Expression of β-Amyloid Precursor Protein-CD3γ Chimeras to Demonstrate the Selective Generation of Amyloid β1–40and Amyloid β1–42 Peptides within Secretory and Endocytic Compartments*

Amyloid β-protein (Aβ) is the main constituent of amyloid fibrils found in senile plaques and cerebral vessels in Alzheimer's disease (AD) and is derived by proteolysis from the β-amyloid precursor protein (APP). We have analyzed the amyloidogenic processing of APP using chimeric proteins stably transfected in Chinese hamster ovary cells. The extracellular and transmembrane domains of APP were fused to the cytoplasmic region derived from the CD3 γ chain of the T cell antigen receptor (CD3γ). CD3γ contains an endoplasmic reticulum (ER) retention motif (RKK), in the absence of which the protein is targeted to lysosomes without going through the cell surface (Letourneur, F., and Klausner, R.D. (1992)Cell 69, 1143–1157). We used the wild-type sequence of CD3γ to create an APP chimera predicted to remain in the ER (γAPPER). Deletion of the RKK motif at the C terminus directed the protein directly to the lysosomes (γAPPLYS). A third chimera was created by removing both lysosomal targeting signals in addition to RKK (γAPPΔΔ). This last construct does not contain known targeting signals and consequently accumulates at the cell surface. We show by immunofluorescence and by biochemical methods that all three APP chimeras localize to the predicted compartments within the cell, thus providing a useful model to study the processing of APP. We found that Aβ1–40 is generated in the early secretory and endocytic pathways, whereas Aβ1–42 is made mainly in the secretory pathway. More importantly, we provide evidence that, unlike in neuronal models, both ER/intermediate compartment- and endocytic-derived Aβ forms can enter the secretable pool. Finally, we directly demonstrate that lysosomal processing is not involved in the generation or secretion of either Aβ1–40 or Aβ1–42.

One of the major features of Alzheimer's disease (AD) 1 neu-ropathology is the deposition of amyloid ␤-peptide (A␤) in brain parenchyma and cerebral vessels. A␤ can be produced as a 40-amino acid peptide (A␤  ) or, occasionally, as a more amyloidogenic form of 42-43 amino acids (A␤  ). Both forms are generated by the activity of two unknown proteases termed ␤and ␥-secretase from a larger amyloid precursor protein (APP), a ubiquitously expressed type I membrane glycoprotein (1). A␤ peptide sequence begins at the extracellular domain of APP and ends within its transmembrane domain. In an alternative pathway, APP may be cleaved within the A␤ sequence by another protease termed ␣-secretase to generate a soluble ϳ100-kDa N-terminal fragment (␣APPs) and a membraneretained ϳ10-kDa C-terminal fragment (2,3). Because ␣-secretase activity takes place within the A␤ sequence, generation of intact A␤ and ␣APPs are mutually exclusive events.
The identities of ␣, ␤, and ␥ secretases are not known, and the subcellular location of their activities is currently unclear. It is generally thought that ␣-secretase cleavage occurs at the trans-Golgi network (TGN) or at a late compartment in the constitutive secretory pathway, as well as from the cell surface (4 -7). Less clear however is the mechanism and intracellular compartments involved in the production of A␤ 1-40 and A␤  . Full-length APP at the plasma membrane may be internalized to generate A␤ in an unidentified intracellular compartment. APP can also be sorted to the endosomal/lysosomal compartments where several A␤ containing C-terminal APP fragments accumulate (8 -11). It is not known whether these potentially amyloidogenic fragments are indeed A␤ intermediates or simply undergo lysosomal degradation. Recently it was shown that in transfected COS cells both A␤ 1-40 and A␤  are produced at the cell surface, although the actual sites of production were not identified (12). In contrast, in transfected neurons, A␤  appears to be produced at the TGN (6) and A␤ 1-42 at the endoplasmic reticulum (12)(13)(14).
To better understand the role of different subcellular compartments in the production of A␤  and A␤ 1-42 we constructed several APP chimeric proteins. We fused the extracellular and transmembrane domains of APP to the cytoplasmic region derived from the CD3␥ chain of the T cell antigen receptor (CD3␥). CD3␥ contains an ER retention motif (RKK), in the absence of which the protein is targeted directly from the TGN to the lysosomes without going through the cell surface by virtue of two different lysosomal signals, LL and YQ (15). We used the wild-type sequence of CD3␥ to create an APP chimera predicted to remain in the ER (␥APP ER ). A second APP chimera was constructed by removing the RKK motif, thus directing the protein directly to the lysosomes. Finally, a third chimera was created by removing both targeting signals in addition to RKK (␥APP ⌬⌬ ). This ⌬LL⌬YQ double mutation renders the cytoplas-mic tail to be devoid of known sorting signals in its sequence, and the resulting mutant protein accumulates at the cell surface at high levels due to impaired internalization (15). This approach therefore provides the opportunity to target APP to multiple intracellular compartments.
