Calcineurin (CaN) activation is critically involved in the regulation of spine morphology in response to oligomeric amyloid-β (Aβ) as well as in synaptic plasticity in normal memory, but no existing techniques can monitor the spatiotemporal pattern of CaN activity. Here, we use a spectral fluorescence resonance energy transfer approach to monitor CaN activation dynamics in real time with subcellular resolution. When oligomeric Aβ derived from Tg2576 murine transgenic neurons or human AD brains were applied to wild-type murine primary cortical neurons, we observe a dynamic progression of CaN activation within minutes, first in dendritic spines, and then in the cytoplasm and, in hours, in the nucleus. CaN activation in spines leads to rapid but reversible morphological changes in spines and in postsynaptic proteins; longer exposure leads to NFAT (nuclear factor of activated T-cells) translocation to the nucleus and frank spine loss. These results provide a framework for understanding the role of calcineurin in synaptic alterations associated with AD pathogenesis.
Calcineurin (CaN) is a calcium-dependent serine/threonine phosphatase that regulates ion channels, cytoskeletal proteins, and transcription factors (Klee et al., 1979; Halpain et al., 1998; Berridge et al., 2000; Winder and Sweatt, 2001). CaN plays fundamental roles in neuronal development, regulating axonal cone outgrowth (Lautermilch and Spitzer, 2000), dendritic complexity, and synaptic function (Schwartz et al., 2009). Alteration of CaN activity was also reported in neuropathological contexts such as Parkinson's disease (Martin et al., 2012) and Alzheimer's disease (Liu et al., 2005).
Several biochemical methods exist to monitor CaN activity (Blumenthal et al., 1986; Enz et al., 1994; Sellar et al., 2006), but they suffer from a lack of anatomical precision and specificity. Translocation of dephosphorylated nuclear factor of activated T-cells (NFAT) to the nucleus is also used to assess CaN activity, but this indirect method requires stimulation for hours and is sensitive to the balance between NFAT dephosphorylation (by CaN) and phosphorylation [by GSK3β (Newman and Zhang, 2008)]. CaN has a wide spectrum of substrates in neurons such as cofilin, AMPA receptors, BAD (Bcl2 antagonist of cell death), and NFAT (Springer et al., 2000; Tavalin et al., 2002; Morishita et al., 2005; Wang et al., 2005; Wu et al., 2010). These downstream targets are localized in different cellular compartments, but current approaches cannot address the dynamics of CaN activation at a subcellular level, and are not suitable to temporally follow the changes in CaN activity.
We postulated that fluorescence resonance energy transfer (FRET) assays, increasingly used to detect molecular interactions in cells with spatiotemporal resolution (Stryer, 1978; Selvin, 2000; Jares-Erijman and Jovin, 2003), might be capable of measuring CaN activation. CaN phosphatase activity is physiologically regulated by calcium, CaN B subunit (CaNB) and calmodulin (CaM) (Klee et al., 1979; Aramburu et al., 2000). Under basal conditions, CaN is inactivated by a regulatory inhibitory domain. After an intracellular calcium influx, CaM and CaNB bind to their target sites on CaNA, triggering a conformational change leading to the displacement of the autoinhibitory domain from the enzymatic site (Wang et al., 1989; Klee et al., 1998). Based on CaNA interactions with CaNB and CaM, we developed a FRET assay to measure CaNA–CaNB or CaNA–CaM interaction or conformational change using either endogenous proteins or fluorescent tagged molecules.
We show that FRET assays can efficiently monitor CaN activation in different subcellular compartments. Imaging two-color fluorescence for CaNA–CaM or CaNA–CaNB pairs allowed us to interrogate the activation status of CaN in single neurons. We observed low levels of activity in the resting state. After a brief exposure to oligomeric amyloid-β (Aβ) prepared from Tg2576 conditioned media or purified from human AD cortex, we detected a rapid CaN activation in spines with a later activation in the cytoplasm and, ultimately, in the nucleus. Washout of the oligomeric Aβ prevented progression to the cytoplasm or nucleus. The rapid activation of CaN in spines alters spine morphology and reduces postsynaptic proteins including F-actin and GluR1-containing AMPA receptors. The distinct spatial and temporal distribution of CaN activation supports the idea that metabolic processes in different cellular compartments are selectively targeted by CaN during Aβ neurotoxicity.
Materials and Methods
Primary neuronal cultures.
All animals were used in accordance with animal protocols approved by the Institutional Animal Care and Use Committee. Neuronal cultures were generated using transgenic embryos from either sex expressing human mutated APP gene [Tg2576 line: transgenic mice overexpressing the 695 aa isoform of human Alzheimer β-amyloid precursor protein containing the double Swedish mutation K670N, M671L with a hamster prion protein gene promoter in B6;SJL F2 mice (Hsiao et al., 1996)]. Primary neuronal cultures were derived from cerebral cortex of embryonic day 15–19 Tg2576 mice (Charles River Laboratories), as described previously with modifications (Wu et al., 2004). Briefly, cortices were dissected, gently minced, trypsinized (0.027%, 37°C; 5% CO2 for 15 min), and then washed with 1× HBSS. Neurons were seeded to a density of 4 × 105 viable cells/35 mm culture dishes previously coated with poly-d-lysine (100 μg/ml) for at least 1 h at 37°C. Cultures were maintained at 37°C with 5% CO2, supplemented with Neurobasal medium with 2% B27 nutrient, 2 mm l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. The cultures were used within 28 d in vitro (DIV). To maintain elevated levels of extracellular Aβ, media were not changed. To identify the genotype of the animals, we used PCR on DNA extracted from sample tail taken after dissection of the cerebral cortex.
