Abstract
Neuronal excitatory synapses are primarily located on small dendritic protrusions called spines. During synaptic plasticity underlying learning and memory, Ca2+ influx through postsynaptic NMDA-type glutamate receptors (NMDARs) initiates signaling pathways that coordinate changes in dendritic spine structure and synaptic function. During long-term potentiation (LTP), high levels of NMDAR Ca2+ influx promote increases in both synaptic strength and dendritic spine size through activation of Ca2+-dependent protein kinases. In contrast, during long-term depression (LTD), low levels of NMDAR Ca2+ influx promote decreased synaptic strength and spine shrinkage and elimination through activation of the Ca2+-dependent protein phosphatase calcineurin (CaN), which is anchored at synapses via the scaffold protein A-kinase anchoring protein (AKAP)150. In Alzheimer's disease (AD), the pathological agent amyloid-β (Aβ) may impair learning and memory through biasing NMDAR Ca2+ signaling pathways toward LTD and spine elimination. By employing AKAP150 knock-in mice of both sexes with a mutation that disrupts CaN anchoring to AKAP150, we revealed that local, postsynaptic AKAP–CaN–LTD signaling was required for Aβ-mediated impairment of NMDAR synaptic Ca2+ influx, inhibition of LTP, and dendritic spine loss. Additionally, we found that Aβ acutely engages AKAP–CaN signaling through activation of G-protein-coupled metabotropic glutamate receptor 1 (mGluR1) leading to dephosphorylation of NMDAR GluN2B subunits, which decreases Ca2+ influx to favor LTD over LTP, and cofilin, which promotes F-actin severing to destabilize dendritic spines. These findings reveal a novel interplay between NMDAR and mGluR1 signaling that converges on AKAP-anchored CaN to coordinate dephosphorylation of postsynaptic substrates linked to multiple aspects of Aβ-mediated synaptic dysfunction.
Significance Statement
Understanding mechanisms of synaptic dysfunction in Alzheimer's disease (AD) is pivotal for therapeutic advances. Amyloid-β oligomers (Aβo), the primary culprits in AD pathology, disrupt critical synaptic plasticity mechanisms, leading to enhanced LTD and synaptic loss. However, the underlying signaling pathways remain elusive. Calcineurin (CaN), localized by AKAP79/150 at synapses, plays a key role in LTD formation. Inhibition of CaN mitigates Aβo-induced synaptic deficits, implicating its involvement in AD pathology. Our study shows that AKAP-anchored CaN is critical in acute Aβo-mediated inhibition of NMDAR Ca2+ signaling and dendritic spine loss. Additionally, we identify mGluR1 as an upstream regulator of these Aβo-induced deficits, highlighting several potential therapeutic targets for AD-related synaptic pathology.
Introduction
Among the earliest manifestations of AD are the dysfunction and loss of excitatory synapses in the hippocampus (Hsieh et al., 2006; Shankar et al., 2007; Wei et al., 2010). One primary pathological agent in AD is amyloid-β (Aβ), a 42 aa peptide generated through enzymatic cleavage of the amyloid precursor protein (APP). While Aβ plaques and fibrils are hallmarks of AD, prior studies established that soluble, oligomeric Aβ (Aβo) is responsible for synapse impairments and loss (Li and Selkoe, 2020). For example, Aβo potently and rapidly inhibits NMDAR-mediated Ca2+ signals (Shankar et al., 2007; Sinnen et al., 2016) and LTP, a key form of plasticity essential for learning and memory (Freir et al., 2001; Walsh et al., 2002; Shankar et al., 2008). Over longer timescales, Aβo triggers dendritic spine loss and synapse elimination (Cleary et al., 2005; Shankar et al., 2007). While Aβo are known to mediate synapse dysfunction, the underlying signaling mechanisms are still not well understood.
Accumulating evidence suggests that Aβo synapse toxicity is mediated at least in part by signaling pathways associated with LTD. Calcineurin (CaN) is a protein phosphatase (also known as PP2B) with an essential role in LTD (Mulkey et al., 1994). Activated by prolonged, low-level Ca2+ signals, CaN promotes LTD by dephosphorylating substrates including the AMPAR subunit GluA1 at S845 resulting in its removal from synapses and internalization (Ehlers, 2000; Lee et al., 2000, 2003, 2010; Man et al., 2007; He et al., 2009). CaN also dephosphorylates and activates slingshot phosphatase (SSH1L), which in turn dephosphorylates and activates the F-actin severing protein cofilin leading to dendritic spine shrinkage and elimination associated with LTD (Wang et al., 2005). Finally, CaN regulates transcriptional pathways associated with LTD maintenance and spine elimination that are chronically engaged by Aβo (Abdul et al., 2009; Espana et al., 2010; Hudry et al., 2012; Martinez et al., 2024).
Given that Aβo blocks LTP and facilitates LTD, numerous groups have investigated whether CaN signaling is responsible for Aβo-induced synaptic dysfunction. Indeed, global inhibition of CaN with the immunosuppressant drugs cyclosporin A (CsA) or tacrolimus (FK506) prevents Aβo-mediated LTP blockade, spine loss (Q. C. Chen et al., 2002; Hsieh et al., 2006; Shankar et al., 2007; Zhao et al., 2010), and cognitive deficits (Dineley et al., 2007, 2010). Intriguingly, retrospective studies of patients taking CaN inhibitors for immunosuppression revealed reduced incidence of dementia (Taglialatela et al., 2015). However, in addition to immunosuppression, there are major side effects of CsA and FK506 that preclude chronic use to treat AD, including renal toxicity (Tedesco and Haragsim, 2012). Thus, a better understanding of how Ca2+–CaN signaling is engaged by Aβo and its downstream targets could inform future therapies.
While CaN is distributed throughout neurons, LTD signaling is specifically mediated by CaN localized to excitatory synapses by AKAP79/150 (human AKAP79/rodent AKAP150; AKAP5 gene), which is linked to NMDARs and AMPARs through association with the PSD scaffolding proteins PSD95 and SAP97 (Jurado et al., 2010; Sanderson et al., 2012). The importance of AKAP-anchored CaN was demonstrated using knock-in mice carrying a mutant version of AKAP150 lacking the PxIxIT CaNA subunit-binding motif (AKAP150ΔPIX). In these mice, total cellular levels of CaN are unperturbed, yet CaN-mediated GluA1 S845 dephosphorylation and removal from synapses is impaired and LTD is absent (Sanderson et al., 2012). Remarkably, AKAP150ΔPIX mice are also resistant to acute, Aβo-induced LTP blockade, at least in part due to altered AMPAR trafficking (Sanderson et al., 2021). Thus, while our prior work identified AKAP-anchored CaN as a critical signaling node for both normal plasticity and Aβo-mediated synapse impairment, we do not know if AKAP-anchored CaN is required for additional synaptotoxic Aβ signaling events that mediate LTP inhibition and spine loss. Accordingly, here we found that AKAP-anchored CaN is also responsible for acute Aβo-mediated inhibition of NMDAR Ca2+ signaling, activation of cofilin, and dendritic spine loss downstream of the Group I metabotropic glutamate receptor mGluR1.
Materials and Methods
Animals
All animal procedures were approved by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee in accordance with National Institutes of Health (NIH)/United States Public Health Service guidelines. Production and initial characterization of AKAP150ΔPIX mice was previously described in Sanderson et al. (2012). Mouse allele (AKAP150ΔPIX, RRID: MGI_5702308) was backcrossed to C57Bl6J several generations but then maintained, along with related WT controls, on a mixed C57Bl6J/129X1/SvJ background. Mice of both sexes were used for this study.
