Aggregation and accumulation of the 42-residue amyloid β peptide (Aβ42) in the extracellular matrix and within neuronal cells is considered a major cause of neuronal cell cytotoxicity and death in Alzheimer's disease (AD) patients. Therefore, molecules that bind to Aβ42 and prevent its aggregation are therapeutically promising as AD treatment. Here, we show that a non-self-aggregating Aβ42 variant carrying two surface mutations, F19S and L34P (Aβ42DM), inhibits wild-type Aβ42 aggregation and significantly reduces Aβ42-mediated cell cytotoxicity. In addition, Aβ42DM inhibits the uptake and internalization of extracellularly added pre-formed Aβ42 aggregates into cells. This was the case in both neuronal and non-neuronal cells co-expressing Aβ42 and Aβ42DM or following pre-treatment of cells with extracellular soluble forms of the two peptides, even at high Aβ42 to Aβ42DM molar ratios. In cells, Aβ42DM associates with Aβ42, while in vitro, the two soluble recombinant peptides exhibit nano-molar binding affinity. Importantly, Aβ42DM potently suppresses Aβ42 amyloid aggregation in vitro, as demonstrated by thioflavin T fluorescence and transmission electron microscopy for detecting amyloid fibrils. Overall, we present a new approach for inhibiting Aβ42 fibril formation both within and outside cells. Accordingly, Aβ42DM should be evaluated in vivo for potential use as a therapeutic lead for treating AD.
Among human neurodegenerative diseases, Alzheimer's disease (AD) is the most common, with nearly 47 million patients worldwide . AD is characterized by a sequestration of amyloid β (Aβ) peptides into fibrillar aggregates that ultimately lead to neuronal cell death and subsequently progressive cognitive dysfunction [2,3]. Aβ peptides that form amyloid fibrils and plaques are processed products of the transmembrane amyloid precursor protein (APP), which is expressed at high levels in the brain and is cleaved into 36–43 residue protein segments by the β- and γ-secretase proteolytic enzymes in neuronal cells . The majority of APP cleavage occurs on the plasma membrane of neuronal cells, with the resulting Aβ peptides being subsequently releasing to the extracellular matrix (ECM) to form plaques. Aβ42-containing extracellular plaques are believed to induce microglia activation, resulting in the release and activation of pre-inflammatory cytokines, such as IL-1β, TNF-α, and IFN-γ, leading to neuro-inflammation [5–7]. Although the majority of the aggregated Aβ fibrils are found in the ECM, emerging evidence indicates that Aβ peptides can also accumulate intracellularly to form fibrillar aggregates. Intracellular accumulation and aggregation of Aβ stems from two sources, namely (i) APP cleavage within cells and (ii) re-uptake of Aβ aggregates from the ECM [8–10]. APP cleavage within cells mainly occurs in the Golgi network, endoplasmic reticulum, and endosomes, as well as on mitochondrial membranes, while Aβ aggregates can be localized with these organelles and in the cytoplasm [11–13]. Aβ accumulation and aggregation within the mitochondria disrupts the activity of this organelle by diminishing the enzymatic activity of respiratory chain complexes III and IV, resulting in increased reactive oxygen species production . Moreover, intracellular Aβ inhibits and disrupts the ubiquitin-proteasome system and alters calcium and synaptic function [15,16]. Together, these effects eventually lead to disruption of calcium homeostasis, altered proteolysis of essential proteins, and increased nitric oxide levels, and eventually, to apoptosis of neuronal cells [11,13,15,16].
Among the numerous inhibitors that bind Aβ and prevent its aggregation, widely used peptide-based inhibitors are designed based on the Aβ oligomer self-assembly motif KLVFF, a hydrophobic cluster on the Aβ peptide surface [17,18]. However, self-association into spherical assemblies of most of these inhibitors reduces their inhibition efficacies and, therefore, diminishes their potential as therapeutics . As Aβ oligomers form high-ordered amyloid fibrils that are enriched in β-sheet elements, other widely used inhibitors that bind Aβ oligomers are structurally designed to target the β-sheet elements of Aβ [19–22], so as to disrupt protein aggregation. Such peptides, like rGffvlkGr-1,5-diaminopentane (where the lower case letters indicate the D form of an amino acid that was used to reduce proteolytic degradation) , HKQLPFFEED (also known as H102) , and NAVRWSLMRPF (also known as NF11) , present moderate to high efficacy in inhibiting Aβ aggregation in vitro. Nonetheless, all these peptide-based inhibitors currently lack clinical outcome due to high toxicity, insufficient inhibition activity, or proteolytic degradation [25–27]. As such, the need for new inhibitors is urgent.
The insufficient inhibitory activities of available peptide inhibitors of Aβ aggregation can be explained by the fact that although both the extracellular and intracellular Aβ aggregation induce toxic effects, no existing inhibitors efficiently inhibit such aggregation at both sites. In addition, most attempts to develop Aβ aggregation inhibitors, such these described above [17–24,26], relied on studies largely performed in vitro, mostly addressing extracellular Aβ aggregation and measuring aggregation inhibition in vitro . In contrast, only few studies have been conducted in cells, such as a recent study in which the human molecular chaperone DnaJB6 was employed as an inhibitor, targeting Aβ intracellular aggregation. DnaJB6 was, however, unable to inhibit aggregation of extracellular Aβ42 .
In this study, we used an Aβ42 variant carrying two mutations, F19S and L34P (designated Aβ42DM) and evaluated its efficiency in inhibiting Aβ42 aggregation both in vitro and in cells. We further tested the ability of Aβ42DM to rescue cell viability in response to Aβ-mediated toxic effects. Aβ42DM was previously characterized in bacteria as a non-aggregating form of Aβ42 [29,30]. Substitution of the hydrophobic phenylalanine at position 19, part of the central KLVFF hydrophobic cluster in Aβ42, with the polar amino acid serine disrupted Aβ42 self-assembly and perturbed its conformation by altering the steric nature and polarity of the peptide, resulting in reduced aggregation [31,32]. Positions 33–36 in the Aβ sequence are also known to form a hydrophobic region, which is involved in Aβ peptide self-assembly. Previous studies have shown that changing position 34 in Aβ affects peptide conformation and polarity, as well as the ability to self-aggregate [30,33,34]. Moreover, a combination of the F19S and the L34P mutations was previously shown to prevent Aβ self-assembly more efficiently than either single-point mutations introduced separately, disrupting the predominantly β-sheet conformation associated with wild-type Aβ42 . Similar characteristics of Aβ42 variants carrying the F19S and additional mutations were also observed in neuronal cells .
