Visualization of Alzheimer’s Disease Related α-/β-/γ-Secretase Ternary Complex by Bimolecular Fluorescence Complementation Based Fluorescence Resonance Energy Transfer

The competitive ectodomain shedding of amyloid-β precursor protein (APP) by α-secretase and β-secretase, and the subsequent regulated intramembrane proteolysis by γ-secretase are the key processes in amyloid-β peptides (Aβ) generation. Previous studies indicate that secretases form binary complex and the interactions between secretases take part in substrates processing. However, whether α-, β- and γ-secretase could form ternary complex remains to be explored. Here, we adopted bimolecular fluorescence complementation in combination with fluorescence resonance energy transfer (BiFC-FRET) to visualize the formation of triple secretase complex. We show that the interaction between α-secretase ADAM10 and β-secretase BACE1 could be monitored by BiFC assay and the binding of APP to α-/β-secretase binary complex was revealed by BiFC-FRET. Further, we observed that γ-secretase interacts with α-/β-secretase binary complex, providing evidence that α-, β- and γ-secretase might form a ternary complex. Thus our study extends the interplay among Alzheimer’s disease (AD) related α-/β-/γ-secretase.


INTRODUCTION
Amyloid plaques primarily composed of amyloid-β peptides (Aβ) have been identified to be one of the major hallmarks of Alzheimer's disease (AD) (Masters et al., 1985;Huang and Mucke, 2012). Amyloid cascade hypothesis suggest that the imbalance between Aβ production and clearance might be the central event and often initiating factor in AD pathogenesis (Hardy and Selkoe, 2002;Selkoe and Hardy, 2016). Aβ peptides are derived from amyloid-β precursor protein (APP) through sequential proteolysis by βand γ-secretase (Vassar, 1999;De Strooper, 2003;Blennow et al., 2006). In the non-amyloidogenic pathway, α-secretase competes with β-secretase in APP ectodomain shedding and cleaves APP within Aβ domain, thus precludes Aβ production (Skovronsky et al., 2000;Colombo et al., 2013). For AD treatment, researchers have put much effort to reduce Aβ generation through inhibiting or modulating βor γ-secretase activities (De Strooper et al., 2010;Yan and Vassar, 2014;MacLeod et al., 2015). However, most of the chemicals targeting the secretases showed side effects in clinical trials due to the diverse substrates of secretases (Graham et al., 2017;Voytyuk et al., 2018). Therefore, it is crucial to gain insights into the molecular mechanism of secretases activities in APP processing.
cDNA sequences of mCherry, human BACE1, BACE1 N-terminal peptide LT52 (amino acids 43-94) and the scramble control peptide LT52-S were cloned into pcDNA3 or pcDNA4 vector. The cDNA sequences of human Aph1aL, Nct, PS1 and Pen2 were subjected to codon optimization (Life Technologies) and cloned into pMlink vector. pMlink-Aph1aL, pMlink-Nct and pMlink-Pen2 were combined by the LIC method to generate pMLink-nicastrin-Pen2-Aph1aL as reported previously (Lu et al., 2014). For BiFC and FRET constructs, the C-or N-terminal fragments of Venus (with I152L mutation), or mTurquoise2 was fused to the C-terminus of ADAM10 or BACE1 or APP with a linker (GSGGGGSGGGGS) and cloned into pMlink vector, and mTurquoise2 was fused to the N-terminus of PS1 and cloned into pMlink vector. We performed site-directed mutagenesis using an overlapping PCR strategy to generate APPαM, APPβM and APPαβM. Flag tag were fused to the C-terminus of wild-type and mutated APP.

Cell Culture and Transfection
HEK 293 cells were obtained from ATCC and cultured in Minimum Essential Medium (MEM) supplemented with 10% FBS at 37 • C under 5% CO 2 . Neuro-2a cells were obtained from ATCC and cultured in MEM supplemented with 10% FBS at 37 • C under 5% CO 2 , and transfected in MEM supplemented with 0.5% FBS to allow differentiation. HEK 293/APPswe cells were stably transfected with APP Swedish mutant (APPswe-HA).
To monitor the expression of BiFC constructs in cells, HEK 293 cells used for Western blots analysis were plated and transfected at the same time with the same transfection mix as BiFC analysis. HEK 293 cells were also harvested 16 h after transfection for immunoblots analysis.
The fluorescence intensity of Venus and total cell number in 20× images and protein expression analyzed by Western blots were quantified respectively using ImageJ software, and relative BiFC signal was calculated as: Venus fluorescence intensity/(total cell number × expression of ADAM10 and BACE1).