Primary neuronal cultures
Briefly, the hippocampus was dissected from postnatal day 0 to 2 AKAP150ΔPIX or AKAP150WT mice of both sexes and dissociated in dissection buffer (1.5 mM CaCl2, 5 mM EDTA, 1 μM NaOH, 200 units papain, 1.15 mM ʟ-cysteine) for 40 min. After five times washes with MEM, hippocampi were triturated, and the single cell suspension was counted. Neurons were plated in MEM with 10% FBS, 25 μM ʟ-glutamine, P/S at a density of 400,000 cells/ml (for QCT recordings, 18 mm glass coverslips in 12-well plates) or 580,000 cells/ml (for biochemistry, 6-well plates) coated with 0.1 mg/ml poly-d-lysine, and 10 μg/ml laminin (BD Biosciences) and maintained at 37°C, 5% CO2 for 14 d. On DIV1, MEM was replaced by Neurobasal Plus B27, GlutaMAX, 12.5 μM ʟ-glutamine, and Pen/Strep (Invitrogen). Antimitotics (1 mM FdUr, 1 mM uridine at 1:100) were added on Day 5. Fifteen percent fresh media with antimitotics was added at DIV7. Neurons were used for experiments at DIV14–16.
Aβ42 oligomer preparation
For electrophysiology experiments, 1 mg of Aβ (1-42) was dissolved in 80 μl of NH4OH (1%) + 120 μl of dimethylsulfoxide (final concentration 1.1 mM), aliquoted, flash frozen, and stored at −80°C until use. Aliquots were then thawed on the day of use and diluted to the appropriate working concentration(s) in artificial cerebrospinal fluid (ACSF). For spine loss and QCT experiments, Aβ(1-42) (AnaSpec) was prepared in aliquots as dried 1,1,1,3,3,3-hexafluoro-2-propanol film and stored at −80°C, as described previously (Klein, 2002). The peptide film was dissolved in 2 μl of anhydrous DMSO by vortexing, diluted to 100 μm with 1× PBS, vortexed, and allowed to oligomerize at 4°C overnight. The preparation was centrifuged at 14,000 × g for 10 min at 4°C to remove insoluble aggregates. The supernatant was kept on ice until use. The concentration of oligomeric Aβ species is likely lower than what is reported due to heterogeneous nature of the preparation and the removal of the insoluble fraction during centrifugation.
Extracellular fEPSP recordings
Unless indicated, all chemicals were from Sigma-Aldrich. For slice preparation, 2–3-week-olds were decapitated under deep anesthesia with isoflurane via inhalation. The hippocampi were removed from the brain, placed in ice-cold sucrose containing cutting buffer (in mM: 2 KCl, 12 MgCl2, 0.2 CaCl2, 1.25 NaH2PO4, 10 d-glucose, 220 sucrose, and 26 NaHCO3), and 400-μm-thick horizontal slices were made using a McIlwain tissue chopper. Slices were recovered at 22–27°C for >80 min in ACSF/cutting solution mixture (ACSF in mM: 126 NaCl, 5 KCl, 2 CaCl2, 1.25 NaH2PO4, 1 MgSO4, 26 NaHCO3, 10 glucose, and 2 N-acetyl cysteine). Following recovery, slices were transferred to a recording chamber and maintained at 29–30°C in ACSF with 0.0025 mM picrotoxin (Tocris Bioscience) minus N-acetyl cysteine. A bipolar tungsten stimulating electrode was placed in the SC pathway 200–300 μm from CA1 cell bodies to evoke field excitatory postsynaptic potentials (fEPSPs) recorded in the stratum radiatum using a nearby glass micropipette filled with ACSF (access resistance, 2–5 MΩ). For studies of LTP, the test stimulus intensity was set to evoke 45% of the maximum slope and delivered at 0.05 Hz. Aβ(1-42) was purchased from AnaSpec, and Aβ oligomers were prepared as previously described (Freund et al., 2016). For experiments with Aβo, slices were recovered for 60 min and then preincubated with 100 nM Aβo for 30 min prior to LTP induction. WinLTP software was used for data acquisition and analysis.
Ca2+ imaging and analysis
To image quantal Ca2+ transients (QCTs), primary mouse hippocampal neurons, transfected with GCaMP6s (plasmid #40753, Addgene; (T. W. Chen et al., 2013a)) 2 d prior to recording, were incubated at DIV14–16 in an ACSF solution containing the following (in mM): 130 NaCl, 5 KCl, 10 HEPES, 30 glucose, 2.5 CaCl2, 0.003 glycine, and 0.002 TTX, pH 7.4. Single z plane images of a portion of the dendritic arbor were acquired at 6.2 Hz for 2 min to record baseline QCTs. Then, vehicle (DMSO/PBS) or Aβ oligomer was added to the recording chamber (for final concentration of 100 or 500 nM) and cells were incubated for 45 min. The same z plane was imaged to record QCTs post-Aβ treatment.
Live-cell imaging was performed at 31°C on an Olympus IX71 equipped with a spinning disk scan head (Yokogawa). Images were acquired using a 60× Plan Apochromat 1.4 numerical aperture objective and collected on a 1,024 × 1,024 pixel Andor iXon EM-CCD camera. Data acquisition and analysis were performed with MetaMorph (Molecular Devices) and ImageJ software. To measure the frequency and amplitude of QCTs, regions of interest (ROIs) were drawn around clearly resolved spines in the baseline movie. The ROIs were saved and the same synapses were analyzed in the posttreatment movie. The mean background-subtracted GCaMP6 fluorescence within each ROI was measured. A baseline of 10 frames was established and each frame was compared with that baseline. A threshold of a 100% increase in fluorescence over baseline was established to remove small variations in fluorescence. Event amplitudes were compared between baseline and post-Aβ oligomer treatment at the same synapses. Data are plotted as the change in amplitude relative to the baseline values. Synapses displaying no events postvehicle/ βo were excluded from analysis.
Western blotting
For Western blot analysis, cell lysates were collected in ice-cold lysis buffer containing the following (in mM): 10 NaPO4, pH 7.4, 5 EDTA, 5 EGTA, 100 NaCl, 1 Na3VO4, 10 sodium pyrophosphate, and 50 NaF, with Protease/Phosphatase Inhibitor Cocktail (Cell Signaling, catalog #5872S), 1% SDS, 1% Triton. After centrifugation at 15 min at 14,000 × g at RT, the samples were incubated with 1× SDS-PAGE loading buffer at 95°C for 5 min. Sample separation was performed on 12% SDS-PAGE gels for cofilin, and 8% SDS-PAGE gels for GluN2B and PKA. Proteins were transferred to a low-fluorescence PVDF membrane, blocked in 3% BSA/TBS-T for 1 h, and incubated overnight at 4°C with primary antibodies (listed in the antibody section) except for anti-β-tubulin, anti-α-tubulin, and PKA RIIα, which were incubated for 20 min at RT while rocking. After three washes with TBS-T, the membranes were incubated with fluorescent secondary antibodies for 1 h at RT while rocking. Membranes were exposed with Azure Imager from Azure Biosystems.