The cDNA of human CaM (NM-001743) was cloned into peGFP-C1 vector (Clontech) using BamH1 restriction site. RFP and mouse CaN were cloned into peGFP-N1 vector (Clontech) using Kpn1/Age1 (RFP) and Age1/Not1(mCaN), respectively. Authenticity of all constructs was confirmed by sequencing.
Immunostaining of HEK cells and cortical primary neurons.
To detect endogenous CaNA, CaNB and CaM, HEK cells or primary neurons were fixed with 4% paraformaldehyde for 15 min and were permeabilized with 0.5% Triton X-100 in PBS for 5 min. After blocking with 3% bovine serum albumin (BSA) at 37°C for 1 h, cells were incubated with primary antibodies: anti-CaNA (1:200; Calbiochem), anti-CaM antibody (1:200; Calbiochem), or anti-CaNB antibody (1:200; Millipore) overnight at 4°C. Cells were then washed four times and incubated with secondary antibodies conjugated to either cyanine 3 (Cy3) (1:500; Jackson ImmunoResearch) or Alexa 488 (1:500; Invitrogen). Each antibody was diluted in PBS containing 1.5% BSA.
Surface AMPA receptors were measured as described below. In brief, 21 DIV primary neurons treated with wild-type or TgCM for 60 min were fixed in 4% paraformaldehyde (20 min; room temperature), but not permeabilized. Following an incubation with 3% bovine serum albumin (1 h) to block nonspecific staining, neurons were incubated with anti-NT-GluR1 (1:500; Millipore) overnight at 4°C. After three washes, neurons were incubated with Cy3-conjugated secondary antibody (Jackson ImmunoResearch) for 2 h at room temperature. After washing in PBS three times, the coverslips were mounted on slides with Vectashield mounting medium. For staining of F-actin, neurons were incubated with Oregon Green 488 phalloidin (1 U/ml; Invitrogen) at room temperature for 20 min. Fluorescent images were captured using an LSM 510 Carl Zeiss microscope with a 25× or a 63× water-immersion objective lens [numerical aperture (NA), 0.8 and 1.2, respectively]. All images were taken at 512 × 512 pixel resolution and quantified with the ImageJ software. To define dendritic clusters, a single threshold was chosen manually so that clusters corresponded to puncta of twofold to threefold greater intensity than the diffuse fluorescence on the dendritic shaft. On each coverslip, the cluster density, size, and fluorescence intensity of four to six GFP-positive neurons (two to three dendritic segments of ∼50 μm/neuron) were compared. Quantitative analyses were conducted without knowledge of the experimental conditions.
Quantitative spectral FRET assays.
Spectral FRET was performed as described previously (Uemura et al., 2009). Briefly, HEK293 cells in 35 mm glass bottom dishes were either transfected with CaNA-RFP and CaM-GFP fused constructs, or immunolabeled for endogenous CaNA, CaNB, and CaM. Cells were imaged 24 h after the transfection. Spectral imaging was performed on a Zeiss LSM510 microscope equipped with a 63× water-immersion objective lens (NA, 1.2) and with 37°C heating chamber containing 5% CO2. The wavelengths used for FRET imaging were 488 nm (excitation of GFP), separated by 488/563 dichroic mirror, and emitted fluorescence was detected by seven channels of the Metadetector with 21.4 nm spectral bandwidth for each channel that was set to detect fluorescence in the 502–651 nm wavelength range. When FRET occurs, the energy is transferred from GFP (donor) to RFP (acceptor); thus, the GFP emission peak weakens, and the fluorescence intensity of RFP increases. Analogous measures were made for Alexa 488 (donor) and Cy3 (acceptor) in immunofluorescent-based assays (Knowles et al., 1999). We used the ratio of 565 to 522 nm as readout of the FRET efficiency, which reflects the relative proximity between each two fluorophores. The ratio value was calculated by dividing the average pixel fluorescence intensity at 565 nm by the intensity at 522 nm. The ratiometric image was produced by dividing the intensity of the image from the 565 nm spectral window by the intensity from the 522 nm window after background subtraction, using ImageJ. For representation of the data, a look-up table was applied by mapping image values to color scale (16 colors) resulting in pseudocolor image. We used 10 and 25× objectives for the average-per-cell analysis, and 63× objective for studying subcellular distribution of CaN activation.
A photobleach experiment was performed as a control for FRET. HEK cells were transfected with GFP-CaM and RFP-CaNA using the same amount of each plasmid. After 2 d, ionomycin (or DMSO) was applied for 30 min to induce CaN activation and cells were then fixed with 4% paraformaldehyde for 15 min. First, the intensity of the GFP signal was measured in a small cytoplasmic region of interest (ROI) of the cell. To bleach the acceptor signal, the selected ROI was scanned with 100% laser at 543 nm for 5 min. The photobleach efficiency was confirmed by the loss of any fluorescent signal at 565 nm, and the intensity of the GFP fluorophore in the same area was measured again. To determine whether the photobleach of the RFP signal leads to an increase of the donor fluorescence intensity, the ratio of the GFP signal after and before photobleach was calculated. This experiment showed enhanced GFP signal within the acceptor photobleached area, consistent with dequenching of the donor fluorophore and FRET. As an internal control, the GFP signal of another ROI where the RFP signal was not photobleached was unchanged.
