Abnormal calcium activity and CREB phosphorylation are associated with motor memory impairment in presenilin-1 mutant knock-in mice

Abstract

Introduction

Presenilin (PS) gene mutations cause memory impairment in early-onset familial Alzheimer’s disease (FAD), but the underlying mechanisms remain unclear.

Methods

We examined the effects of the PS1 M146V FAD mutation on motor learning, motor learning-related changes in neuronal Ca²⁺ activity and CREB phosphorylation in the primary motor cortex.

Results

We found that PS1 M146V knock-in mice displayed long-term deficiencies in motor skill learning. Ca²⁺ levels are altered in a cortical layer and neuron type-specific manner in PS1 mutant mice as compared to WT control mice. Notably, while running caused a significant increase of CREB phosphorylation in WT mice, it led to a significant decrease of CREB phosphorylation in layer 5 neurons of mutant mice.

Discussion

These findings suggest that alterations of Ca²⁺ activity and CREB phosphorylation in deep cortical layers are early events leading to memory impairment in the PS1 mutation-related familial form of AD.

Graphical Abstract

1. Background

Alzheimer’s disease (AD) is a progressive neurodegenerative disease characterized by amyloid plaque deposition, formation of neurofibrillary tangles, synapse loss, and memory impairment[1]. AD occurs in sporadic forms of unknown origin and in less common, early-onset familial forms that have a heritable, genetic basis, accounting for a minority (5-10%) of all AD cases. Mutations of three genes, the Amyloid Precursor Protein (APP) gene, and the PSEN1 and PSEN2 genes, which encode respectively Presenilin 1 (PS1) and Presenilin 2 (PS2) constitute the majority of familial AD (FAD) cases. More than 180 FAD mutations in PSEN1 have been reported, accounting for 30% to 70% of all early-onset FAD cases[2].

It is well established that presenilins serve as the catalytic components of the γ-secretase enzyme, an intramembrane protease that functions in cleaving the amyloid precursor protein (APP) into Aβ peptides of varying lengths[3, 4]. Mutations in PS1 or PS2 are often associated with increased production of Aβ42[5-9]. In addition to its role in APP processing, many lines of evidence have shown that PS1 mutations could cause sustained deregulation of Ca²⁺ in the endoplasmic reticulum (ER), mitochondria and lysosomes, leading to amyloid plaque formation, memory loss and neuronal death in AD[10-13]. Consistently, in the triple transgenic mouse model (3xTg AD) harboring PS1M146V, APPSwe, and tauP301L transgenes, deficits of Ca²⁺-dependent processes, such as long-term potentiation, occur before amyloid plaque and tangle pathology[14].

Mice expressing targeted mutations of FAD genes provide models for studying familial forms of AD[15-17]. One such model is the PS1 M146V knock-in mouse. Homozygous PS1 M146V-KI mice (up to 16 months of age) are viable, fertile, normal in size and do not display any gross physical or behavioral abnormalities[16]. PS1 M146V mice exhibit larger medium and late after hyperpolarization in CA3 pyramidal cells as compared with nontransgenic mice or those transgenic mice expressing wild-type PS1[18]. In cultured PS1 M146V-KI neurons expressing the mutant protein, there is a disruption of Ca²⁺ homeostasis in various organelles and Ca²⁺-dependent synaptic plasticity processes[10-13, 19]. It has been shown that PS1 forms complexes with the IP3 receptor (IP3R), an IP3-regulated channel that conducts Ca²⁺ from the endoplasmic reticulum into the cytosol[11, 20]. PS1 FAD mutations are found to increase IP3R open probability, enhancing neuron Ca²⁺ activity[11]. In addition, PS1 FAD mutations are also reported to abolish a putative Ca²⁺ leak current, and to impair Store Operated Ca²⁺ Entry (SOCE) [12, 21].

Cytosolic Ca²⁺ levels have a major role in controlling neuronal function and activity-dependent synaptic plasticity via two Ca²⁺-regulated enzymes, the protein kinase, CaMKII, and the phosphatase, calcineurin[22-28]. These enzymes respond differentially to Ca²⁺ levels and exert opposing effects on synaptic regulation[29-34]. CaMKII and calcineurin reciprocally regulate downstream signaling pathways that govern synaptic plasticity[28, 35-37]. Among these is the CREB transcription factor, which provides activity-dependent expression of genes necessary for enhancement of dendritic arborization and spines. CREB is activated by CaMKII phosphorylation of CREB S133 and inhibited by dephosphorylation by calcineurin[38, 39]. Consistent with dysregulation of Ca²⁺ signaling in the brain, PS1 M146V KI mice show elevated phosphorylated CREB (pCREB) in whole-brain extracts[40]. However, the impact of PS1 mutation on neuronal Ca²⁺ activity and pCREB regulation in the process of learning and memory formation remain unknown.

The early phase of AD is associated with impairments in memory formation and retrieval and by spatial disorientation and loss of routine cognitive ability, all of which depend upon the hippocampus and associative cortices[41]. In addition to spatial and cognitive memory deficits, impairment of motor functions also takes place in AD [42] and often precedes cognitive impairment, making this impairment a significant marker for the preclinical disease[42]. Consistent with a contribution of loss of motor cortex function to the Alzheimer’s phenotype, senile plaques and neurofibrillary tangles are also observed in the primary motor cortex of Alzheimer’s patients[43].

To better understand learning deficits in the early phase of FAD, we examined the effect of the PS1 M146V FAD mutation on motor learning and on Ca²⁺ activity and CREB phosphorylation in pyramidal and somatostatin-expressing neurons in L2/3 and L5 of the mouse primary motor cortex. Studies were performed in FAD mutant M146V KI mice and in WT mice, either under quiet resting conditions or while running on a treadmill. We show that PS1 M146V mutant mice exhibit deficits in motor skill learning over long term (days, intersessions), but not short-term (tens of minutes, intrasessions). Additionally, the PS1 M146V mutation reduces somatic Ca²⁺ activity and pCREB in layer 5 neurons of the motor cortex under the running condition. These results suggest a model in which FAD PS1 mutation deregulates neuronal Ca²⁺ and uncouples phosphorylation of CREB from its neuronal activity-dependent regulation, thereby contributing to memory deficits.