Dendritic spine loss quantification
For assessment of spine density, 14–16 DIV neurons were transfected with a GFP cell fill using Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations 1 d prior to the experiment. On the day of the experiment, they were either exposed to the indicated concentration of Aβ oligomers (Fig. 4) or pretreated with vehicle (equal volume of water to corresponding drug volume; no more than 6 μl to 2 ml media), 30 μM LY367385 or 25 μM MTEP for 15 min, prior to the addition of Aβ for 24 h. After 24 h of treatment, the coverslips were rinsed with 4% sucrose/PBS, fixed in 4% sucrose/4% PFA/PBS for 10 min at RT, permeabilized with 0.2% Triton X-100 for 10 min, and blocked in 4% BSA/PBS for 1 h. The coverslips were then incubated for 2 h at RT on a rocker with Chicken-anti-GFP 1° antibody, followed by 3× PBS rinses, and 1 h 488-Goat-anti-Chicken 2° antibody. After 3× PBS rinses, one rinse in H2O to remove the salts, the coverslips were mounted on glass slides using ProLong Gold Antifade Mountant (Thermo Fisher Scientific, catalog #P36930) and allowed to cure at RT overnight. For analyses of Aβ-mediated spine loss in WT versus ΔPIX neurons, neurons were visualized using a Zeiss Axiovert 200M inverted microscope equipped with a Sutter Instruments Lambda LS Xenon Arc Lamp (300 W, 330–650 nm) or a Sutter Instruments Lambda XL Illuminator (330–700 nm); a Zeiss 63× Plan Apochromat, 1.4 NA, oil immersion objective; FITC/Alexa 488 filter sets (Chroma); and a 1,024 × 1,024 pixel Photometrics CoolSNAP HQ2 Monochrome CCD camera (Roper Scientific) controlled by SlideBook 5.5–6.0 software (3i-Intelligent Imaging Innovations). Three-dimensional z stacks were collected over the dendritic arbor with a 200 nm step size. z stacks were then deconvolved using the “nearest-neighbor” algorithm and collapsed into maximum intensity projections of the dendritic arbor or sum intensity projections of the soma. Alternatively, for analyses of Group I mGluR regulation of Aβ-mediated spine loss, imaging was performed using Olympus IX71 equipped with a spinning disk scan head (Yokogawa). Images were acquired using a 60× Plan Apochromat, 1.4 NA, oil immersion objective, and collected on a 1,024 × 1,024 pixel Andor iXon EM-CCD camera with 488 nm laser excitation controlled by MetaMorph software (Molecular Devices). Following image acquisition, regardless of the imaging system used, dendritic spines were then counted along 200–300 μm of 2 and 3° apical dendrite, with quantification from at least three separate dendrites per cell performed using Fiji/ImageJ2 software.
Antibodies
For antibodies used, refer to Table 1.
Antibodies used in this study
Experimental design and statistical analysis
LTP time-course data for both male and female juvenile mice combined were analyzed using standard two-way ANOVA in Prism (GraphPad). Group comparisons where indicated were performed in Prism using one-way ANOVA with Tukey's or Dunnett's correction for multiple comparisons with a control group or Bonferroni’s correction for multiple comparisons between pairs of groups. Significance is reported as *p < 0.05, and data are expressed and graphed as mean ± SEM. For one-way and two-way ANOVA, actual p values are reported in the Results section when provided by the software. When actual p values are not provided by the software only, *p < 0.05, **p < 0.01, **p < 0.001, and ****p < 0.0001 are reported in the Results section. In addition, only *p < 0.05, **p < 0.01, **p < 0.001, and ****p < 0.0001 are reported in the figure legends.
Results
AKAP-anchored CaN is required for Aβo-mediated impairment of synaptic NMDAR Ca2+ influx
LTP induction requires Ca2+ entry through NMDARs, which is impaired by Aβo (Sinnen et al., 2016). We tested whether AKAP-anchored CaN signaling is required for Aβo-mediated impairment of NMDAR function using a live-cell microscopy assay we previously developed to directly visualize NMDAR-dependent Ca2+ influx in response to spontaneous, single neurotransmitter vesicle release events (Sinnen et al., 2016). This approach allows us to make paired comparisons of NMDAR-mediated quantal Ca2+ transients (QCTs) at the same individual spines before and at various times after Aβo application (Sinnen et al., 2016). We previously demonstrated that Aβo rapidly (within minutes) reduces QCT amplitude through a mechanism independent of NMDAR internalization, suggesting Aβo could engage signaling mechanisms that directly impact the function of synaptic NMDARs (Sinnen et al., 2016). Since we originally developed this assay in rat dissociated hippocampal neurons, we first confirmed Aβo has a similar effect in mouse neurons. QCTs measured from wild-type (WT) mouse primary hippocampal neurons were similar in magnitude and sensitive to the NMDAR antagonist APV, as previously described in rat preparations (Sinnen et al., 2016; Fig. 1A,B). QCTs in neurons from AKAP150ΔPIX mice were also similarly blocked by APV (Fig. 1C). Furthermore, basal amplitudes of QCTs were similar in WT and AKAP150ΔPIX neurons (Fig. 1D; amplitude: WT, 7 ± 0.18 a.u., n = 577, N = 17 independent cultures; AKAPΔPIX, 7.4 ± 0.19 a.u., n = 607, N = 17 independent cultures, p = 0.0996 vs WT; unpaired t test). We next measured QCT amplitudes before and 45 min following Aβo application in WT mouse neurons. In control (vehicle-treated) neurons, we observed a small but consistent increase in QCT amplitude, compared with baseline, over the time course of the experiment. In contrast, either 100 or 500 nM Aβo treatment led to significantly reduced QCT amplitudes over the same time frame (Fig. 1E,F; control ΔF / F0, 0.123 ± 0.06, n = 104; 100 nM Aβo, −0.935 ± 0.07, n = 104, p = 0.0026 vs control; 500 nM, −0.222 ± 0.007, n = 107, p = 0.0007 vs control; N = 3 independent cultures; one-way ANOVA with Dunnett's). As in our previous study in rat neurons, the GluN2B-selective negative-allosteric modulator ifenprodil (5 μM) largely occluded Aβo-mediated NMDAR impairment, thus indicating Aβo exerts its effects through GluN2B-containing NMDARs (Fig. 1G).
AKAP150-anchored CaN is required for Aβo-mediated reduction in NMDAR Ca2+ signals. A, Time series showing a QCT visualized in a dendritic spine (red box) from a mouse hippocampal neuron transfected with GCaMP6s. The traces below show GCaMP6s signal from a single dendritic spine before (black) and after (red) APV addition. B, Representative effect of APV on QCT frequency in WT neurons (n = 93 spines from 1 neuron). C, Representative effect of APV on QCT frequency in AKAPΔPIX neurons (n = 36 from 3 neurons). D, Comparison of baseline QCT amplitudes from WT and AKAPΔPIX neurons (n = 577, from N = 17 independent cultures). E, Representative QCT traces measured from neurons cultured from WT animals from the same spine before and after a 45 min incubation with vehicle (left), 100 nM (middle), or 500 nM (right) Aβo. F, Compiled QCT data from WT neurons plotting the average change in amplitude 45 min post-Aβo treatment compared with baseline measurements prior to Aβo treatment. Vehicle, n = 104 spines, 8 neurons, N = 3; 100 nM, n = 104, 8 neurons, N = 3; 500 nM Aβo, n = 107, 9 neurons N = 3; frequency (n = 108, 8–9 neurons, N = 3 for all conditions). **p < 0.01, ***p < 0.001 by one-way ANOVA with Dunnett's. G, GluN2B is the primary target of Aβo-mediated Ca2+ amplitude decrease. Data represented as within-single-spine change in QCT amplitude post 45 min Aβo treatment (left panel) or a 5 min, 5 μM ifenprodil treatment followed by addition of 500 nM Aβo for 45 min (right panel), showing no further QCT decrease (n = 26 spines from 1 neuron). ****p < 0.001 by one-way ANOVA with Tukey's. H, Same as E except measurements were made from AKAPΔPIX neurons. I, Same as F, except measurements were made from AKAPΔPIX neurons. Amplitude: vehicle, n = 104, 9 neurons, N = 3; 100 nM Aβo, n = 108, 9 neurons, N = 3; 500 nM Aβo, n = 108, 8 neurons, N = 3).
