Nuclear Translocation Uncovers the Amyloid Peptide Aβ42 as a Regulator of Gene Transcription*♦

Background: Biological activities of nontoxic Aβ42 peptides remain unclear in Alzheimer disease. Results: Aβ species are taken up in the nucleus of cells by a nonregulated mechanism, but only Aβ42 plays a role in gene transcription. Conclusion: Aβ42 may act as a transcriptional regulator, similar to the cytoplasmic fragment AICD. Significance: Genes regulated by nuclear Aβ42 could represent alternative targets for therapeutic approaches. Although soluble species of the amyloid-β peptide Aβ42 correlate with disease symptoms in Alzheimer disease, little is known about the biological activities of amyloid-β (Aβ). Here, we show that Aβ peptides varying in lengths from 38 to 43 amino acids are internalized by cultured neuroblastoma cells and can be found in the nucleus. By three independent methods, we demonstrate direct detection of nuclear Aβ42 as follows: (i) biochemical analysis of nuclear fractions; (ii) detection of biotin-labeled Aβ in living cells by confocal laser scanning microscopy; and (iii) transmission electron microscopy of Aβ in cultured cells, as well as brain tissue of wild-type and transgenic APPPS1 mice (overexpression of amyloid precursor protein and presenilin 1 with Swedish and L166P mutations, respectively). Also, this study details a novel role for Aβ42 in nuclear signaling, distinct from the amyloid precursor protein intracellular domain. Chromatin immunoprecipitation showed that Aβ42 specifically interacts as a repressor of gene transcription with LRP1 and KAI1 promoters. By quantitative RT-PCR, we confirmed that mRNA levels of the examined candidate genes were exclusively decreased by the potentially neurotoxic Aβ42 wild-type peptide. Shorter peptides (Aβ38 or Aβ40) and other longer peptides (nontoxic Aβ42 G33A substitution or Aβ43) did not affect mRNA levels. Overall, our data indicate that the nuclear translocation of Aβ42 impacts gene regulation, and deleterious effects of Aβ42 in Alzheimer disease pathogenesis may be influenced by altering the expression profiles of disease-modifying genes.

Although soluble species of the amyloid-␤ peptide A␤42 correlate with disease symptoms in Alzheimer disease, little is known about the biological activities of amyloid-␤ (A␤). Here, we show that A␤ peptides varying in lengths from 38 to 43 amino acids are internalized by cultured neuroblastoma cells and can be found in the nucleus. By three independent methods, we demonstrate direct detection of nuclear A␤42 as follows: (i) biochemical analysis of nuclear fractions; (ii) detection of biotinlabeled A␤ in living cells by confocal laser scanning microscopy; and (iii) transmission electron microscopy of A␤ in cultured cells, as well as brain tissue of wild-type and transgenic APPPS1 mice (overexpression of amyloid precursor protein and presenilin 1 with Swedish and L166P mutations, respectively). Also, this study details a novel role for A␤42 in nuclear signaling, distinct from the amyloid precursor protein intracellular domain. Chromatin immunoprecipitation showed that A␤42 specifically interacts as a repressor of gene transcription with LRP1 and KAI1 promoters. By quantitative RT-PCR, we confirmed that mRNA levels of the examined candidate genes were exclusively decreased by the potentially neurotoxic A␤42 wild-type peptide. Shorter peptides (A␤38 or A␤40) and other longer peptides (nontoxic A␤42 G33A substitution or A␤43) did not affect mRNA levels. Overall, our data indicate that the nuclear translocation of A␤42 impacts gene regulation, and deleterious effects of A␤42 in Alzheimer disease pathogenesis may be influenced by altering the expression profiles of disease-modifying genes.
The amyloid precursor protein (APP) 4 is first cleaved by the ␤-site APP-cleaving enzyme (BACE1) and sequentially processed by the ␥-secretase complex to generate amyloid-␤ (A␤) peptides of varying lengths encompassing 38, 40, 42, and 43 residues (1). A␤ generation, through amyloidogenic processing of APP, results in the simultaneous production of a C-terminal fragment corresponding to the APP intracellular domain (AICD). This has been reported to translocate into the nucleus and activate gene transcription (2,3).