To test our hypothesis that Aβ42DM can act as efficient inhibitor of Aβ42 aggregation not only in vitro but also in cells, we employed a mammalian cell-based platform in which expression of an Aβ42-GFP fusion protein served as a reporter of Aβ42 aggregation. Our results show that Aβ42DM binds Aβ42, inhibits its aggregation, and reduces Aβ42 cytotoxic effects when co-expressed with Aβ42, or when supplemented externally, in both HEK293T and SH-SY5Y neuroblastoma cells. We also show that Aβ42DM prevents the uptake and internalization of aggregated Aβ42 fibrils into neuronal cells, and thus has the potential to serve as a therapeutic lead in the search for drugs treating AD.
Cells and transfection
HEK293T cells were purchased from ATCC (Manassas, VA). SH-SY5Y neuroblastoma cell line was a generous gift from Prof. Varda Shoshan-Barmatz (BGU). HEK293T and SH-SY5Y cells were grown at 37°C and 5% CO2 in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% tetracycline-free fetal bovine serum (FBS), l-glutamine (2 mM), and penicillin (100 units/ml)/streptomycin (0.1 mg/ml) (Gibco, Israel). Transfection was performed using Lipofectamine 2000 transfection reagent (Thermo-Fisher, Waltham, MA) for HEK293T cells or Turbofect (Thermo-Fisher) for SH-SY5Y cells, according to the protocol of the manufacturer. When co-transfections were performed, empty plasmids were transfected as controls.
The Aβ42-GFP gene where Aβ42 is fused to GFP (a generous gift from Michael Hecht, Princeton University) was cloned into the pHAGE lentivector, in which transcription is under the control of the CMV promoter. The pHAGE vector also expresses a puromycin resistance selection marker that follows an internal ribosome entry site (IRES) sequence. Plasmid pHAGE Aβ42F19S,L34P-GFP (Aβ42DM-GFP) was generated by two consecutive rounds of directed mutagenesis of Aβ42 at positions 19 and 34. We also generated pHAGE HA-Aβ42DM-IRES-BFP lentivector, and plasmids pCDNA3.1-HA-Aβ42DM, pCDNA3.1-HA-Aβ42DM-BFP, and pCDNA3.1-HA-Aβ42DM-Fc. The Aβ42DM gene was amplified by PCR and fused to DNA encoding an N-terminal human influenza hemagglutinin (HA)-tag, and later cloned into the pHAGE2 or pCDNA3.1 expression vectors with one of the indicated markers. All plasmids are shown in Supplementary Figure S1.
Production of pseudo-typed lentivirus
HEK293T cells were transfected with the gene of interest using the packaging HIV expression plasmids pGag-Pol, pRT, pTat, and pREv and the envelope VSV-G protein by the calcium phosphate method. Cell supernatant containing the virus was collected 48 h post-transfection, cleared through a 0.45 µm filter, and further concentrated by centrifugation (2000 rpm, 10 min) to obtain a high titer viral stock that was then aliquoted and stored at −80°C. To determine viral titer, 5 × 104 HEK293T cells were plated on 24-well plates and infected with serial dilutions of the indicated VSV-G pseudo-typed virus CMV-GFP. Cells were analyzed for GFP expression by FACS.
Generation of stable cell lines
5 × 104 HEK293T and SH-SY5Y cells were seeded in 24-well plates and transduced with lentivirus expressing the relevant genes at 0.75 multiplicity of infection levels. After 72 h following transduction, the cells were subjected to 1.5 μg/ml puromycin selection and/or sorted for GFP and/or BFP expression (when Aβ42DM IRES-BFP was used). The following stable cell lines were generated:
(i) HEK293T Aβ42-GFP (HEK- Aβ42-GFP).
(ii) HEK293T Aβ42DM (HEK-Aβ42DM).
(iii) HEK293T Aβ42-GFP and Aβ42DM with low expression (HEK-Aβ42-GFP-Aβ42DML).
(iv) HEK293T Aβ42-GFP and Aβ42DM with high expression (HEK- Aβ42-GFP-Aβ42DMH).
(v) SH-SY5Y Aβ42-GFP (SH-Aβ42-GFP); and (vi) SH-SY5Y Aβ42DM-GFP (SH-Aβ42DM-GFP).
Flow cytometry and cell sorting
For analysis of Aβ42-GFP, Aβ42DM-GFP or Aβ42DM-IRES-BFP expression in HEK293T or SH-SY5Y cells, cells were harvested, re-suspended with phosphate-buffered saline (PBS) and analyzed by a Beckman Gallios FACS (Faculty of Health Sciences, BGU). For sorting of HEK293T or SH-SY5Y cells stably expressing Aβ42-GFP, Aβ42DM-GFP, Aβ42DM-IRES-BFP or both Aβ42-GFP and Aβ42DM-IRES-BFP, cells were harvested and re-suspended with PBS supplemented with 10% FBS. Cells were analyzed for GFP or BFP expression and sorted using the BD Aria FACS (Ilse Katz Institute for Nanoscale Science & Technology, BGU). HEK293T cells stably expressing both Aβ42-GFP and Aβ42DM-IRES-BFP were re-sorted to establish cell populations showing low (L) or high (H) Aβ42DM-IRES-BFP expression and similar Aβ42-GFP expression.
Aβ 1-42 peptide and Aβ 1-42-Hylight 488 labeled peptides were purchased from Anaspec (Fremont, CA). Aβ 1-42 F19S, L34P (Aβ42DM) peptide was synthesized and purchased from GL-Biochem (Shanghai, China).
Aβ seeds preparation
Aβ42 (40 μM) in the presence or absence of Aβ42DM at 1:65 molar ratio (Aβ42DM:Aβ42) was incubated in 500 μl of 20 mM sodium phosphate buffer, pH 7.4, 0.15 M NaCl at 37°C with orbital shaking for 72 h, followed by centrifugation at 20 K G for 1 h. Pellet was dissolved in 50 µl of PBS and aliquoted.
