Turnover of Amyloid Precursor Protein Family Members Determines Their Nuclear Signaling Capability

The amyloid precursor protein (APP) as well as its homologues, APP-like protein 1 and 2 (APLP1 and APLP2), are cleaved by α-, β-, and γ-secretases, resulting in the release of their intracellular domains (ICDs). We have shown that the APP intracellular domain (AICD) is transported to the nucleus by Fe65 where they jointly bind the histone acetyltransferase Tip60 and localize to spherical nuclear complexes (AFT complexes), which are thought to be sites of transcription. We have now analyzed the subcellular localization and turnover of the APP family members. Similarly to AICD, the ICD of APLP2 localizes to spherical nuclear complexes together with Fe65 and Tip60. In contrast, the ICD of APLP1, despite binding to Fe65, does not translocate to the nucleus. In addition, APLP1 predominantly localizes to the plasma membrane, whereas APP and APLP2 are detected in vesicular structures. APLP1 also demonstrates a much slower turnover of the full-length protein compared to APP and APLP2. We further show that the ICDs of all APP family members are degraded by the proteasome and that the N-terminal amino acids of ICDs determine ICD degradation rate. Together, our results suggest that different nuclear signaling capabilities of APP family members are due to different rates of full-length protein processing and ICD proteasomal degradation. Our results provide evidence in support of a common nuclear signaling function for APP and APLP2 that is absent in APLP1, but suggest that APLP1 has a regulatory role in the nuclear translocation of APP family ICDs due to the sequestration of Fe65.


Introduction
The amyloid precursor protein (APP) is a type I transmembrane glycoprotein, encoded by a single gene on chromosome 21q21, which is causally involved in Alzheimer's disease (AD) [1]. Fulllength APP is processed by a series of proteolytic cleavage reactions, mediated by the enzymes a-, band c-secretase [2]. Cleavage by either aor b-secretase results in the liberation of the soluble N-terminal fragments, sAPPa and sAPPb, and the membrane bound C-terminal fragments, C83 and C99. The Cterminal fragments are further cleaved by c-secretase, producing the extensively studied Ab peptide from C99, which is regarded as a central player in AD [3]. Cleavage of APP C-terminal fragments by c-secretase at the e-site releases the APP intracellular domain (AICD), which has been shown to signal to the nucleus and play a role in transcriptional regulation [4,5,6,7]. AICD-regulated genes include KAI1, APP, BACE1, neprilysin, and p53 [4,8,9,10,11].
APP is a member of a highly conserved family of glycoproteins, which includes the APP-like proteins 1 (APLP1) and 2 (APLP2) in vertebrates, APPL in Drosophila, and APL-1 in C. elegans. APP and APLP2 are ubiquitously expressed, with high expression in the brain [12,13], while APLP1 expression is neuron specific [14]. Interestingly, APP-like proteins lack the Ab domain, which is a feature unique to APP itself [15,16]. Human APLP2 is located on chromosome 11q23-q25 and exists in two alternatively spliced forms, one of which, similarly to APP, contains a KPI domain [15,17,18]. Human APLP1 is located on chromosome 19q13.1 and, as yet, no spliced transcripts have been identified [19,20]. All three APP family members have been shown to bind both zinc and heparin [21] and are thought to play an important role in cell adhesion in both a homo-and heterotypic manner [22,23]. Furthermore, all family members interact with PAT1a via their basolateral sorting signal, promoting intracellular transport and increasing the processing of APP/APLPs [24].
Despite the structural homology and conserved domain structure of APP family members, it has been shown that their subcellular localization differs strikingly. A recent study by Kaden et al. showed that APP and APLP2 mainly localize within intracellular compartments, such as the ER and endosomes, with only low levels at the plasma membrane. In contrast, APLP1 was found to mainly localize to the plasma membrane, corresponding to an increased tendency to form in-trans interactions at cell-cell contacts, highlighting an important role for APLP1 in cell adhesion [25]. Further studies investigating the binding of the APP family to adaptor proteins have also identified differences between family members. For example, the binding of JIP-1 to APP has been shown to result in transcriptional activation, whereas expression of JIP-1 with APLP1 or APLP2 showed little or no transcriptional activity [26].
