Background
Alzheimer’s disease (AD) is characterized by amyloid deposits, intracellular neurofibrillary tangles, neuronal loss and a progressive decline in cognitive function [
1,
2]. Much progress has been made towards understanding the physiopathology of the disease, mostly through studies of transgenic mice designed to reproduce, as closely as possible, the histological and behavioral features of AD [
3,
4].
AD is multifactorial, but the abnormal processing of Amyloid Precursor Protein (APP) is a key element in its development [
5]. The physiological functions of APP are unclear, but it has been shown to play crucial roles in spine density, morphology and plasticity [
6]. The injection of anti-APP antibodies into the rat brain is known to induce behavioral impairments [
7,
8]. Furthermore, APP knockout mice have very low levels of dendritic complexity [
9]. Altogether, this suggests that APP has an important physiological role associated with synaptic plasticity as well as trophic properties. Overexpression of wild-type (WT) APP and various mutant forms has been used as a means to model AD in many transgenic mouse lines [
10,
11]. In most of these transgenic strains, the significant increase in APP production beginning
in utero may trigger consequences that are not likely mimicking the biochemical deficit observed in AD. Interestingly, Saito and coworkers recently described a new APP knock-in model without APP overproduction [
12]. This model reproduces the cognitive deficits and amyloid plaques of AD, but unfortunately does not provide information about changes occurring early in the development of the pathology.
If the role of amyloid component is crucial, the role of amyloid plaque deposition in disease development is currently a matter of debate [
13]. The presence of plaques is a diagnostic criterion for AD, but several studies have suggested that the accumulation of amyloid deposits may have a protective function [
14]. Moreover, an absence of plaque has been reported in patients with familial AD and mutant forms of APP [
15], whereas abundant Aβ plaques have been found in brain samples from elderly patients without clinical dementia [
16‐
19]. Plaques appear many years after disease onset and they cannot, therefore, be responsible for the early events in AD development [
20]. By contrast, soluble Aβ may play an important role in the synaptic and cognitive impairments observed in the early stages of AD [
21]. The use of transgenic models displaying higher levels of APP and cleavage products compared to the human situation and inducing artificial phenotypes in few months is therefore likely to be inappropriate for studies of the initial phases of AD. The levels of Aβ produced in these models are much higher than those observed in patients and may have toxic effects unrelated to the early phases of AD. In addition, the negative outcomes of recent clinical trials have fueled debate about the validity of overexpression models. Indeed, most of the therapeutic strategies previously tried and largely unsuccessful, have been tested in such transgenic models. There is a growing body of evidence suggesting that amyloid plaques and tangles occur late in disease progression. Therefore, the development of pertinent protective or disease-modifying therapeutic strategies based on the decrease of these markers does not seem to fit well [
22,
23]. These compelling observations demonstrate the need to develop new alternative models of AD more closely mimicking the human disease and in particular the early events in its development.
The present study is an attempt at developing such an alternative model involving the production,
in vivo, in the mouse hippocampus, of moderate levels of amyloid derivatives, resembling as closely as possible the pattern of expression observed in the hippocampus of human AD patients to study the consequences of initial amyloid pathway engagement. We used this modelling strategy to analyze the events potentially contributing to AD development before the appearance of late hallmarks of the disease, such as amyloid deposits, neurofibrillary tangles and neuronal death. The injection of AAV vectors carrying mutated forms of human APP and PS1 into the mouse hippocampus led to the stable production [
24,
25] of APP, βCTF and Aβ peptides, at levels similar to those observed in the hippocampus from AD patients and significantly lower than those present in the hippocampus of APP/PS1ΔE9 transgenic mice. The data generated demonstrate the importance of the Aβ42/Aβ40 ratio, which has already been identified as a relevant biomarker in AD patients [
26], together with early changes in synaptic functions, Tau phosphorylation and cognitive deficits. These modifications were observed in the absence of plaque formation, or any sign of inflammation, atrophy and/or neuronal death. They were nevertheless capable of inducing cellular changes, such as the abnormal activation of extrasynaptic NMDARs and a decrease in the levels of neuromodulation-associated metabolites, causing memory impairment. These results suggest that APP processing in a limited number of neurons, as recently observed in sporadic forms of AD [
27], may be sufficient to trigger an impairment of hippocampal-dependent behavior.
