γ-Secretase Processing and Effects of γ-Secretase Inhibitors and Modulators on Long Aβ Peptides in Cells*

Background: Aβ peptides are generated by stepwise cleavage of the amyloid precursor protein. Results: Aβs longer than Aβ1–48 are efficiently cleaved by γ-secretase in cells with different response to inhibitors (GSIs) and modulators (GSMs). Conclusion: The initial γ-secretase cleavage does not precisely define subsequent product lines. Significance: Understanding how different species of Aβ are generated and modulated by small molecules has broad therapeutic implications. Understanding how different species of Aβ are generated by γ-secretase cleavage has broad therapeutic implications, because shifts in γ-secretase processing that increase the relative production of Aβx-42/43 can initiate a pathological cascade, resulting in Alzheimer disease. We have explored the sequential stepwise γ-secretase cleavage model in cells. Eighteen BRI2-Aβ fusion protein expression constructs designed to generate peptides from Aβ1–38 to Aβ1–55 and C99 (CTFβ) were transfected into cells, and Aβ production was assessed. Secreted and cell-associated Aβ were detected using ELISA and immunoprecipitation MALDI-TOF mass spectrometry. Aβ peptides from 1–38 to 1–55 were readily detected in the cells and as soluble full-length Aβ proteins in the media. Aβ peptides longer than Aβ1–48 were efficiently cleaved by γ-secretase and produced varying ratios of Aβ1–40:Aβ1–42. γ-Secretase cleavage of Aβ1–51 resulted in much higher levels of Aβ1–42 than any other long Aβ peptides, but the processing of Aβ1–51 was heterogeneous with significant amounts of shorter Aβs, including Aβ1–40, produced. Two PSEN1 variants altered Aβ1–42 production from Aβ1–51 but not Aβ1–49. Unexpectedly, long Aβ peptide substrates such as Aβ1–49 showed reduced sensitivity to inhibition by γ-secretase inhibitors. In contrast, long Aβ substrates showed little differential sensitivity to multiple γ-secretase modulators. Although these studies further support the sequential γ-secretase cleavage model, they confirm that in cells the initial γ-secretase cleavage does not precisely define subsequent product lines. These studies also raise interesting issues about the solubility and detection of long Aβ, as well as the use of truncated substrates for assessing relative potency of γ-secretase inhibitors.

␥-Secretase cleavage of APP CTFs has been proposed to occur in three distinct steps: (i) an initial ⑀-cleavage near the cytoplasmic face of the transmembrane domain that liberates the cytoplasmic tail of the substrate (22,23), (ii) stepwise tri-or tetra-peptide carboxyl-peptidase-like cleavages, and (iii) a final cleavage releasing the A␤ peptide (24). The most commonly observed final products are A␤1-37, 1-38, 1-39, 1-40, and 1-42 (25). A␤1-45 and A␤1-46 that can be generated during the stepwise cleavage have typically only been detected in the presence of certain ␥-secretase inhibitors (GSIs) or in in vitro assays (26 -28), although some studies have provided limited evidence that these species may be present in human and APP transgenic mouse brain (27,29). A␤1-48 and A␤1-49, the main A␤s produced from the initial ⑀-cleavage are typically not detectable in physiological models, but their generation is inferred from the identification of the corresponding APP intracellular domains, C51 and C50 (22,23). These ⑀-sites are considered the main initial cleavage sites within the APP CTF. A third site generating A␤1-51 and the APP intracellular domain C49 has also been identified in broken cell ␥-secretase assays and in presenilin APP mutant cell lines (30). Shorter A␤ peptides such as A␤1-17 through A␤1-20 have also been reported, but whether they are produced by ␥-secretase alone or the combined action of ␥-secretase and other proteases has not been resolved (25,(31)(32)(33).
