|Home | About | Journals | Submit | Contact Us | Français|
Amyloid β (Aβ) peptides, the main pathological species associated with Alzheimer’s disease (AD), disturb intracellular calcium homeostasis, which in turn activates the calcium-dependent phosphatase calcineurin (CaN). CaN activation induced by Aβ leads to pathological morphological changes in neurons, and overexpression of constitutively active calcineurin is sufficient to generate a similar phenotype, even without Aβ. Here, we tested the hypothesis that calcineurin mediates neurodegenerative effects via activation of the nuclear transcription factor of activated T-cells (NFAT).
We found that both spine loss and dendritic branching simplification induced by Aβ exposure were mimicked by constitutively active NFAT, and abolished when NFAT activation was blocked using the genetically encoded inhibitor VIVIT. When VIVIT was specifically addressed to the nucleus, identical beneficial effects were observed, thus enforcing the role of NFAT transcriptional activity in Aβ-related neurotoxicity. In vivo, when VIVIT or its nuclear counterpart were overexpressed in a transgenic model of Alzheimer’s disease via a gene therapy approach, the spine loss and neuritic abnormalities observed in the vicinity of amyloid plaques were blocked.
Overall, these results suggest that NFAT/calcineurin transcriptional cascades contribute to Aβ synaptotoxicity, and may provide a new specific set of pathways for neuroprotective strategies.
Amyloid beta (Aβ) soluble oligomeres are thought to be an important source of neurotoxicity in Alzheimer’s disease (AD) (Walsh et al., 2002; Lesne et al., 2006; Shankar et al., 2007; Koffie et al., 2009). While synaptic loss correlates best with cognitive decline (Terry et al., 1991; DeKosky et al., 1996), the molecular mechanisms underlying Aβ synaptotoxicity and cognitive impairment remain largely unexplained, and new therapeutic approaches acting to protect neurons towards Aβ-related injuries need to be investigated.
Calcium dyshomeostasis and the consequent activation of calcineurin (CaN, or protein phosphatase 2B) may trigger Aβ-related pathological effects (Liu et al., 2005; Shankar et al., 2007; Kuchibhotla et al., 2008; Wu et al., 2010). An increased CaN activity has been described in AD brains (Liu et al., 2005; Wu et al., 2010), and CaN activation was reported in astrocytes surrounding amyloid deposits, thus exacerbating neuroinflammation (Norris et al., 2005). In neurons, CaN activation leads to pathological morphological changes, which could be blocked by abolition of CaN enzymatic activity using FK506 (Wu et al., 2010; Rozkalne et al., 2011). FK506 also improves cognitive deficits in the Tg2576 AD mouse model (Taglialatela et al., 2009), thus making calcineurin a potential therapeutic target. However, because FK506 prevents CaN phosphatase activity in all the cell types and towards all its substrates, the mechanisms underlying FK506 neuroprotection are unclear.
Calcineurin is a unique neuronal Ca2+/Calmodulin-dependent serine/threonine phosphatase that plays fundamental physiological roles during development and regulates processes such as neurotransmitter release, synaptic plasticity and learning (Klee et al., 1979; Groth et al., 2003). Upon activation, CaN leads to post-translational modification of post-synaptic proteins such as cofilin (Zhou et al., 2004) and AKAP79, which is associated with long term depression (Bhattacharyya et al., 2009; Jurado et al., 2010). In addition, CaN dephosphorylates the nuclear factor of activated T cells (NFAT), which induces its nuclear translocation and the expression of target genes implicated in neuronal survival, axonal outgrowth and dendritic complexity (Benedito et al., 2005; Nguyen and Di Giovanni, 2008; Schwartz et al., 2009). CaN activation therefore triggers short-term effects directly at the synapse as well as long-term modifications of synaptic plasticity by modulating gene expression through NFAT. In order to decipher more precisely the role of CaN activation in Alzheimer disease, we tested the hypothesis that neuronal morphological changes caused by Aβ may be improved by inhibiting the CaN-induced NFAT pathway. We used a genetically encoded VIVIT peptide initially developed as a specific competitive inhibitor of the interaction between CaN and NFAT, without affecting CaN phosphatase activity (Aramburu et al., 1999). We established that 1) VIVIT efficiently inhibits the neurotoxic effects associated with a constitutively activated CaN (CACaN); 2) expression of a constitutively active form of NFAT produces a phenocopy of the effects induced by CACaN or by Aβ exposure; 3) Aβ-induced decrease of spine density and dendritic complexity is inhibited by VIVIT; 4) restricted expression of VIVIT in the nucleus retains the beneficial effects of VIVIT, whereas a membrane-bound version of this peptide has no effect; 5) Changes associated with amyloid plaques are alleviated when VIVIT or NLS-VIVIT were introduced by AAV-mediated gene delivery in the cortex mouse model of AD in vivo.
Taken together, our findings offer new therapeutic opportunities by targeting Aβ downstream events through inhibition of NFAT transcriptional pathways.
Primary neurons were cultured from cerebral cortices of embryonic day 16 mice (Charles River Laboratories, Wilmington, MA). Embryos resulted from the mating between a Tg2576 male that heterozygously overexpresses a human mutated APP gene and a wild-type female, thus giving rise to both transgenic (Tg) and littermate (Wt) cultures. The genotype of each embryo was determined by PCR. Neurons were prepared as previously described (Wu et al., 2010) and plated to a density of 6.7×105 viable cells/35-mm culture dishes coated with poly-D-lysine (100μg/ml, Sigma-Aldrich, St. Louis, MO). The medium of the cells was not renewed to allow Aβ peptides to accumulate and conditioned media were collected from 14DIV cultures (Wu et al., 2010). The concentration of Aβ peptides was quantified by a mouse/human ELISA kit (Wako, Japan) and reached approximately 7000pMol. Transfection experiments were done at 5DIV using lipofectamin 2000 (Invitrogen, Carlsbad, CA).
