**2. Architecture of the APP protein**

APP can be divided into three domains (Figure 1). As a single pass type I membrane protein, the N-terminal ectodomain of APP (residues 18 to 624 neglecting the signal peptide, numbers refer to the neuronal splice form APP695, UniPROT entry: P05067-4) locates to the extracellular space. The single hydrophobic transmembrane domain (TMD, residues 625 to 648) is followed by the rather short APP intracellular domain (AICD, residues 649 to 695). More important than this topological classification is the distinction according to the fragments produced by secretase cleavage events [10]. The products produced by ectodomain shedding are sAPPα (residues 18 to 612; cleaved by α-secretases, members of the ADAM family of zinc metallo‐ proteases) and sAPPβ (residues 18 to 596; cleaved by β-secretase, an aspartic protease also known as BACE1 in the nervous system and BACE in peripheral tissue). The C-terminal fragments (CTFs) generated by ectodomain shedding are the still membrane embedded αCTF (CTF83) and βCTF (CTF99), respectively. The CTFs are subsequently cut in the RIP process by the intramembrane aspartate protease presenilin (1 or 2) as part of the γ-secretase complex, with αCTF being split into the p3 peptide and the AICD (residues 646 to 695) and βCTF into the Aβ peptide (Aβ40: residues 597 to 636; Aβ42: residues 597 to 638) and again the AICD.

**Figure 1. Architecture of APP and of its proteolytic fragments**. A. Domain architecture of the neuronal splice var‐ iant APP695. Domains with known atomic structures (E1 and E2) and the TMD are shown as ribbon diagrams in a col‐ our code from blue (N-terminus) to red (C-terminus). Dashed lines give structurally unknown regions. Proposed homodimeric interactions within E1 and E2 are shown in gray. Positions of secretase cleavage events and the respec‐ tive breakdown products are labeled. B. Sequence and proteolytic fragments within βCTF. Aβ peptides, the TMD (gray), and sequence fingerprints within the AICD are colour coded.

In terms of three-dimensional structure, only substructures within the large APP ectodo‐ main have been solved as independently folded subdomains. The N-terminal E1 domain is a two-lobe structure consisting of the growth factor like domain (GFLD, residues 18-123) and a copper-binding domain (CuBD, residues 124 to 189), both comprising mixed αβ topologies rigidified by disulfide bridges [11-13]. The E1 domain is followed by a highly acidic, and probably unfolded, stretch of about 100 residues that passes on to the E2 domain (residues 290 to 495), consisting of two coiled-coils connected through a continuous central helix and resembling a spectrin family fold [14]. E1 and E2 domains have been implicated in APP dimerization [14-16], which is reported to be modified by the extracellular matrix [17], and to have significant impact on localization and cleavage events. In addition, dimerization might also involve the TMD region [16]. Besides dimeri‐ zation, APP architecture (and likely function) is also influenced by a series of post-trans‐ lational modifications, mainly by N- and O-glycosylation and phosphorylation [18], which will be discussed in detail below. The reminder of the ectodomain between E2 and the TMD, the so-called juxtamembrane region (residues 496 to 624), is again intrinsi‐ cally disordered based on secondary structure prediction and contains the cleavage sites for the α- and β-secretases. The single TMD is clearly α helical, although with partial propensity in forming β structures. This propensity extends also to the juxtamembrane region with the fatal consequence, that after secretase cleavage the amyloid peptide folds into a β hairpin structure and aggregates to form the toxic oligomers and finally the amyloid fibrils. Finally, the AICD itself is again intrinsically disordered as shown by NMR and CD experiments [19, 20]. Importantly however, this small C-terminal stub has recently been shown to adopt different conformations reflecting its versatile functions. The structure-function relationship of the AICD shall be described in the following.