Western Blots Analysis
After transfection, HEK 293 cells were lysed in 1× loading buffer. Protein samples were resolved on 10% or 12% SDS-PAGE and transferred to nitrocellulose membranes. After blocked by dried skimmed milk in TBST, membranes were incubated with primary antibodies, washed three times and incubated with HRP-conjugated secondary antibodies. The membranes were incubated with ECL substrates (Bio-rad) and then scanned by Tanon-5200 system. The quantitative analysis was carried out by measuring the gray value of the blot bands using ImageJ software.

Immunofluorescence Staining Analysis
For lysosome staining, cells were incubated with LysoTracker (Molecular probes) for 30 min and fixed with PBS/2% paraformaldehyde followed by Hoechst 33342 staining. For Cis-Golgi complex and early endosome staining, cells were fixed with PBS/2% paraformaldehyde, permeabilized, and blocked with PBS/0.1% Saponin/2% BSA followed by incubation with primary antibodies at room temperature for 2 h. After washed with PBS/0.1% Saponin/2% BSA, cells were incubated with Alex647-labeled secondary antibodies for 1 h and followed by Hoechst 33342 staining in the dark at room temperature. Then cells were mounted on slides for image acquisition under a LAS SP8 confocal microscope (Leica) with a 63×/1.40 NA oil objective (Leica).

Acceptor Photobleaching Fluorescence Resonance Energy Transfer (FRET) Measurement
For BiFC-FRET assay, 16 h (HEK 293 cells) or 24 h (neuro-2a cells) after transfection, cells were fixed with PBS/2% paraformaldehyde, washed and mounted on slides for FRET detection. Samples were then subjected to acceptor photobleaching FRET imaging under a confocal microscope (LAS SP8; Leica) with a 63×/1.40 NA oil objective (Leica) as described previously . Image acquisition, registration, background subtraction, and data analyses were performed with Leica Application Suite Advanced Fluorescence (LAS AF) software. Imaging conditions were set up manually: mTurquoise2 (excitation: 458 nm, emission: 465-505 nm), and Venus (excitation: 514 nm, emission: 525-600 nm). Photobleach was performed using acceptor excitation light and the similar bleach efficiency (∼80%-90%) was achieved. Images of donor and acceptor channels were acquired preand post-bleach, respectively. FRET efficiency was calculated as percentage of enhancement in donor fluorescence (f) after acceptor photobleaching: Five non-bleached regions were selected and the average value was used to correct the FRET efficiency of photobleached region.

Statistical Analysis
All experiments were repeated at least three times and data are presented as Mean ± SEM and analyzed by GraphPad Prism 6.01 (San Diego, CA, USA). Unpaired Student's t-test was used for the comparisons between two groups. One-way analysis of variance (ANOVA) followed Bonfferoni's multiple comparisons test was used for the comparisons among more than two groups.