Studies using knockout mice have revealed important insights into the relationship between different APP family members. Single knockout for APP, APLP1, or APLP2 cause minor phenotypes due to a clear redundancy within the APP gene family [27,28,29]. However, it is also clear that the different family members exhibit different functions. Mice with both APP and APLP1 disrupted are viable and fertile, while those that are double knockout for both APP and APLP2 or APLP1 and APLP2 die shortly after birth [27,28]. These findings point to an important developmental role for APLP2 that is required when either APP or APLP1 are absent. Combined APP and APLP2 double knockout exhibit a poorly formed neuromuscular junction and reduced numbers of presynaptic vesicles [30], which supports findings in Drosophila models with mutant APPL [31], implicating a role for the APP family in the regulation of synaptogenesis. Triple knockout mice, lacking all three APP family members, were found to die soon after birth and display cortical dysplasia, which resembles human type 2 lissencephaly, and is characterized by fragmented basal lamina and over-migration of neurons [32]. This phenotype is very similar to the phenotype reported in Fe65 family knockout mice [33], highlighting the importance of this AICDbinding protein in determining the functions of the APP family members.
Both APLP1 and APLP2, like APP, can be cleaved by a-, band c-secretase, and intracellular domains are released after csecretase cleavage at the e-site [34,35,36,37]. Furthermore, BACE1 cleavage of APLP1 has recently been shown to result in the formation of an Ab-like peptide that does not aggregate and can be a surrogate marker for increased b-cleavage in cerebrospinal fluid [38]. All three APP family members have been shown to be cleaved by caspases at their conserved VEVD motif within the C-terminus, although the significance of this cleavage is still debated [39,40].
Proteolytic processing of APP, APLP1 and APLP2 by csecretase-mediated e-cleavage generates intracellular domains (ICDs), which are stabilized by binding to the adaptor protein Fe65 and, when coexpressed with Fe65, can transactivate Gal4-Tip60 constructs [34,41]. Transgenic mice expressing AICD and Fe65 are reported to show pathological features of AD [42]. We have previously shown that AICD is transported to the nucleus by Fe65 where, together, they bind Tip60 and form spherical nuclear complexes (AFT complexes), which localize in transcription factories [4,43]. We have now investigated the nuclear localization of APLP1 and APLP2 ICDs upon Fe65 and Tip60 coexpression. Furthermore, we analyzed the differences in subcellular localization and turnover of the full-length proteins and their ICDs. We have identified dramatic differences in both the subcellular localization and signaling capability of APLP1 compared to the other two APP family members. Our results provide evidence for a nuclear signaling function of the APP and APLP2 ICDs and support a distinct physiological role for APLP1.

Results
ICDs Derived from APP and APLP2, but not APLP1, Localize to Spherical Nuclear AFT Complexes The members of the APP family-APP, APLP1, and APLP2show a high degree of sequence homology. In particular, the intracellular domain of APP family members is highly conserved and contains common motifs such as a caspase cleavage site and a YENPTY domain for binding of adaptor proteins. We have previously shown that AICD bound to Fe65 translocates to the nucleus and forms, together with Tip60, spherical nuclear AFT complexes [43]. We were therefore interested whether APLP1 and APLP2 intracellular domains (subsequently referred to as AL1ICD and AL2ICD) form nuclear AFT-like complexes similar to AICD.
Coexpression of Fe65, Tip60, and APP in HEK cells resulted in the retention of Fe65-AICD complexes in the nucleus and the formation of AFT complexes as reported previously [4,44] (Fig. 1A, top row). For detection of AFT complexes, Tip60 and APP were fused to fluorescent proteins, while Fe65 was detected by antibody staining. As already shown in an earlier publication, antibody access to AFT complexes is restricted due to their density [44]. Therefore the localization of Fe65 to AFT complexes is sometimes not as clear as observed for the fluorescently tagged proteins. Similar to AICD, AL2ICD also formed spherical AFT complexes with Fe65 and Tip60 in the nucleus (Fig. 1A, bottom row). Surprisingly, AL1ICD did not localize to the nucleus and Fe65 was retained in extranuclear compartments, colocalizing with fulllength APLP1 (Fig. 1A, middle row). Cytosolic retention of Fe65 prevented the redistribution of Tip60 to nuclear spots, resulting in an accumulation of Tip60 in nuclear speckles. (Further confocal fluorescence images of HEK and N2a cells expressing APP family members together with Tip60 and Fe65 are available in Fig. S1 and Fig. S2.).
