Generation of the β-Amyloid Peptide and the Amyloid Precursor Protein C-terminal Fragment γ Are Potentiated by FE65L1*

Members of the FE65 family of adaptor proteins, FE65, FE65L1, and FE65L2, bind the C-terminal region of the amyloid precursor protein (APP). Overexpression of FE65 and FE65L1 was previously reported to increase the levels of α-secretase-derived APP (APPsα). Increased β-amyloid (Aβ) generation was also observed in cells showing the FE65-dependent increase in APPsα. To understand the mechanism for the observed increase in both Aβ and APPsα given that α-secretase cleavage of a single APP molecule precludes Aβ generation, we examined the effects of FE65L1 overexpression on APP C-terminal fragments (APP CTFs). Our data show that FE65L1 potentiates γ-secretase processing of APP CTFs, including the amyloidogenic CTF C99, accounting for the ability of FE65L1 to increase generation of APP C-terminal domain and Aβ40. The FE65L1 modulation of these processing events requires binding of FE65L1 to APP and APP CTFs and is not because of a direct effect on γ-secretase activity, because Notch intracellular domain generation is not altered by FE65L1. Furthermore, enhanced APP CTF processing can be detected in early endosome vesicles but not in endoplasmic reticulum or Golgi membranes, suggesting that the effects of FE65L1 occur at or near the plasma membrane. Finally, although FE65L1 increases APP C-terminal domain production, it does not mediate the APP-dependent transcriptional activation observed with FE65.

Processing of the amyloid precursor protein (APP) 1 results in the generation of the amyloidogenic peptide, A␤, which plays a central role in the pathogenesis of Alzheimer's disease. Cleav-age of C99, the APP C-terminal fragment derived from ␤-secretase processing of APP, by ␥-secretase generates the A␤ peptide. Furthermore, ␥-secretase cleavage of C99 and C83, the ␣-secretase derived APP C-terminal fragment (APP CTF), releases the APP C-terminal domain (AICD), a 6-kDa peptide also called CTF␥ or AID, that regulates transcription after translocation to the nucleus (1)(2)(3)(4).
The majority of proteins reported to bind the 47-amino acid intracellular region of APP (5)(6)(7), including the FE65 protein family members FE65, FE65L1, and FE65L2, bind the YENPTY sorting motif of APP via a phosphotyrosine interaction domain (PID/PTB). YENP is a clathrin-coated pit internalization domain required for trafficking of APP into the endocytic pathway (8,9). Previous studies have shown that FE65 protein family members can alter the processing of APP by influencing APP trafficking. Increased maturation of APP and increased ␣-secretase-cleaved APP (APPs␣) secretion was observed in H4 neuroglioma cells induced for FE65L1 overexpression (10). Furthermore, enhanced secretion of APPs␣ and A␤ was reported in Madin-Darby canine kidney APP695 cells stably overexpressing FE65 (11). Cell surface APP levels were elevated in these cells, and increased routing of APP into the endocytic pathway from the plasma membrane was suggested to account for the observed increase in A␤ (11).
The FE65 proteins are adaptor proteins that have three protein-protein interaction domains, a WW domain and two PID/PTB domains, and are thus capable of mediating the formation of protein complexes. The N-terminal PID/PTB domain (PID1) of FE65 mediates the interaction of FE65 with the low density lipoprotein receptor-related protein, LRP (12), the histone acetyltransferase Tip60 (3), and the transcription factor CP2/LSF/LBP-1 (13), whereas the C-terminal PID/PTB domain of the FE65 protein family members binds APP (14 -16). In addition, the WW domain binds the mammalian homolog of Drosophila Enabled (17), implicated in cell motility (18). Biochemical complexes have been demonstrated for FE65⅐AICD⅐Tip60 (3) and FE65⅐APP⅐Mena (11) but only postulated for APP⅐FE65⅐LRP (12). A biologically relevant role for the FE65⅐AICD⅐Tip60 complex in transcriptional activation is suggested by its ability to regulate expression of two members of the tetraspanin superfamily (4,19). In addition, upregulation of FE65 and Tip60 is observed in mouse brains of APP transgenic mice and results in enhanced expression of the tetraspanin, KAI-1 (4).
Given the importance of the APP CTFs for A␤ generation and transcriptional activation and the fact that the FE65 proteins bind the NPXY motif present in all APP CTFs, we examined the effect of FE65L1 overexpression on APP CTF processing in H4 neuroglioma cells. Our data show that FE65L1 overexpression results in increased generation of AICD and A␤ in these cells by increasing ␥-secretase processing of both C83 and C99. Furthermore, FE65L1 does not directly alter ␥-secretase activity, but binding of FE65L1 to APP and APP CTFs is necessary for FE65L1-dependent ␥-secretase processing of APP CTFs. These data suggest that the interaction of FE65L1 with APP CTFs facilitates access of ␥-secretase to its substrates. In addition, we provide evidence that these events occur at or near the cell surface.
