Alzheimer's disease (AD) is a neurodegenerative disorder that represents the most common form of dementia. The amyloid precursor protein (APP) has been in the limelight of the research on AD pathogenesis for more than 20 years. Many research groups focused their activity on this molecule, because its proteolytic processing by β- and γ-secretases gives rise to the β-amyloid peptides (Aβ). The latter are the main constituents of the senile plaques, which for many years were believed to be the pathologic trigger of AD. A large body of literature suggests that soluble oligomers of Aβ, in particular those involving Aβ42, interfere with synaptic plasticity and/or with memory functions (1). Although several reports questioned the relevance of Aβ in AD pathogenesis (2), the involvement of APP and its processing is undisputed, being supported by solid genetic data, indicating that the rare forms of familial AD are caused by mutations of the APP gene or of the presenilin genes, which encode the catalytic activity of γ-secretase (3). Experimental evidence supports the possibility that APP and/or γ-secretase dysfunction could lead to neurodegeneration (2). Obviously, the understanding of the functions of APP and of its processing is crucial to explore loss of function hypotheses in the pathogenesis of AD.
In 1995, we observed that the NPXY motif present in the small cytodomain of APP interacts with a PTB/PI domain of a protein of unknown function, named Fe65 (4, 5). This finding was followed by several new observations indicating that, further than with Fe65, the same region of APP can also interact with other proteins such as X11, JIP1, mDab, Numb, Shc, and Grb2 (6). Many of these partners have the characteristics of adaptor proteins, thus suggesting the existence of complex molecular machineries that are centered at the APP cytodomain. These findings opened a new promising scenario to understand the functions and the regulation of APP and of its processing, which could contribute to further understand the molecular basis of AD. Several review articles describe the main results on this topic. This review will focus only on one aspect of the entire story that aimed at addressing the function of Fe65.
THE ADAPTOR PROTEIN Fe65
Fe65 belongs to a protein family, evolutionarily conserved from worms to humans, which in mammals includes two more paralogs, Fe65L1 (7) and Fe65L2 (8), which share striking amino acid and domain homology to Fe65. All Fe65 proteins have an identical multimodular structure with three different protein–protein interaction domains, a WW domain and two consecutive PTB domains (9).
During the past years, screenings for Fe65 binding partners led to the identification of many proteins. APP was the first binding partner of Fe65 that has been identified, and this interaction is highly conserved as all Fe65s examined so far are able to bind through their C-terminal PTB domain (PTB2) to the NPTY-containing region of all APP and APP-like proteins. Phosphorylation of Thr-668 of APP has been shown to impair Fe65 binding, suggesting a possible “molecular switch” mechanism to regulate the Fe65–APP interaction (10). Recently, another ligand of Fe65 PTB2 domain, the small G-protein Dexras1, has been shown to bind Fe65, in a noncompetitive way with APP, and this binding is reduced by the phosphorylation of Y547 within the Fe65 PTB2 (11). Differently, the N-terminal PTB domain of Fe65 (PTB1) binds several proteins, such as the transcription factor CP2/LSF/LBP1 (12), the histone acetyltransferase Tip60 (13), and the lipoprotein receptors (14). Furthermore, this PTB1 domain of Fe65 binds the N-terminal domain of Tau protein and their interaction is regulated by Tau phosphorylation (15). The WW domain of FE65 binds the proline-containing motif of Mena, the mammalian homolog of enabled, which is involved in the control of cell motility through regulation of the actin cytoskeleton (16). Another ligand of the WW domain is the c-Abl tyrosine kinase, which phosphorylates APP on Tyr-682 (17, 18) and Fe65 on Tyr-547, within the PTB2 domain (19). The member of INHAT complex SET/TAF1-β binds to a region of Fe65 overlapping the WW domain (20). Fe65 simultaneously recruits SET and Teashirt3, which in turn binds to histone deacetylases (21).
Fe65 interactions could be regulated by conformational changes and/or post-translational modifications. In 2004, Cao and Südhof (22) proposed a model in which Fe65 is inhibited by an intramolecular interaction of the WW domain with the PTB domains that can keep Fe65 in a “closed” inactive state. Data from structural studies showed that this intramolecular interaction is weak and easy to be opened by the binding of WW domain or PTB domains to their high-affinity ligands (23). Therefore, the binding of PTB2 to APP could induce the structural change that opens Fe65, thus allowing to unmask the PTB1 and WW domains that become available for their binding partners.