### **3. Biology of the AICD: Tales of a tail**

6 (DR6) [9], and the APP intracellular domain as created by the ε-cut of γ-secretase during the RIP process [3], which is the topic of this paper. We will start by getting the architecture of

APP can be divided into three domains (Figure 1). As a single pass type I membrane protein, the N-terminal ectodomain of APP (residues 18 to 624 neglecting the signal peptide, numbers refer to the neuronal splice form APP695, UniPROT entry: P05067-4) locates to the extracellular space. The single hydrophobic transmembrane domain (TMD, residues 625 to 648) is followed by the rather short APP intracellular domain (AICD, residues 649 to 695). More important than this topological classification is the distinction according to the fragments produced by secretase cleavage events [10]. The products produced by ectodomain shedding are sAPPα (residues 18 to 612; cleaved by α-secretases, members of the ADAM family of zinc metallo‐ proteases) and sAPPβ (residues 18 to 596; cleaved by β-secretase, an aspartic protease also known as BACE1 in the nervous system and BACE in peripheral tissue). The C-terminal fragments (CTFs) generated by ectodomain shedding are the still membrane embedded αCTF (CTF83) and βCTF (CTF99), respectively. The CTFs are subsequently cut in the RIP process by the intramembrane aspartate protease presenilin (1 or 2) as part of the γ-secretase complex, with αCTF being split into the p3 peptide and the AICD (residues 646 to 695) and βCTF into the Aβ peptide (Aβ40: residues 597 to 636; Aβ42: residues 597 to 638) and again the AICD.

βCTF

…ISEVKMDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIATVIVITLVMLKKKQYTSIHHGVVEVDAAVTPEERHLSKMQQNGYENPTYKFFEQMQN 597 , , 613 , , 639 646 , 653 , 668, , 682 687 , 695

Aβ40 (Aβ42) AICD

**Figure 1. Architecture of APP and of its proteolytic fragments**. A. Domain architecture of the neuronal splice var‐ iant APP695. Domains with known atomic structures (E1 and E2) and the TMD are shown as ribbon diagrams in a col‐ our code from blue (N-terminus) to red (C-terminus). Dashed lines give structurally unknown regions. Proposed homodimeric interactions within E1 and E2 are shown in gray. Positions of secretase cleavage events and the respec‐ tive breakdown products are labeled. B. Sequence and proteolytic fragments within βCTF. Aβ peptides, the TMD

 E1 18-189 E2 290-495 N-APP 18-286 sAPPβ 18-596 sAPPα 18-612 ectodomain 18-624

B β α γ42 ε

(gray), and sequence fingerprints within the AICD are colour coded.

18-123 124-189 190-289 496-624 625-648 646-695

AICD-C31

β 597 α 613

γ40 637 ε 646 γ42 639

APP into place.

4 Understanding Alzheimer's Disease

A

**2. Architecture of the APP protein**

When talking about the AICD, a clear distinction has to be made: the function (and probably also the structure) is different for AICD as part of APP at the membrane and for AICD as peptide generated by ε-cleavage of γ-secretase and first described by Passer *et al.* [21]. Within the AICD three sequence motifs have been identified to be of function‐ al relevance. The first one is the 653YTSI sequence, which has been implicated in the baso‐ lateral sorting of APP in polarized MDCK cells [22] and which is reminiscent to the YXXΦ (X: any residue; Φ: aromatic or large hydrophobic residue) consensus motif as ty‐ rosine-based and clathrin-mediated endocytic sorting signal [23]. Indeed, when Tyr653 is mutated to alanine, APP is equally distributed on apical and basolateral membranes in MDCK cells [24]. Somewhat surprisingly, in neurons polarized sorting occurs independ‐ ently of this signal [25]. Subcellular trafficking and neuronal APP sorting is still poorly understood [26] and remains a topic of intense investigation. This first motif contains three phosphorylatable residues (YTS), and it has been reported that at least Thr654 and Ser655 are phosphorylated in the adult rat brain under physiological conditions [27].