Intriguingly, when we measured NMDAR Ca2+ entry in primary cultured hippocampal neurons from AKAP150ΔPIX mice, we observed no reduction in QCT amplitude mediated by application of either 100 or 500 nM Aβo (Fig. 1H,I; control ΔF / F0, 0.402 ± 0.09, n = 104; 100 nM Aβo, 0.494 ± 0.09, n = 108, p = 0.633 vs control; 500 nM, 0.342 ± 0.062, n = 108, p = 0.81 vs control, N = 3 independent cultures; one-way ANOVA with Dunnett's). Thus, AKAP-anchored CaN is required for Aβo-mediated impairment of synaptic NMDAR Ca2+ entry.
In contrast to our previous experiments in rat neurons (Sinnen et al., 2016), we did not observe reduced QCT frequency, which reports the rate of spontaneous presynaptic vesicle fusion, following Aβo treatment in WT (control Δfreq / freq0, 0.295 ± 0.127; 100 nM Aβo, 0.388 ± 0.179, p = 0.88 vs control; 500 nM, −0.076 ± 0.1, p = 0.0525 vs control; n = 108 for all conditions, N = 3 independent cultures) or AKAP150ΔPIX (control Δfreq / freq0, 0.093 ± 0.083, n = 108; 100 nM Aβo, 0.192 ± 0.105, n = 106, p = 0.681 vs control; 500 nM, 0.412 ± 0.097, n = 108, p = 0.034 vs control; N = 3 independent cultures) neurons.
Aβo regulates phosphorylation of GluN2B at S1166 to impair NMDAR function
We next investigated potential mechanisms for Aβo-mediated impairment of synaptic NMDAR function. Since many synaptic deficits caused by Aβo appear to be mediated through GluN2B-containing NMDARs (Li et al., 2011; Sinnen et al., 2016; Rammes et al., 2017; Taniguchi et al., 2022), we focused on posttranslational modifications of this subunit that affect receptor localization and/or function. Previous studies in rat cortical neurons showed that application of 1 μM Aβ monomer, or medium containing secreted Aβo, led to internalization of NMDARs as evidenced by reduced surface expression of GluN1 and GluN2B subunits and that this effect was due to the dephosphorylation of the GluN2B subunit of NMDAR at Y1472 (Snyder et al., 2005). While our previous experiments suggest this mechanism does not play a role in acute Aβo-mediated impairment of synaptic NMDARs (Sinnen et al., 2016), we examined Y1472 phosphorylation in WT and AKAP150ΔPIX mouse hippocampal cultures treated with 0, 100, 500, and 1,000 nM Aβo under the same conditions we used for QCT measurements. We did not observe significant differences in Y1472 phosphorylation levels via immunoblotting of whole cell lysates for any of the conditions tested in WT (Fig. 2A; % pY1472/total GluN2B, normalized to baseline: 100 nM Aβo: 109 ± 6.98, p = 0.49 vs baseline; 500 nM Aβo, p = 0.5 vs baseline; 1,000 nM Aβo, p = 0.67 vs baseline; N = 6; one-way ANOVA with Dunnett's) or in AKAP150ΔPIX neurons (Fig. 2B; % pY1472/total GluN2B: 100 nM Aβo: 113.6 ± 6.6, p = 0.29 vs baseline; N = 5; 500 nM Aβo, 120.4 ± 18.21, p = 0.60 vs baseline, N = 4, 1,000 nM Aβo, 122.1 ± 14.04, N = 5, p = 0.46 vs baseline; one-way ANOVA with Dunnett's). Another mechanism regulating NMDAR localization is through phosphorylation at GluN2B S1480, which interferes with synaptic anchoring by disrupting PSD95/NMDAR interactions (Chung et al., 2004; Chiu et al., 2019). However, we did not observe significant differences in S1480 phosphorylation levels at any of the Aβo concentrations tested in either WT (Fig. 2C; % pS1480/total GluN2B, normalized to baseline: 100 nM Aβo, 108 ± 10.4, p = 0.79 vs baseline; 500 nM Aβo, 103.4 ± 10.07, p = 0.97 vs baseline; 1,000 nM Aβo, 87.94 ± 8.58, p = 0.47 vs baseline; N = 4; one-way ANOVA with Dunnett's) or AKAP150ΔPIX hippocampal neurons (Fig. 2D; % pS1480/total GluN2B: 100 nM Aβo, 95.16 ± 10.91, p = 0.95 vs baseline; 500 nM Aβo, 92.76 ± 7.42, p = 0.67 vs baseline; 1,000 nM Aβo, 92.33 ± 6.86, p = 0.59 vs baseline; N = 5; one-way ANOVA with Dunnett's).
Aβo application does not alter GluN2B Y1472 or S1480 phosphorylation. A, Levels of GluN2B Y1472 phosphorylation after 45 min Aβo application in WT (vehicle, 100 and 1,000 nM Aβo, N = 7; 500 nM Aβo, N = 6) and (B) AKAPΔPIX (vehicle and 500 nM Aβo, N = 4; 100 and 1,000 nM Aβo, N = 5) hippocampal mouse cultures. Top, Representative Western blot. Bottom, Quantification. C, Levels of GluN2B S1480 phosphorylation after 45 min Aβo application in WT (N = 4 for all conditions) and (D) AKAPΔPIX (N = 5 for all conditions) hippocampal mouse cultures.
Given that both our findings above and our previous work implicate NMDAR channel function rather than localization as being perturbed by Aβo, we investigated whether phosphorylation at sites that directly impact NMDAR function could be regulated by Aβo. Recent studies show that NMDAR Ca2+ entry is highly sensitive to protein kinase A (PKA)-mediated phosphorylation at S1166 on the GluN2B subunit, which increases Ca2+ permeation through the activated receptor (Murphy et al., 2014). Indeed, we observed a dose-dependent reduction in GluN2B S1166 phosphorylation 45 min after Aβo application in WT neurons (Fig. 3A; % pS1166/total GluN2B, normalized to baseline: 100 nM Aβo, 77.62 ± 6.63, N = 7, p = 0.037 vs baseline; 500 nM Aβo, 64.69 ± 7.5, N = 7, p = 0.009 vs baseline; 1,000 nM Aβo, 52.73 ± 2.3, N = 3, p = 0.035 vs baseline; one-way ANOVA with Dunnett's), but not in AKAP150ΔPIX neurons, where the highest Aβo dose instead induced a significant increase in S1166 phosphorylation (Fig. 3B; % pS1166/total GluN2B: 100 nM Aβo, 93.85 ± 3.84, N = 8, p = 0.567 vs baseline; 500 nM Aβo, 119.9 ± 13.59, N = 7, p = 0.325 vs baseline; 136 ± 12.15, N = 3, p = 0.0001 vs baseline; one-way ANOVA with Dunnett's). Thus, Aβo signaling through AKAP-anchored CaN appears to impact NMDAR function through either direct or indirect regulation of GluN2B S1166 phosphorylation.