A␤42 is hypothesized to be the main culprit in the pathogenesis of Alzheimer disease (AD) as it was postulated to impair synaptic function and initiate neuronal degeneration (4). In AD, soluble species of A␤42 are more strongly correlated with disease symptoms than with amyloid plaques (5)(6)(7)(8)(9). A␤42 was also reported to have an effect on differentiation and death of cultured neural stem or progenitor cells (10). Intraneuronal A␤ accumulation in brains of patients with AD, in animal models, and in cultured cells has suggested a pathophysiological role specific for A␤40 and A␤42 (11). Moreover, oxidative DNA damage in guinea pig primary neurons was shown to induce A␤42 accumulation in the cytosol and to activate the p53 promoter (12). Similarly, under conditions of stress, Tau aggregates, which are the neuropathological hallmark of several neurodegenerative diseases, were found in the nucleus of neurons (13). In primary neurons, microinjection of A␤42 rapidly induced cell death, further underlining the neurotoxicity of intracellular amyloid (14). Accordingly, transgenic (tg) mice producing only intracellular A␤ developed neurodegeneration (15).
The presence of intraneuronal A␤ is explained by a dynamic relationship that exists between pools of intracellular and extracellular A␤ (16). Intraneuronal A␤ originates both from APP inside the neurons (17) and via uptake from the extracellular space. A␤ is internalized by neurons as well as by non-neuronal cells in culture, although the molecular events involved remain unclear (18 -20). A␤ has been shown to accumulate within certain organelles, including the endosomes/lysosomes, and mitochondria (21), resulting in endosomal/lysosomal leakage, mitochondrial dysfunction, and apoptosis (22). Less well characterized is the peptide's ability to create channel-like pores in membranes (23)(24)(25)(26).
Channel-like pores may allow direct passage of A␤ oligomers into the nucleus, although at present it is not known whether A␤ peptides have a biological activity in the nucleus. Here, we show that A␤ peptides of varying lengths such as A␤38, A␤40, A␤42, and A␤43 are internalized by neuroblastoma cells and are subsequently detected in the nucleus. The nuclear localization of internalized A␤42 peptides was further confirmed by both confocal and transmission electron microscopy (TEM). We also demonstrated the presence of endogenous A␤42 peptides in the nuclei of neurons of tg APP mice (APPPS1). Using the chromatin immunoprecipitation (ChIP) assay, A␤42 was found to specifically interact with the LRP1 and KAI1 promoter. Thus, at higher concentrations the widely recognized neurotoxic form A␤42 here acted at sublethal concentrations as a repressor of transcription of the genes LRP1 and KAI1. This result was confirmed by quantification of the mRNA levels of the examined candidate genes by qRT-PCR. Furthermore, we found that A␤42 increased the transcription of its own precursor gene APP. The mRNA levels were exclusively altered by the neurotoxic A␤42 wild-type peptide, whereas neither A␤38, A␤40, nor A␤43 had any effect on mRNA levels. Treatment with the nontoxic substitution peptide A␤42 G33A, which was used as a control because it forms ␤-pleated sheet aggregates like the wild-type peptide (5), also did not have an effect.
Although for all A␤ peptides tested a nuclear translocation was observed, albeit to a varying extent, only A␤42 entailed gene regulation. Thus, the major deleterious effects in the pathogenesis could be mediated by A␤42 gene control activity, because it specifically interacts as a repressor or activator, respectively, of gene transcription.

EXPERIMENTAL PROCEDURES
Cell Culture-SH-SY5Y and HEK-293 cells were purchased from DSMZ (ACC 209 and ACC 305) and were routinely cultured as described (5,27). Briefly, SH-SY5Y cells were synchronized in the G 1 phase of the cell cycle by double-thymidine block (2 mM for 15 h; Sigma). After blocks, cells were grown in fresh medium without thymidine for 8 or 2 h (medium containing 20% (v/v) fetal calf serum) and released to progress through the cell cycle. As growth medium of MEF PS1/2 KO cells, we used DMEM high glucose (PAA), 10% fetal calf serum (PAA), 2 mM glutamine (PAA), and 0.1 mM nonessential amino acids (PAA).