Cell Culture, cDNA Constructs, and Transfections-The human H4 neuroglioma cell line, 32-6, is tetracycline-repressible for human HAtagged FE65L1 and expresses endogenous levels of APP (10,26). The tetracycline-repressible HA-tagged mutant FE65L1 cell line (C702V) expressing a mutant allele of FE65L1 bearing a C 3 V substitution at residue 702 of FE65L1 (National Center for Biotechnology accession number XM_051782), a conserved residue located in the APP binding PID2 domain, was established as described (26). The C702V mutation was created in the pHA-FE65L1 plasmid using the primers 5Ј-GCAC-TTCTGATATCGTAACATGACGGCGGCCTGCACCGCT-3Ј and 5Ј-AG-CGGTGCAGGCCGCCGTCATGTTACGATATCAGAAGTGC-3Ј by PCR mutagenesis (QuikChange, Stratagene). Cells induced for 72 h by removal of tetracycline showed similar expression levels for WT HA-FE65L1 and C702V HA-FE65L1. FuGENE 6 (Roche Applied Science) was used for all transient transfections. The mammalian notch N ⌬E construct (a generous gift from R. Kopan) was transiently transfected into H4 neuroglioma cells uninduced and induced for FE65L1 at 24 h post-induction, and cells were harvested at 72 h. Proteasomal, lysosomal, calpain, and ␥-secretase inhibitors in Me 2 SO were added at the concentrations indicated to complete media containing the appropriate antibiotics for cell treatments of 3 h with the exception of the lactacystin treatment, which lasted 24 h.
Immunoprecipitations and Western Blot Analyses-Cells were lysed in extraction buffer (10 mM Tris, pH 8.0, 140 mM NaCl, 5 mM EDTA, and 1% Triton X-100) containing protease inhibitors. Protein concentrations of total cell lysates were determined by the BCA protein assay (Pierce). APP CTFs were immunoprecipitated from 500 g of total protein using the A8717 antibody and protein A magnetic beads (Qiagen). The precipitates were washed three times in extraction buffer and subjected to electrophoresis on 10 -20% acrylamide Tris-Tricine gels and transferred to polyvinylidene difluoride membranes (Millipore), and Western blot analysis was performed using the 13G8 antibody according to standard protocols.
Metabolic Labeling-H4 32-6 cells were grown to 80 -90% confluency in the presence or absence of tetracycline on 150-mm-diameter culture dishes. Cells were trypsinized and resuspended in 1 ml methionine-and cysteine-free Dulbecco's modified Eagle's medium. After a 30-min starvation at 37°C, cells were labeled for 20 min with 250 mCi/ml [ 35 S]me-thionine and [ 35 S]cysteine (PerkinElmer Life Sciences), washed, and chased in complete medium containing excess (5-fold) unlabeled methionine and cysteine. The addition of inhibitors to the growth medium was made 2 h before initiation of each pulse/chase experiment, and tetracycline was added to maintain the uninduced state of the controls throughout. Cell lysate volumes containing 1000 g of protein were immunoprecipitated with APP CTF antibody and 20 l of protein A/G PLUS-agarose (Santa Cruz Biotechnology). The immune complexes were separated on a 10 -20% Tricine gel. The gels were dried after fixation, and the radioactive signal was detected using a Cyclone phosphorimaging device (PerkinElmer Life Sciences).
Cell-free AICD Generation-AICD generation was examined in vitro using an active membrane preparation as described by Pinnix et al. (27). The 10,000 ϫ g active post-nuclear membrane fraction containing the same amount of protein (5 mg) isolated from cells uninduced or induced for FE65L1 was resuspended in 200 l of Buffer A (50 mM HEPES buffer, 150 mM NaCl, 5 mM EDTA, pH 7.4, containing 5 mM 1,10-phenanthroline) and incubated at 37°C or on ice for the indicated times. After the incubation, the released APP CTFs and AICD were separated from the membranes by centrifugation at 10,000 ϫ g. APP CTFs and AICD were immunoprecipitated with A8717 (20 g of purified serum IgG) from the reaction supernatant. The membrane fractions were lysed in radioimmune precipitation assay buffer. Western blot analysis using the A8717 antibody was performed on these immune complexes, and equal volumes of the active membrane fractions were loaded after completion of the reaction.
Subcellular Fractionation-Five 15-cm dishes of cells were resuspended in 3 ml of homogenization buffer containing 10 mM HEPES, pH 7.4, 1 mM EDTA, and 0.25 M sucrose plus protease inhibitor. Cells were disrupted by 10 strokes in a metal homogenizer followed by 5 passages through a 27-gauge needle. Nuclei and unbroken cells were removed by centrifugation at 1,500 ϫ g for 10 min. The post-nuclear supernatants were centrifuged for 1 h at 65,000 ϫ g. The resultant membrane pellets were resuspended in 0.8 ml of homogenization buffer. Discontinuous iodixanol gradients were prepared as described by Xia et al. (28) starting from an OptiPrep (Invitrogen) gradient stock solution containing a final concentration of 50% OptiPrep in 250 mM sucrose, 1 mM EDTA, and 10 mM HEPES, pH 7.4. The resuspended vesicle fractions were loaded on top of the gradients and centrifuged in a SW41 rotor at 40,000 rpm for 2.5 h at 4°C. Fractions of 1 ml were collected.
