A Transcriptively Active Complex of APP with Fe65 and Histone Acetyltransferase Tip60

See allHide authors and affiliations

Science  06 Jul 2001:
Vol. 293, Issue 5527, pp. 115-120
DOI: 10.1126/science.1058783

This article has a correction. Please see:


Amyloid-β precursor protein (APP), a widely expressed cell-surface protein, is cleaved in the transmembrane region by γ-secretase. γ-Cleavage of APP produces the extracellular amyloid β-peptide of Alzheimer's disease and releases an intracellular tail fragment of unknown physiological function. We now demonstrate that the cytoplasmic tail of APP forms a multimeric complex with the nuclear adaptor protein Fe65 and the histone acetyltransferase Tip60. This complex potently stimulates transcription via heterologous Gal4- or LexA-DNA binding domains, suggesting that release of the cytoplasmic tail of APP by γ-cleavage may function in gene expression.

Amyloid-β precursor protein is a cell-surface protein with a large NH2-terminal extracellular sequence, a single transmembrane region (TMR), and a short COOH-terminal cytoplasmic tail (1–4). The α- and β-secretases initially cleave APP at defined extracellular sequences outside of the TMR. Thereafter, γ-secretase cuts APP in the middle of the TMR to generate small extracellular peptides and an intracellular fragment that is composed of half of the TMR (10 to 12 residues) and the cytoplasmic tail (47 residues). The small secreted peptides include amyloid-β peptides (Aβ40 and Aβ42), which are involved in Alzheimer's disease. γ-Cleavage of APP requires presenilins, intrinsic membrane proteins that are mutated in some cases of familial Alzheimer's disease (1–5). In addition to APP, two closely related homologs, APLP1 and APLP2, are expressed in vertebrates and also appear to be cleaved by α- and γ-secretases (6). The structures of APP and APLPs resemble cell-surface receptors whose proteolysis may be triggered by an external ligand; indeed, several binding activities of the extra- and intracellular regions of APP have been identified [e.g., see (7–10)]. Furthermore, triple knockouts of APP, APLP1, and APLP2 in mice are lethal, suggesting that these proteins are essential (11). However, the function of APP and APLPs, and of their proteolytic cleavages, remains unclear. Clues to such a function come from Notch proteins, cell-surface receptors that are also cleaved in the TMR in a presenilin-dependent reaction (12–14). Cleavage of Notch proteins liberates a cytoplasmic fragment that regulates nuclear transcription (15, 16), raising the possibility that cleavage of APP may have an analogous role.

To examine a possible function for APP in transcription, we constructed fusion proteins of APP with the DNA binding domains of yeast transcription factor Gal4 (17) or bacterial transcription factor LexA (18). We engineered Gal4 and LexA into the intracellular tail of full-length APP695 at the cytoplasmic boundary of the TMR or fused them to the NH2-terminus of the isolated cytoplasmic tail. The two distinct DNA binding domains (Gal4 and LexA) were used in these experiments in order to avoid sequence-specific artifacts. We transfected the APP-Gal4 and APP-LexA fusion proteins into PC12, HEK293, COS, or HeLa cells, and measured transactivation of transcription from cotransfected Gal4- and LexA-dependent reporter plasmid expressing luciferase (19, 20). Isolated Gal4 and LexA were used as negative controls, and all cells were cotransfected with a constitutive β-galactosidase expression plasmid to standardize the transfection efficiency. Little transactivation was observed with the APP-Gal4 and APP-LexA fusion proteins over Gal4 or LexA alone (about fivefold; Fig. 1, A and B). Thus, the cytoplasmic tail of APP, when overexpressed in cells as a fusion protein with a heterologous DNA binding domain, does not greatly stimulate transcription by itself, suggesting that APP may require binding of a cofactor. To search for such cofactors, we performed yeast two-hybrid screens for proteins that bind to the cytoplasmic tail of APP (21). Similar to previous screens (7–9), Fe65 was the major interacting protein identified, although it was isolated at an unexpectedly high frequency (90% of all clones).

