A Dominant Role for FE65 (APBB1) in Nuclear Signaling*

FE65 has been described as an adaptor protein; its partners include the β-amyloid precursor protein (APP) and Tip60 (a histone acetyltransferase). Recent evidence suggests that APP may function in a nuclear signaling pathway via formation of APP-FE65-Tip60 complexes. The evidence is largely based on experiments in which APP/Tip60 is fused to the DNA binding domain of a yeast transcriptional factor Gal4 (Gal4DB) that can activate a reporter gene only when FE65 is coexpressed. One interpretation of published experiments has not yet been tested; however, there is the possibility that FE65 itself is the dominant transcriptional activator, whereas APP and Tip60 serve merely as positive/negative modulators or bridges for connecting FE65 to Gal4DB. To test this possibility, we fused Gal4DB directly to either end of FE65 and assessed their nuclear signaling in the presence or absence of exogenous APP/Tip60 or after knockdown of endogenous APP/Tip60. We found that FE65-Gal4DB by itself was able to trigger robust reporter activities. Increasing levels of APP could not further augment the reporter activity, but knocking down endogenous APP or interrupting FE65-APP binding reduced the signaling by up to 2-fold. The magnitudes of the reporter activities did not correlate with relative FE65 affinities for APP. Both overexpression and knockdown experiments showed that Tip60 plays a negative role. The results are consistent with the notion that FE65 is the key agent of Gal4DB-mediated transcriptional transactivation, whereas Tip60 is an FE65-associated repressor. Although APP may have modest stimulating effects, apparently there is no absolute requirement for that molecule for the nuclear signaling pathway.

FE65 is a brain-enriched protein and has the ability to interact with several different proteins via its three protein-protein interaction domains as follows: a WW domain and two phosphotyrosine interaction domains (PID 2 domains) (1,2). PID1 interacts with Tip60 (Tatinteractive protein, 60 kDa); PID2 is the main region bound by the APP intracellular domain (AICD). These interactions may have impacts on APP processing (1,3,4), membrane dynamics (including axonal projections and neuronal positioning) (5-7), learning and memory (8), and transcriptional transactivation (9).
APP is the best studied protein in the field of Alzheimer research because causal relationships have been established with mutations in APP and presenilins, enzymes involved in APP processing (10 -12). Proteolytic fragments of APP are also key components of Alzheimer pathology (13). Despite extensive research efforts, however, APP functions are poorly understood. The protein undergoes two consecutive cleavages at sites near or within its transmembrane domain (12). The first cleavages (by ␣-/␤-secretases) shed APP ectodomains into extracellular environments, perhaps leading to modulations of cell proliferation and adhesion, neurite outgrowth, and synaptogenesis (6,14,15). A subsequent cleavage (by ␥-secretases complexes) liberates P3 or the ␤-amyloid peptides, as well as AICD fragments. The potential importance of AICD peptides is suggested by similarities with the metabolism of other type I transmembrane proteins, the prototype being the Notch pathway of regulated intramembranous proteolysis (14, 16 -18). Notch and APP undergo similar cleavages by similar secretases. Regulated intramembranous proteolysis of the Notch protein releases intracellular domains that associate with specific cellular factors and translocate into the nucleus, where they alter gene transcription (19,20). The intriguing hypothesis that AICD may also function in nuclear signaling was first tested by Cao and Sudhof (9) using fusion proteins of APP/AICD with Gal4DB (APP/AICD-Gal4DB) or the DNA binding domain of a bacterial transcription factor LexA (APP/ AICD-LexADB). They demonstrated that APP/AICD forms a multimeric complex with FE65 and Tip60 that stimulates transcription via the heterologous Gal4DB or LexADB. Although more recent evidence indicates that liberated AICD is dispensable for the signaling mechanism (21), the potential nuclear function mediated by APP/AICD has been reproduced by several other research groups with two major versions of Gal4DB fusion proteins, APP/AICD-Gal4DB and Gal4DB-Tip60 (21-26). Regardless of which version of the fusion proteins is used, cotransfection of FE65 is inevitably required for the transactivation. However, an important hypothesis that has not yet been tested is whether FE65 itself may be the "real" transactivation factor, whereas APP/AICD or Tip60 in Gal4DB fusion proteins merely serve as bridging peptides for connection of FE65 to Gal4DB. They may, perhaps, even negatively regulate FE65 nuclear functions. To test that hypothesis, we developed a similar Gal4DB/luciferase reporter system in which Gal4DB was directly fused to either end of the full-length FE65. We examined the effects of APP and Tip60 on the FE65-dependent nuclear signaling. Our results show that FE65-Gal4DB by itself can robustly activate the reporter gene. Endogenous APP only provided modest enhancement of FE65 nuclear signaling; Tip60 was shown to be a negative factor for this pathway. Our results should be helpful in the design of experiments for the elucidation of downstream genes under the control of this class of nuclear regulation.