Our results suggest that, in transfected Chinese hamster ovary (CHO) cells, A␤ 1-40 is generated in the ER/IC, as well as in the endocytic pathway, whereas, as previously reported for other cell models, A␤ 1-42 is generated in the early secretory pathway, mainly in the ER/IC. More importantly, we provide evidence that, unlike in neuronal models, both ER/IC-and endocytic-derived A␤ forms can enter the secretable pool. Finally, we demonstrate that, in CHO cells, the lysosomes are not involved in the generation or secretion of either A␤ 1-40 or A␤ 1-42 .

MATERIALS AND METHODS
Construction of APP Chimeras-cDNAs encoding the chimeric APP/CD3␥ proteins were generated by oligonucleotide-directed mutagenesis with the expression vector pCI-NEO (Promega) from the parental APP 751 containing the extracellular and transmembrane but lacking the entire cytoplasmic tail (residues 734 -770; APP 770 numbering) fused to the cytoplasmic domain of the ␥ chain of the CD3 (T cell) receptor (15). Three different APP/CD3␥ chimeric constructs were used: full-length tail (Gln-116 to Lys-159) was used to generate ␥APP ER ; Gln-116 to Leu-156 (deletion of the C-terminal RKK residues) was used to generate ␥APP LYS ; and ␥APP ⌬⌬ was obtained by deleting residues L130L131 and Y137QPL140. Both the dileucine-and tyrosine-based targeting motifs were deleted in the last construct. All constructs were verified by DNA sequencing.
Cell Culture-CHO cell lines were transfected using Lipofect-AMINE TM (Life Technologies) reagent and selected by G418 resistance. The stably transfected CHO cell lines expressing wild-type APP and various APP chimeras were chosen with comparable levels of expression of the exogenous gene product. The cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with G418 (200 g/ml) and 10% fetal bovine serum at 37°C, with 5% CO 2 . All experiments using these transfected CHO cells were repeated 3-5 times, and results from either representative experiments or the mean (Ϯ S.E.) of all experiments are shown.
Antibodies-Monoclonal antibodies 5A3 and 1G7 and polyclonal antibody 863 directed to the mid-region of APP have been described (16,17). Anti-␤ tubulin monoclonal antibody was from Amersham Pharmacia Biotech and was used in Western blots at 1:5,000 dilution.
Immunofluorescence Microscopy-APP-transfected cells grown on coverslips were fixed and permeabilized in methanol for 5 min at Ϫ20°C. Following extensive washing in PBS, cells were blocked with 3% BSA in PBS (PBS/BSA) and then incubated with 5A3/1G7 (10 g/ml in PBS/BSA) for 20 min at room temperature. Cells were then extensively washed in PBS and anti-mouse IgG conjugated to fluorescein isothiocyanate (Roche Molecular Biochemicals) for 20 min at room temperature. Control samples were incubated with mouse IgG instead of primary antibody. Immunostained cells were visualized by conventional epifluorescence microscopy.
Metabolic Labeling and Immunoprecipitation-Confluent cultures of stably transfected CHO cells were incubated in methionine-free DMEM for 20 min followed by incubation with methionine-free DMEM supplemented with 250 Ci/ml [ 35 S]methionine for 15 min (pulse labeling) or 8 h with 150 Ci/ml (long labeling). In pulse-chase experiments, cells were lysed immediately after brief pulse-labeling or incubated in DMEM with 1 mM methionine (chase) for the indicated time points. APP was immunoprecipitated with antibodies 5A3/1G7 or 863 (16) and separated by SDS-polyacrylamide gel electrophoresis. Where indicated, cells were incubated for 4 h with 10 M proteasome inhibitor MG-132 (Calbiochem) or 8 h with 100 M leupeptin (Sigma) before immunoprecipitation. Gels were analyzed by phosphorimaging (Bio-Rad).