THE PHENOTYPE OF Fe65 NULL ANIMALS
The study of the alterations observed in KO mice is the ideal shortcut to get information about the function of a protein. Unfortunately, the Fe65 null mice have no obvious phenotypes, perhaps because the two Fe65 paralogs, Fe65L1 and L2, having similar structure and a great sequence similarity, could be able to substitute, at least in part, for Fe65. Therefore, the first data on the effects of Fe65 gene ablation came from Caenorhabditis elegans, in which only one orthologous gene (feh-1) is present. The feh-1 null worms either arrest during the early steps of embryo development or, if this block is escaped, they arrest at the larval stage L1. In these arrested larvae, the functional defect consists of impaired feeding ability due to increased rate of pharyngeal contraction (24, 25), a phenotype similar to that caused by the suppression of the nematode ortholog of APP. To get an evident phenotype in mammals, two genes, Fe65 and Fe65L1, should be ablated. These mice show cortical dysplasia (26), also in this case similar to that observed in the triple KO of the APP family members, APP, APLP1, and APLP2 (27).
Although apparently normal, the Fe65 null mice have some subtle defects. Indeed, they show learning and memory impairment and defective early-phase LTP at the Shaffer collateral-CA1 synapses (28). Moreover, Fe65 null mice show enhanced neurogenesis (29), similar to that observed in APP−/− mice, as well as a slight increase of GnRH-1 neurons during development (30). In another set of experiments, it was demonstrated that Fe65 null mice are more sensitive to DNA-damaging agents, as etoposide and ionizing radiations (31).
The molecular bases of these defects are still not definitively addressed and various possible explanations have been proposed.
Fe65 AS A REGULATOR OF APP PROTEOLYTIC PROCESSING
The discovery of Fe65 and of the other partners of APP cytodomain raised the possibility that these proteins could be involved in the proteolytic processing of APP and in the generation of Aβ, therefore perhaps relevant for AD pathology. APP is a type I membrane protein with a large extracellular/intraluminal (EC/IL) domain, a single transmembrane tract, and a short cytosolic domain. APP EC/IL domain can be cleaved by two enzymes, α-secretase or β-secretase, acting at two different sites near the EC/IL layer of the membrane. These cleavages result in the shedding of the APP N-terminal domain in the EC or IL environment (sAPP). The resulting transmembrane stubs, called C83 or C99, respectively, are then cleaved by γ-secretase giving rise to the P3 fragment or to the Aβ peptides of various lengths, mostly of 40 or 42 residues (1). The cleavage catalyzed by the γ-secretase also causes the formation of a C-terminal fragment, released in the cytosol, called APP intracellular domain (AICD).
The regulation of APP processing is of obvious interest, because APP fragments, including Aβ, could have important functional and/or pathological roles. It is clearly established that APP processing depends on the trafficking of the molecule among various intracellular compartments. In particular, mutation of the internalization signal YENPTY of the cytodomain of APP blocks endocytic trafficking and decreases Aβ production (32). Considering that this motif is the binding site of Fe65, it can be hypothesized that the binding of Fe65 to APP influences the endocytosis and in turn the processing. Experimental evidence did not help to definitively address this point. Indeed, there are reports indicating that Aβ production is increased by Fe65 overexpression (33) and reduced by Fe65 knockdown (34), whereas other studies reported that FE65 increased sAPPα and decreased Aβ production (10). Apparently, the results obtained in animal models are similarly conflicting: APP knock-in mice carrying the Y682G mutation, which abolishes the binding of Fe65, showed decreased levels of Aβ and a massive increase in soluble APPα (35). However, the brain of transgenic mice overexpressing Fe65 contains reduced deposits of Aβ (36). Furthermore, mutations of T688 in the APP cytodomain, whose phosphorylation prevents Fe65 binding, does not affect the physiological processing of APP in knock-in animals (35). Lastly, in the Fe65/Fe65L1 double KO (DKO) mice, no difference was observed in sAPP between wild type (wt) and DKO mice, whereas a slight decrease of Aβ42 was present only in DKO male mice (26). The reasons for these apparently conflicting results are not clear. The complexity of the protein–protein interaction network centered at the YENPTY motif of the cytodomain of APP renders the phenomenon very difficult to dissect with loss or gain of function experiments. Furthermore, the proteolytic processing is also under the control of other likely unrelated events such as cell-type-specific intracellular trafficking and regulation of secretases.
