Calcitonin native prefibrillar oligomers but not monomers induce membrane damage that triggers NMDA-mediated Ca2+-influx, LTP impairment and neurotoxicity

Amyloid protein misfolding results in a self-assembling aggregation process, characterized by the formation of typical aggregates. The attention is focused on pre-fibrillar oligomers (PFOs), formed in the early stages and supposed to be neurotoxic. PFOs structure may change due to their instability and different experimental protocols. Consequently, it is difficult to ascertain which aggregation species are actually neurotoxic. We used salmon Calcitonin (sCT) as an amyloid model whose slow aggregation rate allowed to prepare stable samples without photochemical cross-linking. Intracellular Ca2+ rise plays a fundamental role in amyloid protein-induced neurodegerations. Two paradigms have been explored: (i) the “membrane permeabilization” due to the formation of amyloid pores or other types of membrane damage; (ii) “receptor-mediated” modulation of Ca2+ channels. In the present paper, we tested the effects of native sCT PFOs- with respect to Monomer-enriched solutions in neurons characterized by an increasing degree of differentiation, in terms of -Ca2+-influx, cellular viability, -Long-Term Potentiation impairment, Post-Synaptic Densities and synaptophysin expression. Results indicated that PFOs-, but not Monomer-enriched solutions, induced abnormal -Ca2+-influx, which could only in part be ascribed to NMDAR activation. Thus, we propose an innovative neurotoxicity mechanism for amyloid proteins where “membrane permeabilization” and “receptor-mediated” paradigms coexist.

entry of small molecules and ions, such as Ca 2+ , can reflects in acute excitotoxicity 44 . It's now accepted that lipid-rafts play a pivotal role in the interaction between amyloid oligomers and membranes, as reported in our paper concerning the binding of sCT to a Langmuir membrane model 13 . More in particular, GM1 seems to be a specific target of amyloids. Hong et al. reported that Aβ dimers strongly reduced Long-Term Potentiation (LTP) in mouse hippocampal slices, and this impairment was due to the interaction with membranes mediated by GM1. They also showed that a pre-treatment with Cholera Toxin Subunit β (CTβ), which masks the two terminal sugars of GM1, interfered with this binding and abolishes the LTP depression 24 . (ii) The "receptor-mediated" hypothesis, where amyloid oligomer modulation of Ca 2+ channels such as NMDAR, AMPAR or voltage-dependent (V-dependent) channels, leads to an abnormal intracellular Ca 2+ concentration 5,45 . Several groups reported on the relationship between Aβ oligomers and NMDA excitotoxicity 20,46 . Peters et al. demonstrated that Aβ-evoked ion currents in hippocampal neurons are proportional to neuronal NMDAR expression 47 . It has been also reported that amyloid oligomers can elevate intracellular Ca 2+ , through a signalling pathway depending upon specific Ca 2+ /calmodulin protein phosphatases present at PSD 45,48 . Other groups demonstrated that, in hippocampal neurons, Aβ oligomers activate metabotropic glutamate receptors-5 (mGluR5), resulting in the activation of JNK, Cdk5 and p38 MAPK pathways 31,45,49,50 . Moreover, it has been also suggested that LTP inhibition would be a consequence of Aβ-oligomers interference with glutamate synaptic reuptake mechanism 30 . Notably, it is well known that the Ca 2+ channel occurrence is correlated to the lipid-rafts and then to the GM1 expression.
In the present paper, we used sCT as a tool to investigate the molecular mechanisms of amyloid neurotoxicity in neuronal systems characterized by increasing occurrence of lipid-rafts, marked by GM1. Firstly, in neuronal cultures characterized by increasing degree of differentiation (HT 22, HT-22 DIFF, HpC 6 DIV, HpC 14 DIV) we tested the effects of native sCT PFOs-with respect to Monomer-enriched solutions, in terms of intracellular Ca 2+ -influx, reduced cellular viability and PSD-95 expression. Finally, we assessed LTP impairment in mouse hippocampal slices.
Our results show that native metastable sCT PFOs induced biological effects whose intensity increased with the differentiation degree of neuronal cultures, while Monomers were always ineffective. Notably, we propose for the first time to our knowledge, a possible mechanism of action where both "membrane permeabilization" and "receptor-mediated" paradigms could contribute to explain the experimentally observed intracellular Ca 2+ -influx. We speculate that in mature neurons PFOs, but not Monomers, may induce an initial membrane damage that triggers a subsequent abnormal NMDA-mediated Ca 2+ -influx, leading to the observed LTP impairment and neurotoxicity.