Because there is evidence that the initial cleavage of APP at A␤1-51 or A␤1-48 generates a higher ratio of A␤1-42, whereas cleavage at A␤1-49 generates higher levels of A␤1-40, A␤1-40 and A␤1-42 have been proposed to be generated from two different "product lines" in a stepwise cleavage model. Funamoto et al. (34) provided initial evidence for this model using truncated A␤ minigenes in which A␤ was expressed in mammalian cells as APP signal peptide:A␤ fusion proteins. Cells expressing A␤1-49 generated predominantly A␤1-40 and cells expressing A␤1-48 generated much more A␤1-42, although the absolute amount of A␤ generated was very low (ϳ0.2 pM compared with Ͼ200 pM from A␤1-49). Takami et al. (24) subsequently provided direct evidence for stepwise cleavage using a very elegant strategy, which combined in vitro ␥-secretase cleavage and LC-MS/MS to quantify the specific peptides that were postulated to be released during stepwise processing. They showed that A␤1-40 can be preferentially derived from A␤1-49 through sequential removing of ITL, VIV, and IAT peptides, whereas A␤1-42 is preferentially derived from A␤1-48 through sequential cleavage of VIT and TVI peptides. More recently, additional support for this stepwise cleavage model has come from several studies showing that FAD-linked mutations in APP and PSEN can not only shift the initial ⑀-cleavage site, but also can alter subsequent processivity, thus increasing the relative production of long A␤ peptides (35,36). One of these studies also suggested that A␤1-38 arises from an additional stepwise cleavage of A␤1-42 (36), a finding directly supported by a recent study showing that A␤1-42 could be further processed into A␤1-38 in vitro, whereas A␤1-43 is cleaved into both A␤1-40 and A␤1-38 (37).
In the present study, we further explored the sequential cleavage model initially proposed by Ihara and co-workers (24, 34) using BRI2-A␤1-x fusion proteins that produced A␤ peptides ranging from 1-38 to 1-55. We have previously employed this strategy to efficiently express individual A␤ proteins in cells and in the brains of transgenic mice (38 -40). Because many of the studies supporting a stepwise cleavage model for ␥-secretase processing of APP have been performed in in vitro assays or in cells utilizing signal peptide A␤ peptide expression constructs that produce A␤ peptides that are inefficiently secreted, we hypothesized that the BRI2 fusion protein strategy, which results in efficient processing and secretion of A␤ from the fusion protein, might provide a more physiologic system to assess ␥-secretase processivity. Although our studies further support the sequential ␥-secretase cleavage model, they show that in cells the initial ␥-secretase cleavage site does not absolutely define subsequent product lines. Unexpectedly, we also find (i) that under some circumstances long A␤ peptides generated in cells following processing of the BRI2 fusion protein are efficiently secreted as intact soluble peptides; (ii) that the long A␤ peptides dramatically decrease the relative potency of multiple GSIs compared with the potency of the same GSIs when APP or C99 is used as a substrate; and (iii) that two PSEN1 mutants selectively alter processing for A␤1-51, but not A␤1-49.

EXPERIMENTAL PROCEDURES
DNA and Cell Culture-Fusion constructs encoding the first 243 amino acids of BRI2 protein followed by A␤ peptides encompassing various A␤ species from A␤38 to A␤55 or C99 (99 amino acids on the C-terminal of APP) were generated as previously described (38). The fragments were ligated into the expression vector pAG3. Sequences were verified by DNA sequencing. The overexpression was performed by transiently transfecting human embryonic kidney (HEK 293T) and CHO cells. Cells were grown in either DMEM (HEK) or Ham's F-12 (CHO) media supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin/streptomycin (Invitrogen). Briefly, 2.7 g of DNA was applied to a 75% confluent 6-well plate (Corning) using the polycation polyethylenimine transfection method (41). When BRI2-A␤ constructs were co-transfected with either wild-type PSEN1 or the mutants M139V or ⌬exon9 into CHO cells, 1.35 g of each DNA were used. In some experiments, 10-cm dishes were used with appropriately adjusted amounts of DNA. Cells were incubated with transfection reagent for 12-16 h, after which the growth medium was replaced with fresh medium. DMSO (Sigma), GSIs, or GSMs were added to the appropriate concentration. The GSIs L685,458 and MRK680 were purchased from Tocris; the GSMs sulindac sulfide and fenofibrate were purchased from Sigma. (Z-LL) 2 ketone was from Calbiochem. LY411,575, GSM1, and compound 2 were synthesized by A. Fauq at the Mayo Clinic Chemical Core. 24 h later, the medium was collected for assay by ELISA and MS. Stable CHO cell lines were selected with 0.8 g/ml hygromycin B. To evaluate any potential non-␥-secretase-mediated proteolytic cleavage of the long secreted A␤s, 1 M synthetic A␤1-49 was incubated with CHO cells overnight, and then A␤ in the media was analyzed by immunoprecipitation MALDI-TOF mass spectrometry (IP/MS), as described below. A sample with 1 M synthetic A␤1-49 in fresh media was used as control.