At 5DIV, primary neurons were transfected with either pEGFP-N1 or pDsRed-Express-N1 plasmids (Clontech Laboratories, Mountain View, CA), altogether with pVIVIT-EGFP and pVIVIT-NLS-EGFP constructs generously provided by Dr Norris (Sanders-Brown Center On Aging, University of Kentucky, KT) and Dr Ruthazer (McGill University, Montreal, QC), respectively. For a subset of experiments, cells were also transfected with a wild-type or a constitutively active form of calcineurin previously described (Wu et al., 2010). Live imaging of GFP− or DsRed-expressing primary neurons (18 DIV) was used to analyze the morphological parameters of the cells. Images of the whole cell and of dendritic segments were captured using a LSM 510 Zeiss microscope in order to determine the neuritic complexity and the spine density. GFP and DsRed were respectively excited at 488nm and 543nm and emitted light was collected between 500–550nm and 565–615nm. Dendritic spines were analyzed using the Neuron Studio software (CNIC tools) that automatically detects three different spine types (thin, stubby and mushroom) according to their morphological measures (i.e. length of the “neck” and size of the “head” of each spine) (Rodriguez et al., 2008). The complexity of the neuronal dendritic tree was determined by Sholl analysis (Spires-Jones et al., 2011), reporting the number of branch points with respect to the distance from the cell body. All these parameters were evaluated on transfected cells that did not show any obvious morphological alterations such as dendritic dystrophies.
Immunostaining was performed using a standard protocol (Wu et al., 2010). Briefly, primary neurons were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) (pH 7.4) for 15 min, permeabilized with 0.5% Triton X-100 in PBS for 20 min and blocked with 3% bovine serum albumin at room temperature for 1 h. Primary antibodies to detect NFATc4 (polyclonal sc-13036, 1:200 dilution, Santa Cruz Biotechnology, Santa Cruz, CA), Laminin-B1 (polyclonal ab16048, 1:1000, Abcam, Cambridge, MA), HA tag (Clone 16B12, 1:200 dilution, Covance, Princeton, NJ), GFP (polyclonal ab6556, 1:1000, Abcam, Cambridge, MA or chicken anti-GFP, Aves, Tigard, Oregon) were incubated overnight at 4°C. Secondary antibodies conjugated to either cyanine 3 (Cy3, 1:1000; Jackson ImmunoResearch, West Grove, PA), cyanine 5 (Cy3, 1:500; Jackson ImmunoResearch, West Grove, PA) or Alexa 488 (1:1000; Molecular Probes, Eugene, OR) were then applied. Fluorescent images were obtained using a LSM 510 Zeiss confocal microscope.
The changes in the subcellular distribution of NFATc4 were evaluated after immunostaining of the endogenous protein in primary neurons or in mouse brain slices. GFP positive neurons that were previously transfected or transduced with GFP, VIVIT-GFP or NLS-VIVIT-GFP were specifically analyzed. Using image J (National Institutes of Health open software), the fluorescence intensity of NFATc4 in the nucleus was determined by overlap with the nuclear staining DAPI (Vector laboratories, Burlingame, CA), whereas the fluorescence intensity in the cytoplasm was quantified in the rest of the cell body, excluding the nuclear compartment (see Fig. 1C). The intensity of nuclear NFATc4 was then divided by the intensity of cytoplasmic NFATc4.
Cortical primary neurons (7DIV) were transduced with an AAV-NFAT–TA–Luc, altogether with AAV-wtCaN or AAV-CACaN and AAV-GFP, AAV-VIVIT-GFP or AAV-NLS-VIVIT-GFP. Three days after transduction, cells were harvested and the luciferase activity was measured with a luminometer using a reagent kit (Luciferase Assay System with Reporter Lysis Buffer; Promega). The background luciferase activity calculated when pNFAT-Luc alone was added, was subtracted from all experiments.
In vivo experiments were performed using APPswe/PS1d9 double transgenic mice (APP/PS1, obtained from Jackson laboratory, Bar Harbor, Maine) that overexpress a human mutant amyloid precursor protein gene containing the Swedish mutation K594N/M595L and a variant of the Presenilin 1 gene deleted for the exon 9, both under the control of PrP promoter (Jankowsky et al., 2004). Substantial amyloid deposition is visible by 6 months of age and we used 7-month-old animals. Wild-type littermates were used as controls. Experiments were performed in accordance with NIH and institutional guidelines.
Plasmids containing VIVIT-GFP and NLS-VIVIT-GFP were digested using NotI and NcoI restriction enzymes and subcloned into an AAV backbone containing the chicken β-actin promoter and a Woodchuck Hepatitis Virus Posttranscriptional Regulatory Element (WPRE). The Myr tagged VIVIT-GFP was obtained after amplification by PCR and insertion into an AAV backbone using Xho1 and BsrG1 restriction sites. The AAV-NFAT–TA–Luc plasmid was purchased from Clontech Laboratories (Mountain View, CA) and the constructions of AAV-CBA-HA-wtCaN and pAAV-CBA-HA-CACaN backbones were previously described (Wu et al., 2010). As a control, we used an AAV-GFP that was available at the Harvard Gene Therapy Core. High titer of AAV serotype 2 vectors (AAV2-GFP, AAV2-VIVIT-GFP and AAV2-NLS-VIVIT-GFP, AAV2-WtCaN, AAV2-CACaN and AAV2- NFAT–TA–Luc) were produced after triple-transfection of HEK293 cells by the Harvard Gene Therapy Core, except for the AAV8-tdTomato that was provided by Gene Transfer Vector Core from University of Iowa. Lentiviral vectors coding for WtNFAT and CANFAT were obtained after amplification of HA-mNFATc4 (WtNFAT) and HA-CA-mNFATc4 (CANFAT) by PCR. Both PCR products were then digested by NheI and BsrGI restriction enzymes and cloned into a pCSCW2-IRES lentiviral backbone. Lentiviral vectors were produced by the Vector Development and Production Core at MGH. All the constructs were verified by sequencing.
