Epigallocatechin-3-gallate and related phenol compounds redirect the amyloidogenic aggregation pathway of ataxin-3 towards non-toxic aggregates and prevent toxicity in neural cells and Caenorhabditis elegans animal model

Abstract

The protein ataxin-3 (ATX3) triggers an amyloid-related neurodegenerative disease when its polyglutamine stretch is expanded beyond a critical threshold. We formerly demonstrated that the polyphenol epigallocatechin-3-gallate (EGCG) could redirect amyloid aggregation of a full-length, expanded ATX3 (ATX3-Q55) towards non-toxic, soluble, SDS-resistant aggregates. Here, we have characterized other related phenol compounds, although smaller in size, i.e. (−)-epigallocatechin gallate (EGC), and gallic acid (GA). We analysed the aggregation pattern of ATX3-Q55 and of the N-terminal globular Josephin domain (JD) by assessing the time course of the soluble protein, as well its structural features by FTIR and AFM, in the presence and the absence of the mentioned compounds. All of them redirected the aggregation pattern towards soluble, SDS-resistant aggregates. They also prevented the appearance of ordered side-chain hydrogen bonding in ATX3-Q55, which is the hallmark of polyQ-related amyloids. Molecular docking analyses on the JD highlighted three interacting regions, including the central, aggregation-prone one. All three compounds bound to each of them, although with different patterns. This might account for their capability to prevent amyloidogenesis. Saturation transfer difference NMR experiments also confirmed EGCG and EGC binding to monomeric JD. ATX3-Q55 pre-incubation with any of the three compounds prevented its calcium-influx-mediated cytotoxicity towards neural cells. Finally, all the phenols significantly reduced toxicity in a transgenic Caenorhabditis elegans strain expressing an expanded ATX3. Overall, our results show that the three polyphenols act in a substantially similar manner. GA, however, might be more suitable for antiamyloid treatments due to its simpler structure and higher chemical stability.

Introduction

In recent years, plenty of investigations have highlighted the beneficial effects of polyphenols in preventing and/or mitigating several disorders such as cancer, stroke and neurodegenerative diseases (1–3). Polyphenols are a large and diverse class of chemical compounds found in beverages obtained from fruits, vegetables, teas, cocoa and other plants. They are divided into several subclasses, the largest being represented by flavonoids. These share a basic structure consisting of two aromatic rings bound together by three carbon atoms that form an oxygenated heterocycle. Catechins, quercetin, myricetin are among the most common flavonoids (4). Fresh tea leaves contain a high amount of catechins, known to constitute 25–35% of the solid green tea extract and consisting of eight related compounds, namely (+)-catechin (C), (−)-epicatechin (EC), (+)-gallocatechin (GC), (−)-epigallocatechin (EGC), (+)-catechingallate (CG), (−)-epicatechin gallate (ECG), (+)-gallocatechin gallate (GCG) and (−)-epigallocatechin-3-gallate (EGCG). EGCG is the most abundant catechin, with an estimated content of about 90 mg per cup of green tea (2.5 g of green tea leaves/200 ml of water), and is thought to make a substantial contribution to the beneficial effects ascribed to this beverage, in particular to its neuroprotective properties (5).

It was formerly believed that the protective effect of EGCG against neurotoxicity could be essentially accounted for by its antioxidant activity, with resulting reduction of the harmful effects of oxygen-derived free radicals (6). Actually, oxidative damage and increased accumulation of iron in specific brain areas are considered major pathological features of Parkinson’s disease (7), Alzheimer’s disease (8) and amyotrophic lateral sclerosis (9), so special interest has been given to the therapeutic potential of nutritional antioxidants in neurodegenerative diseases. More recently, however, EGCG was also shown to directly interfere with amyloid fibril formation by several peptides and proteins and to remodel preformed fibrils, thus generating non-toxic species (10–13). In keeping with this idea, we previously observed that EGCG affects the aggregation pathway of ataxin-3 (ATX3), the protein responsible for spinocerebellar ataxia type 3 (SCA3), by a mechanism similar to that reported for the other mentioned amyloidogenic proteins. Specifically, EGCG interferes with the early steps of ATX3 aggregation, thus leading to the formation of off-pathway, non-toxic, SDS-stable final aggregates (14).

Unfortunately, one major constraint in the employment of EGCG as a pharmaceutical tool is its limited bioavailability, mainly due to poor systemic absorption, especially when administered orally (15). Several factors contribute to this outcome: a) its chemical instability at the intestinal pH around 8.5; b) the time required to cross the mucolytic layer and to be absorbed by the small intestinal epithelial cells; c) the transporters involved in its transport and the efflux transporters that recycle EGCG back into the lumen of the intestine; finally, d) the phase II enzymes involved in its transformation (16). EGCG also causes cytotoxicity at high concentrations (17). No less important, it displays a scanty capability to cross the blood-brain barrier (BBB) (18).

Searching for more effective treatments based on the administration of cathechins, we report here a study aimed at assessing the action mechanisms and the neuroprotective effects of two structurally related compounds, i.e. EGC and gallic acid (GA) (Fig.1). This approach is justified by the fact that a controlled conjugation of GA to nanodevices is an attainable goal, due to its structural simplicity and the presence of an unesterified carboxyl group, unlike the case of EGC and EGCG. Conjugation of therapeutic compounds to nanodevices offers a considerable advantage, as it makes it possible, in principle, their targeted delivery to the CNS (19,20). Furthermore, the therapeutic potential of GA has been already reported, in that it is capable of inhibiting fibril formation by amyloid β (Aβ) peptide (10), α-synuclein (21) and insulin (22). Additionally, a comparative analysis of the effects of the three mentioned catechins might provide a deeper insight into their action mechanism, particularly at the molecular level.