Figure 1

Fe65 potently stimulates transcription mediated by APP proteins fused to Gal4 or LexA DNA binding domains. (A to D) Results of transactivation assays obtained with the constructs schematically displayed below the panels. Diagrams exhibit representative experiments in which cells were cotransfected with a Gal4- or LexA-luciferase reporter plasmid (to measure transactivation), a β-galactosidase plasmid (to normalize for transfection efficiency), and the test plasmids identified by numbers below the bars. Constructs marked with an asterisk contain a point mutation in the NPTY sequence of the APP cytoplasmic tail that inactivates Fe65 binding. (A and B) Potent transactivation of transcription is achieved with APP-fusion proteins of both Gal4 and LexA DNA binding domains only when Fe65 is present. (C) This effect is specific for Fe65 because another binding protein to the APP cytoplasmic tail, Mint1/X11, does not achieve stimulation. (D) The Fe65-binding site in the APP cytoplasmic tail confers onto an unrelated sequence (the neurexin cytoplasmic tail) the property of transactivation. Note inhibition of stimulation by mutation of the Fe65-binding site in all instances. The normalized luciferase activity is expressed as fold induction over transcription by Gal4 or LexA alone (A to C), or as fold induction over APP-Gal4 alone (D). Similar results obtained with COS, HEK293, PC12, and HeLa cells (47).

We then tested whether binding of Fe65 to APP could activate transcription. Cotransfecting Fe65 with the APP-Gal4 and APP-LexA fusion proteins greatly stimulated transcription (e.g., >2000-fold over Gal4 in HeLa cells), suggesting that Fe65 is a potent transactivator (Fig. 1, A and B). Strong transactivation by Fe65 was observed in all cell lines tested (PC12, HEK293, COS, or HeLa cells), whereas Mint1/X11, which also binds to the cytoplasmic tail of APP (9–14), had no significant effect (Fig. 1C). Neither Fe65 nor Mint1/X11 affected transcription of the control β-galactosidase plasmid cotransfected into all cells. Because Fe65 stimulated transcription even when Gal4 or LexA was inserted into full-length APP, the APP fusion proteins appear to be cleaved in the transfected cells. Immunoblotting revealed that in transfected COS cells, the size of the APP-Gal4 cleavage products corresponds precisely to those of the α- and γ-cleavage products of APP, indicating that APP is cleaved correctly by γ-secretase (22) and that the hydrophobic NH2-terminal sequence in the γ-cleavage product does not inhibit nuclear translocation.

In the cytoplasmic tail of APP, Fe65 binds to the NPTY sequence (8, 9, 23). To test whether binding of Fe65 to APP mediates transactivation, we mutated the NPTY sequence to NATA. As shown by yeast two-hybrid assays and coimmunoprecipitations, this mutation abolished Fe65 binding (22). The same mutation also abolished the Fe65-dependent stimulation of transcription (Fig. 1, A and B). Furthermore, when we replaced the cytoplasmic tail of APP-Gal4 with that of neurexin 1 (24), Fe65 was unable to stimulate transactivation (about twofold induction; Fig. 1D). However, when we transplanted 32 amino acids from the cytoplasmic tail of APP containing the NPTY sequence into the neurexin cytoplasmic tail, potent transactivaton by Fe65 was recovered (>200-fold induction). These results are consistent with the notion that Fe65 potentiates transcription by directly binding to APP-Gal4 and APP-LexA.

The Fe65 multidomain protein is composed of a negatively charged NH2-terminal sequence with no homology to other proteins, a central WW domain, and two COOH-terminal PTB domains, the PTB1 and PTB2 domains (8, 9). The negatively charged NH2-terminal sequences of Fe65 stimulates Gal4-dependent transcription (25), the PTB1 domain binds to the transcription factor CP2/LSF/LBP1 (26), and the PTB2 domain interacts with the cytoplasmic tail of APP (8,9). As an initial approach to elucidating how Fe65 activates transcription, we examined a series of Fe65 deletion mutants (27). Transactivation assays with these mutants showed that the WW domain and both PTB domains of Fe65 are essential for stimulating transcription, independent of the DNA binding domain used (Fig. 2). By contrast, deletion of the NH2-terminal third of Fe65 with the acidic region suspected of activating transcription (25) had no effect. In addition to deletions, we used point mutations in the WW domain to assess the precise need for the WW domain in transactivation. Substitution of one of the conserved tryptophan residues of the WW domain of Fe65 did not impair transactivation, but replacement of the central YYW motif with alanine residues abolished transactivation (Fig. 2). Immunoblotting confirmed that mutant Fe65 proteins were stably expressed, and immunoprecipitations showed that mutants still interacted with APP as long as the PTB2 domain was present, suggesting that the Fe65 mutants were not inactive because of instability or of inability to bind to APP (22). Thus all three canonical domains of Fe65 (the WW domain and the two PTB domains) likely have to interact with target molecules in stimulating transcription.