Results
Sample preparation and characterization. Native sCT solution was prepared by incubating 2 mg of sCT powder in phosphate buffer (5 mM) at room temperature overnight. In a previous paper 3 , we showed that, under these conditions, the formation of PFOs was favoured with respect to MFs. In order to purify oligomeric species enriched fractions, the aggregated and native solution was then loaded in the SEC column (detailed procedures of SEC are reported by Diociauti et al. 3 ).
The SEC profile of sCT spontaneously aggregated mixture is depicted in Fig. 1a. The absorbance at 280 nm, along SEC fractions, is characterized by a peak located around 29 th fraction (V e = 24.65 mL), followed by a prolonged tail. According to calibration standards elution profile (inset of Fig. 1a), the peak (6450 Da) is equivalent to a dimer.
The tail on the right is due to Monomers that are uniformly distributed in several fractions at a very low concentration. For this reason, we decided to prepare Monomer-enriched sCT solution by dissolving the lyophilized powder in desalted water just before treatments, at concentration of 80 μM, similar to the concentration of the 29 th fraction. This because it is well known that the aggregation dynamics depends from ionic strength and concentration.
A characterization of the PFOs fractions, from the SEC peak and Monomers native samples, has been obtained after PICUP procedure followed by tricine SDS-PAGE (Fig. 1b). In the case of PFOs (fractions 29 th and 30 th ), the gel analysis clearly demonstrated an upwards-shifted pattern as compared to Monomers sample, where dimers and trimers were also present. A strong enhancement of the aggregated species, such as tetramers, pentamers and hexamers, was observed. To better analyse this trend, we calculated two indexes: (i) the Average Molecular Weight index (AMWi); (ii) trimers + tetramers + pentamers + hexamers percentage (PFOs%) for PFOs-enriched fractions and Monomer unfractioned samples (for details see Fig. S1). Results relative to the 29 th and 30 th fractions (V e = 24.65 and 25.50 mL), highlighted in Fig. 1d, gave AMWi = 10.0 KDa for sCT PFOs, which is close to trimer molecular weights (10.2 KDa), while AMWi = 5.8 KDa for native unfractioned Monomers, which is between Monomer and dimer molecular weights (3.4 KDa and 6.8 KDa, respectively). The percentages of trimers, tetramers, pentamers and hexamers rises from about 16.5% for Monomer-to 54.0% for PFOs-enriched samples. We noticed that, even if monomeric species were overexpressed in the Monomer-enriched samples, dimers and trimers were always present (Fig. 1b).
We want to stress that, using sCT we did not obtain aggregates of molecular weight higher than hexamers. As reported earlier, unlike others members of the amyloid family, sCT presents a very low aggregation rate with the longest aggregation lag-phase among CT variants 4 . This rises the possibility to produce stable PFOs-enriched samples made of known aggregates to be tested directly onto cells. Figure 1c shows representative Energy Filtered-Transmission Electron Microscopy (EF-TEM) micrographs relative to the central fraction (29 th fraction) of the PFOs peak, together with images relative to the Monomer-enriched sample. Notably, PFOs appeared as particles of 10 ± 3 nm, not perfectly spherical and www.nature.com/scientificreports www.nature.com/scientificreports/ sometimes characterized by well-defined or incomplete hexagonal shape. Few bigger particles were also present. Conversely, image of Monomer-enriched sample revealed the presence of very small dots of about 3 ± 1 nm, likely Monomers, together with clusters of few dots, dimers and trimers (means of maximum diameters of more than 30 particles per sample type).

PFOs Administration to Cellular Cultures
Intracellular Ca 2+ -influx. Here (Fig. 2a)  In a previous paper, we reported that the Ca 2+ -influx was evoked by not-fractioned mixture of sCT oligomers in 14 DIV hippocampal neurons, which was abolished by disrupting GM1 sialic acid heads by neuraminidase 14 . Thus, we hypothesize that growing levels of this ganglioside would result in the [Ca 2+ ] i increase.
To investigate if the Ca 2+ -influx depends on the GM1 content, we compared undifferentiated HT22 cells, differentiated HT22 cells, primary hippocampal neuronal cultures from rat hippocampi at 6 DIV and 14 DIV. In fact, it is well known that the GM1 content increases with neuronal differentiation degree. We verified this fact in our cells by immunofluorescence (Fig. 2d).
Results are reported in Fig. 2b. Significant differences were obtained between all groups. As it can be observed, shape and intensity of the curves changed with the differentiation degree. The high and sustained Ca 2+ -influx of 14 DIV hippocampal neurons was reduced for 6 DIV neurons. The shape was maintained in the first minutes but the plateau reached in the second part was more than halved. HT22 DIFF cells gave rise to less intense signals, with a shape very similar to that of 6 DIV neurons. Finally, we note that undifferentiated HT22 cells showed the smallest Ca 2+ -influx, vanishing after about 6 minutes.