Nuclei Preparation and ELISA-Nuclei were isolated with the Nuclei EZ Prep nuclei isolation kit according to the manufacturer's instructions (Sigma) with minor modifications. Briefly, pellets containing nuclei were resuspended in PBS, sonicated, incubated at 95°C, and centrifuged for 10 min at 20,000 ϫ g. A␤42 ELISAs using the C-terminal specific G2-13 antibody were performed as described (28). To quantify A␤38, A␤40, A␤42, A␤42 G33A, and A␤43, the antibody 4G8 (HISS Diagnostics) recognizing A␤ residues 17-24 was used.
Chromatin Immunoprecipitation-ChIP was performed as described previously (30) with minor modifications. SH-S5Y5 cells were synchronized and grown on 100-mm dishes to a confluence of 80 -90%. At the indicated times after the addition of 1 M freshly dissolved A␤ peptide, cells were washed twice with PBS, fixed with 2 mM disuccinimidyl glutarate for 45 min at room temperature, and then cross-linked for 10 min at room temperature using 1% (v/v) formaldehyde. For immunoprecipitation, 2 g of anti-NCoR (Santa Cruz Biotechnology), anti-Fe65 (Millipore), anti-Tip60 (Millipore), anti-A␤ (W0 -2; TGC), or anti-NFB p50 (Santa Cruz Biotechnology) antibodies were used. For PCR analysis, 2 l of extracted DNA were used as template. Primers for the p50-binding motif in the KAI1 proximal promoter were described previously (31) and for the LRP1 promoter fragment were forward, 5Ј-GGG AGC CTG AAA TCC TAG AG-3Ј, and reverse, 5Ј-GGA AAG CGG TCC AAG AGT G-3Ј, and for the HES1 promoter fragment were forward, 5-CTC CCA TTG GCT GAA AGT TAC-3Ј, and reverse, 5Ј-TGG CTA CTT GGT GAT CAG TAG-3Ј. PCR cycles were 2 min at 94°C, 28 -32 cycles at 94°C for 15 s, annealing for KAI1 promoter (30 s at 69°C), for the HES1 promoter (30 s at 57°C) and LRP1 promoter (30 s at 60°C), elongation for 45 s at 72°C, final extension for 3 min at 72°C. PCR products were separated on 8% polyacrylamide gels.
Immunofluorescence-SH-SY5Y cells were treated with biotinylated A␤42 peptide 1 M. Cells were fixed and permeabilized with 3.3% formaldehyde containing 0.5% Triton X-100 followed by 125 mM glycine in PBS containing magnesium and calcium. Cells were blocked with 5% fetal bovine calf serum followed by the primary antibody. Biotin-A␤42 was detected with the monoclonal antibody AB (Sigma) or alternatively with Avidin Fluor488 (Sigma). Nuclei were stained with DAPI (Roche Diagnostics). Images were obtained using an LSM 510 meta (Carl Zeiss) confocal microscope.
Animal Studies and TEM-Transgene-negative littermate controls were obtained from the heterozygous breeding of APPPS1 ϩ/Ϫ mice. Genotyping was performed according to published protocols (32). All animal experiments were approved by the Berlin Office for Health and Social Services, Germany (O 0132/09).
Animals (12 months old) were perfused with PBS followed by fixative solution (4% formaldehyde, 0.2% glutaraldehyde in 50 mM sodium cacodylate buffer, pH 7.4). Hippocampi were dissected and post-fixed in the same solution at room temperature. Samples were embedded in LR-Gold resin (Science Services GmbH) and polymerized at 4°C. Ultra-thin sections were incubated with the G2-13 primary antibody labeled with 10 nm colloidal gold (BB International). The sections were counterstained with uranyl acetate followed by lead citrate.
