Characterization of a presenilin-mediated amyloid precursor protein carboxyl-terminal fragment gamma. Evidence for distinct mechanisms involved in gamma -secretase processing of the APP and Notch1 transmembrane domains.

A variety of investigations have led to the conclusion that presenilins (PS) play a critical role in intramembranous, gamma-secretase proteolysis of selected type I membrane proteins, including Notch1 and amyloid precursor protein (APP). We now show that the generation of the S3/Notch intracellular domain and APP-carboxyl-terminal fragment gamma (CTFgamma) derivatives are dependent on PS expression and inhibited by a highly selective and potent gamma-secretase inhibitor. Unexpectedly, the APP-CTFgamma derivative is generated by processing between Leu-645 and Val-646 (of APP(695)), several amino acids carboxyl-terminal to the scissile bonds for production of amyloid beta protein peptides. Although the relationship of APP-CTFgamma to the production of amyloid beta protein peptides is not known, we conclude that in contrast to the highly selective PS-dependent processing of Notch, the PS-dependent gamma-secretase processing of APP is largely nonselective and occurs at multiple sites within the APP transmembrane domain.

Presenilin 1 and 2 (PS1 and PS2) 1 are polytopic membrane proteins that are mutated in the majority of pedigrees with early-onset FAD. Compelling evidence has emerged supporting a role for PS in intramembranous ␥-secretase processing of the type I membrane proteins, including APP (1, 2) and Notch1 (3). ␥-Secretase processing of a set of membrane-tethered APP-CTFs results in the production of A␤ peptides that are secreted and subsequently deposited in brains of patients with Alzheimer's disease. On the other hand, ␥-secretase processing of the Notch1 derivative, termed S2/NEXT, releases the intracellular domain (S3/NICD), a transcriptional coactivator that associates with CBF1/Su(H)/Lag1 (4 -6). The observation that A␤ and S3/NICD production are completely eliminated in cells with compound deletions of PS1 and PS2 (7,8), and the demonstration that PS1 and PS2 can be photocross-linked to ␥-secretase inhibitors (9), have led to the conclusion that PS are the molecules that execute intramembranous cleavage of APP and Notch1. Furthermore, PS shares limited homology with the prepilin peptidases, a family of bacterial aspartyl proteases (10), and PS binds to pepstatin, an aspartyl protease inhibitor (11). Although compelling, several observations have questioned the veracity of this model. First, and despite reports showing PS1 to be resident in late compartments, including endosomes and plasma membranes (12)(13)(14), we and others have revealed that PS are preponderantly localized in the endoplasmic reticulum and associated compartments, sites that are incompatible with the known cellular sites for ␥-secretase cleavage of APP and Notch1 that include the plasma membrane, Golgi, and endosomes (15,16). Second, endocytosis and recycling of APP-CTFs is a major pathway for generating A␤ (17), but production of the Notch S3/NICD does not require endocytic trafficking of S2/NEXT (18). Third, ␥-secretase has contrasting substrate specificities for processing within the APP (19,20) and Notch (5) TM domains, and ␥-secretase processing of the APP TM domain occurs at heterogeneous sites (19,20) whereas ␥-secretase cleavage of Notch1 appears to generate a single S3/NICD species (5).
In the present report, we examine the production of ␥-secretase-generated APP and Notch1 derivatives. We now document, both in living cells and cytosol-free membrane preparations, the generation of APP-CTF␥ derived from membrane-tethered APP-CTFs. In parallel, we document the production of the Notch S3/NICD in these preparations. The production of the APP-CTF␥ and S3/NICD derivatives are dependent on PS1 expression and inhibited by a ␥-secretase inhibitor. Unexpectedly, the APP-CTF␥ derivative produced in these reactions is generated by endoproteolytic processing at a site several amino acids carboxylterminal to the scissile sites for production of A␤ peptides but interestingly, occurs at a site immediately proximal to the analogous site for production of S3/NICD. Although the relationship of APP-CTF␥ to the production of A␤ peptides is not known, we conclude that in contrast to the highly selective PS-dependent proteolysis of the Notch TM domain, the PS-dependent ␥-secretase processing of APP is largely nonselective and occurs at multiple sites within the APP TM domain. APP-CTFs were detected by CT15 (22). For detection of A␤, conditioned medium was immunoprecipitated with 26D6 antibody (23) and immunoblotted with 26D6.
