Wnt Signaling Prevents the Aβ Oligomer-Induced Mitochondrial Permeability Transition Pore Opening Preserving Mitochondrial Structure in Hippocampal Neurons

Alzheimer’s disease (AD) is a neurodegenerative disorder mainly known for synaptic impairment and neuronal cell loss, affecting memory processes. Beside these damages, mitochondria have been implicated in the pathogenesis of AD through the induction of the mitochondrial permeability transition pore (mPTP). The mPTP is a non-selective pore that is formed under apoptotic conditions, disturbing mitochondrial structure and thus, neuronal viability. In AD, Aβ oligomers (Aβos) favor the opening of the pore, activating mitochondria-dependent neuronal cell death cascades. The Wnt signaling activated through the ligand Wnt3a has been described as a neuroprotective signaling pathway against amyloid-β (Aβ) peptide toxicity in AD. However, the mechanisms by which Wnt signaling prevents Aβos-induced neuronal cell death are unclear. We proposed here to study whether Wnt signaling protects neurons earlier than the late damages in the progression of the disease, through the preservation of the mitochondrial structure by the mPTP inhibition. To study specific events related to mitochondrial permeabilization we performed live-cell imaging from primary rat hippocampal neurons, and electron microscopy to analyze the mitochondrial morphology and structure. We report here that Wnt3a prevents an Aβos-induced cascade of mitochondrial events that leads to neuronal cell death. This cascade involves (a) mPTP opening, (b) mitochondrial swelling, (c) mitochondrial membrane potential loss and (d) cytochrome c release, thus leading to neuronal cell death. Furthermore, our results suggest that the activation of the Wnt signaling prevents mPTP opening by two possible mechanisms, which involve the inhibition of mitochondrial GSK-3β and/or the modulation of mitochondrial hexokinase II levels and activity. This study suggests a possible new approach for the treatment of AD from a mitochondrial perspective, and will also open new lines of study in the field of Wnt signaling in neuroprotection.

Introduction Wnt signaling prevents neuronal cell death by protecting the mitochondrial structure and inhibiting mitochondrial permeabilization.

Materials and Methods Animals
Animals were born and maintained at the Animal Facility of the Pontificia Universidad Católica de Chile under sanitary barrier in ventilated racks and in closed colonies. Experimental procedures were approved by the Bioethical and Biosafety Committee of the Faculty of Biological Sciences of the university. Euthanasia were performed using 5-8% isofluorane. Pregnant Sprague-Dawley rats (E18) were used to prepare primary hippocampal neurons in culture as was described previously [24]. For more details, see S1 File.

Formation of amyloid-β oligomers
Synthetic Aβ 1-42 peptides corresponding to wild-type human Aβ were obtained from Genemed Synthesis, Inc. (San Francisco, CA). An Aβ peptide stock solution was prepared by dissolving freeze-dried aliquots of Aβ in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Sigma H-8508) at 1 mM. For oligomer preparation, the peptide film was dissolved in dimethyl sulfoxide (DMSO, Sigma D2650) at 5 mM and was then diluted in PBS to a final concentration of 100 μM. The preparation was incubated overnight to allow Aβos formation and then centrifuged (14,000 rpm, 1 h, 4˚C) to eliminate any formed fibril [28]. An aliquot of Aβos solution is used to quantify protein with the Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA, USA) in order to determine the final concentration (70-80 μM). Aβos were visualized by electron microscopy and analyzed by Tris-Tricine SDS gel electrophoresis, as previously described [29,30].

Aβo treatments in cultured hippocampal neurons and brain slices
For in vitro treatments, freshly prepared Aβos were diluted to 5 μM in Neurobasal medium without supplements and were directly added to the cultured hippocampal neurons, which were chronically stimulated for 24 h to conduct neuronal viability assays. For live-cell imaging experiments, neurons were acutely exposed to 10 μM Aβos for 7 min. Hippocampal slice preparations were treated for 1 h with 5 μM Aβos diluted in artificial cerebrospinal fluid (ACSF). For all treatments, vehicle solution corresponds to a PBS/DMSO solution prepared by using the same volume of DMSO used for the Aβos preparation, and then diluted in Neurobasal or ACSF, as appropriate.

