Background
Lewy bodies (LBs), which are the cardinal histological hallmark of Parkinson’s disease (PD), contain abnormal filamentous α-synuclein (αSYN) aggregates. In addition, a variety of other neurodegenerative diseases are associated with αSYN-positive lesions [
1,
2]. The presence of αSYN in LBs, Lewy neurites or glial cytoplasmic inclusions (GCIs) in PD and related disorders provides a conceptual link that has led to the use of the term ‘synucleinopathy’ to encompass these diseases [
3‐
5]. Emerging evidence from genetic and biochemical studies demonstrates that abnormal αSYN aggregates directly contribute to neurodegeneration in PD and other synucleinopathies [
6‐
10]. Because αSYN is enriched in the presynaptic nerve terminals and is mainly detected in the cytosolic and synaptosomal fractions, it has long been believed that αSYN exerts its physiological as well as pathogenic effects intracellularly [
11,
12]. However, accumulating evidence suggests that both monomeric and oligomeric αSYN can be secreted into the extracellular milieu, thereby affecting the normal physiological state of neighboring cells [
13‐
17]. For example,
in vitro-generated soluble αSYN oligomers can induce the transmembrane seeding of αSYN aggregation and can eventually lead to cell death [
18,
19]. Moreover, exogenous αSYN fibrils have been shown to induce perikaryal LB pathology, which can lead to synaptic dysfunction and neuronal cell death [
20]. The existence of the transcellular spread of αSYN has also been verified by co-culture experiments and
in vivo animal models, which showed that αSYN aggregates released from neuronal cells can be transferred to neighboring cells and form inclusion bodies [
21‐
23]. Finally, the existence of an
in vivo intercellular propagation of αSYN aggregates was supported by recent observations of LB-like inclusions in the grafted neurons of PD patients who had received transplants of fetal mesencephalic neurons more than a decade previously [
24‐
26]. In addition to PD, the intercellular transmission of αSYN pathology can be assumed to be present in multiple system atrophy (MSA), in which widespread αSYN-positive GCIs are found in oligodendroglia, a type of brain cell that does not normally express αSYN [
27‐
29]. Phenomenologically, the propagation theory is also attractive as an explanation for the hierarchical distribution of Lewy pathology in PD, a theory proposed by Braak and colleagues [
30]. To understand how αSYN travels from cell to cell, the underlying mechanisms responsible for αSYN uptake and secretion must be elucidated. In this study, we provide evidence to support the functional role of dynamin-mediated endocytosis in the process of αSYN uptake by neurons and oligodendroglial cells. Furthermore, we propose therapeutic strategies aimed at reducing the propagation of protein misfolding in synucleinopathies.
Discussion
Although αSYN is generally located in the cytoplasm, several studies have demonstrated that αSYN has an affinity for phospholipids and vesicles [
44,
45]. In addition, αSYN is known to be delivered to the plasma membrane through its interactions with the endoplasmic reticulum (ER)-Golgi secretory pathway [
46]. Furthermore, increasing evidence has suggested that nanomolar concentrations of αSYN can be found in neuronal culture medium, as well as in body fluids such as plasma and cerebrospinal fluid [
47,
48]. The secreted monomeric and oligomeric forms of αSYN were shown to re-enter neighboring cells, resulting in various cytotoxic effects, such as the production of reactive oxygen species, the generation of a glial cell inflammatory response [
18,
21], and synaptic malfunction [
20]. In the present study, we showed that, in cultured neuronal and oligodendroglial cells, the incorporated αSYN oligomers were assembled into SDS-stable HMW oligomers and could form LB- and GCI-like cytoplasmic inclusions. Except for KG1C cells, which do not express endogenous αSYN, it is possible that the internalized recombinant αSYN may be entangled with endogenous αSYN, and the assembled HMW oligomers in the SH-SY5Y neuronal cells may consist of both endogenous and recombinant αSYN. Consistent with a previous study, we confirmed that most of the internalized αSYN was located within the intracellular soluble fraction rather than the membrane fraction, indicating that extracellular αSYN rapidly crosses the cellular membrane [
18]. Furthermore, wild-type αSYN has been shown to be degraded by lysosomes through chaperone-mediated autophagy, and decreased lysosomal function has also been observed in PD patients [
49‐
52]. In agreement with this result, we detected internalized αSYN in the endosomal compartment, where it was eventually degraded by the lysosomes. Our data showed that lysosomal inhibition impairs the autophagic flux in αSYN-treated neuronal and oligodendroglial cells. This observation strongly suggests that proper lysosomal machinery is required for the clearance of the incorporated αSYN oligomers and is therefore indispensable for the maintenance of cellular homeostasis.
