Different ataxin-3 amyloid aggregates induce intracellular Ca2 + deregulation by different mechanisms in cerebellar granule cells

Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta journal homepage: www.elsevier.com/locate/bbamcr

Different ataxin-3 amyloid aggregates induce intracellular Ca2 + deregulation by different mechanisms in cerebellar granule cells Francesca Pellistri a, Monica Bucciantini b,e,⁎, Gaetano Invernizzi c, Elena Gatta a, Amanda Penco a, Anna Maria Frana c, Daniele Nosi d, Annalisa Relini a,e, Maria Elena Regonesi c, Alessandra Gliozzi a,e, Paolo Tortora c, Mauro Robello a,e, Massimo Stefani b,e a

Department of Physics, University of Genoa, Genoa, Italy Department of Biomedical, Experimental and Clinical Sciences, University of Florence, Florence, Italy c Department of Biotechnologies and Biosciences, University of Milano-Bicocca, Milan, Italy d Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy e Research Centre on the Molecular Basis of Neurodegeneration, Florence, Italy b

a r t i c l e

i n f o

Article history: Received 15 April 2013 Received in revised form 27 August 2013 Accepted 28 August 2013 Available online 11 September 2013 Keywords: Ataxin-3 Amyloid aggregate Intracellular Ca2 + level Oligomer toxicity Fibril toxicity

a b s t r a c t This work aims at elucidating the relation between morphological and physicochemical properties of different ataxin-3 (ATX3) aggregates and their cytotoxicity. We investigated a non-pathological ATX3 form (ATX3Q24), a pathological expanded form (ATX3Q55), and an ATX3 variant truncated at residue 291 lacking the polyQ expansion (ATX3/291Δ). Solubility, morphology and hydrophobic exposure of oligomeric aggregates were characterized. Then we monitored the changes in the intracellular Ca2+ levels and the abnormal Ca2+ signaling resulting from aggregate interaction with cultured rat cerebellar granule cells. ATX3Q55, ATX3/291Δ and, to a lesser extent, ATX3Q24 oligomers displayed similar morphological and physicochemical features and induced qualitatively comparable time-dependent intracellular Ca2+ responses. However, only the pre-fibrillar aggregates of expanded ATX3 (the only variant which forms bundles of mature fibrils) triggered a characteristic Ca2+ response at a later stage that correlated with a larger hydrophobic exposure relative to the two other variants. Cell interaction with early oligomers involved glutamatergic receptors, voltage-gated channels and monosialotetrahexosylganglioside (GM1)-rich membrane domains, whereas cell interaction with more aged ATX3Q55 pre-fibrillar aggregates resulted in membrane disassembly by a mechanism involving only GM1-rich areas. Exposure to ATX3Q55 and ATX3/291Δ aggregates resulted in cell apoptosis, while ATX3Q24 was substantially innocuous. Our findings provide insight into the mechanisms of ATX3 aggregation, aggregate cytotoxicity and calcium level modifications in exposed cerebellar cells. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Several human neurodegenerative diseases, including spinocerebellar ataxia type 3 (SCA3), Huntington's disease and spinobulbar muscular atrophy (SBMA), are associated with the genetic expansion over a defined threshold of CAG triplet repeats in the encoding genes [1–3].

Abbreviations: SCA3, spinocerebellar ataxia type 3; SBMA, spinobulbar muscular atrophy; JD, N-terminal Josephin domain; CGNs, rat cerebellar granule neurons; ThT, Thioflavin T; ANS, 8-anilino-1-naphthalenesulfonic acid; AMPA-R, α-amino-3-hydroxy-5methyl-4-isoxazolepropionic acid receptor; NMDA-R, N-methyl-D-aspartate receptor; VGCCs, voltage-gated calcium channels; CNQX, cyano-nitroquinoxaline-dione; APV, 2-amino-5-phosphonopentanoic acid; GM1, monosialotetrahexosylganglioside; BSA, bovine serum albumin ⁎ Corresponding author at: Department of Biomedical Experimental and Clinical Sciences, University of Florence, Viale Morgagni, 50, 50134 Florence, Italy. Tel.: +39 055 4598346; fax: +39 055 4598905. E-mail address: monica.bucciantini@unifi.it (M. Bucciantini). 0167-4889/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.bbamcr.2013.08.019

The presence of the resulting pathological polyQ expansion in the encoded proteins favors their polymerization into aggregates displaying amyloid-like morphological and biophysical properties, whose time of appearance seems directly related to the expansion length [1,4,5]. Ataxin-3 (ATX3), the causative agent of SCA3, is a 42-kDa, mainly cytosolic, protein containing an N-terminal Josephin domain (JD) (amino acid 1–182) and a polyQ stretch in the C-terminal region whose expanded pathological variants are responsible for the disease. Both domains appear involved in protein aggregation since it was previously shown that the JD has an intrinsic amyloidogenic potential and expanded ATX3 aggregation occurs through a two-step mechanism involving Josephin self-association, followed by a polyQ-dependent stage [6–8]. Similarly to amyloid aggregates of other peptides/proteins [9–14], early oligomers of polyQ aggregates are considered the primary species responsible for cell toxicity [15,16] which, at least in part, results from biochemical alterations in the affected neurons involving disruption of

3156

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Ca2+ homeostasis and signaling [17]. The latter effect appears to result from the interaction of the earliest assemblies in the aggregation path with the cell membrane either directly (reviewed in [18]) and/or mediated by specific membrane proteins [19,20]. In particular, we have previously reported that amyloid-like oligomers of HypF-N, the N-terminal domain of the Escherichia coli hydrogenase maturation factor F, which is not involved in any amyloid disease, are toxic to rat brain neural cells similarly to Aβ peptides [21,22]. These oligomers interact with glutamatergic receptors of primary rat cerebellar granule cells inducing a detrimental Ca2+ influx from the medium [19] with a mechanism comparable to that observed by exposing rat hippocampal neurons to Aβ oligomers [20]. These similarities provide support to the hypothesis that amyloid interaction with the cell membrane bilayer and/or membrane receptors and the resulting permeabilization is a basic behavior of amyloid oligomers associated with a shared conformational motif in the aggregates [23,22]. A theme of growing interest in the amyloid science is the relation between amyloid structure and cytotoxicity. Recent findings on the biophysical and structural features of aggregates of the same peptide/protein or its variants grown at different conditions both in vitro and in tissue suggest a close relation between the environmental physicochemical properties where protein aggregation occurs and those of the resulting aggregates [24]; those data provide a basic explanation of amyloid aggregate polymorphisms and the resulting variable cytotoxicity [25–30]. The data on amyloid polymorphisms involve a growing number of peptides/proteins either associated or not-associated to amyloid diseases, including Abeta peptides, α-synuclein, prion proteins, huntingtin and HypF-N. Accordingly, amyloid polymorphism has become one of the leading themes in the fields of amyloid structure and cytotoxicity (reviewed in [31]). However, in spite of the increasing importance of the theme, the occurrence of structurally different amyloid aggregates of ATX3 and their structure–toxicity relationship has not been reported yet. To shed light on the mechanism(s) of ATX3 toxicity, in the present work we investigated several ATX3 variants: a non-pathological form (ATX3Q24), a pathological expanded form (ATX3Q55), and a derivative truncated at residue 291 lacking the polyQ expansion (ATX3/291Δ). All these forms aggregate into soluble amyloid assemblies which persist under aggregating conditions at 37 °C, but only ATX3Q55 assembles into bundles of elongated mature fibrils after 48 h of incubation under aggregation conditions [32]. We here provide data on ATX3 polymorphisms. In particular we investigated the relation between the morphological and physicochemical properties of aggregates grown from ATX3 with normal or pathological polyQ expansions, their different interactions with neuronal cell membranes and the mechanisms of the resulting intracellular Ca2+ increase, an important, yet poorly investigated aspect of polyQ aggregate toxicity.

2. Materials and methods 2.1. Production and purification of ATX3 variants Full length ATX3Q24 and ATX3Q55 were cloned, expressed and purified as previously described [32] with minor modifications. Briefly, His-tagged proteins were affinity-purified and then chromatographed in PBS buffer on a Superose 12 10/300GL column (GE Healthcare LifeSciences, England) at a flow rate of 0.5 ml/min. The truncated form ATX3/291Δ was cloned and purified as previously reported [33]. Protein-containing fractions were pooled and stored at −20 °C. The ATX3 variants were centrifuged at 15,000 ×g for 15 min at 4 °C immediately prior to use. Protein content was determined by Coomassie brilliant blue G-250 (Pierce Biotechnology, Rockford, IL) using BSA as a standard protein.