Aβo application results in a dose-dependent decrease in GluN2B S1166 phosphorylation mediated by AKAP-anchored CaN. A, Levels of GluN2B S1166 phosphorylation after 45 min Aβo application in WT (N = 7 for vehicle, 100 and 500 nM Aβo; N = 3 for 1,000 nM Aβo) and (B) AKAPΔPIX (vehicle, N = 5; 100 nM Aβo, N = 8; 500 nM Aβo, N = 7; 1,000 nM Aβo, N = 3). Left, Representative Western blot; right, quantification. *p < 0.05, **p < 0.01, ***p < 0.001 by one-way ANOVA with Dunnett's. C, Hippocampal neuronal lysates were prepared after incubation (45 min) in QCT imaging conditions (in mM: 130 NaCl, 5 KCl, 10 HEPES, 30 glucose, 2.5 CaCl2, 0.003 glycine, and 0.002 TTX, pH 7.4) to assess protein phospho-states corresponding to baseline Ca2+ amplitudes in WT and AKAPΔPIX mouse cultures. Quantification of pS99 PKA RIIα levels normalized to total PKA RIIα (WT, N = 6; PIX, N = 6; ****p < 0.001 by t test). Note: the pS99 blot always contained a prominent smaller band below the phospho band that did not colocalize with the total PKA RIIα protein signal and was excluded from quantification of the phospho band). D, Quantification of pS1166 GluN2B levels normalized to total GluN2B (WT, N = 6; PIX, N = 6). Loading control is α-tubulin. D, Comparison of baseline QCT amplitudes between WT and PIX mouse cultures (WT, n = 577, N = 17 independent cultures; PIX, n = 607, N = 17 independent cultures, p > 0.05.).
Since CaN is known to regulate PKA activity through the dephosphorylation of the regulatory subunit PKA RIIα, enhancing its capture and autoinhibition of the catalytic subunit (Church et al., 2021), and both enzymes are anchored on AKAP150 (Sanderson and Dell'Acqua, 2011), we tested whether loss of CaN anchoring impacted this form of PKA regulation. We found that phospho-PKA RIIα levels were significantly elevated under basal conditions in AKAPΔPIX cultures compared with WT (Fig. 3C; ratio PKA RIIα pS99/total, WT, 0.1074 ± 0.0017, N = 6; AKAPΔPIX, 0.2199 ± 0.0087, N = 6, p < 0.0001 vs WT; unpaired t test). Given that CaN regulation of PKA is impaired in ΔPIX mice, we tested whether GluN2B S1166 phosphorylation was also elevated in ΔPIX neurons under basal conditions. However, we found that S1166 was basally phosphorylated at similar levels in both WT and AKAPΔPIX neurons (Fig. 3D; pS1166/GluN2B ratio: WT, 0.684 ± 0.026, N = 6; AKAPΔPIX 0.623 ± 0.024, N = 6, p = 0.6863 vs WT; unpaired t test), indicating that perturbations induced by Aβo application were necessary to elicit the increased phosphorylation of S1166 observed in AKAPΔPIX neurons (Fig. 3B).
mGluR1 activity is required for acute Aβo-mediated impairment of synaptic NMDAR Ca2+ signals
Our previous work demonstrated that Aβ oligomers rapidly bind to the surface of neurons immediately adjacent to the PSD (Actor-Engel et al., 2021) where AKAP150 is also prominently localized (X. Chen et al., 2022). Thus, we focused our investigation on putative surface receptors with similar perisynaptic localization. Multiple studies have implicated Group I mGluRs as targets of Aβo mediating synaptic dysfunction (Renner et al., 2010; Rammes et al., 2011; Um et al., 2013; X. Chen et al., 2013b; Haas et al., 2014; Taniguchi et al., 2022; Valdivia et al., 2023), and their perisynaptic localization matches that of acutely applied Aβo (Baude et al., 1993; Lujan et al., 1996; Bodzeta et al., 2021; Scheefhals et al., 2023). In addition, Group I mGluRs are coupled to Gq-PLC signaling that mobilizes Ca2+ from intracellular stores, which can contribute to CaN activation (Hogan et al., 2003; Thiel et al., 2021). Thus, we investigated which, if any, Group I mGluRs could be involved in the rapid Aβo-mediated synaptic deficits we observed. We started by testing whether pharmacological inhibition of mGluRs could prevent Aβo-mediated impairments in NMDAR Ca2+ entry. Selectively antagonizing mGluR5 with a 15 min pretreatment with 25 μM MTEP before a 45 min addition of 500 nM Aβo had no effect on QCT amplitude and did not prevent Aβo-mediated QCT amplitude decrease (Fig. 4A; control ΔF / F0, 0.065 ± 0.45, n = 108; Aβo, −0.284 ± 0.04; n = 104, p < 0.0001 vs control; MTEP, −0.087 ± 0.06, n = 104, p = 0.1143 vs control; MTEP + Aβo, −0.264 ± 0.039, n = 107, p < 0.0001 vs control; one-way ANOVA with Tukey's). We then antagonized mGluR1 with a 15 min pretreatment with 30 μM LY367385 before a 45 min application of 500 nM Aβo. Blocking mGluR1 had no effect on its own but in contrast to blocking mGluR5, prevented Aβo-mediated NMDAR impairment (Fig. 4B; control ΔF / F0, 0.153 ± 0.07, n = 95; Aβo, −0.1 ± 0.06; n = 97, p = 0.032; LY367385, 0.09 ± 0.05, n = 105, p = 0.84; LY367385 + Aβo, 0.174 ± 0.08, n = 98, p = 0.033 vs control; one-way ANOVA with Tukey's).
mGluR1 activation is required for Aβo-mediated deficits in NMDAR Ca2+ influx and alterations in GluN2B pS1166. A, Comparison of changes in QCT amplitudes in WT DIV14 neurons treated for 45 min with vehicle (black bar), 500 nM Aβo (red bar), 25 μM MTEP alone (blue bar) or 500 nM Aβo plus MTEP (purple bar). Note MTEP was added 15 min prior to Aβo addition. (Vehicle, n = 108 spines; Aβo, n = 104 spines; MTEP, n = 104 spines, MTEP + Aβo, n = 107 spines; N = 3 independent cultures). *p < 0.05, ****p < 0.0001 by one-way ANOVA with Tukey's. B, Comparison of changes in QCT amplitude of WT DIV14 neurons treated for 45 min with vehicle (black bar), 500 nM Aβo (red bar), 30 μM LY367385 alone (blue bar), or 500 nM Aβo plus LY367385 (purple bar). Note LY367385 was added 15 min prior to Aβo addition. (Vehicle, n = 95; Aβo, n = 97; LY367385, n = 105; LY367385 + Aβo, n = 98; N = 3 independent cultures). *p < 0.05 by one-way ANOVA with Tukey's. C, Levels of pS1166 GluN2B, total GluN2B, with α-tubulin as a loading control. Lysates collected after treatment in the same corresponding conditions as in A and B. (Vehicle, N = 8; 500 nM Aβo, N = 8; LY367385, N = 7; LY367385 + Aβo, N = 8; MTEP, N = 5; MTEP + Aβo, N = 5.) *p < 0.05, ***p < 0.001 by one-way ANOVA with Dunnett's.
Having determined the involvement of mGluR1 activation in Aβo-mediated NMDAR impairment, we sought to determine whether the Aβo-mediated decrease in GluN2B S1166 phosphorylation was prevented by mGluR1 antagonism. Biochemical analysis of whole cell lysates from hippocampal neuronal cultures treated with Aβo under the same conditions as QCT measurements revealed that antagonism of mGluR1 with 30 μM LY367385 prevented Aβo-mediated pS1166 decrease (Fig. 4C; % pS1166/total GluN2B, normalized to baseline: 500 nM Aβo, 80.85 ± 3.08, N = 8, p = 0.021 vs baseline; LY367385, 91.39 ± 3.5, N = 8, p = 0.43 vs baseline; LY367385 + Aβo, 94.63 ± 3.7, N = 7, p = 0.83 vs baseline; one-way ANOVA with Dunnett's). In contrast, and matching our functional data, antagonism of mGluR5 by MTEP did not prevent the decrease in pS1166 levels (Fig. 4C; % pS1166/total GluN2B MTEP, 93.6 ± 6.17, N = 5, p = 0.80 vs baseline; MTEP + Aβo, 81.16 ± 5.08, N = 5, p = 0.011 vs baseline; one-way ANOVA with Dunnett's). These data support Aβo engagement of mGluR1 signaling to limit NMDAR Ca2+ influx through downstream dephosphorylation of GluN2B at S1166 by AKAP-anchored CaN.