Calcein/cobalt imaging for mPTP opening assay
For in vivo cell imaging, neurons were seeded on 25-mm cover slips at a density of 1.5 x 10 5 cells per cover slip. mPTP opening was measured using the Image-iT ™ LIVE Mitochondrial Transition Pore Assay Kit (I35103) (Molecular Probes, Carlsbad, CA) with modifications, as described in previous studies [31,32]. At 10 days in vitro (DIV), hippocampal neurons were treated with recombinant mouse Wnt3a (300 ng/ml, 24 h) (R&D Systems, Inc., MM, cat n1 324-WN/CF) or Neurobasal control medium and were loaded with the labeling mix solution (4 μM calcein-AM, 50 nM MitoTracker Orange, 1 mM CoCl 2 and 1 μM Hoechst 33342 dye in Neurobasal medium) for 30 min at 37˚C. Cultures were washed with the recording solution, Tyrode's buffer (135 mM NaCl, 5 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, 5.6 mM glucose, pH 7.3) supplemented with 1mM CoCl 2 and imaged on an Olympus DSU IX81 spinning-disk confocal microscope in the same conditions. The excitation/emission (ex/em) peaks of calcein after hydrolysis occur at 494/517 nm. The cytoplasmic signal of calcein is quenched by cobalt ions, which do not affect the mitochondrial signal. As a result of stimulating mPTP opening, including the permeabilization of the IMM, calcein is released from the mitochondrial matrix, resulting in the redistribution and rapid decay of mitochondrial calcein fluorescence [32]. After the basal signals were measured, 10 μM Aβos was added at 3 min, and fluorescence was recorded for 10 min. Hoechst staining (λex/em, 350/461 nm) was not used during the experiment but was used to determine the viability of the neurons before and after Aβos exposure. The mitochondrial pattern of calcein staining was verified using MitoTracker as a mitochondrial marker, and mitochondria that were initially positive for both labels were used for the analysis. The ionophore ionomycin (0.5 μM), which induces Ca 2+ overload, was used as a positive control for mPTP induction. Pre-incubation of neurons with cyclosporin A (CsA, 20 μM) for 30 min was used to inhibit pore opening [33,34]. Wnt inhibitors were used as follow: sFRP2 (250 nM) was pre-incubated with the recombinant Wnt3a for 30 min and then applied together to the neurons for 24 h; DKK1 (100 ng/ml) was added for 30 min to the neurons and then co-incubated with the Wnt3a ligand for 24 h; ICG001 (20μM) was co-incubated with Wnt3a for 24 h; and 6-BIO (10nM) was used in the same conditions as Wnt3a, for 24 h. For calcium-induced mitochondrial swelling experiments, cells were treated with control or recombinant Wnt3a, as described above, and washed twice with Neurobasal medium. In order to measure the mPTP opening, neurons were subsequently loaded with calcein/Co 2+ as was explained before. The measurement of mitochondrial buffering ability after calcium application was done permeabilizing plasma membrane of the hippocampal neurons with 9 μM digitonin for 2 min in an intracellular solution (130 mM K-gluconate, 7m M KCl, 4m M ATP-Mg 2+ , 0.3 mM GTP-Na, 10 mM HEPES, 10 mM phosphocreatinedi(Tris), 0.4% biocytin, pH 7.4, 0.308 Osmol/Kg). In these conditions we achieved only the plasma membrane permeabilization without affecting the mitochondria [35]. The recording solution used for the calcium-induced mPTP assay was the intracellular solution supplemented with 1mM CoCl 2 . After 100 s recording the basal signal, we applied 20μM CaCl 2 and after 300 s 100μM CaCl 2 was applied. Time-lapse images were acquired on an Olympus DSU IX81 spinning-disk confocal microscope in the same conditions explained before. The images were analyzed using Stack-T-function/DeltaF plugin from NIH ImageJ software.

Electron microscopy
Hippocampal slices were used for electron microscopy analysis according to standard procedures. Ultra-thin sections were examined using a Phillips Tecnai 12 transmission electron microscope at 80 kV at the Electron Microscope Facility of the Faculty of Biological Sciences, Pontificia Universidad Católica de Chile, Santiago, Chile. For each treatment, 40 to 50 digital images at 16,500X magnification were obtained, and they were manually analyzed with ImageJ in a blinded fashion by measuring the mitochondrial area [32,36]. Ultrastructural features of the mitochondria, such as the integrity of the membrane and cristae, were determined for each mitochondrion concomitantly with the measurement of morphological parameters [37]. Membranes or cristae were considered to be intact when the entire structure was preserved and organized and the mitochondria appeared normal, as previously described [38]. Quantitative analysis was performed with n = 3 using GraphPad Prism 5.01 software. For more details, see S1 File.