To exclude the possibility that the observed different nuclear signaling capabilities of APP family members were due to properties of the chosen cell line, experiments were repeated in N2a mouse neuroblastoma cells. Coexpression of Fe65, Tip60, and APP led to the formation of spherical nuclear AFT complexes (Fig. 1B, top row). Similar to our observations in HEK cells, AL2ICD formed AFT complexes (Fig. 1B, bottom row), while AL1ICD was not detected in nuclear structures and, due to the sequestration of Fe65 by full-length APLP1, Tip60 was localized to nuclear speckles (Fig. 1B, middle row).

APP and APLP2 Exhibit Faster Protein Turnover than APLP1
To further study the subcellular localization of the APP family members we expressed APP/APLP with N-terminal 3 myc tags, preceded by the APP signal peptide to ensure membrane insertion, and C-terminal 3HA tags ( Fig. 2A). A GAPDH promoter was chosen for expression because the GAPDH gene is constitutively expressed at high levels in almost all tissues and expression of transgenes via this mammalian promoter yields good expression that is weaker than with viral promoters. Expression of APP family members in HEK cells showed a clear intracellular localization to vesicular structures for APP and APLP2, whereas APLP1 was mostly localized to the cell membrane (Fig. 2B). The different subcellular localization of APP and APLP1 was also observed with coexpression of Citrine-labeled APP and Cerulean-labeled APLP1 constructs (Fig. S3A). Furthermore, fluorescent resonance energy transfer (FRET) could be observed between APP and APLP2 but not between APP and APLP1 in HEK cells, primary astrocytes, and neurons ( Fig. S3 B-D). Thus, in addition to the colocalization of APP and APLP2, these two family members can form heterodimers.
The N-terminal 3 myc tag-constructs allowed cell-surface labeling of APP/APLPs on living cells (Fig. 2C). Antibody incubations were performed at 4uC to inhibit endocytosis that is dependent on GTPase function. In contrast, exocytosis, which is promoted by zippering of SNARE proteins mediating vesicle fusion, is still able to occur in the absence of enzymatic activity. After 10 minutes of anti-myc antibody incubation we detected surface labeling for all three APP family members. The strongest staining was observed for APLP1. However, after 30 minutes incubation, APP and APLP2 surface staining reached similar levels as APLP1. This points towards a higher turnover of APP and APLP2 compared to APLP1 at the plasma membrane.
To measure the half-lives of the APP family members we inhibited protein synthesis for different durations. HEK cells were transfected with C-terminally 3HA-tagged APP/APLP constructs, all driven by a GAPDH promoter to ensure comparable expression levels. 24 hours after transfection, protein synthesis was inhibited with cycloheximide (CHX). Cells were harvested after the indicated time of cycloheximide incubation and levels of full-length protein determined by Western blot (Fig. 3A). APP/ APLP full-length levels were normalized to the highly stable atubulin. APP and APLP2 displayed short half-lives of 43 minutes and 53 minutes respectively, while half-life of APLP1 was higher than five hours (308 minutes) (Fig. 3B). The much higher stability of APLP1 is also evident in the left column of figure 3A: despite the expression of all three family members under the same promoter, the levels of APLP1 were much higher, again highlighting its slower turnover.

The N-terminus of ICDs Determines their Nuclear Signaling Capabilities
To investigate whether differences in nuclear localization capability of AICD and AL1ICD are mediated by properties of their extracellular or intracellular domains, we constructed chimeric APP/APLP1 expression plasmids. ICDs of APP and APLP1 were joined at the e-cleavage site of APLP1 and APP, resulting in the chimeric constructs APLP1-AICD and APP-AL1ICD, respectively (Fig. 4A). Chimeric constructs were cotransfected with Fe65 and Tip60 in HEK cells and AFT complex formation was observed by confocal microscopy. APP-AL1ICD did not form AFT complexes, despite preserved binding of Fe65 (Fig. 4B, top row). In contrast, nuclear AFT complex formation was observed in cells transfected with APLP1-AICD. These results suggest that the formation of AFT complexes is determined by the properties of the intracellular domain. Interestingly, the size of AFT complexes formed from APLP1-AICD appeared to be decreased when compared to AFT complexes formed from wildtype APP (Fig. 1A, top tow).