Fluorescence lifetime imaging.
Fluorescence lifetime imaging (FLIM) assay was performed in HEK cells and cultured neurons that were previously stained for endogenous CaNA and CaM, as previously described. Briefly, cells were imaged on a Zeiss LCM510 microscope with a 63× water-immersion objective lens (NA, 1.2) and a Chameleon pulsed laser was used to excite Alexa 488 donor fluorophore (two-photon excitation wavelength at 840 nm). Becker & Hickl FLIM hardware and software (Becker & Hickl) were used to acquire the donor lifetime information. Data analysis was performed using SPC Image (Becker & Hickl), in which the donor fluorophore lifetimes are determined by fitting the data to one (negative control) or two (experimental conditions) exponential decay curves. The donor lifetime information was color coded and displayed as pseudocolored images (Berezovska et al., 2005).
Analysis of spine density and spine morphology.
Cultured neurons (7 DIV) were transfected with green fluorescent protein and examined at 21 DIV using a LSM 510 Zeiss confocal microscope (excitation laser at 488 nm). Each image consisted of a z stack of pictures taken at a depth interval of 0.8 μm and then projected into one image. The densities of filopodia and spines were measured separately. For the analysis of density, filopodia or spines were defined as dendritic protrusions of 0.5–3 μm in length, with (spine) or without (filopodia) a head. Density was calculated by dividing the filopodia or spine number by the measured length of the correspondent neurite they were connected to. Filopodia and spines were specifically evaluated in the distal dendrites of the neurons. In each experiment, >200 spines in >12 neurons were examined for each treatment and three experiments were performed. For each condition, measurements of spine densities were averaged per neuron; the means from multiple neurons were then averaged to obtain the mean ± SEM for the population of neurons.
Size exclusion chromatography and quantification of Aβ.
As described previously (Hashimoto et al., 2002), 1 g of cortical gray matter from frontal lobe of AD patient brain or non-demented control brain (from either sex) was homogenized in 5 vol of TBSI (Tris-buffered saline with protease inhibitor mixture; Roche) with 25 strokes on a mechanical Dounce homogenizer and centrifuged at 260,000 × g for 20 min at 4°C. The supernatant was used as TBS-soluble fraction. A volume of 750 μl of TBS-soluble fraction of brain was separated by size exclusion chromatography (SEC) on Superdex 75 10/300 GL column (GE Healthcare) in 50 mm ammonium acetate, pH 8.5, with AKTA purifier 10 (GE Healthcare) (Townsend et al., 2006). To quantify Aβ40 and Aβ42, eluted fractions were diluted and subjected to BNT77/BA27 for Aβ40, or BNT77/BC05 for Aβ42 two-site ELISAs (Wako Chemicals).
Data are reported as mean ± SEM, and statistical significance was defined at p < 0.05. We made comparisons between groups by one- or two-way ANOVA, followed by post hoc Bonferroni's test for comparison among means.
Characterization of Ca2+-induced CaNA–CaM or CaNA–CaNB interaction by spectral FRET in HEK cells
We first examined the interaction of CaNA and CaM in HEK cells, in which CaN activation can be induced by application of a Ca2+ ionophore (Bittinger et al., 2004; Park et al., 2006). If CaN can be activated by transient increases in intracellular Ca2+ following ionomycin application, the interaction of CaNA with CaM should take place because activation of CaN requires binding of CaM (Wang et al., 1989; Klee et al., 1998). We first examined the interaction of CaNA–CaM in the presence of elevated intracellular Ca2+ in living HEK cells expressing GFP-CaM and RFP-CaNA (Fig. 1A,B). To measure spectral FRET, cells were excited with 488 nm laser (donor), and the emission of the acceptor RFP was recorded at 565 nm (emission peak of RFP; Fig. 1A). If CaNA and CaM physically interact (within 10–50 Å), an excitation light at 488 nm should result in a significantly enhanced RFP emission at 565 nm, as a result of the energy transfer from the GFP-CaM donor to RFP-CaNA acceptor. We determined the ratio of fluorescence intensity between 565 to 522 nm as readout for the spectral FRET efficiency: higher 565/522 ratio indicates higher FRET efficiency, and hence increased interaction. Figure 1B shows images of HEK cell expressing GFP-CaM and RFP-CaNA before and after stimulation with 2 μm ionomycin, in which the RFP/GFP emission ratio is represented as pseudocolored images. In cells coexpressing GFP-CaM and RFP-CaNA, high FRET was observed after ionomycin treatment, whereas lower but detectable FRET was observed before ionomycin treatment (Fig. 1B,C). Cells expressing the fluorescent donor CAM-GFP alone showed no change in the 565/522 ratio after stimulation (Fig. 1C), indicating that the change in fluorescence observed at 565 nm emission wavelength is due to the energy transfer from the GFP fluorophore to the RFP fluorophore fused to their respective proteins. As an additional technical control to ensure that the 565/522 ratio change reflected FRET, a photobleach experiment of the acceptor fluorophore was performed. After treatment with ionomycin, the RFP-CaNA was photobleached by scanning a small ROI with 100% laser at 543 nm for 5 min. An increased intensity of the donor fluorophore due to donor dequenching (Fig. 1D) was detected in the same ROI, thus confirming that any increase in the 565/522 ratio is associated with a transfer of energy from the donor to the acceptor fluorophore after stimulation with ionomycin. However, the GFP-CaM signal intensity remained stable when the RFP-CaNA fluorophore was not bleached.