HEK293T or SH-SY5Y cells were grown on 1 cm micro-slides and transfected using Lipofectamine or turbofect, respectively, with a total of 0.5 μg DNA, as indicated in Table S1. For monitoring Aβ42-488 internalization, SH-SY5Y cells were incubated with 500 nM extracellular pre-formed Aβ42-488 labeled peptide fibrils in the presence or absence of 7.75 nM of the Aβ42DM peptide. Forty-eight hours post-transfection of Aβ42, or 18 h after adding the soluble Aβ42-488 peptide, images of live cells were taken using an Olympus FV1000 confocal microscope with a long-working distance ×60/1.35 numerical aperture oiled-immersion objective [The National Institute for Biotechnology in the Negev (NIBN), BGU]. Nuclear staining was conducted using Hoechst dye (Thermo-Fisher, Israel), diluted 1/2000 in PBS, while membrane staining was performed with FM 4-64 dye (10 μg/ml final concentration; Thermo-Fisher).
Intracellular protein aggregation quantification
Confocal images of HEK293T cells overexpressing Aβ42-GFP, Aβ42DM or a combination of both proteins at different expression levels were analyzed as above. Images of the number of cells containing Aβ42-GFP aggregates from 3 independent experiments were analyzed using the ImageJ software. Values were normalized to cells expressing only Aβ42-GFP. Quantitative analysis (n = 3) was performed by unpaired Student's t-test. **, P <0.05; ***, P < 0.005.
Quantification of Aβ42 internalization
SH-SY5Y cells were incubated with 500 nM of Aβ42-488 in the absence or presence of 7.75 nM Aβ42DM. Images were taken 18 h post-incubation with the Operetta high-content imaging system with ×40 magnification (NIBN, BGU). Nuclear and membrane staining were performed as described above. Images were analyzed using the Columbus software and levels of fluorescence intensity at 488 nm in the cytosol were determined. All measurements were done in triplicate and different regions of each well were imaged. Quantitative analysis (n = 3) was performed by unpaired Student's t-test. **, P <0.05; ***, P < 0.005.
Immunoprecipitation and western blot analysis
4 × 106 naïve or transformed HEK293T or SH-SY5Y cells were seeded in 10 cm plates. HEK293T cells were transfected to express Aβ42-GFP, Aβ42DM-Fc, or a combination of both using a total of 15 µg DNA. The cells were harvested and lysed with IP (immunoprecipitation) buffer [20 mM Tris–HCl, pH 7.6, 200 mM NaCl, 0.72 mM EDTA, 10% glycerol, 7 mM DTT, and protease inhibitor (PI) cocktail (Sigma–Aldrich, Israel) which was added fresh before use at a ratio of 1:200, and 0.15% Triton X-100]. The solubilized cell fraction was incubated with pre-blocked magnetic protein G Dynabeads (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The beads were magnetically isolated and washed three times with PBS, and suspended with sample buffer (6.25 mM Tris–HCl, pH 6.8, 10% glycerol, bromophenol blue, 2% SDS, 5% β-mercaptoethanol). The samples were heated for 5 min at 95°C and separated by 15% SDS–PAGE and subjected to WB (western blot). Proteins in the gel were transferred to a nitrocellulose membrane at 110 V for 45 min, which was probed with anti-GFP antibodies (Abcam, Cambridge, MA) diluted 1/1000 in PBS to detect Aβ42-GFP fusion protein, anti-Fc horseradish peroxidase (HRP)-conjugated antibodies (Jackson, West Grove, PA) diluted 1/1000 in PBS to detect Aβ42DM-Fc, or anti-β tubulin antibodies (Abcam) diluted 1/500 in PBS. HRP-conjugated anti-mouse and anti-rabbit antibodies (Jackson) were used as secondary antibodies. Antibody binding was revealed using an enhanced chemiluminescence kit (Biological Industries).
To monitor the level of transiently expressed Aβ42-GFP or Aβ42DM-GFP, cells were grown on 10 cm plates (107 cells per plate), harvested, and washed with PBS. Cell pellets were then suspended with RIPA buffer (25 mM HEPES-NaOH, 150 mM KCl, 1 mM EDTA, 1% Triton X-100, 0.1% NP-40, pH7.4, fresh (PIs cocktail (1:200; Sigma–Aldrich). Protein concentrations were determined by Bradford assay (Bio-Rad, Hercules, CA) and 100 μg of total cell lysates was supplemented with sample buffer before 15% SDS–PAGE. Following transfer to nitrocellulose membranes, the separated proteins were probed with anti-Aβ42 IgG (Sigma–Aldrich) or anti-β tubulin antibodies (Abcam) as described above.
To monitor Aβ42 fibrils and aggregate formation in the absence or presence of Aβ42DM, 360 ng of Aβ42 with or without 6 or 0.75 ng of Aβ42DM, or 6 ng of Aβ42DM alone was supplemented with sample buffer, boiled, and separated on a 4–15% gradient SDS–PAGE gel (Bio-Rad). WB was performed using anti-Aβ42 antibodies (Sigma–Aldrich) as described above.
Dot blot experiment
Aβ42 samples (5 μl) incubated for 18 h were applied to nitrocellulose membranes. The membrane was blocked for 1 h with 5% non-fat milk in PBS followed by incubation with A11 antibodies (Thermo-Fisher) at 1:1000 dilution in PBS containing 5% non-fat milk. HRP-conjugated anti-rabbit antibodies (Jackson) were used as secondary antibodies. Antibody binding was revealed using an enhanced chemiluminescence kit (Biological Industries).
XTT cell viability assay
The viability of HEK293T and SH-SY5Y stable cell lines (described above) or SH-SY5Y cells incubated with Aβ42 fibrils in the absence or presence of different concentrations of Aβ42DM (3.5–32 nM) was assessed using an XTT-based kit (Biological Industries) according to the manufacturer's protocol. Briefly, 104 cells per well were seeded in a 96-well plate. Twenty-four hours after treatment, cell viability was determined. Naïve HEK2393T or SH-SY5Y cells served as controls for these experiments. Data for each assay were normalized relative to the control and are presented as percentage of control. Experiments were performed in triplicate. Quantitative analysis (n = 3) was performed by unpaired Student's t-test. ** P < 0.05; *** P < 0.005.
Thioflavin T aggregation assay
Assays were performed, as described previously . Briefly, Aβ42 (4 μM) with or without Aβ42DM was incubated in 200 μl of 20 mM sodium phosphate buffer, pH 7.4, 0.15 M NaCl to which 10 μM of ThT (thioflavin T) (Sigma–Aldrich, Israel) was added. Reactions were performed in a black 96-well plate at 37°C shaken at 300 rpm in a Tecan Infinite M1000 plate reader with fluorescence excitation and emission wavelengths set at 440 and 485 nm, respectively. ThT fluorescence was measured at 10 min intervals. Each experiment was performed in triplicate.