Immunoprecipitation and Mass Spectrometry-50 l of magnetic sheep anti-mouse IgG beads (Invitrogen) were incubated with 4.5 g of Ab5 antibody for 30 min at room temperature with constant shaking. The beads were then washed with PBS and incubated with 1-10 ml of conditioned medium, to which 0.1% Triton X-100 was added for 1 h. For cell lysate IP, cells from a 10-cm dish were washed twice with PBS and then incubated with 1 ml of PBS with 1% Triton X-100 on ice for 20 min. Lysates were centrifuged at 1500 ϫ g for 10 min at 4°C with the resulting supernatants diluted with 9 ml PBS and incubated with Ab5-pretreated beads. Bound beads were washed sequentially with 0.1% and 0.05% octyl glucoside (Sigma) followed by water. Samples were eluted with 10 l of 0.1% trifluoroacetic acid (Thermo Scientific, Rockford, IL) in water. 2 l of elute was mixed with an equal volume of saturated ␣-cyano-4-hydroxycinnamic acid (Sigma) solution in 60% acetonitrile, 40% methanol. 1 l of sample mixture was loaded to ␣-cyano-4-hydroxycinnamic acid pretreated MSP 96 target plates (Bruker Daltonics Inc., Billerica, MA). The samples were analyzed with a Bruker Microflex (Bruker Daltonics Inc.) mass spectrometer.
In Vitro Assay-A␤1-48, A␤1-49, and C99 (CTF99 plus a methionine at the N-terminal and FLAG tag at the C-terminal) were used for in vitro assay. A␤48 and A␤49 were purchased from AAPPTEC (Louisville, KY). The peptides were pretreated with 1,1,1,3,3,3-hexafluoroisopropanol. After drying, the peptides were dissolved in 0.1% ammonia hydroxide. C99 was purified as described with modifications (46). Briefly, inclusion bodies were dissolved in Tris⅐HCl-urea buffer (25 mM Tris, 8 M urea, pH 8.0). After addition to a HiTrap Q column (GE Healthcare), the column was washed with 10ϫ column volume of Tris⅐HClurea buffer and then Tris⅐HCl-CHAPSO buffer (25 mM Tris, 1% CHAPSO, pH 8.0). C99 was eluted using Tris⅐HCl-CHAPSO buffer with NaCl gradient. Gel filtration and anti-FLAG M2 affinity column can be used if further purification is necessary. CHAPSO-solubilized ␥-secretase was prepared from CHO S-1 cells as previously described (47). Briefly, S-1 cells were washed once with cold PBS and then resuspended in PBS with Complete protease inhibitors (Roche Applied Science). After nitrogen cavitation at 750 p.s.i. on ice for 1 h, cell debris and nuclei were removed by centrifugation at 1000 ϫ g for 10 min. The supernatant was centrifuged at 100,000 ϫ g for 1 h to pellet the total membranes. The membrane pellet was resuspended and homogenized in cold sodium bicarbonate buffer (100 mM NaHCO 3 , pH 11.0). After centrifugation at 100,000 ϫ g for 1 h, the pellet was solubilized with CHAPSO buffer (150 mM sodium citrate, 1% CHAPSO, 5 mM EDTA) by slightly vortex and incubated on ice for 1 h. The insoluble debris was removed by centrifugation at 100,000 ϫ g for 1 h. For in vitro assay, substrates (25 M for MS, 1 M for ELISA) were incubated with CHAPSO solubilized ␥-secretase (100 g/ml protein concentration) in sodium citrate buffer (150 mM, 1ϫ complete protease inhibitor, pH 6.8) for 3 h at 37°C in the presence or absence of ␥-secretase inhibitors.