In addition to the Lewy pathology in PD, the presence of oligodendrocytic αSYN deposition, i.e., GCI, in MSA has attracted a significant amount of attention [
29,
53‐
55]. However, regarding the pathogenesis of MSA, it remains unclear why αSYN accumulates in an oligodendrocytes, which do not normally express endogenous αSYN [
28]. It is possible that αSYN expression may be upregulated and/or inefficiently degraded in affected oligodendrocytes. Alternatively, it is plausible that oligodendrocytes may actively take up αSYN molecules derived from neurons. Indeed, endocytosis regulatory proteins such as Rab5 and Rabaptin-5 are known constituents of GCIs [
56]. It is also possible that an intrinsic protein, such as TPPP/p25α, may promote the aggregation of internalized αSYN within oligodendroglia [
34,
57,
58]. Interestingly, the prion-like hypothesis supports the possibility that aberrant oligodendroglial expression of αSYN may have an ectopic origin. Our study provides the first concrete evidence that extracellular αSYN can be incorporated and assembled into HMW oligomers and inclusions that mimic GCIs in cultured oligodendrocytes. In protein-folding diseases, protein misfolding and aggregation are thought to follow a ‘seeding-nucleation’ mechanism [
59‐
61], whereby misfolded αSYN is transmitted from a LB-bearing donor cell to a neighboring recipient cell and can act as a template for the conversion of native, unfolded αSYN into a β-sheet-rich conformation within the recipient cell. However, given that cultured oligodendroglial cells without inherent αSYN expression can import extracellular αSYN and form HMW oligomers, the existence of endogenous αSYN may not be a prerequisite for the buildup of αSYN aggregates in the recipient cells.
The precise mechanisms by which the intercellular transmission of αSYN occurs remain controversial. Lee and colleagues implicated exocytosis as a plausible mechanism for αSYN release from cultured neuronal cells because its release was effectively blocked under low-temperature conditions [
39]. As brefeldin A, an inhibitor of ER-Golgi-dependent exocytosis, is ineffective at preventing the secretion of nascent αSYN secretion by MES cells, αSYN exocytosis is thought to rely on an unconventional exocytic pathway [
62]. Intriguingly, the small GTPase Rab5, which is a known marker of early endosomes, is critical for the endocytosis of exogenous αSYN into neuronal cells [
18]. In a yeast model, the A30P mutant αSYN was shown to bind the endocytic cargo-transport protein YPP1 at the plasma membrane, which led to the budding of endocytic vesicles via receptor-mediated endocytosis and the subsequent targeting of this form of αSYN to the vacuole for degradation [
63]. Furthermore, previous studies by our lab and others have demonstrated that internalized αSYN is secreted from neurons via a process that is modulated by the recycling endosome regulator Rab11a [
64,
65]. In the case of prion disease, both the normal cellular prion protein (PrPc) and the abnormally folded pathogenic form (PrPsc) are associated with nanovesicles called ‘exosomes’ that are released from non-neuronal and neuronal cells [
66,
67]. However, the involvement of the exosomal vesicle in αSYN secretion remains unclear. αSYN has been shown to be secreted by the exosome, and exosome-containing conditioned medium can induce neuronal cell death [
68,
69]. In contrast, we recently demonstrated that extracellular αSYN was mainly detected in the supernatant fraction rather than in the exosome-containing pellets obtained from neuronal culture medium or CSF [
64]. Moreover, we found that the perturbation of exosome formation by a DN mutant of vacuolar protein sorting 4 (VPS4) not only interfered with lysosomal targeting of αSYN but also facilitated αSYN secretion [
64].