BiFC Reveals α-/β-Secretase Complex
The pairwise interactions of α-, βand γ-secretase have been previously reported, we developed BiFC based FRET (BiFC-FRET) assay to visualize ternary complex in intact cells. First, we studied the interaction between α-secretase ADAM10 and β-secretase BACE1 using BiFC assay. We engineered BiFC constructs by fusing complementary C-fragment (CV) or N-fragment (NV) of Venus to the C-terminal of ADAM10 or BACE1 to generate fusion protein ADAM10-CV and BACE1-NV ( Figure 1A). In HEK 293 cells stably overexpressing APP, we found that Venus fragment tagged ADAM10 and BACE1 possess catalytic activities to cleave APP (Supplementary Figure 1A). We assessed BiFC signal 16 h after transfection to diminish the effect of BiFC complex accumulation. Venus BiFC fluorescence was scarcely seen in cells transfected with BACE1-NV and only CV fragment or in cells transfected with ADAM10-CV and only NV fragment (Figures 1BI,II), while was more frequently observed in cells transfected with ADAM10-CV and BACE1-NV ( Figure 1BIII, quantified in Figure 1C). Besides, we confirmed the expression levels of ADAM10-CV and BACE1-NV to be near those of endogenous proteins to avoid artificial interaction caused by robust expression (Figure 1D). Reversely, Venus fluorescence in cells transfected with ADAM10-NV and BACE1-CV (Supplementary Figure 1BIII, immunoblots analysis shown in Supplementary Figure 1C) was also significantly stronger than those in cells transfected with ADAM10-NV and only CV fragment or BACE1-CV and only NV fragment (Supplementary Figures 1BI,II, quantified in Supplementary Figure 1D). Although Venus fluorescence in cells transfected with ADAM10-NV and BACE1-CV was relative weaker than that in cells transfected with ADAM10-CV and BACE1-NV, these data suggests that ADAM10/BACE1 interaction was more favorable to the complementation of ADAM10-CV and BACE1-NV. To further confirm the BiFC efficiency, we examined ADAM10/BACE1 BiFC signal with the expression of mCherry as a control. As shown in Supplementary Figures 1E,F, ADAM10/BACE1 BiFC signal could be observed in the majority of mCherry-expressing cells, indicating that the interaction between ADAM10 and BACE1 bring the two nonfluorescent halves of Venus together and generate BiFC signal efficiently.
To further testify the BiFC fluorescence is caused by ADAM10/BACE1 interaction, we examined the effect of LT52 (Supplementary Figure 1G), a peptide that was reported to attenuate ADAM10/BACE1 interaction (Wang et al., 2018), on ADAM10/BACE1 BiFC signal. As shown in Figure 1E, the expression of LT52 led to significant reduction of BiFC fluorescence (quantified in Figure 1F), and showed much less effect on ADAM10-CV and BACE1-NV expression level (Supplementary Figure 1H). However, the expression of scrambled peptide LT52-S (Supplementary Figure 1G) that has no influence on ADAM10/BACE1 interaction, showed little effect on BiFC fluorescence (Figures 1E,F). These data suggest that the BiFC recapitulated ADAM10/BACE1 specific interaction.
Next, we inquired into the ADAM10/BACE1 BiFC by high-resolution imaging. As shown in Figure 1B , Venus BiFC fluorescence in cells transfected with ADAM10-NV and BACE1-CV largely distributed in punctate pattern. Then we examined the subcellular localization of ADAM10/BACE1 BiFC signal. HEK 293 cells transfected with ADAM10-CV and BACE1-NV were stained with specific antibody for Golgi complex marker GM130 or early endosome marker EEA1, or incubated with lysosome probe LysoTracker. As shown in Figure 1G, ADAM10/BACE1 BiFC signal co-localized with GM130, while the BiFC puncta showed more significant co-localization with EEA1 but little with LysoTracker (quantified in Figure 1H).
Together, these data indicate that the interaction between ADAM10 and BACE1 could be monitored by BiFC assay in intact cells.

Visualization of APP/α-/β-Secretase Ternary Complex Using BiFC-FRET
We sought to investigate whether APP incorporates into binary α-/β-secretase complex further. In order to visualize the formation of APP/α-/β-secretase ternary complex directly, we utilized BiFC-FRET assay. If APP binds to α-/β-secretase complex, cyan fluorescent protein mTurquoise2 (Tur) tagged APP would be close to reconstituted Venus tagged α-/βsecretase complex, then fluorescent energy transfer occurs between mTurquoise2 and Venus ( Figure 3A). We transfected HEK 293 cells with APP-Tur, ADAM10-CV and BACE1-NV, and observed that the co-localization of APP-Tur and ADAM10/BACE1 BiFC signal was mainly seen in the stacks structure adjacent to nucleus ( Figure 3B). We adopted acceptor photobleaching method to detect the FRET efficiency between APP-Tur and ADAM10/BACE1 BiFC signal, and a low FRET efficiency of 2.86% was obtained (Figures 3D, APP-Tur/A-B  BiFC). However, when treated with GI254023X and BSI-IV, the co-localization of APP-Tur and ADAM10/BACE1 BiFC signal was observed in vesicles scattered throughout the cytoplasm (Figure 3C), and the FRET efficiency between APP-Tur and ADAM10/BACE1 BiFC signal was significantly higher to be 8.85% (Figure 3D, APP-Tur/A-B BiFC/AI/BI). Besides, we assessed the FRET efficiencies between APP and ADAM10 or BACE1 simultaneously (Supplementary  Figures 3A,B), and the FRET efficiencies of 10.34% and 9.93% were detected, respectively (Figure 3D, APP-Tur/A-V/AI and APP-Tur/B-V/BI). These results indicate that similar with its binding to αor β-secretase, APP interacts with α-/β-secretase complex and might form a ternary complex with α-/β-secretase.
To explore whether APP/α-/β-secretase complex form in physiological conditions, we transfected differentiated neuro-2a cells with APP-Tur, ADAM10-CV and BACE1-NV. As shown in Figures 3E,F, when treated with GI254023X and BSI-IV, we observed that FRET could also occur between APP and ADAM10/BACE1 BiFC signal, suggesting that APP, αand β-secretase could form ternary complex in neuronal cells.