3-Dimensional (3D) image reconstruction
The mitochondrial network was 3-dimensionally reconstructed as previously described [39] from mito-Cherry-transfected neurons. To overexpress mito-Cherry, we used NeuroMag (OZ Bioscience, Marseille, France) according to the manufacturer's protocol. Briefly, neurons after 10 DIV at 70,000 cells per well were incubated with the mito-Cherry plasmid and NeuroMag beads as previously described [40]. After 48 h, the neurons were used in treatments. To visualize mitochondria, neurons were illuminated at 563 nm. Serial images (z-stacks) were collected with a Nikon Eclipse C2si spectral confocal microscope with a z-step of 0.35 μm. Z-projection and 3D isosurface reconstructions of the mitochondrial network and individual mitochondria were performed with a manually adjusted threshold level of 76.4582 using Imaris 5.7.0 (Bitplane Inc., Saint Paul, MN) at the Microscope Facility of the University of Concepción, Chile. Mitochondrial volume was calculated based on 3D isosurface reconstructions, as previously described [37,41]. In each condition, 6 to 8 neurons were analyzed, with n = 4 independent experiments.
Mitochondrial membrane potential, m ΔΨ Time-lapse experiments to detect m ΔC were performed using hippocampal neurons loaded with 50 nM of the fluorescent probe MitoTracker-Orange (Molecular Probes) for 30 min at 37˚C [29,30,39]. MitoTracker-Orange fluoresces according to the m ΔC. Fluorescence variations were analyzed in the same manner as the calcein/cobalt measurements, using an Olympus DSU IX81 spinning-disk confocal microscope with λex/em values of 554/576 nm [42].

Cytochrome-c release
Neurons treated with 5 μM Aβos in the presence or absence of Wnt3a for 24 h were loaded with MitoTracker Orange (50 nM), fixed and then analysed by immunofluorescence using a mouse anti-cytochrome-c antibody (BD Pharmingen, San Diego, CA) to detect cytochrome c localization. Hoechst staining was used to detect live cells and only those neurons with no condensed nucleus were considered for the analyses. Images were captured with an Olympus FluoView1000 Confocal Microscope and analyzed using NIH ImageJ software. Manders' Coefficient M2 was calculated to determine the colocalization of cytochrome c with mitochondria.

Hexokinase activity
Hippocampal slices were prepared as described above and treated with recombinant Wnt3a (300 ng/ml), DKK1 (30 ng/mL), rWnt3a+DKK1, and 2-deoxy-D-glucose (2-DG, 7 mM; a competitive inhibitor of HK) [44] and analyzed for hexokinase activity (HK) after 3 h of treatment. Then, the slices were washed with ACSF and subjected to mechanical disaggregation and centrifuged at 500 g for 5 min at 4˚C. The tissue was resuspended in isolation medium (250 mM sucrose, 20 mM HEPES, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM DTT, 2 mg/mL aprotinin, 1 mg/mL pepstatin A, and 2 mg/mL leupeptin) at a 1:3 dilution, sonicated at 4˚C, and then centrifuged at 1,500 g for 5 min at 4˚C. Finally, the HK activity of the supernatant was quantified. For the assay, the purified fraction was mixed with the reaction medium (25 mM Tris-HCl, 1 mM DTT, 0.5 mM NADP/Na + , 2 mM MgCl 2 , 1 mM ATP, 2 U/mL G6PDH, and 10 mM glucose), and the mixture was incubated at 37˚C for 30 min. The reaction was stopped by the addition of 10% trichloroacetic acid, and the generation of NADPH was measured at 340 nm [45].

Quantification and statistical analysis
The data are shown as the mean and standard error of the mean (SEM) from 3-8 independent experiments for live-cell imaging, each with n = 3-4 replicates. Replicates (n = 3-4) were also used for immunofluorescence, mitochondrial network reconstruction and electron microscopy analysis. p values were obtained using two-way ANOVA for grouped data and one-way ANOVA for bar graphs, and the post hoc Bonferroni correction was also used. Error bars indicate SEM. Ã p<0.05; ÃÃ p<0.005; ÃÃÃ p<0.0005.

