A Synthetic Antibody Fragment Targeting Nicastrin Affects Assembly and Trafficking of γ-Secretase*

Background: γ-Secretase is critical in Notch signaling and Alzheimer disease. Results: Coexpression of a nicastrin antibody affects maturation and trafficking of γ-secretase and alters Notch and APP processing. Conclusion: Nicastrin is a feasible target for modulating the targeting and substrate specificity of γ-secretase. Significance: Modulation of γ-secretase through nicastrin may be of therapeutic value in cancer and Alzheimer disease. The γ-secretase complex, composed of presenilin, nicastrin (NCT), anterior pharynx-defective 1 (APH-1), and presenilin enhancer 2 (PEN-2), is assembled in a highly regulated manner and catalyzes the intramembranous proteolysis of many type I membrane proteins, including Notch and amyloid precursor protein. The Notch family of receptors plays important roles in cell fate specification during development and in adult tissues, and aberrant hyperactive Notch signaling causes some forms of cancer. γ-Secretase-mediated processing of Notch at the cell surface results in the generation of the Notch intracellular domain, which associates with several transcriptional coactivators involved in nuclear signaling events. On the other hand, γ-secretase-mediated processing of amyloid precursor protein leads to the production of amyloid β (Aβ) peptides that play an important role in the pathogenesis of Alzheimer disease. We used a phage display approach to identify synthetic antibodies that specifically target NCT and expressed them in the single-chain variable fragment (scFv) format in mammalian cells. We show that expression of a NCT-specific scFv clone, G9, in HEK293 cells decreased the production of the Notch intracellular domain but not the production of amyloid β peptides that occurs in endosomal and recycling compartments. Biochemical studies revealed that scFvG9 impairs the maturation of NCT by associating with immature forms of NCT and, consequently, prevents its association with the other components of the γ-secretase complex, leading to degradation of these molecules. The reduced cell surface levels of mature γ-secretase complexes, in turn, compromise the intramembranous processing of Notch.

In view of evidence that NCT plays a critical role in substrate recognition (14,21), we and others have examined the feasibility of modulating NCT function using antibody-based approaches (21)(22)(23)(24) with the notion that NCT-specific antibodies could bind to and modulate the binding of NCT to individual substrates. In this regard, we reported that a synthetic antibody fragment that targets the NCT ectodomain (ECD) impairs the ␥-secretase-mediated processing of both Notch and APP in in vitro assays (21). In this study, we generated additional NCT-specific synthetic antibodies using phage display technology and then reformatted the cDNAs encoding these antibodies to corresponding cDNAs encoding singlechain variable fragments (scFvs) (25) that were then stably expressed in HEK293 cells that constitutively express the APP "Swedish" (APPSwe) variant that causes early onset familial AD (26). We now describe the analysis of two anti-NCT-specific antibodies that, following conversion to scFvs, bind to the NCT ECD in vitro. In HEK293 cells that stably express one of these antibodies, termed scFvG9, NCT maturation is impaired and leads to reduced level on the cell surface, where the Notch derivative, NEXT, is subject to ␥-secretase processing at the "⑀" site. On the other hand, ␥-secretase that is present in the Golgi apparatus and endosomal/recycling compartments in cells expressing scFvG9 is fully capable of processing APPSwe to generate A␤ peptides. Therefore, we suggest that scFvG9 affects both the maturation of NCT and subcellular distribution of ␥-secretase that leads to differential processing of APP versus Notch.

EXPERIMENTAL PROCEDURES
Cell Lines, cDNA Constructs, and Transfection-Full-length human NCT was C-terminally tagged with a CT11 tag (27). The entire ECD segment or a region corresponding to exons 7-16 (716) of nicastrin were C-terminally tagged with a His 6 tag (21). The mouse N⌬E construct (mN⌬E) was C-terminally tagged with a myc 6 tag (28). HEK293 cells and HEK293 cells stably expressing either wild-type human APP or the human APP Swedish variant were stably transfected with an empty vector or cDNAs encoding an scFv using Lipofectamine Plus reagent (Invitrogen). Stable cell pools were selected and maintained in the presence of 200 g/ml zeocin (Invitrogen). HEK293S GnT1 Ϫ cells (29) and HEK293 cells were maintained in DMEM containing 10% FBS and 1% PS (Invitrogen). To assess ␥-secretase activity in HEK293 cells that stably express APPSwe and scFv, cDNA encoding mouse N⌬E was transiently transfected into these cell pools for 48 h before detergent-solubilized cell lysates were prepared for analysis.