2. Methods

2.1. Experimental Animals

PS1 (M146V) KI mice (Jax stock number: 004193) were used in this study. These mice were maintained on a homogeneous genetic background (C57BL/6). PS1(M146V). KI mice were bred and housed in the animal facility at NYU Langone Medical Center, with 12h/12h light/dark cycle and libitum access to food and water. Mice heterozygous (+/KI) for the PS1 mutation were intercrossed to generate homozygous mutant knock-in mice (KI/KI) and homozygous wild-type mice (+/+). SST-IRES-Cre mice (Sst^{tm2.1(cre)Zjh}, Jax Strain #:013044) were crossed with PS1(M146V) KI mice to generate homozygous mutant knock-in mice (KI/KI) expressing Cre in SST-expressing interneurons. For genotyping of PS1 mutants, the forward primer (5’-3’ AGG CAG GAA GAT CAC GTG TTC AAG TAC) and the reverse primer (5’-3’ CAC ACG CAC ACT CTG ACA TGC ACA GGC) were used. 3-4 months old animals of both sexes were used for behavior tests and in vivo imaging. 5-8 month-old animals of both sexes were used for detecting CREB phosphorylation. All animal studies were performed with an approved protocol (IA16-01351) from New York University Langone Medical Center's Institutional Animal Care and Use Committee.

2.2. Rotarod test

An EZRod system with a test chamber (44.5 cm × 14 cm × 51 cm, Accuscan Instruments) was used in this study. Animals were placed on the motorized rod (30 mm in diameter) in the chamber. The rotation speed of the rod gradually increased from 0 to 100 r.p.m. over the course of 3 min. The rotation speed at which the animal was unable to keep up with the increasing speed and fell was recorded. Rotarod training/testing was performed in one 30-min session at the same time per day (20 trials in total). The maximum speed that animals stayed on the rotarod and the latency to fall (measured in seconds) were measured over each of the 20 trials. Latency to fall was used as the primary performance metric.

2.3. AAV injection

To perform AAV injections in mice, we followed our previously published protocols [44, 45]. In brief, mice were anesthetized by intraperitoneal injection (5–6 μl g⁻¹ body weight) of KX (20 mg ml⁻¹ of ketamine and 3 mg ml⁻¹ of xylazine in 0.9% NaCl solution). The hair of the scalp was carefully shaved with a double-edged razor blade. A midline incision of the scalp was made with sterile surgical scissors. The brain area to be imaged was localized based on the stereotactic coordinates and marked with a fine marker. A small amount of glue was placed around the edges of the internal opening of the skull holding plate, which was pressed against the skull for a few seconds. It was made sure that the area to be injected was exposed in the center of the internal opening of the skull holder. The genetically encoded Ca²⁺ indicator GCaMP6s was used for Ca²⁺ imaging in the primary motor cortex. In general, 0.1–0.2 μl of AAV-synapsin-GCaMP6s (AAV serotype 2/1; >2 × 10¹³ (GC/ml) titer from the University of Pennsylvania Gene Therapy Program Vector Core) was diluted two times in artificial cerebrospinal fluid (ACSF) and slowly injected (Picospritzer III; 20 p.s.i., 20 ms, 0.3 Hz) into L2/3 or L5 of the motor cortex (0–1.5 mm posterior from bregma and 0–1.5 mm lateral from midline) using a glass microelectrode. ACSF contained 124 mM NaCl, 2.5 mM KCl, 2.0 mM MgSO₄, 1.25 mM KH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 4 mM sucrose, and 2.5 mM CaCl₂ and was equilibrated with 95% O₂, 5% CO₂, to a pH of 7.4. For imaging somas of L2/3 SST-expressing interneurons, Cre-dependent AAV-GCaMP6m (AAV-CAG-Flex-GCaMP6m; serotype 2/1; >2 × 10¹³ (GC/ml) titer from the University of Pennsylvania Gene Therapy Program Vector Core) was injected into SST-IRES-Cre mice which expressed Cre recombinase exclusively in SST-expressing interneurons. Post-operative analgesia was administered using ibuprofen (administered via drinking water at 0.2 mg/mL) for 48 hours following surgery. Animals were monitored post-surgery until full recovery. Mice were prepared for Ca²⁺ imaging around 3 weeks after virus injection.

2.4. Two-photon Ca²⁺ imaging of dendrites and somas of layer 2/3 and layer 5 neurons

24 hours before imaging, a head holder with two micro-metal bars was attached on top of the skull with dental acrylic cement (Lang Dental Manufacturing Co.) as described previously[46]. Open-skull surgery was performed to place a glass window over the region of interest of the skull. The center of the region for Ca²⁺ imaging is in the primary motor cortex (~0.5 mm anterior from bregma and ~1.2 mm lateral from the midline). During an imaging session, the head of the mouse was restrained by attaching the head holder to a custom stage fitted with a free-floating treadmill[44]. Mice were kept on the stage for a maximum of 1 hour at a time. Layer 1 dendrites (depth of ~50 μm below the pial surface) projected from L2/3 (depth of ~250 μm below the pial surface) or L5 (depth of ~500 μm below the pial surface) pyramidal neurons and the somas in L2/3 or L5 were imaged using an Olympus Fluoview 1000 two-photon microscope equipped with a Ti:Sapphire laser (MaiTai DeepSee, Spectra Physics). Images were acquired with GaAsP detector with two-photon excitation at 920 nm and emission collection at 495-540 nm. The average laser power on the tissue sample was 30 ~ 50 mW, and the frame rate was 2 Hz for imaging. A 25X objective (N.A. 1.05, Olympus) was used. For the running protocol, animals were forced to run forward on the treadmill (speed increased gradually from 0 to 6 cm/second within ~3 seconds). Image acquisition was performed using FV10-ASW v.3.0 software and analyzed post hoc using NIH ImageJ software.

2.5. Image analysis

The Ca²⁺ activity indicated by GCaMP6 fluorescence changes was analyzed post hoc using Image J software (NIH) as described previously[47]. In Figures 1-4, we used the genetically encoded Ca²⁺ indicator GCaMP6s under the control of the synapsin promoter for Ca²⁺ imaging. As the synapsin promoter is not cell-type-specific, it drives expression across a broad population of neurons, including both excitatory and inhibitory types. An individual pyramidal neuron generally exhibits a large, triangular soma and a prominent apical dendrite, which are visually distinguishable under our imaging conditions. While we cannot entirely rule out the possibility that a small number of SST neurons were inadvertently included in the population classified as pyramidal neurons, such contamination is minimal due to the small percentage of SST neurons relative to pyramidal neurons.

Figure 1. Long-term memory deficits and abnormal dendritic Ca²⁺ transients of layer 2/3 pyramidal neurons in PS1 mutant mice.