Much more attention has been drawn to the second fingerprint 667VTPEER, as this site seems to be also critically involved in pathophysiological processes. While the function of the residues has remained elusive prior to the availability of structural data, Thr668 has since been established as the major phosphorylation site of APP and its physiological function has been investigated in the adult rat brain, post mitotic differentiating neurons and dividing cells [18]. Whereas pT668 in neurons is dominant in the fully-glycosylated mature APP, in differentiating cells the purely N-glycosylated immature protein as present in the endoplasmic reticulum and the early Golgi is of relevance. Accordingly, different kinases are responsible for Thr668 phosphorylation. In neurons, it is glycogen synthase kinase-3β (GSK3β) and cyclin-dependent kinase 5 (Cdk5), while Cdk1 and cdc2 kinase phosphorylate this residue in dividing cells. Moreover, when cells are exposed to stress, phosphorylation is taken over by c-Jun N-terminal kinase (JNK) [28].

phospho-

G G

+ secretases

AICD interactions is strongly implicated in Alzheimer's disease progression.

+ Tip60 (or CP2/LSF/LBP1)

T

αN

G

\_\_\_\_\_

T

αN

Fe65

AICD-C31

± Fe65? + caspases

T

αN

G

apoptosis

containing adaptor protein Fe65.

rylation - Fe65

Fe65

II.

I.

AICD

transcription activation

Tip60 (or CP2/LSF/LBP1)

**Figure 2.** AICD in health and disease. Different fates of the AICD are exemplified for the main AICD interaction with Fe65-PTB2 (red T-box: TPEE, cyan Y-box: NPTY, G: glycine hinge, gray cylinder: C-terminal helix of Fe65-PTB2). In the non-phosphorylated state, AICD forms a stable complex with Fe65-PTB2 that assembles in ternary complexes with i.e. Tip60 or CP2/LSF/LBP1 via Fe65-PTB1. Upon cleavage by the secretases, the liberated complexes are involved in tran‐ scription activation. Alternatively, caspase cleavage within the AICD results in cytotoxic AICD-C31, which might com‐ pete with AICD for Fe65-PTB2 binding and induce apoptosis. Phosphorylation of either Thr668 (I.) or Tyr682 (II.) results in a destabilization of the Fe65-PTB2/AICD interaction (shown in brackets) and results in complex dissociation. Phos‐ phorylation stimulates (I.) neuronal differentiation or (II.) initiates signaling cascades. Deregulation of the Fe65-PTB2/

Instead, the NPXY motif is well known to interact with adaptor proteins containing a domain known as phosphotyrosine-binding (PTB) or phosphotyrosine-interacting domain (PID) [36]. PTB domains reveal a fine tuned plasticity in ligand recognition, and besides recognizing phosphorylated NPXpY motifs, most PTB adaptor proteins can also bind to their ligand in a pY-independent manner. Accordingly, *in vitro* phosphorylation of Tyr687, which does not seem to occur in the brain [18], does i.e. not alter the binding affinity of AICD to its major PTB-

In APP, the NPXY signal is extended by three residues at the N-terminal side (GYE), with especially Tyr682 being most critical for function [31, 37, 38]. The motif is present in many lysosomal glycoproteins that are endocytosed and targeted to the lysosomes [39]. In cellculture studies, Tyr682 can be readily phosphorylated by the nerve growth factor receptor TrkA and the tyrosine kinases Abl and Src [40]. In brains of AD patients, it is known that at least βCTF is phosphorylated, whereas this is not the case for αCTF [41]. In addition, phos‐

disease Fe65 Fe65

endosomal or lysosomal membrane

II.

in excess?

I. (pT668)

Structure and Function of the APP Intracellular Domain in Health and Disease

neuronal differentiation

amyloidogenic processing

pT

http://dx.doi.org/10.5772/54543

I.