Fe65, CYTOSKELETON REMODELING, AND CELL MOVEMENT
Having the characteristics of an adaptor protein, Fe65 was expected to interact with other proteins, further than APP. Mena was the first protein identified to interact with the WW domain of Fe65 (16). Mena belongs to a small family of proteins involved in the control of cell movement and morphology and in chemotactic responses (37). It was demonstrated that Fe65 and APP are present in dynamic adhesion complexes present on the surface of the cells and that their overexpression causes the acceleration of cell migration (38). Therefore, it can be speculated that an APP/Fe65/Mena complex is involved in the regulation of cell movement. This hypothesis is supported by further observations indicating that Fe65 and APP are also present in growth cones and synapses, mostly at the level of actin-rich lamellipodia (39). The neuropathological phenotype of the Fe65/Fe65L1 (26) DKO mice and those of the triple KO of APP/APLP1/APLP2 (27) or Mena/VASP/EVL (36) could be interpreted as defects of cell movement, thus confirming the possible role of Fe65 in this function. The Drosophila orthologs of the mammalian Mena and mDab1 genes are genetic modulators of the phenotype observed in flies null for the Abl tyrosine kinase. Therefore, the observations that mDab1 and Mena, directly or indirectly through Fe65, interact with APP reinforce the possibility that all these proteins are involved in a single pathway. Furthermore, we found that Fe65, through its WW domain, binds in vitro and in vivo the active form of Abl. The latter, recruited by Fe65 close to APP, phosphorylates a tyrosine residue of its cytodomain and then binds to this phospho-APP (17).
One important point that requires to be addressed is the role of the competition among the various players of this machinery, as Mena and Abl are expected to compete for the binding to the WW of Fe65, which in turn should compete with mDab1 and Abl for the interaction with APP. Figure 1 shows only some of the possible complexes that can be formed by the above-mentioned proteins. Among the various hypothetical models that can be assembled, one of particular interest is based on the competition between Dab1 and Fe65 for the binding to APP cytodomain. When Dab1 occupies the APP site, Fe65 is in the “closed” conformation (see above) that is unable to interact with Mena, being the WW domain masked. In these conditions, Mena is available to regulate cytoskeleton. Conversely, when Fe65 displaces Dab1 and binds to APP cytodomain, its WW domain is available for the binding to Mena, which in turn will be sequestered in an inactive form or in an irrelevant compartment. This scenario could be further modified by the activation of Abl, which competes with Mena for the binding to the WW of Fe65, thus liberating Mena in its active form and/or toward its relevant compartment. Obviously, the model could be further modulated by the involvement of other ligands of APP and Fe65.
NUCLEAR Fe65 AND THE REGULATION OF GENE TRANSCRIPTION
The possible nuclear functions of Fe65 are no less intriguing than those described in the previous paragraphs. The starting point in this case was the observation that Fe65 is also present in the nucleus and that its interaction with APP regulates its nuclear translocation (40). Other results indicated that further transcription factors or cofactors could be recruited at gene promoters associated with Fe65. This is the case, among the others, of: (i) late SV40 factor (LSF), whose interaction with the PTB1 domain of Fe65 affects the regulation of the thymidylate synthase gene (41); (ii) estrogen receptor α, through which estrogen appears to modulate the activity of Fe65-containing complex (42); (iii) SET, a component of the inhibitor of histone acetyltransferases (20); and (iv) Teashirt, which, together with SET, regulates the caspase 4 gene (21).
Another important study reported that nuclear Fe65 interacts with the histone acetyltransferases Tip60 and that the Fe65–Tip60 complex is able to activate the transcription from synthetic gene promoters (13). A very interesting point that emerged from this work is that the cytosolic fragment of APP, AICD, may take part in this machinery by forming a heterotrimeric complex with Fe65 and Tip60. The obvious relevance of this finding is that it suggested the possible similarity between the APP-centered complex and Notch. Indeed, Notch is a membrane-tethered transcription factor that, as a consequence of a proteolytic processing catalyzed by γ-secretase, liberates the Notch cytosolic domain (NICD). The latter goes into the nucleus, where in association with other proteins regulates the transcription of key developmental and differentiation genes (43). Therefore, AICD could function like NICD, together with Fe65, Tip60, and other proteins. Many research groups tried to address the AICD-dependent regulation of transcription; however, the results obtained so far still leave numerous doubts. Several genes have been suggested to be direct targets of AICD–Fe65, like APP itself, BACE, Tip60, KAI1, LRP, neprilysin, and Stathmin1 (44–47); however, a systematic analysis of the Fe65-binding sites in the mammalian genome is still not available. However, conflicting results are instead available on the role of the APP processing in gene regulation. Indeed, genetic and/or pharmacological suppression of γ-secretase gave very different results or was not effective on the regulation of the above-mentioned genes (48, 49). Therefore, the demonstration that AICD, through its interaction with Fe65, regulates gene expression is still lacking.