Early Endosome Isolation-Five 15-cm dishes of cells were resuspended in 2-ml homogenization buffer containing 10 mM HEPES, pH 7.4, 320 mM sucrose, 1 mM EDTA, 1 mM EGTA, and 1 mM MgCl 2 with protease inhibitors. Cells were disrupted by 10 strokes in a metal homogenizer followed by 5 passages through a 27-gauge needle. After centrifugation for 10 min at 4,000 ϫ g, the post-nuclear supernatants were collected and loaded on an OptiPrep gradient consisting of 1 ml of 60%, 3 ml of 30%, 3 ml of 20%, and 3 ml of 10% OptiPrep diluted in buffer with 250 mM sucrose, 6 mM EDTA, 60 mM HEPES, pH 7.4 in 13-ml Beckman SW41 centrifuge tubes. After centrifugation at 40,000 rpm for 1 h and 15 min, the endosome-enriched light vesicle fraction was collected by taking the 0 -10% interface. Early endosome vesicles were further enriched from the light vesicle fraction (1 mg) by immunoprecipitation of early endosome vesicles with the EEA1 antibody (1 g) and goat anti-mouse IgG magnetic beads (Polysciences Inc.).
APPs␣ and A␤ Measurements-Media conditioned for 4 h was collected from cells uninduced or induced for HA-hFE65L1 for 65 h, and APPs was enriched by MacroPrep High Q chromatography as previously described (10). Cell lysates corresponding to each of the conditioned media samples were obtained, and Western blot analysis using 6E10 showed that cellular APP levels were no different for cells induced for WT FE65L1 or FE65L1 C702V (data not shown). APPs␣ was detected by Western blot analysis from the MacroPrep High Q eluates using the 6E10 antibody. For A␤ measurement, conditioned media was collected after a 24-h incubation from cells that were either uninduced or induced for FE65L1 expression for 48 h. Endogenous A␤ levels were measured by Sandwich enzyme-linked immunosorbent assay using the antibodies BAN50/BA27 for A␤ 1-40 and BAN50/BC05 for A␤1-42 as previously described (29).
Luciferase assays were performed with the luciferase assay system (Promega), and ␤-galactosidase activity was measured with the luminescent ␤-gal kit (BD Biosciences). Luciferase measurements were normalized for transfection efficiency using the ␤-galactosidase activity values. Results are expressed as fold increase over the luciferase activity obtained with the pMST control plasmid.

RESULTS
Increased ␥-Secretase Cleavage of the ␣-Secretase-derived APP CTF, C83, in H4 Cells Overexpressing FE65L1-We examined APP CTF levels in H4 neuroglioma cells overexpressing FE65L1 to determine whether FE65L1 influenced APP CTF stability. Western blot analysis of total protein lysates obtained from cells uninduced or induced for FE65L1 overexpression using an APP CTF antibody (13G8) showed a single 10-kDa APP CTF fragment corresponding to the ␣-secretasederived APP CTF, C83, that was dramatically decreased in lysates obtained from H4 cells induced for FE65L1 overexpression (Fig. 1A).
To determine whether the effect of FE65L1 on C83 steady state levels was due to increased degradation of C83, we tested lysosome and proteasome inhibitors for their ability to block the FE65L1-dependent decrease in C83. Our data show that neither ammonium chloride nor chloroquine had any effect on the observed decrease in C83 upon FE65L1 overexpression (Fig. 1B), indicating that the lysosomal pathway is not responsible for the observed reduction in C83 levels. MG132 cell treatments resulted in accumulation of APP CTF␤ and A␤40, indicating that the proteasome plays a role in APP CTF␤ degradation (30). Similarly, calpain was previously implicated in the degradation of C83 and C99 and in the generation of A␤, particularly A␤42 (20,31). To test the possibility that calpain or proteasomal processing of C83 was increased in H4 cells overexpressing FE65L1, we treated cells with the calpain inhibitor CI-III and the proteasome inhibitors MG132 and lactacystin and examined C83 levels. Neither proteasomal nor calpain inhibitors were able to abolish the FE65L1-dependent decrease in C83 (Fig. 1B). However, a comparison of the APP CTF levels in Me 2 SO-treated versus drug-treated cells showed that the drug-treated cells had higher APP CTF levels in both uninduced and induced cells than the Me 2 SO controls. These data suggest that these inhibitors affect activities that alter APP CTF levels but do not prevent the FE65L1-dependent C83 decrease. ALLN, reported to inhibit calpain, the proteasome, and cathepsins (31), was able to partially inhibit the FE65L1dependent decrease in C83, but did not completely reverse the effects of FE65L1 on APP CTF processing. For ALLN treatments, C83 levels were higher in the uninduced cells treated with ALLN by 1.8-fold, and accumulation of a higher APP CTF fragment was also observed (possibly C89) (Fig. 1B). We can conclude from these data that the observed decrease in C83 in H4 cells overexpressing FE65L1 is not due to increased catabolism of C83 by calpain, the proteasome, or the lysosome.