We further confirmed the specificity of the interaction of CaNA–CaM or CaNA–CaNB in a native cellular environment by performing an immunolabeling FRET experiment (Knowles et al., 1999). Endogenous CaNA was immunolabeled with Alexa 488 as a donor fluorophore, and endogenous CaM was immunolabeled with Cy3 as an acceptor fluorophore. HEK cells treated with ionomycin showed a significantly higher 565/522 ratio compared with control (Fig. 1E). An alternative technique to measure FRET, FLIM, showed consistent results. In cells treated with ionomycin, the lifetime of Alexa 488-immunolabeled CaNA became significantly shorter compared with the lifetime in cells treated with DMSO, reflecting a transfer of energy between the donor and the acceptor (Fig. 1F). This technique relies only on the characteristics of the donor fluorophore, thus confirming that the results obtained with the ratiometric spectra approach reflects FRET interactions rather than spectral artifacts such as bleed through (Berezovska et al., 2003).
Together, the different FRET approaches used to detect interactions of CaNA–CaM demonstrate that change in FRET efficiency is sufficient to distinguish basal cellular CaN activity from stimulated states. The methods we used here report FRET signal consistently and are useful to monitor live-cell, real-time interactions of RFP-CaNA and GFP-CaM as well as interactions of the endogenous CaNA and CaM proteins, thus providing a great deal of flexibility in assay design.
CaN activation in an in vitro model of AD can be reported at a subcellular level by FRET assay
Previously, we have shown that, compared with wild-type cultured neurons, CaN activity measured by NFATc4 translocation to the nucleus is chronically elevated in neurons prepared from transgenic embryos expressing APP with the familial Swedish mutation (Tg2576 neurons) (Hsiao et al., 1996; Wu et al., 2010). This effect on NFAT nuclear localization could be blocked by immunodepletion of Aβ or treatment with CaN inhibitors (Wu et al., 2010). While these data support the idea that soluble oligomeric Aβ peptides induce CaN activation, they provide little spatial or temporal resolution. To test whether the FRET assay is sensitive enough to detect changes in interaction between CaNA–CaM or CaNA–CaNB under these circumstances, the FRET signal was compared between wild-type and Tg2576 neurons under basal conditions. In primary cultured neurons stained for endogenous CaNA–CaM or CaNA–CaNB, we observed a significant increase in the 565/522 ratio in cultured Tg2576 neurons compared with wild-type cultured neurons (Fig. 2). In wild-type neurons, the pseudocolored image, in which blue represents low FRET signal and red represents high FRET signal, shows only a very faint FRET signal (Fig. 2A,E). However, in accord with our previous observations of NFAT localization in the nucleus, Tg2576 cultured neurons show a strong FRET signal analyzed by both spectral FRET and FLIM-based FRET, indicating an interaction between CaNA and CaM (Fig. 2A–D). The spectral FRET assay indicates substantially increased interaction between CaNA–CaM, confirmed by similar data using the CaNA–CaNB pair in Tg2576 neurons (Fig. 2E,F). Moreover, the enhanced interaction between CaNA and CaM or CaNA and CaNB in Tg2576 neurons was distributed in the cytoplasm, nucleus, and both axon and dendritic processes, but the extent of activation was different in the three compartments. Quantitatively, there was a trend toward having a greater elevation of CaN activity in dendritic spines (1.2-fold for CaNA–CaM; 1.5-fold for CaNA–CaNB; n ≥ 50 neurons and n ≥ 200 spines in each condition; Fig. 2B,F) than in both the cytosol and nuclei (1.1-fold for CaNA–CaM; 1.3-fold for CaNA–CaNB).
Differential kinetics of CaN activation in dendritic spines and soma after applying conditioned media from Tg2576 neurons
We recently reported that exposure of wild-type neurons for 24 h to Aβ-containing conditioned media prepared from Tg2576 culture is capable of inducing CaN activation using NFAT nuclear translocation as a readout of CaN activation (Wu et al., 2010). NFAT translocation could be blocked by immunodepletion of the media with anti-Aβ antibody or by blocking CaN with the pharmacological inhibitor (3S,4R,5S,8R,9E,12S,14S,15R,16S,18R,19R,26aS)-5,6,8,11,12,13,14,15,16,17,18,19,24,25,26,26a-hexadecahydro-5,19-dihydroxy-3-[(1E)-2-[(1R,3R,4R)-4-hydroxy-3-methoxycyclohexyl]-1-methylethenyl]-14,16-dimethoxy-4,10,12,18-tetramethyl-8-(2-propen-1-yl)-15,19-epoxy-3H-pyrido[2,1-c][1,4]oxaazacyclotricosine-1,7,20,21(4H,23H)tetrone (FK506). To characterize the precise spatial and temporal patterns of CaN activation, wild-type cultured neurons were exposed to Aβ-containing conditioned media prepared from Tg2576 neurons. Consistent with our previous finding, application of Aβ-containing conditioned media onto wild-type neurons for 24 h caused a marked increase in the 565/522 ratio of both CaNA–CaM and CaNA–CaNB (Fig. 3A,B), but no changes were observed in neurons exposed to conditioned media prepared from wild-type cultures. Importantly, significantly enhanced 565/522 ratio was detected in the nucleus, cytoplasm, and dendritic spines. This finding shows that application of Aβ-containing TgCM causes CaN activation in all neuron compartments. This activation of CaN was directly related to the presence of Aβ, as immunodepletion of Aβ from Tg2576 conditioned media with 3D6, a specific antibody to Aβ, prevented the increase in the 565/522 ratio of both CaNA–CaM and CaNA–CaNB (Fig. 3A,B).