Monitoring Aβ42 oligomerization by light scattering
Aβ42 peptides (4 μM) were incubated in 200 μl of 20 mM Na+/phosphate buffer, pH 7.4, 0.15 M NaCl, in a black 96-well plate (transparent bottom) at 37°C with 250 rpm continuous orbital shaking using the Infinite M1000 (Männedorf, Switzerland; Cytometry and Proteomics unit, NIBN, BGU) plate reader in the presence or absence of Aβ42DM. Light scattering was monitored at 400 nm over 18 h of incubation.
CD (circular dichroism) spectra were recorded on a Jasco J715 spectropolarimeter over a range of 190–260 nm using a quartz cuvette with a path length of 1 mm, a scanning speed of 50 nm/min, and a data interval of 1 nm, at 25°C. Twenty micromolars Aβ42 samples in the absence or presence of Aβ42DM (62 nM) were incubated for 18 h in 20 mM Na+/phosphate buffer, pH 7.4, 0.15 M NaCl. Each sample was scanned four times at time zero and 18 h post-incubation. Scans were averaged to obtain smooth data and background corrected with respect to protein-free buffer.
TEM sample preparation
Samples for TEM imaging were prepared as described . Briefly, after 70 h of incubation, 2.5 μl aliquots of the samples (see ThT assay samples) were deposited on carbon-coated copper 300 mesh grids. After 1 min, excess liquid was carefully blotted with filter paper, and the grid was held at ambient temperature for another minute. Uranyl acetate (2%, 5 μl) was added for 1 min, after which time any excess solution was carefully removed with filter paper. Imaging was performed using a Tecnai G2 12 BioTWIN (FEI) (Ilse Katz Institute for Nanoscale Science & Technology, BGU) TEM with an acceleration voltage of 120 kV. Variable magnifications were used for visualization, depending on the size of the aggregates.
Surface plasmon resonance spectroscopy
The constants for the binding of Aβ42DM to Aβ42 were determined by SPR (surface plasmon resonance) spectroscopy on a ProteOn XPR36 (Bio-Rad) , as follows. Aβ42DM or Aβ42 was immobilized on the surface of the chip using the amine coupling reagents, sulfo-NHS (0.1 M N-hydroxysuccinimide) and EDC [0.4 M 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide] (Bio-Rad). Aβ42DM or Aβ42 (20 µg) was covalently immobilized on the chip in 10 mM sodium acetate buffer, pH 4.0, to obtain 1663 response unit (RU). Bovine serum albumin (BSA; 3 μg, 3432 RU) was immobilized on the chip as a negative control, using a different flow channel. Unbound esters were deactivated with 1 M ethanolamine HCl at pH 8.5. Before each binding assay, the temperature was set at 25°C. Soluble human Aβ42 or Aβ42DM (the analyte) was then allowed to flow over surface-bound Aβ42DM/Aβ42 at concentrations of 0, 0.125, 0.25, 0.5, 1, and 2 μM and a flow rate of 50 µl/min. While the analyte was flowing over the surface (for 16 min), interactions between Aβ42DM and Aβ42 were determined. Dissociation of the proteins was next examined by allowing PBST (PBS, 50 mM NaCl, 0.05% Tween 20) to flow over the surface for 28 min at 50 µl/min. After each run, a regeneration step was performed with 50 mM NaOH at a flow rate of 100 μl/min. For each protein complex, a sensorgram was generated from the RUs measured over the course of the protein–protein interaction minus the values from the BSA background channel. The kon and koff were determined by fitting the data points using a 1:1 Langmuir kinetic model . The fitted data were considered statistically valid as the χ2 value was less than the 10% of the Rmax in the fitted graphs .
Aβ42DM suppresses Aβ42 aggregation in cells
In the present study, we initially traced Aβ42 intracellular expression and aggregation in human cells. Previous studies showed that an Aβ42-GFP fusion protein could serve as a reporter for protein misfolding and aggregation in bacteria, since Aβ42 aggregates more rapidly than the GFP protein folds, causing GFP to misfold and hence lose its fluorescence [29,30]. In contrast, the Aβ42 variant mutant Aβ42F19S,L34P (Aβ42DM) does not aggregate. Indeed, when fused to GFP, proper folding and preserved GFP fluorescence ensue. To verify that expression levels and aggregation patterns of Aβ42-GFP and Aβ42DM-GFP were similarly maintained in human cells, we expressed Aβ42-GFP or Aβ42DM-GFP in HEK293T cells and followed GFP fluorescence as a marker Aβ42 aggregation by flow cytometry and confocal microscopy (Figure 1). Our results show that in human cells, Aβ42-GFP displayed low intensity fluorescence, indicating that the GFP moiety was not properly folded. This was accompanied by visible intracellular protein aggregation, probably reflecting Aβ42-GFP inclusion bodies found in the cytoplasm of the HEK cells (Figure 1A,B). In contrast, Aβ42DM-GFP exhibited 10-fold higher intensity fluorescence than Aβ42-GFP, implying correct folding. Moreover, Aβ42DM-GFP displayed a diffuse and non-aggregated pattern of expression (Figure 1A,B). These findings confirm the earlier work performed in bacteria and demonstrate the distinctive folding patterns of Aβ42 and Aβ42DM when expressed in mammalian cells [28,39,40].
Aβ42DM prevents Aβ42 aggregation in cells.
To ensure that the differences in GFP fluorescence levels and Aβ42 aggregation (i.e. the amount of inclusions bodies) displayed by Aβ42 or Aβ42DM are not a result of different expression levels of the two proteins in HEK293T, we tested the expression of both proteins by WB analysis. Surprisingly, Aβ42DM was expressed 2.8-fold better than Aβ42 (Figure 1C). Comparing the 2.8-fold difference in the expression of Aβ42DM over that of Aβ42 to the 10-fold difference in fluorescence suggests that the increased fluorescence of Aβ42DM was not only due to augmented expression of this variant. To validate that the differences in aggregation patterns and fluorescence/expression levels between Aβ42-GFP and Aβ42DM-GFP in HEK293T cells were also the case in neuronal cells, we expressed Aβ42-GFP or Aβ42DM-GFP in the SH-SY5Y neuroblastoma cell line, a well-established cell model for neuronal disorders (Supplementary Figure S2). Our results showed that, similar to what was seen in HEK293T cells, Aβ42-GFP expression resulted in low levels of diffused fluorescence in the cytoplasm, which was accompanied by protein aggregates (Supplementary Figure S2B). At the same time, Aβ42DM-GFP expression led to high levels of diffused fluorescence and no inclusions bodies (Supplementary Figure S2D).