2.2. Cell preparation Cerebellar granule neurons (CGNs) were prepared from 8 dayold Sprague–Dawley rats (Harlan, Italy) as previously described [34] and used from the 6th to the 10th day of culture (after the 1st or the 2nd day of culture in the experiments on immature neurons). Experiments with animals were carried out according to international guidelines on the ethical use of animals and every effort was made in order to minimize the number of animals used and their suffering. 2.3. Fluorescence measurements Fluorescence experiments were performed and analyzed as described previously [19]. The KCl depolarizing solution (75 mM, a concentration which causes maximum [Ca2+]i rise by opening all types of voltagegated calcium channels VGCC) [35], was prepared by equimolar substitution of NaCl. Cell response to NMDA was measured in 100 mM NMDA and 50 mM glycine, in the absence of Mg2+. In some experiments, 50 mM CdCl2 or 1.0 mM MgCl2 were added to block VGCC or NMDA channels, respectively. In the Ca2+-free solution, 2.0 mM EGTA was present in place of CaCl2. 2.4. Labeling with Texas red Aggregate internalization was detected by incubating cultured cells for 24 h with protein aggregates labeled with Texas red, as reported previously [36]. 2.5. Apoptosis detection To detect apoptosis, CGNs were exposed for 1 h or 4 h to 0.8 μM (monomer protein concentration) ATX3 aggregates aged for 24 h or to vehicle. Then, the cells were fixed with cold (−20 °C) methanol and their nuclei were stained upon incubation for 30 min with 1.0 μg/ml bisbenzimide (Hoechst 33258, Mol. Probes, Invitrogen), washed three times and analyzed for condensed or disrupted nuclei by a fluorescence microscope. At least 250 cells/coverslip were analyzed and the experiments, performed in duplicate, were repeated at least twice. 2.6. Sialic acid cleavage from membrane GM1 CGNs were starved by incubation with DMEM serum-free media for 3 h at 37 °C, left for 2 h in the presence of a mixture of Vibrio cholerae (11,7 mU) and Clostridium perfringens (100 mU) neuraminidase (Sigma Aldrich), washed and incubated with Oregon Green [19]. 2.7. Confocal analysis CGNs were exposed for 10 min to variously aged 0.8 μM ATX3/291Δ, ATX3Q24 or ATX3Q55 aggregates and fixed as described previously [19]. Then the cells were double immunostained by incubation for 1 h at room temperature with a mouse monoclonal anti-AMPA-R (Chemicon) or NMDA-R (SantaCruz) Ab and a rabbit polyclonal anti-ATX3 Ab in blocking solution (PBS containing 0.5% BSA and 0.2% gelatin). The immunoreaction was revealed with Alexa 488 or Alexa 568-conjugated anti-rabbit or anti-mouse secondary Ab (Molecular Probes) diluted 1:400 in PBS. GM1 labeling was performed by incubating the cells with 10 ng/ml Alexa 488-conjugated CTX-B (Molecular Probes) in complete medium for 20 min at 37 °C. Cell fluorescence was visualized and analyzed as described above [19]. 2.8. Atomic force microscopy The ATX3 variants (25 µM protein concentration in PBS) were aggregated at 37 °C. ATX3 samples at fixed aggregation times were

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

imaged by tapping mode AFM as described previously [32]. Both 100-fold and 400-fold dilutions were used for samples diluted before deposition on mica, Drive frequencies in the 240–300 kHz range and scan rates in the 0.6–2.0 Hz range were used. Aggregate height was measured from the height in cross-section in the topographic AFM images as described previously [24,32].

2.9. Physicochemical characterization of ATX3 aggregates The Thioflavin T (ThT) assay was used to monitor the progress of protein aggregation as previously reported [32] by incubating 25 μM ATX3 variants in PBS with 20 μM ThT at 37 °C. Changes in hydrophobic exposure during protein aggregation were detected by measuring 8-anilino-1-naphthalenesulfonic acid (ANS) extrinsic fluorescence, which rises on interaction with exposed hydrophobic patches [37,38]. The ATX3 variants (25 μM protein concentration) were incubated in PBS buffer at 37 °C and mixed at the indicated times with 2.0 μM ANS. The soluble species were obtained by sample centrifugation at 20,000 ×g for 15 min. In all cases, the protein was diluted in the presence of ANS to a final 2.0 μM concentration. Sample fluorescence was recorded in a Cary Eclipse instrument (Varian, Palo Alto, CA) at 374 nm excitation and 400–600 nm emission and expressed as the peak intensity in 1-cm path length quartz cuvettes. All data were collected at room temperature.

3. Results 3.1. Morphological characterization of ATX3 aggregation intermediates Our aim was to better elucidate the mechanisms of toxicity of ATX3 aggregates grown at different aggregation times. Therefore, we first carried out preliminary in vitro studies to ensure that we could reproduce previous observations [6–8,32] on ATX3 variants with polyQ expansions of varying lengths (depicted in Fig. S1) and investigated the behavior of the truncated form, ATX3/291Δ. In particular, the aggregation kinetics at 37 °C of the ATX3 variants over a time span of 70 h assessed by ThT (Fig. S2a), yielded profiles in good agreement with previous reports [6–8,32]. Next, different ATX3 variants incubated at 37 °C were imaged by tapping mode AFM as a function of the aggregation time (Fig. 1). At early stages of ATX3Q55 aggregation, isolated globular particles, most likely protein oligomers, were seen; after 6 h of aggregation, the AFM analysis showed isolated oligomers coexisting with partially coalesced oligomers (Fig. S3a). A heterogeneous morphology was observed in the ATX3Q55 sample aged for 14 h, where isolated and coalesced oligomers as well as large clusters of aggregates were seen (Fig. S3b,c). The first bundles of fibrils appeared after 24 h and became very abundant after 48 h (Figs. 1 and S3d). The inset in Fig. S3d shows the insoluble fibrillar fraction recovered after sample treatment with SDS and centrifugation. ATX3Q24 and ATX3/291Δ behaved differently: within our experimental aggregation times they grew into globular oligomers and very short beaded protofibrils (arrows in Fig. 1) without forming fibrils. These results match well with previous observations acquired by both TEM and AFM [39,8]. Fig. S2b compares the heights of the oligomers grown from the different ATX3 variants as a function of the aggregation time. The data reported for each variant are restricted to the time interval in which oligomeric structures were the prevailing aggregated species. On the whole, our results confirm that: i) although all the variants assayed undergo aggregation, the rate increases with the size of the protein and ii) only the expanded form gives rise to SDS-insoluble fibrils that pack into large bundles, whereas the two other forms do not exceed the globular oligomer stage.