Labeling of mAb G2-13 was performed with colloidal gold (BB International). After centrifugation (10 min, 6700 ϫ g), the supernatant of colloidal gold was counterstained with uranyl acetate and analyzed by TEM to ensure that the supernatant was free of gold aggregates. An amount of 500 g of mAb G2-13 was dialyzed with 2 mM sodium tetraborate. Equal volumes of G2-13 and colloidal gold were incubated for 20 min at RT. The stability of the gold/protein ratio was assessed by titration with 10% NaCl. Colloidal gold was adjusted to pH 9 in 100 mM potassium carbonate. The protein/gold solution was incubated in 1% BSA for 20 min. After centrifugation, the pellet was resuspended in 20 mM TBS, 1% BSA, and 0.05% sodium azide, pH 8.2, and the solution was centrifuged for 5 min at 6700 ϫ g. The supernatant contained G2-13 antibodies labeled with 10 nm of gold.
Modeling-The structure of the A␤ peptides was modeled using the NMR structure of A␤42 fibrils (Protein Data Bank code 2beg) (33) as a template as described previously (5). All peptides were energetically minimized with help of the GROMOS 43B1 force field to avoid distorted geometries. Electrostatic surface potentials were calculated using the program FIGURE 1. Purity of cytoplasmic and nuclear fractions and nuclear A␤ uptake. A, Western blot analysis of nuclear fractions for cytoplasmic or nuclear membrane marker proteins of SH-SY5Y cells. Cytoplasmic (CP) or nuclear membrane (insoluble nuclear fraction, INF) marker proteins are absent in the soluble nuclear fraction (SNF) that was analyzed for quantitative determination of A␤ by ELISA. Cytoplasmic or nuclear membrane marker proteins were found in the insoluble nuclear fraction (INF) that was not used for A␤ determination by ELISA. B, quantification of nuclear A␤ by ELISA after treatment of SH-SY5Y cells with freshly dissolved A␤ peptides. A␤38, A␤40, A␤42, A␤43, and A␤42 G33A are displayed as detected in nuclear fractions of cells treated with individual peptides for 2 h using the 4G8 antibody to the epitope residues 17-24; number of experiments (n) ϭ 4 -8. C, nuclear fractions were analyzed for A␤42 and A␤42 G33A by ELISA using the A␤42 C-terminal specific antibody G2-13 after treatment of SH-SY5Y cells for the indicated times. The graph shows an increase in A␤ compared with the endogenous level of untreated cells (*, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.0005); n ϭ 4 -7. D, HEK-293 and MEF PS1/2-KO cells were incubated for the indicated time points, and nuclear fractions were analyzed for A␤42 by ELISA using the C-terminal specific antibody G2-13. Compared with the endogenous A␤42 levels, the graph displays an increase over time (*, p Ͻ 0.05; **, p Ͻ 0.005), n ϭ 3-5.
Adaptive Poisson-Boltzmann Solver with nonlinear Poisson-Boltzmann equation and contoured at Ϯ5 kT/e. Statistical Analysis-Statistical significance between treated cells and control groups was determined using the one-way analysis of variance Dunnett's multiple comparison test. The number of experiments and p values are given in the respective figure legends.