Figure 2

The WW domain and both PTB domains of Fe65 are required for stimulation of transactivation. (A andB) Full-length APP-Gal4 and APP-LexA fusion proteins were cotransfected with Gal4- or LexA-reporter plasmids, β-galactosidase control plasmid, and the various Fe65 expression plasmids as indicated. The resulting transactivation is expressed in fold induction over transfection with APP-Gal4 or APP-LexA alone. (C) Domain structure of Fe65 with residue numbers of domain boundaries. The sequences of the point mutants in the WW domain are shown above the diagram, and the locations of the deletion mutants used in (A) and (B) are displayed below the diagram.

In order to activate transcription, Fe65 presumably interacts with DNA binding proteins, histone acetyltransferases, and general transcription factors. A candidate binding partner for Fe65 is the transcription factor LBP/CP2/LSF, which was reported to bind to the PTB1 domain of Fe65 (26). However, we observed only a weak interaction between LBP/CP2/LSF and Fe65 in yeast two-hybrid assays and detected no change in transactivation when we cotransfected LBP/CP2/LSF with Fe65 and APP-Gal4 (27). Therefore, we searched for other potential Fe65-interacting proteins using yeast two-hybrid screens. We identified a single prey clone that strongly bound to the PTB1 domain of Fe65 (28). This clone encodes Tip60, a histone acetyltransferase that is expressed in two alternatively spliced forms [Tip60α and β (29, 30)]. Tip60 is part of a large nuclear protein complex that contains structural DNA binding, ATPase, and DNA helicase activities in addition to histone acetyltransferase activity (31, 32). Quantitative yeast two-hybrid assays and glutathioneS-transferase (GST) pull-down assays confirmed a strong interaction of Fe65 with rat and human Tip60 proteins [Fig. 3, A to C (28,33)]. GST-Tip60 efficiently pulled down APP together with Fe65; conversely, a GST-fusion protein of the APP cytoplasmic tail captured Tip60 together with Fe65, suggesting that the APP cytoplasmic tail, Fe65, and Tip60 form a stable trimeric complex in vitro (Fig. 3, B and C).

Figure 3

Histone acetyltransferase Tip60 binds to Fe65 and forms a multimeric complex with Fe65 and the cytoplasmic tail of APP. (A) Quantitative yeast two-hybrid assays of Fe65 binding to wild-type rat Tip60 (residues 63 to 454) and mutant Tip60 (Tip60*) in which the NKSY sequence in a loop of the histone acetyltransferase domain (residues 257 to 260) was mutated to NASA. The Fe65-bait vectors encode either both PTB domains or only the PTB1 domain; a lamin bait vector was used as negative control as indicated above the bar diagram. (B and C) GST pull-down assays with wild-type and mutant GST-Tip60 (B) or wild-type and mutant GST-APP cytoplasmic tail (C). Proteins from COS cells cotransfected with Myc-Fe65 (128-711) and APP (B) or Myc-Fe65 (128-711) and HA-Tip60 (C) were used for GST pull-down experiments, and bound proteins were analyzed by immunoblotting with antibodies to Myc-Fe65 (128-711), APP, and HA-Tip60, as indicated. (D) Domain structures of Tip60 and Fe65, and locations of their interacting sequences. Tip60 contains an NH2-terminal chromo domain and a COOH-terminal histone acetyltransferase (HAT) sequence that includes the NKSY sequence required for binding to Fe65.

Although PTB domains usually bind to NPXY sequences, variant binding sequences have also been observed (34, 35). In a search for a possible PTB domain target sequence in Tip60, we detected a single motif that is remotely similar to the NPXY sequence (NKSY; residues 257 to 260). Mapping of the NKSY sequence onto the three-dimensional structure of Esa1, a related yeast histone acetyltransferase (36), suggests that the NKSY sequence is located on a surface loop of a conserved domain, and thus it is accessible for a binding partner. To test whether the PTB1 domain of Fe65 binds to this site, we mutated the Tip60 NKSY sequence into NASA. No binding of Fe65 was observed for the mutant as measured either by quantitative yeast two-hybrid assays or GST pull-down assays (Fig. 3, A and B), suggesting that the PTB1 domain of Fe65 binds to the NKSY sequence in Tip60.

