Early Activation of Experience-Independent Dendritic Spine Turnover in a Mouse Model of Alzheimer's Disease

Abstract

Synaptic loss is critical in Alzheimer's disease (AD), but the dynamics of synapse turnover are poorly defined. We imaged dendritic spines in transgenic APPswe/PSen1∆E9 (APP/PS1) cerebral cortex. Dendritic spine turnover is increased far from plaque in aged APP/PS1 mice, and in young APP/PS1 mice prior to plaque formation. Dysregulation occurs in the presence of soluble Aβ oligomer and requires cellular prion protein (PrP^C). APP/PS1 mice lack responsiveness of spine turnover to sensory stimulation. Critically, enhanced spine turnover is coupled with the loss of persistent spines starting early and continuing with age. To evaluate mechanisms of experience-independent supranormal spine turnover, we analyzed the transcriptome of young APP/PS1 mouse brain when turnover is altered but synapse density and memory are normal, and plaque and inflammation are absent. Early PrP^C-dependent expression changes occur in synaptic and lipid-metabolizing genes. Thus, pathologic synaptic dysregulation underlying AD begins at a young age prior to Aβ plaque.

Introduction

Alzheimer's disease (AD) is a common fatal neurodegenerative disease, which entails progressive loss of cognitive function, especially memory. In post-mortem analysis, AD is characterized by the accumulation of extracellular amyloid-beta-(Aβ)-containing plaques and intracellular hyperphosphorylated-tau-containing neurofibrillary tangles (Musiek and Holtzman 2015). Microglial reaction to Aβ and tau pathology contributes in a multifactorial manner to disease progression (Musiek and Holtzman 2015). Throughout the course of disease, progressive loss of synapses correlates strongly with symptomology and memory loss (Scheff et al. 1990; Terry et al. 1991; Masliah et al. 1992; Koleske 2013; Spires-Jones and Hyman 2014). Both the autosomal dominant genetics of early onset cases and biomarker studies of sporadic late-onset disease demonstrate that β-amyloid serves as the initial trigger for disease (Musiek and Holtzman 2015). Among Aβ forms, oligomeric species (Aβo) have direct effects on synapses, disrupting functional plasticity and inducing anatomical loss (Shankar et al. 2008; Koleske 2013).

Dendritic spines form the post-synaptic specialization of excitatory synapses, and their rearrangements, gains and losses have the potential to reorganize network connectivity and modulate neurological function (Grutzendler et al. 2002; Trachtenberg et al. 2002; Holtmaat and Svoboda 2009; Koleske 2013). While overall spine density can be captured by Golgi impregnation or by immunohistology, tracking spine dynamics by chronic in vivo multiphoton confocal imaging of transgenic fluorescent mice is essential to assess evidence for dynamic alteration of connectivity patterns within brain networks. Dendritic spine gains and losses can occur within minutes, but the vast majority of dendritic spines persisting for days form synapses and can remain stable for months in the adult brain (Holtmaat et al. 2006; Holtmaat and Svoboda 2009; Koleske 2013). Critically, spine dynamics in relevant brain regions are regulated by sensory experience and motor learning, such that focal anatomical plasticity contributes to functional adaptation and memory (Trachtenberg et al. 2002; Zuo et al. 2005; Holtmaat et al. 2006; Hofer et al. 2009; Xu et al. 2009; Yang et al. 2009; Lai et al. 2012; Koleske 2013; Condello et al. 2015).

Imaging of transgenic AD mice in vivo has focused on the formation of Aβ plaques, disruption of dendritic arbors near Aβ plaques or loss of dendritic spine density near Aβ plaques (Spires et al. 2005; Grutzendler et al. 2007; Spires-Jones et al. 2007; Meyer-Luehmann et al. 2008; Bittner et al. 2012). These studies have found significant dendritic spine loss and formation of dystrophic neurites within 50 to 100 µm of Aβ plaques. While there are numerous studies describing dendritic spine changes immediately adjacent to plaques in vivo, only a few studies have examined dendritic spine dynamics, e.g. gains and losses over time, in the 90–95% of brain that is not adjacent to plaque (Tsai et al. 2004; Liebscher et al. 2014; Zou et al. 2015). Such analysis is critical in order to understand the cellular basis of the widespread synaptic disruption and memory loss in AD. Furthermore, previous studies have typically examined dendritic spines in mice with developed disease evidenced by memory impairment and synapse loss, and have not assessed dynamics prior to symptoms.

To extend mechanistic understanding, we imaged dendritic spine dynamics in segments of older transgenic mice far from Aβ plaques, and in young presymptomatic AD mice without plaque. We observe that both young and old APPswe/PS1∆E9 mice have substantially enhanced spine turnover in cortical parenchyma without plaque. This aberrantly enhanced dynamic is independent of sensory experience and is coupled with loss of persistent spines. Cellular prion protein (PrP^C) is a high-affinity binding site for Aβo, which is required for Aβo-induced dysfunction in LTP, memory and learning, and dendritic spine loss (Lauren et al. 2009; Gimbel et al. 2010; Um et al. 2012, 2013). Our previous work identified PrP^C as being responsible for mediating dendritic spine loss in live cell imaging of dissociated hippocampal cultures (Um et al. 2012, 2013). Here, we show that the APP/PS1 dendritic spine turnover phenotype is rescued by loss of PrP^C function. Transcriptomic analysis demonstrates that synaptic and lipid metabolism contributes to the first synaptic phenotypes in transgenic mice.

Materials and Methods

Animals

All procedures for in vivo two photon microscopy and surgery were conducted in accordance with protocols reviewed and approved by the Yale Animal Care and Use Committee. APPswe/PSen1∆E9 transgenic mice (Jankowsky et al. 2004; Gimbel et al. 2010) and Prnp-/- (Edinburgh strain) were crossed with Thy1-EGFP-M line mice (Feng et al. 2000). Multiple crosses led to four genotypes of mice : WT Thy1-EGFP-M, APPswe/PSen1∆E9 Thy1-EGFP-M, Prnp-/- Thy1-EGFP-M and APPswe/PSen1∆E9 Prnp-/- Thy1-EGFP-M. Littermate mice on the C57Bl6J background were utilized for these studies. Mice were imaged between 3 and 4 months of age and 10 and 11 months of age (Supplementary Table 1).