Cell Cultures and Transfections-A transformed human embryonic kidney cell line, HEK293E, and brain neuroblastoma cell lines, B103 (rat) and SK-N-SH (human), were cultured as described previously (4). All transfections were done in 12-well plates unless otherwise indicated. For plasmid transfection in the absence of siRNAs, 3 ϫ 10 5 cells/well were transfected with equal amounts of pFRluc (Stratagene, La Jolla, CA), pcDNA3.1-lacZ, and one of the plasmids containing FE65 and Gal4DB fusion proteins; the experiments were performed with or without inclusion of pcDNA3.1-app695 or pcDNA3.1-HA-tip60. Total amount of plasmid DNA was always made up to 1.6 g/well with the pcDNA3.1 vector. Transfection was mediated by 9 l of 1 g/l polyethyleneimines (Polysciences, Inc., Warrington, PA) (4). Plasmid transfection in the presence of siRNAs was mediated by mixing 3 l of Lipofactamine2000 TM (Invitrogen), 50 pmol of double-stranded siRNA, and 1 g of total plasmid DNA per well. Cells were usually harvested 48 h after transfection. Luciferase and ␤-galactosidase activities were determined using a luciferase reporter assay kit (BD Biosciences) and a ␤-galactosidase enzyme assay system (Promega, Madison, WI), respectively, according to the manufacturers' instructions. All samples were examined in quadruplicate, and all results were derived from at least two independent experiments.
GST-AICD Binding Assay and Immunocytochemistry-Expression and purification of GST-C48 (a fusion protein between glutathione S-transferase and the 48 C-terminal amino acids of human APP) and the pull-down conditions were described in a previous study (4). Typically, 20 l of fusion protein-immobilized beads and 40 l of FE65-transfected HEK293E cell lysates were used for the assays. Cell lysates and pulled down proteins were analyzed by Western blotting with an FE65-specific antibody, FE518 (4). Experimental conditions for immunocytochemistry were the same as described previously (4).