Summarizing, we noted a strong correlation between [Ca 2+ ] i signal maintained after 10 ninutes and cell differentiation degree (inset of Fig. 2b). www.nature.com/scientificreports www.nature.com/scientificreports/ To investigate the nature of the Ca 2+ -influx we pre-treated 14 DIV neurons with MK801, a specific blocker of NMDAR and with Verapamil, a blocker of V-dependent channels. Results (Fig. 2c) indicated that, after treatment with MK801, the [Ca 2+ ] i increase was reduced to a small and transitory peak and this observation suggests that the strong and maintained Ca 2+ -influx of the 14 DIV neurons was due to the NMDARs. Conversely, we found a slight dependence on voltage-activated Ca 2+ channels, since their specific blocker (Verapamil) determined only a small decrease of the PFOs-evoked response (Fig. 2c).
We want to stress the similarity in shape and intensity between the peak obtained with undifferentiated HT22 cells (Fig. 2b) and the one observed with 14 DIV neurons in the presence of MK801 (Fig. 2c) where NMDARs were inhibited. This similarity suggests that a mechanism different from NMDAR activation was in part responsible for [Ca 2+ ] i increase. Notably, as pointed out by He et al., these channels were totally missing in the HT22 cells 51 . In order to verify this important issue, we treated both HT22 and HT22 DIFF with NMDA, monitoring the induced Ca 2+ -influx. Results (Fig. S2) clearly showed that Ca 2+ -influx was totally absent in HT22 cells, confirming the lack of NMDARs and demonstrating that the small peak induced by sCT PFOs must be due to a different mechanism.
NMDA induced a gradual raising of [Ca 2+ ] i in HT22 DIFF, demonstrating that, as reported by He et al., the differentiation process leads to the expression of the NMDARs 51 . However, the shape of the Ca 2+ -influx induced by NMDA in HT22 DIFF (Fig. S2) was very different from that induced by sCT PFOs, lacking the transitory features occurring in the first minutes and characterized by a plateau value reached after 10 minutes, of the same intensity of the plateau reached with sCT PFOs (Fig. 2b). This suggests that NMDARs were responsible for the Ca 2+ -influx obtained at the steady state, in agreement with the conclusion drawn in the case of 14 DIV neurons pre-treated with MK801. Finally, to investigate the possible correlation between Ca 2+ -influx and GM1 expression, we examined by fluorescence microscopy the membrane distribution of GM1 in all cell types, by using fluorescence-labelled CTβ, known to selectively bind the GM1 sialic acid heads. In HT22 cells, images showed the characteristic dotted distribution and the increase of GM1 expression due to differentiation (Fig. 2d, upper panel) while, in hippocampal cultures, GM1 was evenly distributed throughout the plasma membranes (Fig. 2d, lower panel) and the GM1 content further increased upon differentiation (lower panel). We conclude that a correlation exists between Ca 2+ -influx and GM1 expression. www.nature.com/scientificreports www.nature.com/scientificreports/ Summarizing, we provide evidences that sCT PFOs evoked strong and sustained intracellular Ca 2+ -influx mediated by NMDARs and that this behaviour was correlated with the GM1 expression. In the absence of NMDARs or with these channels blocked, and with few GM1 (undifferentiated HT22) a small but detectable Ca 2+ -influx was still observed.

Cell viability.
To clarify if neurotoxicity is correlated to the aggregation state of sCT, we studied the dose-response relationship of PFOs-and Monomers-enriched solutions in HT22 DIFF cells, after 24 hours. Results clearly showed that only PFOs-enriched solutions induced cell viability reduction (about 10%) while Monomers-enriched solutions were totally harmless (Fig. 3). Notably, the same result was obtained with 14 DIV hippocampal neurons (Fig. 4a). Based on these results we decided to perform all viability esperiments at 8 μM concentration.
We noted that cell viability reduction depended from the cell differentiation degree. In fact, as reported Fig. 4a, it was higher in hippocampal neurons (about 25%) with respect to HT22 DIFF. Notably, undifferentiated HT22 cells didn't reach a significant toxicity (Fig. 4b). Finally, we noted that the loss of viability was not time-dependent, since after 10, 100 minutes and 24 hours of treatment, the same reduction was observed in 14 DIV hippocampal neurons (Fig. 4c).
