p-Amyloid Peptide and a 3-kDa Fragment Are Derived by Distinct Cellular Mechanisms*

We have analyzed the cellular processing pathways which produce the 4-kDa amyloid @-peptide (A@) and a 3-kDa derivative (p3) of the @-amyloid precursor protein (BAPP) found in conditioned media of tissue culture cells and in cerebrospinal fluid. Pulse-chase experiments reveal that both peptides are secreted in parallel with soluble @APP (APP.); no precursor-prod-uct relation between AB and p3 was found. The pro- tease inhibitor leupeptin did not influence the production of either peptide. In contrast, the weak base am- monium chloride (NH4Cl) showed a dose-dependent inhibition of A@ production with less decrease in p3. A similar effect was observed using the monovalent ion- ophore monensin. Brefeldin A completely inhibited the generation of both peptides, indicating that proteases located in the endoplasmic reticulum or early Golgi are not sufficient for the production of the small peptides. Deletion of the @APP cytoplasmic domain, which re-moves a consensus sequence that probably mediates reinternalization, caused an increase in secretion of both APP. and p3 and did not abolish A@ production. These observations suggest that completely mature BAPP within the late Golgi and/or at the cell surface is a prerequisite for AB production but processing within the lysosome might not be directly required. p3 appears to derive from the 10-kDa N’- N’ synthetic AP”40. tates AD, amounts a PAPP C-terminal Deletion Construct and Trans-fection-A construct designed to

isoforms of PAPP have been described which are all derived by alternative splicing. In addition to the 695-amino acid form (Kang et al., 1987), three other major forms have been reported which contain an additional exon encoding a protease inhibitory domain (@APP 563,751,and 770) (Kitaguchi et al., 1988;Ponte et al., 1988;Tanzi et al., 1988;DeSauvage and Octave, 1989). Upon maturation of pAPP within the endoplasmic reticulum (ER) and Golgi, the precursor is cleaved by an as yet unidentified protease designated bAPP secretase to create the secreted form of pAPP (APP,) and a 10-kDa Cterminal fragment that remains membrane-bound (Weidemann et al., 1989;Oltersdorf et al., 1990;Esch et al., 1990;Wang et al., 1991). Because this cleavage occurs within the A@ domain, this processing pathway inhibits the formation of AP. In contrast, some pAPP molecules are reinternalized from the cell surface and targeted to late endosomes/lysosomes (Haass et al., 1992a), where Ab-containing C-terminal fragments of DAPP accumulate (Golde et al., 1992;Estus et al., 1992;Haass et al., 1992a). These fragments could potentially give rise to the formation of A@. Recently, we and others found that AB is normally present in the media of cultured cells (Haass et al., 1992b;Shoji et al., 1992) and cerebrospinal fluid (Seubert et al., 1992;Shoji et al., 1992), indicating that the production and release of Ab is a normal physiological event. In addition to the 4-kDa A@, we detected a 3-kDa peptide (p3) corresponding to a truncated fragment of Ab as well as a number of minor A@-related peptides (Haass et al. 1992b). To characterize the cellular pathways that produce the two peptides, we analyzed the formation of the peptides in pulse-chase experiments and after treatment of cultured cells with a variety of agents that interfere with cellular processing pathways. In addition, the effect of a C-terminal deletion of pAPP on the formation of A@ and p3 was studied.