Pulse-Chase and Immunoprecipitation-N2aWT.7 or D385A.16 cells were labeled for 15 min with 750 Ci/ml [ 35 S]methionine (PerkinElmer Life Sciences). Cells were either harvested after pulse or chased for 45 min at 37°C in the presence of 0.5 mM cold L-methionine (Life Technologies, Inc.). Conditioned media were collected, centrifuged briefly, and immunoprecipitated with 4G8. Cell pellets were treated with 250 l of 3% SDS in phosphate-buffered saline containing 10 l/ml ␤-mercaptoethanol and subjected to vortexing and heating at 95°C for 10 min, followed by sonication and centrifugation at 100,000 ϫ g for 10 min. The supernatants were diluted 1:4, adjusted to a final concentration of 2% Triton X-100 and 190 mM NaCl, 20 mM Tris-Cl, pH 8.8, and 2 mM EDTA, and subjected to immunoprecipitation with 369 or 4G8.
Purification of Membrane Fractions and in Vitro ␥-Secretase Assay-Purification of membrane fractions from cultured cells was performed as described (24). Briefly, cell pellets from ten 10-cm dishes were resuspended in Buffer A (250 mM sucrose, 20 mM HEPES, pH 7.4) containing protease inhibitor mixture. Post-nuclear supernatant was layered onto 30% Percoll solution and subjected to high speed centrifugation. Cellular membranes was resuspended in 14% iodixanol and overlaid with 12 and 6% iodixanol. Following centrifugation at 52,000 ϫ g for 90 min, membranes were collected from each interface. Collected membranes were incubated at 4 or 37°C for 60 min, and the reactions were terminated by the addition of Laemmli sample buffer. In the ␥-secretase inhibitor experiments, the reaction tubes were preincubated at 4°C for 10 min with 100 nM L-685,458 (25) prepared in Me 2 SO, or an equivalent concentration of Me 2 SO, and then incubated at 37°C for 60 min.
Mass Spectrometric Analysis-The reaction samples were centrifuged at 100,000 ϫ g for 1 h to separate the membrane-bound and soluble APP-CTFs. CTF␥ from the soluble fraction was immunoprecipitated with CT15 and protein A-agarose and analyzed by MALDI-TOF MS (26).

Cellular Production of S3/NICD and APP-CTF␥ Is PS1-dependent-
We demonstrated previously that a Notch6-mycgreen fluorescent protein chimera is subject to PS1-dependent ␥-secretase processing; this processing event is stimulated either by addition of extracellular ligand or by depletion of extracellular calcium (21). To assess the role for PS1 in the production of an APP carboxyl-terminal derivative, termed APP-CTF␥, that would be generated following intramembranous proteolysis by ␥-secretase, we performed Western blot analysis of detergent lysates from PS1-deficient fibroblasts transiently cotransfected with cDNA encoding APPmyc and either wild type PS1 or a PS1 aspartate variant (D385A) that has been shown to reduce A␤ secretion (27,28) (Fig. 1A). As expected, upon transfection of APPmyc cDNA, ϳ12-kDa CTF-myc␣ and ϳ14-kDa CTFmyc␤ appear that are significantly elevated over the ϳ10-kDa endogenous CTF␣ seen in untransfected cells (Fig. 1A, compare lanes 1 and 2). On the other hand, coexpression of human wild type PS1 with APPmyc reduces the levels of these ϳ12and ϳ14-kDa CTFmyc species, commensurate with the accumulation of a novel ϳ7-kDa CTFmyc species, which we propose is CTFmyc␥ (Fig. 1A, lane 3). In support of this notion, coexpression of the D385A PS1 variant and APPmyc restores the levels of accumulated CTFmyc␣ and CTFmyc␤ species seen in APPmyc-transfected cells and most importantly, eliminates