(A) Schematic of rotarod training and behavioral assessments. (B) Latency to fall from the rotarod across three consecutive days showed no significant difference between WT and PS1 mice during the initial two days. However, a significant decline in performance was observed in the PS1 animals on the final day. (WT: 8 mice; PS1: 8 mice, Mann-Whitney unpaired nonparametric test). *P < 0.001. (C) Two representative time-lapse images and fluorescent traces of dendritic Ca²⁺ activity in WT mice during running condition. Scale bar: 10μm. (D) Two representative time-lapse images and fluorescent traces of dendritic Ca²⁺ activity in PS1 mice during running condition. Scale bar: 10μm. (E-H) Cumulative distribution of total integrated activity (E), peak ΔF/F₀ (F), frequency (G), and duration (H) of dendritic Ca²⁺ transients in WT and PS1 during quiet resting (WT: 81 dendrites from 6 mice; PS1: 101 dendrites from 5 mice. Two-sample Kolmogorov-Smirnov Test). (I-L) Cumulative distribution of total integrated activity (I), peak ΔF/F₀ (J), frequency (K), and duration (L) of dendritic Ca²⁺ transients in WT and PS1 during running (WT: 61 dendrites from 6 mice; PS1: 51 dendrites from 5 mice. Two-sample Kolmogorov-Smirnov Test). *P < 0.05, **P < 0.01, ***P < 0.001. n.s., not significant

Figure 4. The somas of layer 5 pyramidal neurons show decreased Ca²⁺ activity during resting and running.

(A) A representative image of somatic Ca²⁺ activity in WT mice (left). Fluorescence traces of 4 different somas during quiet resting and running states (right). Scale bar: 10μm. (B) A representative image of somatic calcium activity in PS1 mice (left). Fluorescence traces of 4 different somas during quiet resting and running states (right). (C-F) Cumulative distribution of total integrated activity (C), peak ΔF/F₀ (D), frequency(E), and duration (F) of dendritic calcium transients in WT and PS1 during quiet resting (WT: 49 somas from 4 mice; PS1: 307 somas from 5 mice. Two-sample Kolmogorov-Smirnov Test). (G-J) Cumulative distribution of total integrated activity (G), ΔF/F₀ (H), frequency (I), and duration (J) of dendritic calcium transients in WT and PS1 during running (WT: 52 somas from 4 mice; PS1: 242 somas from 5 mice. Two-sample Kolmogorov-Smirnov Test). *P < 0.05, **P < 0.01, ***P < 0.001. n.s., not significant

To image Ca²⁺ in SST interneurons, we generated SST Cre:PS1 (M146V) KI mice and expressed GCaMP6 specifically in SST interneurons by injecting AAV encoding floxed GCaMP6 in the motor cortex.

The GCaMP6 fluorescence (F) during quiet resting and running was measured by averaging pixels within each identifiable area (Regions of interests, ROIs). A background fluorescence value of the vessels was subtracted from all F values. Then ΔF/F₀ was calculated as ΔF/F₀ = (F-F₀)/ F₀, in which ΔF was F-F₀, and F₀ was the average of 10% minimum F values over 1 min period, representing baseline fluorescence. The threshold for detecting a Ca²⁺ transient was set at greater than three times the standard deviation of baseline fluorescence noise and duration greater than 1s. The frequency of Ca²⁺ transients was calculated as the number of Ca²⁺ transients per minute on each dendrite. The duration was calculated as the total time of Ca²⁺ transients with ΔF/F₀ above the threshold. The peak amplitude was the highest value of the Ca²⁺ transient. The total Ca²⁺activity was the accumulated Ca²⁺ activities above the threshold per minute.

Cumulative fraction refers to the empirical cumulative distribution function (ECDF) of calcium events including total activity, amplitude, frequency and duration. For each experimental group (PS1 M146V vs. wild-type), all measured values were sorted in ascending order, and the cumulative fraction at any value X represents the proportion of observations ≤ X.

Y-axis unit: Dimensionless (range 0-1)

0: Indicates no calcium events with amplitudes ≤ the minimum observed value

1: Represents 100% of calcium events having amplitudes ≤ the maximum observed value

In calcium signal analysis, cumulative fraction plots display the normalized cumulative probability (0-1 scale) of calcium events, where 0 corresponds to the minimum measurable amplitude and 1 represents full inclusion of all measured events.

2.6. Motor training paradigm to induce CREB phosphorylation

Mice were placed on an EZRod system within a test chamber (44.5 cmx 314 cm x 351 cm, Accuscan Instruments). Both WT and PS1 mice were littermates. Animals were placed on the motorized rod (30 mm in diameter) in the chamber and the rotation speed was gradually increased from 0 to 100 r.p.m over 3 min. Mice were subjected to this motor training paradigm for 1 hour.

2.7. Immunohistochemistry

Immediately after training, mice were deeply anesthetized with a mixture of Ketamine (100 mg/kg) and Xylazine (15 mg/kg) and then perfused with 10 ml phosphate buffered saline (PBS) followed by 10 ml 4% paraformaldehyde (PFA) in PBS. Brains were kept in 4% paraformaldehyde at 4 degrees overnight. Twenty-four hours after PFA fixation, brains were rinsed with PBS and 200 μm sections were prepared with a vibratome (Leica VT1000S). Two coronal sections per brain were permeabilized in a buffer containing 1% Triton X-100 and 5% bovine serum albumin (BSA) in PBS for 3 hours. Sections were incubated overnight with primary antibodies against pCREB (Cell Signaling Cat. No. 9198, 1:800, rabbit) and CREB (Cell Signaling Cat. No. 4820, rabbit, 1:400; Santa Cruz Biotechnology Cat. No. sc-377154 AF594, 1:100, mouse) in a buffer containing 0.1% Triton X-100 and 5% BSA. Sections were then washed 3 times with 0.05% Tween-20 in PBS and incubated overnight with Alexa Fluor-conjugated goat anti-rabbit antibodies (Invitrogen Cat. No. A11008 (red), A21428 (green), 1:500) in a buffer containing 0.1% Triton X-100 and 5% BSA. Sections were then washed 3 times with 0.05% Tween-20 and mounted for imaging. Brain slices were mounted using ProLong Gold antifade reagent (Thermo Fisher Scientific, Cat. No. P10144), and imaged by an LSM 800 Zeiss confocal laser scanning microscope.