RESULTS
p97FE65-Gal4DB or Gal4DB-p97FE65 Alone Can Turn on Transcription of the Luciferase Reporter Gene-Our first question was whether the 97-kDa full-length FE65 (p97FE65) alone, when fused with Gal4DB, could turn on the reporter gene. We fused Gal4DB directly to either the N terminus (Gal4DB-p97FE65) or the C terminus (p97FE65-Gal4DB) of the neuronal form of p97FE65 (30), and we assessed their nuclear signaling in several cell lines. As expected, when Gal4DB-p97FE65 or p97FE65-Gal4DB was coexpressed with the luciferase reporter gene in HEK293 cells, transcription of the reporter gene could be turned on by either fusion protein (Fig. 1B). Levels of the transactivation signals triggered by p97FE65-Gal4DB were ϳ2.5-fold stronger than those by Gal4DB-p97FE65. Western analysis with an FE65-specific antibody (FE518) showed that there was less expression of the fulllength Gal4DB-p97FE65 than of the full-length p97FE65-Gal4DB (Fig.  1B). For unknown reasons, the majority of Gal4DB-p97FE65 molecules was partially degraded, perhaps resulting in impaired transactivation. The diminished activity might also be attributable to proteolytic cleavage of the N terminus, resulting in an N-terminally truncated C-terminal fragment, p65FE65 (4). Such processing would delete the Gal4DB from Gal4DB-p97FE65 as well as the N-terminal region of p97FE65 (Fig.  1, A and B). Although p97FE65-Gal4DB could undergo a similar cleavage, the resulting p65FE65-Gal4DB would still be able to trigger the reporter gene (Fig. 1B). To compare the activities of various isoforms of FE65, we fused Gal4DB to the C termini of the non-neuronal form of p97FE65 (lacking two amino acids, Glu-Arg, in the PID1 domain) (30) and of p97FE65a2, an alternatively spliced isoform containing an altered C terminus with reduced APP affinity ( Fig. 1A) (31). The non-neuronal form of p97FE65-Gal4DB showed levels of transactivational activity comparable with that of the neuronal form of p97FE65-Gal4DB (data not shown). Transactivation mediated by p97FE65a2-Gal4DB, however, was only ϳ10% of that mediated by p97FE65-Gal4DB (Fig. 1B). This may reflect the weak binding of p97FE65a2 for APP and/or may be attributable to other functional consequences of their distinctive C-terminal sequences. Similar results were also observed in other cell lines, including SK-N-SH cells (a human brain neuroblastoma cell line) and B103 (a rat brain neuroblastoma cell line). In sharp contrast, no reporter signals could be detected in COS cells (data not shown), indicating that the transactivation by p97FE65-Gal4DB is cell type-dependent. Taken together, these results show that p97FE65-Gal4DB alone, without coexpressions with APP and/or Tip60, is sufficient to trigger transactivation in appropriate cell types.
Overexpressions of Exogenous APP/AICD or Tip60 Do Not Stimulate FE65-dependent Transcriptional Activation-It has been proposed that a complex of APP, FE65, and Tip60 is required for transactivation (9, 21). Therefore, we examined the effects of coexpressions of APP and/or Tip60 on transcriptional activation by p97FE65-Gal4DB, Gal4DB-p97FE65, or p97FE65a2-Gal4DB in HEK293 cells. Most surprisingly, overexpression of exogenous APP did not further enhance the transactivational activity mediated by any of the FE65 and Gal4DB fusion proteins ( Fig. 2A). Signaling was in fact slightly reduced (ϳ10%). Cotransfections of p97FE65-Gal4DB with increasing amounts of pCDNA3.1-app695 (0 -400 ng) altered reporter activities by Ϯ10 -20% when compared with those transfected by p97FE65-Gal4DB alone ( Fig.  3A; data from HEK293E cells were not shown). Overexpression of APP, however, might tether p97FE65-Gal4DB to cellular membranes, thus preventing the fusion proteins from functioning within the nucleus until the APP C terminus is released by ␥-secretases. We therefore also coexpressed p97FE65-Gal4DB with AICD (C59 or C44). The results were similar to those of experiments utilizing coexpressions with APP (data not shown). These results suggest that APP/AICD may not be required for the FE65-dependent pathway of transcriptional activation.