Discussion
Alzheimer’s disease (AD) is a complex condition. To improve our understanding of the physiopathology of the disease and design new therapeutic strategies, it is crucial to get access to the initial phases of its development. We aimed at recapitulating
in vivo key features that are suspected to account for triggering AD, more specifically an increased Aβ42/Aβ40 ratio, while avoiding major APP overexpression. The importance of the Aβ ratio in the disease onset above their absolute levels has been already proposed [
40] and the absence of APP overproduction is increasingly being recognized as an important factor to take into account when modeling AD [
12]. Towards this goal, we developed an alternative approach to existing models, involving a single co-administration of two AAV vectors encoding human mutated APP (APP
sl) and PS1 (PS1
M146L) in adult wild-type mice hippocampi.
Transgenic mice expressing WT form of the 695-amino acid isoform of APP do not develop AD-like phenotype nor behavioral impairment [
41,
42] in contrast to APP751 transgenic mice which have strong learning deficits at 12 months of age [
11]. Therefore, the APP751 isoform seems more interesting to trigger an AD-like phenotype. In order to specifically enhance amyloidogenic APP processing, we decided to express the APP751 isoform including the Swedish and London mutations [
43]. We simultaneously expressed PS1 with the M146L mutation to increase the specificity of the 42-specific γ-secretase cleavage [
44]. Thus, the co-expression of PS1 (PS1M146L) with APP751SL resulted in an increased production of Aβ with an Aβ42/Aβ40 ratio similar to the one observed in AD patients. Interestingly, increase of APP cleavage products was not associated with amyloid deposition, neurofibrillary tangles and neuronal loss.
AAV-APP/PS1 mice display cognitive impairment as early as three months post-injection. These memory deficiencies reflect a global impairment of neuronal activity, as suggested by 1H-MRS, showing the levels of metabolites related to neurotransmission expressed in the hippocampus globally reduced. These result highlights that there is an early response of neurons to the APP processing affecting their functionality.
To get further insight in the understanding of this neuronal dysfunction, we used electrophysiological recordings on the CA1 layer to characterize synaptic consequences of moderate soluble Aβ42 production. We observed an enhancement of the tonic glutamatergic current generated in CA1 pyramidal cells of AAV-APP/PS1 mice. Two types of N-methyl D-aspartate receptors (NMDARs) have been described (synaptic and extrasynaptic), the tonic current resulting mostly from the activation of the extrasynaptic subgroup, stimulated by ambient glutamate present outside the synaptic cleft [
45,
46]. As we did not observe an increase of glutamate synthesis by
1H-MRS, the higher tonic current observed in AAV-APP/PS1 mice may reflect the accumulation of glutamate in the extrasynaptic space. There are two possible explanations for this possible accumulation. First, lower levels of glutamate transporters, such as GLT-1, may result in weaker levels of glutamate uptake by astrocytes [
35]. Second, lower levels of scaffolding proteins, such as PSD-95, may lead to the internalization of synaptic NMDARs, resulting in the preferential activation of extrasynaptic ones [
47]. AAV-APP/PS1 mice display these features since they express significantly lower levels of GLT-1 and PSD-95. No decrease of GLAST, the second most important glutamate transporter, was observed which might suggest that GLAST expression could be decreased later in the course of the disease [
48]. This hypothesis is not surprising, in light of published data about GLT-1 and PSD-95 dysfunctions, which may play an important role in synaptic dysfunction and, thus, in the pathogenesis of AD [
49,
50]. To our knowledge, we show for the first time that an impairment of the extrasynaptic compartment could precede sustained alterations of the synaptic compartment associated with LTP deficits, during the early stage of AD progression. We also showed that a subtle loss of neuronal integrity in the hippocampus is sufficient to trigger deleterious effects before actual neuronal death occurs. Despite diffusion properties of Aβ42 [
51], no amyloid peptide have been detected outside the hippocampus excluding an Aβ42 direct effect far from its hippocampal production area. By contrast, Aβ altered hippocampus functions could lead to a less efficient communication between hippocampal neurons and connected structures especially cortex and amygdala. This network dysfunction could be responsible of cognitive modifications especially during memory formation [
52,
53] or emotional behaviors [
54].