EXPERIMENTAL PROCEDURES
Drug Treatments-Colchicine was made as a 1 mM stock solution in media. Monensin was made as a 10 mM stock solution in ethanol. Leupeptin was used as described previously (Haass et al., 1992a). NH4CI was added from a 5 M stock solution. Brefeldin A (BFA) was added from a 5 mg/ml stock solution in ethanol. For control experiments in the absence of drugs, the appropriate carrier was added. All drugs and carriers were diluted into the media before the mixture was applied to the cells. Human embryonic kidney 293 cells stably transfected with (3APP 695 (Selkoe et al., 1988) were incubated during a 16-h labeling period in methionine-free media containing 10% fetal calf serum with the corresponding drugs (colchicine, leupeptin, monensin, NH4CI). Identical results were obtained during a 3-h pulse label in the presence of the corresponding drug. Experiments using BFA were carried out only in a 3-h pulse-labeling experiment. All experiments were repeated three to nine times. Metabolic labeling and immunoprecipitations from cell extracts and media were performed as described (Haass et al., 1991(Haass et al., , 1992b. Immunoprecipitated APP. was separated on a 10% SDS-polyacrylamide gel, whereas A@ and p3 were separated on a 10-20% Tris-Tricine gel (Haass et al., 1992b). Autofluorography was carried out as described (Haass et al., 1991). The inhibitory effect of drugs on the formation of A(3 and p3 was quantified by densitometry. Antibodies Used for Immunoprecipitation-The polyclonal antibody C7 (Podlisny et al., 1991) is directed against the last 20 amino acids of the cytoplasmic tail of (3APP. This antibody immunoprecipitates N'-and N' plus 0'-glycosylated full-length (3APP, the 10-kDa and a variety of potentially amyloidogenic C-terminal fragments (Haass et al., 1992a). The affinity-purified polyclonal antibody B5 protein of (3APP44"592 (numbering of (3APP 695; Kang et al. (1987)) (Oltersdorf et al., 1990) was raised to a recombinantly expressed 3021 and immunoprecipitates APP, and N'and N' plus 0'-glycosylated full-length PAPP. The polyclonal antibody R1280 (Tamaoka et al., 1992) was raised to synthetic AP"40. This antibody immunoprecipitates AD, p3, and small amounts of APP, from media of tissue-culture cells (Haass et al., 1992b).
Construction of a PAPP C-terminal Deletion Construct and Transfection-A cDNA construct was designed to encode a truncated species of PAPP by introducing a stop codon after amino acid 653 of PAPP 695. The stop codon was introduced by using the oligonucleotide CCTCTAGACTAGTACTGTTTCTTCTT (underlined sequence = XbaI site; bold letters = STOP codon) as a 3' primer and the 5' oligonucleotide GATGCAGAATTCCGACAT (underlined sequence = EcoRI site) as a primer for a polymerase chain reaction (PCR). The PCR products were digested with EcoRI and XbaI and subcloned into the EcoRIISpeI-linearized CMV695 plasmid (Selkoe et al., 1988) creating CMV695AC. The sequence was confirmed by sequencing both strands of the PCR product. Transient transfections into 293 cells were performed using Lipofectin (GIBCO/BRL) as described by the manufacturer.

RESULTS AND DISCUSSION
As a model system to analyze the cellular pathways which result in the production of AB and p3, we used human kidney 293 cells stably transfected with a BAPP 695 cDNA. These cells are known to process BAPP in a manner similar to primary human endothelial cells (Haass et al., 1992a). I t should be noted that AB and p3 are normally produced by a variety of primary cells and untransfected cell lines, including kidney 293 cells (Seubert et al., 1992;Haass et al., 1992b), thus excluding the possibility that these peptides are produced by aberrant processing due to BAPP overexpression in transfected cells.
T o examine the release of AB and p3, 293 cells stably transfected with a BAPP 695 cDNA (Selkoe et al., 1988) were pulse-labeled with [%]methionine and chased in the presence of excess unlabeled methionine for various times. AB and p3 were immunoprecipitated from the conditioned media using R1280, a high titer antibody against synthetic (Tamaoka et d., 1992). As a control, APP, was immunoprecipitated from equal aliquots of the same samples using antibody B5 to recombinant BAPP444-5y2 (Oltersdorf et al., 1990). AB and p3 were generated at rates similar to that of APP, (Fig.  1). The concomitant appearance of AB and p3 suggests that no precursor-product relationship exists between the two peptides. This conclusion is supported by the observation that even after a 2-h label-free chase no decrease of AB or increase of p3 was observed, indicating that AB was not converted to  (lanes 1-5), and the remaining supernatants were immunoprecipitated with antibody R1280 (lanes 6-10). p3 (Fig. 1). The results suggest that the two peptides are formed via parallel mechanisms from different BAPP molecules or degradative intermediates. Furthermore, the data suggest that p3 could be derived from the 10-kDa C-terminal fragment following secretory cleavage of APP,, and not from AB. This possibility is supported by our previous radiosequencing of p3, which showed that its N terminus is either at the secretase cleavage site or 1 amino acid C-terminal to this site (Haass et al., 1992b). Nevertheless, our experiments do not rule out the possibility that AB and p3 are both derived from one common precursor, e.g. a -100-amino acid-long Cterminal fragment that can be differentially cleaved.