Both C99 and C83 are substrates for ␥-secretase (32)(33)(34). To determine whether FE65L1 can modulate ␥-secretase cleavage of C83, we treated H4 neuroglioma cells uninduced and induced for FE65L1 overexpression with structurally different ␥-secretase inhibitors (22,24). Our results show that all five ␥-secretase inhibitors can restore C83 levels to those of the uninduced controls (Fig. 1C). In fact, ␥-secretase inhibitor treatments produce C83 levels that are higher in cells overexpressing FE65L1 than in the control cells, consistent with our previous observation that APP undergoes increased processing by ␣-secretase (10). These data indicate that C83 processing by ␥-secretase can account for the FE65L1-dependent increase in C83 turnover.
FE65L1 Overexpression Decreases the Half-life of C83 and Facilitates AICD Generation-Cleavage of C83 by ␥-secretase results in the generation of p3 and AICD (35). Given that C83 processing by ␥-secretase is increased in H4 neuroglioma cells overexpressing FE65L1, we predicted that AICD generation from these cells would be increased. To test whether FE65L1 decreases the half-life of C83 and produces higher levels of AICD, H4 cells uninduced and induced for FE65L1 overexpression were metabolically labeled for 20 min and chased with excess methionine for up to 4 h ( Fig. 2A). Our results show that C83 is produced within 30 min of chase in cells uninduced for FE65L1 ( Fig. 2A). However, in cells induced for FE65L1 the levels of C83 detectable at the 30 min chase time point is significantly lower than in uninduced cells, and a 6-kDa band with the mobility of AICD is observed (Fig. 2A). These data suggest that C83 is processed by ␥-secretase to generate AICD in H4 cells overexpressing FE65L1. AICD detection was ob- served as early as 20 min of chase in these cells, which corresponds to the earliest time at which C83 could be detected in cells uninduced for FE65L1 (data not shown), indicating that FE65L1 is available for binding and facilitates AICD generation immediately upon APP CTF generation. In a previous pulse/chase study examining the effect of FE65L1 on APP maturation we observed an increase in the ratio of mature/ immature APP that is seen at the 30-and 60-min chase time points ( Fig. 2A) (10). In addition, we previously showed that the earliest time point at which increased APPs␣ secretion was detected was 1 h (10). Detection of C83 as early as the 20-min chase time point suggests that APP cleavage by ␣-secretase occurs intracellularly in H4 neuroglioma cells, consistent with the results of Kuentzel et al. (36), showing APPs␣ release from H4 microsomal membranes.
The effects of ␥-secretase inhibitors (DAPT and WPE-III-31C), a proteasome inhibitor (MG132), a calpain inhibitor (CI-III), and an inhibitor of lysosomal degradation (chloroquine) on the levels of C83 were examined at the 2-hour chase time point (Fig. 2B). Our results indicate that only the ␥-secretase inhibitors restored C83 levels to those observed in the uninduced cells.
We also examined AICD production in H4 neuroglioma cells induced for FE65L1 with a cell-free AICD generation assay (27,34,37). Our data show that incubation of microsomal membranes over a 2-h period produces similar levels of AICD in cells uninduced and induced for FE65L1 overexpression (Fig.  2C). However, we can conclude that more AICD is generated per molecule of C83 in cells overexpressing FE65L1 because membranes isolated from these cells have lower C83 levels (Fig. 3C). We have also verified that ␣-secretase activity is not saturated in this assay because greater levels of AICD were produced from membranes isolated from H4 cells stably overexpressing APP751 (data not shown). Thus, analysis of the ratio of AICD generated over C83 substrate levels (time 0) indicates that the level of AICD generated is 31% greater for cells overexpressing FE65L1 than uninduced cells (AICD/C83; uninduced ϭ 0.15, induced ϭ 0.22). We also observed C83 in the supernatant fraction at time 0, most likely because of a release of C83-containing vesicles upon resuspension of the membrane pellet. An increase in C83 production over the incubation period was also evident, indicating that C83, like AICD, can be generated in this assay. These data confirm our results of the pulse/chase study showing that FE65L1 potentiates ␥-secretase cleavage of C83.