3157

3.2. ATX3 aggregates increase intracellular Ca2+ in rat CGNs We then looked at the effects in vivo of all ATX3 variants at different aggregation times. Since many amyloid oligomers affect Ca2+ homeostasis and signaling in exposed cells, we monitored intracellular Ca2+ levels of CGNs exposed to ATX3Q55 oligomers aggregated for 4 h by recording the fluorescence intensity of cells loaded with Oregon Green. Fig. 2a reports a typical time course of the average fluorescence intensity of exposed cells. The lower part of Fig. 2a shows the fluorescence images of CGNs at the times indicated in the upper panel. The fluorescence response indicates a sudden rise of the intracellular free Ca2+ followed by a slower drop. The mean duration of the fluorescence signal was 61 ± 8 s, while the maximal value was reached after about 20 s from the onset of the rise (Fig. 2a), a behavior we have indicated as type 1 response. A very similar time course of Ca2+ rise was found in cells exposed to ATX3Q55 aggregates grown at 37 °C for 2, 4 or 6 h or to the protein at the earliest times of incubation. The amplitude of the type 1 response and the number of cells where it was detected were found to depend on the age of protein aggregates, increasing with the age of the aggregates. This finding is particularly intriguing and will be further discussed below. Aggregates aged for longer times (14, 24 or 48 h) had heterogeneous morphologies and induced a more complex fluorescence response in the exposed CGNs. At these conditions, in addition to type 1 responses, we observed a much shorter spike-like response (type 2 response) with a mean time of 13 ± 1 s (Fig. 2a, inset). The occurrence of either response appeared to depend on the morphology of differently aged aggregates. To check whether the type 2 fluorescence response was related to the presence of fibrillar or pre-fibrillar assemblies like those shown in Fig. 1, the soluble and insoluble fraction of 24 h-aged ATX3Q55 aggregates were separated by centrifugation. The soluble fraction still elicited both type 1 and type 2 responses, whereas the insoluble fraction, mainly composed of fibrillar aggregates, was completely inactive, suggesting that (i) the fibrillar fraction is unable to evoke either response in the exposed cells and (ii) a significant correlation does exist between aggregate type (oligomeric or pre-fibrillar) and the type and amplitude of the fluorescence response. Finally, in a Ca2+-free medium no fluorescence increase was recorded in cells exposed to ATX3Q55 aggregates, suggesting that the increased Ca2+ did not came from the intracellular stores but, rather, from the extracellular medium. Next, we investigated the fluorescence response of CGNs exposed to ATX3Q24 or to ATX3/291Δ aggregated at 37 °C for different times. Only a type 1-like response was observed, very similar, though not identical, to that seen in cells exposed to ATX3Q55 oligomers (Fig. S4). In cells exposed to ATX3/291Δ aggregates, the fluorescence signal reached a maximum in about 10–15 s, with a mean duration of 70 ± 4 s; ATX3Q24 displayed a maximum at a comparable time, while the mean duration was 58 ± 2 s. Even in these cases, we did not record any fluorescence response in cells cultured in Ca2+-free medium. Moreover, unlike ATX3Q55-exposed cells, the cells exposed to either ATX3Q24 or to ATX3/291Δ aggregates displayed a fluorescence signal almost independent of the aggregation time, in agreement with the substantial invariance of their morphology with the aggregation time. This behavior can be explained by considering that at the earliest aggregation stages all ATX3 variants display similar intermolecular interactions sustained by main chain hydrogen bonding, whereas only the polyQexpanded variant evolves into mature fibrils, whose hallmark is hydrogen bonding among glutamine side chains [32,33]. After having determined the different behavior of different ATX3 aggregates in terms of Ca2+ dishomeostasis in exposed cells, we sought to correlate structural features of ATX3 aggregates to the physiological responses detected. To this end, we monitored the progress of protein aggregation by recording the changes in ANS fluorescence, which is related to protein hydrophobic exposure (for typical emission spectra, see Fig. S5), at different aggregation times. At 24 h of aggregation, the fluorescence of the soluble fraction was also recorded (Table 1). During

3158

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Fig. 1. Tapping mode AFM images (height data) of the aggregates formed by the investigated ATX3 variants. Columns: representative images of aggregates imaged at t = 0; 6 h; 24 h; 48 h. Rows from top to bottom: ATX3Q55; ATX3Q24, ATX3/291Δ. Scan size 1.0 μm (left columns) and 2.2 μm (right columns). Z range (from left to right): ATX3Q55, 7 nm, 10 nm, 200 nm, 250 nm; ATX3Q24, 12 nm, 10 nm, 20 nm, 55 nm; ATX3/291Δ, 10 nm, 30 nm, 6 nm, 3 nm.

the incubation, the fraction of soluble protein decreased, with no major differences among the three variants (Fig. S6). In particular, after 24 h, the soluble fraction was about 65% of the initial value for ATX3Q55 and 80% for the two other forms. We found a moderate time-dependent fluorescence increase in the ATX3Q55 total fraction, while this increase was negligible in the case of ATX3Q24 and ATX3/291Δ. After 24 h of aggregation the soluble fraction of ATX3Q55 displayed a substantially higher ANS fluorescence relative to the normal and the truncated form. Thus, the increased exposure of hydrophobic patches of ATX3Q55 may result from a different structural arrangement of the monomers and/or their different mode of assembly into aggregates. In any case, at later stages only ATX3Q55 can generate highly hydrophobic soluble assemblies likely responsible for a type 2 response. Taken together, these results suggest that: (i) mature ATX3 fibrils are unable to foster any calcium influx in the exposed cells; (ii) type 1 responses occur in the presence of soluble oligomers whereas (iii) a type 2 response occurs in the presence of soluble pre-fibrillar assemblies representing a later aggregation stage. In line with this interpretation, ATX3Q24 and ATX3/291Δ aggregates whose structural features are apparently similar to those displayed by ATX3Q55 at early aggregation times, caused a nearly constant type 1 fluorescence response (see the following section). 3.3. Ca2+ increase in exposed cells depends on aging of ATX3Q55 oligomers As reported above (Fig. 2a), we found that the fluorescence response of cells exposed to variously aged ATX3Q55 aggregates displayed a different time dependence. However, this was only one of the factors that characterize the intracellular Ca2+ increase in these cells. Two additional parameters were important to fully describe the cell response: (i) the mean value of the relative fluorescence increase,

ΔF/F and (ii) the percentage of exposed cells displaying increased mean fluorescence. Both parameters revealed that different aggregates of the various protein variants were differently effective in inducing free Ca2 + increase in the exposed cells. To describe better the overall response of the aggregate-cell system in terms of intracellular Ca2 + level modification, we introduce the W(F) parameter defined as the product of ΔF/F and the percentage of cells displaying a fluorescence signal. W(F), which is also a function of the aggregation time can be a reliable indicator of the efficacy of the aggregates in inducing intracellular Ca2 + increases. Fig. 2b shows the W(F) value as a function of the aggregation time t of the three ATX3 variants. Profiles I) and II) correspond to the type 1 response stimulated by ATX3/291Δ and ATX3Q24 aggregates, respectively. In both cases, fibrillar structures are absent even after 48 h of aggregation and the W(F) value remains substantially unchanged. Although the experimental conditions may affect the absolute W(F) values, the fact that they are relatively constant clearly confirms that the physiological effect of the two variants is not affected by the aggregation time. In contrast to curves I) and II), curve III), corresponding to a type 1 response in cells exposed to ATX3Q55 aggregates, displayed a remarkable time-dependent increase of the W(F) value, which peaked around 6 h. However, at longer aggregation times, when pre-fibrillar and mature fibrillar structures are increasingly present, W(F) decreased along with the oligomer concentration, approaching the values displayed by ATX3Q24 aggregates. In fact, as the aggregation proceeded, pre-fibrillar structures, which induce a type 2 response, became increasingly more represented. Curve IV) shows the trend of the W(F) value for a type 2 response. W(F) was zero when the cells interacted with oligomers (up to 6 h of aggregation) and increased at longer aggregation times, when pre-fibrillar structures were increasingly enriched. The data from these experiments

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

are summarized in Table 1S. The staggered W(F) peaks in type 1 and type 2 responses only in cells exposed to ATX3Q55 aggregates clearly indicate that either response is evoked by different aggregates while the other two variants showed no change in response with the aggregation time.

3.4. ATX3 aggregates interact with the plasma membrane of rat CGNs at specific sites Next, we investigated the mechanisms of plasma membrane permeabilization to Ca2+ underlying the different responses triggered by cell interaction with ATX3 pre-fibrillar aggregates. To this end, we checked whether ATX3Q55, ATX3Q24 and ATX3/291Δ aggregates interacted with specific membrane regions, particularly those enriched in

Fig. 2. (a) Upper panel, time course of the relative fluorescence intensity ΔF/F of cells exposed to 4 h-aged ATX3Q55, showing a typical type 1 response (see text). Inset, typical type 2 response (see text). Lower panel, fluorescence images of granules loaded with Oregon Green. Times corresponding to the capture of images 1–3 are indicated on the fluorescence curve in the upper panel. Scale bar 10 μm. (b) The W(F) parameter is plotted as a function of the aggregation time, for cells exposed to ATX3/291Δ (I), ATX3Q24 (II) or ATX3Q55 aggregates (III, IV). (I, II, III) Type 1 responses in cells exposed to ATX3 oligomers, (IV) type 2 responses in cells exposed to ATX3Q55 pre-fibrillar and fibrillar assemblies. Errors were calculated according to Student's statistics assuming a confidence level of 95%.