BRI2-A␤1-x Fusion Proteins Efficiently Produce A␤ Peptides
Ranging from 1-38 to 1-55-We had previously employed the BRI2-A␤1-x fusion protein strategy to efficiently generate A␤1-40 and A␤1-42 in the late secretory pathway; additionally we had also shown that a BRI2-C99 (CTF␤) fusion protein efficiently generated C99 that could be subsequently processed by ␥-secretase to A␤ (38,40,48,49). Here we evaluated whether the BRI2 fusion protein strategy could be used to generate A␤ peptides spanning from 1-38 to 1-55 to examine their subsequent proteolyic processing. Expression plasmids encoding the various BRI2-A␤1-x proteins were transiently transfected into HEK-293 cells with A␤ assessed in the cells and in the media by both A␤ ELISAs and IP/MS. Each transfected BRI2 fusion construct resulted in overexpression and secretion of the fulllength A␤1-x peptides as assessed by IP/MS (Fig. 1A). The calculated molecular mass for each peptide and the average observed molecular mass of each peptide are listed in Table 1. Overall, mass accuracy was within 0.05%. To determine whether the secreted A␤1-x peptides were present in exosomes or present as soluble secreted proteins, conditioned media were centrifuged at 100,000 ϫ g for 1 h. A␤1-x peptides remained in the supernatants with no depletion, indicating that they are produced as soluble proteins. Full-length A␤1-x peptides were also detectable in the cell lysate (data shown for a subset of the fusion proteins, Fig. 1B). Western blotting of the cell lysate with BRI2 protein antibody and an antibody recognizing the N terminus of A␤ (6E10) demonstrated that the constructs were efficiently overexpressed (Fig. 1C). In the HEK-293 cell studies, we were not able to detect shorter A␤ cleavage products in the media by IP/MS. However using A␤1-40 and A␤1-42 ELISAs coupled with GSI treatment, we could establish that both A␤1-40 and A␤1-42 were being produced by ␥-secretase processing from constructs longer than A␤1-47 ( Fig. 2A). Compared with other constructs, BRI2-A␤1-51 showed a different production profile, which generated significant amount of A␤x-42. A␤x-42 was detected by ELISA from A␤1-43, 44, 45, 46, and 47, but this likely represents detection of these species by mm2.1.3, the anti-A␤x-42 antibody used for capture. This antibody, like most anti-A␤42 antibodies, shows crossreactivity with longer A␤ peptides. Indeed, GSI treatment did not significantly lower levels of A␤x-42 detected by ELISA from these A␤ species (Fig. 2B).
A␤1-x Peptides Longer than A␤1-48 Are Processed by ␥-Secretase in CHO Cells-In parallel studies using select constructs transfected into CHO cells, we were able to detect processing of the secreted A␤ peptides produced from BRI2-A␤ fusion proteins longer than A␤1-48 by both IP/MS and A␤ ELISA. A␤ ELISA results from CHO cells (both transiently and stably transfected) were similar to those from HEK-293 cells (Fig. 3, A and B). All constructs longer than A␤1-48 appeared to be efficiently processed by ␥-secretase to produce A␤x-40 and x-42 because 1 M of the GSI LY-411,575 almost completely inhibited production of these A␤ peptides (Fig. 3C). Although the small amount of A␤1-40 produced by A␤1-47 or A␤1-48 was also completely inhibited by the GSI, A␤1-42 was not. The lack of complete inhibition of A␤1-42 likely represents some detection of the full-length peptide by the ELISA. By IP/MS from 1 ml of media, intact A␤1-47 from BRI2-A␤1-47 cells and A␤1-48 BRI2-A␤1-48 could be detected. Although BRI2-A␤1-48 was not efficiently processed by ␥-secretase, it appeared to be processed preferentially to A␤1-42, but not other processed A␤ species, and could be detected by IP/MS from 10 ml of conditioned media from BRI2-A␤1-48 cells (Fig. 3D). Although small amounts of full-length A␤1-49 could also be detected from cells expressing the BRI2-A␤1-49 fusion protein, full-length species could not be detected from fusion proteins expressing longer A␤s (A␤1-50 to 1-55). Consistent with the concept of a preferential product line for A␤1-40, A␤1-49 was cleaved predominately to A␤1-40 and a small amount of A␤1-37. A␤1-50 showed a similar profile to A␤1-49, although by ELISA a small amount of A␤1-42 could be detected. In contrast, BRI2-A␤1-51 showed a distinct profile producing a larger proportion of A␤1-42 as assessed by both IP/MS and ELISA. Moreover, minor amounts of A␤1-37, 38,  Table 1). B, selected subset of IP/MS spectra from Triton X-100 cell lysates from the transiently transfected cells expressing the various BRI2-A␤1-x vectors. C, overexpression of the BRI2-A␤1-x constructs as detected by Western blotting of the cell lysates using an anti-BRI2 antibody ITM2b and the anti-A␤ 1-16 antibody 6E10.