2.8. IHC image acquisition and data analysis

Confocal images of cortical slices were obtained using a Zeiss 800 confocal microscope equipped with a 20X lens (N.A. 0.75) with a step size of 2 μm. Image processing and data acquisitions were done using Fiji (version 2.14.0). We subtracted the background signal and generated a Z projection to create a single image for data acquisition. Perpendicularly positioned to the L2/3 and L5 in the primary motor cortex, we placed a box with the following dimensions: 800 μm in length and 200 μm in width. We then measured the fluorescence intensity of anti-pCREB antibody signals from L1 to L6, which was used as the proxy for the pCREB level across different cortical layers. For each animal, the fluorescence intensity traces from L1 to L6 were measured from 1 or 2 brain sections, one per hemisphere, and were averaged per animal. In total, we averaged 15 traces from 6 mice per genotype at resting conditions and 15 traces from 10 mice per genotype at running conditions. For extracting the fluorescence from L2/3 and L5, we created squared ROIs (100 μm x 100 μm in size) positioned over L2/3 or L5 and extracted the fluorescence values within the ROIs. We measured 2-4 ROIs for L2/3 and L5 per slide per mouse. Each dot in the figure represents one ROI (n= WT resting: 20 ROIs from 6 mice; WT running: 20 ROIs from 10 mice; PS1 resting: 20 ROIs from 6 mice; PS1 running: 20 ROIs from 10 mice

2.9. Statistics

Statistical analyses were performed using GraphPad Prism (version 9.0) and IBM SPSS 23. Statistical comparisons were performed across individual somas/dendrites (e.g., N=307 somas from 5 mice), with each soma/dendrite treated as an independent observation for Ca²⁺ signal differences between PS1 and WT genotypes. The two-tailed Student’s t test was used to compare two groups if datasets were normally distributed after running the Kolmogorov-Smirnov test. For those distributions that did not pass the Kolmogorov-Smirnov test, the non-parametric tests were performed instead. The two-sample Kolmogorov-Smirnov (K-S) test was employed to statistically compare the cumulative distribution functions (CDFs) of Ca²⁺ event parameters (total activity, amplitude, frequency and duration) between PS1 and wild-type groups in Figures 1-5. In Figure 6, two-way ANOVA; Tukey's multiple comparison post hoc test was used to quantify pCREB expression between PS1 and WT genotypes. Data are presented as average ± SEM in Figures 1B, 6 and Supplementary Figure 1. A value of P < 0.05 was considered statistically significant.

Figure 5. Reduced somatic calcium of layer 2/3 and layer 5 SST interneurons under running condition.

(A) A representative image of layer 2/3 (upper) and layer 5 (lower) somatic Ca²⁺activity in WT mice. Fluorescence traces of 4 different somas during quiet resting and running states (right). Scale bar: 20μm. (B) A representative image of layer 2/3 (upper) and layer 5 (lower) somatic Ca²⁺ activity in PS1 mice. Fluorescence traces of 4 different somas during quiet resting and running states (right). (C-D) Cumulative distribution of somatic Ca²⁺ activity of layer 2/3 SST neurons in WT and PS1 during quiet resting and running states (WT: 194 somas from 3 mice; PS1: 207 somas from 3 mice. Two-sample Kolmogorov-Smirnov Test). (E-F) Cumulative distribution of somatic Ca²⁺ activity of layer 5 SST neurons in WT and PS1 during quiet resting and running states (WT: 130 somas from 3 mice; PS1: 131 somas from 3 mice. Two-sample Kolmogorov-Smirnov Test). *P < 0.05, **P < 0.01. n.s., not significant

Figure 6. PS1 mice display higher pCREB activity at resting condition but fail to trigger running-induced pCREB activity compared to WT mice.

(A-D) Five to eight-month-old WT and PS1 mice were forced to run in a motorized rod for one hour (B, D) or kept in their home cage for over the same period (A, C). Immediately after the motor running training ended, the mouse brain was sliced and processed to perform immunohistochemical staining against pCREB. The regions of interest (ROIs, 200 μm weight per 800 μm length) are represented by dotted lines over the primary motor cortex. Magnified ROIs with pCREB-positive cell bodies are shown from layer I to layer VI. Scale bar: 200 μm. n= 6 WT resting mice, 10 WT running mice; n= 6 PS1 resting, 10 PS1 running mice. (E-H) Fluorescence profile extracted from the ROIs of the brain slices stained with anti-pCREB. Royal blue trace: WT at resting; Green trace: WT at running; Pink trace: PS1 at resting; Turquoise trace: PS1 at running. (I) pCREB fluorescence in Layer 2/3 at resting and running conditions. (J) pCREB- fluorescence in Layer 5 at resting and running conditions. Two-way ANOVA; Tukey's multiple comparison post hoc test, n= WT resting and PS1 resting: 20 ROIs (100 μm x 100 μm) from 6 mice per genotype; WT running and PS1 running: 20 ROIs from 10 mice per genotype.

3. Results

3.1. Long-term motor learning deficits in PS1 mutant mice.

To investigate the impact of PS1 mutation on motor learning, we examined the performance of 3-month-old PS1 mutant mice using a rotarod training paradigm. Animals were placed on a rotarod, the speed of which increased from 0 to 100 rpm over 3 minutes. The length of time that each animal remained on the rotarod was recorded as a measure of performance.

During the first two days of training, the performance of PS1 mutant mice was comparable to that of wild-type animals (Figure 1A-B), indicating no significant deficits in the early phases of motor skill acquisition. On the third training day, however, the performance of PS1 mutant mice was significantly lower than that of WT mice (Figure 1A-B; WT: n= 8 mice; PS1: n= 8 mice; P < 0.001). Together, these observations suggest that PS1 mutant mice showed deficits in performance after motor skill learning over long-term (days, intersessions), but not over short-term (tens of minutes, intrasessions).

3.2. Abnormally high dendritic Ca²⁺ activity in layer 2/3 pyramidal neurons of PS1 mutant mice during both resting and running

Previous studies have shown abnormal dendritic Ca²⁺ activity in L2/3 neurons of the motor cortex in APPSwe/PS1dE9 mice expressing FAD mutant APP and mutant PS1[48]. Furthermore, on those dendrites with abnormally prolonged dendritic Ca²⁺ activity, the activity and size of postsynaptic dendritic spines were significantly reduced [48], which may contribute to motor learning deficits. To investigate whether and how PS1 M146V mutation affects dendritic Ca²⁺ activity, we performed Ca²⁺ imaging of apical dendrites of L2/3 pyramidal neurons within the primary motor cortex under both quiet resting and running conditions (Figure 1C, D). In the resting state, mutant L2/3 pyramidal neurons exhibited significantly higher total integrated activity and frequency of dendritic Ca²⁺ transients as compared to WT animals (Figure 1E, G and Supplemental figure 1a; WT: n= 81 dendrites from 6 mice; PS1: n= 101 dendrites from 5 mice; P < 0.003). The peak amplitude of dendritic Ca²⁺ transients was significantly lower in PS1 mice (Figure 1F, P < 0.001), while the duration of these transients was comparable between WT and PS1 mice (Figure 1H; P = 0.703). In the running state, PS1 mice displayed a significant increase in total integrated activity, peak amplitude, frequency, and duration of dendritic Ca²⁺ transients relative to WT mice (Figure 1I-L and Supplemental figure 1b; WT: n= 61 dendrites from 6 mice; PS1: n= 51 dendrites from 5 mice; P < 0.038). Consistent with the high level of dendritic Ca²⁺ activity observed in APP/PS1 animals[48], these findings reveal abnormally high Ca²⁺ activity in apical dendrites of L2/3 pyramidal neurons of PS1 M146V knock-in mice during both quiet resting and running conditions.

