Differential Effects of the Swedish Mutant Amyloid Precursor Protein on β-Amyloid Accumulation and Secretion in Neurons and Nonneuronal Cells*

Expression of the Swedish ΔNL mutation in the β-amyloid precursor protein (APPΔNL) dramatically increases Aβ generation in nonneuronal cell lines, although it is unclear whether intracellular levels of β-amyloid (Aβ) are also elevated after APPΔNL expression. Furthermore, the effects of expressing APPΔNL in neurons on the production and secretion of Aβ-(1–40) and Aβ-(1–42) are unknown. To address these issues, we examined the generation of both intracellular and secreted Aβ-(1–40) and Aβ-(1–42) in human neuronal NT2N cells, in primary rat astrocytes, and in Chinese hamster ovary cells engineered to express wild-type APP or APPΔNL using a recombinant Semliki Forest virus expression system. Expression of APPΔNL led to a marked increase in APPβ and the C-terminal fragment containing the entire Aβ sequence (C99) in all cells tested. However, a dramatic elevation of intracellular and secreted Aβ-(1–40) and Aβ-(1–42) was seen only in astrocytes and Chinese hamster ovary cells. The ΔNL mutation did not cause a significant increase in intracellular or secreted Aβ-(1–40) or Aβ-(1–42) in NT2N cells. Since NT2N cells expressing APPΔNL accumulate much higher levels of C99 than cells expressing wild-type APP, we conclude that the rate-limiting step in Aβ production could be the further processing of C99 by γ-secretase in these cells. These results show that the Swedish ΔNL mutation causes nonneuronal cells to process APP via pathways more in common with the metabolism of wild-type APP in neurons.

The 4-kDa amyloid ␤ peptide (A␤) 1 is the principal protein-aceous component of senile plaques, the hallmark pathological feature of Alzheimer's disease (AD). The A␤ peptide varies in length from 39 to 43 amino acids (1)(2)(3)(4)(5) and is derived from post-translational cleavage of the amyloid precursor protein (APP) (6 -9). A variety of proteolytic pathways have been described for the processing of APP, but not all of them result in the production of A␤ (reviewed in Refs. 10 and 11). For example, a portion of APP is processed by the ␣-secretase pathway in which APP is cleaved within the A␤ region at or near the plasma membrane, releasing a large N-terminal ectodomain fragment (APP␣) (12,13), thereby precluding the formation of full-length A␤. The utilization of this pathway appears to be preferred by transfected nonneuronal cells expressing wildtype APP (APPwt), since the N-terminal ectodomain containing the first 17 amino acids of A␤ (i.e. APP␣) and p3, the A␤ fragment beginning at amino acid residue 17 of A␤, are recovered at high levels from media conditioned by these transfected nonneuronal cells (13,14).
Processing pathways that result in the constitutive production of A␤ have also been identified, although their utilization is relatively minor in nonneuronal cells (13,15,16). Cleavage of APP by ␤-secretase at the N terminus of the A␤ sequence releases a soluble N-terminal fragment (APP␤) and generates a C-terminal fragment (C99) that contains the entire A␤ sequence. C99, but not APP␤, has been recovered from transfected nonneuronal cells expressing high levels of APPwt (15,17). A second proteolytic activity termed ␥-secretase, cleaves APP at the C-terminal end of the A␤ sequence, releasing A␤-  or A␤-(1-42) (18). Although secreted A␤-(1-40) and A␤-  are present in media conditioned by APPwt-transfected nonneuronal cells, intracellular A␤ has not been detected (19 -21).
Studies of a Swedish family with familial AD identified a double mutation immediately flanking the N terminus of the A␤ domain (APP⌬NL (27)) that results in elevated plasma levels of A␤-(1-40) and A␤-(1-42) (28). Moreover, nonneuronal cells transfected with APP⌬NL secrete three to six times more A␤-(1-40) and A␤-(1-42) than cells transfected with APPwt (29,30). Unlike nonneuronal cells expressing APPwt, nonneuronal cells transfected with APP⌬NL produce intracellular A␤ and APP␤⌬NL (19 -21, 31, 32), suggesting that the ⌬NL mutation diverts APP into a processing pathway that resembles that found in postmitotic neurons. To test this hypothesis, we compared the processing of APPwt and APP⌬NL in postmitotic NT2N cells, in primary rat astrocytes, and in Chinese hamster ovary (CHO) cells. We found that APP is processed in a distinct manner in NT2N neurons and that the ⌬NL mutation does not cause an increase in secreted or intracellular A␤ despite causing an increase in APP⌬NL and C99 in these cells. However, this mutation does lead to greatly increased levels of intracellular and secreted A␤-(1-40) and A␤-  in both primary astrocytes and nonneuronal cell lines. These data provide evidence that the consequences of this familial AD-associated mutation on APP processing in nonneuronal cells is to increase the utilization of the ␤-secretory pathways at the expense of the ␣-secretory pathway such that they resemble the processing pathways in postmitotic neurons. Finally, we also provide evidence that ␥-secretase, but not ␤-secretase, is a rate-limiting step in the production of A␤ in NT2N neuronal cells.