We next examined the localizations of APP, Fe65, and Tip60 in transfected cells [Fig. 4(37)]. When expressed alone, Fe65 and Tip60 are colocalized in the nucleus as reported previously (29–32, 38). APP probed with an antibody to the cytoplasmic tail exhibits a diffuse vesicular distribution, consistent with the fact that the majority of APP is uncleaved and probably localizes to recycling vesicles. Upon cotransfection of wild-type APP with Fe65 and Tip60, most Fe65 shifts to the same location as APP, presumably because it binds to uncleaved APP in cytoplasmic vesicles (Fig. 4). When Fe65 is coexpressed with mutant APP that is unable to bind Fe65, it resumes a largely nuclear localization. In contrast, Tip60 was always present in the nucleus (Fig. 4). Together these data suggest that overexpressed APP binds to Fe65 so strongly that it captures most of the Fe65 in the cytoplasm, indicating that Fe65 can move between nucleus and cytoplasm. We did not detect significant staining for APP in the nucleus, possibly because the fraction of APP that is cleaved is very small, and because the cytoplasmic APP fragment produced may be unstable, as previously observed for Notch proteins (12–16).

Figure 4

Localization of Fe65, Tip60, and APP in transfected HeLa cells. (A to D) Triple-labeling of cells coexpressing APP (green), Fe65 (red; labeled by an NH2-terminal HA tag) and Tip60 (blue; labeled by an NH2-terminal Myc tag). (E to H) Triple-labeling of cells as in (A to D), except that mutant APP (APP*), which is unable to bind Fe65, was transfected. (I toK) Localization of Fe65 and Tip60 in cells not expressing APP. Note that Tip60 is concentrated in the nucleus (N) under all conditions. In contrast, Fe65 is nuclear in the absence of APP or in the presence of mutant APP*, but largely relocalized to the cytoplasm by wild-type APP. Because of limited sensitivity, the apparent lack of nuclear staining for APP does not exclude the possibility that a fraction of APP (<5%) is localized to the nucleus. Calibration bars, 10 μm.

The interaction of Tip60 with Fe65 is potentially important for the transcriptional activation mediated by Fe65 because it couples Fe65 to the Tip60 complex and thus directly links Fe65 to transcriptional regulation. Our observations could be explained by at least two models that imply different functions for APP in Fe65-dependent transcriptional activation. The first model suggests that the physiological role of APP is to keep Fe65 out of the nucleus, and that γ-cleavage of APP liberates Fe65 for a nuclear function. According to this model, Fe65 transactivates APP-Gal4 or APP-LexA, because it latches Fe65, Tip60, and other nuclear proteins onto the heterologous DNA binding domains that are artificially fused to APP. The second model, by contrast, suggests that the cytoplasmic tail of APP normally functions in transcription when it is bound to nuclear Fe65, and that the cytoplasmic tail of APP released by γ-cleavage has a direct function in transcription. According to the second model, Fe65 stimulates APP-Gal4– or APP-LexA–mediated transcription because APP has a direct function in transcription independent of a heterologous DNA binding domain. Both models link γ-cleavage of APP to transcriptional activation, but assign different roles for the released cytoplasmic tail in transcriptional activation.

To differentiate between these models, we constructed a Gal4-Tip60 fusion protein, and tested the effects of Fe65 and APP on Gal4-dependent transactivation [Fig. 5(39)]. Gal4-Tip60 alone was unable to support significant Gal4-dependent transcription (no activation over Gal4 alone). Expression of either Fe65 or APP alone with Gal4-Tip60 did not enhance transactivation. However, when we coexpressed Gal4-Tip60 with both Fe65 and APP, transactivation was stimulated dramatically (∼100-fold; Fig. 5). Mutant APP that is unable to bind to Fe65 (APP*) was largely inactive (∼10-fold enhancement). Furthermore, little transactivation was observed when Fe65 and APP were coexpressed with mutant Gal4-Tip60 (Gal4-Tip60*) which is unable to bind to Fe65, or with Gal4 only. As described above, Fe65-dependent stimulation of transactivation by APP-Gal4 and APP-LexA proteins requires all three canonical Fe65 domains (the WW domain and the two PTB domains) (Fig. 2). To test whether the same applies for the Fe65- and APP-mediated transactivation of Gal4-Tip60, we examined Fe65 mutants in this assay (Fig. 5). As expected, the NH2-terminal sequence of Fe65 was not needed for potentiating Gal4-Tip60–dependent transactivation, whereas the PTB2 domain that binds to APP was essential. The WW domain of Fe65 was also essential, indicating that Fe65 interacts with additional factors besides Tip60 and APP in stimulating transactivation. Together these data suggest that the cytoplasmic tail of APP has a direct active role in stimulating transactivation and that it collaborates with Fe65 in enhancing transcription by Gal4-Tip60.