the production of the CTFmyc␥ (Fig.  1A, lane 4). We confirmed these findings by showing that expression of APPmyc does not lead to appreciable levels of secreted A␤ but that coexpression of wild type PS1 restores A␤ secretion (Fig. 1A, lanes 2Ј and 3Ј). Moreover, the PS1 D385A variant inhibits A␤ production, as expected (Fig. 1A, lane 4Ј). These results in transient transfection analyses were fully confirmed in biosynthetically labeled N2a cell lines that stably coexpress APPmyc and either wild type PS1 (WT.7) or the PS1 D385A variant (D385A. 16). Pulse-labeling with [ 35 S]methionine for 15 min and immunoprecipitation analysis of detergent lysates using antibody 369 revealed identical synthetic levels of APP in both lines and low but detectable levels of ϳ12and ϳ14-kDa CTFmyc species (Fig. 1B, lanes 1 and 2); intracellular A␤ peptides are not detected at this time point (Fig. 1B,  lanes 1Ј and 2Ј). After 45 min of chase, we observed increased levels of ϳ12-kDa CTFmyc␣ and ϳ14-kDa CTFmyc␤ species and an ϳ7-kDa CTFmyc␥ species in lysates of WT.7 line (Fig.  1B, lane 3). In contrast, we observed very high levels of CTF-myc␣ and CTFmyc␤ species in lysates of cells expressing the D385A PS1 variant (Fig. 1B, lane 4). However, the level of ϳ7-kDa CTFmyc␥ in cells expressing the D385A PS1 variant was markedly lower than the levels observed in cells expressing wild type PS1 (Fig. 1B, compare lanes 3 and 4). Indeed, the levels of A␤ both in cell lysates and in the conditioned medium from cells expressing the D385A variant were substantially reduced compared with those from cells expressing wild type PS1 (see Fig. 1B, compare lanes 3Ј and 4Ј and lanes 3Љ and 4Љ; A␤ secretion is not detectable at the end of the pulse-labeling period).  1 and 2) or 45 min (lanes 3 and 4). Cell lysates were immunoprecipitated with 369 (first and second panels) or 4G8 (third panel), and conditioned media (bottom panel) were immunoprecipitated with 4G8. Immunoprecipitated materials were analyzed by 10 -20% Tris-Tricine (for A␤ and CTFs) or 4 -12% Tris-Glycine (for APP) SDS-PAGE. Molecular mass markers are in kDa. The bands representing full-length APP, transgenic APP-CTFmyc␤ and ␣, and endogenous CTF␣ and A␤ are indicated at right. *, APP-CTFmyc␥.
In Vitro Generation of NICD and APP-CTF␥-To develop a biochemically tractable system to examine ␥-secretase processing of Notch and APP, we isolated membrane fractions from PS1Ϯ fibroblasts that express the Notch-green fluorescent protein chimera and examined EDTA-induced ␥-secretase processing ( Fig. 2A). We show that incubation of these membranes at 37°C led to the production of the S3/NICD fragment ( Fig. 2A,  lanes 1 and 2) and that this reaction was inhibited by addition of the ␥-secretase inhibitor L-685,458 to 100 nM ( Fig. 2A, lane  3). In parallel, we assessed the production of a previously described ϳ6-kDa CTF␥ derived from ␥-secretase cleavage of endogenous ϳ10-kDa CTF␣ (29,30). We detected the CTF␥ only after incubation at 37°C (Fig. 2B, lanes 1 and 2), and this reaction was also inhibited by addition of the ␥-secretase inhibitor (Fig. 2B, lane 3). In titration experiments using varying concentrations of L-685,458, we have demonstrated that the in vitro generation of S3/NICD and APP-CTF␥ are equally sensitive to the inhibitor with an IC 50 of ϳ50 pM, 2 findings that suggest that these derivatives are generated by similar, if not identical, ␥-secretase activities.