3159

Table 1 ANS fluorescence of ATX3 variants during aggregation. The proteins were incubated in PBS at 37 °C and at the indicated times samples were withdrawn and mixed with ANS (2.0 μM final protein concentration). Soluble species are those found in supernatants of the samples after centrifugation. At zero time, only soluble protein was present. The values are the mean of at least three independent experiments ± standard deviations. See also under Section 2.9 of Materials and methods. Time (h)

ATX3Q24 Total

0 8 16 24

97 97 100 101

± ± ± ±

4 2 3 3

ATX3/291Δ Soluble

Total

103 ± 6

93 95 98 100

± ± ± ±

1 1 4 6

ATX3Q55 Soluble

Total

104 ± 4

97 97 99 105

± ± ± ±

Soluble 2 4 5 5

123 ± 7

glutamatergic and voltage-gated (VGCCs) Ca2 + channels. The cells were exposed to different ATX3 aggregates for 10 min, fixed, permeabilized and checked for aggregate co-localization with NMDA-Rs, GluR2/AMPA-Rs and VGCCs by double label confocal microscopy. Figs. 3 to 5 show the membrane distributions of the AMPA-Rs, NMDA-Rs and VGCCs (red channel) and the localization of the different types of ATX3 aggregates (green channel), respectively. We did not detect a massive internalization of the aggregates except for ATX3/291Δ oligomers; in most cases we noticed that these assemblies were embedded within the bilayer and only sporadically did some of them appear bound to the inner leaflet of the plasma membrane. The Pearson correlation coefficient, the overlap coefficient and the Manders' coefficients of channels A (green) and B (red) were used to measure the degree of co-localization as reported in Figs. 3, 4 and 5. The analysis suggests that an interaction between AMPA-Rs and 6 h-aged ATX3Q55 oligomers or ATX3/291Δ oligomers did occur (Fig. 3) and that such co-localization was strongly reduced upon cell pre-incubation with CNQX (an AMPA-R inhibitor). The co-localization between 24 h- and 48 h-aged ATX3Q55 aggregates and AMPA-Rs was only partial (Fig. 3) and quite modest with NMDA-Rs (Fig. 4); the latter disappeared almost completely in cells pre-treated with APV (an NMDA-R antagonist). Interestingly, we did not record any major change of the AMPA-R fluorescence signal or of the redistribution of the receptor signal at the surface of cells exposed to ATX3Q55 aggregates aged 6 h in spite of their remarkable co-localization. These findings suggest that these oligomers do not affect receptor trafficking by endocytosis, as reported in cells exposed to other pre-fibrillar assemblies [40–42], leaving substantially unchanged receptor abundance on the cell membrane. However, we cannot exclude that the short time of cell exposure to the aggregates before measurements (10 min) was not sufficient to cause any detectable alteration of receptor redistribution at the cell membrane. Cell treatment with 24 h-aged ATX3Q24 aggregates (Figs. 3 and 4) resulted in a substantial lack of association with AMPA and NMDA receptors. The Pearson's correlation and Manders' coefficient showed remarkable co-localization of ATX3Q55 and ATX3Q24 oligomers aged 6 h with VGCCs (Fig. 5), in agreement with the finding that VGCC inhibition reduced by 50%, the Ca2+ responses (see below). However, when the cells were exposed to ATX3Q55 aggregates aged for 24 h or 48 h, where heterogeneous species such as pre-fibrillar aggregates and fibrils were present, we found a mild interaction of these assemblies with the cell membrane but no co-localization with NMDA-Rs (Fig. 4). Finally, to investigate whether these aggregates had gained access to the cytoplasmic compartment, the cells were exposed for 10 min to the same aggregates conjugated to Texas red. The data reported in Fig. S7, while showing that these aggregates contact the cell membrane, exclude any aggregate internalization within the investigated time. Overall, the main results of this part can be summarized as follows: i) the ATX3Q55 aggregate–AMPA receptor interaction was higher at short aggregation times and disappeared in cells pretreated with an AMPA-R antagonist, while only a mild interaction occurred with NMDA-

3160

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

R, ii) VGCCs interacted with ATX3Q55 and ATX3Q24 at short aggregation times, and iii) no massive internalization of the oligomers occurred (with the exception of ATX3/291Δ) and neither were receptors internalized.

3.5. Type 1 response to ATX3Q55 aggregates involves the activation of the glutamatergic AMPA-Rs Once assessed that our ATX3 aggregates were able to interact with Ca2+ channels in the exposed cells, we checked whether the glutamatergic AMPA-Rs were really involved in either type 1 or type 2 response, as suggested by the co-localization data. Therefore, we repeated the fluorescence experiments with cells exposed to ATX3Q55 aggregates in the presence of 10 μM CNQX, which hinders membrane depolarization thus maintaining the NMDA-R block by Mg2+. We first analyzed the data related to the type 1 response. When the cells were exposed to 6 h- or 24 h-aged ATX3Q55 aggregates in the presence of CNQX, the W(F) value decreased by about 70 ± 10%; while W(F) decreased by about 75% and 30% with 6 h- and 24 h-aged aggregates, respectively, in the presence of APV, in agreement with the colocalization data. Cell exposure to ATX3/291Δ or ATX3Q24 aggregates in the presence of APV or CNQX yielded effects qualitatively similar to the type 1 response, even though no evident differences between the CNQX and APV effect were observed, suggesting a lower specificity of the aggregate–receptor interaction (Fig. 6a). These results did not exclude the possibility that the intracellular Ca2 + increase in aggregate-exposed cells could result, at least in part, from activation of voltage-dependent Ca2 + channels. Therefore, we performed fluorescence experiments, by adding to the cell culture medium 50 μM Cd2 +, a VGCC blocker. Cell exposure to 6 h-aged ATX3Q55 aggregates in the presence of Cd2 + decreased the response by about 30%, whereas no significant changes of W(F) were detected when the cells were exposed to ATX3Q55 aged 24 h. A decrease by over 70% was recorded in cells exposed to ATX3Q24 aggregated either 6 or 24 h. In contrast with these results, the type 2 response did not significantly change in the presence of CNQX or APV (Fig. 6a), suggesting that this response, evoked by large soluble pre-fibrillar aggregates, is independent of the interaction between ATX3Q55 pre-fibrillar structures and glutamatergic receptors. Rather, it can tentatively be ascribed to some generic perturbation of the bilayer integrity with local disassembly producing discontinuities that hinder membrane selective permeability. This interpretation agrees with the finding that ATX3Q55 is the only variant capable of generating highly hydrophobic soluble aggregates. Our experiments with channel inhibitors show that: i) the VGCC partially participate in free Ca2+ rise in cells exposed to ATX3Q24 and early ATX3Q55 aggregates, ii) only oligomeric ATX3Q55 aggregates affect AMPA-R and, to a minor extent, NMDA-R functionality, that is impaired less specifically by ATX3/291Δ or ATX3Q24 oligomers, and iii) these data along with the ANS data reported in Table 1 lead to the conclusion that these oligomers, though morphologically similar, must have distinct structural and physicochemical features underlying their different behavior. 3.6. GM1 mediates the effects of ATX3Q55 and ATX3/291Δ aggregates in CGNs It is known that metabotropic glutamate receptors are raft-associated membrane proteins [43] and that the sialic acid moiety of GM1, a key raft component, is required for amyloid–membrane interaction [44,45]. Accordingly, to better elucidate mechanistically the cell membrane– Fig. 3. Confocal images of AMPA-Rs in cerebellar granule cells after exposure to ATX3Q55, ATX3Q24 or ATX3/291Δ. The cells were exposed for 10 min to 0.8 μM (monomeric protein concentration) ATX3Q55 aggregates after 6 h, 24 h or 48 h of incubation under aggregation conditions or to ATX3/291Δ or ATX3Q24 oligomers in the absence or in the presence of CNQX. The cells were fixed, permeabilized and stained by anti-ATX3 Ab (green) and anti AMPA-R Ab (red). Colocalization was estimated using specially developed algorithms that calculate a number of respective coefficients, such as Pearson's correlation coefficient, overlap coefficient according to Manders, and colocalization coefficients M1 and M2.

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

ATX3 aggregate interaction, we treated CGNs with a mixture of neuraminidases to specifically cut the sialic acid component of GM1. In neuraminidase-treated cells, the aggregate-driven Ca2+ influx was