Immunoblot Analysis and Antibodies-Cells were lysed in a buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Nonidet P-40, 0.5% sodium deoxycholate, 5 mM EDTA and protease inhibitor mixture (Sigma). Protein concentrations were determined by BCA kit (Thermo Scientific, Rockford, IL). Equal amounts of protein lysates were resolved on SDS-PAGE and transferred to a nitrocellulose membrane. After blocking, the membrane was sequentially incubated with primary and secondary antibodies, and the secondary antibodies were detected with ECL (PerkinElmer Life Sciences). PS1 NT antibody was used to detect full-length PS1 and the N-terminal fragment of PS1 (30). MAB5232 was used to detect the PS1 C-terminal fragment (EMD Millipore, Billerica, MA). PNT-2 antibody (Dr. Gopal Thinakaran) was used for the detection of PEN-2 protein (30). H2D antibody (Dr. Gang Yu) was used to detect endogenous APH-1aL (31). CT11 antibody was used to detect CT11-tagged NCT (30). Nicastrin (N-19) antibody (Santa Cruz Biotechnology) was used to detect endogenous NCT. 9E10 (Santa Cruz Biotechnology) was used to detect myc 6 -tagged mN⌬E and NICD fragments as well as the scFv proteins. Anti-His 6 antibody (Rockland Immunochemicals) was used to detect His 6 tagged ECD, 716, as well as scFv protein. CTM1 polyclonal antibody was used for the detection of fulllength APP and APP CTFs (21). 26D6 monoclonal antibody was used to detect APPs and A␤ (32). 4G8 monoclonal antibody (Covance) was used to immunoprecipitate A␤ from conditioned medium. Actin antibody was used to detect endogenous actin (Santa Cruz Biotechnology).
Synthetic Antibody Generation and Construction of scFv Vectors-Purification of secreted NCT fragments, screening, and expression of anti-nicastrin synthetic antibodies have been described previously (21), except that we used a new antibody phage display library (33) in this study. cDNAs encoding single chain variable fragments were generated by multiple rounds of PCR reactions. Heavy chain and light chain sequences of NCTspecific Fabs A9 and G9 as well as those of the negative control Fab2-2 were used as templates for the amplification of VH and VL regions by PCR. The VH region was amplified using the following primers: human transthyretin-VH, 5Ј-GTATTTGT-GTCTGAGGCTGGCCCTACGGGCACCGGTGAGATCTC-CGAGGTTCAGCTG-3Ј (forward); LK-VH, 5Ј-GCCGCCAG-AACC GCCGC CACCAGAGCCACCACCACCGGCCGAG-GAGACGGTGACCAGGGT-3Ј (reverse). The VL region was amplified using the following primers: LK-VL, 5Ј-GGCTCTG GTGGCGGCGGTTCTGGCGGCGGCGGTTCTTCCGAT-ATCCAGATGACCCAGTCC-3Ј (forward); Xho-VL, 5Ј-AGC-TCTCGAGAGTTCGTTTGATCTCCACCTTGGTACC-3Ј (reverse). The VH and VL regions were linked by a 15-amino acid peptide containing the three repeats of GGGGS. The human transthyretin signal peptide sequence was fused to the N terminus of the VH region, and a myc-His 6 tag was attached to the C terminus of the VL region. The resulting cDNA was cloned into the pAG vector.
Immunoprecipitation of Secreted NCT Fragments with Anti-NCT Fabs or scFvs-Conditioned medium from cells secreting ECD and 716 fragments was subjected to immunoprecipitation with Fabs 2-2, 2, 12, A9, and G9 essentially as described earlier (21). In this study, a mixture of protein A and protein L beads (Thermo Scientific) was used for the capture of Fab-antigen complexes. To obtain secreted scFv proteins, cDNAs encoding scFv 2-2, scFvA9, and scFvG9 were transiently transfected into HEK293 cells for 48 h, and the conditioned media were collected. To assess the interaction between the ECD and these scFv proteins, conditioned medium from stable ECD-expressing cells was mixed with the conditioned medium containing secreted individual scFv protein. The mixture was adjusted with 5ϫ PBS to a final concentration of 1ϫ PBS and incubated at 4°C overnight. The ECD-scFv complex was captured with protein A beads (Thermo Scientific) and detected by immunoblot analysis with anti-His 6 antibody.
