PDZ Domain-dependent Suppression of NF-κB/p65-induced Aβ42 Production by a Neuron-specific X11-like Protein

It is widely believed that one of the causes of Alzheimer's disease (AD) is the generation and secretion of β-amyloid (Aβ) from amyloid precursor protein in the brain. Here we report that a transcription factor, NF-κB/p65, induces increased secretion of amyloidogenic Aβ42 but not Aβ40. The κB motif-dependent production of Aβ42 was suppressed by binding of NF-κB/p65 to the PDZ domain of the X11-like protein (X11L), which a human homologue protein of LIN-10. The results suggest that the PDZ domain of X11L can control the ability of NF-κB/p65 to induce expression of protein(s) involved in Aβ42 production. The amino acids 161–163 in Rel homology domain (RHD) of NF-κB/p65 is important in interaction of NF-κB/p65 with X11L. Another subunit NF-κB/p50 and heterodimers of p65 and p50 do not bind to X11L. Our finding indicates NF-κB and X11L may, in novel way, regulate Aβ production in neuronal cells. Targeting X11L by specific therapy may provide the possibility to control the progression of AD.

X11-like (X11L) is a member of a family of proteins homologous to human LIN10 and is thought to function as a scaffold protein in neuron (reviewed in Ref. 1). Human X11L (hX11L) was originally isolated when it was found to be bound to amyloid precursor protein (APP) 1 (2). APP is a precursor protein of ␤-amyloid (A␤), whose deposition and accumulation in the brain are a hallmark of Alzheimer's disease (AD) (reviewed in Refs. [3][4][5]. While the mechanisms regulating the proteolytic cleavage of APP and secretion of A␤ are as yet not well understood, we and others recently reported that X11 (6,7) and X11L (2) proteins regulate APP metabolism and/or A␤ production. The phosphotyrosine interaction domain of X11L interacts with the cytoplasmic domain of APP, and the PDZ (a repeated sequences in the brain-specific protein PSD-95, the Drosophila septate junction protein disks-large, and the epithelial tight junction protein ZO-1) domains of X11L are essential for the modulation of A␤ production (2). However, the exact regulatory mechanism of A␤ production by X11L has yet to be resolved. In this study, we isolated only NF-B/p65 from the multiple isoforms of NF-B from a human brain cDNA library and found that it interacts with the PDZ domain of hX11L. We also found that NF-B/p65 induces the increased secretion of A␤42, but not A␤40, and that this could be suppressed by the expression of X11L due to trapping of NF-B by its binding to X11L. Since NF-B acts as a transcriptional regulator for genes participating in cell death and/or survival (reviewed in Ref. 8), our findings provide new insights regarding the regulation of NF-B activity for neuronal survival and the mechanisms controlling A␤ production in the brains of AD patients.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid System-The protocol for yeast two-hybrid screening using MATCH MAKER Two-Hybrid System (CLONTECH) has been described previously (2). In brief, cDNA encoding PDZ domain (amino acid sequence 556 -749) from hX11L was inserted into pGBT9, and this was used as a "bait" plasmid to isolate cDNA of PDZ-binding proteins from a human adult brain cDNA library cloned into the vector pGAD424, which contains the GAL4 transactivation domain. Both plasmids were co-transfected into yeast cells, and colonies whose HIS3 and LacZ reporter genes were activated were selected. Plasmids from positive clones were re-transfected into Escherichia coli, the nucleotide sequence was determined, and from this NF-B/p65 was identified as a candidate PDZ-binding protein. To identify the PDZ that interacts with NF-B/p65, the pGAD424NF-B/p65 bait plasmid was co-transfected into yeast with pGBT9 "prey" plasmid containing either a region of PDZ1ϩ2 (amino acid positions 556 -749), PDZ1 (amino acid positions 556 -655), PDZ2 (amino acid positions 656 -749), or plasmid alone. Co-transfectants were streaked on the selective medium lacking Trp, Leu, and His or control medium lacking Trp and Leu and clones growing on the selective medium were observed.