3161

noticeably reduced and type 2 responses elicited by pre-fibrillar ATX3Q55 aggregates were completely abolished (Fig. 6a). This observation strongly supports the hypothesis that electrostatic interactions are the main forces driving aggregate adsorption to the cell surface at the negatively charged sialic acid residues carried by raft GM1 that results in some bilayer disassembly and non-specific permeabilization. The importance of both glutamatergic receptors and membrane gangliosides for membrane permeabilization by ATX3 aggregates was further investigated by experiments performed on immature (DIV1, DIV2) or mature (DIV8) cerebellar granule cells. Immature cells express a reduced density of glutamatergic receptors and gangliosides that increases with maturation. We found that immature cells displayed a much lower sensitivity to ATX3 aggregates, as shown in Fig. 6b, which compares the Ca2+ influx resulting from cell exposure to NMDA with that elicited by ATX3 aggregates in the three types of cells. Table 2S reports these results in detail together with the Ca2+ influx resulting from depolarized cells. As expected, the W(F) value in cells exposed to NMDA increased with cell maturation, suggesting a parallel increase of the number of glutamatergic receptors (Fig. 6b) and confirming that the latter are involved in type 1 responses. This conclusion agrees with the results obtained in the presence of APV and CNQX and with the co-localization data. Interestingly, mature CGNs, but not immature cells showed a type 2 response upon exposure to pre-fibrillar assemblies in 24 h aged ATX3Q55 aggregates. Since type 2 responses are likely due to the interaction of pre-fibrillar aggregates with GM1, our data confirm the absence (or reduced abundance) of this lipid, and possibly of raft domains, in immature CGNs, in agreement with data on immature hippocampal neurons [46]. In conclusion, the comparative study of ATX3Q55 oligomers and large pre-fibrillar aggregates shows a crucial difference in their interaction with the cell membrane. In fact, oligomer interactions can depend on several factors, including the presence of well structured, GM1 enriched lipid rafts and membrane protein components, including glutamatergic receptors, whereas large pre-fibrillar aggregates appear to interact only with ganglioside-rich membrane sites, possibly lipid rafts, without any interaction with raft proteins such as glutamatergic channels. 3.7. ATX3Q55 and ATX3/291Δ aggregates induce apoptosis in CGNs Finally, to explore whether the intracellular Ca2+ rise was associated with cytotoxicity, we checked by nuclear staining whether apoptosis was activated in CGNs either untreated or exposed to ATX3 aggregates for 1 h or 4 h. Cells were then fixed, stained with Hoechst 33258 and observed under fluorescence microscopy. Under these conditions, apoptosis is detected as nuclear condensation or fragmentation characterized by strong light blue fluorescence, whereas spherical/oval nuclei displaying moderate blue fluorescence indicate viable cells. We found that cell exposure to ATX3Q55 or ATX3/291Δ aggregates aged 24 h induced a significant increase of condensed and/or fragmented nuclei as compared to untreated cells or to cells exposed to ATX3Q24 aggregates (Fig. S8a, b). After exposure to 0.8 μM ATX3Q55 or ATX3/ 291Δ aggregates aged 24 h, the number of apoptotic cells (calculated by measuring the percentage of condensed/fragmented nuclei on total nuclei) was much higher than that observed in untreated control neurons or in neurons exposed for 24 h to ATX3Q24 aggregates (Fig. S8c) supporting severe or negligible cytotoxicity, respectively. Fig. 4. Confocal images of NMDA-Rs in cerebellar granule cells after exposure to ATX3Q55, ATX3Q24 or ATX3/291Δ aggregates. The cells were exposed for 10 min to 0.8 μM ATX3Q55 aggregates after 6 h, 24 h or 48 h of incubation under aggregation conditions or to ATX3/291Δ or ATX3Q24 oligomers in the absence or in the presence of APV. The cells were fixed, permeabilized and stained with anti-ATX3 Ab (green) and anti NMDAR Ab (red). Colocalization was estimated using specially developed algorithms that calculate a number of respective coefficients, such as Pearson's correlation coefficient, overlap coefficient according to Manders, and colocalization coefficients M1 and M2.

3162

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Fig. 5. Confocal images of VGCCs in cerebellar granule cells after exposure to ATX3Q55, ATX3Q24 or ATX3/291Δ aggregates. The cells were exposed for 10 min to 0.8 μM ATX3Q55 aggregates after 6 h, 24 h or 48 h of incubation under aggregation conditions or to ATX3/291Δ or ATX3Q24 oligomers in the absence or in the presence of Cd2+. The cells were fixed, permeabilized and stained with anti-ATX3 Ab (green) and anti VGCC Ab (red). Colocalization was estimated using specially developed algorithms that calculate a number of respective coefficients, such as Pearson's correlation coefficient, overlap coefficient according to Manders, and colocalization coefficients M1 and M2.

Finally, to check whether apoptosis was produced by fibrillar or prefibrillar ATX3 assemblies, we separated the soluble and insoluble fraction of 48-h aged aggregates by centrifugation. As expected, negligible damage was found in cells exposed for 1 h to the fibrillar fraction (5 ± 1% apoptotic cells) with respect to the vehicle-exposed control (1.5 ± 0.2% apoptotic cells) (data resulting from about 2000 cells examined). By contrast, cell exposure for the same time to the soluble fraction caused apoptosis of almost all cells. Altogether these results indicate that ATX3Q55 and ATX3/291Δ soluble aggregates produce cell damage and apoptosis.

4. Discussion Understanding the mechanisms by which amyloid aggregates induce permeabilization of membranes causing cytotoxicity [47–49] is of growing interest. However, no common consensus so far has been reached on the molecular details of membrane disruption [50,51]. Interestingly, amyloid-induced lipid membrane permeabilization is modulated by the lipid composition of the bilayer [52] and may significantly increase Ca2+ permeability by a generic mechanism involving local disassembly of the bilayer itself [23]. Moreover, other experimental

Fig. 6. (a, b) Fluorescence responses in cells exposed to 0.8 μM (monomeric protein concentration) ATX3Q55, ATX3Q24 or ATX3/291Δ aggregates aged 24 h. a) Percentage change of the W(F) parameter with respect to the control for type 1 and type 2 responses in the presence of CNQX, APV or neuraminidase. b) The W(F) parameter for immature (DIV1 or DIV2) or mature (DIV 8) granule cells exposed to ATX3 aggregates or to NMDA (for details see text).

findings indicate that amyloid aggregates, particularly oligomers, can also interact selectively with voltage-dependent Ca2+ channels or glutamatergic receptors [53–55,19]. These receptors are required for synaptic targeting of toxic Aβ oligomers [56], which results in intracellular Ca2+ rise and subsequent biochemical modifications eventually leading to dendritic dystrophy and neurodegeneration [57]. It has therefore been suggested that Aβ oligomers bind to synapses in close proximity to NMDA and AMPA receptors, whose deregulation disrupts Ca2+ homeostasis and triggers oxidative stress [20], providing a molecular basis for plasticity failure, synapse loss and memory dysfunction in AD [58–60]. Much less information is currently available on the aggregate– neurotoxicity relationship in polyQ diseases, which is quite complex and only partially understood [61,62]. The proposed mechanisms are based on both gain and loss of function, including some modification of membrane function and permeability. In the case of ATX3, no direct involvement of the plasma membrane in such phenomena has been reported; however the pathogenic expanded forms of ATX3 have been associated with different membrane-related effects, including the interruption of synaptic transmission in Caenorhabditis elegans [63], potassium channel dysfunction and depolarization of resting membrane potential in PC12 cells [64], and increased inactivation of voltage-activated potassium currents in Purkinje neurons in a mouse model [65]. An expanded ATX3 also directly affected Ca2 + signaling through association with type 1 IP3 intracellular calcium release channels, sensitizing activation by IP3 and disrupting calcium homeostasis in rat neuronal cultures [17].