Because the chimeric proteins carry a chimeric e-cleavage site it is possible that absent nuclear signaling of ICDs is due to impaired c-secretase cleavage. To investigate this possibility, HEK cells were transfected with APP, APLP1 or the chimeric APP/APLP1 proteins and treated with the c-secretase inhibitor DAPT (Fig. 4C). DAPT treatment resulted in an accumulation of C-terminal fragments (CTFs) when compared to non-treated cells, demonstrating that c-secretase cleavage is not impaired in the chimeric proteins. In line with the previous experiments using cycloheximide, increased full-length levels and decreased CTF levels of APLP1 indicate the slower turnover of APLP1 compared to APP. Furthermore, these results suggest that the extracellular and/or transmembrane regions mediate the differences in protein turnover. To better understand the turnover of APP family members, wildtype and chimeric APP-transfected HEK cells were treated with cycloheximide to monitor the turnover of full-length proteins. As described earlier, APP turnover is much faster than APLP1 turnover (Fig. 4D). Turnover of the chimeric proteins was in between that of the wildtype proteins but resembled more closely the protein carrying the respective extracellular domain. Together, these results suggest that both extracellular and intracellular regions determine the turnover of APP family members. Moreover, it is likely that the decreased AFT complex formation for APLP1-AICD as compared to APP is due to the decreased turnover of APLP1-AICD, which would also lead to decreased AICD levels.
Our results show that nuclear signaling by the APP family members is determined by properties of the ICDs, which are highly conserved and share common motifs, such as a caspase cleavage site and the YENPTY motif (Fig. 5). We reasoned that the differences in AFT complex formation capability between APP/APLP2 and APLP1 must derive from single amino acid or motif differences between the proteins. We identified 17 conserved residues in AICD and AL2ICD that are not conserved in AL1ICD and could thus be responsible for nuclear signaling (Fig. 5). To test this hypothesis, we exchanged individual or multiple residues of APP by the corresponding residues of APLP1 and investigated AFT complex formation. Mutation of the NPTY motif to NATA abolished Fe65 binding as previously reported [6] and prevented the nuclear translocation of AICD (Fig. 6A, row 2). We subsequently mutated amino acids sequentially, either alone or in combination. (See Fig. S4 for a diagram depicting all mutations and Fig. S5 for corresponding confocal fluorescence pictures.) Changing all seven non-conserved residues in the C-terminal region of AICD that is reported to bind to Fe65 [45] did not disrupt nuclear signaling (Fig. 6A, row 3; APP(7 xmut)). In contrast, when we inserted the first 12 amino acids of AL1ICD into the APP sequence (APP(AL1ICD-AICD38)), nuclear signaling was completely abolished (Fig. 6A, row 4). Further site-directed mutagenesis were done to identify the minimal set of amino acids that prevent nuclear localization of AICD. By this means we discovered that exchange of the N-terminal residues VML to the corresponding APLP1 residues LLR (APP(VML646LLR)) is sufficient to ablate AFT complex formation (Fig. 6A, row 5).
Co-immunoprecipitation of streptavidin-binding protein (SBP)tagged APP and APP(VML646LLR) showed that Fe65 binding is not impaired by exchange of N-terminal ICD residues (Fig. 6B). Since DAPT treatment of cultures transfected with APP(VML646LLR) resulted in an accumulation of CTFs, we can also exclude the possibility that the absence of nuclear AFT complex formation for APP(VML646LLR) is due to disturbed csecretase cleavage (Fig. 6C). Interestingly, in DMSO control treated cells, ICDs derived from APP were detected but not from APP(VML646LLR) or APLP1.