We next compared the spatiotemporal dynamics of CaN activation during Aβ exposure in living neurons expressing GFP-CaM and RFP-CaNA. Figure 4C shows pseudocolor images of a wild-type neuron expressing GFP-CaM and RFP-CaNA before (0 min) and at 15, 30, and 60 min after application of TgCM. Exposure to TgCM triggered a rapid increase in the interactions between GFP-CaM and RFP-CaN in dendritic spines (Fig. 4D), whereas changes in the cytosolic and nuclear compartments were not detected for at least 60 min (Fig. 4C). Importantly, no change in FRET signal was observed after exposure to conditioned media from wild-type cultures (Fig. 4A,B).
The rapid CaN activation in spines in response to TgCM application was further verified by using endogenous CaNA-Alexa 488 and CaNB-Cy3 or CaN-Alexa 488 and CaM-Cy3 as FRET pairs. Significant increases in FRET between endogenous CaNA and CaNB or CaNA and CaM in dendritic spines were observed as early as 15 min after TgCM application (the earliest point examined) and reached their peak after ∼30 min (Fig. 5A,B,D,E). The level of CaNA–CaNB or CaNA–CaM interaction persisted for up to 60 min with no sign of decline. Increases in the interaction of CaNA–CaM or CaNA–CaNB in spines indicate that a rapid and sustained CaNA activation was induced by TgCM. Comparing histograms of high FRET events among different time points for experiments across many spines revealed that FRET efficiency gradually and stably increased over a period of 0–60 min in most spine populations (Fig. 5C,F). High FRET observed in spines from CaNA–CaM or CaNA–CaNB within minutes of TgCM exposure indicates a rapid CaN activation in response to acute Aβ stimulation. The intensity of the 565/522 ratio of both CaNA–CaM and CaNA–CaNB in the cytoplasm and nucleus was unchanged at 60 min after Aβ application but significantly increased at 6 h (FRET ratio increase for the CaNA–CaM pair: cytoplasm, 0.53 ± 0.017; nucleus, 0.50 ± 0.014; FRET ratio increase for the CaNA–CaNB pair: cytoplasm, 0.51 ± 0.008; nucleus, 0.49 ± 0.006; p < 0.001 compared with correspondent baseline; Fig. 6A,B).
Because the role of CaN in neurons involves both structural and transcription factor targets, the subcellularly specialized pattern of CaN activation in response to Aβ-containing TgCM stimulation may suggest that local posttranslational effects regulated by CaN in the spine occur essentially immediately, whereas CaN-mediated transcriptional effects in the nucleus may be a delayed phenomenon.
To further dissect early and late events, CaN activation was measured by FRET in neurons exposed to Aβ-containing conditioned media for 60 min, a period that induces significant CaN activation in the spine, and then washed off, and neurons were recovered in their original media for another 24 h. As expected, neurons maintained in Aβ-containing conditioned media for the entire 24 h had significantly increased FRET ratios of both CaNA–CaM and CaNA–CaNB pairs (Fig. 6C,D). However, neurons exposed to Aβ-containing conditioned media for 60 min, following washing off and incubation for 24 h in original conditioned media, showed basal levels of FRET signal in all three compartments. Thus, the CaN activation induced by Aβ is reversible if Aβ is removed after relatively brief exposure, and the progressive activation of CaN in the cytoplasm and nucleus can be aborted.
Morphological consequences of rapid CaN activation in spines after applying Aβ-containing TgCM
CaN function in neurons has been implicated in various aspects of synaptic plasticity, including regulation of dendritic spine morphology (Halpain et al., 1998; Zhou et al., 2004; Wu et al., 2010). We previously observed that application of TgCM for 24 h decreases spine density in cultured neurons (Wu et al., 2010). To examine whether the rapid CaN activation in spines leads to spine loss or change in morphology, we compared the density of mature spines and of filopodia-like protrusions (evaluated independently) in mature cultured cortical neurons (21 DIV) 60 min after treatment with WtCM or TgCM. Quantitative analysis indicated that spine density in neurons treated with TgCM for 60 min was comparable with neurons treated with wild-type CM (Fig. 7A,B). In contrast to the relatively normal spine density, spine morphology was noticeably altered in neurons treated with TgCM (Fig. 7A,C). Neurons treated with wild-type CM for 60 min had the majority of spines with mushroom shape similar to those spines before treatment. Treatment with TgCM for 60 min led to a significant increase in the density of filopodia-like protrusions (Fig. 7C). The average number of filopodia-like protrusions versus total spines was 11% in neurons (21 DIV) before TgCM application, but dramatically increased to 24% at 15 min, 32% at 30 min, and 45% at 60 min after TgCM treatment, respectively (Fig. 7D). Immunodepletion of Aβ from TgCM with 3D6 or inhibition of CaN with its specific inhibitor FK506 markedly reduced the increase in number of filopodia-like protrusions (Fig. 7C), demonstrating that the effect of Tg2576 conditioned media is mediated by Aβ and consequent to activation of CaN. Treatment with VIVIT, a specific inhibitor for NFAT activation, had no effects on the increase in number of filopodia-like protrusions, suggesting that CaN-mediated NFAT activation does not participate in the acute spine morphological alteration [unlike the effective protection from spine loss at 24 h due to NFAT blockade with VIVIT (Wu et al., 2010)]. This result suggests that, in mature neurons, Aβ-containing CM from Tg2576 culture induces rapid CaN activation in spines, which causes acute spine morphological changes without affecting total spine density.