39, 40, and 41 were produced from the BRI2-A␤1-51 construct. IP/MS profiles and ELISA data from BRI2-A␤1-52, A␤1-55, and C99 were similar, with each producing a typical spectrum of A␤ peptides with A␤1-40 as the major species and minor amounts of A␤1-37, 38, 39, and 42 detected. To evaluate the potential non-␥-secretase-mediated proteolytic cleavage of the long secreted A␤s, synthetic A␤1-49 was incubated with CHO cells overnight, and then A␤ in the media was analyzed by IP/MS. These data revealed that A␤1-49 is quite stable in the media and not cleaved to smaller A␤ species (Fig. 3E).
Truncated Substrates Alter IC 50 values of Several GSIs but Do Not Alter Effect of GSMs-The effects of two GSIs (1 M of L685,458 and 100 nM MRK560 (50, 51)) on production of secreted A␤1-40 from cells transfected with BRI2-A␤49 to BRI2-A␤55, BRI2-C99, and APP was assessed by ELISA (Fig. 5,  A and B). At these concentrations both GSIs showed much less ability to inhibit A␤1-40 production from the BRI2-A␤ constructs compared with GSI effects on full-length APP and for L685,458; the BRI2-A␤ constructs also were less responsive to the GSI than BRI2-C99. To more fully explore this phenomena, we conducted dose-response studies comparing the effects of L685,458 on total A␤ and A␤1-40 production from transiently transfected HEK 293 cells expressing BRI2-A␤1-49 and APP (Fig. 5C). In these studies, the IC 50 of L685,458 for inhibition of A␤1-40 shifted from 6.9 nM for APP to 362 nM for BRI2-A␤49 and for total A␤ shifted from 4.8 nM for APP to 14 M for BRI2-A␤49. Tests of L685,458 with stably transfected CHO cells show similar effects (Fig. 5D), with the IC 50 for inhibition of A␤1-40 shifted from 65.6 nM for APP to 771.3 nM for BRI2-A␤49. Additional dose-response studies using the GSIs MRK560 and LY-411,575 (52) also show that the IC 50 for inhibition of A␤1-40 production from BRI2-A␤1-49 is greatly increased as compared with APP (Table 2). We also assessed whether similar effects could be observed with in vitro ␥-secretase assays using recombinant C99 and synthetic A␤1-49 as substrate. As shown in Fig. 5E and Table 2, inhibition of A␤1-40 production from A␤1-49 (IC 50 637 nM) was markedly less sensitive to inhibition by MRK560 than inhibition of C99 cleavage (0.3 nM).
In contrast to the loss in potency for multiple GSIs, we found no evidence for differential effects of two potent GSMs (GSM1 and compound 2) and two inverse GSMs (fenofibrate and (Z-LL) 2 ketone) on modulation of A␤1-42 generation from cells expressing BRI2-A␤1-49, 51, and 55 (Fig. 6A). Even though the yield of A␤x-42 from BRI2-A␤49 was very low (5-10 pM, which is close to the detection limit of the ELISAs 1-2 pM), we were still be able to evaluate the modulation. At the concentrations used, all modulatory compounds showed a similar effect on A␤x-42 production as we had previously observed for APP or BRI-C99. The lack of differential modulation was further established by directly comparing the effect of varying concentrations of GSM1 (Fig. 6B) and compound 2 (Fig. 6C) on A␤x-42 production in CHO stable lines expressing BRI2-A␤1-49, -A␤1-51, and -A␤1-55 and C99. No significant difference in dose response to these GSMs was observed.