Figure 1

Structures of EGCG, EGC and GA. The EGCG cleavage point leading to the release of EGC and GA is indicated.

We therefore characterized in parallel the effects of EGCG, EGC and GA on the aggregation mode of both a pathogenic variant of ATX3 carrying 55 consecutive glutamines (ATX3-Q55) close at the C-terminus, and its N-terminal globular domain (Josephin domain, JD). In spite of the presence of the polyglutamine stretch, it has been clearly established that the aggregation process starts at the level of the JD (23–25), which prompted us to also analyse the effects of the catechins on this protein moiety in isolation. Furthermore, we assessed the neuroprotective effect of EGCG and GA in a transgenic Caenorhabditis elegans SCA3 model.

The molecular mechanisms of the interaction between the catechins and ATX3 were mainly investigated by molecular docking simulations, Fourier transform infrared (FTIR) spectroscopy, atomic force microscopy (AFM) and saturation transfer difference (STD) NMR experiments. Our results strongly support the idea that the GA moiety of EGCG is the minimal functional unit, although somewhat less neuroprotective than EGCG itself. Thus, thanks to its substantial stability and structural simplicity, it could be employed in the future development of pharmaceutical nanodevices.

Results

EGC and GA affect JD and ATX3-Q55 aggregation kinetics and structural features in the same way as EGCG

To evaluate the effect of EGC and GA on ataxin-3 (ATX3) aggregation, freshly prepared His-tagged JD and ATX3-Q55 (an expanded form carrying 55 consecutive glutamines) were incubated at 37 °C in PBS at a 125 µM and 25 µM final concentration, respectively, in the presence or the absence of EGCG, EGC or GA (1:5 protein:compound molar ratio) (Fig. 2). Aliquots were taken at different times of incubation. Then, the soluble fraction obtained by centrifugation was analysed by SDS-PAGE (Fig. 2A and B) and densitometric quantification of the SDS-soluble fraction (Fig. 2C and D). The latter is identified as the protein migrating in monomeric form in the gel. We confirmed that, for both proteins, EGCG treatment induced a rapid decrease in the SDS-soluble fraction starting from the earliest time of incubation compared with untreated samples (Fig. 2A–D). It also induced the formation of SDS-resistant species that are large and do not enter the separating gel, as previously reported (14,26) (Fig. 2A and B). This pattern can be represented quantitatively as time courses of SDS-soluble, SDS-resistant soluble and insoluble species (Supplementary Material, Fig. S1). In the case of JD, the much scantier amount of such species can be only explained by the formation of large aggregates, which are therefore discarded by centrifugation (Fig. 2A).

Figure 2

SDS-PAGE analysis of JD and ATX3-Q55 soluble fraction. (A) SDS-PAGE (16%) of the soluble protein fraction obtained by centrifugation of aliquots of 125 µM JD incubated at 37 °C in the absence or the presence of EGCG, EGC or GA at a molar ratio protein:compound 1:5. The arrows indicate the monomeric forms. (B) SDS-PAGE (12%) of the soluble protein fraction obtained by centrifugation of aliquots of 25 µM ATX3-Q55 incubated at 37 °C in the absence or the presence of EGCG, EGC or GA at a molar ratio protein:compound 1:5. The gels were stained with IRDye Blue Protein Stain (LiCor, USA). (C, D) SDS-soluble protein amounts of JD (C) and ATX3-Q55 (D) were quantified by densitometry. Signals were normalized at t₀ protein content. Error bars represent standard errors and are derived from at least three independent experiments. *P < 0.05; **P < 0.01.

Furthermore, both EGC and GA promoted a faster decrease in the JD SDS-soluble fraction in comparison with the untreated sample, like in the case of EGCG (Fig. 2A and C), although the latter exerted a much more pronounced effect. At the earliest times (1–6 h of incubation), the decrease in soluble fraction was also paralleled by the appearance of soluble, SDS-resistant species (Fig. 2A).

Likewise, also in the case of ATX3-Q55, the addition of EGC or GA induced a rapid decrease in the SDS-soluble fraction and the appearance of large, soluble SDS-resistant species but to a lesser extent with respect to EGCG (Fig. 2B and D).

EGCG, EGC and GA drastically affect the structural features of JD and ATX3-Q55 aggregation intermediates