To monitor the ability of Aβ42DM to inhibit Aβ42-GFP aggregation, we transiently co-expressed the two proteins in HEK293T cells and monitored Aβ42-GFP aggregation using confocal microscopy (Figure 1D,E). Our results showed that when co-expressed, Aβ42DM suppressed Aβ42-GFP aggregation up to 90%, relative to what was seen in cells that expressed Aβ42-GFP alone. Significantly, this reduction was Aβ42DM-dose dependent, as increasing the amount of Aβ42DM led to a reduction in the numbers of Aβ42-GFP inclusion bodies seen. To ensure that the inhibitory effect of Aβ42DM was not due to interactions between Aβ42DM and GFP, we performed an IP experiment with Aβ42DM and GFP in HEK293T cells; no association was detected (Supplementary Figure S4). These findings highlight the ability of Aβ42DM to inhibit Aβ42 aggregation in human cells.
Aβ42DM associates with Aβ42
Having demonstrated that Aβ42DM effectively inhibits Aβ42 intracellular aggregation in HEK293T cells, we next sought to elucidate the mechanism involved. To this end, we monitored the interaction between Aβ42DM and Aβ42 in cells. Aβ42-GFP and Aβ42DM-Fc-fusion proteins were co-expressed in HEK293T cells, followed by co-IP with protein G beads and followed by WB analysis. Our results showed that Aβ42DM is associated with Aβ42 in cells (Figure 2A).
Interaction between Aβ42DM and Aβ42.
To confirm direct binding between Aβ42 and Aβ42DMin vitro and to further illuminate the effect of Aβ42DM on Aβ42 aggregation kinetics, we used Aβ42DM and Aβ42 peptides and monitored their direct binding by SPR spectroscopy. The parameters of binding were computed by fitting the SPR sensorgrams using a Langmuir 1:1 binding model (Figure 2B,C). Our results showed that when immobilized to the chip, Aβ42DM binds Aβ42 with kon and koff rates of 300 M−1 s−1 and 1.6 × 10−4 s−1, respectively, with the KD for the interaction being 520 nm (Figure 2B). However, when Aβ42 is immobilized, no binding was detected (Figure 2C). The fusion of Aβ42 with the chip forced the protein to persist as a monomer, whereas in solution, the protein started to undergo oligomerization. Consequently, these results imply that Aβ42DM binds to oligomeric but not monomeric Aβ42.
Aβ42DM suppresses Aβ42 aggregation and fibril formation in vitro
To monitor the effect of Aβ42DM on Aβ42 aggregation kinetics in vitro, we co-incubated the two peptides at different molar ratios and followed the progress of Aβ42 amyloid aggregation by staining with ThT, an amyloid-specific staining dye (Figure 3A). Our results show that Aβ42 exhibits a typical aggregation pattern with a ∼2 h lag phase (Supplementary Figure S5A) and that low concentrations of purified Aβ42DM (15 nM) were sufficient to significantly suppress Aβ42 aggregation (at 4 μM) (Figure 3A). Importantly, Aβ42DM at 7.8–62 nM did not self-aggregate or form fibrils, as shown by ThT staining (Supplementary Figure S5B). In addition, TEM images revealed the presence of fibrillar aggregates of untreated Aβ42 after 60 h of incubation. This experiment further showed that adding Aβ42DM and Aβ42 at molar ratios of ∼1:266 or ∼1:65 (Aβ42DM:Aβ42) significantly reduced Aβ42 fibrillation (Figure 3B). This was also shown by WB analysis (Figure 3C), demonstrating significant inhibition of fibril formation in the presence of Aβ42DM at a 1:65 (Aβ42DM:Aβ42) molar ratio. Interestingly, adding Aβ42DM after 24 h of Aβ42 aggregation did not affect its further aggregation, suggesting that Aβ42DM does not interact with high molecular-weight oligomers (probably pre-fibrils) that appeared at this time point (Supplementary Figure S6).
Aggregation kinetics and morphology of Aβ42 aggregates in the presence or absence of Aβ42DM.
Aggregated Aβ42 seeded with monomeric Aβ42 is known to induce and increase the aggregation of monomeric Aβ42 . To further evaluate the ability of Aβ42DM to suppress Aβ42 aggregation, we examined its inhibitory activity by measuring the progression of Aβ42 amyloid aggregation by staining with ThT upon addition of aggregated Aβ42 seeds (Figure 4A). Our results show that the lag phase of Aβ42 was decreased and that aggregation was increased upon the addition of Aβ42 seeds, whereas Aβ42DM was not recruited to aggregate by the seeds. Moreover, Aβ42DM significantly reduced lag phase shortening and the aggregation-inducing effect of the seeds on Aβ42. Importantly, seeds generated from a mixture of Aβ42 and Aβ42DM and reacted with Aβ42 did not increase Aβ42 aggregation and exhibited a strong inhibitory effect (Figure 4B), in terms of both aggregation suppression and lag phase extension, reinforcing Aβ42DM as an efficient inhibitor of Aβ42 aggregation.
The inhibitory effect Aβ42DM has on the aggregation kinetics of seeded Aβ42.
Aβ42DM modulates the Aβ42 aggregation pathway
To better understand Aβ42DM mechanism of activity in both cell model and in vitro, we investigated whether Aβ42DM prevents Aβ42 self-assembly or otherwise allows Aβ42 oligomerization. We showed that Aβ42DM did not prevent Aβ42 oligomerization in the presence of Aβ42DM at a 1:65 molar ratio (Aβ42DM:Aβ42), as shown by light scattering (Figure 5A). Nevertheless, oligomers formed by Aβ42 were recognized by the A11 antibody (Figure 5B), which recognizes toxic ‘on-pathway’ Aβ oligomers [42,43]. However, in the presence of Aβ42DM, the recognition signal was reduced to 83%. These results imply that Aβ42DM did not prevent Aβ42 oligomerization but significantly modulated the aggregation pathway of Aβ42, leading to the formation of ‘off-pathway’ oligomers. This was also reinforced by CD analysis, showing that while being predominantly unstructured at t = 0, the oligomers that formed upon treatment with Aβ42DM for 18 h had less beta sheet content than did untreated Aβ42 (lower value of Aβ42 at 216 nm, when compared with treated Aβ42; Supplementary Figure S7). Considering those results and those depicted in Figure 4 showing that Aβ42 aggregates formed in the presence of Aβ42DM behave differently than do Aβ42 aggregates, it is probable that Aβ42DM modulates the Aβ42 aggregation pathway into less-toxic off-pathway oligomers.