β

γ ε 7

G

I. or II. in excess? Alzheimer's

II. (pY682) + ShcA/Grb2

signaling

Phosphorylation on Thr668 of APP depends on the presence of Pro669 and strongly affects Aβ production [29]. This is reminiscent of the Tau protein, where the phosphorylation of certain serine and threonine residues depends on adjacent proline residues and leads to tangle formation [29]. A first molecular explanation for the proline-dependency was revealed by studies showing that the prolyl isomerase Pin1, catalyzing the *cis-trans* isomerization of the Thr-Pro peptide bond, increases amyloidogenic APP processing and selectively elevates Aβ42 levels. Intriguingly, Pin1 is down regulated and/or inhibited by oxidation in neurons of Alzheimer's disease patients and *Pin1* knockout causes neurodegeneration (and tauopathy). Pin1 binds to Thr668-phosphorylated APP and accelerates Pro669 isomerization (by a factor of 103 ). Thus, the AICD swaps between two conformations, as visualized by NMR [29]. This conformational switch may in turn have crucial consequences with regard to the AICD protein interacting network, as shown for the neuronal adaptor protein Fe65 (Figure 2 and see below) [20, 30]. To evaluate in as much the phosphorylation state of Thr688 controls APP processing *in vivo*, knockin mice were generated in which Thr668 was changed to alanine (APPTA/TA) [31, 32]. The APPTA/TA mutation, and thus absence of phosphorylation, did not significantly alter APP localization, processing, and Aβ generation, thus questioning the i*n vivo* role of Thr668 phosphorylation. However, these studies cannot rule out the possibility that a pathological increase in Thr668 phosphorylation, as found in AD patients [33], will also modulate its function. In line with this notion, Thr668 phosphorylation has also been reported to influence APP cleavage by caspases between residues Asp664 and Ala665, producing the cytotoxic AICD-C31 fragment, a process that has been strongly implicated in AD pathogenesis [34].

The third and most intensely studied fingerprint within the AICD is the 681GYENPTY sequence containing an NPXY motif, a well-established internalization signal for membrane proteins [35]. NPXY is a classical tyrosine-based sorting signal for transmembrane proteins to endo‐ somes and lysosomes [23]. However, the signal has been shown to only mediate rapid internalization of a subset of type I membrane proteins, including APP as well as members of the low-density lipoprotein (LDL) receptor family and integrin β. These proteins are internal‐ ized via clathrin-coated pits. Nevertheless evidence for a direct interaction of NPXY motifs with the coat or the AP-2 adaptor is weak.

Structure and Function of the APP Intracellular Domain in Health and Disease http://dx.doi.org/10.5772/54543 7

Much more attention has been drawn to the second fingerprint 667VTPEER, as this site seems to be also critically involved in pathophysiological processes. While the function of the residues has remained elusive prior to the availability of structural data, Thr668 has since been established as the major phosphorylation site of APP and its physiological function has been investigated in the adult rat brain, post mitotic differentiating neurons and dividing cells [18]. Whereas pT668 in neurons is dominant in the fully-glycosylated mature APP, in differentiating cells the purely N-glycosylated immature protein as present in the endoplasmic reticulum and the early Golgi is of relevance. Accordingly, different kinases are responsible for Thr668 phosphorylation. In neurons, it is glycogen synthase kinase-3β (GSK3β) and cyclin-dependent kinase 5 (Cdk5), while Cdk1 and cdc2 kinase phosphorylate this residue in dividing cells. Moreover, when cells are exposed to stress, phosphorylation is taken over by c-Jun N-terminal

Phosphorylation on Thr668 of APP depends on the presence of Pro669 and strongly affects Aβ production [29]. This is reminiscent of the Tau protein, where the phosphorylation of certain serine and threonine residues depends on adjacent proline residues and leads to tangle formation [29]. A first molecular explanation for the proline-dependency was revealed by studies showing that the prolyl isomerase Pin1, catalyzing the *cis-trans* isomerization of the Thr-Pro peptide bond, increases amyloidogenic APP processing and selectively elevates Aβ42 levels. Intriguingly, Pin1 is down regulated and/or inhibited by oxidation in neurons of Alzheimer's disease patients and *Pin1* knockout causes neurodegeneration (and tauopathy). Pin1 binds to Thr668-phosphorylated APP and accelerates Pro669 isomerization (by a factor