RESULTS
A␤ Peptides Can Be Detected in the Nucleus-We first examined uptake of A␤ into neuroblastoma SH-SY5Y cells by biochemical means. We exposed cells for 2 h to 1 M concentrations of synthetic A␤ peptides with 38, 40, 42, and 43 residues. The purity of cytoplasmic and nuclear fractions was confirmed by Western blot analyses (Fig. 1A) before we analyzed them quantitatively for A␤ by ELISA. The nuclear uptake in SH-SY5Y cells was quantified using a monoclonal antibody recognizing the A␤ epitope formed by residues 17-24 (4G8). All peptides were detected in the nuclear fractions (Fig. 1B). Surprisingly, the nonaggregating A␤38 peptide accumulated equally well as the 42-residue-long substitution peptide A␤42 G33A (Fig. 1B). The latter is nontoxic and easily forms low n oligomers as we demonstrated previously by size exclusion chromatography (5) and presently by Western blot analysis (see below). To analyze the time course of A␤ uptake, SH-SY5Y cells were treated with A␤42 and A␤42 G33A peptides, respectively. The nuclear fractions were quantitatively analyzed by ELISA at different time points using the C-terminal specific A␤42 monoclonal antibody G2-13 (Fig. 1C). In particular, A␤42 and A␤42 G33A began to accumulate in cell nuclei within the first 30 min of treatment and accrued thereafter over 8 h. We repeated the experiments using HEK-293 cells to examine whether the observed nuclear uptake in SH-SY5Y neuroblastoma nuclei was cell type-dependent. In addition, we analyzed nuclei of MEF PS1/2 knock-out cells that lack the catalytic subunit PS1 of the ␥-secretase and treated the cells with A␤42 (34). Nuclear translocation of A␤ in all these cells appeared similar, and results from MEF PS1/2 knock-out cells indicated that at least a portion of the intra-nuclear peptide was not a consequence of cellular ␥-cleavage (Fig. 1D).
Notably, A␤ levels in Fig. 1, B and C, were quantified using monoclonal antibodies against two different epitopes. Values measured at the 2-h time point in Fig. 1, B and C, are similar for A␤42 G33A and A␤42, whereas the amount measured for A␤42 in Fig. 1B is 50% lower compared with Fig. 1C. This reduction might be due to epitope masking of the monoclonal 4G8 through oligomer formation, as observed previously for other conformation-dependent A␤ antibodies (35). This rationale is consistent with our previous study (5) that showed A␤G33A has a different aggregation profile compared with wild-type A␤42.
Although cellular and nuclear uptake is readily facilitated for shorter A␤ segments, there is no obvious rule for longer forms that vary in hydrophobicity or aggregation behavior (5). Our modeling data indicate that longer A␤ peptides such as A␤43 and the substitution peptide A␤42 G33A possess an enlarged hydrophobic surface compared with A␤42 ( Fig. 2A). An enlarged hydrophobic surface patch is effectively shielded from the polar milieu upon oligomerization explaining why A␤42 G33A tends to form higher oligomers compared with A␤42. This rationale is consistent with Western blot data showing that similarly to A␤43, freshly dissolved A␤42 G33A migrates at molecular weights in the range of 12-20-mers and higher (Fig.  2B). Nevertheless, there is no apparent correlation between peptide hydrophobicity and nuclear uptake because the highly hydrophobic A␤42 G33A behaves similarly to less hydrophobic peptides (Fig. 1C). These results imply that aggregation would not interfere with uptake.
To test whether concentrations of 1 M A␤42 were toxic, we monitored the viability of SH-SY5Y cells using the MTT assay, observing no substantial change when cells were exposed to 1 M synthetic A␤ peptides for as long as 12 h (Fig. 3A). At 2.5 M concentration, however, A␤43 reduced viability to 75% and A␤42 to 65% (Fig. 3B). Thus, all experiments executed with A␤42 and other peptides at 1 M concentrations were performed under sublethal conditions. We next studied the presence of intra-nuclear A␤ peptides in cultured cells using confocal laser scanning microscopy. Fig. 4A shows A␤42 localized in both the cytoplasm and nucleus of SH-SY5Y cells after 30 min of incubation and after 8 h. In the latter, large accumulations of A␤42 were seen in the cytoplasm (Protein Data Bank code 2BEG) when the peptides adopt ␤-sheet conformation. This enlarged hydrophobic surface patch is effectively shielded from the polar milieu upon oligomerization explaining why A␤42 G33A tends to form higher oligomers compared with A␤42. Electrostatic surface potentials were calculated using the program Adaptive Poisson-Boltzmann Solver contoured from red (negative) over white (neutral) to blue (positive). B, freshly dissolved synthetic A␤38 and A␤40 mainly yielded monomers and dimers. A␤42 and the substitution peptide A␤42 G33A appear as monomers to tetramers. A␤42 shows slight aggregation into high n oligomers, whereas a higher amount of high n oligomers is detected for the substitution peptide G33A. A␤43 is present at monomers to trimers and high n oligomers. and in the nucleus, as had been described previously for AICD complexes containing the adapter protein Fe65 and the histone acetyltransferase Tip60. Similar to AICD (36 -38), intra-nu-clear A␤ appeared as spherical spots (Fig. 4A). Finally, we confirmed the presence of intra-nuclear A␤ peptides by TEM not only in cultured SH-SY5Y cells (Fig. 4B) but also in brain tissue (Fig. 5). Synthetic A␤ was detected in the cytoplasm, mitochondria, and the nucleus in cultured SH-SY5Y cells by the monoclonal antibody W0-2 (Fig. 4B). In brain tissue from 19-month-old mice, the colloidal gold-labeled monoclonal G2-13 antibody directed against the C terminus of A␤42 specifically recognized both endogenous mouse A␤42 and A␤42 derived from overexpressed human APP. Thus, A␤42 was clearly evident in the nuclei of hippocampal neurons of wild-type mice and more strongly in neurons of tg APPPS1 mice, as expected due to the higher levels of expression (Fig. 5).