Cell Surface Biotinylation-Confluent cultures of stably transfected CHO cells were surface-biotinylated on ice using sulfosuccinimidobiotin (sulfo-NHS-biotin, Pierce, IL) as described previously (16). Cells were lysed, and immunoprecipitations were performed using antibody 863 against APP. Biotinylation of precipitated APP was analyzed by phosphorimaging (Bio-Rad) after enhanced chemiluminescence using an antibody against biotin (Jackson Laboratories). Ratios of biotinylated versus total APP were used to calculate relative amounts of cell surface APP.
Internalization Assay-Internalization of cell surface APP was per-formed as described (5). Briefly, iodinated (3-6 Ci/g) whole 1G7 monoclonal antibody was diluted in binding medium (BM) (RPMI 1640 supplemented with 20 mM Hepes ϩ 0.2% BSA) and added to triplicate cultures of CHO cells grown to confluence in 6-well tissue culture plates. Cells were incubated with radiolabeled 1G7 antibody at 37°C for 30 min, chilled on ice, and washed once with ice-cold BM. After extensive washing with ice-cold Dulbecco's PBS, 1G7 antibody bound to cell-surface APP was uncoupled by two 5-min washes with ice-cold PBS, pH 2.5 followed by cell lysis with 0.2 M NaOH. Radioactivity from both the acid washes and the cell lysates was determined in a ␥ counter. The ratio of radioactivity of acid-resistant to acid-labile fractions constitutes a measure of internalized versus cell-surface pool of APP. Measurement of A␤ by Sandwich ELISA-Sandwich ELISA was performed as described (18,19) using monoclonal antibodies specific for different species of A␤. BAN-50, specific for the N-terminal 10 amino acids of A␤, was used as capturing antibody. Horseradish peroxidaseconjugated BA-27 (A␤ 1-40 ) and horseradish peroxidase-conjugated BC-05 (A␤ 1-42 ) were used as secondary antibodies. The results are shown in Fig. 2, a and b. Essentially, no differences in the levels of accumulated "mature" and "immature" forms of APP were found between APP, ␥APP LYS , and ␥APP ⌬⌬ , indicating that at this level of analysis, all three forms are appropriately post-translationally processed through the secretory pathway. In contrast, very low levels of ␥APP ER mature forms were detected, suggesting that the majority of the newly synthesized protein is, as expected, retained in the endoplasmic reticulum and does not undergo normal maturation.

Expression and Localization of APP and APP/CD3␥ Chimeras-Stably
To further investigate the subcellular localization of the APP chimeras, we first performed immunofluorescence on methanol fixed/permeabilized cells. As expected, APP localized mainly to diffuse vesicular structures with a juxtanuclear distribution (Fig. 3), consistent with Golgi localization (20). In contrast, ␥APP ER -staining pattern showed a strong perinuclear ring ex- ␥APP ER contains the two lysosomal sorting motifs, "LL" and "YQ", as well as the "RKK" ER retention motif and is predicted to stay in the ER. ␥APP LYS lacks the motif RKK and is directed to the lysosomes from the TGN without sorting to the cell surface. ␥APP ⌬⌬ lacks both the lysosomal signals LL and YQ and the ER retention motif RKK. This chimera accumulates at the cell surface and is endocytosis-deficient. The sequence of the cytosolic tail of CD3␥ (amino acids 116 to 159) is represented at the bottom with the motifs LL and YQ in bold.
tending into fine reticular structures, a distribution characteristic of ER resident proteins. ␥APP LYS distribution partially overlapped with that of APP at a juxtanuclear region due to the presence of both proteins in the biosynthetic pathway, but also in larger vesicular structures in the periphery. These latter vesicular structures were more intensely stained after leupeptin treatment (not shown), suggesting that ␥APP LYS does indeed localize to the lysosomes. Finally, ␥APP ⌬⌬ was found at the cell surface (arrowheads, ␥APP ⌬⌬ panel) in addition to the juxtanuclear staining also present in both APP and ␥APP LYS . Double staining experiments also showed that ␥APP ER colocalized with calnexin, whereas ␥APP LYS co-localized with the lysosomal markers LAMP-1 and LAMP-2 (not shown).