Regardless of the mechanisms involved in αSYN secretion, there is evidence to support the uptake of extracellular αSYN by neighboring cells, which subsequently facilitates aggregate formation. Previous reports have suggested that the internalization of αSYN oligomers may be mediated by the endocytic process; the overexpression of a DN-dynamin effectively reduced the extent of incorporated αSYN aggregates in cultured cell lines [
21,
39]. Furthermore, the inhibition of endocytosis by monodansylcadaverine and dynasore has also been shown to decrease αSYN uptake both
in vitro and
in vivo[
22]. Coincubation of αSYN pre-formed fibrils with WGA enhances the αSYN pathology in neuronal cells, indicating adsorptive-mediated endocytosis as the potential mechanism of αSYN internalization [
20]. The importance of the endocytic process in the uptake of extracellular αSYN is further supported by our findings showing that genetic as well as pharmacological disruption of the dynamin GTPases through the administration of sertraline, a widely used antidepressant, significantly decreased the internalization and translocation of αSYN. It should also be noted that the concentration of sertraline used in our study (10 μM = 3 μg/ml) is comparable to the concentrations observed to be therapeutically effective within the CSF and brain (2 μg/ml) for its antidepressant effects [
70]. In fact, neuropsychiatric manifestations such as depression, apathy, and anxiety are frequently encountered as non-motor symptoms in PD patients [
71,
72]. SSRIs are currently used as a first-line therapy for PD-associated depression [
73]. Thus, the identification of novel therapeutic uses for sertraline not only provides a strategy focused on the prevention of the pathological propagation of αSYN but also has the advantage of utilizing time-tested drugs for the benefit of patients. A recent study has shown that the first SSRI, fluoxetine, ameliorated behavioral and neuropathological deficits in an MSA mouse model [
74]. They concluded that the protective effect of fluoxetine might be attributed to the increased level of neurotrophic factors and/or the activation of the ERK signaling pathway; however, the reduction of αSYN-propagation through the inhibition of dynamin may be another underlying mechanism. Indeed, tricyclic antidepressants, which are also known to inhibit dynamin GTPases, have been shown to slow disease progression in PD [
41,
75]. In the case of MSA, a German group has reported the effectiveness of paroxetine in a small, short-term trial with 19 MSA patients [
76]. In this study, motor disabilities and dysarthria were significantly improved compared with the placebo group. It is also interesting to note that paroxetine may prevent the glottis stenosis in MSA patients [
77]. Further investigations with larger samples are necessary to assess the long-term safety and effectiveness of SSRI treatment in MSA. We are currently awaiting for the outcome of a double-blind, multicenter trial using fluoxetine for the treatment of MSA (MSA-Fluox) being conducted in France [
78].
In summary, we demonstrated that αSYN was taken up by neuronal and oligodendroglial cells, assembled into HMW oligomers, and formed cytoplasmic inclusions. Furthermore, we have provided evidence that αSYN uptake by neuronal and oligodendroglial cells is regulated by dynamin GTPases, which implies a role for the endocytic process in this activity. The importance of the vesicular trafficking machinery in the pathogenesis of PD is also highlighted by recent findings that a mutation in the
VPS35 gene, which encodes a retromer complex involved in the retrograde transport of proteins from the endosome to the trans-Golgi network, can cause late-onset familial PD [
79‐
81]. Furthermore, the prevention of αSYN-mediated pathology by sertraline is a potentially promising method for the treatment of PD and other synucleinopathies. Thus, defining the precise mode of intercellular αSYN transmission will shed light on the pathogenic mechanisms involved in synucleinopathies, and this research may pave the way for the identification of novel targets for therapeutic intervention in other neurodegenerative diseases.
Methods
Plasmid construction and preparation
For the bacterial expression of the GST-αSYN fusion protein, human αSYN cDNA was subcloned into the SalI and NotI restriction sites of the pGEX6P-1 vector (GE Healthcare Life Sciences, Piscataway, NJ), which encodes an N-terminal GST fusion tag that is cleavable by a human rhinovirus 3C proteinase recognition site (Leu-Glu-Val-Leu-Phe-Gln-Gly-Pro, referred to as the PreScission® site). An N-terminal mCherry-tagged human wt-αSYN cDNA was introduced into the KpnI and XhoI sites of the pcDNA3.1+ expression vector (Life Technologies/Invitrogen, Carlsbad, CA). The pEGFP-C1 eukaryotic expression plasmids (Clontech, Mountain View, CA) encoding the EGFP-tagged human wt and DN mutant K44A dynamin 1 were kindly provided by Dr. Hiroshi Miyoshi at the St. Marianna University School of Medicine in Kawasaki, Japan. Plasmid DNA used for transfection was prepared with the Genopure Plasmid Maxi Kit (Roche, Basel, Switzerland). The fidelity and orientation of the expression constructs were confirmed by restriction digestion and nucleotide sequencing analyses.