Endogenous APP Provides Only Modest Stimulation of FE65-dependent Transactivation-Although exogenous APP was not required for the FE65-dependent transactivation, perhaps endogenous levels of APP might be sufficient for maintaining such functions. We therefore tested whether endogenous APP was required for FE65-dependent transactivation. Our initial experiments were with rat neuroblastoma B103 cells, a cell line that expresses very little endogenous APP (32). We found that expression of p97FE65-Gal4DB alone in B103 cells could also turn on the reporter gene; coexpressions of increasing amounts of exogenous APP did not augment the reporter signals (Fig. 3A). We then knocked down endogenous levels of APP in HEK293 cells by siRNA. In order to eliminate potential off-target effects, the following two independent targeting sequences were employed: app-siRNA 1059 targeted a sequence in exon 10 (29), and app-siRNA 2404 targeted a sequence within the 3Ј-untranslated region. Both sequences are present in all app splicing variants, and neither sequence has homology to other genes, including the two other members of the APP family, aplp1 and aplp2. A scrambled app-siRNA 1059 sequence, with no homology with other genes in the NCBI Data Base, was used as a control. Western analysis showed that both app-specific siRNAs were equally effective; they knocked down endogenous levels of APP by 80% when compared with the scrambled control (Fig. 3B, bottom panel). Coexpressions of p97FE65-Gal4DB with app-siRNA 1059 or app-siRNA 2404 reduced p97FE65-Gal4DB-dependent signaling by 50 or 15%, respectively, when compared with the scrambled control (Fig. 3B, top panel). These results indicate that modulation of FE65-dependent nuclear signaling by endogenous APP is modest. To assess further whether APP was required for FE65-dependent nuclear signaling, a mutation (C654F) was introduced into p97FE65-Gal4DB, one that had been shown to disrupt the APP-FE65 interaction (21). Affinities for the APP C terminus (GST-C48) were compared by GST pull-down assays among cell lysates expressing p97FE65-Gal4DB, p97FE65a2-Gal4DB, and p97FE65 C654F -Gal4DB. The results confirmed that the p97FE65 C654F -Gal4DB mutant failed to bind GST-C48 (binding was below the detectable level), whereas large amounts of p97FE65-Gal4DB were retained on the GST-C48 beads (Fig.  4B). We also confirmed that p97FE65a2-Gal4DB had very low binding affinity for GST-C48 when compared with the p97FE65-Gal4DB (Fig.  4B). To assess FE65-APP binding in vivo, we also performed immunostaining (Fig. 3D). The results showed that when either p97FE65-Gal4DB or p97FE65 C654F -Gal4DB was expressed alone in HEK293 cells, localization was predominantly nuclear. Coexpressions with APP tethered p97FE65-Gal4DB in the cytoplasm, colocalizing with APP in perinuclear structure (Fig. 3D, top panel). In contrast, APP could not tether p97FE65 C654F -Gal4DB in the cytoplasm (Fig. 3D, middle panel), suggesting a disruption of the interaction between the two proteins. The results of the in vivo immunostaining and in vitro pull-down experiments therefore documented the anticipated functional consequence of the C654F point mutation. Additional luciferase assays revealed that, although the p97FE65 C654F -Gal4DB mutant construct was unable to interact with APP, it still could activate expression of the reporter gene, although at levels ϳ50% of that mediated by the wild type 97FE65-Gal4DB construct (Fig. 3C). Thus, interruption of the interaction of FE65 with endogenous APP only modestly influenced FE65-dependent transactivation; therefore, that interaction may not be essential for the process. Presumably, the transactivation by p97FE65 C654F -Gal4DB is also independent of endogenous APLP1 and APLP2, because FE65 interacts with the APP family members at a common YENPTY motif located in their cytoplasmic domains (33,34). Another interesting finding was that although p97FE65a2-Gal4DB bound to the APP C terminus better than p97FE65 C654F -Gal4DB (Fig. 4B), the transactivation ability of p97FE65a2-Gal4DB was much lower than that of p97FE65 C654F -Gal4DB (Fig. 3C). The results suggest that binding affinity for APP does not correlate with the ability of transactivation among different FE65 isoforms.