Stereotactic intracortical injections of AAV serotype 2 vectors were performed as described previously (Spires et al., 2005). Animals were anesthetized by intraperitoneal injection of ketamine/xylazine (100mg/kg and 50mg/kg body weight, respectively) and positioned on a stereotactic frame (David Kopf Instruments, Tujunga, CA). Injections of vectors were performed in the cerebral cortex of each hemisphere with 3μl of viral preparation (titer 2×1012 vg/ml) using a 33-gauge sharp micropipette attached to a 10-μl Hamilton syringe (Hamilton Medical, Reno, NV) at a rate of 0.12 μl/minute. Stereotactic coordinates of injection sites were calculated from bregma (anteroposterior −1mm, mediolateral ± 1mm and dorsoventral −1.2mm).
Four weeks after AAV intracortical injection, mice were anesthesized with isoflurane (1.5%) and a cranial window was implanted by replacing a piece of skull by a glass coverslip of 8mm diameter (as described previously, (Spires et al., 2005)). For imaging, a wax ring was built on the edge of the window to create a well of water for the objective (Olympus XLPlan N 25x objective with a numerical aperture of 1.05). In order to visualize the amyloid deposits, transgenic animals received an intraperitoneal injection of methoxy-XO4 (5mg/kg) 24hrs prior to surgery, a fluorescent compound that crosses the blood–brain barrier and binds to amyloid deposits (Bacskai et al., 2002; Klunk et al., 2002). When needed, Texas Red dextran (70,000 Da molecular weight; 12.5 mg/ml in sterile PBS; Molecular Probes, Eugene, OR) was injected into a lateral tail vein to provide a fluorescent angiogram.
GFP-filled neuronal processes, amyloid deposits (in the case of APP/PS1 animals) and blood vessels were imaged using the Olympus FluoView FV1000MPE multiphoton laser-scanning system mounted on an Olympus BX61WI microscope (Olympus, Tokyo, Japan). A DeepSee Mai Tai Ti:sapphire mode-locked laser (Mai Tai; Spectra-Physics, Fremont, CA) generated two-photon fluorescence with 860 nm excitation. Emitted light was detected through three filters in the range of 420–460, 495–540, and 575–630nm (Hamamatsu; Ichinocho, Japan). Neurites were imaged at depth of 50 to 200μm from the surface of the brain. High-resolution images were captured using the optical zoom feature in the Fluoview software (63× 63× 1 μm sections; 40–100 sections per stack).
Two-dimensional projections of GFP-filled neurites were obtained using Image J. Only dendrites that were at least 20μm long and had prominent dendritic spine protrusions were considered for analysis. Spine density and morphology were evaluated using the Neuronstudio software, as previously described (Rodriguez et al., 2008). Distances to plaques were evaluated by the average of three different measures between each neurite (two measures at each endpoint and one measure at the midpoint) and the edge of the nearest amyloid deposit.
One month after implantation of the cranial window, mice were sacrificed by CO2 inhalation. One cerebral hemisphere was fixed in 4% paraformaldehyde in phosphate buffer saline whereas the other hemisphere was snap frozen in liquid nitrogen. To detect GFP and NFATc4, paraffin-embedded sections (10 μm) were sequentially deparaffinized in xylene, rehydrated in ethanol, treated in citrate buffer (10mM Sodium Citrate, 0.05% Tween 20, pH 6.0), permeabilized in PBS with 0.5% Triton, blocked in PBS with 3% BSA, and incubated overnight with the primary antibody. Incubation with the secondary antibody was done for 2hrs at room temperature the next day.
Except for the Sholl’s Plot Analyses, statistical analyses were done using Statview and JMP softwares. In vitro experiments were done at least in triplicate. In vivo, at least 5 animals were included per condition. Equality of variances (F-test) was verified between each date of experiment and between animals of the same group. We made comparisons between groups by one- or two - way ANOVA followed by post-hoc Bonferroni correction for multiple comparisons among means. Data are presented as mean ± standard deviation (SD).
Sholl’s Plot Analyses were represented as the mean of the number of intersections according to the distance from the cell body and statistics were done using a newly developed analytical approach (using R software, (Spires-Jones et al., 2011)). To assess the difference in length and complexity of dendritic arborization between each treatment, the area under the curve (AUC) was estimated through a trapezoidal method, which reflects the overall complexity of the dendritic arborization. A Wilcoxon Rank Sum test followed by a Bonferroni p-value adjustment procedure was applied. For all the analyses, differences with a p value of < 0.05 were considered statistically significant.
The use of pharmacological inhibitors of CaN phosphatase activity, such as FK506, is associated with a broad range of side-effects in transplanted patients (Ponticelli and Campise, 2005; Lee et al., 2008). In an effort to develop a safer drug, Aramburu et al. screened for an NFAT-specific reagent and developed a new inhibitor peptide called VIVIT (16 amino-acids, MAGPHPVIVITGPHEE) that was able to interact with the CaN docking motif of NFAT with a high affinity, without interfering with its enzymatic activity (Aramburu et al., 1999).