3.3. Increased somatic Ca²⁺ activity of layer 2/3 neurons of PS1 mutant mice during both resting and running

In addition to dendritic Ca²⁺, we further performed Ca²⁺ imaging of L2/3 neuronal soma in the motor cortex (Figure 2). The total integrated Ca²⁺ activity showed significant increases during both resting and running conditions in PS1 mutant mice compared to WT mice (Figure 2C, G and Supplemental figure 1c-d; Resting: WT (n= 92 somas from 6 mice); PS1 (n= 85 somas from 5 mice); Running: WT (n= 93 somas from 6 mice); PS1 (n= 83 somas from 5 mice); P < 0.02). There was no significant alteration in the peak amplitude, frequency, or duration of somatic Ca²⁺ transients (Figure 2D-F, H-J; P > 0.06). These findings reveal that while specific aspects of Ca²⁺ dynamics such as peak amplitude, frequency, and duration are not significantly altered, PS1 M146V mutation causes a significant increase of the overall somatic Ca²⁺ activity in L2/3 neurons in both resting and running animals.

Figure 2. The somas of layer2/3 pyramidal neurons show increased calcium activity during resting and running.

(A) A representative image of somatic calcium activity in WT mice (left). Fluorescence traces of 4 different somas during quiet resting and running states (right). Scale bar: 10μm. (B) A representative image of somatic calcium activity in PS1 mice (left). Fluorescence traces of 4 different somas during quiet resting and running states (right). (C-F) Cumulative distribution of total integrated activity (C), peak ΔF/F₀ (D), frequency (E), and duration (F) of dendritic calcium transients in WT and PS1 during quiet resting (WT: 92 somas from 6 mice; PS1: 85 somas from 5 mice. Two-sample Kolmogorov-Smirnov Test). (G-J) Cumulative distribution of total integrated activity (G), peak ΔF/F₀ (H), frequency (I), and duration (J) of dendritic Ca²⁺ transients in WT and PS1 during running (WT: 93 somas from 6 mice; PS1: 83 somas from 5 mice. Two-sample Kolmogorov-Smirnov Test). *P < 0.05, **P < 0.01. n.s., not significant

3.4. Increased dendritic Ca²⁺ activity of layer 5 pyramidal neurons in PS1 mutant mice during running

To investigate the impact of PS1 M146V mutation on neuronal Ca²⁺ further, we compared Ca²⁺ activity of apical dendrites in L5 pyramidal neurons between WT and PS1 mice. Unlike in L2/3 neurons, we detected no significant difference in total integrated Ca²⁺ activity between WT and PS1 mutant animals in the resting state (Figure 3A-C and Supplemental figure 1e; WT: n = 44 dendrites from 5 mice; PS1: n = 69 dendrites from 4 mice; P = 0.862). Detailed analysis showed that PS1 mice exhibited significantly higher peak amplitude but lower frequency and duration of dendritic Ca²⁺ transients compared to WT mice at rest (Figure 3D-F; P < 0.005).

Figure 3. Dendritic Ca²⁺ transients of layer 5 pyramidal neurons increased in PS1 mutant mice during running.

(A) Two representative time-lapse images and fluorescent traces of dendritic Ca²⁺ activity in WT mice during running. Scale bar: 10μm. (B) Two representative time-lapse images and fluorescent traces of dendritic Ca²⁺ activity in PS1 mice during running. (C-F) Cumulative distribution of total integrated activity (C), peak ΔF/F₀ (D), frequency (E), and duration (F) of dendritic Ca²⁺ transients in WT and PS1 during quiet resting (WT: 44 dendrites from 5 mice; PS1: 69 dendrites from 4 mice. Two-sample Kolmogorov-Smirnov Test). (G-J) Cumulative distribution of total integrated activity (G), ΔF/F₀ (h), frequency (I), and duration (J) of dendritic Ca²⁺ transients in WT and PS1 during running (WT: 44 dendrites from 5 mice; PS1: 53 dendrites from 4 mice. Two-sample Kolmogorov-Smirnov Test). *P < 0.05, **P < 0.01, ***P < 0.001. n.s., not significant

On the other hand, when animals ran on the treadmill, PS1 animals showed significantly higher total integrated Ca²⁺ activity and peak amplitude in L5 (Figure 3G-H and Supplemental figure 1f; WT: n = 44 dendrites from 5 mice; PS1: n = 53 dendrites from 4 mice; P < 0.020), but lower Ca²⁺ frequency compared to WT animals (Figure 3I; P = 0.020). There was a slight but not significant reduction in the duration of dendritic Ca²⁺ activity in PS1 mice (Figure 3J; P = 0.061). Together, the above results show abnormally high Ca²⁺ activity in apical dendrites of L5 pyramidal neurons of PS1 M146V knock-in mice under the running, but not resting condition.

3.5. Decreased somatic Ca²⁺ activity of layer 5 neurons in PS1 mutant mice during resting and running

In contrast to the increased somatic Ca²⁺ activity of L2/3 neurons, L5 neurons in PS1 mutant mice exhibited reduced total integrated Ca²⁺ activity and duration of somatic Ca²⁺ as compared with their WT littermates during both resting and running conditions (Figure 4A-C, F, G, J and Supplemental figure 1g-h; Resting: WT (n = 49 somas from 4 mice); PS1 (n = 307 somas from 5 mice); Running: WT (n = 52 somas from 4 mice); PS1 (n = 242 somas from 5 mice); P < 0.05). Detailed analysis reveals that the peak amplitude and frequency of somatic Ca²⁺ transients in L5 neurons showed no significant differences between PS1 and WT mice under resting conditions (Figure 4D, E; P > 0.37).

On the other hand, in the running state, the peak amplitude and duration of somatic Ca²⁺ transients of L5 neurons were significantly lower in PS1 mice relative to WT mice (Figure 4H, J; P = 0.024), although the frequencies were comparable between WT and PS1 mice (Figure 5I; P = 0.388). These results reveal a significant reduction of L5 somatic Ca²⁺ in PS1 mutant mice relative to WT in the resting and running conditions, suggesting that the impacts of PS1 M146V mutation on somatic Ca²⁺ transients differ between L2/3 and L5 neurons.