The C Terminus of FE65 Negatively Regulates Its Own Transactivation-As the affinity for APP did not correlate with transactivation activity, additional factors might also contribute to the relatively low transactivation by p97FE65a2-Gal4DB (Fig. 3C). p97FE65a2 differs from p97FE65 by 55 amino acid residues at its C-terminal region (residues 656 -710) (Fig. 1A). A deletion mutant that lacked the 55 C-terminal residues of p97FE65, p97FE65⌬C-Gal4DB, was generated. Transactivation assays showed that the deletion mutant was a robust transactivator when compared with its parental molecule, either p97FE65-Gal4DB or p97FE65a2-Gal4DB, suggesting that the FE65 C-terminal residues negatively regulate its transactivation activity (Fig.  4A). In vitro GST-C48 binding assays were carried out to assess binding of p97FE65⌬C-Gal4DB to GST-C48. Most surprisingly, the results showed that FE65⌬C-Gal4DB was still able to bind to the APP C terminus, although with reduced affinity when compared with p97FE65-Gal4DB (Fig. 4B). Once again, the results indicate that binding affinity for APP does not correlate with FE65-dependent transactivation activity. Immunostaining of the HEK293 cells expressing p97FE65⌬C-Gal4DB alone or also coexpressing with APP revealed that deletion of the C terminus of FE65 altered its cellular localization from mainly in the nucleus to mainly in the cytoplasm (Fig. 3D, bottom panel).
Tip60 Negatively Modulates FE65-dependent Transactivation-As shown earlier (Fig. 2, A and B), overexpression of exogenous Tip60 suppressed FE65-dependent transactivation, which is contradictory to the previous observations by Cao and Sudhof (9,21). To study the effect of endogenous Tip60 on FE65-dependent transactivation, we knocked down levels of endogenous Tip60 with two previously published siRNA sequences, tip60-siRNA 685 and tip60-siRNA 407 (28). The knockdown efficacy was evaluated by cotransfections of tip60-siRNA 685 or tip60-siRNA 407 with pcDNA3.1-HA-tip60, given the lack of reliable antibodies for detection of endogenous levels of Tip60. Western analysis showed that levels of the cotransfected HA-Tip60 were knocked down by 50%  1-gal4DB, pcDNA3.1-p97fe65-gal4DB, or pcDNA3.1-p97fe65⌬C-gal4DB. Reporter assays were performed as described in Fig. 1. Luc, luciferase; ␤-gal, ␤-galactosidase. B, in vitro binding assays for examining the interaction between the APP C terminus (GST-C48) and various FE65-Gal4DB fusion proteins in the samples of the top panel and Fig. 3C. HEK293E cell lysates (50 l) overexpressing indicated proteins were incubated with GST-C48 beads. Proteins in cell lysates and GST-C48 pull down were analyzed by Western blotting with antibody FE518. FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 by tip60-siRNA 685 and by 90% by tip60-siRNA 407 , when compared with the scrambled control (Fig. 5B). Transactivation assays revealed an inverse correlation between levels of Tip60 and magnitudes of p97FE65-Gal4DB-dependent transactivation (Fig. 5A). About 20 and 80% increased transactivations were observed in the presence of tip60-siRNA 685 and tip60-siRNA 407 , respectively, when compared with that from the scrambled control. Consistent with those from Tip60 overexpression (Fig. 2), the results suggest that Tip60 negatively regulates FE65dependent transactivation.

Nuclear Signaling by FE65
Acetyltransferase Activity and Interaction with FE65 Are Involved in the Negative Regulation of p97FE65-Gal4DB Transactivation by Tip60-To investigate the molecular mechanism by which Tip60 negatively regulated FE65-dependent transactivation, two Tip60 mutants were generated. Mutant Tip60HAT Ϫ was generated by changing Gly 463 to Asp 463 . The residue is located in the acetyl-CoA-binding site; the point mutation would destroy acetyltransferase activity (27). Mutant Tip60 NASA was generated by altering the FE65-binding site 309 NKSY 312 to 309 NASA 312 , which has been shown to fail to bind FE65 in vitro (9). p97FE65-Gal4DB was then cotransfected with increasing amounts of wild type (WT) Tip60 or mutant Tip60. pcDNA3.1-Gal4DB was used as a control to show that increasing amounts of WT Tip60 did not affect basal levels of transactivation by Gal4DB (Fig. 5C). Luciferase assays showed that increasing levels of WT Tip60 dramatically reduced transactivation by p97FE65-Gal4DB, and increasing levels of Tip60HAT Ϫ mutant led to reduced transactivation to a lesser extent. In contrast, Tip60 NASA exhibited the least inhibitory effects on p97FE65-Gal4DBdependent transactivation when compared with those of WT Tip60 and Tip60 HAT Ϫ (Fig. 5C). These results suggest that both acetyltransferase activity and interaction with FE65 are involved in the repression of p97FE65-Gal4DB signaling.