Thus, we investigated if a correlation exists between the [Ca 2+ ] i increase and cell viability. In good agreement with the Ca 2+ -influx experiments, MTT results clearly showed a direct relationship between cell differentiation and/or GM1 expression and PFOs induced neurotoxicity.
As in the case of Ca 2+ -influx, before PFOs administration we pre-treated our cell cultures with MK801 and Verapamil, which block Ca 2+ channels. In agreement with Ca 2+ -influx results, we found that these blockers fully reverted neurotoxicity. This indicates a direct relationship between neurotoxicity and levels of [Ca 2+ ] i reached after 10 minutes and more.
In order to confirm that the reduction of cell viability observed with MTT assay was mainly due to the reduction of neuronal cells instead of glia, we performed immunofluorescence analysis and TUNEL assay of PFOs treated hippocampal cultures. Results (Fig. S3) showed a reduced expression of NeuN positive cells, indicating that the damaged cells were neurons. Moreover, the colocalization of NeuN and TUNEL assay confirmed that PFOs treatment induced apoptosis in neurons.

Acute PFOs Administration to Brain Slices
Synaptic plasticity. Over the last decades, the link between oligomers neurotoxicity and synaptic plasticity impairments has been investigated for Aβ 6,30,31,45,48,50,[52][53][54][55][56] . As well known, Aβ is an unstable protein and, in agreement with Benilova et al., sample changed along different preparation making interpretation and direct comparison of data between different research groups very difficult 5 . At the present, any attempt to test the effects of amyloid native PFOs and Monomers stable during the experiment on synaptic plasticity hasn't been obtained yet. Herein, we investigated the effects of native samples, as intended in the introduction 3,4 , on synaptic plasticity.
PFOs and Monomers were diluted in carboxygenate Artificial CerebroSpinal Fluid (ACSF) at a final concentration of about 3 μΜ and used to superfuse mouse hippocampal slices.
sCT PFOs-enriched samples fully abrogated LTP in hippocampal slices, 80 minutes after the tetanus, while native Monomer-enriched solutions did not affect LTP even when compared to control (Fig. 5). Interestingly, LTP reduction in the case of sCT PFOs was very similar to that reported for Aβ 6 . However, for the first time to our knowledge, we report significant differences between effects induced by sCT native amyloid PFOs-and Monomer-enriched solutions.
Western blotting and Immunofluorescence of synaptic proteins. In order to evaluate if the altered activity of LTP was correlated with an impairment of synaptic structures 57,58 and loss of dendritic spines 59 , we studied the expression of a pivotal post-synaptic protein such as PSD-95 and the expression of the pre-synaptic synaptophysin, a component of synaptic vesicles.
In hippocampal slices incubated with sCT PFOs, WB analysis (Fig. 6a) showed a significant decrease in the expression of PSD-95 with respect to untreated slices (Fig. 6b). Under the same conditions, the levels of synaptophysin remained unmodified (Fig. 6c). www.nature.com/scientificreports www.nature.com/scientificreports/ Notably, in agreement with LTP results, our findings showed that the treatment with sCT Monomer-enriched solutions did not modify the levels of these proteins (Fig. 6b,c). The analysis of immunofluorescence confirmed the reduced levels of PSD-95 after treatment with sCT PFOs and the substantial invariance of synaptophysin (Fig. 6d).

Discussion
Our experimental results demonstrated that sCT was a useful tool to prepare native PFOs-or Monomer-enriched samples to be tested in vitro without the need of cross-linking procedure or to apply more sophisticated approaches 9 . Due to its very low aggregation rate, PFOs samples were stable during experiments and rich in aggregates not greater than hexamers. These properties allowed us to overcome difficulties highlighted by Benilova et al. in the identification of the amyloid structures responsible for neurotoxicity 5 . Moreover, the sCT PFOs morphology and, more important, the biologic effects they induced were very similar to those reported for toxic Aβ oligomers 3,5 .
We demonstrated that native soluble sCT PFOs-rich samples, composed of tetramers, pentamers and hexamers but not species of higher molecular weights, were able to induce sustained Ca 2+ -influx in mature mouse primary hippocampal neurons (14 DIV) (Fig. 2a). Interestingly, when NMDARs were blocked by MK801 a small and transitory but detectable Ca 2+ -influx was still observed (Fig. 2c). Notably, in a previous work we observed the same influx by treating with sCT oligomers rat mature primary neurons pre-treated with MK801 (see Fig. 7  www.nature.com/scientificreports www.nature.com/scientificreports/ therein) 14 . This suggests that the main part of the observed Ca 2+ -influx in mature neurons was due to NMDAR activation 35 but that a residual NMDAR-independent component exists.