Tissue Processing for Analysis of Static Dendritic Spine Density and Amyloid Plaques

Mice were perfused transcardially with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) and the whole brain was removed. The whole brain was immersion fixed in 4% PFA in PBS for ~24 h at 4°C. Coronal sections were cut at 40 µm intervals on a vibratome. Sections were mounted on SuperFrost Plus slides, air dried, and subsequently transferred into thioflavin staining. Slides were submerged in 0.1% Thioflavin S (SIGMA T1892) in 70% ethanol for 15 min, then washed for 70% ethanol for one minute two times and with ultrapure water for one minute two times. Slides were air-dried then mounted using vectashield.

Imaging of Static Dendritic Spine Density and Amyloid Plaques

Images of the hippocampus and barrel cortex were acquired using a Carl Zeiss LSM 710 microscope with a 63X oil objective lens and a 1.5X zoom on a GFP 488 nm laser.

Analysis of Static Dendritic Spine Density

EGFP-labeled pyramidal neurons located in the dorsal region of the CA1 of the hippocampus were analyzed. Basal dendrites projecting toward corpus callosum in the proximal region (30–120 µm from soma) were imaged. Accordingly, apical dendrites projecting toward dentate gyrus in the proximal region (30–120 µm from soma) were imaged. The apical dendritic tuft of EGFP pyramidal neurons projecting from layer II/III of the barrel cortex were imaged in layer I of the barrel cortex. In each section, spines on secondary dendrites were counted manually over a dendritic length of >100 µm per image using ImageJ software. Only regions more than 100 µm from any Aβ plaque visualized by thioflavin staining were included in the analysis. Cell density values were obtained by dividing the number of dendritic spines by the corresponding dendritic length. All analyses were performed by a researcher blind to the genotype of the animal.

Analysis of Amyloid Plaque Density in Fixed Tissue

The density of amyloid plaque in the hippocampus and barrel cortex was calculated using ImageJ software. A macro using thresholding overlay for Thioflavin-S-labeled amyloid beta plaques was applied across all images and the overall plaque per region was calculated accordingly.

c-Fos Staining

c-fos expression was analyzed for two age groups of mice, 2–4 months and 11–14 months, housed in standard and enriched environments. Enriched environment mice were housed in an enriched environment for one hour, during which they were able to freely navigate and explore the cage with unimpeded access to food and water. After one hour, the mice were sacrificed for immunohistochemistry. Controls were housed in a standard environment. They were sacrificed for immunohistochemistry and processed alongside the enriched environment mice. All animals were perfused with PBS and 4% PFA. Brains were extracted and immersed in 4% PFA for 3–5 days before long-term storage in 0.01% sodium azide in phosphate-buffered saline. Coronal slices of brain were made at 40 μm steps using a vibratome (VT1000; Lecia Microsystems). The free-floating sections were stained with the primary antibody c-Fos (Santa Cruz, 1:250). Sections were subsequently stained with the secondary antibody biotinylated anti-rabbit IgG made in goat serum (1:250). The chromogenic reporter diaminobenzidine (DAB) was then used to further enhance protein detection. A light microscope (Zeiss) was used to detect stained nuclei in all samples. Images were taken of the barrel cortex at 10X magnifications. The density of c-Fos-labeled neurons (the count per defined area) was quantitatively determined using a contrast mask macro created in the software ImageJ. The density of c-Fos labeling was averaged across two tissue slices to obtain the final value for each sample.

Thin Skull Surgery for Imaging

Thin skull imaging was performed on mice between 3–4 months and 10–11 months of age. The mice were anesthetized using ketamine (100 mg/kg)/ xylazine (10 mg/kg) as previously described (Akbik et al. 2013). The coordinates used to locate the barrel cortex were 3-mm lateral and 1-mm caudal to bregma. The skull overlaying the right barrel cortex was thinned to a thickness of ~200 µm using a pneumatic microdrill (Henry Schein). The skull was then thinned to a thickness of ~20 µm using a fine ophthalmic blade (Surgistar) leaving the skull intact.

Labeling of Aβ Plaque In Vivo

Each of the 10-month-old APP/PS1 mice received a (E,E)-1-fluoro-2,5-bis(3-hydroxycarbonyl-4-hydroxy)styrylbenzene (FSB) tail vein injection intravenously (i.v.)) 1–2 days before the experiment (Sato et al. 2004; Condello et al. 2011, 2015). A 2.4 mM stock of FSB dissolved in DMSO and bacteriostatic saline solution was made and injected intravenously (5 µl/gram of body weight). This provided Aβ plaque labeling for greater than 4 weeks, and regions greater than 100 µm from plaque were selected for dendritic spine imaging.

Imaging Procedure

Transcranial imaging began immediately following thinning. Images were acquired using a 2PLSM. A Ti:Sapphire laser running at 910 nm wavelength powered by a Coherent Chameleon Ultra II laser was used for imaging. The objective used was a water immersion lens (40X aperture 1.0NA) (Zeiss). A digital zoom of 3X was used to attain 120X images of dendritic spines. A 10X water immersion lens (Zeiss) was used to create a vasculature map, which allowed relocation of the same region throughout imaging. Image acquisition was done using Zen software (Zeiss). The dendritic apical tuft of neurons projecting from layer II/III of the barrel cortex was imaged at three time points: the day of initial surgery (day 0), 2 days following initial surgery (day 2), and 14 days following the initial surgery (day 14). Imaging was done to a maximum depth of 100 µm from the cortical surface with 1 µm z-steps used. A summary of all mice imaged is provided in Supplemental Table 1.

Sensory Enrichment and Deprivation for Imaging Experiments

In enriched groups, the animals were placed in an enriched environment immediately following the first imaging session. Enriched environments were created by hanging round faceted beads on wire from a standard wire cage top. This method was modified from previous studies (Yang et al. 2009; Akbik et al. 2013) and eliminated the ability of the mice to destroy the enriched environment and ensured consistent enrichment over the imaging period. Mice were allowed to freely explore the sensory-enriched environment with normal access to food and water.

In the sensory deprived groups, the animals had whiskers plucked bilaterally under light isoflurane anesthesia 2 days prior to the first imaging session. Whiskers were removed every 2 to 3 days as needed. The mice were placed in the enriched environment following the first imaging session and were housed in this environment for the remainder of the deprivation experiment.

Image Analysis

Image analysis was done using Image J. Each dendrite was evaluated in three dimensions with no knowledge of genotype. Dendritic spines were classified as protrusions from the dendritic shaft of at least 0.5 μm with an intensity of at least five times the standard deviation of the background fluorescence (Knott et al. 2009). Considering limitations in spine resolution in the Z-direction, only spines which project laterally from the dendrite were analyzed. Additionally, any filipodia-like projections which were greater than 2.5 µm in length and had no defined spine head were not counted as spines (Kwon and Sabatini 2011). Dendritic spine formation and loss were calculated for each animal. The dendritic spine turnover ratio was calculated as (N_gained + N_loss)/(N_Total + N_gained), where N_gained is the number of spines gained at indicated time point, N_loss is the number of spines lost at indicated time point, and N_Total is total spines at indicated time point. Only dendritic segments more than 100 µm from an Aβ plaque stained with FSB were included in the analysis. In Supplemental studies, images from 10-month mice housed in a standard environment were reanalyzed requiring a 1 µm rather than the standard 0.5 µm protrusion from the dendritic shaft (Knott et al. 2009; Swanger et al. 2011).