Aβ42DM modulates the Aβ42 aggregation pathway.
Aβ42-induced cell toxicity is suppressed by treatment with intra- and extracellular Aβ42DM
Aβ42 intracellular aggregation is considered a pivotal step in inducing neuronal cell death . As our results showed that Aβ42DM inhibits Aβ42-GFP intracellular aggregation in HEK293T cells (Figure 1), we further tested whether Aβ42DM can also protect these cells from intracellular Aβ42-induced toxicity. To do so, we compared the viability of cells expressing Aβ42 to that of cells expressing supposedly non-toxic Aβ42DM, or to that of cells expressing both proteins, when naïve HEK293T cells served as control. Aβ42-GFP and Aβ42DM-IRES BFP were stably expressed in cells respectively designated as HEK-Aβ42-GFP and HEK-Aβ42DM (Figure 6A). Similarly, stable expression of Aβ42-GFP and Aβ42DM-GFP was established in the SH-SY5Y neuroblastoma cell line (Figure 4C,D). In addition, we also generated cells expressing both proteins (designated as HEK-Aβ42-GFP-Aβ42DM). Two cell populations that equally expressed Aβ42-GFP but differentially expressed Aβ42DM-IRES-BFP were FACS-sorted on the basis of their Aβ42DM-IRES-BFP expression levels. The two sorted populations correspond to cells expressed either low (HEK-Aβ42-GFP-Aβ42DML) or high (HEK-Aβ42-GFP-Aβ42DMH) Aβ42DM levels (Supplementary Figure S8A–E), the latter expressing Aβ42DM ∼2.5-fold over the level of the low expression population. We then followed Aβ42-GFP-induced cell toxicity by the XTT cell viability assay. In agreement with the known toxic effects of Aβ42, our results showed that HEK-Aβ42-GFP cells were ∼30% less viable, relative to naïve HEK293T cells or to HEK-Aβ42DM cells (Figure 6A). Moreover, as the toxic effect of Aβ42 is known to cause neuronal cell death, we also examined the toxicity of stable Aβ42-GFP expression by the SH-SY5Y neuroblastoma cell line and noted a toxic effect similar to that seen in HEK293T cells (Figure 6C). Importantly, upon monitoring cell toxicity in HEK-Aβ42-GFP-Aβ42DML and HEK-Aβ42-GFP-Aβ42DMH cells, as opposed to HEK-Aβ42-GFP cells, where no inhibitory effect was observed, we found that Aβ42-induced toxicity was reduced in a manner dependent on Aβ42DM expression levels, being completely abolished in the presence of high Aβ42DM expression (Figure 6A). This implies that Aβ42DM is highly efficient as an Aβ42 inhibitor, blocking Aβ42-mediated aggregation and cell toxicity.
Aβ42DM reduces Aβ42 toxicity in cells.
In addition to intracellular Aβ42 toxicity, Aβ42 extracellular fibrils and aggregates are also known to induce neuronal cell death . Therefore, we tested the ability of Aβ42DM to inhibit Aβ42-mediated toxicity upon incubation with extracellular aggregates in SH-SY5Y cells incubated with Aβ42 fibrils in the presence or absence of Aβ42DM peptide, with cell viability being monitored by the XTT assay. The assay was performed at 1:571, 1:266, and 1:65 molar ratios of Aβ42DM and Aβ42 (2 µM of Aβ42 and 7, 15, or 31 nM of Aβ42DM, respectively; Figure 6B). Our results demonstrated that Aβ42 fibrils that were pre-formed extracellularly (as confirmed by ThT and TEM, Figure 3A,B, respectively), reduced cell viability by ∼40–50%, relative to control SH-SY5Y cells that had not been treated with aggregated Aβ42 fibrils. Not surprisingly, 32 nM Aβ42DM alone (the highest concentration tested) had only minor effects on cell viability (∼10% reduction). Importantly, in SH-SY5Y cells introduced to Aβ42 that had been allowed to aggregate in the presence of different concentrations of Aβ42DM, Aβ42-mediated toxicity was significantly reduced in a dose-dependent manner, and exhibited ∼90% of the viability of naïve cells (Figure 6B).
Having demonstrated that Aβ42DM can prevent Aβ42-mediated toxicity when the two were co-expressed in the cell or when added externally together following Aβ42 aggregation (Figure 6A,B), we tested whether the external addition of Aβ42DM could prevent the intracellular Aβ42-mediated toxicity in neuronal SH-SY5Y cells stably expressing Aβ42, and whether neuronal cells stably expressing Aβ42DM were less affected by the external addition of Aβ42 fibrils. For this, we generated SH-SY5Y cells stably expressing Aβ42-GFP (designated SH-Aβ42-GFP) or Aβ42DM-GFP (designated SH-Aβ42DM-GFP), as confirmed by WB analysis (Supplementary Figure S8F). An XTT cell viability assay was performed with SH-Aβ42-GFP cells incubated in the presence of extracellular Aβ42DM peptide at different concentrations (10–500 nM; Figure 6C). Our results indicated that SH-Aβ42-GFP cells were ∼35–40% less viable, relative to naïve SH-SY5Y cells, or to naïve SH-SY5Y cells treated with externally added Aβ42DM (500 nM). However, upon the addition of Aβ42DM to SH-Aβ42-GFP cells at concentrations of 10, 50, or 500 nM, Aβ42-GFP-mediated cell toxicity was significantly reduced in a dose-dependent manner (Figure 6C). In particular, when treated with 500 nM Aβ42DM, SH-Aβ42-GFP cells exhibited increased viability up to 91%, relative to control cells (naïve SH-SY5Y cells) (Figure 6C).
We next monitored the viability of SH-Aβ42DM-GFP cells incubated with 2 µM extracellular purified Aβ42 fibrils (Figure 6D). Our results show that upon extracellular addition of 2 µM Aβ42 peptide, the viability of naïve SH-SY5Y cells was reduced by ∼40–50%. SH-Aβ42DM-GFP cells were, however, resistant to the toxic effects of Aβ42 and their viability was similar to that of non-treated SH-Aβ42DM-GFP cells (∼5% reduction).