Recombinant α-synuclein purification
The GST-αSYN fusion construct (pGEX6P-1/αSYN) was transformed into the BL21(DE3)pLysS E. coli strain for protein expression. The transformed bacteria were grown in LB medium containing 100 μg/ml ampicillin and 35 μg/ml chloramphenicol (for pLysS) at 37°C until reaching an A600 of 0.4. The bacteria were then cultured for an additional 5 hours following induction with 0.5 mM IPTG. The bacteria were harvested by centrifugation, resuspended in ice-cold PBS (pH 7.4), and disrupted by ultrasonication (Smurt NR-50, Microtec, Chiba, Japan). After removal of the cell debris, the supernatant was loaded onto a glutathione-Sepharose 4B column (GE Healthcare) equilibrated with PBS. The GST-αSYN fusion protein was washed with PBS three times and was then eluted with 10 mM glutathione elution buffer. The final eluate that flowed through the column was collected as the control specimen and was further dialyzed against PBS overnight. Next, the GST tag was cleaved overnight at 4°C on a carousel in the presence of the PreScission® protease (2 units for 100 μg fusion protein, GE Healthcare). After cleavage, the sample was re-loaded onto the glutathione-Sepharose 4B column, and the flowthrough containing the tag-free αSYN was collected. The purity and the identity of the recombinant αSYN were verified by CBB staining and western blot analysis. The recombinant βSYN and γSYN, as well as A30P and A53T mutants of αSYN, were purchased from ATGen Co., Ltd. (Gyeonggi-do, Korea).
Cell culture and plasmid transfection
Dopaminergic neuronal cells (human SH-SY5Y and rat PC12) and human cells of oligodendroglia-lineage (KG1C and MO3.13) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM; Life Technologies/GIBCO, Carlsbad, CA) containing 4.5 g/l glucose, 2 mM L-glutamine (Life Technologies) and 10% FBS (PAA Laboratories, Pasching, Austria) at a temperature of 37°C under conditions of humidified 5% CO2/air. The SH-SY5Y cells were purchased from American Type Culture Collection (ATCC, Vienna, VA). The PC12 cells were kindly given to us by Dr. Katsutoshi Furukawa, Department of Geriatrics and Gerontology, Tohoku University, Sendai, Japan. The KG1C and MO3.13 cells were purchased from RIKEN Cell Bank (Tsukuba, Japan) and Cosmo Bio (Tokyo, Japan), Respectively. Plasmid DNAs (2 μg DNA for 1.5X106 cells) were introduced into the SH-SY5Y cells using the 4D-Nucleofector™ device with the CA-137 program (LONZA AG, Cologne, Germany). The cells were harvested 48 hours post-transfection unless otherwise stated. For the suppression of dynamin GTPases, cells were treated with 5 μM αSYN for 30 min with varying concentration of sertraline as indicated. Likewise, the SH-SY5Y cells were treated with 5 μM αSYN (for 30 min) 48 hours after transfection with the DN-dynamin1 or the siRNA silencing of dynamin 1. Stable transfection of mcherry-αSYN or EGFP in the cultured cells was achieved using the FuGENE HD® Transfection Reagent (Roche) according to the manufacturer’s instructions. For the stable transfection of mcherry-αSYN or EGFP, the transfected cells were maintained under selective pressure with 300 μg/ml G418 sulfate (InvivoGen, San Diego, CA).