In our opinion, this component could be ascribed to the formation of amyloid pores. We base this hypothesis on our data relative to undifferentiated HT22 cells (Fig. 2b). As shown by He et al., these cells do not express cholinergic and glutamatergic receptors, and in particular NMDAR 51 . We carefully checked this fact by treating cells with NMDA and we did not observe any Ca 2+ -influx (Fig. S2). PFOs were able to induce a Ca 2+ -influx transitory peak even in undifferentiated HT22, where NMDAR was absent. Thus, we hypothesize that the influx can be due to the formation of amyloid pores.
We note that several groups proposed that the presence of GM1 is necessary for pores formation 13,54-56 . Here we report that undifferentiated HT22 cells expressed low but detectable levels of GM1 (Fig. 2d). Thus, we hypothesize that in these cells few amyloid pores can be formed, with the result to create the observed weak and transitory Ca 2+ -influx, insufficient to affect cells viability (Fig. 4c). Notably, we observed the same peak in 14 DIV primary neurons pre-treated with MK801, where NMDARs were blocked (Fig. 4d).
It is interesting to note what happened working with HT22 DIFF cells that express a higher level of GM1 (Fig. 2d) and, at the same time, express glutamatergic receptors 51 (Fig. S2). The Ca 2+ -influx was more than doubled in intensity and remained sustained beyond 10 minutes (Fig. 2b) and was sufficient to induce a significant (about 10%) cell viability decrease (Fig. 4b).
We speculate that the formation of more amyloid pores, determined by the higher GM1 content, induced a stronger Ca 2+ -influx leading to the activation of the NMDARs. This phenomenon can lead, in our opinion, to the observed cell death through mitochondrial dysfunction and oxidative damage, as proposed by Angelova and Abramov 37 .
In good agreement, when primary neurons (6 DIV, 14 DIV) were studied, the high GM1 content (Fig. 4b) leads to the formation of many amyloid pores, enough to boost up the abnormal activation of the NMDARs described before in differentiated HT22 cells. In 6 DIV and more in 14 DIV, the Ca 2+ -influx curves were very similar to that observed for HT22 DIFF but more than doubled (Fig. 2b) and a cell viability reduction of about 25% (Fig. 4b).
The crucial role played by GM1 in the formation of amyloid pores by PFOs, has been recently highlighted by Hong et al. for Aβ 24 . Furthermore, in a previous paper we showed that masking GM1 in rat primary hippocampal neurons totally prevents Ca 2+ -influx and toxicity induced by sCT oligomers 14 .
However, the results we are presenting now clearly indicate that the formation of amyloid pores alone is not enough to explain neurotoxixcity and that NMDARs must be involved, as suggested by the protection exerted by MK801. Finally, we note that none of the effects described before was induced by sCT Monomer-enriched www.nature.com/scientificreports www.nature.com/scientificreports/ samples, containing a minority of dimers and trimers. As proposed by Angelova and Abramov, this can be due to the inability of Monomers to form stable amyloid pores and to switch on the pathological Ca 2+ -influx and the consequent neurotoxicity 37 .
For what concerns LTP experiments, we showed that sCT PFOs abrogated synaptic plasticity after 100 minutes of treatment while, in good agreement with neurotoxicity results described before, Monomer-enriched solutions were totally ineffective (Fig. 5). Similar results have been reported for aggregates of others amyloid proteins 5,6,30,50,53 . However, in our knowledge, the direct comparison between the effects induced by PFOs and Monomers has been never investigated for sCT.
Interestingly, we provide evidences that cells were damaged after 10 minutes as after 24 hours of PFOs treatment (Fig. 4c). This suggests that, in the LTP experiments, also neurons can be damaged after 10 minutes of PFOs treatment. However, neurons were still alive since responded to the tetanic stimulation producing population spikes. Angelova and Abramov have shown that ROS production in primary neuronal culture is significantly higher after 8-10 minutes of oligomer exposure, compared to Monomer treatment 37 .
Thus, we interpret our LTP observations as a consequence of the Ca 2+ -influx mechanism proposed before. In our hypothesis, once Ca 2+ ions were allowed to flow into the cells, the following membrane depolarization with the consequent abnormal Ca 2+ conductance, impaired synaptic transmission and produced vesicle depletion resulting in synapse silencing 31 . This interpretation is supported by our results relative to the PSD-95 expression in brain slices used in the LTP experiments. Notably, we found reduced levels of PSD-95 and the invariance of synaptophysin expression, induced by sCT PFOs-but not by Monomer-enriched solutions (Fig. 6a,b).