Wnt3a prevents Aβo-induced mitochondrial permeability transition pore opening
To directly monitor the mPTP opening in hippocampal neurons in vitro we used a live-cell imaging assay, based on the labeling of mitochondria with calcein-Co 2+ [31,32]. Wnt3a ligand was used to induce the activation of the Wnt signaling pathway in hippocampal neurons, as its neuroprotective properties against Aβ and different apoptotic insults are well described [24,46]. The mPTP opening was stimulated with Aβos, and as a result of permeabilization of the mitochondrial inner membrane the mitochondrial calcein fluorescence rapidly decayed in control neurons, whereas Wnt3a-treated neurons did not show any obvious changes in response to Aβos (Fig 1A). Quantification showed significant differences between control and Wnt3a-treated neurons in response to Aβos stimulation ( Fig 1B). Ionomycin was used to completely induce mitochondrial permeabilization; and Cyclosporin A (CsA), a specific inhibitor of mPTP [33], to inhibit Aβos-induced pore opening. CsA treatment showed the same protective effect as Wnt3a on mPTP inhibition, based on the absence of significant differences between the two treatments after Aβos exposure (Fig 1B, compare the white and red curves), revealing the specificity of the effect of Wnt3a on mPTP inhibition, which will be confirmed hereinafter.
In order to verify the potential effect of Wnt3a on mPTP inhibition, we evaluated whether Wnt signaling activation render mitochondria more resistant to calcium-induced mPTP, as has been described before using the same assay in intact cells [47] (Fig 2). Mitochondrial calcein intensity transiently decreased in response to calcium pulses (20 μM and 100 μM) in control neurons, immediately returning to basal levels after each pulse ( Fig 2B). This transient response was less in Wnt3a-treated cells in response to the 20μM CaCl 2 , indicating that Wnt signaling activation render mitochondria more resistance to calcium-induced mPTP (Fig 2). The protective effect produced by Wnt3a was inhibited when neurons were preincubated with DKK1 ( Fig 2C), a well characterized Wnt signaling inhibitor [48]. These results suggest that Wnt3a protects mitochondria from mPTP induction and confirm the results above-mentioned in Fig 1. Classical mitochondrial morphological changes in Aβ-induced permeability transition are prevented by Wnt signaling Mitochondria undergo several morphological changes through the mPTP induction, including mitochondrial swelling. To assess these alterations we measured the mitochondrial area using electron microscopy [32], and two integrity parameters: the membrane disruption and the cristae disorganization. The analysis was performed on the CA1 region of the hippocampus and specifically focused on mitochondria at the synaptic contacts between CA3 axons and CA1 dendrites from the stratum radiatum zone (S1A Fig). Hippocampal slices were acutely treated with Wnt3a for 4h and then co-incubated with Aβos for 1h, enough time to produce mitochondrial damage but not neuronal death (S2 Fig). Only Aβos-treated slices showed enlarged and swollen mitochondria compared with control slices. Wnt3a treatment prevented these morphological changes, and mitochondria with or without Aβos had the same normal shape as under control conditions ( Fig 3A and S1B Fig). Compared with the control group, the mean mitochondrial area value in the Aβos group increased by 31.1% (Fig 3B). The increase in the mean value for the Aβos group reflected an augmentation in the size of a subset of mitochondria in the upper 50th percentile rather than an upward size shift of the entire population. This result is demonstrated in Table 1 by the 37.1% and 104.2% increases in the 75th percentile and in the maximum values, respectively, compared with the negligible changes in the minimum values and the 10.2% increase in the 25th-percentile values when the control and Aβos groups are compared. Similar results were obtained for mitochondrial diameter and perimeter (not shown). By contrast, mitochondria from the Wnt3a+Aβos group were structurally similar to mitochondria from the control group; they showed non-significant changes compared with control mitochondria (Fig 3B, Table 1). No changes in mitochondrial biomass were observed with any of the treatments by the measurement of the number of mitochondria An ultrastructural analysis