Binding of FE65L1 to APP Is Required for Its Effect on APPs␣ Secretion and C83 Processing-FE65L1 binds the YENPTY motif of APP via its C-terminal phosphotyrosine interaction domain (PID2) (15). A cysteine to alanine substitution at residue 702 of the FE65L1 protein sequence located in the APP binding domain of FE65L1 was generated. This cysteine residue is conserved in all six PID/PTB domains of the three FE65 protein family members and was previously shown to be required for binding of FE65 to APP (38). Our data show that APP is not recovered in anti-HA immune complexes containing the HA-FE65L1 C702V mutant protein even though similar amounts of the WT and mutant FE65L1 protein are immunoprecipitated (Fig. 3A). In contrast to WT FE65L1, overexpression of the FE65L1 C702V mutant does not result in increased APPs␣ secretion (Fig. 3B). APP steady state levels are unaltered by overexpression of the FE65L1 C702V mutant (data not shown), similar to what was previously reported for WT FE65L1. Furthermore, C83 levels are not decreased in the presence of the mutant protein (Fig. 3C), indicating that binding of FE65L1 to APP is also required for its effect on the ␥-secretase-dependent C83 processing. These data indicate that binding of FE65L1 to APP is required for its effects on APPs␣ secretion and ␥-secretase-dependent C83 processing.
FE65L1 Overexpression Does Not Directly Influence the Activity of ␥-Secretase-␥-Secretase also cleaves Notch to generate the Notch intracellular domain, NICD (37). To confirm our observation that FE65L1 modulation of ␥-secretase activity is ]cysteine for 20 min and chased for the times indicated. APP full-length and APP CTFs were immunoprecipitated from 1000 g of total protein using 1 l of A8717 antibody, which recognizes the last 20 C-terminal residues of APP, and 20 l of protein A/G PLUS-agarose followed by electrophoresis in 10 -20% Tris-Tricine gels. Detection of labeled APP and APP CTFs was achieved using a Cyclone phosphorimaging system. In the top panel a 1-h exposure of the dried gel shows mature and immature full-length APP, and the bottom panel shows a 3-day exposure of the same gel for detection of APP CTFs and AICD. B, cells were pretreated with the inhibitors indicated for 2 h before labeling and throughout the pulse/chase experiment, which included a 2-h chase interval. The inhibitors used include the ␥-secretase inhibitors, WPE-III-31C and DAPT, the calpain inhibitor, CI-III, the proteasomal inhibitor, MG132, and the lysosomal inhibitor, chloroquine. C, AICD generation was examined in a cell-free assay by incubating membranes (10,000 ϫ g postnuclear pellet) in reaction buffer for 2 h at 37°C. APP CTFs released in the reaction supernatant were immunoprecipitated with the A8717 antibody, and the immune complexes as well as the APP CTFs present in the membrane pellet were detected by immunoblotting with the A8717 antibody. The membrane blot was reprobed with the anti-PS1 N-terminal antibody, Ab14. These data show a representative example of three independent experiments.

FIG. 3. Binding of FE65L1 to APP is required for the APP and APP CTF processing effects that occur upon FE65L1 overexpression.
A, substitution of the conserved cysteine within the FE65L1 PID2 domain to valine (C702V) blocks its binding to APP. Immunoprecipitates were obtained from lysates of uninduced (Ϫ) or induced (ϩ) 32-6 and C702V cell lines using an anti-HA antibody. The precipitated proteins were immunoblotted with the APP antibody (369W) and reprobed with the anti-HA antibody. B, 32-6 and C702V cell lines were either induced (ϩ) or uninduced (Ϫ) for FE65L1, and APPs␣ was detected by Western blot analysis using the anti-APP antibody 6E10 after enrichment of APPs from conditioned media as described under "Experimental Procedures." C, APP CTF␣ levels were analyzed in cell extracts obtained from 32-6 and C702V cell lines that were either induced (ϩ) or uninduced (Ϫ) for FE65L1 by immunoprecipitation with the A8717 antibody followed by immunoblotting with the 13G8 antibody.
specific for the substrates that it binds, APP and APP CTFs, we examined the effects of FE65L1 overexpression on NICD generation in cells expressing the Notch N ⌬E proteins (39). Our data show that FE65L1 overexpression does not alter NICD generation in H4 cells (Fig. 4), whereas an expected increase and decrease in NICD is observed in double stable CHO APP751/PS1WT and CHO APP751/PS1D385A, respectively. These data suggest that FE65L1 does not have a direct effect on ␥-secretase activity and are consistent with an effect of FE65L1 on ␥-secretase APP CTF processing that is dependent on the interaction of FE65L1 with APP CTFs.
FE65L1-dependent ␥-Secretase Processing of C83 Occurs at the Plasma Membrane or in the Early Endocytic Pathway-As a first step to determining the subcellular compartment for the ␥-secretase-dependent C83 processing, we performed ER/Golgi subcellular fractionations and examined C83 levels in each of these fractions. Although full-length APP was readily detected by Western blot analyses of the ER/Golgi fractions showing both immature and mature APP in the Golgi fractions and immature APP in ER vesicles (Fig. 5A), APP CTFs were not detected (data not shown).