Pulse Labeling-To assess the conversion of full-length mouse N⌬E into the NICD fragment, HEK293 APPSwe/scFv cells were transfected with cDNA encoding mouse N⌬E and labeled for 40 min with 250 Ci [ 35 S]methionine (PerkinElmer Life Sciences). Then the cells were chased for 2 h. In experiments to analyze the processing of APP, HEK293 APPSwe/scFv cells were either labeled for 10 min or 4 h with 250 Ci [ 35 S]methionine. Cells were lysed in lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, and 0.5% sodium deoxycholate), and the total radioactivity (cpm) was determined after TCA precipitation by a liquid scintillation counter. Cell lysates of equal cpm were used for the immunoprecipitation of mouse N⌬E, NICD, and full-length APP and its derivatives. Immunoprecipitated mouse N⌬E and NICD fragments were resolved on 6% SDS-PAGE; soluble APP was resolved on 10% SDS-PAGE; and full-length APP, APP CTFs, and A␤ peptides were resolved on a Tris-Tricine gel. Resolved proteins were then transferred to a nitrocellulose membrane and exposed to a PhosphorImager screen or to an x-ray film. The PhosphorImager screen was scanned using a Storm Imager (GE Healthcare). Imagequant 5.0 software was used for the image analysis.
Quantification of A␤ Peptides by ELISA-A␤38, A␤40, and A␤ 42 peptides in conditioned media of HEK293 APPSwe/scFv stable cells were quantitated by ELISA utilizing the Meso Scale Discovery A␤ triplex immunoassay with 4G8 detection (catalog no. K15141E-1, MSD, Rockville, MD). The kit was run according to the instructions of the manufacturer. Briefly, 25 l of MSD SULFO-TAG-conjugated 4G8 detection antibody, and 25 l of sample was added to a blocked MSD multiplex plate precoated with capture antibodies specific for A␤ 38, A␤ 40, and A␤ 42 and incubated for 2 h with shaking at room temperature.
The plate was read on an MSD SECTOR Imager 2400 after addition of read buffer. Concentrations of A␤38, A␤40, and A␤42 in samples were back-fitted from the respective standard curves using MSD Workbench software.
In Vitro ␥-Secretase Activity Assay-Membranes were prepared from HEK293swe/scFv-stable cells. Reactions were per-formed at 37°C, and the production of NICD was determined using AlphaLISA signals (21).
Mass Spectrometry Analysis of Immunoprecipitated A␤ Peptides-Immunoprecipitation MS measurement of A␤ peptides was carried out as described previously using 4G8 antibody (34). A␤ peptides in 300 l of conditioned medium from HEK293 cells stably expressing APPSwe were immunoprecipitated by incubating overnight with 1.5 l of 4G8 antibody and 5.0 l of protein A/protein G plus beads. Mass spectra were collected using a TOF/TOF 5800 mass spectrometer (AB Sciex). Each mass spectrum was accumulated from 2000 laser shots and calibrated using bovine insulin (an internal mass calibrant).
Coimmunoprecipitation of ␥-Secretase Complex Subunits under Native Conditions-HEK293 APPSwe/scFv cells were lysed in a buffer containing 50 mM Tris-Cl (pH 8), 150 mM NaCl, 1% CHAPSO, 1 mM DTT, 5 mM EDTA, and 1ϫ protease mixture (Sigma). For immunoprecipitation, 3 l of 9E10 or N-19 antibody or 1 l of PS1 NT antibody was added to each reaction, respectively. A comparable amount of normal mouse IgG or rabbit IgG was added to the control reactions. The mixtures were then incubated overnight in a cold room. The immunoprecipitated complexes were captured either by protein G beads (for 9E10 and N-19) or by protein A beads (for PS1 NT ) that were washed three times with lysis buffer. ␥-Secretase complexes were released from the beads by incubation in Laemmli buffer at 37°C for 15 min.
Cell Surface Biotinylation-HEK293 APPSwe/scFv cells were grown to near confluence in a 10-cm plate and subjected to cell surface biotinylation with 0.8 mM sulfosuccinimidyl-2-[biotinamido]ethyl-1,3-dithiopropionate (Thermo Scientific) essentially as described previously (35). Cells were then lysed in immunoprecipitation buffer, and biotinylated proteins were captured with NeutrAvidin beads (Thermo Scientific). Biotinylated molecules were eluted from the beads by incubation in Laemmli buffer at 37°C for 15 min.
Quantification of Levels of PS1NTF, PS1 CTF, PEN-2, NCT, APH-1, and Full-length APP by ImageJ-Immunoblot images were scanned, and the densities of ␥-secretase components, APP, and actin were quantified by ImageJ. Levels of ␥-secretase components and APP were normalized to the levels of actin. Data are represented as mean Ϯ S.E.