Co-immunoprecipitation (Pull-down) Assay-COS7 cells (ϳ1 ϫ 10 7 cells) were co-transfected for 18 h in LipofectAMINE (Life Technologies, Inc.) with 2 g each of the following plasmids: pcDNA3-hX11L and pcDNA3-NF-B/p65 (Fig. 1c), pcDNA3-hX11L and pcDNA3-FLAG-RHDp65 (Fig. 1d) and/or pcDNA3-FLAG-RHDp50 (Fig. 2a), and pcDNA3-hX11L, pcDNA3-p65, and pcDNA3-FLAG-RHDp65 or pcDNA3-FLAG-RHDp50 (Fig. 2b). After changing the medium, the cells were cultured for 48 h. The cells were then harvested, lysed in 1 ml of lysis buffer (phosphate-buffered saline containing 10 mM CHAPS, 5 g/ml chymostatin, 5 g/ml leupeptin, 5 g/ml pepstatin A, 1 mM Na 3 VO 4 , and 1 mM NaF) for 30 min on ice, and centrifuged at 12,000 ϫ g for 10 min at 4°C. The polyclonal antibodies (20 g) or anti-FLAG monoclonal antibody M2 (40 g) were added to the supernatant (which contained approximately 1 mg of protein) and incubated on ice for 1 h. Immunocomplexes binding to the rabbit and mouse antibodies were * This work was supported by a grant-in-aid for Scientific Research from The Japanese Ministry of Education, Science, Sports and Culture (to T. S.) and in part by a grant from Program for Promotion of Basic Research Activities for Innovative Biosciences. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Recipient of Japan Society for the Promotion of Science Research Fellowships for Young Scientists.
Quantification of A␤ with a Sandwich ELISA-HEK293 cells stably expressing human APP695 (9, 10) were transfected as above. The cells were supplied with fresh growth medium 12 h after the start of transfection, and conditioned medium from cells (2 ϫ 10 6 cells) was collected 72 h after the medium change. A␤40 and A␤42 peptides secreted into the medium (100 l) were quantified by sandwich ELISA (9,10). Expression of hX11L and FLAG-p65 was examined by immunoblot analysis of cell lysate with UT-29 and anti-FLAG antibody M2. The relative ratios of the levels of A␤ observed using cells transfected with the pcDNA3 vector alone were determined.
Luciferase Activity Assay-One of the p55IgkLuc, pActLuc, and pIFNLuc reporter construct plasmids (1.25 g) together with any two of the effector plasmids (each 1.25 g), pcDNA3-FLAG-NF-B/p65, pcDNA3-hX11L, and pcDNA3, were triply transfected into either HEK293 cells stably expressing human APP695 or human neuroblastoma SH-SY5Y cells (ϳ5 ϫ 10 5 cells/well in 24-well dishes). The cells were supplied with fresh growth medium 12 h after the start of transfection. After 48 h, the culture medium was removed and cell lysis buffer (100 l) was added. Luciferase activity in the cell lysate was assayed using Luciferase Assay System (Promega) and measured with Luminometer (11). Results were expressed in an arbitrary unit.

RESULTS
Interaction of NF-B/p65 with X11L-Using the yeast twohybrid system, cDNA encoding full-length NF-B/p65 interacting with the PDZ domain of hX11L was isolated. The hX11L protein contains two PDZ domains ( Fig. 1a) (2), and the His reporter gene assay demonstrated that it was the most carboxyl-terminal domain (PDZ2) that was responsible for the interaction with NF-B/p65 (Fig. 1b).
Interaction of hX11L and NF-B/p65 in mammalian cells was demonstrated by co-immunoprecipitation experiments (Fig. 1c). COS7 cells that transiently express hX11L and NF-B/p65 were lysed and the co-precipitation assay performed. The hX11L protein was detected in the sample immunoprecipitated with the anti-NF-B/p65 antibody, sc-109. NF-B/p65 was also detected in the sample immunoprecipitated with two of the anti-hX11L antibodies UT-29 and UT-50, but detection with the third, UT-30, was weak because UT-30 competes with NF-B for the recognition sequence. Neither hX11L nor NF-B/p65 could be detected in the sample immunoprecipitated using non-immune rabbit antibodies, indicating that the coimmunoprecipitation of hX11L and NF-B/p65 was specific.