We have previously reported the presence of FE65L1 in a crude light membrane fraction obtained from FE65L1-induced H4 cells (10). To further examine the association of FE65L1 with a subcellular membrane compartment, we performed Western blot analysis of the ER/Golgi fractions with an anti-HA antibody and found that FE65L1 localizes to both ER and Golgi membranes (Fig. 5A). In addition, FE65L1 was detected in association with early endosome vesicles enriched from a light vesicle fraction by immunoprecipitation with EEA1 or Rab5 antibodies (Fig. 5B). These data indicate that FE65L1 localization to membranes is not restricted to a specific membrane compartment and is consistent with the distribution of APP and APP CTFs, its binding partners.
Given that low levels of APP CTFs may not be detectable by Western blot analyses of ER/Golgi fractions, immunoprecipitations were performed on pooled ER fractions (calnexin positive) and Golgi fractions (GM130 positive) using an APP CTF antibody. Using this approach C83 levels were detectable in ER and Golgi pools but were found to be no different in cells overexpressing FE65L1 than in uninduced cells (Fig. 6A), suggesting that the FE65L1-dependent ␥-secretase cleavage of C83 does not occur in these compartments. The subcellular localization of constitutive ␣-secretase cleavage of APP has been reported to occur in a post-Golgi compartment in the secretory pathway of H4 neuroglioma cells (36), suggesting that C83 is generated in the late secretory pathway. Localization of C83 in ER/Golgi membranes suggests that a small fraction of C83 may be traf-ficked back into the ER and Golgi after it is generated.
In contrast to the ER/Golgi fractions where only C83 is detected, several APP CTFs were detected in early endosome vesicles. Their mobility on SDS/PAGE gels between the 7-and 17-kDa protein molecular mass markers suggests that these APP CTFs represent C83, C89, and C99 (Fig. 6B). The APP CTF fragment migrating at 14 kDa showed 6E10 immunoreactivity, confirming that it is in fact C99 (Fig. 6B). Interestingly, although APP levels are unchanged in early endosome vesicles, the levels of all three APP CTFs are reduced upon FE65L1 induction. These data suggest that FE65L1-dependent ␥-secretase cleavage of the APP CTFs occurs either at the plasma membrane or in early endosomes (Fig. 6B). The localization of PS1 to the early endosome compartment, determined by detection of the N-terminal PS1 fragment with Ab14, suggests that APP CTFs may be cleaved by ␥-secretase at this subcellular site (data not shown).
Secreted A␤40 Levels Are Increased in Cells Induced for FE65L1 Expression-To examine the effect of FE65L1 on A␤ generation we collected conditioned media from cells uninduced and induced for FE65L1 overexpression and measured A␤ by Sandwich enzyme-linked immunosorbent assay. A 26% increase in secreted A␤1-40 (p Ͻ 0.00000001) is observed for cells induced for FE65L1 compared with the uninduced control (Fig. 7). The data in Fig. 7 show a representative experiment that approximates the average increase in A␤1-40 observed upon FE65L1 overexpression for four independent experiments. The increased A␤40 secretion from cells overexpressing FE65L1 is significant for all four experiments. When FE65L1 C702V mutant protein was overexpressed, an increase in A␤1-40 was still observed (6%), but the increase in A␤1-40 is not statistically significant (Fig. 7). The average increase observed in four independent experiments for FE65L1 C702V is   FIG. 4. FE65L1 overexpression does not alter NICD production. The N ⌬E notch construct was transiently transfected into 32-6 cells, uninduced (Ϫ) and induced (ϩ), for 24 h and CHO cells stably overexpressing PS1 WT or PS1 D385A. Cell lysates were collected at 48 h post-transfection. Total protein (100 g) was separated on 4 -12% Tris-glycine gels and analyzed for NICD by immunoblotting with an NICD-specific antibody. The same blot was reprobed with an anti-Myc antibody, 9E10, recognizing N ⌬E and an anti-␤-tubulin antibody to verify protein loads.
4%, and only one experiment showed a significant difference in the levels of A␤1-40 (data not shown). The increase in A␤1-40 observed with the mutant protein most likely reflects residual binding of the mutant protein to APP that is not detected in other assays (see Fig. 3A). No statistically significant differences were observed for secreted A␤1-42 levels, but an average 17% increase was observed, suggesting that the trend is similar to that of A␤1-40 (data not shown). Our observations that A␤40 levels are increased in FE65L1-overexpressing cells, that APP levels in the early endosome vesicles are no different for cells induced for FE65L1 compared with the uninduced control, that C99 was only detected in early endosome vesicles, and that FE65L1 influence on ␥-secretase cleavage of APP CTFs is only detected in early endosome vesicles suggest that FE65L1 potentiates APP CTF cleavage at or near the cell surface. This interpretation is consistent with data indicating that secreted A␤ in non-neuronal cells is generated at the cell surface (40) and that A␤40 generation, unlike A␤42, occurs predominantly in the endocytic pathway (9,41).