3.6. Reduced somatic Ca²⁺ activity of both layer 2/3 and layer 5 somatostatin-expressing interneurons in PS1 mutant mice under running condition

Previous studies have shown that somatostatin-positive (SST+) interneurons form GABAergic synapses onto the dendrites of pyramidal neurons to inhibit their activity [49, 50]. To further understand the effect of the PS1 M146V FAD mutation, we generated SST ^Cre:PS1 (M146V) KI mice and expressed GCaMP6 specifically in SST interneurons by injecting AAV encoding floxed GCaMP6 in the motor cortex (see Methods). SST+ interneuron Ca²⁺ activity was examined during resting and running in both L2/3 and L5 (Figure 5A-B). We found no significant difference in somatic Ca²⁺ activity of SST+ interneurons between WT and PS1 animals under the resting conditions (Figure 5C, E; L2/3, WT: 3 mice, 194 somas; PS1: 3 mice, 207 somas; L5, WT: 3 mice, 130 somas; PS1: 3 mice, 131 somas. P > 0.05). However, somatic Ca²⁺ activity of SST+ interneurons during running was significantly lower in both L2/3 and L5 of PS1 mice as compared to WT mice (Figure 5D, F; P < 0.019). Because axonal terminals of SST+ neurons form inhibiting synapses onto apical dendrites of pyramidal neurons[51, 52], the reduced activity of SST+ interneurons likely contributed to the increased dendritic Ca²⁺ activity of L2/3 and L5 pyramidal neurons (Figures 1 and 3)

3.7. PS1 mice fail to trigger running-induced pCREB activity in layer 5 of the motor cortex

Previous studies have reported that the brains of PS1 M146V KI mice show elevated phosphorylation of CREB[40], which plays a key role in the processes of learning and memory[53, 54]. Given our findings of memory deficits and alterations of Ca²⁺ activity in the motor cortex of PS1 mutant mice, we investigated whether motor training affects CREB phosphorylation in the motor cortex of PS1 mutant compared to WT littermates by examining CREB Serine 133 phosphorylation, under resting and running conditions. In WT mice, we detected low levels of pCREB staining in both L2/3 and L5 under resting conditions (Figure 6A, E, I, J, n= 6 mice), and increased pCREB upon motor running activity (Figure 6B, E, I, J, n= 10 mice). This suggests that in WT, activity of motor cortex neurons induces CREB phosphorylation.

Consistent with previous studies showing elevated CREB phosphorylation in whole PS1 M146V KI mouse brain[40], under the resting condition, the levels of pCREB in both L2/3 and L5 of PS1 mice were significantly higher than in the respective layers in WT mice (Figure 6C, G, I, J; L2/3: WT resting vs PS1 resting: P = 0.0097; L5: WT resting vs PS1 resting: P = 0.0014). Notably, in L2/3 of PS1 mice, the pCREB signal did not further increase in the running state when compared to the resting state (Figure 6D, F, I; n= 10 mice). Moreover, the pCREB activity in L5 under the running condition was significantly lower when compared to the resting state in PS1 mice (Figure 6D, F, J; PS1 resting vs PS1 running: P = 0.0215). Under the running condition, the level of pCREB in L5 of PS1 mice was also significantly lower than that in WT mice (Fig. 6B, D, H, J; WT running vs PS1 running: P = 0.005).

Together, these findings reveal layer-specific deregulation of CREB phosphorylation in neurons of mutant motor cortex, with abnormally higher pCREB activity levels under resting conditions in both L2/3 and L5, as well as lower pCREB activity levels in L5 under running conditions in PS1 mice as compared to WT mice.

4. Discussion

AD is a progressive neurodegenerative disorder characterized by memory impairment and formation of extracellular plaques and intracellular neurofibrillary tangles[55-57]. We studied a mouse AD model expressing a knock-in of the M146V FAD mutant PS1 gene. This line exhibits memory impairment, although it does not form the Aβ plaques that are characteristic of human AD[58]. Our study examined acquisition of motor skill during 90 min rotarod training sessions, and retention of motor skill between sessions that were spaced apart by 24-48 hrs. The study indicated that while short-term (minutes) motor learning was intact, retention of longer-term (days) motor learning was impaired in PS1 FAD mutant mice. Furthermore, motor training-induced dendritic and somatic Ca²⁺ transients in cortical neurons are altered in a manner specific to the cortical layer and the neuron type. Most significantly, the activity-dependence of phosphorylation of the CREB transcription factor, a major step in the regulation of synaptic plasticity during memory formation, is compromised. In WT motor cortex, treadmill running elevates CREB phosphorylation in cortical neurons relative to the resting state. In mutant motor cortex, however, CREB phosphorylation is constitutively elevated in the resting state in layer 2/3 and layer 5 neurons, but is reduced in response to motor activity relative to the resting state in layer 5 neurons. This diminished dependence of CREB phosphorylation on neuronal activity in the PS1 M146V FAD mutant may reduce the specificity of synapse strengthening by CREB and contribute to long-term memory deficits in the PS1 mutation-related, familial form of AD.

4.1. PS1 M146V mice show long-term, not short-term, deficits in memory formation

Our study showed that short-term motor learning, during each of the three 90 min training sessions, was similar between WT and mutant mice at 3-4 months of age, indicating the M146V mutation did not impair short-term motor learning processes at the behavioral level. In contrast, retention of motor skill was impaired during the 24 h period separating training sessions, specifically during the 24 h period separating the 2^nd and 3^rd training sessions, but not the interval between the 1^st and 2^nd sessions. These findings suggest that the late stage of motor memory formation, which has been shown to require synthesis of new proteins[59], was specifically impaired.

Our findings are consistent with a previous study showing that homozygous PS1 M146V/M146V knock in mice showed impaired contextual fear memory in one of the five epochs during the 24h-recall test as compared with WT mice at 3 months of age[15]. In addition, 24 hours after contextual fear conditioning, PS1 M146V/- mice exhibited a significantly reduced freezing behavior in all the five epochs of memory recall as compared with PS1+/− littermates[15]. These findings also suggest that defective contextual fear memory caused by the PS1 M146V mutation is likely due to the partial loss of PS1 function. In addition to the contextual fear learning paradigm, previous studies have found various alterations in Morris Water Maze behavior between PS1-M146V mice and WT mice[7, 60]. Specifically, Sun et al. reported that in the Morris water maze task, PS1 KI mice at 3 and 9 months of age exhibited no difference on day 2 for time in quadrant occupancy and platform crossing, but decreases in the time in target quadrant and passes over the platform at day 6[7]. In another study by ODERO et al.[60], the authors reported that 3-month-old PS1M146V mutant mice did not show deficits in the formation of a cognitive map, however 6-month-old PS1M146V mice showed significantly increased path length, potentially related to impaired cognitive functioning[60].

Together with aforementioned behavioral studies of fear and Morris Water Maze learning paradigms using PS1 M146V mutant mice, our findings indicate that PS1 M146V mutant mice show long-term, but not short-term deficits in motor, contextual fear and cognitive memory at relatively young ages (~3 months).