The effects of GA and EGC on JD secondary structure and aggregation were also investigated by Fourier transform infrared (FTIR) spectroscopy and were compared with those exerted by EGCG (Fig. 3). The FTIR spectra were collected in Attenuated Total Reflection (ATR) on protein hydrated film (27) as previously reported (14,26,28,29). In the Amide I spectral region (1600–1700 cm⁻¹), the second derivative spectrum of the freshly purified JD displayed minima, corresponding to maxima of the absorption spectrum, at ∼1635 cm⁻¹ and ∼1690 cm⁻¹, both assigned to intramolecular β-sheets, and at ∼1657 cm⁻¹ manly due to the α-helices, with the contribution of random coil structures of the native protein (28). During the incubation at 37 °C in PBS, the component assigned to the protein native secondary structures decreased in intensity and two new peaks appeared in the spectra at ∼1693 cm⁻¹ and ∼1623 cm⁻¹, in the typical spectral region of intermolecular β-sheets in protein aggregates (Fig. 3A). These results indicate the loss in native secondary structures and the formation of protein aggregates, in agreement with previous reports (26,28). FTIR analysis was also performed in the presence of GA at 1:5 protein:compound molar ratio (Fig. 3D). Compared with JD alone, GA induced an earlier intensity decrease in the native β-sheet peak (Fig. 3D and E) and an increase in the intermolecular β-sheet component (Fig. 3D and F), which at the end of the incubation reached a lower intensity compared with the control. Moreover, the peak position of the intermolecular β-sheets was upshifted to ∼1627 cm⁻¹ in the sample incubated with GA (Fig. 3G and H). Amide I shifts can be induced by an altered coupling with neighbouring amide oscillators, due to backbone conformational changes, and by a different degree of hydrogen bonding of the peptide bonds (30). As suggested by computational and experimental studies, an upshift of the main β-sheet peak can be ascribed to a decreased average number of strands per sheets and to larger twist angles (31,32). Furthermore, weakening of the backbone H-bonding can also cause an upshift of the main β-sheet peak (30,33). Therefore, the observed upshift in the sample incubated with GA points to the formation of aggregates with a slightly different conformation, comprising loosely packed structures and/or a reduced average number of strands per sheets in the presence of the compound.

Figure 3

FTIR analyses of the effects of EGCG, EGC and GA on JD misfolding and aggregation. (A–D) Second derivatives of absorption spectra of JD (150 µM) in the presence or the absence of compounds (750 µM), collected at different incubation times in PBS at 37 °C. Peak positions of the main components and their assignment to the protein secondary structures are reported in (A). The arrows point to increasing incubation times. (E) Time course of the component assigned to native β-sheets. (F) Time course of the component assigned to intermolecular β-sheets. (G) Peak position of the component assigned to intermolecular β-sheets taken from second derivatives of spectra collected from JD samples incubated for two weeks (2 ws) at 37 °C in the presence or the absence of compounds. (H) Second derivative spectra of the pellet collected from JD samples incubated for two weeks at 37 °C in the presence or the absence of compounds. Second derivative spectra were normalized at the tyrosine peak at ∼1515 cm⁻¹.

Similar effects on JD misfolding and aggregation were also observed in the presence of EGC (Fig. 3C, E–H) and, more markedly, in the presence of EGCG (Fig. 3B–H). In previous works (14,26), we showed that EGCG is able to induce JD and full-length ATX3-Q55 misfolding, leading to the formation of aggregates that are off-pathway with respect to fibrillogenesis. The FTIR analyses here reported indicate that both GA and EGC are able to exert effects similar to those of EGCG on the structural properties of JD aggregation intermediates, albeit to a reduced extent, with the following order of decreasing efficacy: EGCG > EGC > GA (Fig. 3E–H).

FTIR analyses were also performed on full-length, expanded ATX3-Q55. During the incubation at 37 °C in the absence of any compound added, the IR peak at ∼1635 cm⁻¹, due to native β-sheets, decreased in intensity and a new peak appeared at ∼1624 cm⁻¹ (Fig. 4A). This new component, along with that detected at ∼1694 cm⁻¹, has been assigned to the formation of intermolecular β-sheet structures in the protein aggregates (14,28). The ∼1657 cm⁻¹ component (due to α-helical and random coil structures in the freshly purified protein) decreased in intensity at the beginning of the incubation, then increased again at later times, with a parallel appearance of a new component at ∼1604 cm⁻¹. In a previous work (28) we unambiguously assigned the ∼1657 cm⁻¹ and ∼1604 cm⁻¹ bands to glutamines involved in strong side chain-side chain (and possibly side chain-backbone) hydrogen bonding in the ATX3-Q55 mature amyloid aggregates. In the presence of GA at a 1:5 protein:compound molar ratio, a faster intensity decrease of the native β-sheet peak at ∼1635 cm⁻¹ (Fig. 4D and E) was observed in comparison with ATX3-Q55 alone. Moreover, the components assigned to intermolecular β-sheets and side-chain H-bonded glutamines, respectively at ∼1624 cm⁻¹ and at ∼1604 cm⁻¹, in the presence of the compound reached a lower intensity compared with the control (Fig. 4D and F). These results suggest that GA is able to induce a partial unfolding of ATX3-Q55 and to decrease the fraction of the protein forming mature amyloid aggregates. Similar effects on ATX3-Q55 aggregation were observed for EGC at 1:5 of protein:compound molar ratio (Fig. 4C, E and F). These effects were more evident in the case of EGC compared with GA. In the presence of EGCG at a 1:5 ATX3-Q55:compound molar ratio, the 1635 cm⁻¹ component immediately decreased in intensity with a simultaneous appearance of a shoulder at lower wavenumbers, which slightly increased with times (Fig. 4B–F). These results are in agreement with a previous study, whereby it was shown that EGCG induces the formation of ATX3-Q55 soluble, SDS-resistant aggregates, with low β-sheet content and without ordered side-chain hydrogen bonding (14). Overall, the FTIR data indicate that GA, EGC, and EGCG are able to induce ATX3-Q55 misfolding, redirecting its fibrillogenic process toward off-pathway aggregates. As in the case of JD, the efficacy of these compounds is in the following order: EGCG > EGC > GA (Fig. 4). Under all conditions, a fraction of ATX3-Q55 formed mature amyloid aggregates with an ordered array of H-bonded glutamine side chains, as indicated by the FTIR spectra of the pellet collected after a two-weeks incubation at 37 °C of in the presence or the absence of the different compounds (Supplementary Material, Fig. S2).