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Here, we have investigated whether pathogenic and harmless variants of the cytosolic protein ATX3 induce abnormal intracellular Ca2 + increases in exposed neural cells and the underlying molecular mechanisms. In our experimental conditions, only the expanded ATX3Q55 form grew into large pre-fibrillar aggregates and mature fibrils, whereas the two other variants stopped at the stage of small oligomers apparently similar to the soluble aggregates appearing at the beginning of ATX3Q55 aggregation; however, ANS fluorescence measurements showed that at later aggregation stages (24 h incubation), the surface hydrophobicity of the latter was higher than that of the two other variants. ATX3Q55, ATX3Q24 and ATX3/291Δ aggregates trigger a transient increase in the intracellular Ca2+ level of the exposed cells, resulting from Ca2+ influx from the external medium. The results, summarized in Table 1S and Fig. 2b show that the value of the W(F) parameter was substantially independent of the aggregation time when assaying ATX3Q24 and ATX3/291Δ aggregates (Fig. 2b, curves I and II), which retained the oligomeric form over the investigated aggregation times. However, ATX3Q55 aggregates displayed a unique behavior eliciting two different time-dependent responses (Fig. 2) related to the aggregate morphological and physicochemical features of differently aged aggregates (Figs. 1 and S3). The highest values of Ca2+ influx were observed in cells exposed to samples enriched in oligomeric assemblies, whereas the W(F) value declined in cells exposed to aged samples, enriched in pre-fibrillar and fibrillar structures but with fewer oligomers. These data are in agreement with the ability of the aggregates to induce apoptosis, which was triggered by the soluble fraction of ATX3Q55 aggregated 48 h, whereas the precipitated fibrillar material was substantially harmless. These data are in concordance with those showing that the soluble but not the fibrillar fraction of ATX3Q55 aggregated 48 h induced apoptosis. When type 2 responses became detectable (Fig. 2b, curve IV), the W(F) values showed a staggered time course with respect to the type 1 response, supporting the hypothesis that the two responses were triggered by different types of aggregates. At longer aggregation times, the isolated fibrillar fraction of ATX3Q55 did not elicit any fluorescence response; while a type 2 response was induced by the soluble pre-fibrillar assemblies. The different effects of the early oligomers grown from the three ATX3 variants is likely to result from their different intrinsic structural features and biophysical properties, as detected by AFM, ThT and ANS assays. In the ATX3Q55 variant, the expanded polyQ tract dramatically favors irreversible aggregation and fibril growth, an event related to hydrogen bonding between glutamine side-chains [32]. Based on these observations and on the results presented here, we suggest that an expanded polyQ is needed to generate pre-fibrillar, soluble aggregates endowed with a distinctly higher surface hydrophobicity, a property that correlates with the unique capability of the expanded form to trigger type 2 responses. A moderate Ca2+ unbalance was detected with ATX3Q24, as compared to that induced by the other variants; accordingly, we expected much milder ATX3Q24 neurotoxicity with respect to ATX3Q55 and ATX3/291Δ. This was confirmed by the apoptosis results summarized in Fig. S8c, where cells exposed to ATX3Q24 aggregates displayed negligible damage compared to those resulting from exposure to the other variants. The observed toxicity of the truncated form, ATX3/291Δ, albeit lower than that of the expanded one, also deserves some comment. A recent study showed that Tg mice either homozygous or heterozygous for an ATX3 variant truncated at Gly259 developed severe motor dysfunction and altered behavior followed by premature death [66] suggesting that even the N-terminal region of ATX3 can play a pathological role in agreement with our data. We have also shown that truncation can speed up protein aggregation (Fig. S2a) [33] possibly by unmasking aggregation-prone regions located in the JD (unpublished results), most likely the stretches involved in the overall protein toxicity mechanism. Our data suggest that aggregate neurotoxicity could be related to either non-specific membrane permeabilization or aggregate binding

3163

to specific membrane targets, while a pore-like mechanism priming cell leakage seems unlikely, based on the results obtained by exposing artificial bilayers to polyQ peptides [62]. In the presence of CNQX, Ca2+ influx was remarkably reduced suggesting AMPA-R involvement in the permeability changes induced by ATX3Q55 oligomers (type 1 response), similarly to cells exposed to Aβ oligomers [58–60,67]. The co-localization experiments support this hypothesis and show that NMDA-Rs are involved in type 1 responses as well as suggested by APV inhibition. Immature CGNs exposed to NMDA showed a rise of the W(F) value with cell maturation, a process that results in a remarkable increase in NMDA-R expression (Fig. 6b). Interestingly, the fluorescence response of cells exposed to ATX3/291Δ and ATX3Q24 oligomers in the presence of APV and CNQX was similar, albeit not identical, to that seen with ATX3Q55 oligomers. This result indicates a basic similarity in the way the oligomers of the ATX3 variants interact with glutamatergic receptors, whereas the subsequent cascade of events can be substantially different, resulting in apoptosis only for ATX3Q55 and ATX3/291Δ. Finally, we have shown that voltage-dependent Ca2+ channels are also involved in Ca2+ rise as indicated by the fluorescence response to early ATX3Q55 and ATX3Q24 aggregates in the presence of Cd2+, a Ca2+ channel blocker, in agreement with the co-localization data, while no changes were recorded when the cells were exposed to 24 h-aged aggregates. Increasing evidence shows that lipid rafts play a crucial role in amyloid-associated neurodegenerative diseases [68,69]. Our results indicate that chemical modification of GM1, a typical raft ganglioside, with cleavage of its sialic acid moiety, significantly reduced the type 1 response and completely abolished the type 2 response in cells exposed to ATX3 oligomers. It therefore seems that the interaction of amyloid aggregates with glutamatergic receptors and/or the membrane bilayer requires well organized membrane rafts. In particular, GM1, and its sialic acid moiety appear to be needed for optimal interaction of ATX3 oligomers with the cell membrane, as previously reported for oligomers of other peptides/proteins [44,46,70]. That lipid rafts and their glutamatergic receptors are involved in the fluorescence response to the ATX3 oligomers is further supported by the increase in W(F) with increasing maturity of exposed CGNs, where lipid raft organization as well as NMDA-R expression rises (Fig. 6b and Table 2S). The set of protein variants we chose for this study allowed us to single out the contribution to the Ca2+ rise induced by early oligomers from that of more complex soluble later pre-fibrillar assemblies. In fact, in the investigated time interval, these latter structures were formed only by the ATX3Q55 variant and caused type 2 responses. These responses were not affected by the presence of glutamatergic receptor inhibitors such as APV or CNQX or by Cd2+, a blocker of voltage dependent Ca2+ channels. We conclude that, unlike amyloid oligomers, prefibrillar assemblies arising later do not interact with glutamatergic receptors or with Ca2+ channels but, possibly, with the charged sialic acid moiety of raft GM1. Taken together, our findings support the increasingly recognized idea that the same protein can form structurally distinct aggregates with different effects and toxicities to the exposed cells depending on aggregation conditions, and on structural modification of the aggregating protein. They also add to the body of knowledge highlighting the importance of the membrane lipid composition and structural organization for aggregate cytotoxicity, further highlighting the importance of both oligomer and membrane properties in determining the extent of the eventual toxicity to exposed cells. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.bbamcr.2013.08.019.

Acknowledgments This work was partially supported by MIUR (PRIN 2007, PRIN 2008), Fondazione CARIGE, Regione Lombardia (“NEDD”, Network-Enabled