According to the scenario depicted before, even if data reported here are not sufficient to demonstrate the formation of amyloid pores in a definitive manner, we hypothesize (Fig. 7) that their formation could represent the initial phenomenon able to trigger the observed subsequent NMDA-mediated abnormal Ca 2+ -influx, leading to the neuronal impairment and damage.
Thus, if a common amyloid-related excitotoxicity mechanism exists, as proposed by Glabe 2 , our results can be important in the design of novel therapeutic approaches that should follow three main routes: i) to avoid the PFOs formation before they might interact with neuronal membranes; ii) to avoid the binding between amyloid

Photo-induced Cross-linking of Unmodified Proteins (PICUP) and tricine SDS-PAGE Gel characterization.
Aiming to obtain an electrophoretic characterization (tricine/SDS-PAGE) of the fraction eluted, samples were stabilized in order to prevent oligomers misfolding in a reducing environment. We therefore treated them for PICUP. The original protocol was adapted by Diociauti et al. 3 . Briefly, for each sample, we prepared a 20 μl volume containing 80 μM Oligomers, 50 μM Tris (2,2 bipyridyl) dichlororuthenium (II) hexahydrate) and 1 mM ammonium persulfate (SIGMA). Cross-linking reaction occurred irradiating samples for 2 s with a 100 W white lamp in a dark room, reaction was quickly quenched adding 20 μl of reducing sample buffer containing 5% β-MeOH and boiled for 5 minutes. Samples were finally analysed by tricine/SDS-PAGE. Separating gel: 10% Acrylamide/Bis (32:1) (ICN Biomedicals, Inc., Aurora, Ohio USA/Fluka, Buchs, CH); spacer gel 6.5% Acrylamide/Bis and stacking gel: 2.5% Acrylamide/Bis. For each lane were run 40 μl of the sample buffer, or 8 μl of molecular weight markers (Color Marker Ultralow Range -SIGMA, cat n° C6210-1VL). Then gels were stained by silver procedure. Finally, gel band densitometry has been performed using the imaging freeware ImageJ. For each lane, grey-scale optical density profile has been obtained, including molecular weight markers (1.3 KDa, 3.5 KDa, 6.5 KDa, 14.4 KDa, 17.0 KDa, 26.6 KDa). Peaks corresponding to sCT prefibrillar oligomers: Monomers, dimers, trimers-tetramers, pentamers and hexamers has been individuated according to the molecular weight marker profile. Gel densitometry was used to compute two estimators, named PFOs % indexand Average MW index, taking in account for the oligomers species that prevalently populated the fractions. The values were used to evaluate statistical differences between sCT monomeric and aggregated samples. www.nature.com/scientificreports www.nature.com/scientificreports/ Energy Filtered-Transmission Electron Microscopy (EF-TEM). We used a Transmission Electron Microscope model TECNAI 12 G2 Twin (FEI Company, Hillsboro, OR, USA) equipped with a thermionic gun (single-crystal lanthanum hexaboride) and energy dispersive X-ray spectrometer (model Genesis 4000, EDAX Inc., Mahwah, NY, USA) and post-column electron energy filter (Bio filter, GATAN Inc., Pleasanton, CA, USA). The energy filtered images were acquired by the use of a slow-scan CCD camera (model 794 IF, GATAN Inc., Pleasanton, CA, USA). Conventional imaging was performed in energy-filtered image mode configuration at electron energy of 120 keV, with a collection angle of about 20 mrad. To enhance image contrast and resolution, chromatic aberrations were reduced by collecting only elastic electrons (ΔE = 0) 60 .
Cell cultures. HT22 cells were developed from their analogous HT4 cells, immortalized from primary mouse hippocampal neurons. If grown without establishing synaptic connections, both HT22 and HT4 hippocampal cells are able to develop LTP, in terms of neurotransmitter release 29 . HT22 neuronal cell line in proliferating conditions, doesn't express cholinergic and glutamatergic receptors, although HT22 cells can also be differentiated in a selective medium, changing their morphology and inducing expression of cholinergic markers like: choline acetyl transferase (ChAT), vesicular acetylcholine transporter (VAChT), high affinity choline transporter (HACT), muscarinic M1 and M2 subunit of Ach-receptors. Moreover, they become susceptible to glutamate excitotoxicity 51,61 _ENREF_3 that is mainly mediated by NMDARs 62 via Ca 2+ -influx 63 , but also by several second messengers like nitric oxide 64 , calpain-1/poly-(ADPribose) polymerase1/apoptosis inducing factor 65 , free radicals and mitochondria 66 . NMDAR antagonists, such as Dizocilpine (MK801), can effectively prevent glutamate-induced excitotoxicity 51 .