Figure 4

FTIR analyses of the effects of EGCG, EGC and GA on ATX3-Q55 aggregation. (A–D) Second derivatives of absorption spectra of ATX3-Q55 (25 µM) in the presence or the absence of compounds (125 µM), collected at different incubation times in PBS at 37 °C. The assignment of the main components is reported in (A). The arrows point to increasing incubation times. (E) Time course of the component assigned to native β-sheets. (F) Time course of the component assigned to intermolecular β-sheets. Second derivative spectra were normalized at the tyrosine peak at ∼1515 cm⁻¹.

EGCG, EGC and GA inhibit ATX33-Q55 fibrillogenesis

Tapping mode atomic force microscopy was employed to obtain information on the effects of EGCG, EGC and GA on ATX3-Q55 aggregate morphology. Figure 5 compares representative images obtained after a 48-h aggregation of 25 µM ATX3-Q55 in the absence or the presence of EGCG, EGC and GA at a molar ratio protein:compound of 1:5. ATX3-Q55 alone formed fibrils arranged into bundles of height between 20 and 60 nm (Fig. 5A), in agreement with previous observations (14,28). The mean bundle length was 1100 ± 200 nm and the mean bundle height, measured at the middle of the bundle, was 43 ± 8 nm. In the presence of EGCG (Fig. 5B), fibrillation was completely suppressed and clusters of non fibrillar material were observed, with cluster height between 20 and 80 nm and typical cluster size in the scan plane of 0.5 − 1 µm, in agreement with previous findings (14).

Figure 5

AFM analysis of AT3XQ55 aggregates obtained in the absence and in the presence of EGCG, EGC and GA. Tapping mode AFM images (height data) of ATX3Q55 aggregates obtained after 48-h incubation of 25 µM ATX3Q55 in the absence (A) and in the presence of EGCG (B), EGC (C), GA (D) at a molar ratio protein:compound 1:5. Scan size 3.0 μm. The color bar corresponds to a Z range of 120 nm (A), 80 nm (B, C, D).

In the presence of EGC and GA, fibrillation was not completely suppressed. The fibril bundle lengths were 900 ± 100 nm for EGC and 800 ± 200 nm for GA, and the bundle heights were 40 ± 10 nm and 44 ± 10 nm, respectively. These results indicate that these compounds were less effective than EGCG. However, in the presence of EGC, bundles of modified morphology were often observed (Fig. 5C) and for both EGC and GA, non-fibrillar and relatively flat aggregates of height between 10 and 12 nm were also found (Fig. 5C and D). These results point to the formation of off-pathway species.

Molecular docking investigations show that EGCG, EGC and GA bind the JD in the same regions

To get insight into the binding mode of the ligands EGCG, EGC and GA to the JD, molecular docking investigations were carried out for all ten conformations presented in the 1YZB pdb file, previously obtained by NMR spectroscopy (34,35). Our results reveal that the interaction of GA, EGC and EGCG with the protein is non-specific, as supported by the fact that different binding sites were found over the whole surface of JD for all ligands (Fig. 6; Supplementary Material, Fig. S3). We focused our attention on the best pose of each ligand in all JD NMR conformations (Table 1). This analysis highlights, for all compounds, three binding sites located at the top, middle and bottom regions of the protein, respectively. In the case of GA and EGCG, the best binding site is located in the middle region comprising the amino acids 120–166 (Fig. 6A and C). This region is involved in the binding event in seven out of ten NMR conformations, in the case of GA, and in eight in the case of EGCG. In contrast, EGC showed a higher affinity for the top and bottom region, comprising the amino acids 37–78 and 74–118, respectively (Fig. 6B). In such case, EGC binding to the central region was observed in two NMR conformations only.

Table 1

For each ligand the calculated binding energies and binding regions of the bes poses for all NMR conformations of JD are shown

	Binding Energy (kcal/mol)	Region of interaction
GA
NMR1	−5.4	3
NMR2	−5.9	3
NMR3	−4.6	3
NMR4	−4.0	3
NMR5	−5.0	3
NMR6	−4.9	3
NMR7	−4.5	1
NMR8	−4.6	2
NMR9	−4.8	1
NMR10	−6.1	3
EGC
NMR1	−5.7	2
NMR2	−6.7	1
NMR3	−6.0	3
NMR4	−5.3	2
NMR5	−4.6	1
NMR6	−4.7	2
NMR7	−4.8	1
NMR8	−6.7	2
NMR9	−6.0	1
NMR10	−7.0	3
EGCG
NMR1	−5.7	3
NMR2	−7.0	3
NMR3	−7.0	3
NMR4	−7.2	1
NMR5	−6.0	3
NMR6	−7.0	3
NMR7	−6.6	3
NMR8	−6.8	3
NMR9	−7.0	2
NMR10	−5.5	1

Figure 6

Best scoring poses for EGCG (A), EGC (B) and GA (C) docked on JD structure derived from 1YZB-PDB code. Molecular docking was performed with Glide software. Highlighted in color are the three regions where the compounds bind the ten NMR structures.