Taken together, the three independent methods consistently detected synthetic A␤ in the nucleus. Furthermore, the presence of human A␤42 in the nucleus derived from transgenic mice overexpressing human APP was shown by TEM. These data indicate that exogenously added A␤42 as well as endoge-  nous A␤42 are taken up by cells, and low amounts are always present in the nucleus.
A␤42 Associates with Gene Regulatory Elements-We next interrogated possible gene regulatory effects of nuclear A␤42 on promoters known to be targets of the AICD (2, 3). To assess possible effects of nuclear A␤42, we conducted ChIP assays to test whether A␤42 peptides affected the same promoters as described previously for AICD. AICD is a component of the AFT complex containing AICD, Fe65, and Tip60, which has been shown to bind the KAI1 (CD82) promoter and increases its transcription (36). Furthermore, the AFT complex was found to bind the LRP1 promoter and suppress LRP1 mRNA synthesis (39). As a control, we used the HES1 promoter that reportedly interacts with the Notch intracellular domain, which is generated by the ␥-secretase complex and is known to modulate transcription of downstream genes, including HES1 (39).
We found that treatment of SH-SY5Y cells with A␤42 peptides resulted in the association of Fe65 and Tip60 with both LRP1 and KAI1 promoters after 1 h (Fig. 6, A and B). Association of A␤42 with these promoters was already detectable after 30 min, indicating that A␤ peptides did not translocate into the nucleus with Fe65, unlike what has been reported for the AICD (2). The KAI1 promoter is under the control of p50 and the co-repressor N-CoR (31). Interestingly, A␤ seemed to cause the dissociation of N-CoR. Similarly, a subset of NFB-regulated genes was described to be activated upon export of the N-CoR corepressor complex as a consequence of high concentrations of AFT complexes (40). After 6 h of A␤ exposure, Fe65 and Tip60 were released from both promoters, whereas A␤ remained associated. Also, the HES1 promoter, which is activated by the ␥-secretase-derived Notch intracellular domain (39), was unaffected by A␤42 treatment (Fig. 6C), further suggesting the specificity of A␤42 activation indicating that the effects of A␤42 on the tested promoters were specific. Thus, treatment of SH-SY5Y cells with A␤42 resulted in the early association of the peptide with both LRP1 and KAI1 promoters, while the association of Fe65 and Tip60 with these promoters was delayed.
To investigate whether A␤42 association to promoters impacts gene regulation, we quantified the corresponding mRNA levels of these genes by qRT-PCR. The early downregulation of LRP1 and KAI1 mRNA by A␤42 (Fig. 7, A and  B) suggests a pathway different from AICD-regulated transcription (36,39), perhaps because the AICD must first form a complex with Fe65 before entering the nucleus (41) where Tip60 is then recruited (40,42). HES1 mRNA levels were not changed upon A␤42 treatment (Fig. 7C), further confirming our ChIP results. No changes in mRNA expression were detected upon treatment with the reversed A␤ sequence 42-1 peptide.