Fe65 IN THE CELLULAR RESPONSE TO DNA DAMAGE
The presence of Fe65 in the nucleus, associated with Tip60, does not imply its involvement exclusively in gene regulation. Indeed, Tip60 has a key role in the response to DNA damage, with the dual role of histone and ataxia telangiectasia mutated (ATM) acetyltransferase (50). The study of the phenotype of Fe65 null mice led to the observation that these mice are more sensitive to DNA-damaging agents (31). Low doses of etoposide or H2O2, which have only marginal effects on wt mouse embryo fibroblasts (MEFs), induced high levels of DNA damage and accumulation of γ-H2AX in Fe65 KO mouse embryo fibroblasts (MEFs). Irradiated mice showed a similar accumulation of γ-H2AX. Genotoxic stress induced by sorbitol causes the nuclear translocation of Fe65 in an APP-dependent manner (51). In this experimental system, Fe65 nuclear translocation precedes γ-H2AX accumulation, and the latter is decreased in the absence of Fe65. Therefore, it seems that nuclear Fe65 plays a key role in the response of the cell to DNA damage. Furthermore, the results available so far suggest that nuclear Fe65 should be in a specific conformation, which could depend on its interaction with APP (induction of open conformation; see ref.22). Another possibility is that the activation of Fe65 also depends on post-translational modification on DNA damage. This possibility is supported by the observation that the electrophoretic migration of nuclear Fe65 is modified, likely due to phosphorylation, within few minutes from DNA damage (31).
The molecular mechanisms underlying the function of Fe65 in the nucleus of cells exposed to DNA-damaging agents are still not completely understood. A first set of experiments demonstrated that the Fe65 knockdown hampers the recruitment of Tip60 at DNA strand breaks and in turn the acetylation of histone H4 (52). This event could explain the defects in the DNA damage response and in the DNA repair observed as a consequence of Fe65 suppression. One possibility is that active Fe65 functions as a scaffold protein that interacts with chromatin, thus favoring the recruitment of soluble Tip60. In addition, nuclear Fe65 stabilizes p53 (53), and this could be a further mechanism through which Fe65 contributes to the response of the cells to DNA damage.
In this case, the crosstalk between Fe65 and APP/AICD also remains to be addressed. The first issue is concerned with the possible regulation of Fe65 nuclear translocation on DNA damage. One interesting possibility is that Fe65, bound to APP cytodomain and thus in its “open” active conformation, is released from APP as a consequence of APP phosphorylation at Thr668, possibly catalyzed by c-Jun N-terminal kinase JNK (54). Nuclear Fe65 is then rapidly modified, likely phosphorylated, and it is necessary for Tip60 recruitment at the DNA breaks (see Fig. 2). Of course, Fe65 suppression leads to reduced phosphorylation of H2AX and decreased efficiency of DNA repair (52). The crucial involvement of APP in this machinery is supported by the observation that APP/APLP2 knockdown has the same effects observed with Fe65 suppression, that is, reduced recruitment of Tip60 at DNA breaks and reduced histone H4 acetylation (52). Considering that AICD seems not to be necessary in this machinery, the most likely explanation for the role of APP in the mechanism is that the interaction of Fe65 with APP is necessary to allow Fe65 to acquire the proper conformation.
Despite the large literature on the function of Fe65 and on the consequences of its interaction with APP, many questions are still waiting for a definitive answer. Similarly, the possibility that Fe65 dysfunction could have any role in AD is far to be addressed. The information accumulated up to now is surprising, because Fe65 seems to be involved in several different functions. The number of proteins that appear to have more than one function is growing; however, in many cases, this multiplicity of functions is probably due to the incomplete understanding of the protein. This is obviously the case of Fe65. Robust data indicate that Fe65 is an adaptor protein with a wide choice of possible interactors. It shuttles between two intracellular compartments: in the cytosol, it could be involved with APP, Mena, mDab1, and Abl in the regulation of cell movement; and in the nucleus, in connection with transcription factors and/or histone-modifying enzymes, it could play a role in the response to DNA damage and/or in transcription regulation. In both cases, the proteolytic processing of APP and/or its post-translational modifications could be affected by Fe65 and regulate Fe65 intracellular trafficking. A speculation that deserves to be explored in detail is that Fe65 and its partners are part of a molecular machinery working during development and differentiation and are involved in a concerted cellular response to stress conditions.
The work performed in the authors' laboratory was supported by the Associazione Italiana Ricerca sul Cancro, Italy (AIRC) and the Italian Ministry of University and Research (PON1_02782).