STD NMR analysis provides evidence of EGCG, and EGC binding to monomeric JD

Further insight into the binding mode of EGCG, EGC and GA to JD was provided by Saturation Transfer Difference (STD) NMR experiments (36,37). STD NMR is a robust method allowing the detection and characterization of receptor-ligand interactions in solution, based on the observation of the signals of the small molecule (ligand). In previous works, we exploited this approach to characterize the molecular interactions of different natural (38–40) and synthetic ligands (41–43) with amyloid peptide and proteins, including ATX3-Q55 (44) and JD (26).

Here, we carried out experiments on ligand/protein mixtures dissolved in deuterated phosphate buffer, pH 7.2, 5 °C. The selective saturation of some aliphatic resonances of JD was achieved by irradiating at −1.00 ppm (on-resonance frequency). In fact, when the experimental conditions are chosen properly to assure the absence of direct irradiation of the test compound (verified through blank experiments on a sample containing the potential ligand only), the presence of its NMR signals in the STD spectrum unequivocally indicates its interaction with the receptor. Concurrently, any signal coming from non-binding compounds is erased in the STD spectrum, thus demonstrating that the molecule is not a ligand (Fig. 7).

Figure 7

STD NMR characterization of EGCG, EGC and GA binding to JD. (A) ¹H NMR spectrum of a mixture containing 7 µM JD and 1.5 mM EGCG. (B) STD NMR spectrum of the same mixture of the spectrum A with a saturation time of 3 s. (C) ¹H NMR spectrum of a mixture containing 7 µM JD and 1.5 mM EGC. (D) STD NMR spectrum of the same mixture of the spectrum C with a saturation time of 3 s. (E) ¹H NMR spectrum of a mixture containing 7 µM JD and 1.5 mM GA. (F) STD NMR spectrum of the same mixture of the spectrum with a saturation time of 3 s. All the samples were dissolved in PBS solution, pH 7.2, 5 °C. The spectrometer frequency was 600 MHz. The EGCG H₂ signal is overlapped by water resonance.

STD NMR spectra recorded on samples containing 7 µM JD and each of the three potential ligands EGCG, EGC and GA, are depicted in Figure 7B, D and F, respectively. In agreement with data reported on the interaction of these compounds with ATX3-Q55 (44), EGCG and EGC are ligands of JD monomers, as demonstrated by the presence of some of their resonances in STD spectra B and D, while no evidence of binding to the protein was obtained for GA, whose STD spectrum (Fig. 7F) shows no resonances from this molecule. As GA also has an effect comparable to, although somewhat weaker than that of EGCG and EGC in redirecting amyloid aggregation, we suggest that the failure to detect binding under these experimental conditions may be accounted for by a transient interaction by the former.

EGCG, EGC and GA inhibit ATX3-Q55 aggregate toxicity in neural cells

It was previously reported that the interaction of ATX3-Q55 amyloid aggregates with neural cell membranes induces an increase in membrane permeability, resulting in the disruption of calcium homeostasis and subsequent cytotoxicity (45). To assess the ability of EGCG, EGC and GA to inhibit such cytotoxic effects of ATX3-Q55, rat cerebellar granule cells were loaded with the calcium-sensitive fluorescent dye Oregon Green, to monitor calcium influx induced by protein aggregates (Supplementary Material, Fig. S4). Cells were incubated for 24 h with 48 h-aged ATX3-Q55 aggregates formed in the absence or the presence of EGCG, EGC and GA. Figure 8 reports the corresponding fluorescence increase, resulting from calcium influx, compared to that of control cells incubated in the absence of protein aggregates. In the presence of EGCG or EGC, the increase in intracellular calcium levels elicited by aggregates was completely suppressed. However, even GA reduced calcium influx by as much as 75%.

Figure 8

Fluorescence analysis of the effect of EGCG, EGC and GA on ATX3-Q55 aggregates toxicity in primary neuronal cells. Oregon Green fluorescence increase, indicating Ca²⁺ influx, was recorded in rat cerebellar granule neurons after 24 h incubation with 48 h-aged ATX3-Q55 aggregates obtained in the absence or the presence of EGCG, EGC, GA. Fluorescence changes were measured with respect to control cells incubated without protein aggregates.

EGCG, EGC and GA ameliorate the pathological phenotype of a SCA3 C.elegans model

A SCA3 C. elegans model (14) was used to evaluate the effects of EGCG, EGC and GA on worm pathological phenotype. A wild type variant (ATX3Q17-GFP) and a pathological one (ATX3Q130-GFP) were expressed in the nervous system in fusion with GFP and under the control of the unc-119 promoter. Wild type Bristol N2 strain was used as a control. The compounds were directly added to E. coli OP50 suspension used to feed worms. Their effect was evaluated by monitoring life span and quantifying body bends frequency after 24 h (Fig. 9A) and 48 h (Fig. 9B) of incubation. None of these compounds was able to increase in a statistically significant manner worms’ survival (data not shown). However, after 24 h of treatment, all compounds promoted a statistically significant increase in worm mobility in the Q130-GFP strain with respect to untreated worms. This increment was 30, 17 and 12% in the presence of EGCG, EGC and GA, respectively (Fig. 9C). A statistically significant increase in mobility was also observed after 48 h of treatment in the Q130-GFP strain (25, 12, 8% in the presence of EGCG, EGC and GA, respectively; Fig. 9D).