Generation of Synthetic Antibodies
Recognizing the Nicastrin ECD-We have previously reported the identification and characterization of two NCT-specific synthetic antibodies, Fab2 and Fab12, using phage display approaches (21,25). In this work, we identified two additional anti-NCT antibodies using an NCT fragment encompassing the entire ectodomain, termed ECD, purified from the conditioned medium of N-acetylglucosaminyltransferase 1-deficient (GnT1 Ϫ ) HEK293S cells that stably express the ECD fragment, as the antigen. To map the epitopes of these Fabs, termed FabA9 and FabG9, we performed immunoprecipitation studies under native conditions using His 6 -tagged ECD or an NCT fragment encompassing sequences encoded by exons 7-16, termed 716. As we showed earlier (21), Fab12 and Fab2 immunoprecipitated ECD (Fig. 1A, top panel, lanes 3 and 4), whereas Fab2, but not Fab12, also recognized 716. A negative control, Fab2-2, failed to immunoprecipitate either ECD or 716, as expected (Fig.  1A, lane 2). In contrast, FabA9 and FabG9 recognized ECD but not 716, suggesting that these new Fabs recognize epitopes in the region encoded by exons 1-6. To examine whether these new Fabs bind to linear epitopes or conformational epitopes, purified ECD and 716 fragments were subjected to Western blot analysis using FabA9 or FabG9 as primary antibodies. FabA9 detected a denatured ECD fragment on the blot (Fig. 1B, top panel, lane 1), suggesting recognition of a linear epitope. In contrast, FabG9 failed to detect the NCT fragment in the Western blot, strongly suggesting that it recognizes a conformational epitope. Therefore, we generated two novel antibodies to the NCT ECD.
Stable Expression of scFvG9 in HEK293 Cells Decreases the Steady-state Levels of All ␥-Secretase Components Except APH-1-To examine the function of NCT-specific synthetic antibodies in mammalian cells, we reformatted the Fabs into scFvs. The scFvs were fused to human transthyretin signal peptide to target the polypeptide to the lumen of the endoplasmic reticulum. To facilitate scFv detection and purification, we attached myc and His 6 epitope tags at the very C terminus of the scFv. As a negative control, we prepared scFv 2-2, a non-binding antibody. The expression of scFvs and their secretion into conditioned medium were confirmed by immunoblot analysis (Fig. 1C, lanes [3][4][5]. To confirm the function of scFvA9 and scFvG9, we tested their ability to capture ECD. Conditioned medium containing ECD and those from HEK293 cells expressing individual scFv were mixed and incubated at 4°C overnight. A significant amount of ECD was immunoprecipitated by both scFvA9 and scFvG9 (Fig.  1C, lanes 4 and 5), confirming that both antibodies retain their binding function in the scFv format.
We then expressed scFvA9 and scFvG9 in naïve HEK293 cells and examined their effects on the maturation of endogenous NCT and levels of ␥-secretase components in stable pools. We chose to generate stable pools to avoid significant caveats associated with variable expression between individual cell lines and genomic insertion effects on endogenous genes. In the case of stable pools, these issues are essentially "normalized." In the stable pools, the most striking observation is the accumulation of immature NCT and a reduction in levels of mature NCT in scFvG9 cells compared with naïve 293 cells or cells expressing scFvA9 ( Fig. 2A, first panel, compare lane 4 with lanes 1 and   3). In parallel, we observed a reduction in steady-state levels of PS1 and PEN-2 but with no obvious decrease in the levels of APH-1 in lysates prepared from scFvG9 cells compared with naïve 293 cells ( Fig. 2A, second through fifth panels, compare lanes 1, 3, and 4). To examine the effects of expressing the scFvs on APP processing, we stably transfected HEK293 cells that constitutively express WT APP or the familial AD-linked APPSwe variant with cDNAs encoding scFvA9, scFvG9, and scFv2-2 (control). As we observed in naïve 293 cells, APPwt cells and APPSwe cells stably expressing scFvG9 also exhibited partial impairment of NCT maturation ( Fig. 2A, first panel,  lanes 8 and 12), which was paralleled by a reduction in the steady-state levels of PS1 and PEN-2 ( Fig. 2A, second, fourth, and fifth panels, lanes 8 and 12).
Expression of scFvG9 Does Not Enhance the Turnover of ␥-Secretase Components-To assess the possibility that scFvG9 binding to NCT might lead to destabilization of the ␥-secretase complex, we treated APPSwe/scFv stable cells with the protein synthesis inhibitor cycloheximide to examine the turnover rate of ␥-secretase components. Immunoblot analysis showed that PS1 NTF, a subunit of the mature ␥-secretase complex, was as stable in scFvG9 cells as in other stable cell pools (Fig. 3A, first  row, and B). Similarly, NCT also exhibited a similar level of stability in scFvG9 cell pools as in other stable cell pools (Fig.  3A, first row, and B). On the other hand, cycloheximide treatment lead to a rapid decrease in full-length APP in all cell pools, with a half-life of ϳ2 h, as described previously (36) (Fig. 3A,  third row, and B). To provide additional confirmation of these results, we then asked whether the levels of ␥-secretase components could be elevated by the proteasome inhibitor MG132. In this case, we observed a modest elevation of PS1 NTF in all stable cell pools after treatment over 8 h, which likely reflects new synthesis of PS1 and conversion (Fig. 3, C and D). Collectively, these results suggest that, in scFvG9 cells, the stability of mature ␥-secretase components is no different from that observed in other stable cells analyzed, and we infer that the decrease in steady-state levels of ␥-secretase complex components observed in scFvG9 cells is likely the result of rapid degradation of immature NCT molecules following synthesis, leading to reduced association and, hence, stabilization of subunits of the ␥-secretase complex. This view further suggests that scFvG9 binds to NCT early in its biosynthetic processes.