PLISA

Brain tissue was homogenized in TBS and centrifuged at 100 000 × g for one hour at 4°C. The supernatant containing soluble Aβ was analyzed by PLISA as described (Um et al. 2012; Kostylev et al. 2015). This assay captures Aβo with immobilized PrP^C and detects bound species with anti-Aβ antibody.

RNAseq

RNA samples from the forebrain of 4 mice of each of the 4 genotypes at 4 months age and each sample was analyzed independently (n = 16) at the Yale Center for Genome Analysis. Total RNA was isolated using TRIzol and purified by RNeasy (Qiagen, Valencia, CA). The RNAseq library was prepared using mRNA Seq Kit supplied (Illumina). Each RNAseq library was layered on one of the eight lanes of the Illumina flow cell at appropriate concentration and bridge amplified to obtain raw reads. The quality of the sequence data was assessed using FastQC, after trimming adaptor sequences. Reads that passed the quality control were aligned to reference genome using Tophap2. The reads per sample averaged greater than 20 000 000 with 92–94% of reads mapped to the genome and multiply mapped reads averaging <8%. The Count number (Number of reads mapped to each transcripts) was obtained using htseq-count. Only the transcripts with ≥1 read in at least eight samples were kept for further analysis. Out of the total of 39 179 genes, 19 394 genes were kept in this step. Count numbers were normalized using R package RUVseq. Read counts are further normalized using trimmed Mean of M-values method in R package edgeR. Differential expression was tested from normalized count data using a generalized linear model. P values and Log2(fold-change) were calculated using the normalized count data for each gene. Those genes with False Discovery Rate P value < 0.05 were identified as differentially expressed genes. For the APP/PS1 versus WT comparison, 15 genes were up regulated in the transgenics and 17 were down regulated.

Mouse Brain Tissue Collection for Biochemistry

Mice used for biochemical analysis were 4 months old on average at the time of sacrifice. n = 6 WT, Prnp-/-, and APP/PS1 Prnp-/- mice, as well as n = 7 APP/PS1 mice were used. Mice were euthanized by rapid decapitation, and the dissected cerebral cortices were flash-frozen in dry ice. Cortices were thawed and homogenized in three times the brain tissue weight in radio immunoprecipitation assay (RIPA) lysis buffer (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1% Triton X-100; 1 mM EDTA; 0.1% SDS; 0.5% deoxycholic acid) containing 1X cOmplete-mini protease inhibitor cocktail (Roche). Lysates were centrifuged for 30 minutes at 100 000 × g and 4°C. The supernatants were collected as RIPA-soluble fractions.

Immunoblots

The protein concentration in RIPA-soluble fractions was measured by Bradford assay (Bio-Rad Protein Assay) prior to analysis by SDS-PAGE. Proteins were electrophoresed through precast 4–20% tris-glycine gels (Bio-Rad) and transferred with an iBlot^™ Gel Transfer Device (Novex-Life Technologies) onto nitrocellulose membranes (Invitrogen). Membranes were blocked (Blocking Buffer for Fluorescent Western Blotting, Rockland MB-070-010) for 1 h at room temperature prior to incubation overnight in primary antibodies at 4°C. The following primary antibodies were used: Goat anti-LDLR (Santa Cruz sc-11824; 1:200), mouse anti-β-Amyloid (Covance SIG-39300; clone 6E10; 1:500), and rabbit anti-β-Actin Antibody (Cell Signaling 4967; 1:2 500). Actin was used as loading control and run on the same gel. Odyssey donkey anti-goat, donkey anti-mouse, and donkey anti-rabbit conjugated to IRDye 680 or IRDye 800 (LI-COR Biosciences) secondary antibodies were applied for 1 h at room temperature. Proteins were visualized with a LI-COR Odyssey infrared imaging system, and quantification of band intensities was performed within a linear range of exposure.

Statistics

For all in vivo imaging, data was analyzed using one way ANOVAs with pairwise comparison to WT control. The “n” value is the number of mice in each group. For static spine imaging, statistical comparisons included one-way ANOVA and repeated-measures ANOVA with post hoc Tukey pairwise comparisons. The “n” value is the number of separate dendrites in each group. For c-Fos analysis, data was analyzed using one-sided t-test. The “n” value is the number of mice in each group. All results are reported as mean ± SEM.

Results

Loss of transgene-tagged dendritic spines in hippocampus and barrel cortex of AD mice

Previous studies have shown that there is decreased dendritic spine density in certain Alzheimer's disease (AD) mouse strains (Perez-Cruz et al. 2011). To verify that decreased dendritic spine density occurs in 10-month-old APPswe/PSen1∆E9 mice (APP/PS1) (Jankowsky et al. 2004) in the subset of neurons tagged by the Thy1-EGFPM transgene (Feng et al. 2000), we analyzed the static dendritic spine density in the barrel cortex and hippocampus. We perfused mice at 10 months of age and sectioned tissue to visualize spines on the apical dendritic tuft of EGFP-labeled pyramidal neurons projecting from layer II/III of the barrel cortex and pyramidal neurons in the CA1 region of the hippocampus (Fig. 1A,B,E). Importantly, double staining of Aβ plaque with thioflavin S confirmed that >90% of imaged dendrites were in zones more than 100 µm distant from any plaque (data not shown). We analyzed only the dendritic population more than 100 µm from plaque. Compared to controls, transgenic APP/PS1 mice possess significantly lower dendritic spine density in the apical and basal pyramidal neurons of the hippocampal CA1 region in zones not immediately adjacent to plaque. To determine if synapse loss in vivo is dependent on PrP^C, we crossed APP/PS1 mice with Prnp-/- mice (Manson et al. 1994). The APP/PS1-induced decrease in dendritic spine density is fully corrected in APP/PS1 mice lacking PrP^C (Fig. 1C,D). This pattern is not confined to the hippocampus. PrP^C-dependent decreased dendritic spine density in APP/PS1 mice is also seen in the apical tuft of pyramidal neurons projecting from layer II/III of the barrel cortex (Fig. 1F). These results validate previous findings of decreased dendritic spine loss in transgenic AD strains (Knafo et al. 2009; Perez-Cruz et al. 2011) and a role for PrP^C in mediating these changes (Manson et al. 1994; Chung et al. 2010; Gimbel et al. 2010; Um et al. 2012). Importantly, the use of transgenic EGFP-M mice allows an extension to chronic in vivo imaging.