of each mitochondrion per group showed a clear deterioration of both the mitochondrial membranes and the cristae in hippocampal slices treated with Aβos ( Fig 3A). Vesicular cristae and vesicular swollen mitochondria were also observed in Aβosexposed slices (Fig 3A, red asterisk); these are features of the mitochondrial morphological changes that occur during apoptotic processes [38]. The percentage of mitochondria with intact membranes and cristae significantly decreased in Aβos-treated slices (44.8 ± 8.1% and 35.5 ± 6.1% decrease in membrane and cristae integrity, respectively) compared with control samples, whereas slices treated with Wnt3a+Aβos did not show significant changes from the control (Fig 3C and 3D). CsA was used as a control to inhibit Aβo-induced mPTP and to assess the mitochondrial structure in this condition. CsA protection against Aβos reached similar levels to those observed in hippocampal slices treated with Wnt3a+Aβos, specifically in the analysis of the membranes (Fig 3B; hatched bar), and confirmed the in vitro results obtained above in the mPTP assay ( Fig 1B).
These results suggest that the activation of Wnt signaling prevents the size changes during mitochondrial swelling triggered in Aβos-induced mPTP opening and protects mitochondria from permeabilization and cristae disorganization, favoring the maintenance of mitochondrial integrity.
Due to the fact that mitochondrial swelling involves changes in the volume of the organelle, we performed 3D reconstructions of individual mitochondria from hippocampal neurons ( Fig  4A), as has been previously described [41]. Aβos induced a significant increase in the average volume of mitochondria, which was prevented by Wnt3a to reach control levels ( Fig 4B). Together, these findings indicate that the morphological changes that occur in Aβo-exposed mitochondria correspond to mitochondrial swelling. Furthermore, the Wnt3a-protective effect on mitochondrial morphology is directly associated with its ability to inhibit the mPTP opening, as was described above in Fig 1. Integrity-dependent functions of mitochondria are affected during mPTP opening and also prevented by Wnt signaling The induction of the mPTP facilitates the disruption of the mitochondrial membranes, thereby affecting the mitochondrial membrane potential ( m ΔC). As a consequence, mitochondria release apoptotic factors that trigger neuronal cell death cascades [49]. Incubation of control neurons with Aβos induced a significant loss of the m ΔC. No statistical differences were observed between control and Wnt3a neurons in the presence of Aβos (Fig 5A). In addition, we evaluated cytochrome c release from mitochondria in response to Aβos and Wnt3a treatments. Biochemical analyses did not show any differences in neither mitochondrial nor cytoplasmic cytochrome c levels in response to any treatment (not shown). However, when cytochrome c is specifically measured in the neurites of cultured neurons by immunofluorescence, we observe a slight but significant effect on cytochrome c release from mitochondria in Aβos treated neurons. Manders' Coefficient M2 was used to analyse the overlapping of the  cytochrome c labeling with the mitochondria. A significant decrease in the M2 coefficient was observed in neurites evaluated from Aβos-treated neurons, which was prevented by Wnt3a incubation reaching control levels (Fig 5B). In these conditions we also observed a significant decrease in the percentage of live neurons (+calcein/-EthD1) in the Aβos treatment compared  with controls (44.09 ± 3.09% vs. 72.97 ± 2.74%, respectively) ( Fig 5C). Neurons that were coincubated with Aβos+Wnt3a did not show significant differences from control neurons (77.40 ± 1.49% vs. 72.97 ± 2.74%, respectively). The same results were observed by counting apoptotic nuclei by Hoechst staining (S3 Fig). CsA treatment was used to determine whether this Wnt3a-mediated protection is related to the inhibition of permeability transition. Neurons that were co-incubated with Aβos and CsA did not show significant differences compared with control or Wnt3a+Aβos-treated neurons (Aβos+CsA: 70.26 ± 1.71) (Fig 5C), suggesting that both Wnt3a and CsA protect neurons by the same mechanism, the inhibition of the mPTP. These results suggest a novel mechanism by which the Wnt signaling pathway might protect neurons from Aβos-induced cell death.