4.2. State- and cell type-dependent alterations of Ca²⁺ activity in pyramidal neurons in the primary motor cortex of PS1 M146V mice

We previously employed transcranial two-photon Ca²⁺ imaging to study dendritic Ca²⁺ activity of layer 2/3 pyramidal neurons in the primary motor cortex of awake, behaving, 3-month-old APPSwe/PS1dE9 mice, which express FAD mutant APP and PS1[48]. We found increased duration and high peak Ca²⁺ amplitude in apical dendrites of layer 2/3 pyramidal neurons in the primary motor cortex during treadmill running[48]. The increase of dendritic Ca²⁺ transients was also induced by the Aβ peptide[48]. The Ca²⁺ activity and size of dendritic spines were significantly reduced in those dendritic branches with abnormal prolonged dendritic Ca²⁺ transients in AD mice. In addition to increased Ca²⁺ transients of apical dendrites and dendritic spines in the motor cortex of APPSwe/PS1dE9 mutant mice[48], neuronal hyperactivity and hypoactivity have been observed in various brain regions of several AD models, including APP/PS1 and PS1 mutant mice[59, 61-67], as well as in AD patients[68-70], underscoring changes of network activity during early stage of AD pathogenesis.

Consistent with the dysregulation of neuronal activity and Ca²⁺ levels observed in various mouse models of AD, we found that in the motor cortex of PS1 M146V mutant mice, the total level of Ca²⁺ transients in apical dendrites of layer 2/3 pyramidal cells consistently exceeded those of the WT, under both resting and treadmill running conditions. In layer 2/3 soma, total Ca²⁺ activity of the mutant also exceeded that of the WT, but to a lesser extent than in apical dendrites, and only during treadmill running. In layer 5 pyramidal neurons, the total level of Ca²⁺ activity in apical dendrites of the PS1 mutant exceeded that of the WT, under treadmill running but not resting conditions. Notably, in contrast to layer 2/3, the total Ca²⁺ activity in layer 5 somas was significantly reduced in the mutant relative to WT mice, under treadmill running but not resting conditions. Together, our findings reveal overall hyperactivity of neuronal networks in the primary motor cortex of the PS1 M146V mutant mice, especially under running conditions. In addition, the reduction of motor running-triggered somatic Ca²⁺ activity of layer 5 neurons in mutant mice as compared to WT mice suggests the disruption of neuronal activity caused by the PS1 M146V mutation is cell type- (layer 2/3 versus layer 5) and state- (resting versus running) dependent.

4.3. Multiple mechanisms may underlie the dysregulation of dendritic and somatic Ca²⁺ activity in PS1 M146V mice

The mechanisms underlying the alterations of Ca²⁺ activity in apical dendrites and somas of layer 2/3 and 5 neurons in PS1 mutant mice remain to be studied. Our results indicate that layer 5 SST inhibitory neurons show diminished somatic Ca²⁺ activity during treadmill running. Because SST interneurons exert strong inhibition on pyramidal neuronal dendrites[44, 47], the reduced SST neuronal activity in the mutant mice could contribute significantly to the increased dendritic activity and Ca²⁺ elevation during running that we observed. Furthermore, SST interneurons also inhibit parvalbumin (PV)-expressing interneurons, which in turn exert strong inhibition on the somas of pyramidal neurons[71]. It is thus possible that reduced SST interneuron activity contributes to the increased PV interneuron activity, which in turn may result in the increased somatic inhibition and reduced Ca²⁺ activity of layer 5 pyramidal neurons as observed in our study. If so, it would suggest that reduced SST interneuron activity caused by the PS1 mutation could be an early and critical step in changes of excitatory neuronal activity. Consistent with this possibility, the number of SST neurons was found to be reduced significantly in the frontal and temporal lobes of AD patients[72-75]. Loss of SST and PV positive interneurons was found in 5xFAD mice[76], in the perirhinal cortex of 6-month-old APP/PS1 mice, and in AD patients[77]. Increased excitability of PV interneurons was reported in the hippocampus of 3–4-month-old APP/PS1 mice[78]. Future studies are needed to investigate whether changes of SST interneuron activity precede those of pyramidal neurons and what is the impact of such changes on the somatic firing of pyramidal neurons in the PS1 M146V mice.

In addition to the reduction in SST neuronal activity, the increase of Ca²⁺ release from the ER via the IP3 receptor in the PS1 mutant neurons[79] may also contribute to increased cytosolic Ca²⁺ activity observed in pyramidal neurons. Various mechanisms have been proposed to underlie altered ER and cytosolic Ca²⁺ levels in PS1 M146V mutant mice. Presenilins have been suggested to function as Ca²⁺ leak channels in the ER[12], and their mutations result in the loss of ER Ca²⁺ leak function activity and ER Ca²⁺ overload, as well as synaptic loss[21]. Interaction of mutant PS1 with the IP3 receptor enhances the IP3R current[11, 80] and genetic reduction of IP3R expression normalizes the increased Ca²⁺ transients in the PS1 mutant neurons and restores normal expression of CREB-dependent genes as well as restoring normal LTP[79]. It has also been reported that lack of PS1 and PS2 gene functions in fibroblasts reduced the activity of the sarcoendoplasmic reticulum Ca²⁺ ATPase (SERCA) pump and Ca²⁺ levels in ER stores[64]. Lastly, PS1 mutations have been suggested to increase γ-secretase cleavage of the STIM1 protein and reduce store-operated Ca²⁺ entry[81]. Thus, the large magnitude of Ca²⁺ transients seen in the PS1 M146V mutant mice could be in part triggered by local, synaptically evoked NMDA receptor Ca²⁺ currents that activate the IP3 receptor and the Ryanodine receptor via elevated Ca²⁺ -induced Ca²⁺ release (CICR)[82], and in part triggered by reduced store-operated Ca²⁺ entry[81].

Regardless of the detailed mechanisms underlying the effect of PS1 mutations on Ca²⁺ activity elevation in ER of neurons, it has been shown that emptying intracellular Ca²⁺ stores reduced Ca²⁺ transients in presynaptic axonal boutons as well as the number of hyperactive neurons in PS45 mice, which contain the PS1G384A mutation without amyloid deposition[66]. These results suggest that elevation of ER and cytosolic Ca²⁺ in presynaptic boutons in PS1 mutant mice likely contribute to increased excitatory neurotransmitter release and inputs to postsynaptic neurons, and consequently, the increased network activity.