Molecular consequences of rapid CaN activation in spines after applying Aβ-containing TgCM
Application of Aβ-containing TgCM results in marked changes in spine morphology in cultured neurons that is blocked by inhibiting CaN activation (Fig. 7). As spine morphology and synaptic function are coordinated (Lippman and Dunaevsky, 2005; Dickstein et al., 2007; Bourne and Harris, 2008), we next examined whether Aβ-containing conditioned media-induced CaN activation in spines also regulates glutamate receptors at synapses. To address this question, we labeled the surface GluR1 subunit of the AMPA receptor, which is a major mechanism contributing to synaptic plasticity, a key process for learning, memory, and other cognitive functions of the nervous system (Aramburu et al., 2000; Malenka and Bear, 2004). AMPA receptors are linked to F-actin by scaffold proteins, and actin depolymerization triggers loss of AMPA receptors (Allison et al., 2000; Shen et al., 2000; Zhou et al., 2001; Gomez et al., 2002). In cultured neurons, CaN activation induced by Aβ-containing Tg2576 conditioned media or overexpression of a constitutively active CaN could lead to dephosphorylation of cofilin, a protein that triggers depolymerization of F-actin (Bailly and Jones, 2003) and regulates spine shape (Halpain et al., 1998). Consistent with this idea, in neurons treated with Aβ-containing conditioned media for 60 min, we noticed that the intensity ratio of phalloidin-stained F-actin in the spine to the dendritic shaft decreased markedly compared with neurons treated with wild-type conditioned media (spine/shaft ratio, wtCM, 3.20 ± 0.12; TgCM, 1.62 ± 0.06; p < 0.001; Fig. 8A,B). The significant decrease in spine/shaft fluorescence ratio is indicative of spine shrinkage and actin depolymerization (Gomez et al., 2002), which is in accord with our observations in GFP-expressing neurons (Fig. 7A). Immunodepletion of Aβ or inhibition of CaN prevented the ratio reduction (spine/shaft ratio, 3D6, 2.96 ± 0.11; FK506, 3.48 ± 0.13), suggesting that CaN activation induced by Aβ leads to F-actin reorganization, which causes a change in spine morphology (Halpain et al., 1998; Bailly and Jones, 2003).
To determine whether the rapid CaN activation alters the GluR1 subunit of AMPA receptor content in spines, we measured the mean intensity of GluR1 immunofluorescence signal in individual spines by staining fixed, nonpermeabilized cultured neurons with an antibody against an extracellular epitope of the GluR1 AMPA receptor subunit. In neurons treated with wild-type conditioned media, line scan fluorescence intensity profiles through dendritic spines and adjacent regions of the dendritic shaft revealed the greatest enrichment of surface GluR1 in a punctuated pattern mainly distributed around dendritic spines with much lower levels present in the dendritic shafts, suggesting that the majority of surface GluR1 occurs in a synaptic spine localization. Following application of Aβ-containing conditioned media for 60 min, the amount of the punctate GluR1 around dendritic spines was dramatically reduced, and increased in the dendritic shaft. Quantification of fluorescence mean intensity ratios for GluR1 puncta around dendritic spines versus the dendritic shaft confirmed enrichment of GluR1 in synaptic spines in neurons treated with wild-type conditioned media, and the ratio was decreased dramatically in neurons treated with Aβ-containing conditioned media for 60 min (spine/shaft ratio, wtCM, 1.42 ± 0.15; TgCM, 0.62 ± 0.07; p < 0.001; Fig. 8A,B). These effects were specifically dependent on Aβ exposure-induced CaN activation; removal of Aβ from conditioned media with 3D6 immunodepletion or pretreating neurons with the CaN inhibitor FK506 prevented the Tg2576 conditioned media-induced ratio decreases (spine/shaft ratio, 3D6, 1.35 ± 0.09; FK506, 1.41 ± 0.08). Thus, rapid activation of CaN in dendritic spines by Aβ-containing conditioned media strongly reduces surface GluR1 and removes GluR1-containing AMPA receptors from synapses, potentially leading to weaker synaptic connections. Overall, these results indicate that Aβ-containing conditioned media activates CaN rapidly in the spine, which causes similarly rapid changes in the synapse both structurally (F-actin reorganization) and biochemically (surface GluR1 reduction).