In relation to the diffuse non-aggregating expression pattern of Aβ42DM seen in cells (Figure 1D), the results presented in Figure 6A–D indicated that the non-aggregating variants of Aβ42, such as Aβ42DM, had a lower toxic effect in the cells, when compared with Aβ42. These results further confirm that protein aggregation plays a major role in induced neuronal cell death.
Aβ42DM peptide prevents Aβ42 uptake by cells
Given our demonstration that Aβ42DM significantly reduced the toxicity of intracellular and extracellular Aβ42 in both neuronal and non-neuronal cells (Figure 6), we looked to further elucidate the mechanism of the inhibitory effect. As re-uptake of extracellular Aβ42 protein aggregates is considered to be a key determinant for Aβ42-induced neuronal cell death , we next analyzed whether Aβ42DM had an effect on Aβ42 uptake by cells. For this, we used Hilyte-488-labeled Aβ42 peptide (designated Aβ42-488), which was allowed to extracellularly aggregate and form fibrils in the absence or presence of Aβ42DM at a molar ratio of 1:65 (Aβ42DM:Aβ42-488) before being introduced into SH-SY5Y neuronal cells. Aβ42-488 entry into and localization within cells was monitored by confocal microscopy (Figure 7). Our results showed that 18 h post-incubation, the Aβ42-488 peptide had formed aggregates and was localized as inclusion bodies inside the cells (Figure 7). However, when Aβ42-488 was incubated with Aβ42DM, most of the labeled peptide was localized on the cell membrane and fewer aggregates were seen, suggesting that the labeled peptide could not enter the cell (Figure 7). Quantifying the accumulation of Aβ42-488 within the cells using the Operetta high-content imaging system showed a decrease in ∼50% in Aβ42-488 entry into cells in the presence of Aβ42DM (Supplementary Figure S9). We conclude that Aβ42DM prevents Aβ42 uptake by the cells, providing a possible explanation for the Aβ42-mediated inhibition of cell toxicity observed in Figure 6B.
Aβ42DM prevents Aβ42 penetration into cells.
In the present study, we investigated the potential of Aβ42F19S,L34P (Aβ42DM), an Aβ42 variant carrying two mutations, as an inhibitor of Aβ42 aggregation. We showed that Aβ42DM suppressed Aβ42 aggregation in vitro and modulated its aggregation pathway into non-toxic off-pathway aggregates, as well as intracellular Aβ42 aggregation, and reduced Aβ42-mediated cell toxicity in neuronal and non-neuronal cells. We also demonstrated that Aβ42DM prevented internalization of pre-formed Aβ42 aggregates, providing a mechanistic explanation for its inhibitory effects.
Aβ42DM was previously characterized as a non-aggregating mutant of Aβ42 that cannot self-assemble within the cell when expressed in bacteria or in mammalian cells [30,39]. We were thus interested in testing whether Aβ42DM still binds Aβ42 and prevents Aβ42 aggregation by acting as a dominant-negative inhibitor with antagonistic activity. This assumption was based on a recent study highlighting the potential of the Aβ42 variant Aβ42G37L to efficiently inhibit Aβ42 aggregation and the toxicity that follows . Accordingly, we examined the ability of Aβ42DM to efficiently inhibit Aβ42 aggregation both in vitro and in a variety of cell model-based assays that mimic and address both the intracellular and extracellular aggregation of Aβ42 and further examined the effects of Aβ42DM on both intracellular and extracellular aggregated Aβ42-mediated cell toxicity.
It has been previously shown that the self-assembly of Aβ42 is crucial for its cellularly mediated toxic effects. A recent study showed that an Aβ42 variant (Aβ42F19S,G37D) that is incapable of self-assembly and oligomerization is non-toxic to cells . This observation is with agreement with our results which also show that Aβ42DM lacks the ability to aggregate in cells (Figure 1) or in vitro (Supplementary Figure S4) and is also non-toxic to cells (Figure 6). Importantly, by disturbing Aβ42 aggregation (Figures 1 and 3), Aβ42DM also significantly reduced both intracellular and extracellular Aβ42-mediated toxicity (Figure 6).
The classical amyloid cascade hypothesis suggests that extracellular amyloid plaques are the main driving force in AD pathogenesis [44,47]. However, more recent data also highlight the crucial role of intracellular Aβ42 accumulation and aggregation [9,11,45,48]. Our results show that intracellular Aβ42 is toxic to cells and reduces both neuronal and non-neuronal cell viability (Figure 6). As it is challenging to model the toxic effects of Aβ42 in cell culture, we believe that it is important to develop cell-based assays to better understand the pathogenic mechanism of Aβ42 at the cellular level.
To provide a mechanistic explanation for the effects of Aβ42DM and to elucidate whether Aβ42DM suppresses intracellular and extracellular Aβ42 aggregation through direct binding, we tested whether direct interactions between Aβ42DM and Aβ42 exist by SPR (Figure 2B,C). We found that Aβ42DM does not bind monomeric Aβ42 (Figure 2C), yet it does bind oligomeric Aβ42 (Figure 2B). We further analyzed Aβ42DM activity against Aβ42 aggregation by in vitro ThT staining, TEM, and assessing oligomerization levels (Figure 3). As we showed that Aβ42DM suppressed Aβ42 aggregation, we aimed to better understand the mechanism of Aβ42DM-mediated inhibition. We showed that Aβ42DM allowed the formation of Aβ42 oligomers, as revealed by light scattering (Figure 5A); however, these oligomers were significantly less well-recognized by the A11 antibody, which recognizes toxic ‘on-pathway’ Aβ oligomers [42,43]. Moreover, in contrast with Aβ42 aggregates that induce monomeric Aβ42 aggregation and shorten the lag phase when added as seeds to the Aβ42 monomers (Figure 4A), Aβ42 aggregates formed in the presence of Aβ42DM did not induce aggregation of the monomers or shorten the lag phase, and surprisingly, acted as potent inhibitors, reducing the Aβ42 monomer aggregation and extending the lag phase (Figure 4B). Finally, aggregates formed in the presence of Aβ42DM suppressed toxicity in a dose-dependent manner, in comparison with untreated Aβ42 aggregates (Figure 6B). As we also showed by ThT that Aβ42DM does not affect Aβ42 aggregation when added after 24 h (Supplementary Figure S6), we conclude that Aβ42DM interacts with low molecular-weight oligomers but not with either high molecular-weight (or prefibrillar) oligomers or monomers of Aβ42 and thus directs these low molecular-weight oligomers into ‘off-pathway’ non-toxic aggregates.