Genetic disruption of AKAP–CaN anchoring protects against acute LTP inhibition by Aβo
While our previous studies implicated AKAP–CaN regulation of AMPA receptor trafficking in Aβ-mediated inhibition of LTP, our current results indicate AKAP–CaN signaling-dependent suppression of NMDAR Ca2+ influx may also contribute to this effect. However, our previous work implicating AKAP–CaN signaling in Aβ inhibition of hippocampal CA1 LTP was performed ex vivo in brain slices from adult 8–12-week-old mice (Sanderson et al., 2021). Importantly, CA1 synaptic plasticity mechanisms themselves are plastic and change over the course of postnatal development (Purkey and Dell'Acqua, 2020). Thus, we wanted to determine if AKAP-anchored CaN is also required for Aβo-mediated inhibition of LTP in ex vivo brain slices from juvenile 2–3-week-old mice that are developmentally more similar to the 14 DIV dissociated hippocampal neurons cultures we used above to characterize Aβ–AKAP–CaN regulation of NMDAR Ca2+ influx. Accordingly, we pretreated ex vivo hippocampal slices from 2–3-week-old wild-type and AKAP150ΔPIX mice with vehicle or 100 nM Aβo for 30 min prior to inducing LTP at CA1 synapses by 2 × 100 Hz, 1 s stimulus trains spaced 20 s apart. We found that in wild-type slices, Aβo treatment reduced fEPSP slope compared with vehicle after 60 min post-LTP induction (Fig. 5A; % of baseline fEPSP slope: vehicle, 200.6 ± 8.70, N = 6; 100 nM Aβo, 122.3 ± 6.56%, N = 6; p = 0.0031 vs vehicle; two-way ANOVA; Bonferroni's multiple-comparisons test). However, in slices from AKAP150ΔPIX mice, 100 nM Aβo pretreatment did not result in significant difference in fEPSP slope from vehicle treatment (Fig. 5B; % of baseline fEPSP slope: vehicle, 159.2 ± 3.11%, N = 6; 100 nM Aβo, 152.5 ± %, N = 6; p > 0.999 vs vehicle, two-way ANOVA; Bonferroni's multiple-comparisons test). Thus, Aβo also engage synaptically anchored CaN to drive functional CA1 hippocampal LTP deficits also in ex vivo slice preparations from juvenile mice.
Disruption of AKAP–CaN anchoring protects hippocampal CA1 LTP from inhibition by Aβo. A, Acute WT hippocampal slices (2–3-week-old) were either treated with 100 nM Aβo (red) or vehicle (black) for 30 min prior to LTP induction at t = 0 (vehicle, N = 7; Aβo, N = 5). Representative EPSP traces recorded from vehicle-treated and Aβo-treated WT slices pre- and 55 min post-LTP induction are shown below. B, The effect of 30 min pre-incubation with 100 nM Aβo on LTP in AKAPΔPIX mice (N = 6 for both conditions). Representative EPSP traces recorded from vehicle-treated and Aβo-treated AKAPΔPIX slices are shown below. **p < 0.01 by two-way ANOVA.
AKAP–CaN anchoring is required for acute Aβo-mediated spine loss
Aβo are well known to lead to synaptic spine loss, as demonstrated by multiple research groups (Hsieh et al., 2006; Shankar et al., 2007; Wu et al., 2010; Zhao et al., 2010). In addition, we recently implicated AKAP–CaN activation of NFAT transcription factors in Aβ-mediated spine loss that occurs in response to chronic overexpression of APP containing familial AD mutations for several days (Martinez et al., 2024). However, we also wanted to determine whether AKAP–CaN signaling is also required for more acute, initial signaling events that can promote dendritic spine loss downstream of reduced NMDAR Ca2+ influx. Importantly, partially blocking NMDAR Ca2+ influx is sufficient to promote not only LTD but also dendritic spine loss (Mulkey and Malenka, 1992; Shankar et al., 2007). Thus, we first examined whether AKAP–CaN is also critical for spine loss that is observed after only 24 h of exposure to Aβo in hippocampal neuron cultures. Importantly, 24 h application of 500 and 1,000 nM, but not 100 nM, Aβo led to significantly reduced spine density in cultured neurons from WT (Fig. 6A; spines/10 μm dendrite: WT control, 3.57 ± 0.15, n = 31, N = 3; 100 nM Aβo, 3.43 ± 0.21, n = 32, N = 3, p = 0.95 vs control; 500 nM Aβo, 2.39 ± 0.17, n = 34, N = 3, p < 0.0001 vs control; 1 μM Aβo, 2.55 ± 0.19, n = 27, N = 3, p = 0.0006; one-way ANOVA with Dunnett's) but not AKAP150ΔPIX mice (Fig. 6B; spines/10 μm dendrite: AKAP150ΔPIX control, 3.40 ± 0.15, n = 36, N = 4, 100 nM Aβo, 3.4 ± 0.15, n = 35, N = 4, p > 0.99 vs vehicle; 500 nM Aβo, 3.14 ± 0.13, n = 38, N = 3, p = 0.43 vs control; 1 μM Aβo, 2.86 ± 0.23, n = 16, N = 3, p = 0.11 vs vehicle, one-way ANOVA with Dunnett's). This result demonstrates that synaptically localized CaN is also critical for Aβo to acutely induce dendritic spine loss.
AKAP–CaN is required for Aβ-mediated dendritic spine loss and activation of the F-actin severing protein cofilin. A, Spine density following 24 h of exposure to vehicle or varying Aβo concentrations was quantified in 14–16 DIV hippocampal neurons cultured from WT mice (left, representative images; right, quantification; control, n = 31 cells, N = 3 independent cultures; 100 nM Aβo, n = 32, N = 3; 500 nM Aβo, n = 34, N = 3; 1 μM Aβo, n = 27, N = 4). B, Spine density following 24 h of exposure to vehicle or varying Aβo concentrations was quantified from AKAPΔPIX mouse hippocampal neurons (left, representative images; right, quantification; control, n = 36, N = 4; 100 nM Aβo, n = 35, N = 4; 500 nM Aβo n = 38, N = 4; 1 μM Aβo, n = 16, N = 4). ***p < 0.001 or ****p < 0.0001 by one-way ANOVA with Dunnett's. C, Levels of phosphorylated (inactive) cofilin after Aβo treatment in WT and (D) AKAPΔPIX mouse neurons (left, representative images; right, quantification; N = 5 for all conditions). *p < 0.05 or **p < 0.01 by one-way ANOVA with Dunnett's. Scale bar, 10 μm.