Soluble Aβ oligomers isolated by size exclusion chromatography from human postmortem AD brain induce rapid CaN activation in spines with consequent alteration of spine morphology and reduction of synaptic surface GluR1
To further characterize the specific role of Aβ oligomers on CaN-mediated changes in spines, we next asked whether soluble Aβ isolated directly from human AD cortex activates CaN in spines. Nondenaturing SEC was used to isolate various Aβ species from brain homogenates of human control or AD. Both ELISA and Western blot analysis showed that SEC fractions 18–19th ml of AD brain, but not the same fractions of control brain, were markedly enriched in oligomeric forms of Aβ40 and Aβ42 (Wu et al., 2010). Application of oligomeric Aβ-enriched SEC fractions 18–19th prepared from AD brain samples onto wild-type neurons for 60 min caused a rapid significant increase in the 565/522 ratio of both CaNA–CaM and CaNA–CaNB in dendritic spines with no measurable changes in the soma including cytoplasm and nucleus (for the CaNA–CaM pair: cytosol, control, 0.372 ± 0.005; nuclei, 0.468 ± 0.011; spine, 0.353 ± 0.004; AD, cytosol, 0.385 ± 0.002; nuclear, 0.467 ± 0.013; spine, 0.383 ± 0.005; p < 0.001, control spine vs AD spine; for the CaNA–CaNB pair: control cytosol, 0.374 ± 0.007; nuclei, 0.376 ± 0.004; spine, 0.364 ± 0.007; AD, cytosol, 0.369 ± 0.009; nuclei, 0.371 ± 0.005; spine, 0.451 ± 0.017; p < 0.001, control spine vs AD spine; Fig. 9A,B). In contrast, wild-type neurons exposed to the same SEC fractions prepared from control brain samples showed low FRET efficiency. Immunodepletion of Aβ from fractions 18–19th of AD samples with 3D6 prevented the increase in the 565/522 ratio of both CaNA–CaM and CaNA–CaNB (for the CaNA–CaM pair: AD plus 3D6, 0.349 ± 0.003; p < 0.001, AD spine vs AD plus 3D6 spine; for the CaNA–CaNB pair: 0.374 ± 0.117; p = 0.002, AD spine vs AD plus 3D6; Fig. 9A,B). This result provides a direct demonstration that Aβ oligomers from the human AD brain induce rapid CaN activation in spines.
With the phalloidin-stained F-actin fluorescence signal, we observed that neurons treated with Aβ-enriched SEC fractions prepared from AD brain samples for 60 min appeared to have more filopodia-like outgrowths, and the density of filopodia-like protrusions was significantly increased. The average number of filopodia-like protrusions versus total spines was 46% in neurons treated with fractions of AD brain samples compared with 12% in neurons treated with the same fraction of control brain samples (Fig. 9C). This morphological phenotype in neurons exposed to Aβ-enriched SEC fractions of AD brain samples was also confirmed by measuring the fluorescence mean intensity ratios of F-actin for spine versus dendritic shaft. The ratios were significantly reduced in neurons treated with AD SEC fractions compared with neurons treated with the same SEC fractions prepared from control brain samples (spine/shaft ratio, control, 2.39 ± 0.09; AD, 1.67 ± 0.07; p < 0.001; Fig. 9E,F). In agreement with data from Figure 7B, incubation of neurons with SEC fractions from both control and AD brain samples for only 60 min did not affect total spine density (Fig. 9D). SEC fractions from AD brain samples were incapable of inducing spine morphological changes if these SEC fractions were immunodepleted with 3D6 (percentage of filopodia-like protrusions, 12.5 ± 2.7%; spine/shaft ratio, 2.22 ± 0.07; Fig. 9C,F), demonstrating that Aβ is required for the structural modification in spines. Moreover, inhibition of CaN with its specific inhibitor FK506 markedly reduced the increase in number of filopodia-like protrusions and significantly prevented the decreased ratios of F-actin (percentage of filopodia-like protrusions, 13.7 ± 2.8%; spine/shaft ratio, 2.23 ± 0.11; Fig. 9C,F). This result suggests that Aβ isolated from AD brain induces rapid CaN activation in spines, which causes acute spine morphological changes without immediately affecting spine density.
We also compared levels of surface GluR1 in neurons treated with SEC fractions from both control and AD brain samples. Neurons treated with Aβ-enriched SEC fractions prepared from AD brain samples for 60 min showed decreases in mean surface GluR1 fluorescence in dendritic spine compared neurons exposed to the same SEC fractions prepared from control brain samples (spine/shaft ratio, control, 1.33 ± 0.13; AD, 0.88 ± 0.04; p < 0.001; Fig. 9E,F). This effect was caused by Aβ-induced local CaN activation as immunodepletion of Aβ from AD SEC samples or coapplication of FK506 with AD SEC samples blocked the change in surface levels of GluR1 in synaptic sites (spine/shaft ratio, AD plus 3D6, 1.30 ± 0.12; AD plus FK506, 1.21 ± 0.13; Fig. 9E,F). Overall, this observation provides further evidence for the involvement of local CaN activation by toxic Aβ species in AD pathogenesis.
We describe a spectral FRET-based method to quantitatively measure CaN activation dynamics at a subcellular level in cultured neurons. The spectral FRET-based assay defines two localization states of CaN activation in response to soluble Aβ oligomer stimulation: first, a robust CaN activation is initiated in spines after a brief Aβ exposure; second, a delayed but sustained CaN activation is observed in the cytoplasm and the nucleus by prolonged Aβ exposure. Acute activation of CaN in spines caused by Aβ-containing conditioned media is sufficient to induce spine structural changes and affect levels of postsynaptic proteins including F-actin and GluR1-containing AMPA receptors. Notably, the disease-associated Aβ species, which are isolated by SEC from human AD brain, induces a rapid CaN activation specifically and locally in spines with consequent alteration in spine morphology and reduction in postsynaptic proteins. Our results provide the first demonstration that the temporal and spatial precision of CaN activation in response to Aβ stimulation involves compartmentalized modulation; the disease-associated Aβ oligomers initiate rapid, local posttranslational responses downstream of CaN in spines.