The effect of Wnt3a on mPTP inhibition is not dependent on the transcription of Wnt target genes
In order to determine the mechanism by which Wnt signaling pathway inhibits the mPTP, we blocked Wnt signaling in different steps ( Fig 6A) and we evaluated their effects with the mPTP opening assay. The first step of the signaling, that is, the binding of the ligand with the Fz receptor, was blocked using two classical antagonists: sFRP2, which binds to the Wnt ligand; and DKK1, which blocks the receptor complex [48]. Both inhibitors significantly decreased the protective effect of Wnt3a over Aβos-induced mPTP opening (Fig 6B). The ICG001, an inhibitor of Wnt/β-catenin-dependent transcription [50] was not able to prevent Wnt3a-protection (Fig 6B), which suggests that the protective effect of Wnt3a on mPTP opening is not dependent on the transcription of Wnt target genes. By contrast, the direct inhibition of GSK-3β, one of the main components of the "destruction complex" of β-catenin, with 6-bromoindirubin-3 0 -oxime (6-BIO) [30], showed the same protective effect as Wnt3a against Aβos ( Fig  6B), suggesting that this kinase could be involved in the protection that Wnt signaling produces on mitochondria. Same results were observed with all the inhibitors when the m ΔC was evaluated (Fig 6C).

Wnt3a increases mitochondrial phosphorylated GSK-3β: Possible mechanisms for Wnt-dependent mPTP regulation
Mitochondrial GSK-3β has been described as a trigger of the permeability transition [51]. The inhibition of GSK-3β through serine 9 phosphorylation is associated with cardioprotection and with the regulation of kidney and liver injury through the inhibition of mPTP opening [52][53][54]. Interestingly, mPTP-related GSK-3β phosphorylation occurs in mitochondria at the same residue (Ser9) that leads to Wnt signaling activation in the intracellular compartment [55]. To study GSK-3β behavior in brain mitochondria, we isolated mitochondria from hippocampal slices. We found that Wnt3a triggered the accumulation of phosphorylated GSK-3β (p-GSK3β-S9) in the mitochondrial fraction, with a peak of GSK-3β inactivation after 4 h of treatment (Fig 7A). To evaluate the specificity of this effect, we inhibited the activation of Wnt signaling using DKK1. The accumulation of p-GSK3β-S9 in the mitochondrial fraction was significantly prevented when Wnt signaling was inhibited, even in the presence of the Wnt3a ligand (Fig 7B).
To evaluate the possibility that the Wnt3a-induced mitochondrial p-GSK3β-S9 might also interact with some of the protein components of the mPTP, as has been previously described in other models [56], we performed immunoprecipitation assays for p-GSK3β-S9 with ANT and CypD. Wnt signaling induced an increase in the association between p-GSK3β-S9 and ANT, whereas no significant changes were observed with CypD ( Fig 7C).
Another regulator of mPTP opening is the mitochondrial hexokinase II (HKII), which stabilizes the mPTP in its closed conformation; HKII detachment from mitochondria propagates a conformational change that leads to pore opening [57]. Interestingly, GSK-3β activity has been shown to regulate the release of HKII from mitochondria in response to Aβ, thereby enhancing susceptibility to cell death through the mitochondrial permeability transition [58].
To correlate both events we tested Wnt3a effect over mitochondrial HKII levels. A significant increase of HKII in mitochondrial fractions was observed in Wnt3a-treated mitochondria, which were also prevented by DKK1 (Fig 8A). Furthermore, it has been described that HKII activity is also required for its protective effect [59]. We used hippocampal slices treated with Wnt3a +/-DKK1 to test whether Wnt signaling also modulates the activity of HK. The activity assay showed increased HK activity in response to Wnt3a, which was inhibited by co-incubation with DKK1, reinforcing our hypothesis that HKII could be a good candidate to mediate the inhibitory effect of Wnt3a on mPTP.
Together, these results suggest that Wnt signaling could regulate inhibition of the mPTP through a mechanism that involves mitochondrial GSK-3β inactivation. It may thereby act by regulating the detachment/translocation process of HKII to mitochondria and/or by GSK-3β binding to the ANT protein to directly inhibit pore opening.