Primary cortical neuron culture
Primary cultures of rat cortical neurons were prepared according to a previous method, with slight modifications [
82]. The dissociated cortical neurons were plated at a density of 0.5 × 10
6 cells on a poly-D-lysine-coated disc (Sumilon Celldesk LF, Sumitomo Bakelite Co., Ltd., Tokyo, Japan) in a 24-well plate and cultured in Neurobasal A (Life Technologies/GIBCO) medium supplemented with 2% B27 (Life Technologies/GIBCO), 25 mM glutamate, 18 mM glucose and 0.5 mM L-glutamine. Half of the culture medium was replaced with fresh medium excluding glutamate every 3 days. Six days after the initiation of the culture, cells were exposed to 5 μM recombinant αSYN for 24 hours and were then subjected to immunocytochemical analysis.
siRNA knockdown of endogenous dynamin in SH-SY5Y cells
To suppress endogenous dynamin 1 expression in cultured neuronal cells, siRNA specifically targeting human dynamin 1 (sc-43737) was used; this specific siRNA and a scrambled control siRNA (sc-36869) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). SH-SY5Y cells in log-phase growth were nucleofected with target-specific or control-scrambled siRNAs (300 nM for 2 × 106 cells) using the 4D-Nucleofector device with the CA-137 program and SF solution (LONZA AG). Seventy-two hours after gene silencing, the cells were harvested and subjected to further studies.
Co-culture experiments
For the mixed culture of αSYN donor and acceptor cells, SH-SY5Y neuronal cells overexpressing mCherry or mCherry-tagged αSYN (2 × 105 donor cells in a 3.5-cm dish) were co-cultured with neuronal PC12 or oligodendroglial MO3.13 cells stably expressing EGFP (2 × 105 acceptor cells in a 3.5-cm dish) in culture medium with or without 10 μM sertraline for 5 days. Incorporated mCherry-αSYN in the acceptor cells was evaluated using a FluoView™ FV300 confocal microscope system equipped with HeNe-Green (543 nm) and Ar (488 nm) laser units (Olympus, Tokyo, Japan). To quantify the mCherry-positive cytoplasmic inclusions in the acceptor cells, the total number of cells containing aggregates was counted per 250–300 cells from each of five randomly selected fields. From this, the percentages of cells with red-fluorescent inclusions were calculated. Data pooled from four independent experiments were statistically analyzed with the Student's t-test. The data were presented as the mean ± standard errors. For the double labeling experiments, the images were collected using a single excitation for each wavelength individually and were then merged using Fluoview image analyzing software (version 4.1, Olympus).
Cell fractionation and western immunoblot analysis
Mechanically harvested cells were washed twice with ice-cold PBS. Next, the hydrophobic membrane fraction and hydrophilic fraction were prepared using the Mem-PER® Eukaryotic Membrane Protein Extraction Reagent kit (Thermo Scientific, Waltham, MA) according to the manufacturer’s instructions. In some experiments, the cells were pretreated with bafilomycin A1 (0–5 nM for 24 hours) or sertraline (0–10 μM for 30 min) with or without exposure to αSYN. Successful separation of the hydrophilic fraction from the membrane fraction was verified by immunoblot analyses using Abs against the cytosolic Hsp90 and the plasma membrane localized protein Na
+/K
+ ATPase α, respectively. In some experiments, cells were directly solubilized in radio-immunoprecipitation assay (RIPA) buffer (1% NP-40, 0.5% deoxycholate, 0.1% SDS, 1 mM EDTA, 10 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM sodium orthovanadate, 150 mM sodium chloride, 50 mM Tris–HCl (pH 8.0) plus 1x Cømplete® protease inhibitor cocktail; Roche). Samples containing 50 μg total protein were electrophoresed on SDS-polyacrylamide gels (12.5%) and transferred onto polyvinylidene difluoride (PVDF) membranes (Immobilon-P; Merck Millipore, Billerica, MA). Native-PAGE was performed on 10-20% polyacrylamide gradient gels, according to the standard protocol. In some experiments, the separated proteins were visualized by staining with CBB R-250 (MP Biomedicals, Solon, OH). After blocking with 5% (w/v) nonfat dry milk (Wako Pure Chemical Industries, Ltd., Osaka, Japan) in