Biochemical Characterization of APP/CD3␥ Chimeras-Because accurate targeting of APP chimeras to different subcellular locations is essential in this study, additional biochemical approaches were used to further define the subcellular localization of the three APP/CD3␥ chimeras. Recently, it became clear that selective degradation of ER membrane proteins occurs mainly through the ubiquitin-proteasome pathway (21,22). Accordingly, if ␥APP ER is an ER resident protein, it should accumulate after treatment with MG-132, a specific proteasome inhibitor that has no apparent effect on lysosomal proteases (23). As shown in Fig. 4, there was no effect on APP (Fig.  4b), ␥APP LYS or ␥APP ⌬⌬ (not shown), but ␥APP ER increased substantially after MG-132 treatment, indicating that the latter chimera is retained in the ER.
To further confirm that the vesicular staining seen for ␥APP LYS corresponds to the lysosomes, the transfected cells were treated with leupeptin to inhibit lysosomal proteases. As presented in Fig. 4a, leupeptin treatment results in accumulation of ␥APP LYS , but not APP (Fig. 4a) or the other ␥APP chimeras (not shown). A control immunoblot for ␤-tubulin (Fig. 4a) shows that the modest increase in ␥APP LYS after leupeptin treatment is not due to differences in the amount of protein loading.
In the absence of targeting signals, ␥APP ⌬⌬ is predicted to be mainly cell surface distributed and deficient in endocytosis (15). Consequently, cell surface accumulation of ␥APP ⌬⌬ was further examined by surface biotinylation, followed by lysis and immunoprecipitation of total cellular APP. Surface and total amounts of APP were estimated by dividing the amount of reactivity of the biotinylated versus total pools of APP (see "Materials and Methods"). The results showed that cell surface levels of ␥APP ⌬⌬ were ϳ2.5-fold higher than those seen in APP At the end of the incubation period, equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis, and the levels of APP were analyzed by Western blotting. A modest accumulation of ␥APP LYS , but not APP, is apparent after leupeptin treatment. Parallel samples were blotted for ␤-tubulin to confirm protein loading. Panel b, proteasome inhibitor MG-132 (10 M) was added to cells for 4 h, and levels of APP were analyzed as above. The strong accumulation of ␥APP ER after MG-132 indicates that its predominant degradation route is, as expected, in the ER. Results from a representative experiment are shown. (Fig. 5), a value comparable with that seen in a C-terminally truncated APP (5). The levels of cell surface ␥APP ER were essentially undetectable, whereas ␥APP LYS was too low for accurate quantification, estimated to be approximately onetenth of the amount of wild-type APP. The lack of cell surface molecules in ␥APP ER and ␥APP LYS forms provided further confirmation of the intracellular localization of these chimeric proteins. Lastly, cell surface accumulation of ␥APP ⌬⌬ is correlated with a marked decrease in endocytosis, as measured by uptake of radioiodinated 1G7 antibody (5) (not shown). Moreover, secretion of sAPP (Fig. 6) was increased ϳ2.5-fold, a value also similar to what has previously been described for C-terminally truncated mutant APP (10). Therefore, these results show that the APP-CD3␥ chimeras are appropriately targeted to their predicted subcellular locations.
Levels of A␤  and A␤  in CD3␥/APP Chimeras-We next examined the levels of A␤ 1-40 and A␤ 1-42 in cell lysates and conditioned media from cells stably transfected with APP or ␥APP chimeras using sandwich ELISA (see "Materials and Methods"). In comparison with APP, all ␥APP chimeras showed differences in the levels of A␤ 1-40 and A␤ 1-42 , although in different ways. First, with respect to APP, there was a dramatic decrease in the secreted and intracellular pools of A␤  in ␥APP LYS -and ␥APP ⌬⌬ -expressing cells (Fig. 7a). Because all three proteins presumably share similar processing steps through the secretory pathway up to the TGN, the differences in A␤ 1-40 production are likely to originate late in the secretory pathway (i.e. a post-TGN compartment) and/or the endocytic pathway. Neither ␥APP LYS nor ␥APP ⌬⌬ undergo efficient endocytosis, because in the former construct, the molecules were sorted away from the cell surface levels (Fig. 5) and in the latter construct, internalization was substantially impaired. Taken together, these results suggest that the endocytic pathway is a major site of production and subsequent release of A␤ 1-40 (ϳ60 -70% of the total detected), the remaining presumably originated in the early secretory pathway (i.e. ER, Golgi, and/or TGN, the three compartments shared by both chimeras).