In addition to these consequences, AAV-APP/PS1 mice displayed higher levels of Tau phosphorylation (thr181, AT270 antibody) compared to the other groups. It has been reported that soluble Aβ and extrasynaptic NMDAR activation contribute to Tau protein phosphorylation [
55,
56]. Our results are consistent with these previous reports. The pattern of Tau phosphorylation was shown to be correlated with the multiple steps of neurofibrillary tangle development [
57]. Thr181 immunoreactivity is detected earlier than thr212/Ser214 (AT100 antibody) or ser199/ser202/thr205 (AT8 antibody) immunoreactivity. We could not detect some AT8 or AT100 positive cells (data not shown), suggesting that the modifications observed in AAV-APP/PS1 mice might be compared to early phases of AD in humans. These data suggest that the moderate AD-like APP processing engagement observed in AAV-APP/PS1 mice is sufficient to trigger the Tau pathway engagement.
A particular feature in our experimental strategy is the localized production of human APP and its cleavage products. Contrasting with the ubiquitous overproduction of human APP in mouse transgenic lines, injection of AAV-APP and AAV-PS1 vectors in the stratum lacunosum moleculare region, leads to the transduction of neurons in restricted regions of the hippocampus (CA2 and the subiculum). Interestingly, this pattern of expression may mimic the genomic mosaicism recently described in AD, in which an increase in copy number was observed for the APP gene in a limited number of neurons, in sporadic forms of AD [
27]. The presence of rare neurons with APP amplification may be sufficient to trigger a dysregulation of APP processing with aging. Our findings also point to this direction. The second particular feature in our experimental design is the co-injection of the PS1 vector. Indeed, we demonstrated that AAV-APP injection alone was not sufficient to induce hippocampal alteration. Thus, despite the production of βCTF, Aβ40 and Aβ42 in AAV-APP mouse hippocampus, no change in the levels of phosphorylated Tau, PSD-95, GLT-1 or tonic current were observed, compared to the AAV-APP/PS1 group. There are two possible reasons for this difference between the AAV-APP and AAV-APP/PS1 groups. First, APP acts as a trophic factor [
58] and an increase in the levels of this molecule may increase neuron viability [
59], countering Aβ toxicity. Second, the Aβ42/Aβ40 ratios in the AAV-APP/PS1 group were highly similar to those found in AD patients and different from those of the AAV-APP group or human controls.
Our data raise several questions. This configuration, with no plaque nor tangle and with levels of APP cleavage products close to the human hippocampal condition, is sufficient to induce cognitive impairment three months after injection. It takes about 13 months to obtain an equivalent defect in APP/PS1ΔE9 mice [
60]. This delayed cognitive impairment may result from the trophic roles of APP and the low Aβ42/Aβ40 ratio observed in 5- to 16-month-old APP/PS1ΔE9 mice. These findings confirm the greater importance of the Aβ ratio rather than the absolute amounts of Aβ42 and Aβ40 [
40] to induce synaptic and extrasynaptic impairments. Finally, neurons may become brittle, leading to cognitive impairments, with defects in emotional behavior and long-term memory.
Methods
Plasmid design and vector production
We used a double-mutant human APP751 cDNA containing the Swedish and London mutations (codon optimized and containing a Kozak sequence; GeneArt, Life Technologies, Saint Aubin, France), and a human PS1 cDNA containing the M146L mutation (pENTR4-PS1-S182M146L
). The APP
SL and PS1M146L sequences were cloned in an AAV2 plasmid with CAG promoter to generate the AAV2-CAG-APPSL or -CAG-PS1M146L. AAV vectors were produced as previously described [
61], except that the AAV packaging plasmid was replaced with a plasmid construct containing the
rep gene of AAV2 and the
cap gene of AAVrh10.