Tris-buffered saline with 0.1% Tween 20 (TBST), the membranes were incubated with the following Abs: anti-synuclein-1 (63320, 1:1000; BD Bioscience, San Jose, CA), anti-αSYN (#2628, 1:1000; CST, Danvers, MA), anti-GST (#2625, 1:1000: CST), anti-LC3 (PM036, 1:1000; MBL, Nagoya, Japan), anti-caspase 3 (H-277, 1:2000; Santa Cruz), anti-cleaved caspase 3 (Asp 175, 1:1000; CST), anti-dynamin 1 (3G4B6, 1:1000; CST), anti-dynamin 2 (610263, 1:4000; BD Transduction Lab, Franklin Lakes, NJ), anti-dopa decarboxylase (DDC) (AB1569, 1:1000; Millipore), anti-TPPP/p25α (EPR3315, 1:1000; Epitomics, Burlingame, CA), anti-GFP (M048-3, 1:2000; MBL) anti-Na
+/K
+ ATPase α (D154-3, 1:20000; MBL), anti-Hsp90 (AC88, 1:2000; Stressgen, Farmingdale, NY), anti-α-tubulin (clone DM1, 1:1000; Sigma) and anti-mCherry (5993, 1:1000; BioVision, Mountain View, CA). The primary antibody labeling was followed by the addition of HRP-conjugated secondary Abs (1:10000; Jackson ImmunoResearch, West Grove, PA). The bands were visualized with the Luminata™ Forte Western HRP substrate (Millipore), and the images were captured using the LAS-3000 mini image analyzer (Fujifilm, Tokyo, Japan). Immunoblotting was performed at least three times. In some experiments, the blots were scanned and densitometric analyses were performed using Image J (
http://www.rsb.info.nih.gov/ij/) [
83]. The intensity unit of the αSYN monomer was divided by that of Hsp90. Data are expressed as the mean ± standard errors.
Immunofluorescence confocal microscopy
Cells were fixed in 4% (w/v) paraformaldehyde in PBS for 30 min and then permeabilized with 0.5% Triton X-100 in PBS for 5 min. After blocking with 3% normal goat serum (Wako Pure Chemical Industries) in PBS for 30 min, the cells were incubated with the following primary antibodies: anti-synuclein-1 (1:1000; BD Bioscience), anti-αSYN (#2628, 1:2000; CST), anti-Serine 129 phospho-αSYN (EP1526Y, 1:1000; Epitomics), anti-γ-tubulin (GTU-88, 1:4000; Sigma, St. Louis, MO), anti-peripherin (AB1530, 1:1000; Millipore-Chemicon), anti-vimentin (V9, 1:500; Sigma), anti-ubiquitin (P4D1, 1:1000; Santa Cruz), anti-MAP2 (#4542, 1:1000; CST), anti-TPPP/p25 (EPR-3315, 1:1000; Epitomics), anti-Rab5A Ab (S-19, 1:1000; Santa Cruz) and anti-Lamp-1 (H4A3, 1:1000; DSHB, University of Iowa, USA). For thioflavin S staining, the coverslips were immersed in 0.03% (w/v) thioflavin S (Sigma) solution for 5 min and were then extensively washed with 70% ethanol. Positive immunostaining was detected after a 1-hour incubation with Alexa 488- and Alexa 568-conjugated secondary Abs (1:4000; Life Technologies/Molecular Probes, Carlsbad, CA). Nuclei were counterstained with TO-PRO3 iodide (1:1000; Molecular Probes) and were pseudocolored in blue. The fluorescent images were analyzed with the FluoView FV300 confocal laser scanning microscope system (Olympus). The crossed lines indicate the positions of the xz- and yz-planes. To quantify the αSYN-immunopositive inclusions, the total number of cells with any aggregates was counted across 250–300 cells in five randomly chosen fields. From this, the percentages of cells with inclusions were calculated. Pooled data from four independent experiments were statistically analyzed using the Student's t-test. The data are expressed as the mean ± standard errors.
Acknowledgements
This work was supported in part by a Grant-in-Aid for Scientific Research (C) (23591228), a Grant-in-Aid for Scientific Research (B) (24390219), a Grant-in-Aid for Exploratory Research (24659423) from The Ministry of Education, Culture, Sports, Science and Technology (MEXT), a Grant from the Research Committee for Ataxic Diseases, a Grant-in-Aid for Scientific Research on Innovative Areas (Brain Environment; 24111502) from the Ministry of Health, Labor, and Welfare, Japan, and a Grant from the Symposium on Catecholamine and Neurological Disorders, Japan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MK, TH, TB, and EM performed the experiments. NS, AK, and AT analyzed the data. FCF, TS, MA, and YI contributed reagents/materials. MK and TH designed the study and wrote the paper. AT is the principal investigator. All authors read and approved the final manuscript.