NF-B/p65 consists of a Rel homology domain (RHD), a nuclear localization signal (NLS), and transcription activator domain (TA) (Fig. 1a) (reviewed in Refs. 14 and 15). To determine whether the RHD participated in the binding of NF-B/p65 to hX11L, we performed another co-immunoprecipitation experiment using COS7 cells co-expressing hX11L and the RHD tagged with a FLAG sequence at their amino-terminal ends (Fig. 1d). The expression of the FLAG constructs of RHD and hX11L could be detected in the crude cell lysate (data not shown). When the lysate from cells expressing FLAG-RHD was subjected to immunoprecipitation with anti-FLAG antibody M2, the hX11L protein could also be detected (Fig. 1d, wt), indicating that some sequence in RHD can interact with PDZ2 of hX11L. It is known that PDZ generally binds to S/T-X-I/V (class I), F/Y-X-A/F/V (class II), and D-X-V or Y-X-A/F motifs (class III) within target proteins (16,17); and based on this, four potential PDZ-binding sequences, 75-SLV-77, 141-FQV-143, 161-FQV-163, and 217-DKV-219, were found in the NF-B/p65 RHD (Fig. 1d). To determine which of these sites was important in PDZ binding, an alanine mutation was introduced in each candidate motif. When the 161-FQV-163 motif was altered to AQA, the anti-FLAG antibody failed to pull down hX11L (Fig. 1d, 161-FQV-1633 AQA). The other motif mutants pull down hX11L as well as the non-mutated FLAG-RHD (wt). Thus, PDZ2 of hX11L recognizes and binds to the 161-FQV-163 motif in the RHD of NF-B/p65.
Since NF-B can function as homo-and heterodimers of p65 and/or p50 (reviewed in Ref. 18), we assessed whether the p50 subunit is as able to bind to the PDZ domain of X11L as the p65 subunit. For this purpose, FLAG constructs of RHD from NF-B/p50 (FLAG-RHD(p50)) and hX11L were expressed transiently in COS7 cells. When FLAG-RHD(p50) was isolated from cell lysate by immunoprecipitation with anti-FLAG antibody, hX11L could not be detected in the precipitate (Fig. 2a). Thus, X11L associates with the p65 but not the p50 subunit of NF-B. Consistent with this finding is the observation that the p50 subunit does not contain the PDZ binding motif found in p65 (data not shown).
To determine whether the heterodimer of NF-B/p65 and p50 can associate with X11L, COS7 cells were transiently and triply transfected with pcDNA3-hX11L and pcDNA3-NF-B/ p65 together with either pcDNA3FLAG-RHD(p50) or pcDNA3FLAG-RHD(p65). The protein lysates were immunoprecipitated with anti-FLAG antibody M2 and immunoprecipitates were analyzed by immunoblot with the anti-X11L antibody UT-29, the anti-p65 antibody sc-109 and M2 (Fig. 2b). When cells expressed FLAG-RHD(p65), both NF-B/p65 and X11L could also be detected in the precipitate. However, when cells were expressed with FLAG-RHD(p50), NF-B/p65, but not X11L, could be detected in the precipitate. Thus, NF-B/ p65 associates with FLAG-RHD(p50), but this heterodimer cannot bind to X11L. Although we did not determine whether the NF-B/p65 homodimer is able to associate with X11L, the present results suggest that only monomeric NF-B/p65, not homo-and heterodimers, can interact with X11L.