Human brain samples
Postmortem samples were obtained from brains collected as part of the Brain Donation Program of the GIE-Neuro-CEB Brain Bank. Autopsies were carried out by accredited pathologists, after informed consent had been obtained from the relatives, in accordance with French Bioethics laws. Five hippocampal samples from five patients with sporadic forms of AD (male and female; Braak 6 Thal 5; aged between 69 and 89 years, with a postmortem interval (PMI) of 30 to 59 h) and five hippocampus samples from five age-matched control subjects (male and female, aged between 69 and 92 years, PMI of 6 to 63 h) were used in this study.
Animals
We used 90 male C57Bl/6 J mice (eight-week-old; SARL JanvierLabs, Le Genest Saint Isle, France) and 19 male APP/PS1ΔE9 mice [
31]. All experiments were conducted in accordance with ethical standards and French and European regulations (Directive 2010/63/EU).
Stereotactic injections of AAVs
Mice were anesthetized by an intraperitoneal injection of ketamine/xylazine (0.1/0.05 g/kg body weight) and placed in a stereotactic frame (Stoelting, Wood Dale, IL, USA). Stereotactic intracerebral injections of AAVs into the hippocampi of both hemispheres were performed, using the following coordinates: antero-posterior: -2 mm, lateral: ± 1 mm, ventral: -2 mm relative to bregma. We injected 2 μl of viral preparation into each site (5 x 108 vg/site and 109 vg/site for AAV-PS1 and AAV-APP vectors, respectively) at a rate of 0.2 μl/min. Four groups were designed, a non-injected wild-type (non-injected WT mice) and three injected groups: AAV-CAG-PS1M146L (AAV-PS1 mice), AAV-CAG-APPSL (AAV-APP mice), AVV-CAG-APPSL + AAV10-CAG-PS1M146L (AAV-APP/PS1 mice).
Tissue collection and sample preparation
Mice were anesthetized with ketamine/xylazine and transcardially perfused with 20 ml ice-cold phosphate-buffered saline (PBS). One hemisphere was post-fixed by incubation for 48 h in 4 % PFA, cryoprotected in 30 % sucrose in PBS and cut into 40 μm sections with a freezing microtome (Leica) for histological analyses. The contralateral hemisphere was dissected for isolation of the hippocampus and cortex. Samples were homogenized in a lysis buffer (150 mM NaCl and 1 % Triton in Tris-buffered saline, referenced as TBS-Tx) containing phosphatase (Pierce) and protease (Roche) inhibitors and centrifuged for 20 min at 15000 g. The same procedure was applied to human samples (GIE NeuroCEB Brain Bank).
ELISA
The Aβ extracted was quantified with the MSD Human Aβ42 V-PLEX Kit and the triplex Aβ Peptide Panel 1 (6E10) V-PLEX Kit (Meso Scale Diagnostics, Rockville, USA). βCTF was determined with the IBL Human APP βCTF Assay Kit (IBL International GmbH, Hamburg, Germany). Hyperphosphorylated Tau was determined with the Innogenetics Phospho-Tau 181P kit (Fujirebio Europe, Ghent, Belgium). sAPPβ was determined with the MSD sAPPalpha/sAPPbeta Kit. ELISA was performed according to the kit manufacturer’s instructions in each case.
Western blotting
Equal amounts of protein (30 μg) were separated by electrophoresis in NuPAGE Bis-Tris Gels (Life Technologies) and transferred to nitrocellulose membranes. The membranes were hybridized with various primary antibodies (APP 6E10, 1/500, Covance; PS1, 1/1000, Millipore; APP C-ter, 1/500, Millipore; Actin, 1/2000, Abcam; GAPDH, 1/1000 Abcam; Total Tau, 1/1000, Santa Cruz; PSD-95, Invitrogen, 1/2000; Synaptophysin, Santa Cruz, 1/200; GAD65, Abcam, 1/2000; GLT-1, 1/1000, Frontier Science; GLAST, 1/1000, Frontier Science; NeuN, Millipore, 1/1000). Various secondary antibodies was also used (ECL Anti-rabbit Horseradish Peroxidase linked, 1/2000, GE Healthcare; ECL Anti-mouse Horseradish Peroxidase linked, 1/2000, GE Healthcare; ECL Anti-rat Horseradish Peroxidase linked, 1/2000, GE Healthcare).