FE65L1 Does Not Result in AICD-dependent Transcriptional Activation-FE65 was reported to promote AICD-dependent transcriptional activation (3). To determine whether FE65L1 plays a similar role, we tested whether binding of FE65L1 to APP-GAL4 could activate a GAL4-dependent luciferase reporter construct in H4 cells. Our data show that both rat FE65 and human FE65 are able to activate transcription (Fig. 8A). However, neither mouse nor human FE65L1 promoted transcriptional activation (Fig. 8A). The same results were obtained in COS7 and CHO cells and for non-HA tagged human FE65L1 in H4 cells (data not shown). To ensure that the FE65 proteins were expressed, we performed Western blot analysis of the transient transfections using an antibody that recognizes both FE65 and FE65L1. The FE65 and FE65L1 proteins were detectable in these transfections (Fig. 8B), supporting our conclusion that FE65L1 is not able to promote AICD-dependent transcriptional activation.

DISCUSSION
Members of the FE65 protein family that bind the NPXY motif of APP have the capacity to modulate APP processing by influencing trafficking of full-length APP molecules (10,11). Increased routing of cell surface APP into the endocytic pathway, where A␤ is known to be generated, was proposed as a mechanism for increased A␤ secretion in Madin-Darby canine kidney cells (11). However, an increase in A␤ production was reported in HEK293/APP695 cells overexpressing FE65L2 without any observed effects on APPs␣ secretion (42), suggesting that effects on full-length APP trafficking are not required for increased A␤ production. Our data suggest that enhanced ␥-secretase cleavage of C99 rather than processing of full-FIG. 6. The FE65L1-dependent reduction in APP CTF␣ levels is detected in early endosome vesicles but not in ER or Golgi vesicles. A, APP CTF␣ levels were analyzed by immunoprecipitation of pooled Golgi-rich (fractions 2-6) and ER-rich (fractions 9 -12) containing 1.2 mg of protein isolated from either uninduced (Ϫ) or induced 32-6 cells using the A8717 antibody and Western blot analysis using the 13G8 antibody as in Fig. 1C. B, early endosome vesicles from uninduced (Ϫ) or induced (ϩ) 32-6 cells were immunopurified with the EEA1 antibody, and full-length APP and APP CTF levels were examined by immunoblotting with 13G8. C99 levels were examined with the 6E10 antibody. The blots were reprobed with anti-Rab5 to verify equal protein loads. Conditioned medium from H4 cells overexpressing either wild type FE65L1 or the C702V mutant FE65L1 were analyzed for A␤ levels. Compared with uninduced control cells (-), A␤40 was significantly increased (p Ͻ 0.00000001) in H4 cells induced (ϩ) for WT FE65L1 expression. The C702V mutant FE65L1, which does not interact with APP, showed a small increase in A␤40 that was not significantly different from the uninduced control. FIG. 8. FE65L1 does not promote AICD-dependent transcriptional activation. A, results of transactivation assays obtained from luciferase activity assays. All cells were transiently transfected with the APP-GAL4, GAL4-luciferase, and pCMV-lacZ constructs previously described (3) and the FE65 plasmids shown. The luciferase activity was normalized for transfection efficiency measured by lacZ activity assays, and the data are reported as -fold increase over GAL4, which corresponds to the ratio of the normalized luciferase activity over the luciferase activity obtained with the control pMST plasmid. B, Western blot analysis of total protein (20 g) from transiently transfected cells in A using an anti-FE65L1 antibody that recognizes human (h) and rodent (r) FE65 and FE65L1. length APP in the endocytic pathway is responsible for the higher A␤40 levels measured in conditioned media of H4 cells induced for FE65L1 overexpression. This conclusion is based on our observation that FE65L1 overexpression does not produce increased levels of full-length APP in early endosome vesicles, whereas C99 levels, which were detected solely in this compartment, are decreased. Given that cleavage of APP by ␣-secretase precludes the formation of A␤, our results showing that FE65L1 modulates both C99 and full-length APP processing provides a means by which increases in A␤ levels can be observed in the presence of elevated levels of APPs␣.