In vitro, we first tested if the overexpression of VIVIT-GFP in cortical primary neurons was able to counteract the nuclear translocation and the transcriptional activity of NFATc4 (the most prominent NFAT isoform in neurons, (Ho et al., 1994), as well as the neuronal morphological changes induced by a constitutively active form of calcineurin (CACaN). The constitutively active CaN was initially identified as a calpain-dependent truncated product of calcineurin (45kDa), in which the regulatory autoinhibitory domain was removed, so that CACaN has an enhanced phosphatase activity (Wu et al., 2004). By contrast, overexpression of wild-type calcineurin (WtCaN) is tightly regulated and does not lead to an increase of calcineurin activity. At 5DIV, cortical primary neurons were cotransfected with either CACaN or WtCaN and with either GFP or VIVIT-GFP. Cells were cultured for 16–18 DIV before analysis. Because of the presence of an HA-tag in the WtCaN and CACaN constructs, we verified that more than 90% of GFP-filled neurons also coexpressed each form of CaN (not shown). As expected, the nuclear/cytoplasmic ratio of NFATc4 immunofluorescence intensity was significantly higher when CACaN was overexpressed (0.94 ± 0.025) compared with WtCaN (0.62 ± 0.028), whereas WtCaN by itself did not affect NFAT subcellular localization compared to GFP transfected cells (0.67 +/− 0.08) (Fig. 1A, B). This suggests that increased CaN activity leads to an accumulation of NFATc4 in the nucleus. By contrast, co-transfection of CACaN with VIVIT-GFP prevents the nuclear accumulation of NFATc4, as indicated by lower nuclear/cytoplasmic ratios (0.62 ± 0.036) that were comparable to WtCaN transfected neurons (Fig. 1A, B). When VIVIT-GFP alone was overexpressed, no change could be detected in the subcellular localization of NFATc4 and the nuclear/cytoplasmic ratio remained comparable to GFP-transfected cells (not shown). To further evaluate the effect of VIVIT on NFAT transcriptional activity, a reporter system containing several NFAT transcriptional response elements driving the expression of luciferase was used. Primary neurons were transduced with AAV-GFP, AAV-WtCaN+GFP, AAV-CACaN+AAV-GFP, AAV-CACaN+AAV-VIVIT-GFP or at 10DIV and luciferase activity was quantified after three days. Overexpression of a constitutively active form of calcineurin led to an increase of NFAT transcriptional activity compared with WtCaN. When VIVIT-GFP was co-transduced with CACaN, a significant decrease of the Luciferase activity was observed, suggesting that VIVIT efficiently inhibits NFAT dependent transcriptional activity (Fig. 1C).
To determine if VIVIT-GFP was able to prevent the pathological morphological changes induced by CACaN, we analyzed the spine density and dendritic complexity of GFP or VIVIT-GFP expressing neurons that were also transfected with WtCaN or CACaN plasmids. The density of spines in CACaN/GFP overexpressing neurons (0.18 ± 0.046 spines/μm) was significantly lower compared with cells overexpressing WtCaN/GFP (0.42 ± 0.09 spines/μm, Fig. 1D, E). Sholl’s Plot analyses also indicated that uncontrolled activation of calcineurin in neurons induced a marked dendritic simplification compared with wild-type calcineurin (Fig. 1D, G). However, when VIVIT-GFP was introduced together with CACaN, a significantly higher spine density (0.4 ± 0.081 spines/μm) and a more complex neuritic arborization were observed, so that these morphological parameters reached those of WtCaN transfected neurons (Fig. 1D, E, G). Importantly, no difference was observed between cells that were transfected with GFP alone (0.43 ± 0.13 spines/μm) or WtCaN+GFP (0.42 ± 0.09 spines/μm), suggesting that WtCaN did not change the spine density and dendritic complexity of cortical neurons in culture (Fig. 1D, E and G). Similarly, overexpression of VIVIT-GFP alone did not affect the morphological parameters of the cells (not shown). The beneficial effect of VIVIT-GFP was therefore specifically related to the inhibition of CACaN-dependent NFAT activation.
Interestingly, when the different spine types were compared, a decreased proportion of “mushroom” type spines was associated with CACaN+GFP (38% ±1.8%) compared to GFP (47% +/− 2.8%), WTCaN+GFP (44% ± 2.3%) or CACaN+VIVIT-GFP (46% +/−1.4%) (Fig. 1F), implicating calcineurin activation in the collapse of mature, mushroom-shaped spines that are known to be more stable than thin spines (Holtmaat et al., 2005).
As both spine loss and dendritic simplification were rescued when a genetically encoded VIVIT-GFP was overexpressed in neurons, we postulated that the morphological effects of CACaN were related to NFAT activation. We therefore investigated the specific effects of NFATc4 on these morphological parameters. We transduced 14DIV cortical neurons with either a lentiviral vector encoding a HA-tagged wild-type (WtNFAT) or constitutively activated (CANFAT) form of NFATc4 that lacks its N-terminus regulatory CaN binding domain. This truncated NFATc4 is therefore activated without calcineurin (Molkentin et al., 1998). In order to detect morphological variables, neurons were previously transfected with GFP at 5DIV and we verified that a high percentage (80%) of the GFP-filled cells were also positive for HA. As observed in Figure 2A, WtNFAT was mainly detected in the cytoplasm and was excluded from the nucleus (delimited by the marker of the nuclear membrane Laminin-B1) whereas, as expected, the constitutively active form of NFATc4 was concentrated in the nucleus. Importantly, CaN activity was not previously induced in these cells (either by using a calcium ionophore or CACaN overexpression), demonstrating that CANFAT did not depend upon any other upstream events to be activated. In contrast to CACaN-induced NFATc4 nuclear recruitment, accumulation of CANFAT into the nucleus could not be prevented by VIVIT-GFP (Fig. 2A). Overexpression of CANFAT led to a significant decrease of spine density (0.22 ± 0.045 spines/μm) and dendritic simplification compared with WtNFAT (0.39 ± 0.07 spines/μm). None of these changes could be improved by co-transfection with VIVIT-GFP (0.24 ± 0.023 spines/μm) (Fig. 2B–D), thus demonstrating that VIVIT acts upstream of NFAT activation and does not have an effect non-related to calcineurin. We therefore concluded that ectopic expression of a constitutively active form of NFATc4 is sufficient to induce spine loss and dendritic simplification in primary neurons, suggesting that NFATc4-related transcriptional events are critical for CaN-induced neurotoxicity in vitro.
We previously observed that both transgenic neurons cultured from Tg2576 embryos and wild-type neurons treated with Aβ-containing conditioned media develop AD-like pathological morphological changes that include dendritic spine loss and neuritic simplification. The involvement of CaN in these processes was suggested by the fact that both phenotypes could be rescued when FK506 was added to the medium or used in transgenic mice (Wu et al., 2010; Rozkalne et al., 2011). Here we showed that the overexpression of consitutively activated CaN or NFAT phenocopy the morphological changes induced by Aβ. Based on these observations, we asked whether the inhibition of NFAT activation by calcineurin can also alleviate Aβ synaptotoxicity.