). Thus, the AICD swaps between two conformations, as visualized by NMR [29]. This conformational switch may in turn have crucial consequences with regard to the AICD protein interacting network, as shown for the neuronal adaptor protein Fe65 (Figure 2 and see below) [20, 30]. To evaluate in as much the phosphorylation state of Thr688 controls APP processing *in vivo*, knockin mice were generated in which Thr668 was changed to alanine (APPTA/TA) [31, 32]. The APPTA/TA mutation, and thus absence of phosphorylation, did not significantly alter APP localization, processing, and Aβ generation, thus questioning the i*n vivo* role of Thr668 phosphorylation. However, these studies cannot rule out the possibility that a pathological increase in Thr668 phosphorylation, as found in AD patients [33], will also modulate its function. In line with this notion, Thr668 phosphorylation has also been reported to influence APP cleavage by caspases between residues Asp664 and Ala665, producing the cytotoxic AICD-C31 fragment, a process that has been strongly implicated in AD pathogenesis [34].

The third and most intensely studied fingerprint within the AICD is the 681GYENPTY sequence containing an NPXY motif, a well-established internalization signal for membrane proteins [35]. NPXY is a classical tyrosine-based sorting signal for transmembrane proteins to endo‐ somes and lysosomes [23]. However, the signal has been shown to only mediate rapid internalization of a subset of type I membrane proteins, including APP as well as members of the low-density lipoprotein (LDL) receptor family and integrin β. These proteins are internal‐ ized via clathrin-coated pits. Nevertheless evidence for a direct interaction of NPXY motifs

with the coat or the AP-2 adaptor is weak.

kinase (JNK) [28].

6 Understanding Alzheimer's Disease

of 103

**Figure 2.** AICD in health and disease. Different fates of the AICD are exemplified for the main AICD interaction with Fe65-PTB2 (red T-box: TPEE, cyan Y-box: NPTY, G: glycine hinge, gray cylinder: C-terminal helix of Fe65-PTB2). In the non-phosphorylated state, AICD forms a stable complex with Fe65-PTB2 that assembles in ternary complexes with i.e. Tip60 or CP2/LSF/LBP1 via Fe65-PTB1. Upon cleavage by the secretases, the liberated complexes are involved in tran‐ scription activation. Alternatively, caspase cleavage within the AICD results in cytotoxic AICD-C31, which might com‐ pete with AICD for Fe65-PTB2 binding and induce apoptosis. Phosphorylation of either Thr668 (I.) or Tyr682 (II.) results in a destabilization of the Fe65-PTB2/AICD interaction (shown in brackets) and results in complex dissociation. Phos‐ phorylation stimulates (I.) neuronal differentiation or (II.) initiates signaling cascades. Deregulation of the Fe65-PTB2/ AICD interactions is strongly implicated in Alzheimer's disease progression.

Instead, the NPXY motif is well known to interact with adaptor proteins containing a domain known as phosphotyrosine-binding (PTB) or phosphotyrosine-interacting domain (PID) [36]. PTB domains reveal a fine tuned plasticity in ligand recognition, and besides recognizing phosphorylated NPXpY motifs, most PTB adaptor proteins can also bind to their ligand in a pY-independent manner. Accordingly, *in vitro* phosphorylation of Tyr687, which does not seem to occur in the brain [18], does i.e. not alter the binding affinity of AICD to its major PTBcontaining adaptor protein Fe65.