Induction of A␤ Production by NF-B/p65 and Suppression of the A␤ Production by X11L-It is known that X11L is a potential modulator of A␤ production (2) and that NF-B acts as a transcriptional activator of many genes (reviewed in Refs. 14 and 15). Therefore, we examined the effect of NF-B/p65 on the generation of both A␤40 and A␤42. HEK293 cells that stably expressed human APP695 were transiently co-transfected with any two combinations of cDNA encoding hX11L, NF-B/p65 with a FLAG tag (FLAG-p65) and vector alone. At 72 h, a sandwich ELISA was used to quantify A␤40 and A␤42 secreted into the medium. That hX11L and FLAG-p65 were expressed was verified by immunoblot analysis using UT-29 and M2 (data not shown). The negative control consisted of the A␤ released by cells containing pcDNA3 together with pcDNA3. This amount was set at 1.0 and levels in test samples related to this (Fig. 3). NF-B/p65 alone increased the level of A␤42 but not A␤40. As expected and reported previously, hX11L did not alter the level of A␤42 but slightly decreased the level of A␤40 (2). Surprisingly, however, when hX11L was co-transfected with NF-B/p65, the NF-B/p65-induced secretion of A␤42 was suppressed. The secretion level of A␤40 was not, however, altered in this situation. Thus, NF-B/p65 may activate the expression of one or more genes participating in A␤42 production and hX11L is able to suppress this. Whether hX11L can suppress B motif-dependent gene activation was examined with a reporter gene assay using nonneuronal (HEK293) and neuronal (SH-SY5Y) cells (Fig. 4). HEK293 cells stably express human APP695 and SH-SY5Y cells express endogenous APP. These cells were transfected with either the positive reporter plasmid p55IgLuc, which consists of three tandemly repeated B motifs upstream of a minimal interferon-␤ promoter (Ϫ55 to ϩ19) and the luciferase structural gene (Fig. 4a) (11), or the B-independent reporter plasmids pActLuc and pIFNLuc as negative controls. The reporter plasmid was co-transfected transiently with NF-B/p65 and hX11L genes (Fig. 4, b and c). When FLAG-p65 was present, transcription of the reporter gene p55IgkLuc was activated. When pActLuc and pIFNLuc, reporter genes containing defective B motifs, were used instead of p55IgLuc, no effect on gene activation by NF-B was observed, indicating that the activation of the reporter gene depends on the B motif. The B motif-dependent activation of genes was also supported with an electrophoretic mobility-shift assay (Fig. 5). When p65 was present, specific complex composed of p65 and a probe containing B motif was observed, which definitely demonstrated NF- B/p65 in the nucleus.
Transfection with hX11L alone did not activate any of the reporter genes used (data not shown). However, when hX11L was co-transfected with FLAG-p65, activation of the reporter gene p55IgLuc was partly suppressed. Identical results were obtained with non-neuronal HEK293 (Fig. 4b) and neuronal SH-SY5Y (Fig. 4c) cells, which do not express the endogenous X11L. These results show that NF-B may increase the production of A␤42 via B motif-dependent transcriptional activation of certain gene(s) and further suggest that hX11L may inhibit this process by binding NF-B to its PDZ domain, thereby retaining NF-B in the cytoplasm and impeding its translocation into the nucleus. Further supporting the notion that A␤42 production depends on genes activation by NF-B is the finding that the mutant NF-B containing an ALA substitution in the 75-SLV-77 sequence (whose binding to X11L is not impaired (Fig. 1d)) failed to activate B motif-dependent genes and did not enhance A␤42 production (Fig. 6). This mutant NF-B has lost an ability to bind to B motif (Fig. 5). DISCUSSION In our previous study (2), X11L was found to bind to the cytoplasmic domain of APP and regulate A␤ production. It has also been demonstrated by others that X11 modulates APP metabolism (6,7). From these observations, we hypothesized that X11L may regulate A␤ production by complexing with APP and that the PDZ domains in X11L may play an important role in this process (2). It was, however, not clear whether X11L regulates APP processing by directly interacting with APP or via other, as yet unidentified, proteins. To this aim, we used yeast two-hybrid screening to identify proteins able to bind to the PDZ of X11L and thus identified NF-B/p65 as a PDZ-binder.
NF-B/p65 could induce the production of A␤42 through B motif-dependent gene activation. The 5Ј-regulatory region of the APP gene contains two NF-B binding sequences (18) and NF-B might thus be expected to increase the expression of APP and concomitantly result in increased levels of A␤42. However, this is not the case because when HEK293 cells stably expressing APP695 under the control of cytomegalovirus enhancer-promoter are transiently transfected with NF-B/ p65, the production of A␤40 is not changed, while the production of A␤42 is increased. Furthermore, overexpression of NF-B/p65 did not alter the synthesis and maturation of APP695 (data not shown), and the B sites of APP gene interact with NF-B/p50 but not NF-B/p65 (18). The observations suggest the significance that NF-B signaling system connects to A␤42 generation by activating expression of gene(s), which encodes protein concerning with a ␥-secretase pathway, and also support the concept of two independent ␥-secretases.
