cover clusters of different genes (125). Three major mechanisms have been proposed for rearrangements in the human genome: non-allelic homologous recombination (NAHR), non-homologous end-joining (NHE) and the Fork Stalling and Template Switching (FoSTeS) models (125). Recently, Chen et al. (131) proposed the serial replication slippage (SRS) model to explain the ‘smaller’ DNA rearrangements (between 21 bp and up to 10 kb), including duplications, deletions, insertions and inversions, collected in the Human Gene Mutation Data base (HGMD). The SRS model proposed for small gene mutations shares some general features with the FoSTeS model proposed for the larger rearrangements. Both models assume serial replication slippage, and both stress the importance of the genomic architectural elements, such as palindromic DNA, stem-loop structures, repeats and so on, that may facilitate the initial stalling of the replication fork. Whereas the SRS model assumes that replication slippage occurs on closely adjacent sites (possibly inside the same replication fork) and causes DNA rearrangements of small sizes, the FoSTeS model emphasizes that the template switch can occur over long distances (120 kb to 550 kb observed to date) to another replication fork (given the spatial closeness of the two forks) and cause DNA rearrangements on a much larger scale (124). Recently, in the research work of searching for a link between APP and LNS (108), I have also identified the presence of genomic rearrangements of APP gene. For the first time, three APP-mRNA isoforms resulted from DNA rearrangements of APP gene with a deletion followed by an insertion (INDELS) via the SRS mechanism leading to non-functional APP or loss of APP. This is due to the early translational stops which resulted in transcript instability c.19_2295delC166TG…TTT2442insG166TT…GAGTCC… CTTAGTC…TCT489, p.Leu7Valfs*2 for sample #7, LNS; c.19_2295delC166TG…TTT2442insG169TT…GAGACC… CTTGGTC…TCT492, p.Leu7Valfs*2 for sample #15, LNV; and c.16_2313delG163CA…TAG2460insG84CC…CAT616, p.Leu7Hisfs*45 for sample #14, LNV. The underlined letters T, A, and A, G indicate the difference in nucleotides in the sequence of 324 bp inserted in these two APP-cDNA (Figure 3A, B and C, see Supplementary Material for additional details about the experimental procedures). Indeed, as shown in Figure 3A and B, there was a deletion between exon 1 and exon 18 of the APP gene starting after nucleotide A₁₆₅ or A₁₆₈ in exon 1 of APP-cDNA and substituted by an insertion of 324 bp in which there were five Opal stop codons TGA located after nucleotides T₁₆₈, C₂₁₀, T₂₇₆, A₃₀₀, T₃₃₀ and one Amber stop codon TAG located after nucleotide T₄₈₃ for sample #7, LNS (Figure 3A), and five Opal stop codons TGA located after nucleotides T₁₇₁, C₂₁₃, T₂₇₉, A₃₀₃, T₃₃₃ for sample #15, LNV (Figure 3B), respectively. In fact, the sequence of 324 bp inserted in these two APP-cDNA is identical except for nucleotides T₃₅₁ (in triplet AGT, see horizontal bar mark) and A₄₈₅ (in triplet TAG, see horizontal bar mark) in sample #7 (Figure 3A) and A₃₅₄ (in triplet AGA, see horizontal bar mark) and G₄₈₈ (in triplet TGG, see horizontal bar mark) in sample #15 (Figure 3B). Also, as shown in Figure 3C, there was a deletion in exon 1 of the APP gene starting after nucleotide G₈₃ in exon 1 of APP-cDNA and substituted by an insertion of 533 bp in which there were five Opal stop codons TGA located after nucleotides G₁₁₈, C₂₆₀, C₂₈₁, A₂₈₄, C₃₆₅ and two Amber stop codons TAG located after nucleotides A₂₃₉, C₃₅₃ for sample #14, LNV. Here, all other exons of the APP gene were totally excluded. Only the first stop codons presented in the inserted sequence of all three APP-cDNA are shown in Figure 3A, B and C because only the first is a real stop codon, the others are part of the 3′unstranslated region (UTR) and will not function as stop codon (see horizontal bracket marks). As a result, APP produced from these three APP-mRNA isoforms could not function. Early translational stops can lead to transcript instability, leading to the loss of APP. As a consequence, there is a disease risk. Here, experiments conducted on the controls (normal subjects) have not shown any clones of the APP-mRNA isoforms with the genomic rearrangements comparable to those observed in LNS and LNVs patients. Given the small size of the DNA rearrangements deduced from the three APP-mRNA isoforms isolated in this study, the SRS model is likely the one responsible for the present genomic rearrangements of the APP gene. The complex gene rearrangements caused by the SRS mechanism for diverse mutations found in different genes, such as GALNS, HPRT1, IDS, MECP2, OFD1, SALL1, etc., have been reported (131). Duplication of the APP locus resulting from the genomic rearrangements has been documented [see Ref. (132) and the AD Mutation Database, http://molgen-www.uia.ac.be/ADMutations]. Recently, epigenetic changes controling genomic rearrangements of the IgH locus has been reported (133). It is therefore conceivable that changes in epigenetic regulation caused by genomic rearrangements of the APP gene could result in an imbalance between different APP-mRNA isoforms, and some of these isoforms may be disease risk factors. Epistasis between mutated HPRT1 and APP genes could be one of the factors of epigenetic modifications responsible for genomic rearrangements of the APP gene.

Figure 3:

Automated direct APP-cDNA sequence analysis for screening of insert into the pcDNA^TM3.1/V5-His-TOPO^®vector.

(A) Sample #7, LNS: Chromatogram of the APP-cDNA sequence read from left to right (5′→3′) showed the nucleotide A₁₄₈ of the ATG start codon of exon 1 of APP gene (↑); the beginning of the deletion and substituted by an insertion of 324 bp between exon 1 and exon 18, located after nucleotide A₁₆₅ (↑); the first Opal stop codon TGA, located after nucleotide T₁₆₈ (see horizontal bracket mark); the presence of nucleotides T₃₅₁ (in triplet AGT, see horizontal bar mark) and A₄₈₅ (in triplet TAG, see horizontal bar mark); the nucleotide G₄₉₀ of the GAG start codon of exon 18 (↑); the nucleotide G₅₀₇ of the Amber stop codon TAG end codon of exon 18 at the end of APP-cDNA (↑). (B) Sample #15, LNV: Chromatogram of the APP-cDNA sequence read from left to right (5′→3′) showed the nucleotide A₁₅₁ of the ATG start codon of exon 1 of APP gene (↑); the beginning of the deletion and substituted by an insertion of 324 bp between exon 1 and exon 18, located after nucleotide A₁₆₈ (↑); the first Opal stop codon TGA, located after nucleotide T₁₇₁ (see horizontal bracket mark); the presence of nucleotides A₃₅₄ (in triplet AGA, see horizontal bar mark) and G₄₈₈ (in triplet TGG, see horizontal bar mark); the nucleotide G₄₉₃ of the GAG start codon of exon 18 (↑); the nucleotide G₅₁₀ of the Amber stop codon TAG end codon of exon 18 at the end of APP-cDNA (↑). (C) Sample #14, LNV: Chromatogram of the APP-cDNA sequence read from left to right (5′→3′) showed the nucleotide A₆₉ of the ATG start codon of exon 1 of APP gene (↑); the beginning of the deletion and substituted by an insertion of 533 bp in exon 1, located after nucleotide G₈₃ (↑); the first Opal stop codon TGA, located after nucleotide G₂₁₈ (see horizontal bracket mark); the nucleotide T₆₁₆ at the end of APP-cDNA (↑).