T o analyze the cellular mechanisms involved in the formation of the two peptides in greater detail, we studied the effect of a variety of drugs on the production of AB and p3 by 293 cells stably transfected with BAPP 695. Since microtubule depolymerizing agents are known to inhibit transport vesicles from fusing with prelysosomal compartments (Kelly, 1990), we analyzed the formation of AD and p3 upon treatment of cells with colchicine and nocodazole. Colchicine (Fig. 2 A ) and nocodazole (data not shown) had no significant effect on the production of AB and p3, despite the fact that immunocytochemical experiments using anti-tubulin antibodies revealed a complete depolymerization of the microtubular cytoskeleton (data not shown).
T o determine whether lysosomal proteases may be involved in the generation of AB and p3, 293 cells were treated with leupeptin. Increasing amounts of leupeptin had no influence on the generation of both peptides (Fig. 2B), despite the fact that this drug causes a substantial accumulation of AB-containing C-terminal fragments of DAPP in the lysosomes of these cells (Haass et al., 1992a). In contrast, increasing amounts of NH,Cl clearly inhibited the formation of AB (-78% at the maximum concentration) with less inhibition of p3 formation (-49% at the maximum concentration; Fig.  2C), indicating that an acidic cellular compartment is involved in the generation of AD. The monovalent ionophore monensin, which is known to abolish H', Na', and K' gradients and thus inhibit late Golgi and lysosomal functions (Tartakoff, 1983), had a strong dose-dependent inhibitory effect on the formation of A@ (-96% at the maximum concentration), with less effect on p3 formation (-67% at the maximum concentration; Fig. 20). In parallel, monensin also down-regulated APP, secretion (Fig. 2 0 ) a n d inhibited the maturation of intracellular PAPP, resulting in a marked accumulation of an incompletely mature form of BAPP within the cell (Fig. 2E) and low level secretion of incompletely mature APP. (see down-shift in apparent molecular mass of APP, upon monensin treatment in Fig. 20). The small amount of C-terminal 10-kDa fragment seemed to be stabilized by monensin (Fig.  2E), suggesting an inhibition of the lysosomal degradation of this fragment. The effect of monensin on the maturation of intracellular BAPP is similar to that reported by Caporaso et al. (1992). The drug experiments described above were carried out by incubating the cells with drugs during a 16-h period of metabolic labeling. Repetition using a 3-h pulse-labeling experiment gave essentially the same results (data not shown). The latter experiments rule against the possibility that a 16h treatment with these drugs leads to reduced viability of the tissue culture cells.
The possibility of AB formation within the ER or early Golgi was excluded by the use of brefeldin A (BFA), which causes a redistribution of Golgi into ER (Pelham, 1991). Under these conditions, AD and p3 could not be detected either in the media (Fig. 2F) or inside the cell. As expected, maturation of full-length BAPP was inhibited, resulting in an  (Haass et al., 1992b).
incompletely glycosylated BAPP molecule (Fig. 2G). These experiments also exclude the possibility that AB might be formed within the cytoplasm after incomplete translocation of BAPP into the ER, since such a process would be unaffected by BFA.
Interestingly, we could not detect any intracellular AB in the extracts of the cells treated with each of the different drugs described above (data not shown), a finding consistent with previous data (Shoji et al., 1992;Haass et al., 1992b).