Immunohistochemistry and image acquisition
Cryosections were washed with 0.25 % Triton in PBS and saturated by incubation (0.25 % Triton in PBS/5 % goat serum). They were then incubated with primary antibodies (APP C-ter, 1/500, Millipore; NeuN-Biotin, 1/1000, Millipore; 4G8-Biotin, 1/1000, Covance). Images were taken with a Nikon Eclipse Ti Microscope or a Leica TCS SP8 confocal microscope and analyzed with ImageJ software (NIH).
Behavioral assessment
Open-field
The apparatus consisted of an open-topped, clear Plexiglas box measuring 50 x 50 x 38 cm placed in a room with controlled dim lighting (25 lux) and constant white noise at 60 dB. The mice were placed in the center of the arena and a video recording was made over a period of five minutes. The behavior of the animals was analyzed from this video. The arena was divided into a central region (20 x 20 cm) and a peripheral region, and the time spent in the center and periphery of the open field was measured. The ratio of time spent in the periphery to that spent in the center was calculated as an index of emotional behavior.
Ymaze
The apparatus consisted of three identical arms separated by 120°. Each arm of the Y maze was 37 cm long, and 8 cm wide, with 12.5 cm-high opaque walls. Various extra maze cues were placed on the surrounding walls. One arm of the Y-maze was blocked and the subject was allowed to explore the other two arms for 10 min. The starting position was varied pseudorandomly, between the three arms. The animal was then returned to its home cage. Fifteen minutes later, the mouse was placed in the maze again, this time with all three arms open, and allowed to explore for an additional five minutes. The distance traveled and the number of times the mouse entered each arm were measured both during initial exposure to the maze and during testing.
Morris water maze
Experiments were performed in a tank 120 cm in diameter and 50 cm deep, filled with opacified water kept at 21 °C and equipped with a platform 10 cm in diameter, kept submerged 1 cm below the surface of the water. Visual clues were positioned around the pool, to provide the mouse with spatial landmarks, and luminosity was maintained at 350 lux. The mice were initially exposed to a learning phase, which consisted of daily sessions (three trials per session) on five consecutive days. The starting position was varied pseudorandomly, between the four cardinal points. A mean interval of 15 min was left between trials. The trial was considered to have ended when the animal reached the platform. A 60-s cutoff was used, after which the mice were gently guided to the platform. Once on the platform, the animals were allowed to rest for 30 s before being returned to their cage. Long-term spatial memory was assessed 72 h after the last training trial (fifth day), in a probe trial in which the platform was no longer available. Animals were monitored with ANY-maze video tracking software (Stoelting Co, Wood Dale, USA).
Ex vivo electrophysiology
Mice were anesthetized with halothane and decapitated. The brain was rapidly removed from the skull and placed in chilled (0–3 °C) artificial cerebrospinal fluid (ACSF) containing 124 mM NaCl, 3.5 mM KCl, 1.5 mM MgSO4, 2.5 mM CaCl2, 26.2 mM NaHCO3, 1.2 mM NaH2PO4, 11 mM glucose. Transverse slices (300–400 μm thick) were cut with a vibratome and placed in ACSF in a holding chamber, at 27 °C, for at least one hour before recording. Each slice was individually transferred to a submersion-type recording chamber and submerged in ACSF continuously superfused and equilibrated with 95 % O2, 5 % CO2.