Figure 1.

Dendritic spine density in the hippocampus and barrel cortex of APP/PS1 mice is decreased by a PrP^C-dependent mechanism on prion protein. (A) By imaging Thy1-EGFPM fixed tissue with an LSM, the apical and basal dendrites of the hippocampal CA1 region were visualized. (B) Each image is a maximum intensity projection of serial sections of EGFPM-labeled pyramidal neurons, and all spines counted were > 100 µm from thioflavin-positive plaque. (C) Dendritic spine density in the apical dendrites of 10-month-old WT, APP/PS1, prnp-/-, and APP/PS1 prnp-/- mice. (D) Dendritic spine density in the basal dendrites of 10-month-old WT, APP/PS1, prnp-/-, and APP/PS1 prnp-/- mice. (E) Imaging of fixed tissue in layer I of the barrel cortex shows decreased spine density. Each image is a maximum intensity projection of serial z-steps. (F) Spine densities of apical dendritic tufts projected to layer I of the barrel cortex. All spines counted were > 100 µm from Aß plaque. N reflects separate dendritic segments from 3 to 6 mice per group: WT (n = 30), APP/PS1 (n = 18), prnp-/- (n = 30), and APP/PS1 prnp-/- (n = 18).*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001 (one-way ANOVA with Tukey's post hoc test).

Dendritic Spine Turnover is Elevated in the Barrel Cortex of Aged AD Mice

Having validated the detection of decreased dendritic spine density by Thy1-EGFPM in 10-month-old transgenic APP/PS1 mice in zones without plaque, we evaluated spine dynamics to determine whether net loss occurs via increased rate of loss, decreased rate of gain, or some more complex combination. We used chronic in vivo two-photon laser scanning microscopy (2PLSM) to image the S1 barrel cortex of anesthetized Thy1-EGFPM transgenic mice through a thinned skull. The thinned skull method was chosen to avoid altering any transgene-related inflammatory reaction (Xu et al. 2007; Akbik et al. 2013). The fluorescently labeled apical dendritic tufts of neurons projecting from cortical layers II/III were analyzed (Fig. 2A). Mice were injected with (E,E)-1-fluoro-2,5-bis(3-hydroxycarbonyl-4-hydroxy)styrylbenzene (FSB) prior to spine imaging to visualize Aβ throughout the experiment (Sato et al. 2004; Condello et al. 2011, 2015). This provided confirmation that the vast majority of imaged spines were present in zones farther than 100 µm distance from Aβ plaque, and only spines meeting this criteria were included in the analysis. During a 14-day (14D) imaging window, the turnover of dendritic spines in 10-month-old APP/PS1 mice is twice as great as in WT mice (Fig. 2B; P < 0.01). Because static spine loss in APP/PS1 mice is dependent on PrP^C expression, we assessed its role in the dynamic changes of spine turnover. The 14D turnover rate in Prnp-/- and APP/PS1 Prnp-/- is indistinguishable from WT mice, demonstrating an essential role for PrP^C in this phenotype. Our standard cutoff for counting dendritic spines required 0.5 µm protrusion from the dendritic shaft (Knott et al. 2009). Using a more stringent cutoff of 1 µm (Swanger et al. 2011) to capture the turnover of larger spines yielded slightly lower turnover rates but the identical genotype pattern (Suppl. Fig. 1). Thus, the net synapse loss in aged APP/PS1 mice does not reflect a simple increased rate of loss from a fixed population of spines, and instead suggests a more complex altered dynamic.

Figure 2.

10-month-old transgenic AD model mice show a PrP^C-dependent increase in dendritic spine turnover in S1 cortex. (A) Representative images of chronic transcranial 2PLSM of dendritic apical tufts of neurons projecting from layer II/III of the barrel cortex. Red and green arrows indicate losses and gains, respectively. (B) Quantification of the percent turnover rate of dendritic spines over 14 days. All spines counted were > 100 µm from FSB-positive plaque. N reflects separate mice: WT (n = 7), APP/PS1 (n = 6), prnp-/- (n = 6), and APP/PS1 prnp-/- (n = 6). **P ≤ 0.01 (one-way ANOVA with comparison to WT control and Dunnetts post hoc test).

Elevated Spine Turnover in Aged AD Mice is Unresponsive to Sensory Enrichment

Previous studies have demonstrated a marked increase in synaptic remodeling for mice placed in a novel sensory-enriched environment or during motor task learning (Holtmaat et al. 2006; Hofer et al. 2009; Yang et al. 2009; Akbik et al. 2013). Specifically, increased whisker stimulation of mice housed in an environment with multiple hanging beads doubles dendritic spine turnover in the apical tuft of S1 cortical neurons (Yang et al. 2009; Akbik et al. 2013). Of note, enriched environments have been reported to lessen deficits in AD transgenic mice (Lazarov et al. 2005). Considering that aged APP/PS1 mice have elevated turnover rates in a standard cage environment (SE) similar to WT mice in a sensory enrichment environment (EE), we sought to determine whether or not the APP/PS1 mice are capable of further increased synaptic remodeling in response to an EE. We placed mice into an EE immediately following the first imaging session (Fig. 3A,B). Over 14 days of EE, the turnover rate significantly increases (P < 0.01) for WT mice, but there is no detectable change for spines more than 100 µm from Aβ plaque in the APP/PS1 mice (Fig. 3C). Deletion of Prnp does not alter the response to EE on the WT background, but normalizes EE responses on the APP/PS1 background (Fig. 3C).

Figure 3.