Figure 5

Fe65 and APP are required for transactivation of Gal4-Tip60. Gal4-fusion proteins with full length human wild-type Tip60 (Gal4-Tip60), mutant Tip60 (Gal4-Tip60*), or Gal4 alone were cotransfected into COS cells with a Gal4-reporter plasmid, a β-galactosidase control plasmid, and the indicated APP and Fe65 expression plasmids. The mutant Tip60 used carries a point mutation that inactivates Fe65 binding (see Fig. 3). Cotransfected Fe65 proteins include deletions of the PTB2 domain, which binds APP (Fe65ΔPTB2), and a point mutation in the WW domain that inactivates the ability of Fe65 to stimulate APP-Gal4 mediated transcription (Fe65mW4). The APP mutant used (APP*) is unable to bind Fe65. Note that Gal4-Tip60 transactivation depends on the presence of both wild-type APP and Fe65.

APP is physiologically processed by α- or β-secretases followed by γ-secretase (1–4). However, the orderly and regulated degradation of APP has been a puzzling phenomenon ever since it was initially described because it has no apparent biological benefit. We now report three observations that link γ-cleavage of APP to transcriptional activation. First, we show that the cytoplasmic tail of APP forms a multimeric complex with Fe65, a multidomain adaptor protein (8, 9), and Tip60, a histone acetyltransferase that functions in chromatin remodeling, DNA repair, and transcription (31, 32). Second, we demonstrate that APP fused to DNA binding domains from yeast Gal4 or bacterial LexA exhibits only weak transcriptional activity (about fivefold over Gal4 alone) but is dramatically stimulated by Fe65 (>2000-fold in the case of APP-Gal4). This stimulation requires intact binding sites for APP and for Tip60 in Fe65 and a functional WW domain in Fe65. Third, we find that Gal4-Tip60 is transcriptionally inactive by itself but can be potently transactivated upon coexpression of APP and Fe65. Both APP and Fe65 are required, and any mutation that disrupts the complex of Fe65 with the APP cytoplasmic tail and with Tip60 abolishes this effect.

The most parsimonious explanation for these observations is that the complex of Fe65 with the cytoplasmic tail of APP directly acts in transcription when bound to Tip60. However, several alternative hypotheses are also possible. For example, APP may initiate a nuclear signal without participating in transcription, although this seems less likely in view of the fact that Tip60 and Fe65 can form a direct complex with the cytoplasmic tail of APP. A direct function of the cytoplasmic tail of APP in transcription would agree remarkably well with that of Notch proteins (15, 16). The proposed similarity between APP and Notch provides evidence for a general role of presenilin-dependent proteolysis in regulating transcription (40, 41). According to this hypothesis, cytoplasmic tails of cell-surface proteins are released by presenilin-dependent proteolysis so that they can then regulate transcription. In spite of this similarity, however, the upstream regulation of proteolysis and the downstream target transcription factors are distinct and place APP and Notch into different functional contexts.

The proposed function of APP and its homologs may help to explain why double or triple knockouts in these proteins are lethal (11), and why overexpression of COOH-terminal APP-fragments is cytotoxic (42, 43). However, the current studies have limitations that must be overcome before a function for APP in transcription is assured. Our experiments were performed with chimeric proteins containing exogenous DNA binding domains (LexA and Gal4). If the APP-Fe65-Tip60 complex physiologically regulates transcription, it must be coupled to an endogenous DNA binding protein that remains to be identified. This DNA binding protein may be a component of the Tip60 complex that exhibits DNA binding activity (32), or it may interact with the WW domain of Fe65. Furthermore, all of the current experiments were performed in transfected cells. It will be necessary to demonstrate that endogenous proteins perform similar functions, which will require identification of endogenous genes that are activated by the complex. In addition, demonstration that the cytoplasmic tail of APP normally enters the nucleus, which was also long elusive for Notch, is still lacking. Finally, Notch cleavage is regulated by ligand binding (15,16), suggesting that APP cleavage may also be regulated. Because regulation of APP cleavage will not only control the release of the cytoplasmic tail but also secretion of amyloid β-peptides, this regulation could be important in Alzheimer's disease. Thus, insight into how APP cleavage is regulated represents another major goal of future research. The approaches and observations described in the current study may be helpful for these objectives that will not only be important for knowledge of the normal function of APP, but also for the pathogenesis of Alzheimer's disease.

  • * To whom correspondence should be addressed. E-mail: Thomas.Sudhof{at}


View Abstract

Stay Connected to Science

Navigate This Article