It is worth noting that PS1 is present not only in neurons, but also in astrocytes [83]. Our preliminary data suggest that Ca²⁺ in astrocytes was also elevated in the motor cortex during rest and running conditions in PS1 mutant mice (not shown). Many lines of evidence suggest that astrocytic Ca²⁺ is important for modulating neuronal excitability and synaptic transmission via gliotransmitters [84]. Thus, the observed increase in neuronal Ca²⁺ transients in PS1 mutant mice may not simply reflect neuron-type specific effects of the PS1 mutation. The elevated Ca²⁺ in astrocytes could also contribute to changes in Ca²⁺ transients in different types of neurons. Future studies are needed to investigate the glial-neuron interactions to better understand how the PS1 mutation alters neuronal activity and Ca²⁺signalling and contributes to early cognitive deficits in FAD.

4.4. Uncoupling of CREB phosphorylation from neuron activity in PS1 M146V mice

It is well established that neuronal activity-dependent elevation of cytosolic levels of Ca²⁺ induces CaMKII-dependent phosphorylation of the CREB transcription factor, which in turn stimulates the expression of genes that are critical for synaptic plasticity and memory formation[85, 86]. Consistent with our observation of impairment of long-term motor memory in M146V FAD mutant mice, we observed a loss of neuronal activity-dependence of CREB phosphorylation in layer 2/3 and layer 5 neurons. In the WT mice, motor training elevated CREB S133 phosphorylation of neurons in layer 2/3 and layer 5 of the motor cortex. However, in the PS1 mutant, CREB phosphorylation was constitutively elevated in both layer 2/3 and layer 5 in the resting state relative to WT, and treadmill running did not increase CREB phosphorylation. In fact, treadmill running decreased CREB phosphorylation in layer 5 relative to resting, which is consistent with the decrease in Ca²⁺ activity in layer 5 during running versus resting. This further suggests that the M146V FAD mutation releases CREB function from its dependence on neuronal activity, compromising the role of CREB in synaptic plasticity and long-term memory formation. It is important to note that the increased CREB phosphorylation in whole brain of PS1 mutant mice has been previously reported[40]. However, this is the first report that the PS1 M146V mutation releases CREB phosphorylation from its dependence on neuronal activity related to motor training and that the uncoupling is most prominent in layer 5 neurons.

It is possible that the greater Ca²⁺ activity of cortical neurons in resting PS1 mutant mice is involved in the deregulation of CREB phosphorylation seen in the mutant. Neuronal activity regulates Ca²⁺ transients that in turn regulate CaMKII, which activates CREB by phosphorylating CREB serine 133. The ability of basal neuron firing in resting mutant PS1 mice to increase CREB phosphorylation could result from more efficient activation of CaMKII by dendritic Ca²⁺ activity. This would be the case if, as previously suggested[80], interaction of mutant PS1 with the IP3R enhances the CICR that is necessary to communicate neuron activity down the length of dendrites to the nucleus. Ca²⁺ also regulates the Ca²⁺-dependent phosphatase, calcineurin, which represses CREB by S133 dephosphorylation[87]. Thus, the relative levels of activation of CaMKII versus activation of calcineurin determine CREB phosphorylation. The activation of CaMKII versus calcineurin is determined by the frequency, magnitude and duration of Ca²⁺ transients[88], with high frequency Ca²⁺ spikes favoring CaMKII activation and longer duration spikes of lower amplitude favoring calcineurin activity. The enhanced Ca²⁺ activity in resting M146V mutant animals that we observed likely underlies the constitutive elevation of CREB phosphorylation in these mice.

The role of CREB in memory formation is well established, and memory formation is strongly impaired by CREB knockout in mice[89]. The constitutive and reduced phosphorylation of CREB during resting and running conditions in mutant mice suggest that in the mutant, CREB functional dependence on neuron activity is reduced or eliminated, particularly in layer 5 neurons. As layer 5 pyramidal neurons are the major output cells in the motor cortex, the loss of activity dependent CREB phosphorylation in response to motor activity would disrupt the expression of genes that are critical for synaptic structural plasticity and long-term memory formation as observed in this study. Future studies are needed to establish mechanistic links between PS1 mutations and reduced CREB phosphorylation and to test whether correcting abnormal Ca²⁺ signaling and CREB phosphorylation would alleviate long-term memory deficits caused by PS1 mutations.

It is important to mention that we observed alterations in CREB phosphorylation in PS1 mutant mice at 5-8 months of age. Previous studies have shown that the constitutive hyperphosphorylation of CREB in cells expressing the PS FAD mutations in different cell lines, as well as in brain lysates from transgenic mice at 4- to 6-weeks of age [90]. Because abnormal Ca²⁺ release from ER stores is the major mechanism for constitutive CREB in FAD PS-expressing cells and 4- to 6-eek-old transgenic mice [90], we expect that the same mechanism also underlies altered CREB phosphorylation in PS1 mutant mice at 5-8 months of age, and possibly at 3-4 months of age when behavioral deficits and abnormal Ca²⁺ increase in neurons were observed. Nevertheless, the relationship among abnormal increase in Ca²⁺ transients, alterations in CREB phosphorylation, and behavioral deficits likely undergo age-dependent changes. Future studies are needed to investigate this issue to better understand how PS1 mutations contribute to age-dependent memory decline and AD pathology.

Supplementary Material

1

Supplemental Figure 1. Abnormal Ca²⁺ activity of layer 2/3 and layer 5 pyramidal neurons occur in PS1 mice during both resting and running

(A-B) Total calcium activity of layer 2/3 dendrites in WT and PS1 mice during resting (A) and running (B). Resting (WT: 81 dendrites from 6 mice; PS1: 101 dendrites from 5 mice). Running (WT: 61 dendrites from 6 mice; PS1: 51 dendrites from 5 mice). (C-D) Total calcium activity of layer 2/3 somas in WT and PS1 mice during resting (C) and running (D). Resting (WT: 92 somas from 6 mice; PS1: 85 somas from 5 mice). Running (WT: 93 somas from 6 mice; PS1: 83 somas from 5 mice). (E-F) Total calcium activity of layer 5 dendrites in WT and PS1 mice during resting (E) and running (F). Resting (WT: 44 dendrites from 5 mice; PS1: 69 dendrites from 4 mice). Running (WT: 44 dendrites from 5 mice; PS1: 53 dendrites from 4 mice). (G-H) Total calcium activity of layer 5 somas in WT and PS1 mice during resting (G) and running (H). Resting (WT: 49 somas from 4 mice; PS1: 307 somas from 5 mice). Running (WT: 52 somas from 4 mice; PS1: 242 somas from 5 mice). Mann-Whitney unpaired nonparametric test, *P < 0.05, **P < 0.01, ***P < 0.001. n.s., not significant

Highlights:

• Long-term motor memory impaired in PS1 mutant Familial Alzheimer’s disease model.

• Ca²⁺ activity is altered in a cortical layer- and neuron type-specific manner.

• PS1 M146V mutation alters activity-dependent CREB phosphorylation

• Altered CREB phosphorylation could underly the long term memory deficits.