Aged APP/PS1 mice do not respond to a sensory-enriched environment. Mice are placed in a sensory-enriched environment immediately following first imaging time point. (A) Representative images of chronic 14-day transcranial 2PLSM of dendritic apical tufts of neurons projecting from layer II/III of the cortex for sensory-enriched environment mice. Red and green arrows indicate losses and gains, respectively. (B) Percent turnover of mice housed in sensory-enriched environment. Quantification of the percent turnover rate of dendritic spines over 14 days, > 100 µm from plaque. Sensory enriched mice: WT (n = 6), APP/PS1 (n = 5), prnp-/- (n = 7), APP/PS1 prnp-/- (n = 5). (C) Comparison of Standard (SE) versus Enriched (EE) environment turnover rates by genotype. Gray bars are standard environment data re-graphed from Figure 2B for comparison. Red bars are turnover rates in sensory-enriched environment from panel B. (D) The percentage of gains over 14 days in WT and APP/PS1 mice housed in an SE or EE. (E) The percentage of losses over 14 days in WT and APP/PS1 mice housed in an SE or EE. (F) The net losses (Losses-Gains) over 14 days in WT and APP/PS1 mice housed in either an SE or EE. N reflects separate mice: for panels D ,E, F: SE (WT n = 7 and APP/PS1 n = 6) and EE (WT n = 6 and APP/PS1 n = 6). *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. (Panel B, D, E, and F: One-way ANOVA with comparison to WT control and Dunnetts post hoc test) (Panel C: one tailed t-test).

The 14D EE exposure increases both the spine gain and spine loss rate in WT mice relative to SE and mediates balanced rearrangement of circuits (Fig. 3D–F) (Yang et al. 2009; Akbik et al. 2013). In SE caging, APP/PS1 spine losses exceed WT levels (Fig. 3E, P < 0.01) and the losses exceed the gains in the APP/PS1 mice (Fig. 3F, P < 0.01), providing a basis for chronic net loss of synapses. Exposure of 10-month-old APP/PS1 to EE shows non-significant trends to increase gain rate (Fig. 3D) and reduce loss rate (Fig. 3E), such that the net loss returns to zero (Fig. 3F). Thus, while overall dendritic spine turnover rate in APP/PS1 mice does not increase in EE housing (Fig. 3C), there is a shift to normalize the elevated net loss rate (Fig. 3F). These results indicate that APP/PS1 mice have a deficit in baseline turnover while housed in an SE as well as abnormal responses to EE.

Plaque and Aβ oligomer levels as a function of age

The earliest changes in dendritic spine dynamics have the potential to be the most informative regarding mechanism. The data above were collected at 10 months, an age at which we confirmed the presence of Aβ plaque in the hippocampus and barrel cortex (Suppl. Fig. 2A). In this strain, the earliest reliable detection of Aβ plaque occurs at 5–6 months, and we detected no Aβ plaque in 4-month-old APP/PS1 brain sections (Suppl. Fig. 2A). Since Aβo are hypothesized to be the synaptotoxic species leading to dendritic spine loss, we considered whether they might be present at this earlier stage. Using an assay which detects PrP^C-interacting high-molecular-weight Aβo (Um et al. 2012; Kostylev et al. 2015), we detected low levels of transgene-specific Aβo in APP/PS1 mice as early as 1.5 months with an increase by 4 months of age to 0.2 ng/g of cortex (Suppl. Fig. 2B). As previously reported, Aβ levels are not affected by the expression of PrP^C (Gimbel et al. 2010). Thus, at 4 months of age in APP/PS1 mice, Aβo are present but Aβ plaques are not.

Dendritic spine dynamics disrupted in young APP/PS1 mice without plaques

The 4-month-old APP/PS1 mice provide an opportunity to assess whether dendritic spine dynamics are regulated by low levels of Aβo prior to Aβ plaque deposition with attendant neuroinflammation. Any changes would reflect silent decompensation prior to manifestation of memory deficits or synapse loss (Park et al. 2006; Gimbel et al. 2010). In order to assess the effect of age on dendritic spine dynamics, 3.5-month-old APP/PS1 mice with or without PrP^C were imaged over 14 days in either an SE or an EE (Fig. 4A). Surprisingly, young APP/PS1 mice in SE have a significant increase in baseline dendritic spine turnover as compared to WT controls (Fig. 4B, p < 0.001). The high turnover phenotype is rescued in APP/PS1 mice lacking PrP^C. When placed in an EE, turnover rates increase in every group except for the APP/PS1 mice, with the result that the four genotypes become indistinguishable (Fig. 4C). As for the 10-month mice, the turnover rate significantly increases in EE (p < 0.01) for WT mice, but there is no detectable EE-induced change for spine turnover in the APP/PS1 mice more than 100 µm from Aβ plaque (Fig. 4D). Separate evaluation of dendritic spine gains and losses reveals that the increased turnover seen in young APP/PS1 mice is due to a significant increase in both gains and losses in the SE (Fig. 4E,F). There is a non-significant trend toward greater losses than gains in SE APP/PS1 mice (Fig. 4G) that reaches significance at later ages (Fig. 3F) to generate net synapse loss. The gains and losses in APP/PS1 mice placed in the EE were not significantly different from APP/PS1 mice in an SE, while WT mice show significant increases in both measures (Fig. 4E,F). Thus, the majority of the deranged dendritic spine dynamics observed in 10-month-old APP/PS1 mice are already present in young, behaviorally normal, plaque-free APP/PS1 mice. We also observed that spine turnover is normal in young APP/PS1 mice lacking PrP^C, reflecting a necessity of this binding site for dendritic spine phenotypes in both 3.5- and 10-month-old mice.

Figure 4.

Dendritic spine dynamics are disrupted in 3-month-old transgenic AD mice. (A) Representative images of 14-day chronic transcranial 2PLSM of dendritic apical tufts of neurons projecting from layer II/III of the cortex. Red and green arrows indicate losses and gains, respectively. (B) Percent turnover of spines in 3-month-old mice housed in standard environment (SE). WT (n = 6), APP/PS1 (n = 6), prnp-/- (n = 6), APP/PS1 prnp-/- (n = 6). (C) Percent spine turnover for mice housed in an Enriched Environment (EE). EE mice: WT (n = 6), APP/PS1 (n = 6), prnp-/- (n = 6), APP/PS1 prnp-/- (n = 6). (D) Comparison of SE versus EE turnover rates by genotype. SE data re-graphed from panel B. EE data re-graphed from panel C. (E) The percentage of gains over 14 days in WT (n = 6) and APP/PS1 (n = 6) mice housed in an SE or EE. (F) The percentage of losses over 14 days in WT (n = 6) and APP/PS1 (n = 6) mice housed in an SE or EE. (G) The net losses (Losses-Gains) over 14 days in WT (n = 6) and APP/PS1 (n = 6) mice housed in either an SE or EE. N reflects separate mice. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001 (Panel B-C, E-G: one-way ANOVA with comparison to WT control and Dunnetts post hoc test) (Panel D: one tailed t-test).