Next, we investigated what signaling pathway CaN could be involved in driving the observed acute spine loss. CaN is known to dephosphorylate and activate Slingshot 1L (SSH1L), a phosphatase that subsequently dephosphorylates cofilin, an actin-binding protein that severs actin filaments (Wang et al., 2005). We examined the involvement of cofilin by treating primary hippocampal cultures from WT or AKAP150ΔPIX mice with the same concentrations of Aβo as above. After 24 h of treatment, the cells were collected and lysed, and the ratio of inactive phospho-cofilin to total cofilin was analyzed by Western blot. In a dose-dependent manner, Aβo treatment decreased phospho-cofilin/cofilin ratio in WT neurons (Fig. 6C; % pS3/total cofilin, normalized to baseline: 100 nM Aβo, 79.65 ± 6.37, p = 0.065 vs baseline; 500 nM Aβo, 70.76 ± 7.15, p = 0.028 vs baseline; 1,000 nM Aβo 49.47 ± 7.63, N = 5, p = 0.0047 vs baseline by one-way ANOVA with Dunnett's) but not in AKAP150ΔPIX neurons (Fig. 6D; pS3/total cofilin: 100 nM Aβo, 94.60 ± 4.97, p = 0.72 vs baseline; 500 nM Aβo, 101.3 ± 7.67, p = 0.99 vs baseline; 1,000 nM Aβo 91.87 ± 10.42, N = 5, p = 0.84 vs baseline by one-way ANOVA with Dunnett's). This biochemical data parallels our immunocytochemical data and supports the hypothesis that Aβo could initiate dendritic spine loss through activation of synaptically localized CaN, which leads to downstream cofilin activation in dendritic spines.
Having previously examined the importance of mGluR1 signaling in Aβo-mediated synaptic NMDAR deficits, we next sought to test whether the loss of dendritic spines after acute Aβo administration similarly involved mGluR1 activity. Previous studies have implicated Group I mGluR signaling in spine shrinkage (Oh et al., 2013) and elimination (Ramiro-Cortes and Israely, 2013; Wilkerson et al., 2014; Speranza et al., 2022). Antagonism of mGluR1 with a 15 min pretreatment with 30 μM LY367385, followed by a 24 h application of 1 μM Aβo, prevented spine loss seen in WT neurons (Fig. 7A; spines/10 μm dendrite: WT, 6.8 ± 0.28, n = 27; 1 μM Aβo, 5 ± 0.23, n = 27, p < 0.0001; LY367389, 6.8 ± 0.3; n = 26, p = 0.9994; LY367389 + Aβo, 7.2 ± 0.27, n = 27, N = 3 independent cultures; p = 0.7685 vs vehicle by one-way ANOVA with Tukey's). In contrast, antagonism of mGluR5 by a 15 min pretreatment with 25 μM MTEP, followed by a 24 h application of 1 μM Aβo only, partially prevented spine loss (Fig. 7A; spines/10 μm dendrite: MTEP, 6.8 ± 0.37, n = 27, p = 0.9906; MTEP + Aβo, 5.8 ± 0.32, n = 27, p = 0.1072 vs vehicle by one-way ANOVA with Tukey's, p = 0.0245 by t test vs vehicle, and p = 0.0458 by t test vs Aβo; N = 3 independent cultures). Thus, mGluR1 and to some degree mGluR5 both contribute to Aβo-mediated dendritic spine loss.
mGluR1 activation is critical for Aβo-mediated dendritic spine loss. A, Top, Representative dendrites of WT neurons pretreated with vehicle, 30 μM LY367385, or 25 μM MTEP for 15 min, followed by a 24 h vehicle or Aβo application in the continued presence of inhibitors. Bottom, Quantification of spines/10 μm dendrite (WT, n = 27 cells; Aβo, n = 27; LY367385, n = 26; LY367385 + Aβo, n = 27; MTEP, n = 27, MTEP + Aβo, n = 27; N = 3 independent cultures). LY, LY367385. B, Levels of phospho-cofilin, cofilin, and α-tubulin after treatments described in A. N = 6 independent cultures. *p < 0.05, **p < 0.01, ***p < 0.001; ****p < 0.0001 by one-way ANOVA with Tukey's.
mGluR-dependent LTD can drive the activation of cofilin (Zhou et al., 2011), so we next examined Aβo-mediated activation state of cofilin in conditions where either mGluR1 or mGluR5 was blocked. As above, 15 min pretreatment with 30 μM LY367389, followed by 24 h application of 1 μM Aβo, prevented Aβo-mediated decrease in cofilin phosphorylation (Fig. 7B; % pS3/total cofilin, normalized to baseline: 1,000 nM Aβo, 80.95 ± 1.94, p = 0.0013; LY367385, 101.4 ± 1.83, p = 0.9636; LY367385 + Aβo, 105.8 ± 5.43, p = 0.8768, one-way ANOVA with Tukey's, N = 6 independent experiments). Paralleling the partial protection against spine loss above, antagonism of mGluR5 with 25 μM MTEP prior to a 24 h Aβo application prevented pS3 cofilin dephosphorylation, although with much greater variability compared with LY367385 (Fig. 7B; % pS3/total cofilin, normalized to baseline: MTEP, 94.38 ± 8.94, p > 0.9999; MTEP + Aβo, 106.1 ± 8.221, p > 0.9999, one-way ANOVA with Tukey's; N = 6 independent experiments). Thus, our findings correlate Aβo-mediated activation of cofilin and spine loss via AKAP-anchored CaN, with both processes clearly requiring mGluR1, and to some degree mGluR5, signaling.
Discussion
The present study highlights the critical importance of scaffolding and localization of kinases and phosphatases in proximity to their targets for regulating highly localized signaling events in both health and disease states. We employed a specific genetic disruption that prevents CaN binding to AKAP150 to probe whether localized CaN is necessary for key signaling events underlying Aβo-induced synaptic dysfunction. We found that Aβo-mediated synaptic deficits, including both reduced NMDAR Ca2+ influx that promotes LTD over LTP and increased cofilin activation that promotes dendritic spine loss, both critically rely on a specific pool of CaN, anchored by AKAP150 (Fig. 8).
Model for mGluR1 and AKAP150-anchored CaN signaling in Aβ regulation of NMDAR Ca2+ influx and F-actin dynamics in dendritic spines. AKAP–CaN regulation of PKA RIIα, GluN2B, and cofilin phosphorylation under basal conditions of homeostasis (left) and in response to acute application of Aβo (right). Created with BioRender.com.
NMDARs are a vital source of Ca2+ in dendritic spines for numerous mechanisms of synaptic homeostasis and plasticity and thus subject to complex regulation. Here we directly visualized single-spine NMDAR-mediated Ca2+ events before and after Aβo exposure at the same spines, making it possible to probe the mechanism underlying Aβo-mediated NMDAR dysfunction. We demonstrate that Aβo primarily targets NMDAR Ca2+ influx through GluN2B-containing NMDARs. Along with the obligatory GluN1 subunit, NMDARs can exist as diheteromeric 2B or 2A-only receptors or as triheteromeric 2A and 2B containing receptors (Thomas et al., 2006; Sinnen et al., 2016). The presence of GluN2A-only NMDARs could explain the partial loss of NMDAR function as a result of Aβo application. Indeed, it will be interesting to investigate whether Aβo has any effect on NMDAR function in the absence of the GluN2B subunit.
While there is some evidence of direct Aβo–NMDAR interaction that could be responsible for NMDAR impairment (Taniguchi et al., 2022), we find this to be an unlikely explanation since Aβo only impacted NMDAR function in WT but not in AKAPΔPIX neurons. Moreover, previous studies show that Aβo binding to the neuronal surface occurs within tens of seconds, yet NMDAR impairment takes 10–30 min to manifest (Sinnen et al., 2016; Actor-Engel et al., 2021). Additionally, using superresolution imaging to localize Aβo binding, we previously ruled out direct binding to synaptic receptors by demonstrating that Aβo primarily binds at perisynaptic sites immediately adjacent to, but not overlapping with, the PSD (Actor-Engel et al., 2021). Together, these observations are consistent with a model where Aβo engages a perisynaptic surface receptor(s) to generate a Ca2+ signal that activates AKAP-anchored CaN to promote LTD-like signaling, which is then reinforced by CaN negative feedback that reduces NMDAR Ca2+ to further favor LTD over LTP signaling.