Discussion
The permeabilization of mitochondrial membranes is determinant for cell fate decision between life or death, and has been described as "a point of no return" in mitochondrial cell death [60]. The mPTP plays a central role in alterations of mitochondrial structure and function leading to neuronal damage related to several neurodegenerative diseases [34,[61][62][63], including AD [7]. This is why the mPTP has been consistently proposed as molecular target for the development of therapeutic approaches in neurodegeneration [64]. In AD the permeability of the mitochondrial membranes has been described as an early event in the pathogenesis of the disease, preceding even the synaptic dysfunction and the neuronal cell death [14,65]. For this reason, we have been interested in determining how to protect the mitochondria, to prevent the toxic events that finally produce the neuronal damage. Our previous studies have Input corresponds to whole-slice lysate, and IPP: p-GSK3β-S9 corresponds to the fraction immunoprecipitated with a phosphorylated GSK3β-S9 antibody. Immunoblotting (IB) was performed to detect p-GSK3β-S9, ANT and CypD. Total GSK-3β and GAPDH were used as positive and negative controls, respectively, for the immunoprecipitation assay. Lane C: control; 3a: recombinant Wnt3a. The densitometric analysis was performed from 4 independent experiments. Statistical analyses were conducted using one-way ANOVA and post hoc Bonferroni correction: *p<0.05; **p<0.005.
shown that activation of the Wnt signaling pathway has a neuroprotective role against Aβ toxicity [21,25]; however, there is no direct evidence that explains whether this neuroprotective effect is mediated by a regulation of cell death cascades that are initiated in the mitochondria. Our data demonstrate that the Wnt3a ligand prevents the Aβo-induced mPTP opening and thereby all of the consequent effects, such as mitochondrial morphological changes, structural alteration, m ΔC collapse, cytochrome c release and, therefore, neuronal cell death. The results obtained in the mPTP opening assay suggest that Wnt signaling could protect neurons from Hippocampal slices were treated with DKK1 (100 ng/mL) for 30 min before and during Wnt3a treatment. (A) Western blot analysis of mitochondrial fractions shows increase HKII levels induced by Wnt3a (300 ng/mL) and the inhibition produced by DKK1 (100 ng/mL) at 4 h of co-treatmet (n = 3). Graphs represent the densitometric analysis of and HKII levels in mitochondrial fractions. Protein levels were normalized to COXIV. (B) HK activity was measured in hippocampal slices treated with recombinant Wnt3a for 3 h. DKK1 was used to inhibit the effect of Wnt3a, and the competitive inhibitor of HK, 2-DG (7 mM), was used as an assay control to inhibit basal HK activity. Statistical analyses were conducted using one-way ANOVA and post hoc Bonferroni correction: *p<0.05; ***p<0.0005. doi:10.1371/journal.pone.0168840.g008 cell death by inhibiting the mPTP induction, as has been described for the classical mPTP inhibitor, CsA [33]. Both, CsA and Wnt3a, provide the same level of mPTP inhibition (Fig 1). Equivalent results were obtained in the ultrastructural analyses and the neuronal viability assay where Wnt-mediated protection reached the same levels as CsA to prevent Aβos damage (Figs  3 and 5C, respectively). The magnitude of the effect of both agents suggests that Wnt signaling prevents neuronal cell death by regulating mPTP opening.
Consistent with the inhibition of the mPTP, we also showed that the morphological changes induced in mitochondria during the permeability transition are prevented by Wnt3a (Fig 2). The electron microscopy analysis on synaptic mitochondria showed dramatic changes in the mitochondrial integrity in slices that were acutely exposed to Aβos only for 1 h. These alterations are correlated with the enhanced sensitivity of synaptic mitochondria to Aβ-induced damage [66] specifically at the CA1 region, where we performed the analysis (S1 Fig) [67]. The early sensitivity of mitochondria to amyloid damage compared with any other compartment in the cell has been extensively discussed by other groups that have proposed the "AD mitochondrial cascade hypothesis", wherein the common feature of disease progression is early mitochondrial dysfunction [6,14]. Such dysfunction includes the high sensitivity of synaptic mitochondria to permeability transition [68]. Interestingly, at 1 h of Aβos treatment, we did not observe neuronal cell death (S2 Fig), reinforcing the notion that the damage observed in mitochondria is not a consequence of a general neuronal impairment; instead, Aβos only affected mitochondria as an initial step towards cell death, as we observed in the 24-h treatments ( Fig 5C).
How can Wnt signaling regulate the permeability transition and pore inactivation?
Most of the effects that has been described for the Wnt signaling pathway are β-catenindependent and/or are mediated by the diversity of Wnt target genes [19]. Likewise, there are several mPTP modulators that have been described to act through interactions with mPTP components that could be related to Wnt signaling and their genes [13,69,70]. One of these modulators is the anti-apoptotic protein Bcl-2 [71], which is downregulated in AD models [72]. We have previously shown that Bcl-2 is a Wnt target gene [73] that participates in mitochondrial stabilization against Aβ damage [74]. In the context of mPTP-induced apoptosis, Bcl-2 has been shown to prevent cell death by interacting with CypD [70], revealing a novel protective function of CypD. However, we observed that the Wnt-mediated mPTP inhibition was not dependent on the activation of Wnt target gene transcription (Fig 6), strengthening the idea that protection is not related to the action of Bcl-2, or to any other Wnt-regulated gene.
It is well-documented that Aβ induces mPTP activation in an AD mouse model and in vitro [75,76] by the direct interaction with CypD [77]. In an AD mice model, which presents mitochondrial ultrastructural abnormalities [78] similar to those shown in Fig 2, specifically mitochondrial swelling, CypD deficiency attenuates the cellular death induced by Aβ and improves memory and synaptic function [77]. In the same way, increased levels of CypD in synaptic mitochondria are also important for their sensitivity to Aβ damage, making them more vulnerable to permeability transition processes [79]. These antecedents suggest CypD as the best candidate to control the open conformation of the pore [8], however we did not observe any changes in CypD protein levels (input) in hippocampal slices treated with Wnt3a ( Fig 7C). This result does not allow us to discard the participation of CypD in the Wnt-mediated protection, but it enables us to explore other possible candidates. Indeed, CypD has been identified, at least in AD, as a main contributor of neuronal damage through the induction of the mPTP [77,80], however many other proteins have been implicated in the formation of the pore, some of them dependent on CypD, as Bcl-2 [70], and others that are able to form the pore regardless of CypD, such as Bax [12] and ATP synthase [10,81]. Furthermore, glutamate-triggered mPTP has been shown to be produced in both CypD-dependent and -independent manner, according to the concentrations used [82]. This scenario widens the range of candidates that could participate in the Wnt3a-mediated inhibition of the pore.