DISCUSSION
FE65 has been recognized as a key component of a transcriptional activating complex consisting of FE65, APP, and Tip60 (9,21,35). Surprisingly little consideration has been given to the possibility that FE65 (also known as APBB1, amyloid ␤-precursor protein-binding, family B, member 1, and usually described as "an adaptor protein") is in fact the dominant transactivating protein. This study provides strong evidence that this is in fact the case. APP is found to have only weak modulating effects, whereas Tip60 is shown to be a strong suppressor of FE65mediated transcription. These findings have important implications for future research on such basic neurobiological issues such as synaptic remodeling and the pathogenesis of dementias of the Alzheimer type. FE65 is part of a protein network that includes several mutant or polymorphic proteins with proven or suspected roles in the modulation of susceptibility to that highly prevalent disorder (36).
Cao and Sudhof (21) and Telese et al. (35) fuse Gal4DB to the N terminus of the full-length FE65, leading to either no detectable reporter activity in COS cells or weak activity in HeLa cells. A strong transcriptional transactivation is, however, achieved after extensive deletions of the N and/or C termini of FE65 (21, 35). Our results show that Gal4DB-p97FE65 is less stable than p97FE65-Gal4DB, as the full-length fusion protein can be partially degraded to a slightly smaller fragment and/or cleaved in a region before the WW domain. Such processing may interfere with FE65-mediated transactivation when Gal4DB is fused to the N terminus of FE65. We have also demonstrated that the use of different cell lines may lead to variable outcomes. When p97FE65-Gal4DB was tested in COS cells, we could not detect any signal, not even when cotransfected with APP and/or Tip60 (data not shown). COS cells (a heteroploid cell line derived from the kidneys of African green mon-  1-Gal4DB or pcDNA3.1-97FE65-Gal4DB, with or without 50 pmol of scramble siRNA (scramble), tip60-siRNA 685 (685), or tip60-siRNA 407 (407). Reporter assays were performed as described in Fig. 1. Luc, luciferase; ␤-gal, ␤-galactosidase. B, knockdown efficacy of cotransfected pcDNA3.1-HA-tip60 was evaluated by Western analysis with a monoclonal antibody specific for the HA tag. Levels of ␤-actin were used as loading controls. C, top panel, acetyltransferase activity and interaction with FE65 are involved in the negative regulation of p97FE65-Gal4DB-mediated transactivation by Tip60. HEK293E cells were cotransfected with pFRluc, pcDNA3.1-lacZ, and pcDNA3.1-97FE65-Gal4DB and 0 -600 ng/well of pcDNA3.1-HA-tip60, pcDNA3.1-HA-tip60HAT Ϫ , or pcDNA3.1-HA-tip60 NASA . pcDNA3.1-Gal4DB was used as a control to show that increasing amounts of wild type Tip60 had no effect on basal levels of transactivation. Total amounts of plasmid DNA at 1.6 g/well were made up with pcDNA3.1. Reporter assays were performed as described in Fig. 1 (n ϭ 4). Bottom panel, expressions of wild type Tip60 (WT), Tip60HAT Ϫ (HAT Ϫ ), and Tip60 NASA (NASA) in representative samples of the top panel were evaluated by Western analysis with an antibody against the HA tag. keys) may either lack cellular components required for that function or express inhibitory factors that could block the pathway.