We first tested the efficiency of VIVIT-GFP in transgenic neurons that were cultured from Tg2576 embryos. Because the mutated APP transgene was present as a heterozygous state, transgenic (Tg) or littermate (Wt) cultures could be analyzed in parallel. Cells were transfected with either GFP or VIVIT-GFP at 5DIV and analyzed at 18 DIV. A decreased density of dendritic spines was observed in Tg neurons (0.21 ± 0.11 spines/μm) compared with their wild-type counterpart (0.45 ± 0.1 spines/μm) (Fig. 3A, B). The spine loss detected in Tg neurons was also associated with a simplification of the dendritic tree (Fig. 3C). Importantly, when VIVIT-GFP was overexpressed in Tg neurons, both parameters were rescued to normal wild-type levels (spine density: 0.39 ± 0.12 spines/μm) and significantly improved compared with Tg neurons transfected with GFP alone.
To study the effect of exogenous Aβ peptides on neuronal morphological parameters, Aβ-containing conditioned medium (TgCM) was obtained by collecting the medium of Tg neurons after two weeks in culture and used to treat naive wild-type cells. The concentration of Aβ40 was estimated as 7000 pM and was comparable to the levels observed in AD brains (Ingelsson et al., 2004). Western blot analysis showed that TgCM mainly contained soluble oligomeric species (not shown) that have been described to have important neurotoxic effects (Walsh et al., 2002; Lesne et al., 2006; Shankar et al., 2007; Shankar et al., 2008). Wild-type primary neurons cultured in standard NB/B27-serum free medium for 16 DIV were exposed for 24hrs to either wild-type (WtCM) or transgenic (TgCM) conditioned media diluted 1:1 in the initial medium. As observed in Figure 3D and 3E, 24hrs exposure to Aβ was sufficient to induce a significant loss of spines in wild-type primary neurons (0.28 ± 0.075 spines/μm) compared to WtCM treatment (0.48 ± 0.12 spines/μm). However, no change of the dendritic complexity was observed, suggesting that such a short exposure with oligomeric Aβ could not affect the whole dendritic arbor of the cells. Aβ immunodepletion (using the anti-amyloid-β monoclonal antibody 6E10) significantly inhibited the effect of TgCM on spines (0.39 ± 0.09 spines/μm), demonstrating that the spine loss was directly related to the presence of amyloid β species (immunodepletion led to a 70% decrease in TgCM Aβ content). Remarkably, when cortical primary neurons were transfected with VIVIT-GFP, prior to exposure to TgCM, the effect on spine density was greatly ameliorated (0.41 ± 0.108 spines/μm), despite the persistence of Aβ species in the medium. This suggests that blocking CaN-mediated activation of NFAT protects against neurotoxic effects of soluble exogenous Aβ oligomers.
VIVIT was initially identified as a selective blocker of the CaN/NFAT interaction, mimicking the PxIxIT sequence of NFAT that is required to bind calcineurin. Although Aramburu and colleagues (Aramburu et al., 1999) showed that calcineurin activity on critical substrates (such as CREB) was not affected by VIVIT, several recent studies suggested that it can also disrupt the interaction between CaN and other important cytosolic targets implicated in the regulation of synaptic activity such as AKAP or cabin (Dell’Acqua et al., 2002; Liu, 2003). In an attempt to precisely decipher the role of the transcription factor NFAT in Aβ-related neurotoxic events, we used a nuclear-targeted version of VIVIT that was shown to have a restricted localization within the nucleus and that was previously used to specifically counteract NFATc4-related transcriptional activity (Schwartz et al., 2009). As a negative control, we fused a myristoylation (Myr) tag to the VIVIT-GFP plasmid in order to direct the localization of this peptide towards the cell membrane. Thus, these constructs would disambiguate the transcriptional effects of VIVIT from other non transcriptional functions at the spines. We first tested if these two constructs, NLS-VIVIT-GFP and Myr-VIVIT-GFP, showed the expected sub-cellular localization. At 5DIV, primary cortical neurons were transfected with GFP, NLS-VIVIT-GFP or Myr-VIVIT-GFP. As observed in Figure 4A, NLS-VIVIT-GFP was strictly restricted to the nuclear compartment and limited by the nuclear membrane marker Laminin B1. By contrast, the fluorescent signal of Myr-VIVIT-GFP was completely excluded from the nucleus. When each plasmid was transfected altogether with DsRed (as a common denominator to obtain the full morphology of each cell), both GFP and Myr-VIVIT-GFP could be detected in the processes of neurons and in the dendritic spines, which was not the case for NLS-VIVIT-GFP (Fig. 4B).
To determine the effect of NLS-VIVIT and Myr-VIVIT on NFAT transcriptional activity, primary neurons were transduced with an AAV-pNFAT-Luc reporter vector, altogether with AAV-WtCaN+GFP, AAV-CACaN+AAV-GFP, AAV-CACaN+AAV-NLS-VIVIT-GFP or AAV-CACaN+AAV-Myr-VIVIT-GFP. After three days, the luciferase activity was quantified. As previously observed, overexpression of a constitutively active form of calcineurin led to an increase of NFAT transcriptional activity compared with WtCaN. This increase was significantly diminished when NLS-VIVIT-GFP was overexpressed, but no change could be detected with Myr-VIVIT-GFP. This suggested that only a nuclear-targeted inhibitor was able to inhibit NFAT transcriptional activity in vitro (Fig. 5A). We next determined if CaN-induced morphological changes could be rescued by the different VIVIT-GFP constructs by co-transfecting neuronal cultures with either WtCaN or CACaN, altogether with GFP, NLS-VIVIT-GFP or Myr-VIVIT-GFP. Because dendrites and spines were not detectable from the GFP signal when NLS-VIVIT-GFP was overexpressed, we used DsRed as a common denominator to evaluate spine density and dendritic complexity. At 18DIV, both spine loss and dendritic simplification associated with CACaN overexpression were significantly improved when NLS-VIVIT-GFP was transfected (Fig. 5B–D). By contrast, no effect could be detected when Myr-VIVIT-GFP was used, consistent with the hypothesis that VIVIT’s neuroprotection against CACaN-mediated neurodegeneration are related to transcriptional effects in the nucleus rather than non-transcriptional effects in the cytoplasm or dendrites.