Figure 9

Pharmacological assay on ATX3 transgenic worms. One-day synchronized adult worms (wild type N2 worms as a control and Q17-GFP and Q130-GFP strains carrying the ATX3 variants) were placed on a plate seeded with the OP50 E. coli strain in the presence or the absence of 0.1 mM EGCG, EGC or GA, and cultured at 25 °C. Number of body bends/20 s were scored after 24 h (A) and 48 h (B) of treatment. Data were also expressed as percentage of motility increase with respect to the untreated animals after 24 h (C) and 48 h (D) of treatment. Error bars represent standard errors. Plots are representative of at least three independent experiments. *P < 0.05; **P < 0.001.

Discussion

In a previous report, we demonstrated EGCG’s capability to redirect ATX3 aggregation pathway towards non-amyloid, non-toxic aggregates and to ameliorate the ataxic phenotype of C. elegans (14). However, in view of its employment as a drug for the treatment of the relevant disease, its major flaw is that it can undergo chemical modifications with resulting loss of antiamyloid efficacy and scanty bioavailability (15,16). This prompted us to assay the efficacy of simpler compounds, yet structurally related to EGCG, i.e. EGC and GA. Furthermore, simpler structures should make possible, in principle, to easily accomplish their derivatization to specific vectors, such as nanoparticles, with enhanced brain delivery (43,46). This encouraged us to perform an extensive characterization of their action at the molecular and cellular level, as well as on the C. elegans animal model.

In particular, we analysed the effects of the three compounds on expanded ATX3-Q55 and JD aggregation, the latter being assayed because of its well-known involvement in the earliest events of aggregation (23–25). On the whole, our SDS-PAGE and FTIR analyses have clearly shown that the effects exerted by the three compounds on ATX3 amyloidogenesis are qualitatively similar, as all of them redirect the aggregation process towards soluble, SDS-resistant and non-amyloid off-pathway aggregates. In particular, FTIR spectra collected during the progress of ATX3-Q55 aggregation displayed a strong reduction of the 1604 cm-1 peak, assigned to glutamine side-chain hydrogen bond network, which is the hallmark of amyloid aggregation of proteins carrying expanded polyglutamine stretches (14). AFM data also support this hypothesis as clear morphological differences were observed between aggregates generated by ATX3-Q55 in the presence or the absence of the compounds. Indeed, all treatments resulted in the formation of clusters of non-fibrillar material. Nevertheless, the data show a different efficacy among the treatments with the following order: EGCG > EGC > GA.

To better understand the mode by which the three compounds prevent amyloidogenesis, we also performed molecular docking and NMR analyses on the JD. Molecular docking analyses highlighted three interacting regions, which points to a non-specific binding mode. All three compounds bound to each of them, although with different patterns. Noteworthy, all three compounds bound to the previously identified aggregation-prone regions (APR) (23,24). These results provide a possible interpretation of the mechanism by which these compounds prevent ATX3 fibrillogenesis, given their capability of directly binding to the APR, which suggests that the interaction occurs at the very beginning of the aggregation process.

STD NMR also confirmed that EGCG and EGC are capable to bind to the monomeric form of the protein, whereas the failure of detecting GA binding is quite likely due to a weaker and/or transient interaction by this compound. A previous work reports a similar failure to detect in NMR spectroscopy the interaction of this compound with monomeric α-synuclein (47).

Provided that the compounds’ antiamyloid action relies upon their capability of preventing intermolecular interactions among APR, a still unanswered issue is how this initial effect give rises to non-amyloid aggregates. Wobst and coworkers recently reported that EGCG can prevent the tau protein aggregation into toxic oligomers (48), by interfering with the formation of β-sheet-rich oligomeric tau species. Thus, it might be that EGCG’s primary effect on the APR, and/or subsequent as yet unidentified interactions, results in an altered pattern of β-sheet formation, as actually showed by our FTIR data. It should be also mentioned, however, that, according to a recent analysis on fourteen disease-related proteins and peptides, EGCG would bind to cross-ß sheet aggregation intermediates (49). Thus, the case of ATX3 does not conform to the more generally observed pattern, given the capability of the compound to first bind to monomeric protein, although this is the case of also other proteins (50).

Although our attempts to assess the binding stoichiometry of EGCG to ATX3 failed, we suggest that the ligand does not saturate the protein even at the highest stoichiometric ratio adopted (5:1). This assumption justified by: i) the substantially non-specific mode of binding of the ligand, as attested by its capability to bind to a very large number of both folded and disordered proteins (50); ii) a saturated binding stoichiometry even higher than 20, as assessed in the case of the small beta-amyloid peptide (51).

Noteworthy, tetracycline, another well-known antiamyloid agent, was proven to interfere with ATX3 fibrillogenesis and prevent toxicity by a substantially different mechanism compared with that displayed by EGCG, EGC and GA. In fact, tetracycline did not change the structural features of the aggregated species, but drastically increased their solubility (14). This effect is mediated by its capability to only bind oligomers (26), in keeping with our docking results that did not detect any possible binding mode between the antibiotic and JD in monomeric form (unpublished results).

The protective effect of the three compounds was demonstrated in vivo using both neural cells and the C. elegans animal model. In the first approach, we preincubated ATX3-Q55 with EGCG, EGC and GA, which resulted in the formation of aggregates displaying a substantially lower calcium-mediated cytotoxicity, compared with those formed by the untreated protein. Furthermore, all three compounds ameliorated the pathological phenotype of diseased worms, as shown by their improvement in locomotion. Both experimentations also confirmed the aforementioned order of efficacy, i.e. EGCG > EGC > GA. It should be stressed, nevertheless, that even GA displayed a significant capability of preventing cytotoxicity.