As the full-length FE65 fused with Gal4DB alone could trigger robust reporter activities, we were able to assess the influences of APP and Tip60 on these assays. Our results show that endogenous APP may modestly stimulate the FE65-dependent transactivation but that APP or AICD appears not to be required under our experimental conditions. First, we found that FE65-Gal4DB signaling is not sensitive to a broad range of APP concentration ( Fig. 2A and Fig. 3A). Second, a reduction of the levels of endogenous APP by 80% (Fig. 3B) had no more than a 2-fold effect upon reporter activity. Third, we found that the magnitudes of the reporter activities did not correlate with relative binding affinities of FE65 for APP. For instance, mutant p97FE65 C654F -Gal4DB, which does not bind APP in vitro and in vivo, can still trigger the reporter gene. On the other hand, mutant p97FE65⌬C-Gal4DB, which has a lower affinity for APP, could trigger more than 10-fold higher reporter activities than its wild type control (Fig. 4).
The strength of the reporter signal appears to depend upon the conformation of FE65 itself rather than upon the FE65-APP interaction. Altered conformations may allow FE65 to access or withdraw from a set of molecules that are required for the transcriptional transactivation. Previous evidence has shown that the N-terminal WW domain and its flanking sequences are required for transactivational activity (9,21,35). In this study, we have also found that the 55 C-terminal residues of FE65 negatively regulate its nuclear signaling. Deletion of this region resulted in translocation of the majority of the protein from the nucleus to the cytosol (Fig. 3D, bottom panel), leaving only a small portion of p97FE65⌬C-Gal4DB available in the nucleus (confirmed by z-series images of DeltaVision deconvolution microscopy (data not shown)). This small portion is apparently sufficient, however, to enhance the reporter-gene expression when compared with the wild type control (Fig. 4A). The C terminus of p97FE65a2, a minor allelic isoform variant found only in a small portion of human populations (31), has particularly strong negative effects upon transcriptional activation as compared with its counterpart region in p97FE65, the product of the major human FE65 allele. It is possible that the FE65 nuclear function in vivo is regulated by cellular factors interacting with its C-terminal region. Our results, however, do not implicate APP as a particularly important candidate in such interactions.
More importantly, our results consistently show a negative regulation of the signaling pathway by Tip60. Overexpression of Tip60 strongly represses the reporter activity induced by both p97FE65-Gal4DB ( Fig.  2A) and APP-Gal4DB (in the presence of coexpressed FE65) (Fig. 2B). On the other hand, knockdown of endogenous Tip60 by siRNA enhances the activity (Fig. 5A). Similar repressive effects on AICD-Gal4DBϩFE65 signaling have been observed previously. The results were interpreted as the repression might be because of inefficient recruiting of untagged Tip60 to the site of the promoter; the free Tip60 would bind to other components of its multimolecular complex, thereby leading to inhibitory transcription at the reporter gene (25). However, our results show that the repression is, in part, mediated via direct binding to FE65, suggesting that the untagged-Tip60 is able to form a complex with FE65-Gal4DB at the promoter region. In addition, knockdown of endogenous Tip60 enhances the reporter activity, reinforcing the notion that Tip60 is a repressor. Tip60 is a conserved protein and possesses an intrinsic acetyltransferase activity. It can acetylate histones and possibly other cellular proteins (37). Proteins containing such activities are often involved in chromatin remodeling and gene regulation. It has been shown that Tip60 can associate with either transcriptional activators or repressors. It thus may either stimulate or repress gene expression, depending upon the particular gene or cell type (27,38,39). Although the precise mechanism by which Tip60 represses the FE65-Gal4DB-dependent transactivation is not clear, our mutational analysis suggests that both acetyltransferase activity and FE65-Tip60 interaction are involved.