Proteasomal Degradation Differs between AL1ICD and AICD
Several reports have demonstrated that AICD is degraded by the proteasome [46,47], but little is known about AL1ICD and AL2ICD degradation. We hypothesized that APP family ICDs have a different protein turnover, resulting in different capabilities to form nuclear AFT complexes. To confirm proteasomal degradation of AICD, we treated HEK cells transfected with APP-Cit with two previously described proteasome inhibitors, epoxomicin and MG-132. Treatment with epoxomicin strongly increased AICD levels, confirming proteasomal degradation of AICD, while treatment with MG-132 resulted in an increase in CTF and AICD levels, suggesting that MG-132 inhibits the proteasome as well as c-secretase when used in higher concentrations, as reported previously [48] (Fig. 7A). To avoid interference with secretase processing of APP, epoxomicin was used for proteasomal inhibition in the following experiments. HEK cells were transfected with APP-Cit, APP(VML646LLR)-Cit, or APLP1-Cit and treated with epoxomicin for six hours. In control treated cells, only ICDs generated from APP were detectable, while epoxomicin increased the ICDs to detectable levels for the different constructs (Fig. 7 B-C). These results show that the halflife differs between APP family ICDs and suggest that proteasomal degradation efficacy is dependent on the N-terminus generated after c-secretase cleavage.
Next, we asked whether the localization of APP family ICDs that do not form AFT complexes is changed after epoxomicin treatment. Both APP(VML646LLR)-Cit and APP-AL1ICD-Cit, coexpressed with Fe65 and Tip60, formed AFT complexes when treated with epoxomicin (Fig. 7D). These results show that both AICD and AL1ICD can be transferred to the nucleus and form AFT complexes but under normal conditions AL1ICD is degraded very quickly by the proteasome due to the identity of its Nterminal residue. In line with this, we showed that AL1ICD with an N-terminal fusion of Citrine that prolongs the half-life, resulted in nuclear localization of AL1ICD to AFT-like complexes when coexpressed with Fe65 and Tip60 (Fig. 8A).

APLP1 Regulates APP Nuclear Signaling
APLP1 does not signal to the nucleus but it nevertheless binds to Fe65. We therefore investigated the influence of APLP1 expression on the nuclear signaling of AICD. We used a myc-tagged Tip60 to be able to coexpress Cerulean and Citrine-tagged APLP1 and APP. AICD cleaved from APP again translocated to nuclear AFT complexes in cells coexpressing Tip60 and Fe65 (Fig. 8B). APLP1 bound to Fe65 with higher affinity than Tip60, as seen by the relocalization of Fe65 away from Tip60 in the nucleus. Consequently, coexpression of APLP1 together with APP, Fe65 and Tip60, prevented the formation of nuclear AFT complexes that are clearly seen in cells not expressing APLP1 (Fig. 8B). These results suggest that although APLP1 does not directly signal to the nucleus, it influences AICD nuclear signaling via the sequestration of Fe65.

Discussion
Although the APP family members share a high sequence homology and undergo similar processing, different properties and functions of the three proteins have been described [25]. Here, we provide further experimental evidence for a distinct function of APLP1 among APP family members, by reporting a unique nuclear signaling capability for the ICDs of APP and APLP2, but not APLP1.
We show that the ICDs released from APP and APLP2 localize together with Fe65 and Tip60 to spherical nuclear complexes. In contrast, the ICD released from APLP1 does not localize to the nucleus, although it is able to bind Fe65. We demonstrate that nuclear localization of AL1ICD is prevented at two steps of APLP1 processing. Firstly, slower turnover of full-length APLP1 compared to APP results in lower levels of CTFs, which are the direct precursors of ICDs. Secondly, AL1ICD undergoes faster proteasomal degradation compared to AICD. Both steps ultimately result in lower levels of AL1ICD. Our results also indicate that, although APLP1 does not signal directly to the nucleus, it has a regulatory function in AICD nuclear localization.