In the previous experiment, we determined that a nuclear version of VIVIT was almost as potent as a non-targeted VIVIT at blocking the pathological changes induced by CACaN. Considering the fact that calcium dyshomeostasis and CaN activation have been shown to be downstream events to Aβ neurotoxicity (Busche et al., 2008; Kuchibhotla et al., 2008), we next asked whether a nuclear-targeted version of this peptide may be protective against Aβ toxicity as well.
Following the same procedures as described above, we tested the NLS-VIVIT-GFP or Myr-VIVIT-GFP constructs in Wt and Tg primary neurons that were co-transfected with DsRed (Fig. 6A). Both spine density and dendritic complexity were significantly improved when NLS-VIVIT-GFP was overexpressed in Tg neurons, but no improvement was detected with Myr-VIVIT-GFP (Fig. 6B–C).
Similarly, when Wt primary neurons were transfected with GFP+DsRed, NLS-VIVIT-GFP+DsRed or Myr-VIVIT-GFP+DsRed and then treated for 24hrs with Aβ-containing TgCM (Fig. 6D), the TgCM-induced spine loss was attenuated by NLS-VIVIT-GFP (Fig. 6E), but no change could be observed in TgCM-treated neurons transfected with GFP or Myr-VIVIT-GFP. Again, this short-term exposure to Aβ oligomers did not affect the dendritic branching of the cortical neurons (not shown).
Taken together, these observations strongly suggest that NFAT-related transcriptional events play a role in Aβ-related neurotoxicity, thus predicting that inhibition of the CaN/NFAT pathway may be a protective strategy against Aβ-induced neuronal injuries in vivo.
In vitro, we observed that a selective disruption of the interaction between CaN and NFAT may alleviate Aβ-related neurotoxicity. We asked whether these neuroprotective effects could be reproduced in vivo. Neurites surrounding amyloid deposits have an increased calcium overload, decreased spine density and an increased probability to develop neuritic dystrophies (Spires et al., 2005; Kuchibhotla et al., 2008). Interestingly, overexpression of a constitutively activated CaN in vivo leads to similar abnormal neuronal changes (Wu et al., 2010). We therefore hypothesized that amyloid aggregates may be responsible for an abnormal increased activation of CaN/NFAT pathway in vivo, which would compromise neuronal integrity. We revisited this hypothesis by delivering VIVIT in the brains of APP/PS1 transgenic mice and examining its effects on neuritic degeneration.
To investigate the effect of VIVIT in vivo, AAV vectors encoding for VIVIT-GFP, NLS-VIVIT-GFP (both inhibitors having proven beneficial effects in vitro) and GFP (as a control) were stereotacticaly injected in the cortex of 7 month-old littermates or APP/PS1 mice, when amyloid deposits are already present. In the case of AAV-NLS-VIVIT-GFP injected mice, we first verified that the recombinant protein was properly addressed to the nuclear compartment and colocalized with Hoechst-stained nuclei (Hoechst solution was applied topically, Fig. 7A). AAV-NLS-VIVIT-GFP and AAV-GFP were then co-injected with a ratio of 3:1, so that a high proportion of GFP filled neurons were also transduced by the AAV-NLS-VIVIT-GFP vector. One month later, a cranial window was implanted. Neurites and spines were detected and quantified by multi-photon imaging in the living animal (Fig. 8A). High-magnification images were taken in order to visualize dendritic spines. Spine density was quantified for each neuritic segment using the Neuronstudio software (see Methods). Importantly, no significant difference was observed when comparing the spine density in wild-type littermate mice injected with either AAV-GFP (0.47 ± 0.11 spines/μm), AAV-VIVIT-GFP (0.49 ± 0.107 spines/μm) or AAV-NLS- VIVIT-GFP (0.53 ± 0.1 spines/μm) (not shown), indicating that VIVIT-GFP does not have an effect on spine density by itself, i.e. independently of Aβ.
Compared with dendrites from GFP-injected wild-type littermates (0.47 ± 0.11 spines/μm), dendrites in the vicinity of amyloid deposits (<100μm away from plaque) in APP/PS1 mice exhibited a decreased spine density (0.31 ± 0.12 spines/μm). This amyloid-associated spine loss was restored to essentially normal levels when either AAV-VIVIT-GFP (0.49 ± 0.13 spines/μm) or AAV-NLS-VIVIT-GFP+AAV-GFP (0.44 ± 0.1 spines/μm) were injected in APP/PS1 mice (Fig. 8B). In both APP/PS1 animals and human AD brains the spine densities are correlated with the distance from the edge of the amyloid deposits (correlation coefficient: 0.41) (Koffie et al., 2009) (Spires et al., 2005). However, this local effect of amyloid plaques was nearly abolished when AAV-NLS-VIVIT-GFP (correlation coefficient: 0.22) or AAV-VIVIT-GFP was expressed (correlation coefficient: 0.09; Fig. 8C).
Overall, the injection of AAV-VIVIT-GFP was associated with a marked recovery of spine density around amyloid deposits in vivo. A significant beneficial effect was also observed in AAV-NLS-VIVIT injected mice, even though this improvement did not reach the levels of VIVIT-GFP. This difference might be due to the fact that some of the GFP-filled neurites observed in animals co-injected with AAV-NLS-VIVIT-GFP and AAV-GFP had only been transduced by the later vector. In order to examine this possibility, we co-injected the AAV-NLS-VIVIT-GFP and an AAV-TdTomato with the same 3: 1 ratio as previously used, and we observed that 86% of the red fluorescent cells also contained a nuclear GFP signal (Fig. 7B), suggesting that the vast majority of transduced neurons expressed both recombinant proteins.