Because these BRI2-A␤ constructs were still sensitive to modulation, we used IP/MS to explore the changes in secreted A␤ profiles induced by GSM1, compound 2, fenofibrate, and (Z-LL) 2 ketone from cells expressing BRI2-A␤1-49, 1-51, and 1-55. Representative spectra from these studies are depicted in Fig. 7A, and the data are summarized in Fig. 7B. Studies of modulatory effects on BRI2-A␤1-49 showed that GSM1 slightly increased relative levels A␤1-37 and enabled the detection of A␤1-38. Compound 2 dramatically shifted the spectra with large increases in A␤1-36 and 1-37 noted and also a decrease in 1-40. In contrast, both iGSMs appeared to decrease the detection of shorter A␤ peptides but did not increase A␤1-42. The effects of the compounds on the spectra from BRI2-A␤1-51 were more dramatic; both GSM1 and compound 2 decreased A␤1-39, 40, and 42 and increased A␤1-37 and 38, although the effects of compound 2 were more dramatic in terms of reduction of A␤1-39, 40, and 42. iGSMs decreased shorter A␤ levels and increased A␤1-42. For BRI2-A␤1-55, the results were similar to BRI2-A␤1-51, but less dramatic for most compounds. GSM1 lowered A␤1-42 and increased A␤1-37 and 1-38. Compound 2 increased A␤1-36, 37, and 38 and lowered A␤1-39, 40, and 42. Both iGSMs increased the level of A␤1-42.

DISCUSSION
Understanding the precise mechanism by which ␥-secretase generates different A␤ peptide species has both pathologic and therapeutic relevance. Our data, which use both ELISA and IP/MS methods to evaluate ␥-secretase processing of A␤1-x peptides ranging in length from 38 to 55 amino acids in cells, support and extend a number of recent studies demonstrating that stepwise cleavage by ␥-secretase results in the generation of different A␤1-x peptides (24, 27, 34 -36). Although our data in general are consistent with the stepwise model of ␥-secretase cleavage of APP along different product lines that preferentially produce A␤1-40 or A␤1-42 (34), they demonstrate a number of novel observations regarding the solubility of these long A␤ peptides and their processing by ␥-secretase. First, we find that following transient transfection in HEK cells, all of the A␤ pep-tides are efficiently secreted as intact proteins that are soluble and easily detectable by IP/MS. Notably, both the long peptides derived from the BRI constructs and synthetic A␤1-49 are quite stable in the media. This result is unexpected given that A␤s longer than A␤1-42 have been reported to be difficult to detect by mass spectrometry and might be expected to be membrane-associated because they contain a portion of or the entire  transmembrane domain of APP. Although we do detect fulllength A␤1-47 and A␤1-48 secretion from the stable CHO cell lines, we do not detect longer full-length A␤ peptides. The reasons for this difference between the two different experimental paradigms is not clear at this time but perhaps might reflect subtle differences in membrane lipid composition influencing both ␥-secretase processing and membrane binding; it has been shown that lipid composition can influence ␥-secretase processivity (53). Because these long A␤ peptides are challenging to produce synthetically, the ability to produce soluble long A␤ peptides in HEK cells could be exploited as a source of standards for other biological studies. Second, although shorter A␤ peptides can be shown to be processed in vitro by detergentsolubilized ␥-secretase, we find that in cells, peptides shorter than A␤1-49 are not efficiently processed by ␥-secretase; pre-sumably they do not efficiently insert into intact non-detergent-solubilized membranes. In stable CHO cell lines, A␤1-47 and 1-48 undergo very limited ␥-secretase processing, most of the secreted peptides detected in these cells are the full-length peptides and not shorter A␤s. However, based on GSI inhibition studies, it does appear that the small amount of A␤1-40 and 1-42 produced is generated by ␥-secretase cleavage. In contrast, A␤1-49 is efficiently cleaved in the stable CHO line to A␤1-40 and 1-37. Third, although our data support the concept of product lines with the initial ⑀-cleavage generating A␤1-49 cleavage primarily leading to A␤1-40 and initial ⑀-cleavages at A␤1-48 or 1-51 generating primarily A␤1-42, it is clear that these product lines are not completely distinct. A␤1-49 generates small amounts of A␤1-42, detectable by ELISA; A␤1-48 generates A␤1-37 and small amounts of A␤1-40 and 1-42; and A␤1-51 generates A␤ peptides 1-37 to 1-42. Notably, the diversity of species generated from A␤1-51 would suggest that when produced in cells, the stepwise cleavage is more heterogeneous than what is observed for A␤1-49. When BRI2-A␤1-49 was co-transfected with either the M139V or ⌬exon9 PSEN1 mutants, neither altered the ratio of A␤42:40 from BRI2-A␤1-49. In the present study, M139V only increased A␤42 from BRI2-C99 and BRI2-A␤1-51 but not from BRI2-A␤1-49. Although these data do not evaluate the effects on ⑀-cleavage, our data suggest that M139V and ⌬exon9 mutants alter A␤42 by specifically inhibiting processivity by one cycle along the A␤1-48/51 but not the A␤1-49 production line. We would note that previous studies on M139V have led to different conclusions; a cell-free assay had shown that M139V does not alter ⑀-cleavage efficiency but does impair the fourth stepwise cleavage along both product lines when C99 was used as an in vitro substrate (36). In a separate report using a cell model, M139V increased A␤42 but also reduced ⑀-cleavage efficiency (54). However, our present results are consistent with a previous study from our group where we found that the M139V mutant selectively raised A␤42 but did not alter levels of other peptides in CHO cells stably overexpressing this mutant (55). Discrepancies between these results might be explained by the variable methodology, especially in cell culture systems where there are admixtures of wild-type and mutant PSEN1.