While the upstream receptor that binds Aβo has remained contentious, previous studies have implicated mGluR1/5 receptors, which are localized to perisynaptic regions, in Aβo-mediated synaptic dysfunction (Renner et al., 2010; Rammes et al., 2011; Um et al., 2013; X. Chen et al., 2013b; Haas et al., 2014; Taniguchi et al., 2022). Group I mGluRs are known to stimulate PLC-mediated production of IP3, which would release Ca2+ from internal stores, activate calmodulin, and subsequently activate CaN (Hernandez-Lopez et al., 2000; Fujii et al., 2020). Paradoxically, direct activation of Group I mGluRs with the pan-Group I agonist DHPG can either negatively regulate NMDAR function by promoting endocytosis (Snyder et al., 2001; Jin et al., 2015), which we have already shown is not the reason for decreased Ca2+ influx following Aβo exposure or in other cases positively regulate NMDAR currents (Fitzjohn et al., 1996; Mannaioni et al., 2001; Skeberdis et al., 2001; Benquet et al., 2002). However, the fractional NMDAR Ca2+ current (which accounts for only ∼10–20% of total current) can be regulated independent of the total current through phosphorylation of GluN2B S1166. Thus, alterations in NMDAR Ca2+ influx may be missed using purely electrophysiological approaches that measure total NMDAR current. Indeed, previous work from our group has shown that Aβo does not acutely impair synaptic NMDAR responses measured using electrophysiology despite decreasing Ca2+ responses measured using imaging (Freund et al., 2016; Sinnen et al., 2016).
GluN2B phosphorylation at S1166, which regulates Ca2+ permeability, is mediated by PKA, which is anchored to a site adjacent to CaN on the AKAP150 scaffold (Dell'Acqua et al., 2002; Oliveria et al., 2003, 2007). Proximity between PKA and CaN, mediated by AKAP150, has been shown to aid bidirectional regulation between the two enzymes: PKA can phosphorylate RCAN1 (Regulator of Calcineurin 1), which inhibits CaN activity (Kim et al., 2015; Dudilot et al., 2020); similarly, CaN can dephosphorylate the RIIα regulatory subunit of PKA, which enhances its capture of the PKA catalytic subunit and inhibition of its activity (Blumenthal et al., 1986; Church et al., 2021). In the absence of anchored CaN, PKA RIIα would remain phosphorylated favoring increased PKA signaling. We previously found that Aβo stimulates PKA as well as CaN signaling (Sanderson et al., 2021); thus, loss of PKA negative regulation by CaN could explain the higher levels of pS1166 and the increase in NMDAR QCT amplitudes we observed in AKAPΔPIX neurons following Aβo treatment. Indeed, phospho-PKA RIIα was upregulated under basal conditions in AKAPΔPIX neurons. However, in contrast, while we observed no differences in pS1166 GluN2B levels under basal conditions between WT and AKAPΔPIX neurons, we did observe an Aβo-mediated increase in pS1166 in AKAPΔPIX neurons, indicating that a PKA-activating stimulus, such as Aβo, is necessary to elicit the differential responses we see for these two substrates when CaN anchoring is disrupted.
We also found that genetic disruption of synaptic CaN anchoring and pharmacologic blockade of mGluR1 signaling fully prevented Aβo-mediated activation of cofilin and downstream dendritic spine loss. In contrast, blockade of mGluR5 signaling prior to Aβo application variably blocked cofilin dephosphorylation and led to only a partial prevention of spine loss. These findings are somewhat in agreement with Um et al. (2013) where 500 nM Aβo modestly reduced the spine density by 10% after 5 h, and this small decrease was prevented by mGluR5 antagonism. Here, we applied 1,000 nM Aβ oligomers for 24 h and observed a larger ∼30% reduction in spine density that was only partially prevented by mGluR5 antagonism but fully prevented by mGluR1 antagonism. What is the reason for the differential contributions of mGluR1 and mGluR5 signaling in Aβo regulation of synaptic Ca2+ balance and spine density? In many studies, these two receptors are grouped together as simply Group I mGluRs and their functions are presented as similar, since they both couple to the same G-protein effectors. Still, other studies illuminate distinctions between mGluR1 and mGluR5 (Manzoni and Bockaert, 1995; Mannaioni et al., 2001; Benquet et al., 2002), though the molecular mechanisms potentially underlying any functional segregation (such as differential localization to their respective targets) is unknown. mGluR5 is more highly expressed in hippocampal neurons (Baude et al., 1993; Fotuhi et al., 1994) and has been previously implicated in Aβo-mediated synaptic deficits (Renner et al., 2010; Rammes et al., 2011; Um et al., 2013; Haas et al., 2017). The role of mGluR1 in AD, in contrast, has been described as being upstream of amyloid β processing (Ovsepian et al., 2017), though there is a possible positive feedback mechanism, as seen in aged APP/PS1 mice and Aβo-treated cortical neuronal cultures where mGluR1 expression was increased (Ostapchenko et al., 2013). There is also evidence for mGluR1 colocalizing with Aβ oligomers and NMDAR GluN2B subunits (Taniguchi et al., 2022) and contributing to neuronal hyperactivity in proximity to Aβ plaques (Ovsepian et al., 2017). Thus, it is likely that Aβo engage both mGluR5 and mGluR1 signaling but perhaps over somewhat different time scales. The 45 min Aβo treatment that resulted in decreased NMDAR Ca2+ influx was prevented by blocking mGluR1, but not mGluR5; in contrast, we did observe some role for mGluR5 signaling when using a longer 24 h Aβo treatment that results in spine loss. It is also possible that mGluR1 and mGluR5 couple to and engage different targets with distinct efficacies. Accordingly, Um et al. found that both mGluR1 and mGluR5 promoted basal Fyn activation, but only mGluR5 physically associated with PrPc and Fyn, to mediate Aβ-induced Fyn activation (Um et al., 2013).
Finally, some previous studies used MPEP instead of MTEP to block mGluR5 activity; unfortunately, MPEP has been shown to have off-target effects, such as inhibition of NMDARs at 20 μM in rat cortical neurons (Movsesyan et al., 2001; for review, see Lea and Faden, 2006). Given that NMDAR blockade prevents Aβo-mediated spine loss (Shankar et al., 2007; Martinez et al., 2024), using MPEP could falsely attribute mGluR5 contribution to signaling events resulting from NMDAR activity. By using MTEP, we hoped to discriminate the specific contribution of mGluR5 to Aβo-mediated synaptic dysfunction. Indeed, MTEP alone did not have an appreciable effect on NMDAR Ca2+ influx and only had partial effects on spine density. Nonetheless, our finding that Aβo instead utilizes mGluR1 and anchored CaN signaling to decrease NMDAR Ca2+ influx and reduce spine density provides further clarity into the molecular mechanisms associated with synaptic dysfunction in AD and opens new avenues for investigating potential new therapeutics, including those targeting mGluR1.
Footnotes
We thank Steven Coultrap for generous donation of reagents; Christina Winborn, Nicole Rumian, and Jennifer Sanderson for technical guidance; and Tyler Martinez, Brian Ware, and Chelsie Kadgien for important intellectual discussions.
This work was supported by the National Institutes of Health (F32AG071073, R01NS110383, R01NS040701, R35NS116879) and gift funds from the University of Colorado Alzheimer’s and Cognition Center.
The authors declare no competing financial interests.
- Correspondence should be addressed to Mark L. Dell’Acqua at mark.dellacqua{at}cuanschutz.edu or Matthew J. Kennedy at matthew.kennedy{at}cuanschutz.edu.