Expression of scFvG9 Decreases Notch Processing in HEK293 Cells-To assess the impact of these scFvs on ␥-secretase activity, we first analyzed Notch processing in the APPSwe cells that stably express vector, scFv2-2, scFvA9, or scFvG9. For these studies, we transiently transfected APPSwe/scFv stable cells with cDNA encoding myc6-tagged mN⌬E, a Notch derivative corresponding to NEXT (28). Western blot studies revealed that ␥-secretase-mediated cleavage of NEXT at the ⑀ site that leads to production of NICD was reduced in scFvG9 stable cells compared with the other stable cells (Fig. 4A, top panel, compare NICD in lanes 7 and 8 with lanes 1-6). To confirm this result, we performed pulse-chase analyses. The results revealed that NICD production is reduced in scFvG9 cell pools compared with the other cell pools (Fig. 4A, center and bottom panels, compare NICD in lanes 7 and 8 to lanes 1-6). The production of NICD during the 40-min pulse in scFvG9 cell pools was about 26% of that in vector pools (Fig. 4B), and the production of NICD in scFvG9 cell pools after the 2-h chase was about 56% of that in vector pools (Fig. 4C). We considered the possibility that the reduction in NICD production in these cellular assays may reflect alterations in the trafficking of ␥-secretase to subcellular compartments where the Notch derivative, NEXT, is processed to NICD. Therefore, we analyzed the production of NICD in in vitro ␥-secretase activity assays in which whole cell membranes prepared from HEK293Swe/scFv double-stable cells were incubated with a truncated Notch substrate similar to NEXT. These assays confirmed a significant reduction in the production of NICD in membranes prepared from scFvG9 cells compared with membranes prepared from all other cell pools (Fig. 4D). These findings argue that the total ␥-secretase activity is reduced in these cells.
Expression of scFvG9 Has No Effect on A␤ Production but Impairs the Generation of AICD-We next examined APP processing in APPSwe cells that stably express vector, scFv2-2, scFvA9, or scFvG9 and observed a modest accumulation of the ␣and ␤-secretase-generated APP CTFs in scFvG9 cells compared with CTFs in the other cell pools (Fig. 5A, second panel,  lane 4), suggesting an impairment in ␥-secretase processing in these cell pools. Parallel analysis of full-length APP in cell lysates and soluble APP in conditioned medium showed no differences between each of the pools (Fig. 5A, first and third  panels). Much to our surprise, immunoblot analysis failed to reveal any differences in the levels of secreted A␤ in the conditioned medium of the cell pools (Fig. 5A, fourth panel). Consistent with the immunoblot analyses, ELISA assays also revealed that the levels of secreted A␤38, A␤40, and A␤42 in the conditioned medium of all APPSwe/scFv stable cell pools were indistinguishable (Fig. 5B). APP CTFs are also subject to processing by ␥-secretase at the ⑀ site (37), which is distal to the carboxyl terminus of A␤ peptides, to generate 49-to 50-amino acid APP intracellular domain fragments, termed AICDs, that are virtually undetectable because of rapid intracellular turnover. To visualize this APP derivative, we performed pulse-chase analyses. APPSwe/scFv stable cells were pulse-labeled with [ 35 S]me-  11 and 12). Empty vector and non-binding scFv2-2 were used as controls. GnT1 Ϫ HEK293S cell lysate was used as a control for underglycosylation of NCT (lane 2). Immunoblot analyses were performed on cell lysates from these stable cells. The expression levels of endogenous PS1 NFT, PS1 CTF, APH-1, PEN-2, and NCT are shown in each panel. Stably expressed scFv2-2, scFvA9, and scFvG9 were detected by anti-myc antibody. The actin level was used as a loading control. B-F, quantification of steady state levels of PS1 NTF, PS1 CTF, PEN-2, NCT, and APH-1, respectively. Data were collected from APPwt/scFv stable cell pools and APPSwe/scFv stable cell pools. Levels of ␥-secretase components were normalized to the actin level and are shown as the percentage of that in vector stable cell pools. Data are mean Ϯ S.E. (n ϭ 2). mat, mature; imm, immature. thionine for either 10 min or 4 h. In the 10-min pulse-labeled samples, the levels of immature full-length APP were similar in all cell pools (Fig. 5C, first panel). Similarly, after 4 h of labeling, the levels of both immature and mature APP in cell lysates (Fig.  5C, second panel) and soluble