EXPERIMENTAL PROCEDURES
Antibodies-The epitopes recognized by the various APP-specific antibodies used in this study are depicted in Fig. 1. The goat polyclonal antisera Karen was the generous gift of Dr. Barry Greenberg (Cephalon, Inc., West Chester, PA) (23). Rabbit polyclonal antisera 192 and 192SW were generously provided by Drs. D. Shenk and P. Seubert at Athena Neurosciences (San Francisco, CA) (16,31). The rabbit polyclonal antisera 369W was provided by Dr. S. Gandy (Cornell University, New York, NY) (33), monoclonal antibody 6E10 was purchased from Dr. K. S. Kim (Institute for Basic Research in Developmental Disabilities, New York, NY) (34), and monoclonal antibodies BAN-50, BC-05 and BA-27 were the generous gift of Dr. N. Suzuki of Takeda Pharmaceutical and have been described previously (35,36).
Plasmid Constructs and Virus Production-The Semliki Forest virus (SFV) expression system was used to express APPwt and APPwt bearing the Swedish mutation (APP⌬NL) in a variety of cell types. pSFV1, pSFV-Helper 2, and pSFV3-lacZ were purchased from Life Technologies, Inc. (37). The pSFV-1 polylinker was modified to include ClaI as an unique cloning site. cDNA clones encoding APPwt and APP⌬NL were provided by Dr. T. Golde (University of Pennsylvania, Philadelphia, PA) (29). APPwt was ligated into the modified pSFV1 at the restriction enzyme sites BamHI and ClaI. APP⌬NL was blunt-end-ligated into pSFV1 at the restriction enzyme site SmaI. Recombinant virus vectors were produced as described previously (37). Briefly, recombinant and helper plasmids were linearized with Spe I and used as a template for the production of RNA using the SP6 polymerase. Approximately 25 g of recombinant and helper RNA were then electroporated into 1 ϫ 10 7 BHK-21 cells in PBS. The electroporated cells were incubated for 24 h in 24 ml of complete Glasgow minimal essential medium (Life Technologies, Inc.) and subsequently, the viral supernatant was harvested and stored frozen at Ϫ70°C.
Virus titer was determined by infection of BHK-21 cells with 10-fold serial dilutions of recombinant virus. After 16 h of infection, cells were fixed with ice-cold 95% ethanol, 5% acetic acid and blocked with 5% FBS in Dulbecco's PBS. The cells were subsequently incubated with the antibody Karen followed by horseradish peroxidase-conjugated rabbitanti-goat immunoglobulin. Staining was developed with 0.5 mg/ml 3,3Јdiaminobenzidine in 0.1 M Tris, pH 7, 0.01% Triton X-100, 10 mM imidazole, and 0.01% H 2 O 2 .
Infection with SFV-Cells were washed with PBS, and virus was added in 1 ml of PBS at a multiplicity of infection of 5-10 viral particles/ cell. The cells were incubated at 37°C for 1 h with intermittent agitation. The remaining virus was aspirated, and the cells were cultured at 37°C in 1 ml of complete media for different lengths of time. At the time indicated, the media was harvested, the cells were washed twice with PBS and lysed in 1% Nonidet P-40 in PBS with 5 mM EDTA, and a mixture of protease inhibitors including 500 M phenylmethylsulfonyl fluoride, 1 g/ml each leupeptin, pepstatin, soybean trypsin inhibitor, and sodium-p-tosyl-L-lysine chloromethyl ketone. Media and lysates were centrifuged at 100,000 ϫ g for 30 min. Proteolytic APP fragments in media and cell lysates were evaluated by A␤ sandwich ELISA and immunoblotting as described below.