In contrast, however, levels of A␤ 1-40 from ␥APP ER -expressing cells showed different changes with respect to wild-type APP. Specifically, although intracellular levels of A␤ 1-40 were almost 75% higher, secreted levels were approximately onehalf those of APP cells (Fig. 7a). This suggested that only a portion of intracellularly generated A␤ 1-40 molecules are released. Moreover, because ␥APP ER is retained in the ER and endocytic processing is largely absent, the high levels of intracellular A␤ 1-40 are unexpected. We hypothesized that the accumulation of A␤ 1-40 in ␥APP ER transfected cells derives from its retention in the ER, where there is an enhancement of substrate available for ␥-secretase activity. To test this hypothesis, we measured the half-life of APP and the three ␥APP chimeras by pulse-chase labeling. Indeed, both ␥APP ⌬⌬ and ␥APP ER showed markedly longer half-lives than wild-type APP (135.1 Ϯ 3.38 min and 86.6 Ϯ 4.3 min, respectively, versus 53.3 Ϯ 6.9 min, n ϭ 3, p Ͻ 0.001, analysis of variance) (Table I).
In contrast to A␤ 1-40 levels, A␤ 1-42 levels (Fig. 7b) were surprisingly unchanged in both ␥APP LYS -and ␥APP ⌬⌬ -expressing cells as compared with wild-type APP. This was seen in both intracellular and secreted pools of A␤ 1-42 , a finding that suggests that even the secreted pool of A␤ 1-42 was generated predominantly in the ER/IC. ␥APP ER behaved very differently from the other three constructs. As with A␤ 1-40 , the intracellular levels of A␤ 1-42 were substantially higher in ␥APP ER cells as compared with wildtype APP and the other two APP/CD3␥ chimeras (Fig. 7b). In contrast to A␤ 1-40 , even the secreted levels of A␤ 1-42 were also higher. In summary, with respect to wild-type APP, levels of A␤ 1-42 from ␥APP ER -expressing cells were increased ϳ10and ϳ4-fold in the intracellular and secreted pools, respectively.
Are Lysosomes Involved in the Generation of A␤?-Lysosomes have been hypothesized as a possible site of A␤ production. Several studies have indirectly addressed this question, but, to date, the role of the lysosomes in A␤ production is still unclear (24 -26). Efficient targeting of APP to lysosomes en- abled us to directly address this question. Accordingly, production of A␤ 1-40 did not increase when APP was directly sorted to the lysosomes from the TGN, indicating that lysosomes are not involved in A␤ generation (Fig. 7a). The same can be concluded of A␤  . Because no differences are found between APP, ␥APP LYS , and ␥APP ⌬⌬ (Fig. 7b), the lysosomal compartment is unlikely to be a major source of either A␤ 1-40 and A␤ 1-42 generation.
A␤ 1-42 /A␤  Ratios Increase in All ␥APP Chimeras-Combining all the A␤ results of our study (Fig. 7), we have analyzed the relative ratios of A␤ 1-42 /A␤ 1-40 expressed in CHO cells. Extracellularly, A␤ 1-42 is consistently a minor fraction of total A␤, and wild-type APP showed the expected ϳ1:10 ratio of A␤ 1-42 to A␤  . The intracellular ratio is ϳ6:10, a value similar to that reported in neurons (13). However, there was an increase in the 42:40 ratios in both intracellular (between 2-3:1) and secreted (ϳ4:10) pools for all ␥APP chimeras. And despite the differences in absolute levels of each A␤ species, the relative levels are surprisingly similar to each other. Intracellularly, this was achieved by either a large increase in A␤ 1-42 (␥APP ER ) (although attenuated by a slight increase in A␤ 1-40 ) or by a decrease in A␤ 1-40 (␥APP LYS and ␥APP ⌬⌬ ). A similar, although more modest trend, is seen extracellularly. The secreted 42:40 ratio increase was achieved by either a decrease of A␤ 1-40 and an increase in A␤ 1-42 (␥APP ER ) or by a decrease in A␤ 1-40 secretion (␥APP LYS and ␥APP ⌬⌬ ). DISCUSSION A␤ is the major component of senile plaques in the Alzheimer's brain, and it is thought to play an important role in the pathogenesis of the disease (1). Consequently, much effort has been dedicated to the study of A␤ generation in a variety of models. Here, we have chosen to generate several chimeric APP proteins predicted to localize to particular organelles, to study the role of such subcellular locations in the generation and subsequent secretion of A␤ 1-40 and A␤  . Although several recent reports have successfully used chimeric APP proteins to study the proteolytic processing of APP (27)(28)(29), our study is the first to use a common approach to target APP to multiple organelles and directly analyze the intracellular formation and secretion of A␤ 1-40 and A␤  . The strategy that we followed for APP subcellular targeting is based on a well characterized model, that of the ␥ chain of the CD3 receptor ( Fig. 1) (15). Because the targeting of the ␥APP chimeras to specific organelles is central to our approach, we ascertained by morphological and biochemical studies the predicted localization of the chimeric proteins. Accordingly, several lines of evidence demonstrated that the targeting motifs from ␥CD3 are fully functional when fused to the APP transmembrane and extracellular domains, thereby providing a valuable tool for the analysis of A␤ formation in different subcellular organelles. Thus, we showed that ␥APP ER is localized to the ER with impaired post-translational processing and is turned over more slowly. In the case of ␥APP LYS, we showed that this chimeric protein is predominantly sorted to lysosomes, bypassing the cell surface, after maturation. Finally, we demonstrated that in the absence of the two targeting motifs, ␥APP ⌬⌬ accumulates at the cell surface and, not surprisingly, showed impaired internalization.