Whisker to S1 Cortex c-fos Activation Occurs normally in APP/PS1 mice

The abnormal dendritic spine turnover of APP/PS1 mice may reflect cortical pathology with derangements of activity-dependent regulation of anatomical plasticity and synapse stability, or may be secondary to an unexpected alteration in somatosensory processing. c-fos is an immediate early gene, which marks elevated neuronal activity in vivo. Whisker stimulation is known to induce the c-fos activation in the barrel cortex (Filipkowski et al. 2000). In order to assess whether a deficit in whisking-induced neuronal activation is present in APP/PS1 mice, we compared c-Fos protein level in mice housed in SE or EE at 3.5 months and 10 months of age (Suppl. Fig. 3). There is a significant increase in c-fos activation in all four genotypes at both 3.5 months and 10 months of age (Suppl. Fig. 3B, 3D). Thus, neurons of barrel cortex are activated by the increased whisker stimulation in the EE, so that the deranged spine turnover likely reflects altered anatomical plasticity in the cerebral cortex.

Dendritic Spine Turnover rates remain elevated in sensory-deprived young AD mice

Dendritic spine turnover is elevated in APP/PS1 but is not increased further by EE, raising the issue of whether low level stimulation in SE drives maximal spine turnover. Alternatively, APP/PS1 turnover may be fully experience-independent. It is documented that experience dependent increases in dendritic spine turnover in the barrel cortex of WT mice is whisker dependent (Holtmaat et al. 2006). Thus, when the whiskers of WT mice are trimmed or removed, the mice have no detectable change in spine density or turnover while housed in an EE, and their turnover rates are comparable to those mice housed in an SE (Holtmaat et al. 2006; Akbik et al. 2013). We surgically removed whisker input to the S1 cortex in young APP/PS1 and WT mice. Whiskers were plucked 2 days prior to the first imaging session and were continually removed every 2 days until imaging was completed. The mice were placed in an EE immediately following the first imaging session and imaged again at day 14 (Fig. 5A,B). The sensory-deprived APP/PS1 mice exhibit unusually elevated turnover rates (17.2%), which are indistinguishable from those of whisker-intact APP/PS1 mice housed in an SE (23.4%) or an EE (22.5%) (Fig. 5C). The controls undergoing deprivation showed turnover rates (10.0%) similar to those in an SE (8.35%) but significantly different from WT mice housed in an EE (19.1%), confirming that deprivation of whisker stimulation decreases spine turnover in WT mice. The changes are not limited to a 14-day imaging window, because a very similar pattern is observed when 2-day turnover is analyzed (Suppl. Fig. 4). Thus, we conclude that young AD mice have an inherently high turnover rate not driven by sensory stimuli or deprivation.

Figure 5.

Dendritic spine turnover rates remain elevated after whisker removal in 3-month-old transgenic AD mice. (A) Schematic of imaging timeline. Whiskers are plucked 2 days prior to first imaging session and every 2 days thereafter. Mice are placed in a sensory-enriched environment immediately following first imaging time point (B) Representative images of chronic 14-day transcranial 2PLSM of dendritic apical tufts of neurons projecting from layer II/III of the cortex. Green arrows indicate gains. (C) Black bars show turnover rates in animals undergoing deprivation over 14 days of imaging. Gray bars (SE) and red bars (EE) are re-graphed from Figure. 4 for comparison. Deprived mice: WT (n = 6) APP/PSEN (n = 6). N reflects separate mice. **P ≤ 0.01 (one-way ANOVA with Dunnetts post hoc test).

Persistent Spine Survival is Decreased in Young and Aged Transgenic AD Mice

A persistent spine is defined as a dendritic spine that is present for more than 2 days, and the vast majority of persistent spines form a synapse by ultrastructural correlation studies (Holtmaat et al. 2006). In this sense, the loss of a persistent spine is considered to be the loss of a synapse. Considering the high turnover rate in 3.5-month-old and 10-month-old APP/PS1 mice, we evaluated the persistent spine loss present in such mice with and without PrP^C (Fig. 6A). There is a significant decrease in persistent spine survival for both 3.5-month-old and 10-month-old APP/PS1 mice, which is fully rescued by the loss of PrP^C (Fig. 6B,C). The 10-month measurements were collected from regions more than 100 µm distant from Aβ plaque. Thus, even for 3.5-month-old APP/PS1 mice in which net synapse density is not yet decreased, the persistence of existing synapses is decreased.

Figure 6.

Persistent spine survival is significantly decreased in both young and aged transgenic AD mice in a PrP^C-dependent manner. (A) Representative images of chronic imaging over 14-day transcranial 2PLSM of dendritic apical tufts of neurons projecting from layer II/III of the cortex. Yellow arrows indicate persistent spines and red arrows indicate persistent spines which are lost on day 14. (B) Percent persistent spine survival of 3.5-month-old mice housed in SE. WT (n = 6), APP/PS1 (n = 6), prnp-/- (n = 6), APP/PS1 prnp-/- (n = 6). (C) Percent persistent spine survival of 10-month-old mice housed in SE. All spines counted were >100 µm from Aß plaque. N reflects separate mice: WT (n = 7), APP/PS1 (n = 6), prnp-/- (n = 6), APP/PS1 prnp-/- (n = 6). **P ≤ 0.01, ****P ≤ 0.0001 (one-way ANOVA with comparison to WT control and Dunnetts post hoc test).

Transcriptome Analysis of Mouse forebrain from Young APP/PS1 Mice

The dendritic spine data provide clear evidence of altered dendritic spine dynamics in APP/PS1 mice at 3.5 months of age with increased turnover, lack of sensory responsiveness, and loss of persistent spines. Because this phenotype occurs prior to Aβ plaque accumulation and inflammation, we sought to define the basis of these early changes. Cortical samples were processed by RNAseq to profile the transcriptome at this age. Because the phenotypes require PrP^C expression, we included comparison with Prnp null state to assess specificity. The number of genes differentially expressed between APP/PS1 and WT by False Discovery Rate analysis was limited to 32 genes, 15 upregulated and 17 downregulated (Fig. 7A,B). It should be noted that the APP/PS1 transgenes are driven by the Prnp promotor and the transgenic transcripts include untranslated regions of the mouse Prnp mRNA. Thus, the signal for Prnp is elevated in both the APP/PS1 samples and the APP/PS1, Prnp-/- samples.

Figure 7.