Primary hippocampal co-cultures (neurons and glia) have been prepared from postnatal day 2-mouse brain. All experimental procedures were carried out according to the Italian law and to "Ethical guidelines for scientific experimentation on animals". Experimental protocol was approved by the Italian Ministry of Public Health (authorization n. 86/2018-PR), and was in accordance with guidelines of the European Union Council Directive (86/609/European Economic Community). After dissection, hippocampi were incubated 15 minutes at 37 °C, with 0.25% trypsin (Gibco, 15090-046) and then dissociated in NeuroBasal medium (Gibco, 21103-49) containing 10% heat-inactivated FBS, 50 µg/mL gentamicin (Gibco, 15750-037), 1x Glutamax (Gibco, 35050-038). Cells were seeded on 48-wells plates for MTT assay or on sterile glass coverslips (diameter 12 mm), previously coated with 1x poly-L-lysine (Sigma-Adrich), in 24-wells plates, for Fura2-AM experiments or GM1 evaluation. After plating cell cultures were rapidly stored in the cell incubator (5% CO 2 , 37 °C). After 2 hours, cell culture medium was replaced with 700 μl/well of Neurobasal medium containing 1X B27 supplement (Gibco, 17504-044) instead of FBS. Hippocampal co-cultures were used at 6 DIV (not fully mature culture) or 14 DIV (fully mature culture). . Cell-seeded coverslips were incubated with Fura-2AM working solution for 50 minutes at 37 °C, 5% CO 2 in darkness. After three washes in Ringer solution, coverslips were rapidly placed in the cell bath with fresh Ringer solution on the microscope stage, for Ca 2+ imaging recordings. All recordings were performed in dark conditions. After 2 minutes of adaptation (baseline), samples from different native sCTs fractions were used to treat the cells, at a final concentration of 8 µM. Fluorimetric recordings with Fura-2AM were performed along the experiment to obtain ratiometric measurements of the [Ca 2+ ] i that is the ratio between the emission intensities measured at 510 nm, stimulated at 340 nm and 380 nm 67 . One acquisition was done each 6 second, along 15 minutes, on a region of interest in the cellular bodies. Ratios and 340 nm and 380 nm background signals, from time laps sequence images were obtained by the imaging freeware ImageJ. Data were plotted by the software: Microcal Origin 8.

Fura-2AM Ca
MTT and TUNEL assay in cell cultures. Cell viability was evaluated by the 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT) assay; the MTT assay has been widely used to assess cell viability and is based on the ability of viable cells to reduce MTT, giving rise to an insoluble purple formazan salt. Briefly, the cultures were incubated for 20 minutes at 37 °C with 0.5 mg/ml MTT in Hank's balanced salt solution (Life Technologies). The reaction product was dissolved in dimethyl sulphoxide. The spectral photometric absorbance of the samples was determined at a wavelength of 540 nm. The amount of MTT conversion was evaluated as a percentage of the absorbance measured in treated cells relative to the absorbance of control cells. After fixation in 4% paraformaldehyde in PBS, 0.12 M in sucrose, apoptosis was evaluated in mixed hippocampal cultures by the terminal transferase-mediated dUTP-biotin nick end-labeling (TUNEL) assay (DeadEnd kit, Promega, Madison, WI).

Long-Term Potentiation (LTP) in mouse hippocampal slices.
Wild type BALB/c mice aged 6 to 9 weeks were used in accordance with guidelines and regulations of the European Union Council Directive (86/609/ European Economic Community). All the experimental protocols were approved by the Italian Ministry of Public Health (authorization n. 86/2018-PR). Under anesthesia with halothane (2-Brom-2-chlor-1,1,1-trifluor-ethan), the animals were decapitated and brains were quickly removed and placed in cold, oxygenated artificial cerebral spinal fluid (ACSF), whose composition in mM was: NaCl 124, KCl 2, KH 2 PO 4 1,25, MgSO 4 2, CaCl 2 2, NaHCO 3 (2019) 9:5144 | https://doi.org/10.1038/s41598-019-41462-0 www.nature.com/scientificreports www.nature.com/scientificreports/ 26, and Glucose 10. The hippocampal slices were prepared according to conventional procedures 68 . The hippocampus was rapidly dissected and slices (450 µm thick) were cut transversely by a chopper (McIlwain Tissue Chopper) and transferred into an interface tissue chamber constantly perfused by a flow of 1.2 mL/min of ACSF and humidified gas (95% O 2 -5% CO 2 ) at 32-34 °C (pH 7.4). According to the original protocol 69 , extracellular recordings of the population spikes (PSs) were made in the stratum pyramidale of the CA1 subfield using glass microelectrodes filled with 2 M NaCl (resistance 5-10 MΩ). Orthodromic stimuli (10-500 mA, 20-90 ms, 0,1 Hz) were delivered through a platinum electrode placed in the Schaffer collateral commissural pathways in the stratum radiatum. The test stimulus intensity of 50 ms square pulses was adjusted to elicit a PS of 2-3 mV at 0,03 Hz. Each minute, a trace was calculated as the average of six recordings every 10 seconds. After recording stable signals (20-30 minutes), the hippocampal slices were treated with Monomer-or PFOs-enriched solutions of sCT, in order to assay their effects on synaptic plasticity. PFOs and Monomers were diluted in carboxygenate ACSF at a final concentration of about 2 μM: the slices where then superfused.