In fact, these features above make Aβ42DM different from other Aβ peptides, such as Aβ40, for example. Although sharing high sequence similarity with Aβ42 and Aβ42DM, Aβ40 can act as both an inhibitor and an inducer of Aβ42 aggregation [41,50–52]. Specifically, Aβ40 inhibits the aggregation of Aβ42 monomers in a concentration-dependent manner [50–52]. On the other hand, it has been shown that cross-seeding of Aβ40 fibrils and Aβ42 monomers promotes Aβ42 aggregation . The cross-seeding effects of Aβ40 and Aβ42 probably result from the sequence similarity and structural plasticity of the two proteins, allowing Aβ42 monomers to adopt the fibril structure of Aβ40 seeds and promoting Aβ42 aggregation in the presence of these seeds . In contrast with Aβ40 seeds that induce Aβ42 aggregation, the seeds formed in the presence of Aβ42DM did not induce Aβ42 aggregation but instead acted as a potent inhibitor. This difference in mode of action of Aβ40 and Aβ42DM is probably a result of the two Aβ42DM mutations at positions 19 and 34 (both lie within the β-sheet hydrophobic/aggregation-prone regions 17–20 and 31–36  of Aβ42), which do not allow fibril formation . In contrast, these two regions are intact and have wild-type sequences in Aβ40.
Our results show that the binding affinity between Aβ42DM and Aβ42 is relatively low (KD = 520 nM), as compared with those of other inhibitors. For example, a recent study describing new inhibitors of Aβ42-mediated aggregation showed that several Anticalins, which are based on the human lipocalin affinity scaffold and were screened against Aβ40, displayed binding affinities to Aβ40 and Aβ42 of less than 100 pM . However, for these inhibitors to be efficient in inhibiting amyloid aggregation and reducing cell toxicity, a 1:1 molar ratio of the inhibitor and its target, Aβ40, was required. In contrast, Aβ42DM is effective in reducing Aβ42 fibril formation in vitro, even at a molar ratio of 1:266 (Aβ42DM:Aβ42).
Considering the binding kinetics of Aβ42 to Aβ42DM and the inhibition efficiency of Aβ42 aggregation by Aβ42DM, i.e. low binding affinity of Aβ42DM to Aβ42 (KD of 520 nM) and high inhibition efficiency for Aβ42DM against Aβ42, we predict that Aβ42DM rapidly associates with and dissociates from Aβ42 (with kon and koff rates of 300 M−1 s−1 and 1.6 × 10−4 s−1, respectively), such that each Aβ42DM molecule potentially binds to and dissociates from multiple Aβ42 oligomers in a sequential manner, thereby disturbing and modulating oligomer self-assembly. Consequently, the very low molar ratio of 1:266-1:65 (as seen in Figure 3) is sufficient to suppress Aβ42 aggregation and to reduce Aβ42 toxicity, as seen in Figure 6B. Since other inhibitors with higher binding affinity, such as the anticalins noted above , require higher molar ratios (in favor of the inhibitor) to bind and efficiently inhibit their targets, Aβ42DM offers a major advantage in terms of activity efficiency and potential therapeutic advantage, as the low doses required result in reduced toxicity and side effects.
Finally, re-uptake of extracellular aggregates into neuronal cells is thought to be a key factor in inducing neuronal cell death . As such, we explored whether Aβ42DM affects Aβ42 uptake and internalization by cells to better understand the inhibition of extracellular Aβ42-mediated cell toxicity by soluble Aβ42DM, as shown in Figure 6B. Aβ42 interacts with several different transporters on the membrane that permit entry into neuronal cells [10,54–56]. Our results show that labeled Aβ42 peptide (Aβ42-488) entered SH-SY5Y neuronal cells after 18 h of incubation with the cells (Figure 7). However, the presence of Aβ42DM in a molar ratio of 1:65 (Aβ42DM:Aβ42-488) significantly prevented Aβ42-488 internalization (Figure 7). This result matches our data demonstrating that external addition of synthetic Aβ42 caused a toxic effect that was abolished upon expression of Aβ42DM or upon addition of Aβ42DM to the medium as soluble protein (Figure 6B,D).
These results provide a possible explanation for the inhibition of Aβ42-mediated toxicity, namely as the result of Aβ42 uptake and internalization into cells, where it induces toxic effects and cell death. The presence of Aβ42DM not only disturbed the aggregation process and modulated the aggregation pathway but also prevented Aβ42 internalization and thus protected cells from Aβ42-mediated toxicity. Considering our results showing the protective activity of Aβ42DM against intracellular and extracellular Aβ42 when expressed in cells or added externally as soluble protein, it would appear that Aβ42DM has the potential to be used therapeutically as soluble protein or as intracellularly expressed protein.
In summary, our results highlight the efficient activity of Aβ42DM against both intracellular and extracellular aggregation of Aβ42. In the past, many inhibitors were designed to disrupt Aβ42 aggregation; however, most were addressed solely in terms of their ability to prevent extracellular aggregation [57,58], although some were addressed in terms of their ability to prevent intracellular aggregation [28,39]. Still, there is a lack of knowledge on inhibitors which target both intracellular and extracellular aggregates. As emerging evidence suggests that both intracellular and extracellular aggregation induces neuronal cell death, it is important to develop inhibitors able to inhibit both the intra- and extracellular aggregation, such as Aβ42DM. We thus suggest further exploration of Aβ42DM as a novel AD therapeutic.
amyloid precursor protein
fetal bovine serum
internal ribosome entry site
surface plasmon resonance
transmission electron microscope
O.O., R.T., and N.P. designed the research; O.O. and V.B. performed the research; O.O., R.T., and N.P. analyzed the data; O.O., R.T., and N.P. wrote the paper. All authors edited the manuscript and approved the final version.
This work was supported by the US-Israel Binational Science Foundation [contract grant number: 2015134] to N.P. R.T. acknowledges support from the Israel Science Foundation [contract grant number: 755/17].
The authors thank Ravit Malishev, Prof. Raz Jelinek, Shiran Lacham, Bar Dagan, Dr Alon Zilka, and Dr Uzi Hadad for their technical assistance. FACS experiments were performed at the Ilse Katz Institute for Nanoscale Science and Technology, BGU. Confocal microscopy experiments were performed at the NIBN proteomics unit, BGU.
The Authors declare that there are no competing interests associated with the manuscript.