Importantly, we detected increased APP mRNA levels upon A␤42 treatment (Fig. 7D). Similarly to the detected effects on LRP1 and KAI1 mRNA levels, neither A␤42 G33A nor A␤38, A␤40, and A␤43 affected APP mRNA. Remarkably, the nontoxic A␤42 G33A peptide (5), which is also highly present in the nucleus (Fig. 1), did not affect mRNA expression levels (Fig. 7,  A-D). The latter can be regarded as the more rigid control, compared with the reverse sequence, because it forms ␤-pleated sheet aggregates like the wild-type peptide (5). Nota- FIGURE 5. Electron micrographs of hippocampal neuronal nuclei. Wild-type and APPPS1 double tg mice were analyzed for A␤42 with the colloidal gold-labeled antibody G2-13 against the C terminus of A␤42 that recognizes both endogenous mouse A␤42 and human A␤42. APPPS1 mice express both mouse and human APP. Upper insets show entire nuclei at lower magnification of selected areas stained with immunogold-labeled A␤42, and lower insets show higher magnifications of individual labeling (10 nm of gold, arrow). A considerably higher amount of A␤42 was detected in the nuclei of APPPS1 mice, as expected due to the overexpression of human APP yielding a higher level of A␤42. Scale bar, 210 nm. FIGURE 6. Association of A␤42 in SH-SY5Y cells with promoters analyzed by ChIP. Shown are PCR amplificates of DNA precipitated with the respective antibodies. Upon A␤42 treatment, nuclear A␤42 associated with the promoters of LRP1 (A) and KAI1 (B), followed by the adapter protein Fe65 and the histone acetyltransferase Tip60, both of which dissociated after 4 or 6 h, whereas A␤42 remained. Upon association of A␤42 to the KAI1 promoter, N-CoR dissociates within 1 h after treatment with A␤42 and re-associates after 8 h. C, A␤42 had no effect on the HES1 promoter during time course analysis. JULY 18, 2014 • VOLUME 289 • NUMBER 29 bly, other tested A␤ species such as A␤38, A␤40, and A␤43 also failed to alter mRNA levels (Fig. 7, A-D). Overall, the observed results indicate that the gene regulatory effects are specific to A␤42.

DISCUSSION
Intraneuronal A␤ aggregates in the brains of AD patients and animal models have been detected with antibodies specific for A␤40 and A␤42. This suggests a pathophysiological role for the A␤ pool (11,43,44). Our present study demonstrates that in addition to intracellularly produced A␤ peptides, exogenous A␤ species of varying lengths can be taken up from the medium by the cells and translocated to the nucleus. Previously, intracellular soluble A␤ was shown not to colocalize with internalized transferrin, excluding clathrin-mediated endocytosis as the primary uptake mechanism (45). Nevertheless, clathrinmediated A␤ endocytosis is possible and involves receptors that bind apolipoprotein E (apoE) (20). A␤ internalization by microglia was found nonsaturable, excluding receptor-mediated internalization (45). ApoE accelerates neuronal A␤ uptake, and the majority of endocytosed A␤ is suggested to traffic through early and late endosomes (46). Our data indicate that the uptake process may only be partially influenced by the overall peptide length and hydrophobicity and cannot be correlated with one of the reported mechanisms in the literature. Although the molecular mechanism of A␤ uptake may depend on the cell type and could be influenced by changes in A␤ composition and prevailing aggregation states, we successfully detected all major A␤ peptides (i.e. present in body fluids and cell culture supernatants (47)) in the nucleus of SH-SY5Y cells. Additionally, this study also detected A␤42 peptides in the nuclei of HEK293 and MEF PS1/2 knock-out cells. A␤ entry into cells could potentially occur through ligand-receptor type interactions as follows: e.g. A␤ interacts directly or indirectly with integrins, receptor for advanced glycation end products (RAGE), and APP itself (48); interactions via TrkA, p75NTR, some G-proteins, NMDA, and AMPA receptors (49), and/or interactions via the prion protein (50,51). Entry and transport could alternatively be limited to misfolded forms (e.g. aggregated Tau (52)), possibly involving a transcellular propagation (53) that could lead to wider cerebral A␤ distribution (54). Most likely, peptide conformation plays a role as revealed by the nontoxic control A␤42 G33A, which easily forms low and high n oligomers (5). For example, in astrocytes, it was shown that oligomeric A␤ is more efficiently taken up than fibrillar A␤ (55). Thus, A␤ might be able to enter the nucleus as an oligomeric complex comprising as many as nine A␤ peptide molecules as subunits. It is known that protein complexes of such sizes can pass through nuclear membrane pores because only translocation into the nucleus of proteins larger than 40 kDa requires specific transport receptors (56). In any case, given that (i) we did not apply stress conditions such as oxidative (12) or heat stress (13) and (ii) A␤ does not possess a canonical nuclear localization signal or a nuclear export signal, these peptides are most likely diffusing freely into the nucleus. They are likely retained by binding to nondiffusible nuclear components and might accumulate there.