The presence of amyloid deposits not only affects the density of dendritic spines, but is also known to disturb the morphology of the neuritic shaft itself (Knowles et al., 1999). We therefore asked whether overexpression of VIVIT or NLS-VIVIT might have an additional rescue effect in regards to these morphological parameters. As previously shown, amyloid deposits are associated with the development of neuritic dystrophies, which can be visualized by AAV-GFP injection in APP/PS1 mice. Even though the size of the amyloid plaques that were analyzed was comparable between all the injected animals, plaques from AAV-VIVIT-treated mice were associated with fewer dystrophies (5 ± 4 dystrophies/plaque) compared with those from AAV-GFP (14 ± 9 dystrophies/plaque) and AAV-NLS-VIVIT (18 ± 10 dystrophies/plaque, Fig. 9A, B) injected mice. Because the density of GFP-filled neurons around plaques was similar among all the groups (not shown), we conclude that the difference in the number of dystrophies was not due to a decreased amount of GFP-filled neurites between AAV-GFP and AAV-VIVIT-GFP treated animals. This finding suggests that dystrophies, unlike dendritic spine loss, might not be recovered by a restricted inhibition of NFAT transcriptional activity in the nucleus. Interestingly, like VIVIT, calcineurin inhibition with FK506 led to an improvement in neuritic dystrophies (Rozkalne et al., 2011). The remaining dystrophies in AAV-VIVIT-treated mice were similar in size to those from AAV-GFP and AAV-NLS-VIVIT-injected mice.
A third morphological characteristic of neurites around amyloid deposits is a subtle change in their trajectories (Knowles et al., 1999). This abnormal neuritic curvature was assessed on paraffin-embedded sections after immunostaining for GFP and amyloid plaques. The curvature ratio was calculated by reporting the length of a neurite divided by the end-to-end length of the same segment. We observed that the average neuritic curvature of both AAV-VIVIT-GFP (1.054 ± 0.038) and AAV-NLS-VIVIT-GFP (1.059 ± 0.048) injected mice were improved around amyloid deposits compared with AAV-GFP-treated animals (1.081 ± 0.07) (Fig. 9C, D).
Using an in vivo imaging approach, we observed that several neuronal morphological parameters (i.e. spine density, neuritic dystrophies and curvature) that are abnormal in the vicinity of amyloid deposits were significantly improved when either AAV-VIVIT-GFP or AAV-NLS-VIVIT-GFP was injected in the cortex of APP/PS1 mice. To verify that these beneficial effects were related to the efficient inhibition of NFATc4 recruitment into the nucleus, we performed a post-mortem immunohistological analysis.
Brain sections of injected mice were stained to detect both GFP and endogenous NFATc4 in order to determine the ratio of NFATc4 in the nucleus vs cytoplasm in transduced neurons. We found that the distribution of NFATc4 nuclear/cytoplasmic ratios was shifted toward lower values in APP/PS1 mice injected with AAV-VIVIT-GFP treated animals (0.8 ± 0.08) compared with AAV-GFP (0.89 ± 0.06). AAV-NLS-VIVIT-GFP-injected mice exhibited an intermediate NFATc4 nuclear/cytoplasmic ratio between AAV-GFP and AAV-VIVIT-GFP-injected animals (0.83 ± 0.09) (Fig. 10A, B).
The biological mechanisms that sculpt the fine structure of the adult brain and their alterations in neurodegenerative diseases remain largely unknown. In this report we demonstrated that Aβ neurotoxic damage can be prevented or even reversed by inhibiting calcineurin-mediated activation of NFAT. Importantly, the same profound morphological changes occur when a constitutively active form of the transcription factor NFAT is introduced. The results are consistent with the hypothesis that the neurotoxic effects of Aβ are mediated, at least in part, by the transcriptional activation of NFAT downstream target genes. Although the contribution of NFAT to neurological function or dysfunction has received little attention, recent findings have demonstrated a role of NFAT in synaptic plasticity during development (Nguyen and Di Giovanni, 2008; Schwartz et al., 2009). Moreover, NFAT shows an abnormal nuclear accumulation in the brains of AD patients (Wu et al., 2010), which directly correlates with the levels of soluble Aβ(1–42) and the severity of cognitive impairment (Abdul et al., 2009). Our study provides evidence supporting the idea that NFAT is a specific molecular mediator of transcription-dependent modifications of the CNS structure in the context of Alzheimer’s disease.
In vitro, we observed that the morphological changes induced by constitutively active calcineurin or by constitutively active NFAT replicate the decrease of spine density and dendritic simplification induced by the neurotoxic Aβ peptides. Along with recent data indicating that calcineurin is activated both in human AD brain and in transgenic models of AD (Liu et al., 2005; Dineley et al., 2007; Wu et al., 2010), we tested the hypothesis that inhibition of the excessive NFAT activity might improve the neuritic abnormalities observed in AD models. We demonstrate that VIVIT, a genetically encoded inhibitor of the interaction between calcineurin and NFAT, increases both the spine density and the complexity of dendritic arbors in APPswe-transgenic neurons or in TgCM-treated wild-type neurons. In vivo, a gene transfer approach using adeno-associated vectors enabled the delivery of VIVIT to neurons and significantly increased the spine density in the vicinity of senile plaques. Transduction with an inhibitor that, due to a nuclear localization signal (NLS-VIVIT), specifically blocked the nuclear activation of NFAT, was also able to restore to nearly normal the morphological abnormalities associated with Aβ. The ability of NLS-VIVIT to reproduce the same beneficial effects as VIVIT is intriguing, as this peptide is thought to interact primarily with the docking site of NFAT upon CaN. However, several studies demonstrated that activated CaN is able to translocate to the nucleus in neurons (Pujol et al., 1993; Sola et al., 1999; Yang et al., 2005; Schwartz et al., 2009), a phenomenon we also previously described in AD patients (Wu et al., 2010). We therefore propose that both CaN and NFAT translocate to the nuclear compartment, where VIVIT would compete with NFAT for CaN binding. Considering the higher affinity of VIVIT for CaN (about 25-fold compared with NFAT (Aramburu et al., 1999)), the CaN/NFAT interaction would thus be interrupted and NFAT would rapidly be phosphorylated and shuttled back to the cytoplasm. This hypothesis is in agreement with the previous observation that NLS-VIVIT efficiently inhibited the expression of NFAT-target genes (Schwartz et al., 2009), and with our observation of a significant decrease in the NFAT nuclear/cytoplasmic ratio not only in AAV-VIVIT-GFP-injected animals but also in AAV-NLS-VIVIT-treated mice.