RNA sequencing comparison of 4-month-old AD mice with and without prion protein indicates potential players in deranged spine turnover of young AD mice. (A and B) Heat map showing normalized transcript count per million (cpm) in 4-month-old mice. Each genotype has 4 replicates. Colors indicate the range of expression in each mouse. All genes with FDR P < 0.05 and more than 1 count per million are shown. Downregulated genes in A and upregulated in B. (C) Transcript cpm for Npas2. (D) Transcript cpm for Homer1. (E) Transcript cpm for Ldlr. (F) Transcript cpm for Stard4. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 (one-way ANOVA with post-hoc Tukey pairwise comparison). (G) LDLR protein expression in 4-month-old APP/PS1 brain compared to WT, with and without PrP^C. Representative immunoblots of mouse cortices. Genotypes are indicated above each lane, and each lane is from a separate mouse. (H) Densitometric analysis of the immunoblots from G analyzed by one-way ANOVA with Fisher's post-hoc comparison. Data are mean ± SEM, n = 6 WT, Prnp-/-, and APP/PS1 Prnp-/- mice, and n = 7 APP/PS1 mice. LDLR expression is significantly enhanced in APP/PS1 cortex compared to WT (*,P < 0.05). LDLR expression is not significantly different between WT and Prnp-/- or WT and APP/PS1 Prnp-/- cortex (ns, P > 0.05).

A number of the differentially expressed genes are known to influence neuronal anatomy and plasticity, including Homer1, Npas2, Dpysl3, Rims1, and Arpp21. Lipid metabolism regulating genes include LDLR (an ApoE receptor) and Stard4. The gene Cirb is of potential interest because it encodes a cold-induced RNA binding protein related to RBM3, which was recently shown to preserve synapses in Prion disease and AD transgenic mice (Peretti et al. 2015). The heatmaps show that the APP/PS1 samples are distinct from the other three genotypes, and separate statistical analysis of selected genes demonstrates statistically significant rescue of altered expression by loss of PrP^C expression (Fig. 7C–F). We verified changes in protein for LDLR by immunoblot (Fig. 7G,H). Pathway analysis of protein-protein network interactions links the products of many of these differentially expressed genes (Suppl. Fig. 5). Importantly, the network includes APP and Prnp as well as genes previously linked to synaptic dysfunction in AD models, namely Grm5, Fyn, and Ptk2B.

Our transcriptome study focused on pre-symptomatic stage in APP/PS1, but additional transgenic mouse RNAseq data are available from the AMP-AD project, covering CRND8 mice and later ages of APP/PS1 mice. We assessed the differentially expressed genes identified here across this extended analysis (Suppl. Fig. 6A–F). It is clear that early and age-persistent dysregulation of Homer1, Ldlr, and Stard4 occur in both transgenic models. Analysis of 4-month APP/PS1 here or 3-month CRND8 in the AMP-AD project does not detect inflammatory or lysosomal changes. For a number of genes in this category, including cathepsins, cytokine receptors, and acute phase reactants, there is no differential expression at young ages when dendritic spine dynamics are dysregulated, but as Aβ plaque and microgliosis occur at later ages, there is upregulation (Suppl. Fig. 6G,H).

Discussion

The major finding of this study is that AD transgenic mice have dysregulated dendritic spine dynamics as early as 3.5 months of age, when Aβ plaque, neuro-inflammation, memory deficits, and synapse loss are not yet present. To investigate the mechanisms of altered dendritic spine dynamics in vivo, we housed mice in either a standard or sensory-enriched environment. The greater spine turnover in APP/PS1 mice is minimally altered by sensory input. Importantly, persistent spine loss is increased in both young and aged transgenic mice, indicating that synapse stability is decreased even before there is net synapse loss. In addition, the loss of PrP^C rescues the dysregulated dendritic spine dynamics at both 3.5 months and 10 months of age. Changes in the expression of synapse and lipid metabolism genes are correlated with the spine turnover phenotype. These findings demonstrate that AD model mice have abnormally elevated dendritic spine gains and losses, which ultimately lead to loss of synapse density, and that this is rescued by deletion of PrP^C.

The loss of synapses and dendritic spines has been documented in a range of mouse AD models and human AD (Tsai et al. 2004; Spires et al. 2005; Bittner et al. 2009; Knafo et al. 2009; Tackenberg et al. 2009; Bittner et al. 2012; Hanson et al. 2014; Liebscher et al. 2014; Zou et al. 2015). A priori, the net loss might occur by simple increase of spine loss or a decrease in spine gains, but the current data reveal a more complicated situation in which both gains and losses are accelerated but the balance is tipped toward losses during aging. A striking phenomenon observed is widespread and early changes in cortical spine stability in the AD transgenic mice studied. At initial stages, there is clearly compensation such that there is no net spine loss and memory performance of APP/PS1 on multiple tasks is normal (Lalonde et al. 2004, 2005; Park et al. 2006; Gimbel et al. 2010; Couch et al. 2013; Um et al. 2013) at a stage when dendritic spine dynamics are already abnormal. This may be similar to the clinical notion of cognitive reserve. Here, the young APP/PS1 brain is not normal in its anatomical plasticity, but function and steady state synapse numbers are preserved. A process that starts at 3.5 months progresses to frank spine loss and memory impairment over months. Thus, the early changes in spine dynamics in the APP/PS1 cortex reflect successful compensatory mechanisms, which are overwhelmed as symptoms develop at latter ages. Fortunately, this provides a potential window for intervention if the early presence of altered anatomical plasticity can be recognized by clinical tests and treated.

Previous studies have shown that increased whisker stimulation results in increased turnover of dendritic spines in the barrel cortex. We housed the young and old transgenic APP/PS1 mice in a sensory-enriched environment (Yang et al. 2009; Akbik et al. 2013). Strikingly, the dendritic spine turnover of the AD model mice does not respond to increased whisker stimulation. To determine if the observed high dendritic spine turnover rates could be due to enhanced sensitivity to whisker stimulation, we deprived young APP/PS1 mice of all whisker input. The continued observation of elevated spine turnover in whisker-deprived APP/PS1 mice demonstrates sensory input independence. In contrast, c-Fos profiling of barrel cortex showed normal biochemical responsiveness of young and old APP/PS1 mice to whisker stimulation. Taken together, these data show that the highly elevated dendritic spine turnover rates are not due to increased sensitivity or altered whisker pathway signaling, but instead to pathologic and unregulated dendritic spine dynamics at the cortical neuron level.