APP derivatives (APPs and A␤) in the conditioned medium were essentially identical between all the cell pools (Fig. 5C, fifth and sixth panels). Notably, the maturation of APP was unaffected in scFvG9 cells compared with the other cell pools (Fig. 5, A, first panel, and C, second panel), corroborating the selective effect of scFvG9 on NCT maturation. The increase of APP CTF levels in scFvG9 cells (Fig. 5A) was confirmed by radiolabeling and immunoprecipitation (Fig.  5C, third panel, lanes 7 and 8). The sensitivity of the prolonged 4-h labeling allowed us to detect AICD in all of the stable cells, and, importantly, the AICD levels in radiolabeled scFvG9 cell lysates were lower than those found in other stable cell pools (Fig. 5C, fourth panel, compare AICD in lanes 7 and 8 to lanes  1-6). Quantification of the labeled AICD fragment showed an ϳ54% reduction in scFvG9-expressing cells (Fig. 5D), similar to the reduction in the generation of NICD (Fig. 4). Finally, immunoprecipitation studies demonstrated that the levels of newly synthesized, radiolabeled, A␤-related peptides that were secreted from the stable cell pools were equal between the cell pools, findings that fully corroborate the results obtained by Western blot analysis (Fig. 5A) and ELISA (Fig. 5B) studies of accumulated A␤ in the medium. Although we did not observe any significant differences in levels of A␤ between the stable cell pools, two polypeptides that migrated between the A␤ and p3 appeared to be more pronounced in the conditioned medium of scFvG9 cells compared with the conditioned medium of the other cell pools (Fig. 5C, sixth panel). To determine the spectrum of secreted A␤-related peptides, we immunoprecipitated A␤-related species from the conditioned medium and subjected the recovered material to MALDI-TOF mass spectrometry. This approach failed to identify any differences in the spectrum of A␤-related peptides between cells expressing scFvG9 and cells expressing scFv2-2 or scFvA9 (Fig. 5E). scFvG9 has no perceptible effect on ␥-secretase-mediated APP processing to generate A␤ peptides, but expression of this scFv reduces AICD production.
scFvG9 Interacts with Immature but not Mature ␥-Secretase Complexes-The studies shown in Fig. 3, A and B, clearly demonstrate that the reduced levels of ␥-secretase components in scFvG9 cell pools (Fig. 2) is not the result of enhanced degradation or decreased stability of the mature complex. Therefore, we asked whether the effect of scFvG9 on the steady-state levels of ␥-secretase components may occur at an earlier stage of complex assembly. To address this issue, we solubilized membranes from APPSwe/scFv stable cells with 1% CHAPSO, conditions under which ␥-secretase remains assembled (38). As expected, anti-myc antibody immunoprecipitated scFv2-2, scFvA9, and scFvG9 from corresponding scFv-expressing cell lysates (Fig. 6A, sixth row, lanes 7, 11, and 15), whereas scFv2-2, a negative control antibody, did not coimmunoprecipitate any endogenous ␥-secretase component (Fig. 6A, lane 7). scFvA9 immunoprecipitated only a small amount of endogenous, immature NCT (Fig. 6A, first row, lane 11) but no other ␥-secretase component (Fig. 6A, lane 11). In contrast, scFvG9 immunoprecipitated elevated levels of immature NCT species (Fig.  6A, first row, lane 15). Moreover, scFvG9 coimmunoprecipitated full-length PS1 and APH-1 that would be expected if early assembled NCT-associated subcomplexes are being recognized. In support of this view, scFvG9 failed to coimmunoprecipitate PEN-2 or PS1 NTF, components of mature, active ␥-secretase complexes (Fig. 6A, lane 15). Conversely, an anti-NCT N-terminal antibody immunoprecipitated endogenous NCT from all APPSwe/scFv stable cell lysates (Fig. 6A, first row,  lanes 4, 8, 12, and 16), but, more importantly, the only scFv that was coisolated was scFvG9 (Fig. 6A, sixth row, lane 16). To validate the findings above, we performed coimmunoprecipitation studies with an antibody raised against an amino-terminal epitope of PS1. As expected, this antibody coimmunoprecipitated PS1 CTF and other ␥-secretase components from all of the APPSwe/scFv stable cells (Fig. 6B, lanes 3, 6, 9, and 12), but in no case did we detect scFv coisolated with this antibody (Fig.  6B, fourth row, lanes 3, 6, 9, and 12). These results strongly suggest that scFvG9 binds to immature NCT in isolation and/or in immature ␥-secretase complexes.