Based upon this study and previous results (9, 21), we can now summarize the relative transcriptional activities of permutations and combinations of FE65, APP, and Tip60 as revealed in the Gal4DB experimental paradigm ( Table 1). The analysis leads us to the following tentative conclusions. 1) Given that FE65-Gal4DB by itself can trigger the reporter gene, unlike either APP-Gal4DB or Gal4DB-Tip60, FE65 is most likely to be the dominant transcriptional activator. 2) Given the above conclusion and the evidence that FE65 is required for APP-Gal4DB to trigger the reporter gene via direct FE65-APP binding, the APP moiety of APP-Gal4DB constructs is most likely to serve as a linker connecting FE65 to Gal4DB, thus leading to expression of the reporter gene. 3) Because Tip60 exhibits similar repressive effects on both FE65-Gal4DB and APP-Gal4DB when all three proteins are coexpressed, and because the repression is mediated by FE65-Tip60 binding, Tip60 is an FE65-associated repressor in this pathway. One must explain a puzzling phenomenon, however. Although Gal4DB-Tip60 by itself is unable to trigger the reporter gene, it can turn on the reporter gene when both APP and FE65 are coexpressed. This appears to contradict to conclusion 3 above. One potential explanation derives from the fact that Tip60 can involve several transactivation pathways (38 -42), including repressing the basal transcriptional activity when directly fused to Gal4DB (41). There may, therefore, be pleiotropic consequences resulting from the fusion of Tip60 to Gal4DB.
Additional molecular and cellular properties also support the view that FE65 is the key transcriptional activator. We note, for example, that exogenously expressed FE65 is primarily located in the nucleus and that APP is the molecule that anchors FE65 in the cytosol (4,9,43) (Fig. 3D). Immunostaining also points to a nuclear localization for endogenous FE65 in various cultured cells and in the neuronal cells of mouse brain tissues. 3 It should also be noted that many proteins containing WW domains participate in the coactivation of transcription and in the modulation of RNA polymerase II activity (44). There are thus independent lines of evidence that support a physiological function of FE65 in the nucleus. There are two other FE65 family members, FE65L1 (45) and FE65L2 (46). These do not function in the Gal4DB-dependent pathway (47), although both contain a WW domain and bind to AICD. FE65 also possesses an acidic residue cluster, a feature that is characteristic of nuclear chaperones (4,48). Such clusters are absent in FE65L1 and  Data are from Refs. 9, 21, and this study. NT indicates not tested.

Coexpressions
Reporter activities APP Gal4DB-Tip60 ϩϩϩϩϩ Nuclear Signaling by FE65 FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 FE65L2. Thus, a nuclear function for FE65 may be the principal feature that differentiates it from other family members.
We have provided strong evidence that FE65 is the principal transcriptional transactivator in the commonly used Gal4DB-luciferase reporter system for assessing APP nuclear signaling. Tip60 emerges as a major FE65-associated repressor. Endogenous APP may modestly enhance the transactivation of FE65 but does not appear to be essential for this signaling pathway. Several putative "target" genes in this pathway have been claimed; none of them can be confirmed by independent studies, however (49). This may be due in part to the uncertainty about the functional roles of APP, FE65, and Tip60 in the signaling pathway.
Our results may provide useful information for researchers in choosing adequate approaches to search for the endogenous genes that are targeted by FE65 and modulated by APP and Tip60. The Gal4-reporter system and its equivalents have been extensively used for studying transcriptional activators, repressors, and their modifiers. Some of these factors have been shown to function both in vitro and in vivo (50).
A next logical step in our research would be a search for the target genes activated by FE65. Once such genes are identified, the effects of APP and/or Tip60 could then be evaluated. Despite its modest enhancement of the p97FE65-Gal4DB nuclear signaling, APP may have unexpected impacts on the endogenous targets of FE65, given its robust ability to anchor this nuclear protein in the cytosol. Processing of APP may release such anchorage, modulating availability of FE65 in the nucleus. Such modulations may be key events during discrete times of neuronal differentiation or synaptic remodeling (51).