FRET-based assays have been extensively used as a method to image molecular processes in the natural environment such as protein–protein interaction of several signaling pathways, primarily taking advantage of overexpressed tagged proteins (Stryer, 1978; Selvin, 2000; Jares-Erijman and Jovin, 2003). We have extended this technology to CaN and have used different approaches to detect CaN activation in vitro for live-cell imaging using GFP-RFP tags and endogenous proteins after immunostaining. CaN activation is widely studied in the immune system and in numerous disease processes, including congestive heart failure (Clipstone and Crabtree, 1992; Crabtree and Schreiber, 2009). In the CNS, activation of CaN has been implicated in the context of normal learning and memory, as well as diseases including glaucoma, spinal cord trauma, kainate injury, and ischemia (Morioka et al., 1999; Springer et al., 2000; Wu et al., 2004; Huang et al., 2005). In the context of Alzheimer's disease, CaN was shown to be abnormally activated in neurons but also in glia, thus participating to neuroinflammatory processes that are tightly associated with the progression of the disease (Abdul et al., 2009). The tools we describe here may therefore prove useful in monitoring the role of CaN in different cell types, with improved spatial and temporal resolution.
The FRET-based tools described here allow us to specifically address questions such as the kinetics of CaNA/CaM and CaNA/CaNB interactions following stimulation in neuronal cells and provides a tool for evaluating the kinetics of spine or nuclear activation of calcineurin after physiologic or pathologic stimuli. A previous attempt to monitor CaN activation using a phosphatase activity-dependent molecular switch was based on the N-terminal regulatory domain of the NFAT1 that were sandwiched between the enhanced cyan fluorescent protein (ECFP) and a circularly permuted version of the yellow fluorescent protein, Venus, as a specific substrate of CaN (Newman and Zhang, 2008). However, the overexpressed ECFP-Venus FRET pair has a limited response, is indirect in that it is restricted to report only CaN-mediated NFAT activation, is sensitive to NFAT phosphorylation events, and does not provide spatial information about CaN activation. By contrast, the conformational changes induced by the interaction of CaNA–CaM or CaNA–CaNB directly report CaN activation toward all its substrates. This assay can also be adapted for transfected or endogenous protein assays and provides outstanding spatial and temporal resolution. As demonstrated by Stemmer and Klee (1994) in a previous study, the Kd of the CaNB Ca2+ binding sites is very high (10 nm), so that CaNA and CaNB may be associated even at baseline. Recent pull-down/mass spectroscopy data suggest that the CaNA–CaNB interaction increases significantly with cellular stress (Kozubowski et al., 2011). From our data, we cannot distinguish whether CaNA and CaNB interact at baseline and then change their conformation when activated by calcium, or whether the association increases with an influx of calcium, or both; regardless, activation of calcineurin leads to a change in the conformation so that the fluorophores come close enough to one another to support FRET.
Our results indicate that CaN activity is significantly increased in spines over a period of 0–60 min after Aβ oligomer application. Although overt spine loss does not occur acutely, substrates located in the spine compartment are immediately targeted by CaN and synaptic function could be substantially affected during the early phase of CaN activation. For example, the early phase of CaN activation alters spine morphology and affects levels of surface GluR1-containing receptors in spines, confirming a strong local cellular consequence of CaN activation caused by acute Aβ oligomer stimulation (Hsieh et al., 2006). However, neurons have a late-phase CaN activation in soma that starts 6 h after Aβ oligomer application, lasts for at least 24 h, and induces NFAT nuclear translocation (Wu et al., 2010). Neuron function could be potentially affected by both CaN-mediated posttranslational and transcriptional processes. For example, spine loss occurs in 24 h and can be blocked by VIVIT, which prevents NFAT transcriptional activation (Wu et al., 2010), whereas changes in spine morphology, which is associated with loss of postsynaptic proteins including F-actin and GluR1 subunit of AMPA receptors, occurs within the first hour and is not affected by VIVIT. Moreover, the FRET ratios of both CaNA–CaM and CaNA–CaNB obtained from all three compartments in neurons exposed to Aβ oligomers for 60 min followed by washing off and then returned to original conditioned media for 24 h showed only basal level CaN activation (Fig. 6C,D). This result indicates that the effect of the 60 min exposure to Aβ oligomers on CaN activation is temperate and reversible, and is a completely local event, which is not sufficient to extend its effect to the soma and initiate the late-phase CaN activation. This also suggests that prolonged Aβ oligomer exposure is required for CaN activation in cytoplasm and nuclei. As altering CaN phosphatase activity is considered to be critical for physiological or pathological neuronal plasticity, the subcellular temporal properties of CaN activation in response to Aβ oligomer stimulation we observed here may be important to understand synaptic alterations during AD pathogenesis.
This work was supported by National Institutes of Health Grants AG08487, EB000768, EY13399, and 1R01AG041507-01. We thank Dr. Jeffery D. Molkentin (Cincinnati Children's Hospital, Cincinnati, OH) for providing CaN constructs. We thank Daniel Joyner for primary neuron preparation.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Bradley T. Hyman, MassGeneral Institute for Neurodegenerative Disease, 114 16th Street, Charlestown, MA 02129.