As expected, a consequence of the latter abnormality is an increase in APPs in the medium (16).
Recently, it has become clear that the sites of generation and the rates of secretion of different A␤ forms are complex and may be cell type dependent. Specifically, in neuronal cells, A␤ 1-42 is produced in the ER/IC compartment (12,13,30), whereas A␤ 1-40 is apparently derived from the TGN or beyond (12). In contrast, APP 695 transfected COS-7 cells generate both A␤ forms at the plasma membrane (12) and are undetectable intracellularly. A nonneuronal cell type, kidney 293 cells, when stably transfected with APP, shows detectable levels of intracellular A␤ 1-42 but not A␤   (31). Therefore, our analysis is important from the standpoint that the different intracellular processing pathways can be simultaneously analyzed with respect to both A␤ 1-40 and A␤  . Our results indicate that in CHO cells (1) both secretory and endocytic processing contribute to the production of A␤ 1-40 , whereas A␤ 1-42 derives mainly from the secretory pathway (2) both endocytic and, to a lesser extent, ER/IC-derived A␤ forms can enter the secretable pool, and (3) lysosomes are not a major site of A␤ generation.
Our studies showed clearly that both A␤ 1-40 and A␤ 1-42 species are readily detectable in APP-transfected CHO cells and that the secretory pathway plays a role in the generation of both forms (Fig. 7, a and b). This is apparent from the fact that both A␤ 1-40 and A␤ 1-42 are increased when APP is artificially retained in the ER (␥APP ER ) via the presence of a retention signal engineered into the cytoplasmic domain. However, we reasoned that the majority of A␤ 1-40 (ϳ70 -75%) originates from endocytic processing because of the loss of A␤ 1-40 in the ␥APP LYS and ␥APP ⌬⌬ cells as compared with wild-type APP cells (Fig. 7a). Furthermore, the fact that retention of APP in the ER results in accumulation of intracellular A␤   (Fig. 7a, compare ␥APP ER with APP) indicates that the ER itself contains an A␤ 1-40 -specific ␥-secretase activity. This is a surprising finding, because A␤ 1-40 -specific ␥-secretase activity has only been reported previously at the TGN or at the plasma membrane. One explanation of our finding is that this population of ER-derived A␤ 1-40 is generated from secondary cleavage of A␤ 1-42 , a postulate that is consistent with evidence arguing distinct A␤ 1-40 and A␤ 1-42 ␥-secretase activities (32).