Another unexpected result from our study was the finding that the long A␤ peptides that are efficiently processed by ␥-secretase in cells (e.g., A␤1-49 to 1-55) are much less sensitive to inhibition by various GSIs. This loss of sensitivity to GSIs was observed in cells and in vitro using recombinant substrates. The basis for the differential sensitivity of these long A␤ peptides to GSI inhibition is not clear. However, given the rather dramatic effects observed, in some cases as much as a 300-fold differential sensitivity between A␤1-49 and C99 in vitro, these data suggest that IC 50 measures established for a GSI with any truncated substrate for ␥-secretase could be misleading. This could be of importance when investigating effects of GSIs on other substrates; if one does not establish an IC 50 for cleavage using a full-length substrate, such data could prove to be misleading. Indeed, some GSIs that have been reported to be "Notch-sparing" appear to have no substrate selectivity when evaluated using identical assay conditions in vitro (36,56,57).
In contrast to the differential sensitivity to GSIs, long A␤ peptides appear to remain sensitive to modulation of cleavage by both GSMs and iGSMs. Previous findings have shown that acidic (e.g., GSM1) and nonacidic GSMs (e.g., compound 2) have different modulatory effects. For the most part, acidic GSMs selectively lower A␤1-42 and increase A␤1-38, whereas nonacidic GSMs lower both A␤1-40 and 1-42 and raise both A␤1-37 and 1-38 (44,58). Our current data provide some insight into the differences between these two GSM classes that  Although the effects of iGSMs on A␤1-42 derived from A␤1-49 were only detectable by ELISA, iGSMs showed a consistent effect on the three longer A␤ substrates decreasing the shorter A␤ peptides and increasing A␤1-42. Overall these data suggest a model whereby the acidic GSMs preferentially enhance processivity along the A␤1-42 product line, whereas nonacidic GSMs promote processivity along both A␤1-40 and 1-42 product lines, with iGSMs decreasing processivity in both product lines. Building on the pioneering work of Ihara and co-workers (24), multiple groups have now explored the mechanisms through which ␥-secretase generates A␤ peptides with differing C termini. Using multiple different systems, ranging from homogenous purified reconstituted ␥-secretase assays to cell culture studies, these studies provide detailed insights into how ␥-secretase cleavage of APP generates A␤ peptides with differing C termini and how various factors such as mutations in APP, PSEN1/2, small molecules, or other factors can alter the profile of A␤ produced (35)(36)(37)61). Although the data generated from the different systems is not identical, the differences for the most part are quantitative in nature and not qualitative. Through both heterogeneous ⑀-cleavage and differential stepwise processing, there can subtle to quite dramatic effects on the final profile of A␤ peptides produced. For example, several studies now show that FAD-linked mutations in APP and PSEN1 and PSEN2 alter ␥-secretase by altering ⑀-cleavage site utilization, processivity, or some combination of these two mechanisms (35,36). In contrast, the action of GSMs and iGSMs seems to be restricted to effects on processivity, with GSMs promoting additional stepwise cleavages and iGSMs partially inhibiting the stepwise cleavage (36). Further study of ␥-secretase cleavage of other substrates and the effect of GSMs and iGSMs using methods such as those established here should provide additional data regarding whether the cleavage mechanism observed for APP is utilized for other substrates.