The expression of full-length APP or APLP2, together with Fe65 and Tip60, results in the translocation of their ICDs to nuclear AFT complexes, which is dependent on cleavage by csecretase [4]. In contrast, the ICD derived from APLP1 does not translocate to the nucleus. APLP1 binds and sequesters Fe65 outside of the nucleus, while Tip60 remains localized in nuclear speckles. These speckles represent a different nuclear compart-   to that occupied by nuclear AFT complexes, which themselves are thought to correspond to sites of transcription [43]. Of note, AICD and AL2ICD occupy the same nuclear sites as the transcriptional activator NICD and interactions of AICD and NICD nuclear signaling have been described [49] (data not shown). This could indicate that similar genes are regulated by the ICDs of APP and APLP2. In fact, the analysis of candidate genes regulated by AICD has shown that neprilysin expression and activity is reduced in fibroblasts derived from APP or APLP2 knockout mice and is dramatically diminished in cells from double knockout mice [9]. Likewise, for genes suppressed by AICD, as reported for LRP1, the expression levels were even stronger in APP/APLP2-deficient cells compared to single APP knockout cells [50]. Taking this combined evidence into account, these results point towards an exclusive function of APP and APLP2 in transcriptional regulation.
Using cell surface labeling of living cells and determination of protein half-life time after cycloheximide-induced inhibition of translation, we clearly show that full-length APP and APLP2 have a much faster turnover than APLP1. The stability of APLP1 is evident in different experimental settings and in stark contrast to the rapid turnover of APP and APLP2, whose rapid turnover kinetics have also been reported in vivo [51]. Investigation of the nuclear signaling capabilities of chimeric APP/APLP1 proteins suggested that nuclear signaling capability is an intrinsic property of the ICD sequence that is present in APP but not in APLP1. We  aligned the ICD sequences and identified 17 amino acids that are conserved between APP and APLP2 but differ in APLP1. To identify sequence determinants for nuclear signaling capability we individually or in combination exchanged all amino acids of AICD with the corresponding APLP1 residues. Out of all single amino acid and motif changes, only the exchange of the N-terminal amino acids VML to LLR ablated nuclear signaling of AICD. APP(VML646LLR) binding to Fe65 and processing by the secretases were intact, but the resulting ICD levels were much lower than for the APP-derived ICDs, suggesting different degradation rates.
It has been reported that AICD is degraded by insulindegrading enzyme (IDE), the lysosomal pathway, and the proteasome, whereas degradation of ALICDs has only been poorly investigated [47,52]. Here, we show that the ICDs of both APP and APLP1 are degraded by the proteasome. Proteasomal degradation of AL1ICD is faster than for AICD. Inhibition of proteasomal degradation by epoxomicin and the concomitant increase in ICD levels resulted in the translocation of ICDs derived from APP(VML646LLR) and the chimeric construct APP-ALICD to the nucleus. Under normal conditions the ICDs of these constructs are degraded and do not reach the nucleus. However, ICDs released from APLP1, even with epoxomicin treatment, did not translocate to the nucleus. This is probably due to the fact that APLP1, because of its slower turnover, accumulates to higher levels that bind and sequester Fe65, making it unavailable for the transport of AL1ICD to the nucleus. These results suggest that: (a) nuclear localization is an intrinsic property of all APP family ICDs, (b) a certain threshold level for ICDs has to be reached for nuclear translocation to occur, and (c) slow turnover of full-length APLP1 sequesters Fe65, rendering it unavailable for nuclear translocation of ICDs.
Since the levels and proteasomal degradation of ICDs of APP(VML646LLR) and APLP1 behave in a similar way, our results suggest that the proteasomal degradation rate of ICDs is dependent on their N-terminal residues. Interestingly, work by Alexander Varshavsky and colleagues has revealed a proteasomal pathway in which the degradation of a protein is related to the identity of the protein's N-terminal residues [53]. According to this N-end rule, leucine, the N-terminal residue of APLP1, is a destabilizing residue, whereas valine and methionine, the Nterminal residues of APP, are stabilizing. Since both, APP and APLP1 are degraded by the proteasome and carry multiple lysines that could potentially be ubiquitinated, it is possible that APP family ICDs, as suggested previously [54], are substrates of the Nend rule pathway.
Walsh et al. reported higher stability for AL1ICD than AICD and AL2ICD when ICDs were expressed as soluble forms [37]. Similar to Walsh et al., we observe nuclear signaling for ICDs of all three APP family members when ICDs are expressed as soluble proteins (Fig. 8A). The apparent differences in localization for AL1ICD derived from full length APLP1 and soluble expressed AL1ICD are not surprising since much higher levels of ICDs are obtained from soluble expressed AL1ICD. Furthermore, we have shown that AL1ICD generated by c-secretase cleavage carries Nterminal residues that influence degradation rate. However, the expression of soluble ICDs necessitates the addition of a methionine at the N-termini that is likely to neutralize the different degradation times of AICD and AL1ICD reported here.