Immunoblotting-Quantitative immunoblotting was performed as described (41). Cell lysates or media were resolved on either 7.5% Laemmli SDS-polyacrylamide gels or 16% Tris-Tricine gels. The resolved proteins were transferred to nitrocellulose in 0.2 M glycine and 25 mM Tris base at 1.5 A for 1 h at 4°C. Blots were blocked with 5% nonfat powdered milk in Tris-buffered saline and incubated overnight with the antibody as indicated in 5% milk/Tris-buffered saline). Subsequently, the blots were washed and incubated with either 1 Ci/ml protein A or 1 Ci/ml 125 I goat-anti-mouse IgG (8 -10 Ci/g; NEN Life Science Products) diluted in Tris-buffered saline for 1 h at room temperature. The blots were washed and exposed to PhosphorImager plates (Molecular Dynamics) for 24 -72 h. Quantitation of immunoblots was performed using ImageQuant software (Molecular Dynamics).
ELISA-To quantify A␤ in conditioned media and cell lysates, an antigen-capture ELISA was performed as described (35,36). Briefly, ELISA plates were coated with 1 g of Ban-50 in 0.1 M sodium carbonate buffer, pH 9.6, and blocked with 1% Block Ace (Snow Brand, Sapporo, Japan) in PBS. Plates were washed with PBS, and media or lysate was added with the appropriate synthetic A␤ standards. After 16 h, the plates were washed and horseradish peroxidase-conjugated BA-27 or BC-05 diluted in PBS supplemented with 2 mM EDTA, 1% bovine serum albumin, and 0.005% thimerosal were added. After 24 h, the plates were washed and developed with 3,3Ј,5,5Јtetramethybenzidine and 0.1% H 2 O 2 (Gingardin-Perry, Gaithersburg, MD). The reactions were stopped with 1 M H 3 PO 4 and read at 450 nm on a Dynatech MR 4000 spectrophotometer (Dynatech Laboratories, Chantilly, VA).

Expression of Human APPwt and APP⌬NL in Neuronal and
Nonneuronal Cells-To express wt and mutant forms of APP, we utilized recombinant SFV vectors. SFV efficiently expresses proteins in NT2N cells without cytopathic effects for at least 48 h (data not shown), and APP expressed by recombinant SFV vectors is expressed normally in a variety of cell types including CHO cells and primary rat astrocytes (24,25,42). Furthermore, expression of APP in rat hippocampal neurons by SFV vectors results in amyloidogenic processing of APP (43,44). Thus, SFV vectors provide an opportunity to express APP under conditions that result in normal proteolytic processing.
To determine if there are cell-type-dependent differences in the processing of wt and mutant APP, we expressed APPwt and APP⌬NL in CHO cells, NT2N neurons, and primary rat astrocytes. A recombinant SFV vector that expresses lacZ served as a control. APP expression was optimized for each cell line by varying the length of infection. After 24 h infection (16 h for the CHO cells), cell lysates and media were collected and analyzed by SDS-PAGE for the presence of APP and secreted APPS by immunoblotting with antisera directed against the APP ectodomain ( Figs. 1 and 2). Comparable amounts of full-length APP were detected in the cell lysates with both endoglycosidase H-sensitive (lower band) and resistant (upper band) forms being present in CHO cells and astrocytes. Cells expressing APP⌬NL expressed slightly reduced amounts of endoglycosidase H-resistant APP (data not shown). By contrast, the APP recovered from NT2N cell lysates was predominantly endoglycosidase H sensitive (see also Ref. 24). The amount of APPS secreted into the media was largely independent of the cell type and whether APPwt or APP⌬NL were expressed. This experiment demonstrates that APP was readily expressed and detected in all three cell types using the SFV system and that both APPwt and APP⌬NL were processed, resulting in the secretion of APPS. Significantly, we observed a reliable downward shift in the mobility of secreted APPS derived from APP⌬NL compared with APPwt. This shift, most pronounced in the nonneuronal cells, is consistent with a shift in APP processing that favors APP␤ cleavage at the expense of APP␣ (24).