3164

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165

Drug Design), Fondazione Cariplo (N.O.B.E.L. project), and the University of Genoa (Fondi di Ateneo). We are grateful to Marco Raimondo and Daniela Nichino for technical help, and to Nadia Marano for critical reading of the manuscript. G.I. acknowledges a postdoctoral fellowship of the University of Milano-Bicocca. References [1] H.Y. Zoghbi, H.T. Orr, Glutamine repeats and neurodegeneration, Annu. Rev. Neurosci. 23 (2000) 217–247. [2] J.F. Gusella, M.E. MacDonald, Molecular genetics: unmasking polyglutamine triggers in neurodegenerative disease, Nat. Rev. Neurosci. 1 (2000) 109–115. [3] C.J. Cummings, H.Y. Zoghbi, Fourteen and counting: unraveling trinucleotide repeat diseases, Hum. Mol. Genet. 9 (2000) 909–916. [4] W. Yang, J.R. Dunlap, R.B. Andrews, R. Wetzel, Aggregated polyglutamine peptides delivered to nuclei are toxic to mammalian cells, Hum. Mol. Genet. 11 (2002) 2905–2917. [5] I. Sanchez, C. Mahlke, J. Yuan, Pivotal role of oligomerization in expanded polyglutamine neurodegenerative disorders, Nature 421 (2003) 373–379. [6] L. Masino, G. Nicastro, R.P. Menon, F. Dal Piaz, L. Calder, J. Molloy, A. Pastore, Characterization of the structure and the amyloidogenic properties of the Josephin domain of the polyglutamine-containing protein ataxin-3, J. Mol. Biol. 344 (2004) 1021–1035. [7] A.M. Ellisdon, B. Thomas, S.P. Bottomley, The two-stage pathway of ataxin-3 fibrillogenesis involves a polyglutamine-independent step, J. Biol. Chem. 281 (2006) 16888–16896. [8] L. Masino, G. Nicastro, A. De Simone, L. Calder, J. Molloy, A. Pastore, The Josephin domain determines the morphological and mechanical properties of ataxin-3 fibrils, Biophys. J. 100 (2011) 2033–2042. [9] C. Nilsberth, A. Westlind-Danielsson, C.B. Eckman, N.M. Condron, K. Axelman, C. Forsell, C. Stenh, J. Luthman, D.B. Teplow, S.G. Younkin, J. Näslund, L. Lannfelt, The ‘Arctic’ APP mutation (E693G) causes Alzheimer's disease by enhanced Abeta protofibril formation, Nat. Neurosci. 9 (2001) 887–893. [10] M. Bucciantini, E. Giannoni, F. Chiti, F. Baroni, L. Formigli, J. Zurdo, N. Taddei, G. Ramponi, C.M. Dobson, M. Stefani, Inherent toxicity of aggregates implies a common mechanism for protein misfolding diseases, Nature 416 (2002) 507–511. [11] D.M. Walsh, I. Klyubin, J.V. Fadeeva, W.K. Cullen, R. Anwyl, M.S. Wolfe, M.J. Rowan, D.J. Selkoe, Naturally secreted oligomers of amyloid beta protein potently inhibit hippocampal long-term potentiation in vivo, Nature 416 (2002) 535–539. [12] Y. Gong, L. Chang, K.L. Viola, P.N. Lacor, M.P. Lambert, C.E. Finch, G.A. Krafft, W.L. Klein, Alzheimer's disease-affected brain: presence of oligomeric Abeta ligands (ADDL) suggests a molecular basis for reversible memory loss, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 10417–10422. [13] J.P. Cleary, D.M. Walsh, J.J. Hofmeister, G.M. Shankar, M.A. Kuskowski, D.J. Selkoe, K.H. Ashe, Natural oligomers of the amyloid-beta protein specifically disrupt cognitive function, Nat. Neurosci. 8 (2005) 79–84. [14] S. Lesnè, M.T. Koh, L. Kotilinek, R. Kayed, C.G. Glabe, A. Yang, M. Gallagher, K.H. Ashe, A specific amyloid-beta protein assembly in the brain impairs memory, Nature 440 (2006) 352–357. [15] M. Arrasate, S. Mitra, E.S. Schweitzer, M.R. Segal, S. Finkbeiner, Inclusion body formation reduces levels of mutant huntingtin and the risk of neuronal death, Nature 431 (2004) 805–810. [16] H. Mukai, T. Isagawa, E. Goyama, S. Tanaka, N.F. Bence, A. Tamura, Y. Ono, R.R. Kopito, Formation of morphologically similar globular aggregates from diverse aggregation-prone proteins in mammalian cells, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 10887–10892. [17] X. Chen, T.S. Tang, H. Tu, O. Nelson, M. Pook, R. Hammer, N. Nukina, I. Bezprozvanny, Deranged calcium signaling and neurodegeneration in spinocerebellar ataxia type 3, J. Neurosci. 28 (2008) 12713–12724. [18] M. Stefani, Generic cell dysfunction in neurodegenerative disorders: role of surfaces in early protein misfolding, aggregation and aggregate cytotoxicity, Neuroscientist 13 (2007) 519–531. [19] F. Pellistri, M. Bucciantini, A. Relini, D. Nosi, A. Gliozzi, M. Robello, M. Stefani, Nonspecific interaction of prefibrillar amyloid aggregates with glutamatergic receptors results in Ca2+ increase in primary neuronal cells, J. Biol. Chem. 44 (2008) 29950–29960. [20] F.G. De Felice, P.T. Velasco, M.P. Lambert, K. Viola, S.J. Fernandez, S.T. Ferreira, W.L. Klein, Abeta oligomers induce neuronal oxidative stress through an N-methyl-D-aspartate receptor-dependent mechanism that is blocked by the Alzheimer drug memantine, J. Biol. Chem. 282 (2007) 11590–11601. [21] S. Baglioni, F. Casamenti, M. Bucciantini, L.M. Luheschi, N. Taddei, F. Chiti, C.M. Dobson, M. Stefani, Prefibrillar amyloid aggregates could be generic toxins in higher organisms, J. Neurosci. 26 (2006) 8160–8167. [22] F. Tatini, A.M. Pugliese, C. Traini, S. Niccoli, G. Maraula, T. Ed Dami, B. Mannini, T. Scartabelli, F. Pedata, F. Casamenti, F. Chiti, Amyloid-β oligomer synaptotoxicity is mimicked by oligomers of the model protein HypF-N, Neurobiol. Aging 34 (2013) 2100–2109. [23] A. Demuro, A.E. Mina, R. Kayed, S.C. Milton, I. Parker, C.G. Glabe, Calcium dysregulation and membrane disruption as a ubiquitous neurotoxic mechanism of soluble amyloid oligomers, J. Biol. Chem. 280 (2005) 17294–17300. [24] S. Campioni, B. Mannini, M. Zampagni, A. Pensalfini, C. Parrini, E. Evangelisti, A. Relini, M. Stefani, C.M. Dobson, C. Cecchi, F. Chiti, A causative link between the structure of aberrant protein oligomers and their toxicity, Nat. Chem. Biol. 6 (2010) 140–147.

[25] K.M. Danzer, D. Haassen, A.R. Karow, S. Moussaud, M. Habeck, A. Giese, H. Kretzschmar, B. Hengerer, M. Kostka, Different species of alpha-synuclein oligomers induce calcium influx and seeding, J. Neurosci. 27 (2007) 9220–9232. [26] A. Deshpande, E. Mina, C. Glabe, J. Busciglio, Different conformations of amyloid beta induce neurotoxicity by distinct mechanisms in human cortical neurons, J. Neurosci. 26 (2006) 6011–6018. [27] Y. Nekooki-Machida, M. Kurosawa, N. Nukina, K. Ito, T. Oda, M. Tanaka, Distinct conformations of in vitro and in vivo amyloids of huntingtin-exon 1 show different cytotoxicity, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 9679–9684. [28] A.R.A. Ladiwala, J. Litt, R.S. Kane, D.S. Aucoin, S.O. Smith, S. Ranjan, J. Davis, W.E. Van Nostrand, P.M. Tessier, Conformational differences between two amyloid β oligomers of similar size and dissimilar toxicity, J. Biol. Chem. 287 (2012) 24765–24773. [29] R. Krishnan, J.L. Goodman, S. Mukhopadyyay, C.D. Pacheco, E.A. Lemke, A.A. Deniz, S. Lindquist, Conserved features of intermediates in amyloid assembly determine their benign or toxic states, Proc. Natl. Acad. Sci. U. S. A. 109 (2012) 1172–1177. [30] N. Cremades, S.I.A. Cohen, E. Deas, A.Y. Abramov, A.Y. Chen, A. Orte, M. Sandal, R.W. Clarke, P. Dunne, F.A. Aprile, C.W. Bertoncini, N.W. Wood, T.P.J. Knowles, C.M. Dobson, D. Klenerman, Direct observation of the interconversion of normal and toxic forms of α-synuclein, Cell 149 (2012) 1048–1059. [31] M. Stefani, Structural features and cytotoxicity of amyloid oligomers: implications in Alzheimer's disease and other diseases with amyloid deposits, Prog. Neurobiol. 99 (2012) 226–245. [32] A. Natalello, A.M. Frana, A. Relini, A. Apicella, G. Invernizzi, C. Casari, A. Gliozzi, S.M. Doglia, P. Tortora, M.E. Regonesi, A major role for side-chain polyglutamine hydrogen bonding in irreversible ataxin-3 aggregation, PLoS One 6 (2011) e18789. [33] C. Santambrogio, A.M. Frana, A. Natalello, E. Papaleo, M.E. Regonesi, S.M. Doglia, P. Tortora, G. Invernizzi, R. Grandori, The role of the central flexible region on the aggregation and conformational properties of human ataxin-3, FEBS J. 279 (2012) 451–463. [34] M. Robello, C. Amico, A. Cupello, Regulation of GABAA receptor in cerebellar granule cells in culture: differential involvement of kinase activities, Neuroscience 53 (1993) 131–138. [35] C. Marchetti, C. Amico, C. Usai, Functional characterization of the effect of nimodipine on the calcium current in rat cerebellar granule cells, J. Neurophysiol. 73 (1995) 1169–1180. [36] M. Bucciantini, G. Calloni, F. Chiti, L. Formigli, D. Nosi, C.M. Dobson, M. Stefani, Prefibrillar amyloid protein aggregates share common features of cytotoxicity, J. Biol. Chem. 279 (2004) 31374–31382. [37] B. Bolognesi, J.R. Kumita, T.P. Barros, E.K. Esbjorner, L.M. Luheshi, D.C. Crowther, M.R. Wilson, C.M. Dobson, G. Favrin, J.J. Yerbury, ANS binding reveals common features of cytotoxic amyloid species, ACS Chem. Biol. 5 (2010) 735–740. [38] V. Vetri, C. Canale, A. Relini, F. Librizzi, V. Militello, A. Gliozzi, M. Leone, Amyloid fibrils formation and amorphous aggregation in concanavalin A, Biophys. Chem. 125 (2007) 184–190. [39] M.K. Chow, A.M. Ellisdon, L.D. Cabrita, S.P. Bottomley, Polyglutamine expansion in ataxin-3 does not affect protein stability: implications for misfolding and disease, J. Biol. Chem. 279 (2004) 47643–47651. [40] E.M. Snyder, Y. Nong, C.G. Almeida, S. Paul, T. Moran, E.Y. Choi, A.C. Nairn, M.W. Salter, P.J. Lombroso, G.K. Gouras, P. Greengard, Regulation of NMDA receptor trafficking by amyloid-beta, Nat. Neurosci. 8 (2005) 1051–1058. [41] Z. Gu, W. Liu, Z. Yan, Regulation of N-methyl D-aspartic acid (NMDA) receptors by metabotropic glutamate receptor 7, J. Biol. Chem. 284 (2009) 10639–10649. [42] M.L. Wisniewski, J. Hwang, B.A. Bahr, Submicromolar Aβ42 reduces hippocampal glutamate receptors and presynaptic markers in an aggregation-dependent manner, Biochim. Biophys. Acta 1812 (2011) 1664–1674. [43] A. Francesconi, R. Kumari, R.S. Zukin, Regulation of group I metabotropic glutamate receptor trafficking and signaling by the caveolar/lipid raft pathway, J. Neurosci. 29 (2009) 3590–3602. [44] E. Evangelisti, C. Cecchi, R. Cascella, C. Sgromo, M. Becatti, C.M. Dobson, F. Chiti, M. Stefani, Membrane lipid composition and its physicochemical properties define cell vulnerability to aberrant protein oligomers, J. Cell Sci. 125 (2012) 2416–2427. [45] M. Bucciantini, D. Nosi, M. Forzan, E. Russo, M. Calamai, L. Pieri, L. Formigli, F. Quercioli, S. Soria, F. Pavone, J. Savistchenko, R. Melki, M. Stefani, Toxic effects of amyloid fibrils on cell membranes: the importance of ganglioside GM1, FASEB J. 26 (2012) 818–831. [46] F. Malchiodi-Albedi, V. Contrusciere, C. Raggi, K. Fecchi, G. Rainaldi, S. Paradisi, A. Matteucci, M.T. Santini, M. Sargiacomo, C. Frank, M.C. Gaudiano, M. Diociaiuti, Lipid raft disruption protects mature neurons against amyloid oligomer toxicity, Biochim. Biophys. Acta 1802 (2010) 406–415. [47] R. Kayed, E. Head, J.L. Thompson, T.M. McIntire, S.C. Milton, C.W. Cotman, C.G. Glabe, Common structure of soluble amyloid oligomers implies common mechanism of pathogenesis, Science 300 (2003) 486–489. [48] J.I. Kourie, C.L. Henry, Ion channel formation and membrane-linked pathologies of misfolded hydrophobic proteins: the role of dangerous unchaperoned molecules, Clin. Exp. Pharmacol. Physiol. 29 (2002) 741–753. [49] A. Quist, I. Doudevski, H. Lin, R. Azimova, D. Nq, B. Frangione, B. Kagan, J. Ghiso, R. Lal, Amyloid ion channels: a common structural link for protein-misfolding disease, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 10427–10432. [50] A. Relini, O. Cavalleri, R. Rolandi, A. Gliozzi, The two-fold aspect of the interplay of amyloidogenic proteins with lipid membranes, Chem. Phys. Lipids 158 (2009) 1–9. [51] S.M. Butterfield, H.A. Lashuel, Amyloidogenic protein–membrane interactions: mechanistic insight from model systems, Angew. Chem. Int. Ed. 49 (2010) 5628–5654. [52] M. Stefani, Biochemical and biophysical features of both oligomer/fibril and cell membrane in amyloid cytotoxicity, FEBS J. 277 (2010) 4602–4613.