In accordance with our findings, we anticipate that low n oligomers of aggregation-prone A␤ species, as well as monomers (e.g. of A␤34), are able to enter the nucleus. Also, it seems that peptide length, conformation, and oligomerization are important determinants of A␤ uptake into the nucleus, although their biophysical traits inconsistently influence the process.
Our ChIP results revealed that A␤42 specifically associated with AICD-regulated promoters of LRP1 and KAI1. Subsequent qRT-PCR analyses demonstrated that A␤42 diminished mRNA levels of LRP1 and KAI1. Previously, A␤ was found to localize to the nucleus under conditions of oxidative stress (57). Moreover, A␤42 was shown to bind to the APP promoter sequence using ChIP analysis and electrophoretic mobility shift assays (57,58). Here, we demonstrate an increase of APP mRNA levels upon A␤42 treatment. A␤ likely contains a helix-loop-helix structure (59), which is common to certain transcription factors. Thus, a direct binding of metastable oligomeric structures of A␤, described in preparations of amyloid-forming peptides such as ␣-synuclein Tau, prion, and A␤42, could mediate an interaction with DNA. In vitro, DNA binds all soluble aggregated forms of A␤42 indicating that DNA interaction is a general property of different soluble forms of A␤42 unrelated to the extent of aggregation (60). The less-or non-neurotoxic A␤ species, including A␤38, A␤40, A␤42 G33A, and A␤43 (5,61,62), although possessing the ability to assemble into cross-␤-fibrils (63), did not influence gene regulation. Thus, our novel findings suggest that there might be a functional activity of A␤42 in the nucleus that is different from AICD (2, 3). Among the several A␤ species that can accumulate in the nucleus, only A␤42 appears to impact the expression of LRP1, KAI1, and APP.
Most significantly, it is the overproduction of A␤ (e.g. mutations associated with early onset AD) that invariably leads to the modulation of A␤ load (6). This regulatory mechanism could induce APP transcription, thereby enhancing APP processing, A␤ production and uptake, and subsequent up-regulation of APP synthesis. Because the detection of intracellular A␤ is always accompanied by increased extracellular A␤ (64), such a proposed mechanism would account for the favored uptake of A␤ from the extracellular pool.
Notably, A␤42, which is toxic in vivo at lower concentrations than applied here (49) for in vitro studies, specifically modulated gene regulation as reported herein. Accordingly, we hypothesize that the neurotoxic A␤42 low n oligomers provoke changes in gene transcription in vivo at concentrations that are subtoxic in vitro. This view is supported by the consideration that soluble low n oligomers of A␤42 peptides are neurotoxic, in particular A␤42 dimers and tetra-/hexamers (5,6,9,65).
In conclusion, our data indicate that A␤42 has a specific transcriptional regulatory function. Our data imply that deregulation of A␤ target genes could be an alternative pathway for A␤-induced neurotoxicity starting with processes such as A␤ accumulation.