The biophysical properties of the tonic current generated by the activation of extrasynaptic NMDA receptors with ambient glutamate were evaluated. Whole-cell patch-clamp recordings of CA1 pyramidal cells were performed at room temperature, with borosilicate patch pipettes (5 MΩ) filled with 140 mM CsCH
4O
3S, 6 mM CsCl, 2 mM MgCl
2, 10 mM HEPES, 1.1 mM EGTA, 5 mM QX-314 5, 4 mM ATP, (pH 7.3; 290 mosm). Transmembrane currents were acquired and filtered through an amplifier (AxoPatch 1-D, Axon Instruments), stored on a computer and digitized with WinLTP software for analysis [
62]. The tonic current was recorded at a holding potential of +40 mV, in the presence of TTX (1 μM), NBQX (10 μM), and bicuculline (10 μM), to isolate the NMDA component of the holding current (hc). After the recording of a stable control hc for three to five minutes, APV (50 μM) was added to the superfusion medium. The hc fell to a new stable value under the effect of APV, and the difference between the control hc and that recorded in the presence of APV determined the amplitude of the tonic current.
Two kinds of electrically induced long-term potentiation (LTP) were studied: a strong, saturating LTP consisting of 3x100 Hz (3x 100 pulses, 1 s, with 20 s between pulses), and a weaker stimulation, theta-burst stimulation (TBS), mimicking the natural stimulation at the theta frequency from the medial septum to the hippocampus, consisting of five trains of four 100 Hz pulses each, separated by 200 ms and delivered at the test intensity. The sequence was repeated three times, with an interburst interval of 10s. Testing with a single pulse was resumed for 60 min (TBS) or 75 min (3x100 Hz), to determine the level of LTP.
Proton magnetic resonance spectroscopy (1H-MRS)
Magnetic Resonance Imaging and Magnetic Resonance Spectroscopy were performed with a horizontal 11.7 T scanner (Bruker, Ettlingen, Germany) and a quadrature cryoprobe was used for radiofrequency transmission and reception. A 37.4 μl voxel (7.2 × 2 × 2.6 mm
3) was placed over both hemispheres, such that it contained essentially hippocampal tissue (Fig.
5), and the signal of this voxel was then averaged over 10 min. T
2-weighted images were acquired with a 2D TurboRARE sequence (80 × 80 μm
2 in-plane resolution, and 300 μm slice thickness) and manually segmented to measure hippocampal volume.
1H-MRS acquisitions were performed with a LASER sequence, with echo time (TE)/repetition time (TR) = 20/5000 ms and a 10 kHz bandwidth for the hyperbolic secant pulses. LCModel was used to determine metabolite concentrations. The macromolecule (MM) spectrum of a control mouse was determined by metabolite nulling and included in the base set for LCModel. The following metabolites were systematically quantified (Cramér-Rao lower limits <5 % in all experiments): total choline (glycerophosphocholine + phosphocholine + choline, tCho), total creatine (creatine + phosphocreatine, tCr), glutamate (Glu), glutamine (Gln), myo-inositol (Ins), N-acetyl-aspartate + N-acetyl-aspartyl-glutamate (NAA + NAAG, tNAA), taurine (T) and γ-aminobutyric acid (GABA). Metabolite concentrations were normalized with respect to 8 mM tCr.
Statistical analysis
Data are expressed as mean ± SEM. Statistical analyses were performed with GraphPad Prism (GraphPad Software, La Jolla, CA, USA) or Statistica (StatSoft, Inc., Tulsa, OK, USA) software. One-way ANOVA and Tukey’s post-hoc test were used to determine the significance of differences between groups. Student’s t test was used when only two groups were analyzed (AAV-PS1 vs AAV-APP/PS1), except for tonic glutamatergic current recording, for which Chi2 tests were used. Two-way ANOVA with repeated measure using Group and Metabolite as effect factors, was used for NMR spectroscopy analysis.
Competing interest
The authors have no competing financial interests to declare.
Authors’ contributions
MA, NC and JB carried out the design of the study and wrote the manuscript. MA and JB performed biochemical, behavioral, histological and statistical analyses. RF, SA, MAB and GDC participated in the biochemical analyses and helped to draft the manuscript. PD, BP and JMB performed the electrophysiological recording and helped to design the study and draft the manuscript. JF and JV performed the MRI analyses. APB helped with the viral production. PH and ND helped to draft the manuscript. All authors read and approved the final manuscript.