DISCUSSION
Previous studies showed that FRET could be applied to investigate interactions between αand β-secretase, and between βand γ-secretase. In this study, we corroborate the interaction between αand β-secretase using BiFC assay. By comparing BiFC signals in cells expressing ADAM10-CV and BACE1-NV or BACE1-CV and ADAM10-NV, and examining the effect of ADAM10/BACE1 interaction-interfering peptide, we confirmed that the BiFC signals are primarily generated from specific interaction between ADAM10 and BACE1. Further, we provide evidence for the first time that APP and γ-secretase could form ternary complex with α-/β-secretase complex taking advantage of the combination of BiFC and FRET, and reinforce the idea that secretases may form functional multiprotease complex (Chen et al., 2015). However, the mechanism about the formation of these substrate-multiprotease complexes needs to be explored further. Our previous study showed that the extracellular domains of ADAM10 and BACE1 directly interact (Wang et al., 2018), thus it is possible that ADAM10 might bind directly to BACE1 and then APP or γ-secretase would attach to the complex. On the other hand, our previous study also suggests that ADAM10 and BACE1 partially co-localize, and the interaction between APP and BACE1 was observed in many studies (Kinoshita et al., 2006;Das et al., 2013Das et al., , 2016. Therefore, it is also possible that APP might directly interact with BACE1 or ADAM10 then the other enzyme bind to the complex. It could be speculated that different multiprotease complexes exist according to the reported studies. BiFC assay enables direct visualization of protein interactions without any complicated post analysis, which is also favorable to determine the subcellular locations where interactions happen (Kerppola, 2006). We investigated the location of ADAM10/BACE1 binary complex and observed the BiFC signals were primarily located in early endosomes, which is much consistent with previous reports about the sites of APP processing (Sannerud et al., 2011;Choy et al., 2012). Although it is supposed that ADAM10 cleaves APP at plasma membrane (Xu et al., 2009), ADAM10 was also reported to remove from plasma membrane through clathrinmediated endocytosis (Marcello et al., 2013), locate in early endosome (Escrevente et al., 2008) and exert proteolysis activity at endosomal pH (Mathews et al., 2010). Further, whether ADAM10/BACE1 binary complex exists in plasma membrane or other activity subcellular organelles needs to be explored.
To probe into the formation of secretases complex further, we examined the effects of αand β-secretase inhibitors on ADAM10/BACE1 BiFC signal. However, little effect was observed, which is in accordance with our previous results that α-secretase inhibitor TAPI had no influence on ADAM10/BACE1 FRET efficiency (Wang et al., 2018), indicating that the binding of inhibitor to the catalytic pocket might not affect ADAM10/BACE1 interaction. Previous study also reported that the interaction between βand γ-secretase was not affected by classic βand γ-secretase inhibitors . These data suggest that the conformational changes of secretases induced by the binding of inhibitors may be insufficient to alter the interactions between them, or maybe the catalytic pockets were not much crucial to the interactions. Although much effort has been put to reduce Aβ generation to cure AD, most of the inhibitors of β-secretase or γ-secretase failed in the clinical trials due to their side effects (Graham et al., 2017;Voytyuk et al., 2018). Targeting the APP-secretase or secretase-secretase complex might be a novel therapeutic strategy, but the classic secretase inhibitors don't seem to work in this way. To discover small molecules that interfere with the interactions between secretases or between APP and secretase might modulate APP cleavage and Aβ generation.
Although our previous study showed that APP processing was hardly affected by ADAM10/BACE1 interaction interference, we observed the binding of APP to α-/β-secretase binary complex by BiFC-FRET. The FRET efficiency between APP and α-/β-secretase binary complex is low in perinuclear structure, consistent with the notion that ADAM10 and BACE1 might not be active in this compartment (Capell et al., 2000;Yan et al., 2001;Noy et al., 2016). However, when cells were treated with ADAM10 and BACE1 inhibitors, higher FRET efficiency was detected in the scattered puncta that might be some vesicle structures, suggesting that the interaction between APP and α-/β-secretase complex in the puncta might be functional. Besides, APP mutants showed different effects on ADAM10/BACE1 BiFC signal, suggesting that the mutations might influence the binding of APP to the protease complex. Moreover, we observed that APP exerted an effect on α-/β-/γ-secretase FRET, indicating that APP might bind to the ternary complex formed by α-/β-/γ-secretase as well, further supporting that the secretases complex might be functional. Nevertheless, the binding of APP to the triple secretase complex needs to be explored further.