As regards the mechanisms underlying the effects detected, the protective action we observed on neural cells must be only fulfilled via prevention of toxic amyloid aggregate formation. In contrast, the outcome detected in C. elegans quite likely results from a combination of effects exerted at the cellular level, along with those specifically acting on aggregate formation. This hypothesis is in agreement with other literature reports, wherein the cytoprotective role of phenolic compounds mainly occurs via protection against oxidative stress and/or autophagy stimulation (52–54).

In conclusion, we have demonstrated that EGC and GA display a mechanism of action similar to that observed for EGCG, irrespective of whether they act at the molecular, cellular, or whole animal level. However, we observed a different efficacy by the three compounds, as above outlined. It is worthwhile to mention that GA represents the minimal functional unit of EGCG and, in general, of related phenolic compounds. This finding stimulates our interest in GA-mediated effects because, due to its relatively simple structure, it is suitable for conjugation to nanovectors, which may be bound to molecules enabling them to cross the blood brain barrier.

Materials and Methods

ATX3 purification

ATX3-Q55 cDNA was previously subcloned in the pQE30 vector and the protein expressed in the Escherichia coli strain SG13009 (E. coli K12 Nals, StrS, RifS, Thi2, Lac2, Ara+, Gal+, Mtl2, F2, RecA+, Uvr+, Lon+; Qiagen Hamburg GmbH, Hamburg, Germany) as His-tagged protein (28). JD cDNA was previously subcloned in pET21a vector and the protein expressed in the E. coli strain BL21 Tuner (DE3) pLacI (E. coli B F⁻ ompT hsdSB (rB⁻ mB⁻) gal dcm lacY1(DE3) pLacI (CamR); Novagen, Germany) as His-tagged protein (26). Proteins were purified as previously described (14,26).

SDS-PAGE and densitometry analysis of soluble protein fraction

Freshly purified ATX3-Q55 (25 µM) or JD (125 µM) was incubated at 37 °C in PBS solution (25 mM potassium phosphate, pH 7.2, 150 mM NaCl) in the presence or the absence of EGCG, EGC or GA (Sigma-Aldrich Inc., St Louis, MO, USA) at a protein:compound molar ratio of 1:5. Protein aliquots at different times of incubation (0, 1, 3, 6, 24, 48, and 72 h) were centrifuged 15 min at 14000×g and 10 µl for the expanded form or 3 µl for the JD of the supernatants were subjected to SDS-PAGE. The gels were stained with IRDye Blue Protein Stain (LiCor Biosciences, Lincoln, NE, USA), scanned at 700 nm with the Odyssey Fc System and analysed with the Image Studio software (LiCor Biosciences, Lincoln, NE, USA).

FTIR spectroscopy

FTIR analyses were performed using a previously optimized approach (14,26,28,29). Briefly, after vortexing, 2 µl of the sample solution, taken at different times of incubation at 37 °C in PBS, were deposited on the diamond surface of the single reflection ATR device (Quest, Specac, UK). After solvent evaporation, in order to obtain a protein hydrated film (27), the ATR/FTIR spectra were collected using a Varian 670-IR spectrometer (Varian Australia Pty Ltd,Mulgrave, VIC, Australia) equipped with a nitrogen-cooled mercury cadmium telluride detector under the following conditions: 2 cm⁻¹ spectral resolution, 25 kHz scan speed, 1000 scans co-addition and triangular apodization. The protein spectra were obtained by subtraction of the proper reference spectra (14,26,28). Second derivatives of the spectra were obtained after Savitzky-Golay smoothing. Spectra collection and analyses were performed using the ResolutionsPro software (Varian Australia Pty Ltd, Mulgrave, VIC, Australia). We previously reported that the transmission-mode FTIR spectra of freshly purified ATX3-Q55 in the form of a semi-dry film on a BaF₂ window display the same Amide I features of that collected in ATR modes indicating that the interaction of the protein with the diamond surface do not affect the protein secondary structures (14). Moreover, similar Amide I spectral features have been observed for the protein measured in ATR mode with and without solvent evaporation (data not shown and 29).

Atomic force microscopy (AFM)

ATX3-Q55 was incubated at 37 °C in PBS buffer at a concentration of 25 μM in the presence or the absence of EGCG, EGC or GA at a molar ratio protein:compound of 1:5. At fixed aggregation times, a 10 μl aliquot was withdrawn, incubated on a freshly cleaved mica substrate for 5 min, then rinsed with Milli-Q water and dried under mild vacuum. AFM images were acquired in tapping mode in air using a Multimode Scanning Probe Microscope equipped with ‘‘E’’ scanning head (maximum scan size 10 μm) and driven by a Nanoscope V controller (Bruker). Single beam uncoated silicon cantilevers (type OMCL-AC160TS, Olympus) were used. The drive frequency was between 270 and 320 kHz, the scan rate between 0.25 and 0.5 Hz. Aggregate size was measured from the cross-sections of topographic AFM images. Errors were calculated according to Student’s statistics assuming a confidence level of 95%.

Docking analysis

The NMR-resolved three-dimensional structure of JD protein was retrieved from the protein data bank (PDB ID: 1YZB) (34,35). Docking analyses were performed using Glide software from Schrödinger suite (48). All 10 NMR conformations of the PDB file were used in the docking procedure. The ligands EGCG, EGC and GA were prepared with the specific tool LigPrep, generating any possible protonated state at pH 7.2± 0.2. The XP Scoring function was used to score and rank of the compounds poses (55). All calculations were performed using the OPLS3 force field (56).