Biochemical Characterization of CTF␥-As the levels of endogenous APP-CTF␥ generated in the fibroblasts were extremely low, we chose to analyze the CTF␥ fragment generated in membranes from a cell line (N2aWT.11) that stably expresses APPmyc and wild type PS1. Incubation of membranes at 37°C for 60 min from N2aWT.11 cells leads to unchanged levels of full-length APP but a diminution in levels of the ϳ12and ϳ14-kDa CTFmyc derivatives. Commensurate with reduction in levels of these CTFmyc species was the appearance of an ϳ7-kDa CTFmyc␥ derivative, similar to the PS1-dependent CTFmyc␥ fragment observed in living cells (Fig. 3A, lanes 1  and 2; see Fig. 1B). With the assumption that the CTFmyc␥ fragment might be extruded into the cytoplasmic compartment in a manner similar to the S3/NICD fragment of Notch, we prepared membrane and soluble fractions. The ϳ12and ϳ14-kDa CTFmyc fragments were membrane-bound prior to, or after, incubation at 37°C (Fig. 3B, lanes 2 and 4), and as expected, the CTFmyc␥ fragment was found exclusively in the soluble fraction (Fig. 3B, lane 6). To establish the identity of the CTFmyc␥ fragment in the soluble fraction, we immunoprecipi-tated the peptide with CT15 and subjected recovered material to MALDI-TOF MS (Fig. 3C). We observed principal peptides of M r 7341.3 and 7110.4, which includes the mass of the 12-amino acid myc epitope tag; the prominent m r 7341.3 peptide is generated by endoproteolytic cleavage between Leu-645 and Val-646 (of APP 695 ), whereas the m r 7110.4 peptide is generated by cleavage between Met-647 and Leu-648 (Fig. 3C). DISCUSSION A series of genetic and pharmacological approaches have revealed that PS are required for ␥-secretase processing of Notch1 and APP (31). In the case of Notch1, a membranetethered S2/NEXT is subject to processing by ␥-secretase, leading to the generation of a cytoplasmic fragment, S3/NICD (4). APP is processed by ␣and ␤-secretases to generate a set of membrane-tethered CTFs, and these derivatives are subsequently processed by ␥-secretase to generate a spectrum of secreted A␤ peptides (see below). However, the residual CTF␥ resulting from ␥-secretase processing of the CTF␣ and -␤, has been elusive. Very recently, McLendon et al. (29) and Pinnix et al. (30) reported on the generation of a CTF␥ derivative in isolated membranes from brain or cultured cells. This derivative was not biochemically characterized, but its production was inhibited by high concentrations of pepstatin A, MG132, and a substrate-based difluoroketone (30).
To these latter observations, the present report offers several novel insights relevant to ␥-secretase processing of the Notch S2/NEXT and the APP-CTFs. First, we document that a peptide corresponding to APP-CTF␥ can be detected in lysates of transfected cells and that the production of this derivative is dependent on PS1 expression and largely eliminated by expression of the dominant negative D385A PS1 variant. Second, we establish that both the APP-CTF␥ and S3/NICD can be generated in cytosol-free membranes and that the production of these derivatives is inhibited by a ␥-secretase inhibitor. Third, and most surprising, is our demonstration using MALDI-TOF MS that the prominent in vitro-derived APP-CTF␥ is generated by proteolysis between the 21st (Leu-645) and 22nd (Val-646) amino acids of the 24-amino acid APP TM domain. Notably, the P1Ј Val-646 is one residue more carboxyl-terminal to the P1Ј Val that is necessary for Notch processing (5). Our studies are consistent with recent results from Sastre et al. (32) and Gu et al. (33).