Our results indicate that (1) NFAT activation is a likely consequence of Aβ accumulation in Alzheimer’s disease, as suggested by its prominent nuclear localization in the brain of AD patients (Abdul et al., 2009; Wu et al., 2010), and by the ability of VIVIT to restore Aβ-associated morphological neurodegenerative changes; (2) once activated and translocated to the nucleus, NFAT presumably induces the transcription of target genes in mature neurons that lead to a pathological remodeling of dendrites and dendritic spines. The beneficial effect of a nucleus-directed NFAT inhibitor supports the role of transcriptional events in the regulation of dendritic spine stability in vitro and in vivo. To some extent, these findings are unexpected, as the modulation of spine morphology has been largely focused on endocytosis and stability of AMPAR and NMDAR receptors, rather than transcriptionally regulated phenomena (Lau and Zukin, 2007; Zhang et al., 2008; Bhattacharyya et al., 2009; Goebel-Goody et al., 2009). Although local spine-specific phenomena will ultimately determine the stability of individual spines, the current data emphasize that dendritic trees and spine density are also impacted at a transcriptional level. This conclusion is in accordance with a growing literature suggesting that transcriptional modulation can affect memory function (Saura and Valero, 2011), and with recent data demonstrating that alterations in mRNA trafficking and stability can lead to neural system alterations and degeneration in disorders as disparate as autism and amyotrophic lateral sclerosis (Gatto and Broadie, 2010).
The literature addressing the question of NFAT target genes specifically regulated in neurons is quite limited, but few interesting findings might be relevant in the context of Alzheimer’s disease. For example, NFAT was shown to contribute to the induction of apoptosis in neuroblastoma cells through up-regulation of Fas Ligand (Luoma and Zirpel, 2008; Alvarez et al., 2011), a gene that was also reported to increase in AD brains, especially in the hippocampal formation and entorhinal cortex that are primarily involved in memory encoding. In addition, Aβ peptides induce a pathological increase in Fas-L in cortical primary neurons and its accumulation was demonstrated in neuritic dystrophies surrounding the amyloid deposits (Su et al., 2003). Even though no direct connection between Fas-L and spine morphology has been reported so far, one of the downstream target of Fas-L, capsase-3, was recently shown to trigger early synaptic dysfunction and spine loss in AD (D’Amelio et al., 2011). This would suggest that activation of caspase-3 may not only be involved in cell death but is also closely associated with the regulation of synaptic plasticity. It is therefore possible that the activation of Fas-L expression by NFAT indirectly causes a synaptic failure via the induction of a non-apoptotic caspase-3 pathway. The expression of the potassium channel Kv2.1 can also be driven by NFAT and was shown to be upregulated by 72% in the hippocampus of rat injected with Aβ25–35. (Pan et al., 2004). Interestingly, the use of several potassium channel openers (minoxidil, pinacidil, cromakalim) can induce amnesia in mice (Ghelardini et al., 1998), thus suggesting that the regulation of the activity of the potassium channels might impair memory encoding, one of the best hallmark of AD. As learning and memory deficits mainly reflects a default in spine–mediated plasticity, we can hypothesize that disregulation of the potassium channel Kv2.1 might lead to the alteration of dendritic spine morphology. Lastly, the up-regulation of the inositol 1,4,5-trisphosphate type 1 receptor (InsP3R) gene by NFAT is of particular interest in the context of Alzheimer’s disease (Genazzani et al., 1999; Amberg et al., 2004). Indeed, InsP3R was shown to directly interact with presenilins 1 and 2, increasing its activity in response to Inositol-3-phosphate, thus leading to an excessive release of Ca2+ (Muller et al., 2011). The up-regulation of InsP3R might therefore participate to exacerbate the increase of the resting calcium levels. The few examples mentioned above might be of interest in the context of AD, but other NFAT target genes previously identified does not seem relevant in a pathological context. Indeed, NFAT transcriptional activity was associated with an increased expression of the neurotrophic factor BDNF in Purkinje cells, an important mediator of axonal outgrowth during development (Graef et al., 2003). These varied results could be explained by the ability of NFAT to interact with other transcriptional factors (Crabtree and Olson, 2002; Ogata et al., 2003; Kao et al., 2009), a mechanism by which the pattern of NFAT-induced target genes can be modulated depending on the developmental state, on the cell type and on the physiological or pathological context. It is therefore conceivable that a particular set of NFAT co-activators are present in cortical neurons exposed to Aβ, so that its activation leads to the up-regulation of genes implicated in dendritic spine shrinkage. Although the transcriptional effects of NFAT (and its inhibition) await further investigation, the present study provides a novel neuroprotective therapeutic target against the downstream neurotoxic effects of amyloid-β and confirms the feasibility to pharmacologically arrest the “amyloid cascade” of neurodegeneration at a step after amyloid deposition has already occurred. The latter has critical therapeutic implications, since most amyloid deposition is thought to occur years or even decades prior to the onset of cognitive decline, and therefore, prior to AD diagnosis, whereas synaptic and neuron loss largely occurs in a second stage and correlates with the severity of cognitive impairment (Sperling et al., 2009). In this scenario, amyloid-directed therapies would not be sufficiently effective once cognitive symptoms (and the underlying neurodegenerative processes) have begun. Consequently, combined synergistic therapies aimed at reducing Aβ content (either by inhibiting the production of Aβ peptides or by increasing their clearance) and at restoring Aβ-induced neural damage would be needed.
This work was supported by National Institutes of Health Grants AG08487, EB000768, and EY13399, the Lefler fellowship (Harvard Medical School) and the French Bettencourt-Schueller award for young scientist. We thank Dr. Norris (Sanders-Brown Center On Aging, Lexington, KY) for providing the VIVIT-GFP constructs and Dr Ruthazer (Montreal Neurological Institute, McGill University, Montreal, QC, Canada) for providing the NLS-VIVIT-GFP plasmid. We thank Dr Alberto Serrano-Pozo for helpful comments during manuscript preparation and Daniel Joyner for primary neuron preparation.
The authors declare no conflict of interest.