Our results help to explain the profound effects seen in knockout mouse models of the APP family. Animals with a double deficiency in both APP and APLP2 die shortly after birth [27,28]. Knock-in mice have been generated to identify the domain of APP responsible for these effects. Mutation of the first tyrosine residue, Tyr-682, of the YENPTY motif abolished Fe65 binding, and thus nuclear signaling by AICD [55]. APP Y682G knock-in mice crossed to an APLP2 knockout strain were found to be postnatally lethal with neuromuscular synaptic defects, resembling APP/ APLP2 double knockouts [56]. This points towards a requirement for AICD/AL2ICD-mediated nuclear signaling in development, a function that cannot be performed by APLP1. This is further supported by the fact that knock-in of the sAPPb or sAPPa ectodomains alone is not sufficient to rescue the synaptic deficits in APP/APLP2 knockout mice [57,58]. Although the phenotypes in these mouse models could also originate from other aspects of APP biology-the Y682G knock-in has a dramatically increased production of sAPPa and the sAPP knock-ins also lack the Ab sequence in addition to AICD-the current data combined with our experimental evidence strongly support a role of nuclear signaling by AICD and AL2ICD in synapse formation that cannot be performed by the AL1ICD.
We propose that AL1ICD does not have a direct physiological role in transcriptional regulation, but through sequestration of Fe65 by APLP1 this family member might act as a repressor of AICD-mediated nuclear signaling. We conclude that nuclear signaling is a prime function of the APP and APLP2 ICDs. Our results also caution the use of drugs inhibiting APP processing as treatment for AD because of possible interference with APP/APLP transcriptional function.

Ethics Statement
All animal experiments were performed according to the Swiss animal protection law (TschG) and approved by the local animal committee (Kantonales Veterinä ramt Zürich; license 38/2011).

Expression Constructs
APP-Citrine, HA-Fe65, and CFP-Tip60 expressing constructs have been described previously [4]. Clones of APLP1 (RZPD, Heidelberg, Germany) and APLP2 (courtesy of Stefan Kins, TU Kaiserslautern, Germany) were used to amplify open reading frames and ICDs and cloned into the pUKBK expression vector system [59] with either CMV or GAPDH promoters. Proteins were C-terminally tagged with Citrine (Cit), Cerulean (Cer), or streptavidin-binding protein (SBP), generating APP-Cer, APLP1-Cit, APLP1-Cer, APLP2-Cit, APLP2-Cer, and APP-SBP. For the constructs Cit-AICD, Cit-AL1ICD, Cit-AL2ICD, and Cer- AL2ICD fluorescence proteins were fused to the N-terminus of the respective ICD sequence. To generate Myc-Tip60, CFP was exchanged with a myc tag. GAPDH promoter driven 3myc-APP-3HA plasmid was constructed by inserting a 3myc-tag coding oligonucleotide after the APP signal peptide (SP) by restriction based cloning. For the corresponding APLP constructs, APP was replaced by open reading frames of APLP1 and APLP2. APP residues were changed to APLP1 residues by site-directed mutagenesis. Chimeric APP/APLP1 constructs, APP-AL1ICD and APLP1-AICD were derived from APP and APLP1 constructs by PCR-driven overlap extension as described [60].

Live Antibody Incubation
HEK cells were, 24 hours after transfection, incubated with mouse anti-myc antibodies (1:100; Roche) at 4uC for 10 or 30 minutes. After washing with ice-cold PBS, cells were fixed and stained for HA and myc tags.

Western Blotting
Cell were lysed in a buffer, containing 150 mM NaCl, 10 mM Tris pH 7.6, 1% Triton X100, 0.25% NP40, and protease inhibitor (Roche) and centrifuged for 5 minutes at 14000 rpm [61]. Equal amounts of supernatants were separated on 10-20% Tricine gels (Invitrogen). Protein bands were visualized by ECL (Thermo Scientific, Rockford, IL, USA) using a Fuji LAS 3000 Imager. The following primary antibodies were used in this study: