Differential Regulation of Amyloid Precursor Protein/Presenilin 1 Interaction during Ab40/42 Production Detected Using Fusion Constructs

Beta amyloid peptides (Aβ) play a key role in the pathogenesis of Alzheimer disease (AD). Presenilins (PS) function as the catalytic subunits of γ-secretase, the enzyme that releases Aβ from ectodomain cleaved amyloid precursor protein (APP) by intramembrane proteolysis. Familial Alzheimer disease (FAD)-linked PSEN mutations alter APP processing in a manner that increases the relative abundance of longer Aβ42 peptides to that of Aβ40 peptides. The mechanisms by which Aβ40 and Aβ42 peptides are produced in a ratio of ten to one by wild type presenilin (PS) and by which Aβ42 is overproduced by FAD-linked PS variants are not completely understood. We generated chimeras of the amyloid precursor protein C-terminal fragment (C99) and PS to address this issue. We found a chimeric protein where C99 is fused to the PS1 N-terminus undergoes in cis processing to produce Aβ and that a fusion protein harboring FAD-linked PS1 mutations overproduced Aβ42. To change the molecular interactions within the C99-PS1 fusion protein, we made sequential deletions of the junction between C99 and PS1. We found differential effects of deletion in C99-PS1 on Aβ40 and 42 production. Deletion of the junction between APP CTF and PS1 in the fusion protein decreased Aβ40, while it did not decrease Aβ42 production in the presence or absence of FAD-linked PS1 mutation. These results are consistent with the idea that the APP/PS interaction is differentially regulated during Aβ40 and 42 production.


Introduction
Alzheimer disease (AD) is a neurodegenerative disease characterized by the presence of senile plaques, of which beta amyloid peptide (Ab) is the primary component [1]. Ab is considered to be involved in the pathogenesis of AD, because familial AD (FAD) has been linked to mutations in the genes that encode amyloid precursor protein (APP) [2,3,4] and presenilin (PS) [5,6,7]. Mutations in these genes result in most cases in a relative increase in the production of Ab42 [8,9], which is the predominant form found in senile plaques [10]. Genetic ablation revealed that PS is required for cleavage of the APP C-terminal fragment (CTF), which is generated by aor b-secretase cleavage of full-length APP within the extracellular/luminal domain [11,12]. It is clear now that PS proteins are the catalytic subunits of c-secretase [13,14,15]. Identification of the mechanisms by which intramembrane cleavage is performed has been a topic of intensive investigation. Two aspartate residues in the predicted transmembrane domains 6 and 7 in PS are essential for c-secretase activity [13]. Moreover, at least three molecules, termed anterior pharynx defective phenotype 1 (APH-1), PS enhancer 2 (PEN-2), and nicastrin (NCT), are required for c-secretase activity [14]. The electron microscopic structure of purified, active c-secretase complex revealed a pore in the structure [16,17], but the structural resolution was too low to allow predictions about the working mechanisms. Indeed, it is unknown how Ab40/Ab42 are produced in a ratio of ten to one by wild type PS and how Ab42 is overproduced by FAD-linked PS variants. The nature of APP-CTF/PS interaction during Ab40 and 42 production is not fully understood.
In the present study, we examined the mechanism by which Ab40 and Ab42 are produced. To address this issue, we generated a series of constructs encoding chimeric proteins where APP CTF is fused to the N-terminus of PS, taking advantage of previous findings that deletion of the N-terminal domain of PS does not affect c-secretase activity [18]. Here, we show that fusion of APP CTF to the N-terminus of PS produces Ab and that fusion of APP CTF to FAD-linked presenilin 1 variants overproduces Ab42(43).
Deletion analysis of the junction between APP CTF and PS1 in the fusion protein suggests differential regulation of APP/PS interaction underlying Ab40 and 42(43) production.

Western Blot
SDS-PAGE was carried out on Tris/Tricine gel (Invitrogen, Carlsbad, CA, USA). Samples were mixed with 26SDS sample buffer (Invitrogen) and incubated at 37uC for 30 min immediately prior to electrophoresis. The samples were transferred onto a polyvinylidene difluoride (PVDF) membrane (Millipore, Bedford, MA, USA). Western blotting was performed using the antibodies described above. Enhanced ChemiLuminescence (ECL) was used as the detection system.

Ab ELISA
Analysis of secreted Ab40 and 42(43) in the medium of PS dKO MEF expressing C99-PS1 chimeras was performed using a human-specific Ab ELISA (Wako, Japan). According to the manufacturer's information, the sensitivity of human specific Ab40 and 42 ELISA is 0.12 and 0.08 pM, respectively. DAPT treatment clearly decreased Ab 40 and 42 levels from the basal levels in cells expressing the fusion proteins ( Figure S4D, E). Moreover, the deletion mutants C99/PS1D37/D71 and D37/D72, which have defects in the transmembrane I domain of PS, secreted significantly less Abeta42 than did C99-PS1 ( Figure S4G). These results indicate that our assay had sufficient specificity and sensitivity to detect different values of Ab at these low levels. Because PS(2/2) cells have no c-secretase activity and PS(2/2) cells have no human Ab, the value of [Ab level of vector plus Ab level of PS1wt]/2 was set as 0.

Statistical Analysis
All values are expressed as mean 6 SEM. For comparisons of the means between two groups, the data were statistically analyzed by Student's t-test. Comparisons of the means among three or more groups were performed by analysis of variance (ANOVA). Values of p less than 0.05 were considered significant.

Fusion of APP CTF to presenilin 1 N-terminus produces Ab
A PCR-based method was used to generate a construct that encodes the C99-PS1 fusion protein, where the C-terminus of C99 and the N-terminus of PS1 were fused, with deletion of the last three amino acids of C99 and the first 16 amino acids of PS1 ( Figure 1A). Dominant negative D385A or FAD-linked G266S mutations of PS1 were introduced into the C99-PS1 fusion protein. First, we checked the successful generation and expression of the fusion plasmid in COS cells, which are often used for this purpose. When transiently expressed in COS cells, C99-PS1 fusion protein was detected as an ,50 kDa polypeptide using 6E10, a monoclonal antibody (mAb) raised against the human Ab sequence. Presenilins undergo endoproteolysis within the loop region connecting transmembrane domains 6 and 7, resulting in an ,27-30 kDa NH 2 -terminal fragment (NTF) and an ,16-20 kDa COOH-terminal fragment (CTF) [20]. Consistent with endoproteolysis of the C99-PS1 fusion protein, we found that 6E10 mAb detected an ,35 kDa NTF ( Figure 1B). PS1NT mAb, which is raised against a peptide sequence behind the 17 th amino acid, also reacted with these ,50 kDa and ,35 kDa species, indicating that the C99-PS1 fusion protein undergoes endoproteolysis as observed in wild-type PS1 protein ( Figure 1B). The stability of this fusion protein was comparable to that of PS1 protein, as suggested by the experiments using cycloheximide (30 mg/ml) ( Figure S2). Moreover, we did not observe polypeptides smaller than 30 kDa in size when blots were probed with 6E10 mAb, indicating no fragmentation of APP CTF from the fusion protein. Immunoblot analysis using a mAb raised against the PS1 loop domain revealed that PS1 CTF was also detected in fibroblasts derived from PS1 (2/2) PS2 (2/2) mouse embryos (PS dKO MEF), transfected with the C99-PS1 construct ( Figure 1F). Compared to COS cells, when C99-PS1 plasmid was transfected into PS dKO cells, we did not observe 6E10 immunoreactivity with a molecular weight corresponding to PS1NTF, suggesting efficient cleavage in these cells ( Figure S1). These results suggest that the C99-PS1 fusion protein undergoes endoproteolysis similarly to wild-type PS1 protein.
Next, we examined whether the C99-PS1 fusion protein can restore the loss of c-secretase function in PS dKO MEF. We cotransfected PS dKO MEF with plasmids encoding the APP ''Swedish'' mutant (APPswe) along with an empty vector, C99-PS1, C99-PS1D385A or C99-PS1G266S plasmid. Co-transfection of C99-PS1 and APPswe produced Ab40 and Ab42 ( Figure 1C), while PS1D385A, a dominant negative mutation, failed to do so. Co-transfection of C99-PS1G266S, which harbors an FAD-linked PS1 mutation, and APPswe overproduced Ab42 ( Figure 1C). Taken together, these results indicate that the C99-PS1 fusion protein restores c-secretase activity in PS dKO MEF. Next, we investigated whether C99 within C99-PS1 could be a substrate for c-secretase. Analysis using a human Ab40-specific ELISA clearly showed that expression of C99-PS1, but not PS1, produced human Ab40 ( Figure 1D), suggesting that C99 within C99-PS1 undergoes in cis processing by c-secretase-like activity. To confirm whether a substrate fused to PS1 could be cleaved, we also performed an experiment to investigate whether the fusion protein of F-NEXTDC, a mouse Notch1 [15], another type I membrane protein, derivative that lacks the majority of its extracellular and intracellular domains [21], to PS1 produces an Ab-like peptide, Nb [22]. We found that Nb is indeed produced in K293 cells stably expressing F-NEXTDC-PS1, suggesting that a substrate fused to PS1 could be cleaved ( Figure S3). To confirm whether Ab was indeed produced by c-secretase processing in PS dKO MEF, we treated the cells with the c-secretase inhibitor, DAPT. We found that DAPT inhibited Ab production in a dose-dependent manner ( Figure 1E). Taking these results together, the PS1 protein module within the C99-PS1 fusion protein functions to restore csecretase activity in PS dKO MEF, and the C99 sequence serves as a substrate for c-secretase processing.
Differential effects of deletion in C99-PS1 wild type on Ab40 and 42 production To change the molecular interaction between APP and PS1, we made sequential deletions of the hydrophilic region between the transmembrane domain of C99 and transmembrane I domain of PS1 in the chimeric proteins. We characterized Ab40/42 production in PS dKO MEF transiently expressing deletion mutants of C99-PS1 WT and R278I (Figure 2A, B). Ab40 production was reduced when C99D37-PS1D70 was transiently expressed in PS dKO MEF as compared with expression of C99-PS1 ( Figure 2C), whereas Ab42 production was unaffected. Moreover, Ab42 production did not significantly decrease when C99D37-PS1R278ID70 was transiently expressed in PS dKO MEF compared to C99-PS1R278I ( Figure 2D). These results indicate that the APP/PS interaction is differentially regulated during Ab40 and 42 production. Next, to confirm this hypothesis, we made a series of shorter deletions within the PS1 N-terminus and chose clones with Ab42 levels equivalent to that with the original C99-PS1 fusion protein, to exclude clones with low expression of the transgene from further analysis. We found differential effects of serial deletions in C99-PS1 on Ab40 and 42 production. Deletion of C99-PS1 decreased Ab40, while it did not decrease Ab42 ( Figure 2E). Moreover, deletion of C99-PS1R278I also did not decrease Ab42 ( Figure 2F). While C99-PS1R278I does not produce Ab40 as shown in Fig. 2D, deletions within C99-PS1R278I did not affect Ab40 production ( Figure 2F). These data support the hypothesis that the APP/PS interaction is differentially regulated during Ab40 and 42 production.
Next, to confirm that Ab42 production by the C99-PS1R278I fusion protein was strictly c-secretase dependent, and not due to non-specific degradation, we treated dKO MEF expressing C99-PS1R278ID38/D59 with a selective c-secretase inhibitor, DAPT. Ab42 production was inhibited by DAPT in dKO MEF expressing C99-PS1R278ID38/D59 ( Figure 2G), suggesting that Ab42 production by the C99-PS1R278I fusion protein was indeed csecretase-dependent. Finally, we generated C99-PS2, a fusion protein of C99 and the PS1 homolog that can also function as the catalytic subunit of c-secretase, and found that this chimeric protein also produced Ab, when expressed in dKO MEF (Figure 3).

Discussion
The mechanisms by which Ab40 and Ab42 peptides are produced in a ratio of ten to one by wild type PS and Ab42 is overproduced by FAD-linked PS variants are not clearly elucidated ( Figure 4A). To understand the mechanisms by which Ab40 and Ab42 are produced, we generated chimeras by fusing APP C99 with the N-terminus of presenilin (PS1 or PS2). We found that expression of this fusion protein in MEF lacking PS1 and PS2 expression generated Ab. Furthermore, a fusion protein harboring the FAD-linked PS1 mutation overproduced Ab42. To experimentally alter the intermolecular interactions between the  C99 and PS1 modules within the fusion protein, we introduced sequential deletions between C99 and PS1. We found differential effects of deletion in C99-PS1 wild type on Ab40 and 42 production. While deletions of C99-PS1 decreased the level of Ab40, there was no effect on the level of Ab42. Moreover, deletions within C99-PS1R278I also did not decrease Ab42. These data suggest differential regulation of the APP/PS interaction during Ab40 and 42 production.
Based on our findings, we suggest that the chimeric proteins characterized in this study mimic the native interaction between APP CTF and presenilin, and would be useful tools for further exploration. First, we showed that fusion of APP CTF to the Nterminus of PS1 produced Ab by intrinsic processing upon reconstituion of c-secretase activity in PS dKO cells. Moreover, a c-secretase inhibitor, DAPT, inhibited Ab production, suggesting that fusion of APP CTF to the N-terminus of PS1 produced Ab in a c-secretase-dependent manner. We also found that Nb is indeed produced in K293 cells stably expressing F-NEXTDC-PS1, the fusion protein of F-NEXTDC, a mouse Notch1 derivative that lacks the majority of its extracellular and intracellular domains [21], to PS1 suggesting that a substrate fused to PS1 could be cleaved ( Figure S3). The maximum amount of Ab that could be produced would be no greater than the amount of APP-PS1 fusion protein produced. The expression level of PS1 is regulated by a limiting factor, i.e. the c-secretase complex. Thus, this might be the reason why the amount of Ab that is produced in cells expressing only the fusion protein is very low. However, DAPT treatment clearly decreased Ab40 and 42 levels from the basal levels in cells expressing the fusion proteins ( Figure S4D, E). Moreover, the deletion mutants C99/PS1D37/D71 and D37/D72, which have defects in the transmembrane I domain of PS, secreted significantly less Abeta42 than did C99-PS1 ( Figure S4G). These results indicate that our assay had sufficient specificity and sensitivity to detect different values of Ab even at these low levels. Second, fusion of APP CTF to the FAD-linked PS1 variant overproduced Ab42 (Figure 1). Since APP CTF and PS are transmembrane proteins, fusion of the APP cytosolic C-terminus to the hydrophilic N-terminus of presenilin might allow them to maintain the right positional relationship on the membrane bilayer via hydrophobic transmembrane domains. This strategy could be utilized to investigate the mechanism of other intra-membranous cleavage or other associations between two proteins within or on the membrane. Recent findings that the first transmembrane domain of PS1 plays an important role in the enzyme/substrate relationship [25,26] might support our strategy of fusion of APP CTF adjacent to the first transmembrane domain of presenilin. Third, we found differential effects of deletion in C99-PS1 wild type on Ab40 and 42 production. Deletion of residues between the transmembrane domain of APP and the first transmembrane domain of PS1 decreased Ab40 production, while having no effect on Ab42 production. Moreover, deletions within C99-PS1R278I also did not reduce Ab42(43) production. It is reported that FADlinked PS1 mutations cause alterations in the conformation of PS and interactions with APP [27]. Interestingly, heterogeneity of the c-secretase complex, i.e. Aph1B or Aph1A, causes a change in the relative ratio of Ab40 and 42 production [28], associated with Ab42 level (right) by Ab ELISA in medium of PS1 (2/2) PS2 (2/2) cells transfected with C99-PS1 deletion mutants (C) or C99-PS1R278I (D). ** p,0.01.  conformational change of PS1. It is essential to investigate whether cleavage of the C99-PS1 fusion protein requires binding to the other members of the c-secretase complex. We observed replacement of endogenous PS1 when the fusion proteins C99-PS1 and F-NEXTDC-PS1 were stably expressed in stable K293 cells (Figures S3, S5). These data suggest that the fusion protein is incorporated into the c-secretase complex. We are currently investigating whether the production of Ab from the C99-PS1 fusion protein requires binding to the other members of the csecretase complex. In addition, NSAIDs reduce Ab42 production, also associated with alterations in the conformation of PS and interactions with APP. Our data suggest that conformational change of the APP/PS interaction might underlie cleavage site selection leading to Ab40 and 42 production ( Figure 4B). It might be also interesting to test the effects of the FAD-linked mutations in the APP gene in this fusion system. Introduction of APP Iberian [29] mutation which overproduce Ab42 [30] might result in a similar effect of deletion of residues between C99 and PS1 on Ab production to C99-PS1R278I. It is proposed that APP CTF is cleaved first at the membrane-cytoplasm boundary, producing two longer Ab species, Ab48 and Ab49, which are processed further by releasing three residues at each step to produce Ab42 and Ab40, respectively [31], suggesting that the mechanisms by which Ab42 and Ab40 are generated utilize differential substrate-catalytic site interaction to yield two distinct product lines. Our data presented here are also not inconsistent with this hypothesis. Taken together, these data suggest that the APP/PS interaction is differentially regulated during Ab40 and 42 production.
In summary, we generated fusion proteins of APP CTF to the N-terminus of PS and observed differential regulation of the APP/ PS interaction during Ab40 and 42 production. Because the crystal structures of other intramembrane proteases have been reported [32,33], further analysis including high resolution of crystal structures or NMR of C99-PS1 chimeras would provide further insights into the mechanisms by which Ab40/Ab42 are produced in a ratio of ten to one by wild type PS and Ab42 is overproduced by FAD-linked PS variants, and open the door to structure-based design of pharmacological modulators of this protease.

Supporting Information
Text S1 (DOCX) Figure S1 Comparison of expression level of C99-PS1 in COS cells and PS(2/2) cells. A higher level of protein expression was observed in COS cells than in PS(2/2) cells. Note that no reactivity with 6E10 antibody was observed in cells that expressed C99-PS1 fusion protein in PS(2/2) cells, with even high exposure. (TIF) Figure S2 Stability of C99-PS1 fusion protein in COS cells. The stability of the fusion protein was investigated using cycloheximide (30 mg/ml). Note that that stability of C99-PS1 was comparable to that of PS1. We observed that the immunoreactivity of the fusion protein was not very different between PS1 NT and 6E10 in COS cells. However, this result does not exclude that the fusion protein is being cleaved efficiently, because it is highly overexpressed in COS cells. (TIF) Figure S3 Generation of F-NEXTDC-PS1 and detection of Nb. The fusion protein of Notch1, another type I membrane protein, to PS1 was generated. F-NEXTDC-PS1/K293-clone #15 expressed a high level of full-length F-NEXT-DC-PS1, and replacement of endogenous PS1 NTF and PS2 by full-length F-NEXT-DC-PS1 was observed in this cell line, but less replacement was observed in clone #4 with low expression of full-length F-NEXT-DC-PS1 (left panel). IP-Mass experiment revealed that Nb was secreted in F-NEXTDC-PS1/K293-clone #15. Thus, we confirmed that the fusion protein of flag-tagged NotchDE fused to PS1 could be cleaved, resulting in Nb secretion (right panel), suggesting that substrate fused to PS1 could be cleaved.