Expression of the neuron-specific protein X11L suppressed the transcriptional activity by NF-B/p65 and inhibited A␤42 production. This suggests that X11L binds to NF-B/p65 with its PDZ domain, thus suppressing the capacity of NF-B to induce expression of protein(s) involved in A␤42 production in the neuron.
We were not able to detect a triple complex consisting of APP, X11L, and NF-B/p65 by co-immunoprecipitation with their respective antibodies (data not shown), indicating that NF-B/p65 binds to X11L in the absence of APP. In addition, a mutant NF-B/p65 with alanine substitutions in the 75-SLV-77 sequence failed to induce both A␤42 production and B motif-dependent gene activation despite its ability to associate with X11L. Thus, it is unlikely that X11L acts to suppress NF-B/p65-mediated activation of A␤42 production by forming a triple complex with APP and NF-B/p65 and also unlikely that NF-B/p65 regulates APP metabolism directly in the absence of transcriptional activation. There may be another regulatory mechanism, independent of gene activation, that suppresses A␤40 production through a triple complex consisting of APP, X11L, and an unidentified PDZ-binding protein (2).
It is well known that target genes of NF-B participate in the immune system (reviewed in Refs. 14 and 15), and increasing evidence indicates that a wide variety of members of the NF-B family play a role in the neuronal system (reviewed in Ref. 19). Since inflammation is though to be a cause of AD, it may be that brain inflammation activates NF-B and thus induces A␤42 production. Although the activity of NF-B/p65 is generally regulated by its association with IB (reviewed in Refs. 14 and 15), a constitutive, uncomplexed form of NF-B has been observed in cortical and hippocampal neurons (20,21), which are the cells that undergo the characteristic degeneration in AD. Furthermore, A␤ can activate NF-B in neurons (22,23), and activated NF-B is found only in cells in close vicinity to early senile plaques of AD patients (24). In addition, NF-B contributes to neuronal cell death (8). These observations together with our present results argue strongly for the partici-pation of NF-B in AD pathogenesis. Given the inhibitory role of X11L on NF-B-induced A␤42 production, this neuron-specific protein (2) could be used as a specific drug target for the treatment of AD rather than NF-B, which is expressed ubiquitously.
NF-B can form homo-or heterodimers consisting of p65 and/or p50 subunits and binds to DNA (reviewed in Refs. 14 and 15). We showed that the p50 subunit and a heterodimer consisting of p65 and p50 could not bind to the PDZ domain of X11L. X11L may thus be an anchor protein specific for monomeric NF-B/p65, although it is remains to be seen if X11L can bind p65 homodimers.
In many protein-protein interactions involving PDZ domain recognition, the PDZ domain associates with carboxyl termini of proteins such as those occurring in receptors or channels (25,26). However, a recent report has revealed that PDZ domains can also bind to non-terminal consensus motifs such as those of neuronal nitric-oxide synthase (nNOS) (15,16). The PDZ domains of X11L also recognized to non-terminal motif in NF-B/ p65. The interaction between NF-B/p65 and X11L indicates a novel function for the PDZ domain in that it is involved in regulation of protein metabolism via transcriptional activation. Since NF-B is thought to regulate cell death or survival through activation of several types of genes, analysis of the molecular mechanism(s) of NF-B activation in neuronal cell is important in the understanding of neurologic disease such as AD.
FIG. 6. NF-B-dependent gene activation and A␤42 production. HEK293 cells stably expressing human APP695 were transfected with either pcDNA3-NF-B/p65 (p65wt) encoding NF-B/p65, pcDNA3-NF-B/p65mut (p65mut) encoding p65 carrying the ALA substitution of the 75-SLV-77 sequence or pcDNA3 (vector), together with either the positive p55IgkLuc or the negative pActLuc reporter gene. a, luciferase activity was assayed and shown is the average of four independent assays (n ϭ 4). b, A␤42 in the medium was measured (n ϭ 5). The quantity of A␤42 is indicated as the ratio of the levels of A␤42 observed relative to that produced by cells transfected with a vector only (a reference value of 1.0). The error bar indicates S.D. (**, p ϭ 0.0011; ***, p Ͻ 0.001).