Effects of the APP⌬NL Mutation on A␤-  and A␤-(1-42) Secretion-To assess the effect of the APP⌬NL mutant on A␤ secretion in the different cell types, we infected human NT2N neurons, undifferentiated NT2-cells (from which NT2N neurons are derived), CHO cells, and primary rat astrocytes with SFV vectors expressing either APPwt, APP⌬NL, or lacZ. At 16 (CHO cells) or 24 h (NT2N, NT2-, and astrocytes) post-infection, the media was collected and assayed for the presence of A␤-(1-40) and A␤-(1-42) using a well characterized ELISA assay (36). In all cases, the quantity of A␤ was normalized to the total amount of APP expressed in each cell line as determined by quantitative Western blotting. Despite the diversity of cell types used and the fact that the absolute amount of A␤ produced by each cell line varied widely, we found that expression of APP⌬NL caused a 4 -8-fold increase in the levels of secreted A␤-(1-40) and A␤-(1-42) compared with APPwt in primary rat astrocytes, CHO cells, and NT2-cells (Fig. 3). In all these cells, the ratio of A␤-(1-40) to A␤-(1-42) was approximately 10:1, consistent with other reports (23,36). In marked contrast, expression of APP⌬NL in NT2N neurons resulted in only a modest increase (0 -35%) in both A␤-(1-40) and A␤-(1-42) secretion compared with APPwt (Fig. 3). By comparison to To determine if the modest increase in A␤ secretion from NT2N cells was merely a result of the overexpression APP, we measured the levels of secreted A␤ at 6 h post-infection when the expression of APP is 5-10-fold less than at 24 h postinfection. As expected, the levels of A␤-(1-40) and A␤-(1-42) secreted at 6 h was very low compared with that seen at 24 h. Nonetheless, expression of APP⌬NL resulted in only a 28% increase in A␤-(1-40) and a 30% decrease in A␤-(1-42) at 6-h post-infection compared with APPwt (Fig. 4). Thus, even under conditions where APP processing levels were far from maximal, expression of APP⌬NL in NT2N neurons did not cause the pronounced elevation of A␤ secretion observed in nonneuronal cells. Qualitatively comparable results were obtained in three independent experiments.
Effects of APP⌬NL on Intracellular A␤ Production in CHO and NT2N Cells-To assess the effect of the APP⌬NL mutation on intracellular A␤-(1-40) and A␤-(1-42) levels in neurons and a nonneuronal cell line, we expressed APPwt, APP⌬NL, or lacZ in NT2N and CHO cells using SFV vectors. The cells were lysed 24-h post-infection and assayed for A␤-(1-40) and A␤-(1-42) by ELISA. In the NT2N cells, expression of APP⌬NL resulted in only a 40% increase in intracellular A␤-(1-40) and a 10% increase in A␤-(1-42) compared with that seen with APPwt (Fig. 5). In contrast, expression of APP⌬NL in CHO cells resulted in a 5-fold increase in intracellular A␤-(1-40) and a 2-fold increase in A␤-(1-42) APPwt (Fig. 5). Comparable results were obtained in four of four independent experiments. The small increase in intracellular A␤ observed in NT2N cells and the much larger increase seen in CHO cells correlated with the levels of A␤ secreted by these cell types (Fig. 3). Thus, the increased levels of A␤ observed in the media of CHO cells after expression of APP⌬NL was a result of increased A␤ production rather than decreased turnover or loss following secretion.
Effects of APP⌬NL on APP␤ Production and Secretion-We next investigated the efficiency of ␤ cleavage following APP⌬NL expression in NT2N neurons, astrocytes, and CHO cells to determine if this step was rate-limiting for A␤ production in NT2N cells. Cells were infected with the appropriate SFV vector, cell lysates and media were collected, and aliquots of each were subjected to SDS-polyacrylamide gel electrophoresis and quantitative immunoblotting to measure total APP expression and APP␤ secretion, respectively. To determine if the shift toward slightly lower molecular weight-secreted APPS seen in CHO cells, primary rat astrocytes, and NT2N cells after APP⌬NL expression was due to increased levels of APP␤ relative to APP␣, we used antibodies specific for either APP␣ (6E10), APP␤ (192), or APP␤⌬NL (192SW) (Fig. 1). We found that expression of APP⌬NL led to a reduction in the amount of APP␣ secreted, although the level of reduction varied between the three different cell lines (Fig. 6). Concomitant with the reduction in secreted APP␣, a large increase in the amount of secreted APP␤ was observed from NT2N cells compared with the other cell types. By contrast, APP␤ was below the level of detection in media collected from astrocytes and CHO cells expressing APPwt. The large increase in APP␤ released from NT2N cells suggests that the rate-limiting step in A␤ production after APP⌬NL expression is more likely to involve the ␥-secretase cleavage. Effects of APP⌬NL Mutation on Intracellular APP␤ and APP␤⌬NL Levels-To compare the effect of the APP⌬NL mutation on the intracellular processing of APP in neuronal and nonneuronal cells, we examined the intracellular levels of APP␤ and APP␤⌬NL in astrocytes, CHO, and NT2N cells after expression with either APPwt or APP⌬NL. In all cells tested, intracellular APP␣ was not detected after APPwt expression (data not shown), consistent with previous data showing that ␣-secretase cleavage of APP occurs at or near the plasma membrane (12,15,24). Furthermore, although APP␤ was not detected in cell lysates of astrocytes and CHO cells expressing APPwt, intracellular APP␤ was readily detected in NT2N neurons, consistent with their constitutive production and secretion of A␤ and APP␤ (Fig. 7). However, expression of APP⌬NL sharply increased the amount of detectable APP␤⌬NL in the NT2N cells, although it is evident that this antibody crossreacts to a limited degree with full-length APP⌬NL (see asterisk in Fig. 7). Significantly, intracellular APP␤⌬NL was detected in astrocytes and CHO cells after APP⌬NL expression (Fig. 7). The increase in intracellular APP␤⌬NL in NT2N cells again suggested increase APP processing by the ␤-secretory pathway after expression of APP⌬NL.