After 20 minutes from administration of sCT, a tetanic stimulation (100 Hz, 1 s) was delivered to induce LTP at the same stimulus intensity used for the baseline responses. Field potentials were fed to a computer interface (Digidata 1440 A, Axon Instruments, Foster City, CA) for subsequent analysis with the software PCLAMP10 (Axon Instruments).
Western blot (WB) analysis. Hippocampal 450 μm-thick transversely coronal slices were homogenized in extraction buffer (25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 1 mM PMSF, and a protease inhibitor cocktail) on ice for 30 minutes and centrifuged at 100.000 × g for 1 h, at 4 °C. The protein concentration was determined using the Micro BCA Protein Assay Kit (Pierce, Rockford, IL, USA). Proteins (30 µg) were separated on 12% SDS-PAGE and transferred to nitrocellulose membranes at 35 V overnight. The membranes were blocked at room temperature in 3% BSA and incubated overnight at 4 °C with the following primary antibodies: mouse monoclonal anti-PSD-95, rabbit polyclonal anti-synaptophysin (home-made) and, as a control for protein loading, mouse monoclonal anti-β-actin (Santa Cruz). The membranes were washed and incubated with the appropriate peroxidase-labelled secondary antibody (Bio-Rad, Hercules, CA, USA) for 1 h at room temperature. After extensive washes in TTBS (20 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 0.1% Tween 20), the immunoreactive bands were detected by enhanced chemiluminescence coupled to peroxidase activity (Santa Cruz Biotech) and imaged with a ChemiDoc XRS system (Bio-Rad Laboratories Inc.).
Immunofluorescence. Hippocampal 450 μm-thick transversely coronal slices were treated or not with sCT PFOs-or Monomer-enriched solutions for 100 min. After treatment, the slices were fixed over night with 4% paraformaldehyde in PBS, 0.12 M in sucrose. After fixation, the samples were rinsed three times in PBS with 5% sucrose and 0.15 mM CaCl 2 and left overnight in sucrose buffer (PBS with 30% sucrose and 0.15 mM CaCl 2 ). Samples were then embedded in Tissue Freezing Medium (Jung, Germany), frozen at −30 °C in isopentane and stored at −80 °C. Sections (8 μm) were cut at a Leica CM 1860 UV cryostat and labelled with primary antibodies overnight at 4 °C. The following primary antibodies were used: monoclonal anti-NeuN (Millipore, USA) rabbit anti-PSD-95 (Cell Signaling Technology, Danvers, MA) and monoclonal anti-synaptophysin (BD Transduction Laboratories, Franklin Lakes, NJ). Primary antibodies were revealed with secondary antibodies coupled to Alexa Fluor ® 488 and Alexa Fluor ® 546 (Invitrogen), diluted 1:250 PBS (45 min, 37 °C). Sections were counterstained with Hoechst 33258; the dye, which binds specifically to A-T base regions in DNA and emits blue immunofluorescence at 350 nm, was administered at 1 ng/ml for 1 minute. Sections were observed at an Eclipse 80i Nikon Fluorescence Microscope (Nikon Instruments, Amsterdam, The Netherlands), equipped with a Video Confocal (ViCo) system.

GM1 expression.
To analyse GM1 localization in HT22 cells and hippocampal neurons we used Alexa Fluor 488-conjugated cholera toxin-β (CT β, Molecular Probes, Eugene, OR) (10 μg/mL). After fixation in 4% paraformaldehyde in PBS, 0.12 M in sucrose, cultures were stained with CTβ at room temperature for 30 minutes. To highlight nuclei, cultures were counterstained with Hoechst 33258. Cultures were observed at an Eclipse 80i Nikon Fluorescence Microscope (Nikon Instruments, Amsterdam, The Netherlands), equipped with a Video Confocal (ViCo) system.