Having established the identity of CTF␥, we are now faced with the conundrum that the processing sites for generation of this CTF are distinct from the scissile sites for A␤40 or A␤42 peptides, which occur between the 12th (Val-636) and 13th (Ile-637) or 14th (Ala-638) and 15th (Thr-639) amino acids, respectively. Taken together with our present data showing that CTF␥ production is dependent on PS1 expression, and inhibited by a selective ␥-secretase inhibitor that also blocks production of A␤, we offer three mechanistic scenarios.
The first of these is that CTF␥ is derived by secondary endoproteolysis of the CTF57 or CTF59 derivatives that are the residual fragments from proteolysis at the scissile bonds that generate A␤. We consider this unlikely, as we have detected neither the CTF57 nor CTF59 derivatives in in vivo pulsechase studies nor in kinetic studies in cytosol-free membrane preparations. The second possibility is that proteolysis between Leu-645 and Val-646 is obligatory for subsequent endo-or exoproteolytic activity events necessary for generating A␤40/42 peptides. This notion can be tested by assessing CTF␥ and A␤ production in cells expressing APP variants with mutations surrounding the scissile site. In this regard, and in view of the finding that a V1744G substitution at the P1Ј site in the Notch blocks NICD production and Notch activity (34), Sastre et al. (32) have recently examined CTF␥ production in membranes 2 T. I. and S. S. S., submitted for publication.
prepared from cells expressing APP with a V646G substitution. Surprisingly, the production of CTF␥ was unimpaired. Clearly, a more rigorous mutational analysis is required before a conclusion can be drawn about the relationship between the generation of CTF␥ and A␤ production. The third alternative is that proteolysis giving rise to A␤ does not require prior generation of the CTF␥ or vice versa. In this scenario, the membranetethered APP-CTFs are randomly processed at multiple sites. In view of the multiplicity of A␤-related peptides with termini at 34, 37, 38, 39, 42, and 43 that are invariably detected (19,20,26), and the demonstration that extensive mutagenesis of the APP TM domain have little effect on the production of A␤related peptides (19,20), we would argue that the PS-dependent ␥-secretase activity is largely nonselective. We would offer the proposal that the APP TM domain is subject to processing by a presently unidentified membrane-resident, multicatalytic protease similar to the proteosome. We envision that the function of this protease would be to introduce nicks in TM segments of many membrane proteins, resulting in slippage of the residual TM segments into either the cytoplasmic or lumenal space wherein the entire domain is subject to the proteolytic activities resident within those compartments. In this vein, it is equally conceivable that regulated forms of this protease class would be responsible for generating cytosolic fragments that play critical roles in nuclear signaling events (35). The recent demonstration of a potential nuclear signaling role for the cytoplasmic domain of APP (36) underscores the importance of understanding the mechanisms responsible for intramembranous processing of APP.
Despite the differences in the precise sites of proteolysis and sequence requirements within the APP and Notch TM domains, we have shown that L-685,458 is an equally potent inhibitor of the reaction(s) that generate Notch S3/NICD and CTF␥ (see Ref. 21 and this study). 2 These data would imply that if ␥-secretase is a single entity, it must be highly unusual. In this regard, it is not clear how expression of PS1 harboring a D257A substitution is still capable of generating A␤ but fails to generate Notch S3/NICD (28). Similarly, expression of FAD-linked L166P PS1 variant or the experimental L286E or L286R PS1 mutants leads to overproduction of A␤42 but fails to generate S3/NICD (37). 3 Finally, recent studies have revealed that processing of APP and Notch1 can be discriminated by a JLK family of non-peptidic inhibitors (38). These inhibitors block A␤ production but have very little, if any, effect on production of S3/NICD (38). Collectively, these experiments prove unequivocally that ␥-secretase processing of APP and Notch can be dissociated, leading us to conclude that the catalytic activities responsible for processing these substrates are not one in the same. Thus, whereas PS are critical for regulating ␥-secretase activities, it is our view that these polypeptides are unlikely to be the sole effectors of intramembranous proteolysis of APP and Notch1. In the final analysis, it will be critical to develop in vitro reconstitution systems with purified components to establish the role(s) of PS and their interacting components in facilitating ␥-secretase processing of APP and Notch1.