The Swedish Family Mutation Increases C99 Production Although Decreasing C83-The dramatic increase in APP␤⌬NL without an accompanying increase in A␤ levels in NT2N cells expressing APP⌬NL suggested that either the processing by ␥-secretase(s) is rate-limiting or that the carboxyl-terminal C99 fragment is rapidly degraded. To distinguish between these possibilities, we analyzed cell lysates prepared from CHO and NT2N cells expressing either APPwt or APP⌬NL for C99 and C83 with the antibodies 369W and 6E10. Antibody 369W is specific for the carboxyl terminus of APP and thus recognizes both C99 and C83, whereas antibody 6E10 only recognizes C99 since it binds to an epitope within the first 10 amino acids of A␤ (Fig. 1). In CHO cells expressing APPwt, the C83 fragment was the predominant species detected (Fig. 8), whereas expression of APP⌬NL lead to the production of the C99 fragment as well, consistent with increased ␤-secretase processing (Fig. 8). In contrast, the C99 fragment was detectable after APPwt expression in NT2N neurons (Fig. 8) at levels of approximately 30 -50% of that of C83. When APP⌬NL was expressed in the NT2N cells, there was a dramatic increase in the level of C99 with a concomitant decrease in C83 (Fig. 8). The amount of C99 pres-  (NT2N cells and astrocytes) post-infection, the media was harvested. Media normalized for total APP expression was resolved on a 7.5% polyacrylamide gel and immunoblotted with antibodies as indicated. ent in NT2N cells expressing APP⌬NL was approximately 3-5-fold higher than in cells expressing APPwt. These data suggest that the lack of an increase in A␤ levels after the expression of APP⌬NL in NT2N neurons is likely due to limited processing by ␥ secretase and not due to the rapid degradation of the C99 fragment. DISCUSSION Expression of APP⌬NL in nonneuronal cells leads to a 5-10fold increase in the amount of secreted and intracellular A␤ relative to that seen after expression of APPwt, perhaps explaining how this mutation results in accelerated AD pathology (19 -21, 29 -31, 45, 46). However, expression of APP⌬NL in NT2N neurons resulted in only a minimal increase (Ͻ35%) in the amount of both secreted and intracellular A␤-  and no increase in A␤-(1-42) production. This differential effect of the Swedish ⌬NL mutation on APP processing and A␤ production in NT2N neurons as compared with nonneuronal cell lines demonstrates that cells can process APP⌬NL differently, provides insight into the mechanisms by which A␤ is generated from APP⌬NL, and raises the question as to the cell type(s) specifically affected by this mutation in vivo that contribute the most to the formation of senile plaques.
The failure of the ⌬NL mutation to result in greatly enhanced A␤ production in NT2N neurons could be the result of several factors, including reduced processing by ␤or ␥-secretases. However, we found that expression of APP⌬NL in all cell types examined, including NT2N cells, caused a marked shift from the ␣-secretory to the ␤-secretory pathway as evidenced by reduced secretion of APP␣ and increased secretion of APP␤⌬NL. The expression of APP⌬NL also shifted the production of the C-terminal fragments from C83 (generated by ␣secretase cleavage) to the C99 fragment (generated by ␤-secre-tase cleavage). This shift was especially pronounced in CHO cells, which produced little or no C99 after expression of AP-Pwt. A similar though less dramatic effect on C99 production was observed in NT2N cells after APP⌬NL expression, again indicating that processing by ␤-secretase(s) is increased in NT2N cells expressing APP⌬NL. Thus, the ⌬NL mutation leads to increased processing of APP by ␤-secretase(s) in both neuronal and nonneuronal cells.