F. Pellistri et al. / Biochimica et Biophysica Acta 1833 (2013) 3155–3165 [53] B.I. Kelly, A. Ferreira, Beta-amyloid-induced dynamin 1 degradation is mediated by N-methyl-D-aspartate receptors in hippocampal neurons, J. Biol. Chem. 281 (2006) 28079–28089. [54] H. Hsieh, J. Boehm, C. Sato, T. Iwatsubo, T. Tomita, S. Sisodia, R. Malinow, AMPAR removal underlies Abeta-induced synaptic depression and dendritic spine loss, Neuron 52 (2006) 831–843. [55] X. Hou, H.C. Parkington, H.A. Coleman, A. Mechler, L.L. Martin, M.I. Aguilar, D.H. Small, Transthyretin oligomers induce calcium influx via voltage-gated calcium channels, J. Neurochem. 100 (2007) 446–457. [56] H. Decker, S. Jürgensen, M.F. Adrover, J. Brito-Moreira, T.R. Bomfim, W.L. Klein, A.L. Epstein, F.G. De Felice, D. Jerusalinsky, S.T. Ferreira, N-Methyl-D-aspartate receptors are required for synaptic targeting of Alzheimer's toxic amyloid-β peptide oligomers, J. Neurochem. 115 (2010) 1520–1528. [57] R. Resende, C. Perreira, P. Agostino, A.P. Vieira, J.O. Malva, C.R. Oliveira, Susceptibility of hippocampal neurons to Abeta peptide toxicity is associated with perturbation of Ca2+ homeostasis, Brain Res. 1143 (2007) 11–21. [58] W.L. Klein, Synaptic targeting by Abeta oligomers (ADDLS) as a basis for memory loss in early Alzheimer's disease, Alzheimers Dement. 2 (2006) 43–55. [59] S.T. Ferreira, M.N. Vieira, F.G. De Felice, Soluble protein oligomers as emerging toxins in Alzheimer's and other amyloid diseases, IUBMB Life 59 (2007) 332–345. [60] S.J. Liu, R.S. Zukin, Ca2+-permeable AMPA receptors in synaptic plasticity and neuronal death, Trends Neurosci. 30 (2007) 126–134. [61] A.J. Williams, H.L. Paulson, Polyglutamine neurodegeneration: protein misfolding revisited, Trends Neurosci. 31 (2008) 521–528. [62] L. Mapelli, C. Canale, D. Pesci, S. Averaimo, F. Guizzardi, V. Fortunati, L. Falasca, M. Piacentini, A. Gliozzi, A. Relini, M. Mazzanti, C. Jodice, Toxic effects of expanded ataxin-1 involve mechanical instability of the nuclear membrane, Biochim. Biophys. Acta 1822 (2012) 906–917.

3165

[63] L.A. Khan, P.O. Bauer, H. Miyazaki, K.S. Lindenberg, B.G. Landwehrmeyer, N. Nukina, Expanded polyglutamines impair synaptic transmission and ubiquitin–proteasome system in Caenorhabditis elegans, J. Neurochem. 98 (2006) 576–587. [64] M. Jeub, M. Herbst, A. Spauschus, H. Fleischer, T. Klockgether, U. Wuellner, B.O. Evert, Potassium channel dysfunction and depolarized resting membrane potential in a cell model of SCA3, Exp. Neurol. 201 (2006) 182–192. [65] V.G. Shakkottai, M. do Carmo Costa, J.M. Dell'Orco, A. Sankaranarayanan, H. Wulff, H.L. Paulson, Early changes in cerebellar physiology accompany motor dysfunction in the polyglutamine disease spinocerebellar ataxia type, J. Neurosci. 31 (2011) 13002–13014. [66] J. Hübener, F. Vauti, C. Funke, H. Wolburg, Y. Ye, T. Schmidt, K. Wolburg-Buchholz, I. Schmitt, A. Gardyan, S. Driessen, H.H. Arnold, H.P. Nguyen, O. Riess, N-terminal ataxin-3 causes neurological symptoms with inclusions, endoplasmic reticulum stress and ribosomal dislocation, Brain 134 (2011) 1925–1942. [67] G.M. Shankar, B.L. Bloodgood, M. Townsend, D.M. Walsh, D.J. Selkoe, B.L. Sabatini, Natural oligomers of the Alzheimer amyloid-beta protein induce reversible synapse loss by modulating an NMDA-type glutamate receptor-dependent signaling pathway, J. Neurosci. 271 (2007) 2866–2875. [68] F. Malchiodi-Albedi, S. Paradisi, A. Matteucci, C. Frank, M. Diociaiuti, Amyloid oligomer neurotoxicity, calcium dysregulation and lipid rafts, Int. J. Alzheimers Dis. 906964 (2011) 1–17. [69] J.V. Rushworth, N.M. Hooper, Lipid rafts: linking Alzheimer's amyloid-β production, aggregation, and toxicity at neuronal membranes, Int. J. Alzheimers Dis. 2011 (603052) (2010) 1–14. [70] M. Molander-Melin, K. Blennow, N. Bogdanovic, B. Dellheden, J.E. Månsoon, P. Fredman, Structural membrane alterations in Alzheimer brains found to be associated with regional disease development; increased density of gangliosides GM1 and GM2 and loss of cholesterol in detergent-resistant membrane domains, J. Neurochem. 92 (2005) 171–182.