On the basis of the cycloheximide/MG132 experiments (Fig.  3) and coimmunoprecipitation studies (Fig. 6, A and B) and the earlier demonstration that NCT maturation was impaired in scFvG9 cells (Fig. 2), we asked whether the binding of scFvG9 to either NCT alone or to NCT-containing immature complexes leads to retention of the complexes in intracellular compartments of the secretory pathway, therefore excluding residence on the cell surface where intramembranous processing of NEXT occurs (39). To test this possibility, we used the membrane-impermeable cross-linker sulfosuccinimidyl-2-[biotinamido]ethyl-1,3-dithiopropionate to assess the levels of ␥-secretase complexes on the surface of APPSwe/scFv stable cells. Although the total levels of NCT in all scFv cells were similar, the levels of surface NCT in scFvG9 cells were reduced significantly (Fig. 6C, compare lane 4 with lanes 1-3 and lane 8  with lanes 5-7). Collectively, these findings suggest that, although APP-CTFs can be efficiently processed by ␥-secretase complexes that are present in endosomal and/or recycling compartments in cells expressing scFvG9, these complexes fail to mature and accumulate on the cell surface, therefore limiting intramembranous processing of NEXT.

DISCUSSION
The ␥-secretase complex, composed of PS, PEN-2, APH-1, and NCT, catalyzes the intramembranous proteolysis of over 90 membrane-tethered substrates, including APP and Notch. Although ␥-secretase inhibitors have been developed that effectively block the production of pathogenic A␤ peptides in vivo, clinical trials of such agents have been terminated because of Notch-based liabilities. Other strategies under consideration include the development of ␥-secretase "modulators" that inhibit processing of APP to generate A␤ peptides but spare processing of Notch (6). Both classical ␥-secretase inhibitors and ␥-secretase modulators appear to target PS, the catalytic center of the enzyme complex (40). Alternatively, we and others have examined the feasibility of modulating NCT function using antibody-based approaches (21)(22)(23)(24) with the notion that NCT-specific antibodies could bind to and modulate the binding of NCT to individual substrates. The rationale for this approach is the finding that the activity of ␥-secretase-containing NCT subunits harboring a series of experimentally generated mutations can differentially affect the processing of APP versus Notch (41).
In this report, we examined the mechanism of action of two novel anti-NCT Fabs and offer several important insights regarding their modulation of ␥-secretase function. First, we show that one of these, FabG9, binds to a conformational epitope in an amino-terminal segment of NCT contained within the region encoded by exons 1-6 and mammalian cells that stably express the scFv derivative, scFvG9, leading to lowered steady-state levels of PS, PEN-2, and mature NCT glycoforms. These results resemble the findings by Hayashi et al. (22) that described a NCT-specific scFv, termed 5201F, but we note several significant differences. First, 5201F was shown to bind to multiple regions of the NCT ECD. Expression of this antibody in mammalian cells lowered the levels of mature NCT, PEN-2, PS, and APH-1, whereas we did not observe any changes in APH-1 in cells expressing scFvG9. The levels of immature NCT are elevated in scFvG9 cells, a finding that does not parallel those seen in cells expressing 5201F, where both immature and mature forms of NCT were virtually depleted. Second, we report that, although the levels of mature NCT, PS, and PEN-2 are reduced in scFvG9 cells, treatment of these cells with the proteasome inhibitor MG132 fails to elevate the steady-state levels of these polypeptides over the treatment period. Therefore, we argue that reduced expression of the ␥-secretase components is the result of enhanced degradation of these polypeptides immediately after biosynthesis. Unfortunately, the synthetic rate of ␥-secretase components is extremely low (42), therefore precluding any estimation of degradation rates following synthesis of these polypeptides. Third, we demonstrate that intramembranous processing of Notch to generate NICD is reduced markedly in scFvG9 cells. Moreover, the production of the APP derivative AICD is also reduced in these cells. Most interestingly, the processing of APP to A␤ peptides appears to be unimpaired. In sharp contrast, Hayashi et al. (22) demonstrated a clear decrease in A␤ production from cells expressing the NCT-specific scFv 5201F, whereas Notch processing was not evaluated. These studies demonstrate that NCT is a feasible target to selectively modulate processing of ␥-secretase substrates.