In APP, the NPXY signal is extended by three residues at the N-terminal side (GYE), with especially Tyr682 being most critical for function [31, 37, 38]. The motif is present in many lysosomal glycoproteins that are endocytosed and targeted to the lysosomes [39]. In cellculture studies, Tyr682 can be readily phosphorylated by the nerve growth factor receptor TrkA and the tyrosine kinases Abl and Src [40]. In brains of AD patients, it is known that at least βCTF is phosphorylated, whereas this is not the case for αCTF [41]. In addition, phos‐ phorylation regulates both AICD peptide formation and AICD-dependent cellular responses (Figure 2). These data point to a sorting function regulated by Tyr682 phosphorylation, with non-phosphorylated APP kept at the plasma membrane and therefore processed by αsecretase, and a phosphorylation-dependent re-localization resulting in β-cleavage. Sorting implies docking to respective intracellular trafficking machineries and their adaptors, includ‐ ing PTB domain containing proteins. Consistently, an APPYG/YG mutation introduced into the endogenous APP locus by knock-in led to a marked shift toward the non-amyloidogenic pathway in brain with increased levels of full length APP, sAPPα, αCTF, unaltered βCTF and reduced sAPPβ and Aβ40 levels [31].

Although NMR experiments revealed the AICD to be intrinsically disordered, the TPEE and NPTY motifs where found to form type I β-turns and TPEE forms part of a helix-capping box [19] (Figure 3). Type I turns are the most frequent reverse turns in protein structures, which in total involve about 1/3rd of all residues. Turns usually occur on the exposed protein surfaces and represent molecular recognition sites. In a capping box, the side chain of the first helical residue forms a hydrogen bond with the backbone of the fourth helical residue and, recipro‐ cally, the side chain of the fourth residue forms a hydrogen bond with the backbone of the first residue [45]. These boxes are known to stabilize the N-termini of α-helices, and preordering of the elements is thought to guide recognition of the intracellular protein network and to reduce the entropic costs for complex formation, a feature that applies as well for APP. In addition, the conformation of the TPEE motif and the propensity of forming the N-terminally capped α helix critically depend on the phosphorylation status of Thr668 [20, 46]. This structure-function relationship can be explored by the study of the AICD with its cytoplasmic

Structure and Function of the APP Intracellular Domain in Health and Disease

http://dx.doi.org/10.5772/54543

9

**Figure 3. The TPEE and NPTY motifs.** A. The TPEE motif forms a type I β-turn and a helix capping box with two char‐

More than 20 proteins have been reported to interact with the AICD [47] (Table 1). However, little is known whether these complexes occur also *in vivo* and what relevance they may have for cell physiology or AD pathogenesis. Basically, they can be classified in modifying, sorting, or signaling interactions. The modifying enzymes have been already mentioned and account for phosphorylation and prolyl *cis/trans* isomerization events. Basolateral sorting is guided by the protein PAT1, which is the only protein that has been shown to directly interact with the

acteristic hydrogen bonds (dashed yellow lines). B. The NPTY motif forms a similar type I β-turn.

**5. Interaction partners of the AICD**

653YTSI motif and is associated with microtubules [48].

interaction partners.

Sorting due to differentially phosphorylated residues is one side of the medal, signaling is the other [40]. Two signaling proteins are well known to require Tyr682 phosphoryla‐ tion for binding to APP-CTFs, namely ShcA and Grb2. ShcA is a member of a family of cytoplasmic adaptor proteins (ShcA, ShcB, ShcC) that interacts with its PTB and Src ho‐ mology2 (SH2) domains with receptor tyrosine kinases (RTKs) and activated growth fac‐ tor receptors, which is the case also for SH2/SH3 domains containing Grb2 [42]. The initiated cascades are involved both in cell proliferation and gene transcription events, like i.e. the MAP kinase pathway. Again, binding occurs only to pTyr682 of βCTFs but not of αCTFs [41] (Figure 2). Whereas the reasons for the different binding preferences remain elusive, the underlying structural transitions within the AICD itself modulating sorting and signaling have been studied in some detail.