Recent studies support the idea that GSK-3β, one of the central components of the Wnt signaling pathway, plays a crucial role in the regulation of the mPTP in a variety of disease models [83][84][85]. Several years ago a mitochondrial GSK-3β pool was detected in cerebellum and since then, GSK-3β has been proposed to participate and regulate the induction of the permeability transition, being a point of convergence of different survival signals [86]. The inactive form of GSK-3β (phosphorylated at Ser9) interacts with ANT, a proposed component of the mPTP [52,56]. This interaction has been correlated with a decrease in the CypD-ANT binding, preventing CypD translocation from the matrix to the mitochondrial inner membrane [87,88], which is a necessary event for Cyp-dependent mPTP formation [77]. Interestingly, the activation of Wnt signaling results in GSK-3β inactivation by its phosphorylation at Ser9, as occurs in mitochondria, producing the inactivation of the cytoplasmic GSK-3β pool, without affecting total GSK-3β levels [24,89]. Our results showed an increase in mitochondrial p-GSK3β-S9 levels in response to Wnt3a (Fig 7A and 7B) and a Wnt-induced interaction between p-GSK3β-S9 and ANT, which suggest that the inhibition of the pore opening, in response to Wnt signaling activation, could be mediated by mitochondrial GSK3β through the regulation of the interaction between the mPTP protein components.
Reinforcing the idea of GSK3β participation in the inhibition of the pore, we showed that inhibiting its kinase activity with the specific inhibitior 6-BIO, which has been widely used as a Wnt signaling activator [90], the Aβos-induced mPTP opening was inhibited at the same magnitude that Wnt3a was in our live cell mPTP-assay (Fig 6B). These results support the antecedents related to the role of p-GSK3β-S9 in the inhibition of the pore, and strengthens our hypothesis of its participation in the protective effect of Wnt3a.
Another known mPTP modulator is the mitochondrial hexokinase II (HKII). Detachment of HKII from the mitochondria induces mPTP opening and cell death [57,91]. Interestingly, and consistent with the proposed role of p-GSK3β-S9 in mPTP inhibition, it has been described that the inactivating phosphorylation of GSK-3β by Akt favors the association of HKII with the outer mitochondrial membrane [92]. By contrast, activation of GSK-3β has been shown to induce the release of HKII, increasing susceptibility to cell death [13,93]. Here, we observed increased levels of HKII on the mitochondria in response to Wnt3a (Fig 8A). Moreover, not only mitochondrial binding of HKII is required to inhibit mPTP. As has been shown, HKII activity also contributes to the protective effect of the kinase [59]. We explained that Wnt3a increases HK activity in hippocampal slices, possibly favoring the protective effect of Wnt3a on mPTP inhibition. These results suggest another possible mechanism for Wnt/ GSK-3β signaling in the regulation of mPTP opening, and it allows us to propose two distinct but complementary modes of regulation: (1) via phosphorylation of the cytoplasmic pool of GSK-3β and thereby the translocation and attachment of HKII to the mitochondria to inhibit mPTP opening and (2) via regulation of mitochondrial p-GSK3β-S9 levels to directly interact with the ANT protein in the mPTP complex (Fig 9).
We proposed, for the first time, a possible mechanism to explain the neuroprotective action of Wnt signaling, which involves the regulation of the mPTP through the inhibition of GSK-3β, thereby preventing the Aβos neurotoxicity. This study suggests a possible new approach for the treatment of AD, based on the preservation of the mitochondrial structure as a target, and opens a new line of study in the field of Wnt signaling in neuroprotection. Proposed mechanism for the action of Wnt3a in regulating mitochondrial mPTP opening. Wnt signaling is activated by binding of the Wnt3a ligand to its receptor, Fz, and to its co-receptor, LRP6. This interaction activates Dvl, which causes the dissociation of the destruction complex, which includes Axin, APC, and GSK-3β. This dissociation prevents the proteasomal degradation of βcatenin, thereby inducing its cytoplasmic accumulation and eventual translocation to the nucleus, where it regulates the expression of Wnt target genes. This pathway is known as Wnt/β-catenin signaling. However, Wnt can also act through another signaling mediator, GSK-3β, which has functions that are independent of β-catenin and thus of Wnt target gene transcription [94]. The activation of Wnt signaling triggers the inhibitory phosphorylation of GSK-3β at serine 9 in the cytosol. In our proposed model, the inhibited GSK-3β accumulates in mitochondria in response to the activation of Wnt signaling by Wnt3a. Mitochondrial p-GSK3β-S9 then interacts with ANT to inhibit the opening of the mPTP. This interaction has also been correlated with inhibition of the binding between ANT and CypD that is necessary for the conformational shift and opening of the mPTP. GSK-3β activity can also control the detachment or translocation of HKII to the mitochondria. Activation of Wnt signaling increases mitochondrial HKII, probably through the inhibition of GSK-3β activity, thus also favoring the closed conformation of the pore. (Fz: Frizzled; Dvl: Dishevelled; APC: adenomatous polyposis coli; GSK-3β: glycogen synthase kinase-3β; p-GSK3β-S9: GSK-3β phosphorylated at Ser9; ANT: adenine nucleotide translocase; CypD: cyclophilin D; VDAC: voltage-dependent anion channel; HKII: hexokinase II). Mouse hippocampal slices (400 μm) were pre-incubated for 4h with Wnt3a and then treated with 5μM Aβo for 1 h. Slices were fixed and processed for Hoechst staining. Images show a representative hippocampal slice stained with Hoechst (a-d). Graph shows the quantification of percentage of apoptotic nuclei in each condition (e). Non-significant changes were observed between each condition using one-way ANOVA test with a post hoc Bonferroni. Quantifications represent the results of three independent experiments. (TIFF)