NMR analysis

NMR experiments were recorded on a Bruker 600 MHz Avance III equipped with a QCI cryo-probe, with a z-axis gradient coil. EGCG EGC and GA were dissolved in PBS, pH 7.2 and an aliquot of protein solution, dissolved in the same buffer, was added to reach the final concentration required. Basic sequences were employed for ¹H and STD-NMR experiments. Solvent suppression was performed by excitation sculpting. ¹H spectra were acquired with 128 scans and 2 s recycle delay. For STD-NMR experiments, a train of Gaussian-shaped pulses each of 50 ms was employed to saturate selectively the protein envelope; the total saturation time of the protein envelope was varied between 3 s and 0.15 s. Each STD spectrum was acquired with 1024 scans; acquisitions were performed at 5 °C.

Granule cell preparation

Sprague-Dawley rats were housed in the animal facility of the Department of Pharmacy, Section of Pharmacology and Toxicology of Genoa University. Experimental procedures and animal care complied with the EU Parliament and Council Directive of 22 September 2010 (2010/63/EU) and were approved by the Italian Ministry of Health (protocol number 2207) in accordance with D.M. 116/1992. All efforts were made to minimize animal suffering and to use the minimum number of animals necessary to produce reliable results. Granule cells were prepared from cerebella of 7–8-day-old rats as previously described (57). The cells were plated at a density of 1x10⁶ per dish on 20 mm poly-L-lysine-coated glass coverslips and maintained in Basal Eagle’s culture medium, containing 10% fetal calf serum, 100 μg/ml gentamicin and 25 mM KCl, at 37 °C in a humidified 95% air, 5% CO₂ atmosphere. Cultures were treated with 10 μM cytosine arabinoside from day 1 in order to minimize proliferation of non-neuronal cells. Experiments were performed in cultures between days 6 and 10 after plating.

Intracellular Ca²⁺ concentration measurements in rat cerebellar granule cells

Granule cells were incubated at 37 °C for 40 min in a 6 μM solution of the cell-permeant AM ester of Oregon Green (Molecular Probes, Eugene, OR) and then washed several times with washing solution (135 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl₂, 1.0 mM MgCl₂, 5.0 mM Hepes, 10 mM glucose, pH 7.4). Then, they were transferred to a recording chamber mounted onto a Nikon Eclipse TE300 inverted microscope. Cells were visualized using a ×100 objective in oil (N.A. 1.3) as previously described (58). The fluorescence signal was detected using a Hamamatsu digital CCD camera with a 450–490 nm excitation filter, a 505 nm dichroic mirror, and a 520 nm emission filter (Nikon Italia, Florence, Italy). Images were acquired with the Simple PCI software (Hamamatsu, Sewickley, PA). Fluorescence intensity was calculated in arbitrary units by building a scale of the pixel intensity located in the region of interest. Fluorescence intensity changes were calculated as

$$(\mathrm{F}-{\mathrm{F}}_0)\times 100/{\mathrm{F}}_0$$

where F is the fluorescence intensity measured after 24 h treatment with protein aggregates and F₀ the basal fluorescence level (Supplementary Material, Fig. S4).

C.Elegans strains and maintenance

ATX3Q17 and ATX3Q130 cDNAs were previously cloned in pPDP 95.77 vector as GFP fusion protein under control of pan neural unc-119 promoter as reported in (14). Bristol N2 wild type was used as control. Worms were cultured at 25 °C on solid nematode growth medium (NGM: 50 mM NaCl, 2.5 g/l peptone, 17 g/l agar; 1 mM CaCl₂, 1 mM MgSO₄, 5 µg/ml cholesterol in ethanol) and seeded with OP50 E. coli as food source, according to the standard procedure (59).

Worms age synchronization

To generate an age-synchronized population, a small plate (3 ml) of nematodes was transferred onto a new large plate (25 ml) to obtain many eggs. After 2 d, the population was collected in 2.5 ml of M9 buffer (42 mM Na₂HPO₄, 22 mM KH₂PO₄, 86 mM NaCl, 1 mM MgSO₄) and an equal volume of 4% glutaraldehyde was added. After 4 h of incubation at 4 °C, the suspension was centrifuged 5 min at 1500 g. The eggs were washed twice in M9 buffer and plated in the half uninoculated sector of a moon large plate (plate seeded with E. coli OP50 only in half section of plate) (58). Fluorescent ATX3Q17 and ATX3Q130 worms were selected using SteREO Discovery.V12 (Zeiss, Oberkochen, Germany).

Pharmacological assays

One day-synchronized adult worms were placed onto a new plate in the presence or the absence of 0.1 mM EGCG, EGC or GA. All compounds were added to the E. coli OP50 suspension before seeding the plates. Body bends were recorded for 20 s after 24 and 48 h of treatment. For each treatment, at least 20 worms were used.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank Alessandra Gliozzi and Mauro Robello for helpful discussion.

Conflict of Interest statement. None declared.

Funding

University of Milano-Bicocca (Fondo di Ateneo per la Ricerca), University of Genoa (PRA 2015 and FRA 2016) and Fondazione Cariplo for funding project 2015-0763.

Author notes

†

Present address: SYSBIO.IT, Centre of Systems Biology, 20126 Milan, Italy.