On a molecular level, the basis for experience-independent elevation of dendritic spine clearly requires PrP^C, and by extension is likely to involve mGluR5 and Fyn kinase (Larson et al. 2012; Um et al. 2012, 2013; Hu et al. 2014; Haas et al. 2016; Kaufman et al. 2015). Indeed, pharmacological blockade of either mGluR5 or Fyn permits rapid restoration of synaptic marker density in APP/PS1 mice (Um et al. 2013; Kaufman et al. 2015). At a cellular and network level, the alterations in spine dynamics may relate to hyperactivation and excitability in the AD transgenic mice. Calcium imaging studies have revealed subsets of neurons with excess activation in the presence of Aβ oligomers or APP transgenes (Busche et al. 2008; Kuchibhotla et al. 2008; Busche et al. 2015). Such abnormal activation may drive anatomical changes, as well as epileptic phenomena (Palop et al. 2007; Sanchez et al. 2012; Um et al. 2012; Nygaard et al. 2015). It will be of interest to determine whether hyperactivation in young AD mice is correlated in a cell autonomous fashion to the altered dendritic spine dynamic phenotype. While we have imaged excitatory neuron dendritic spines, the primary deficit may exist in inhibitory neurons (Verret et al. 2012).

While memory deficits occur selectively in older AD transgenic mice, abnormal ocular dominance plasticity and developmental metaplasticity have been reported for young AD mice (William et al. 2012; Megill et al. 2015). These alterations of plasticity may be secondary to the abnormally high turnover rate that we observed in young AD mice. Further evaluation of the pathway that underlies the high turnover rates in young mice may provide a link to understanding why young AD mice have these abnormalities. The data suggest that there is a compensatory mechanism at a young age, which becomes overwhelmed in aging mice and results in the loss of synapses.

The high neuronal activity rate in young AD mice could contribute to the progressive loss of synapses over time by increasing Aβ release (Bero et al. 2011) and generating higher Aβo levels (Lesne et al. 2006; Kostylev et al. 2015). We investigated the levels of Aβ plaque and Aβo in the barrel cortex of our mice and we found that Aβ plaques were present in 10-month-old mice, but not 4-month-old AD mice. Additionally, PLISA revealed that Aβo were already present at 1.5 months and that Aβo levels had nearly doubled by 4 months of age. This is consistent with a correlation between rising Aβo levels, intensity of neuronal activity, and dendritic spine turnover rates in young behaviorally normal mice.

The majority of in vivo imaging studies that have been conducted have focused on dendritic spine dynamics and axonal varicosities within 50 µm of amyloid-plaques (Spires et al. 2005; Grutzendler et al. 2007; Spires-Jones et al. 2007; Kuchibhotla et al. 2008; Bittner et al. 2012; Hanson et al. 2014; Zou et al. 2015). Comparison of these studies has proven difficult as each study varies by strain of mouse, imaging technique, imaging age, labeling of neurons, labeling of plaques, and length of time imaged over (Dorostkar et al. 2015). We did not evaluate amyloid-plaque associated spine loss. Instead we studied the dendritic spine dynamics that were ongoing far from plaque (> 100 µm) to gain a better understanding of dendritic spine dynamics in regions without dense Aβ accumulation. Our results showed increased turnover rates in both young and old transgenic Alzheimer's mice. Future studies of spine dynamics near plaques should evaluate if prion protein plays a role in these dynamics. One previous experiment reported that the turnover rates near and far from plaque were not significantly altered in 4–5 month old APP/PS1 mice over 1 week of imaging, which differs from our findings in young AD mice (Zou et al. 2015). The different observations regarding turnover rates far from plaque could be due to the use of a cranial window, which is known to cause astrocytosis and gliosis to elevate WT turnover rates (Xu et al. 2007). In addition, their study used a shortened period of imaging, in which transient spine changes may obscure persistent spine dynamics. Since Aβo concentration varies inversely with distance plaque (Koffie et al. 2009; Kirkwood et al. 2013), our studies may reflect changes occurring at low Aβo concentrations, whereas observations near plaque may reflect alternate mechanisms triggered by higher Aβo concentrations and may include Aβ fibril and microglial action.

To further evaluate potential pathways that could underlie the high turnover rates in young AD mice, we employed RNAseq to compare AD mice to WT mice and AD mice with and without PrP^C. The two most interesting downregulated genes in the young AD mice are Npas2 and Homer1. Npas2 is a transcription factor that may play a role in the acquisition of certain types of long memory (Garcia et al. 2000). Mice lacking Npas2 show deficits in complex emotional long-term memory tests, but not in non-emotional memory tests. A homologue, Npas4, plays a role in regulating the development of inhibitory synapses (Lin et al. 2008; Spiegel et al. 2014). Homer1 is a postsynaptic density scaffolding protein, which is involved in the downstream signaling pathway of the Aβ-PrP^C-mGluR5 interaction (Um et al. 2013; Haas et al. 2014, 2016). Other differentially expressed genes linked to synaptic function of neuronal morphology include Rims1, Dpysl3, and Arpp21. Cirbp is a hypothermia-induced protein and the related protein RBM3 has been documented to play a role in synapse preservation in neurodegenerative models (Morf et al. 2012; Peretti et al. 2015). Together, the identification of this group of genes suggests a direct effect of the APP/PS1 transgene on synaptic function. Since the changes are PrP^C dependent and since Aβ oligomers bind selectively to PrP^C (Lauren et al. 2009), we hypothesize that Aβo drive these expression changes.

Importantly, at the stage when spine stability is altered in APP/PS1 mice, Aβ plaques have not formed, microgliosis is not yet evident, and expression profiles do not show evidence for the cytokine, acute phase reactants, and lysosomal changes that occur later. Thus, initial stages of pathophysiology driven by Aβo may trigger later glial reaction. Consistent with an initial neuronal phase, deletion of PrP^C expression rescues the spine dynamic and expression phenotypes.

In addition to synaptic markers, Ldlr and Stard4 are dysregulated early in APP/PS1 and CRND8 mice. It is remarkable that Prnp deletion corrects these values. The basis for the connection is not clear at this stage. It has been reported that PrP^C can interact with LRP1 and with ApoE, and these are potentially relevant interactions (Rushworth et al. 2013; Kanekiyo and Bu 2014). Cholesterol and lipid metabolism have not been linked directly to dendritic spine turnover rates.

In conclusion, the current work documents a widespread early dendritic spine phenotype in the APP/PS1 mouse model. This occurs in the presence of Aβo and requires PrP^C but precedes plaque formation and inflammation. The transgene-induced phenotype causes rapid spine turnover, a loss of persistent spines, and renders connections in the somatosensory cortex insensitive to an increase or decrease in sensory stimuli levels. Consistent with an early shift in spine dynamics, several synapse-relevant genes show altered expression. Early intervention or prevention strategies for AD may focus in these synaptic phenotypes.

Funding

National Institutes of Health, BrightFocus Foundation, Alzheimer's Association and Falk Medical Research Trust. Support for the AMP-AD study was provided by NIH U01 AG046139.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/

Disclosure

S.M.S. is a co-founder of Axerion Therapeutics seeking to develop PrP-based therapeutics for Alzheimer's disease.