Although we cannot exclude the possibility that the observed increase in A␤40 is due to increased production of C99 through effects of FE65L1 on ␤-site APP cleaving enzyme activity, we believe that the increased A␤ production is due to effects of FE65L1 on C99 trafficking or enhanced access of C99 to ␥-secretase for the following reasons. First, C99 levels detected in the early endosomes, like those of C83, are decreased by the presence of excess FE65L1, and we have demonstrated that the decrease in C83 levels is due to increased ␥-secretase cleavage of this fragment. Second, all APP CTFs have the FE65L1 binding site, and binding of FE65L1 to APP is required for both increased ␥-secretase processing of C83 and A␤40 generation. Third, we demonstrate that at least one secretase activity involved in A␤40 generation, ␥-secretase, is not directly affected by FE65L1 because NICD production from the ␥-secretase substrate Notch, which does not bind FE65L1, is not affected by FE65L1 overexpression. Although the identification of the JLK protease inhibitors, which block A␤ generation but not NICD production, suggested that the enzyme responsible for A␤ generation may be distinct from the enzyme generating NICD (43), subsequent studies using these inhibitors showed that they do not block ␥-secretase activity (44). Taken together, these data suggest that it is unlikely that FE65L1 directly alters any of the secretase activities.
The 27% increase in A␤40 secretion we observed in H4 neuroglioma cells overexpressing FE65L1 is comparable with the 40% increase reported for HEK293APP695 cells overexpressing FE65L2 using a Sandwich enzyme-linked immunosorbent assay (42). These increases in A␤40 are modest compared with the reported 4.2-fold increase in A␤ observed for Madin-Darby canine kidney APP695 cells overexpressing FE65 detected by autoradiography of secreted metabolically labeled A␤ (11). This discrepancy may in part be due to the different methods used for measurement of A␤ or to cell type-dependent differences in secretase and APP substrate subcellular localization. Our study and those of others (11,42) with the exception of one published report on FE65 (45) show that all three members of the FE65 protein family increase A␤ production.
In a separate report we demonstrated that FE65L1 overexpression in H4 neuroglioma cells reduces functional endogenous cellular levels of the LRP endocytic receptor protein (46). A reduction in LRP levels has previously been reported to influence Kunitz protease inhibitor-APP processing, the major APP isoforms in H4 neuroglioma cells (39), by increasing APPs␣ and decreasing A␤ secretion in receptor-associated protein-treated cells (47). These data raise the possibility that modulation of APP processing by FE65L1 in H4 neuroglioma cells is mediated through its effects on LRP. However, the data presented here suggest that the FE65L1-dependent decrease in LRP is unlikely to contribute to the effects of FE65L1 on APP processing because the effects of FE65L1 on APPs␣ and A␤ secretion can be blocked by abrogation of the binding of FE65L1 to APP. Furthermore, contrary to the reported effects of LRP reduction on APP processing, FE65L1 increases both APPs␣ and A␤.
Our data do not conclusively define the subcellular site for the FE65L1-dependent APP CTF processing by ␥-secretase. However, given what is known about the subcellular sites of action of the secretases, the simplest explanation accounting for all the effects of FE65L1 on APP and APP CTF processing in H4 neuroglioma cells is that FE65L1 overexpression facilitates the cleavage of APP and APP CTFs (both C83 and C99) by ␣and ␥-secretase by increasing the residence time of the substrates at or near the cell surface. In support of this notion are studies showing that constitutive ␣-secretase cleavage and biologically active ␥-secretase occur at the cell surface (48 -51).
In addition to increased production of A␤40 upon FE65L1 overexpression, we also observed FE65L1-dependent generation of AICD in pulse/chase experiments and an increase in the ratio of AICD-generated/APP CTF substrate in cell-free assays. These data indicate that FE65L1 also potentiates AICD generation in H4 cells. FE65 was previously reported to stabilize AICD in COS cells (1). The contribution of FE65L1 to AICD stabilization could not be determined from our pulse/chase experiments because the substrate for AICD, C83, is generated from full-length APP for several hours during the chase, whereas AICD production from C83 could only be detected for up to 1 h of chase, making it difficult to separate the effect of FE65L1 on generation of AICD from its possible effects on AICD stabilization. All APP CTFs are decreased as a result of FE65L1 overexpression, indicating that they all contribute to increased AICD generation. However, because C83 is the most abundant APP CTF in these cells, it is likely that AICD is predominantly derived from ␥-secretase processing of C83, as previously reported for SH-SY5Y cells (32). Although AICD generation is elevated in cells overexpressing FE65L1, unlike FE65, FE65L1 does not mediate transcriptional activation. A similar lack of transcriptional activation was recently reported for FE65L2 even though FE65L2 was detected in nuclear fractions (42). Thus, if the cellular role of the FE65 protein family is AICD-dependent transcriptional regulation, only FE65 appears to take part in transcriptional activation.
A fraction of the FE65 and FE65L1 proteins has previously been shown to be associated with membranes (10,11). In this study, we show that membrane-associated FE65L1 localizes to all membrane compartments examined. This is likely due to APP and APP CTF localization in membranes throughout the secretory and endocytic pathways.
In conclusion, the results presented here show that FE65L1 overexpression facilitates ␥-secretase cleavage of APP CTFs, resulting in increased production of A␤40 and AICD, that these events occur at or close to the cell surface either at the plasma membrane or early in the endocytic pathway, and that FE65L1, like FE65L2, does not promote AICD-dependent transcriptional activation.