The increased processing by ␤-secretase(s) observed in all cell types after APP⌬NL expression could be the result of several factors. At present, at least three different ␤-secretase processing pathways have been reported. The endosomal/lysosomal pathway, which processes APP after re-internalization from the cell surface into endosomes and lysosomes, is the most ubiquitous since both neurons and nonneuronal cells utilize this pathway to produce A␤. However, the contribution of this pathway to the overall production of A␤ is relatively minor since nonneuronal cells transfected with APPwt produce mostly p3 and very little A␤ (15,20,47,48). In addition, it is unlikely that APP⌬NL expression increases A␤ production via this pathway, since the expression of an APP⌬NL construct lacking the cytoplasmic tail, which eliminates re-internalization of cell surface APP⌬NL, does not reduce A␤ secretion (31,49). A second ␤-secretory pathway, active primarily in Golgiderived vesicles, is more likely to process APP⌬NL (31). Previous studies have suggested that nonneuronal cells utilize this pathway at the expense of the ␣-secretory pathway when APP⌬NL is expressed (31). However, it is unclear whether A␤-(1-40) and A␤-(1-42) are both produced by this route, since a recently identified ␤-secretory pathway localized to the ER/IC results in exclusive production of A␤-(1-42) (25,26). Since the secretion of both A␤-(1-40) and A␤-(1-42) is elevated in nonneuronal cells expressing APP⌬NL, it is possible that both the Golgi-associated and the ER ␤-secretase pathways are affected by this mutation. Alternatively, it is possible that both A␤-(1-40) and A␤-(1-42) can be produced by Golgi-derived vesicles and that the ⌬NL mutation only affects this pathway selectively.
The increased processing of APP⌬NL by the ␤-secretase pathway in all cells tested coupled with markedly increased levels of A␤ production in all cells except NT2N neurons suggests that ␥-secretase processing may be rate-limiting for A␤ production in NT2N cells by one of several mechanisms. For example, C99 may be more rapidly degraded in NT2N cells than in nonneuronal cells, precluding or minimizing the possibility of ␥-secretase cleavage of C99 and concomitant A␤ production. However, if this were true, we would not expect the observed increase in the C99 fragment after APP⌬NL expression in NT2N cells. Another possibility is that the high levels of expression resulting from the SFV vector system saturates the ␥-secretase. However, we found no difference in A␤ levels between NT2N cells expressing APPwt or APP⌬NL at early time points when APP expression levels were low. In addition, De Strooper et al. (43) found that expression of APP⌬NL in rat hippocampal neurons caused only a 2-fold increase in A␤ levels relative to APPwt as judged by immunoprecipitation. Although larger than the increase we observed in this study using a more quantitative ELISA approach, the increase in A␤ was still significantly less than the 5-10-fold increase observed in nonneuronal cells. Finally, the intracellular processing pathways utilized by the NT2N cells may be distinct from that of the nonneuronal cells (12,13,22,23,25). Thus, the possibility exists that the ␥-secretase did not have access to the carboxylterminal fragments generated when APP⌬NL was expressed. Identification of the ␥-secretase(s) will help distinguish these possibilities. Several studies suggest that the cellular source of A␤ deposition in senile plaques in nonfamilial cases of AD is the neuron (50,51). Our data supports this hypothesis because only neuronal cells constitutively generate A␤. This production of A␤ is accompanied by the generation of intracellular APP␤ and C99 in NT2N cells. By contrast, intracellular APP␤ and C99 were absent from nonneuronal cells expressing APPwt. This suggests that commitment to neuronal differentiation results in the acquisition of the ability to utilize the ␤ pathway of APP processing, a necessary prerequisite for A␤ production. However, expression of APP⌬NL led to markedly increased production of A␤, APP␤, and C99 in nonneuronal cells, indicating that the Swedish ⌬NL mutation alters APP processing in nonneuronal cells such that APP is now processed in a more "neuronal" fashion. NT2N cells may not exhibit such marked alterations in APP and A␤ metabolism after APP⌬NL expression, because they are already committed to processing APP in pathways that favor increased production of both intracellular and secreted A␤. The relatively modest effects of this mutation on A␤ production in NT2N neurons raises the question as to what role neurons play in elevating A␤ levels in individuals with the ⌬NL mutation. It is possible that in such individuals cell types other than neurons, such as astrocytes, may significantly contribute to the marked increases in central nervous system A␤ levels and the formation of senile plaques.