To determine the influence of the cytoplasmic domain of BAPP, which contains a potential consensus sequence for coated pit-mediated reinternalization of cell surface proteins (NPXY; Chen et al. (1990)), on the formation of AB and p3, a stop codon was inserted after position 653 (numbering according to BAPP 695; Kang et al. (1987)), giving rise to Cterminal truncated BAPP. This construct (named CMV695AC) and the wild type cDNA (CMV695) were transfected into 293 cells, followed by metabolic labeling and immunoprecipitation by R1280 and B5 (Fig. 3). Immunopre- cipitation of APP, by antibody B5 revealed that deletion of the cytoplasmic tail resulted in increased secretion of APP, (Fig. 3A). This observation may be explained by the fact that the truncated protein has been shown to undergo a reduced rate of reinternalization from the cell surface,' thus providing more substrate to BAPP secretase, which has been reported to cleave BAPP at least in part on the cell surface (Sisodia, 1992;Haass et al., 1992a). Such an increase of APP. secretion following the deletion of the C terminus has also been mentioned by Sisodia (1992). This increase in APPs is paralleled by increased release of p3 (Fig. 3B). This is consistent with the hypothesis that p3 is derived from the C-terminal fragment of BAPP created by secretase, because enhanced cleavage of full-length BAPP should create more of the 10-kDa Cterminal fragment. Furthermore, in the absence of the C terminus, AB is still produced, although in smaller amounts. Interestingly, radiosequencing of this 4-kDa band demonstrated that, in addition to AD, significantly increased amounts of 4-kDa AB-related peptides were observed. These peptides began at Val (-3) and Arg (5), relative to the AB sequence (data not shown). These data show that an intact cytoplasmic domain of BAPP is not absolutely necessary for the formation and release of either p3 or AB or AB-related peptides. In addition, these experiments strongly suggest that p3 and AB are derived from different precursor molecules, and that p3 does not derive simply by proteolytic cleavage of AB at the secretase site, since the amounts of p3 and AB do not change in parallel.
Our data indicate that an acidic compartment is involved at some point in the formation of AB and also in part in the formation of p3. This is supported by the fact that NH4Cl and monensin reduce the amounts of both peptides. Furthermore, the pulse-chase experiment together with the results of monensin treatment suggest that only completely mature BAPP can give rise to the formation of AB. The production of AB and p3 in the ER or early Golgi is excluded by the use of BFA. Shoji et al. (1992) have postulated that lysosomes may * E. Koo, personal communication.
be involved in the generation of AB. This is based on the observation that NH,Cl decreases the amounts of AB. However, leupeptin did not have any effect on the formation of AB (Shoji et al., 1992). The fact that leupeptin and colchicine/ nocodazole do not inhibit the formation of both peptides in the present experiments indicates that lysosomes may not be directly involved in the proteolytic processing that results in the formation of either p3 or AB. This conclusion is further supported by the fact that we were unable to detect the two peptides within isolated lysosomes purified from metabolically labeled 293 cells (data not shown). Furthermore, leupeptin strongly increases the amounts of the 10-kDa C-terminal fragment and larger potentially amyloidogenic fragments within the lysosome (Caporaso et al., 1992;Golde et al., 1992;Haass et al., 1992a). Despite this accumulation of very high amounts of the 10-kDa C-terminal fragment and slightly larger AB-containing fragments in the lysosome, we do not observe higher amounts of p3 or AB in the media. We hypothesize that only prelysosomal 10-kDa fragments (most likely on or near the cell surface) and larger AB-containing precursor fragments can give rise to p3 and AB, respectively. One could argue that leupeptin treatment could inhibit the formation of both peptides within the lysosome, but this should result in reduced amounts of the two peptides upon leupeptin treatment, a finding which is clearly not the case (see Fig. 2B).
The hypothesis that lysosomal degradation of BAPP is not an absolute prerequisite for AB production is further supported by the finding that cultured fibroblasts from patients with I-cell disease, which are unable to target proteases directly to lysosomes, still produce Our data do not exclude the possibility that full-length BAPP or AB-bearing fragments thereof might be reinternalized from the cell surface, giving rise to AB and p3 within an early endocytic vesicle that subsequently recycles to the cell surface and releases the peptides into the medium. As a working model, we propose the possibility that the cleavage producing the N terminus of AB may be mediated by alternative cleavage of mature BAPP by the secretase or a closely related enzyme within late Golgi and/or cell surface. This event could take place either on the cell surface or within C. Haass, T. Oltersdorf, I. Lieberburg, and D. Selkoe, unpublished data.
the acidic compartments of late Golgi or transport vesicles derived from a reinternalization pathway carrying BAPP or fragments thereof. Indeed, evidence has recently been presented that alternative secretory cleavage of BAPP may potentially create the N terminus of AB (Seubert et al., 1993).