In contrast, consistent with published reports, A␤ 1-42 may be derived primarily from processing in the early secretory pathway (12,13). As shown in Fig. 7b, APP and both ␥APP LYS and ␥APP ⌬⌬ produce and secrete comparable amounts of A␤ 1-42, indicating that a main site of production is in a compartment common to all three APP forms, i.e. ER, Golgi, or the TGN. Again, the fact that retention of APP in the ER causes intracellular accumulation of A␤ 1-42 points to that organelle as a major site for A␤ 1-42 production. However, our results cannot rule out the late secretory and/or the endocytic pathways as additional sites of production for A␤  . Specifically, because the ␥APP chimeras have very different cytosolic tails, direct comparison to APP may be misleading. This is evident from the fact that all three ␥APP chimeras show substantially higher A␤ 1-42 /A␤ 1-40 ratios. Therefore, the possibility remains that the decrease seen in A␤ 1-42 production from ␥APP LYS and ␥APP ⌬⌬ with respect to ␥APP ER , rather than to wild-type APP, may be due to the deficient endocytosis from the former chimeras and not only to the accumulation of ␥APP ER in the ER. This concept would be consistent with the observation that secretion of both A␤ 1-40 and A␤ 1-42 is diminished when the endocytic signal is removed from the cytoplasmic domain of APP (33). We should emphasize that retaining APP in the ER is equally artificial, and we cannot ascertain the degree to which A␤ 1-42 generation has been abnormally increased. Furthermore, the normal interactions with proteins that associate with the APP cytoplasmic domain are absent in the chimeras. Thus, the Fe65/X11 family of proteins that bind to APP at or near the YENPTY domain and alter APP trafficking, translocation to the plasma membrane, sAPP secretion, and A␤ production (34,35) are ineffectual in the APP chimeric molecules. Interestingly, overexpression of X11 as well as the Y743A mutation (in the NPTY motif) both decreased the turnover of APP but with different effects on A␤ generation (33,35). Although ␥APP ER also increased the APP half-life, direct comparison between these three conditions is not possible. First, as mentioned above, ␥APP ER contains a completely different cytoplasmic domain. Second, the other studies did not examine the levels of intracellular A␤. Third, the mechanisms underlying the reduced turnover of APP are different in all three cases: in ␥APP ER , the molecule is postulated to remain in the ER; in APP Y743A mutant, a lysosomal targeting signal may be lost (33); and in X11, the effect is presumably due to delayed protein maturation (35).
Our results also showed that not all intracellular A␤ 1-40 and A␤ 1-42 is released. It has been argued that in neurons, the secreted and the intracellular pools of A␤ are produced independently and that the secreted pool is endocytosis-dependent (12,36). Subsequent studies (13) showed that the intracellular pool of A␤ 1-42 derives from the ER and does not enter the secreted pool. This was shown by insertion of a KK ER-retention motif to the cytosolic tail of APP (13), an approach similar to the ␥APP ER construct, and by BFA treatment of wild-type APP-expressing cells, although protein traffic distal to the ER is inhibited, and therefore is more appropriate to study A␤ production than A␤ secretion. In our study, we showed that secretion was impaired but not abrogated for both A␤ 1-40 and A␤ 1-42 from ␥APP ER . Note that, although there is an increase in secretion of A␤ 1-40 from ␥APP ER when compared with ␥APP LYS and ␥APP ⌬⌬ , there is a decrease with respect to wildtype APP, indicating that accumulation of A␤ 1-40 in the ER results in its impaired secretion. Similarly, A␤ 1-42 from ␥APP ER accumulates in the ER in much higher proportions than it can be released (10-fold accumulation versus 3-fold secretion increase when compared with wild-type APP). These findings are of interest, because it has been proposed that generation of A␤ 1-42 in the ER, at least in neurons, is a saturable process (13). In other words, it was suggested that retention of APP in the ER in neurons does not result in an increase in A␤ 1-42 levels and, importantly, secretion of A␤ 1-42 is abrogated. In contrast, we show that in our model both A␤ 1-40 and A␤  can accumulate in the ER, indicating that saturation seems to be reached at the secretion rather than at the level of production.
Finally, we presented evidence that the lysosomes are not the site of production of either A␤ 1-40 or A␤  . Lysosomes contain APP degradation products, especially the C-terminal fragments that may be direct precursors to A␤. Although there is indirect evidence suggesting that an acidic compartment, such as the lysosomes, is necessary for A␤ production, attempts to isolate A␤ from endosomes/lysosomes fractions have not been successful (24). Therefore, the role of the lysosomes in A␤ production remains poorly defined. This issue is both important and perplexing in view of the disruptions of the endosomal/lysosomal system that accompany AD (37,38). We addressed this question by targeting APP to the lysosomes and directly measuring A␤ 1-40 and A␤ 1-42 production. Our results showed that in CHO cells the lysosomes are not likely to represent a major site of production for either A␤ 1-40 or A␤ 1-42 .
In summary, we have used three different APP chimeric proteins to study the intracellular amyloidogenic processing of APP and the subsequent release of both A␤ 1-40 and A␤ 1-42 into the medium in cultured CHO cells. We found that secretory and endocytic processing contribute to different degrees to the production and release of both A␤ 1-40 and A␤  . Moreover, our results show that the production and secretion pathways may be substantially more complex than previously thought.