Our finding that production of AICD (37) is reduced in scFvG9 cells, whereas A␤40 and A␤42 peptide production appears spared, is seemingly discordant with earlier studies from Ihara and co-workers (43), which showed that AICD and A␤ peptides are produced at equimolar ratios. This has led to the view that the amino-terminal derivative generated following ⑀ cleavage of C99, the product of BACE 1 cleavage of APP, is successively trimmed to generate the termini of A␤40 and A␤42. However, it should be noted that these studies were conducted in cell-free assays with purified membranes and C99. On the other hand, AICD that is generated in cells and tissues is extremely difficult to detect because, like NICD, these derivatives are subject to rapid intracellular degradation. The mechanistic relationship of ⑀-generated AICD to ␥-secretase-mediated cleavage to generate the termini of A␤ peptides is largely unclear, but several independent studies have suggested that ␥ and ⑀ cleavages are independent. For example, expression of several familial AD-linked PS1 variants elevate the production of A␤42 but fully block AICD formation (44 -46). Similarly, Hecimovic et al. (50) have reported that mutations within the APP transmembrane domain that elevate A␤42 production have no effect on production of AICD. In this regard, He et al. (47) demonstrated that a ␥-secretase-activating protein increases A␤ production but appears to reduce the generation of AICD.
The mechanism by which scFvG9 reduces levels of mature NCT, PS, and PEN-2 is not fully understood. However, our demonstration that scFvG9 can be coimmunoprecipitated with immature NCT would lead us to propose that antibody binding results in unfavorable folding kinetics and/or glycosylation of a population of these molecules, leading to retention and subse-FIGURE 6. The single-chain variable fragment G9 binds to the immature ␥-secretase complex. A, cell lysates from HEK293 APPSwe/scFv double-stable cells were immunoprecipitated with anti-myc or anti-NCT N terminus antibodies. Normal mouse IgG was used as a control antibody. Immunoprecipitated protein-antibody complexes were eluted from protein G beads and resolved on SDS-PAGE. Immunoblot analyses were used to detect NCT, full-length PS1 (PS1 FL), PS1 NTF, APH-1, PEN-2, and scFv. B, cell lysates from HEK293 APPSwe/scFv double-stable cells were immunoprecipitated with anti-PS1 N terminus antibodies. Normal rabbit IgG was used as a control antibody. Immunoprecipitated protein-antibody complexes were eluted from protein A beads and resolved on SDS-PAGE. Immunoblot analyses were used to detect NCT, PS1 CTF, PEN-2, and scFv. C, cell surface biotinylation with sulfosuccinimidyl-2-[biotinamido]ethyl-1,3-dithiopropionate on HEK293 APPSwe/scFv stable cells. Both total cell lysates and biotinylated proteins from the cell surface were resolved by SDS-PAGE. The NCT level was detected by immunoblot analysis. mat, mature; imm, immature.
Nicastrin scFv Affects ␥-Secretase Assembly and Trafficking DECEMBER 12, 2014 • VOLUME 289 • NUMBER 50 quent degradation in the endoplasmic reticulum. In turn, a fraction of full-length PS that associates with this population of immature, misfolded NCT is also subject to degradation, as are PEN-2 molecules, which have been shown to also associate with full-length PS1 (48). In parallel, a population of appropriately folded NCT can associate with full-length PS and PEN-2 and escape the endoplasmic reticulum degradation pathway, leading to endoproteolysis to generate PS1 NTF and CTF (Fig. 7). It is presently unclear whether scFvG9 is still associated with this "active" complex that would accumulate in late compartments of the secretory apparatus where APP-CTFs and NEXT are subject to intramembranous processing to generate A␤, AICD, and NICD, respectively. Interestingly, we find a paucity of NCT on the surface of cells expressing scFvG9 despite the fact that the levels of NICD are only reduced by ϳ50% in these cells. The absence of NCT on the cell surface would argue that a fraction of the constitutively activated Notch derivative mN⌬E, which corresponds to NEXT, is processed not at the cell surface, where full-length Notch is subject to ADAM10 and ␥-secretase processing (39), but during transit through endosomal and recycling compartments, organelles where APP-CTFs are also subject to intramembranous processing. Delineation of the exact mechanisms that lead to differential ␥-secretase processing of APP-CTFs versus NEXT will require high-resolution crystallographic analysis of the NCT ECD in the presence or absence of scFvG9. In this regard, Xie et al. (49) have recently reported the crystal structure of the Dictyostelium purpureum NCT homologue at 1.95-Å resolution. FIGURE 7. Proposed model of scFvG9 binding to NCT and immature ␥-secretase complexes. The single-chain fragment variable G9 is depicted as two attached small ovals (VH and VL), NCT is shown as a hammer, APH-1 is shown as a square, PS1 is shown as an irregular quadrilateral shape, and PEN-2 is shown as an oval. Our data suggest that scFvG9 binds to NCT alone or immature ␥-secretase subcomplexes rather than a mature complex, leading to retention and subsequent degradation of these components in the endoplasmic reticulum.