|Home | About | Journals | Submit | Contact Us | Français|
Alzheimer’s disease (AD) is hallmarked by amyloid plaques, neurofibrillary tangles, and widespread cortical neuronal loss (Selkoe, 2001). The ‘amyloid cascade hypothesis’ posits that cerebral amyloid sets neurotoxic events into motion that precipitate Alzheimer dementia (Hardy and Allsop, 1991). Yet, faithful recapitulation of all AD features in widely used transgenic (Tg) mice engineered to overproduce Aβ peptides has been elusive. We have developed a Tg rat model (line TgF344-AD) expressing mutant human amyloid precursor protein (APPsw) and presenilin 1 (PS1ΔE9) genes, each independent causes of early-onset familial AD. TgF344-AD rats manifest age-dependent cerebral amyloidosis that precedes tauopathy, gliosis, apoptotic loss of neurons in the cerebral cortex and hippocampus, and cognitive disturbance. These results demonstrate progressive neurodegeneration of the Alzheimer type in these animals. The TgF344-AD rat fills a critical need for a next-generation animal model to enable basic and translational AD research.
Alzheimer’s disease (AD) is the most common form of dementia in elderly populations and is hallmarked by progressive: 1) deposition of amyloid-β peptides (Aβ; cleaved from amyloid precursor protein, APP) as β-amyloid plaques, 2) formation of neurofibrillary tangles (NFTs; chiefly comprised of the hyperphosphorylated microtubule-associated protein tau), 3) chronic neuroinflammation, and 4) neuronal injury and loss (Selkoe, 2001). As initially proposed, the ‘amyloid cascade hypothesis’ purports that Aβ deposition as ‘senile’ amyloid plaques is the principal etiopathological event in AD (Hardy and Allsop, 1991). The hypothesis further posits that Aβ aggregation sets downstream pathophysiologic processes into motion that culminate in neuronal injury and loss and precipitate dementia of the Alzheimer type (Hardy and Allsop, 1991; Rozemuller et al., 1989). The strongest support for the hypothesis comes from human genetic evidence: all mutations in APP or presenilins 1 or 2 (PS1/PS2) that drive Aβ accumulation invariably cause early-onset AD (Selkoe, 2001).
Yet, one of the key critiques of the hypothesis is that the ‘gold standard’ animal models of the disease, Aβ-overproducing Tg AD mice (Games et al., 1995; Holcomb et al., 1998; Hsiao et al., 1996; Jankowsky et al., 2001; Mucke et al., 2000; Sturchler-pierrat et al., 1997), do not demonstrate robust tauopathy and neuronal loss unless additional human transgenes are included that are not linked to familial AD (Colton et al., 2008; Oddo et al., 2003; Padmanabhan et al., 2006; Wilcock et al., 2008). Thus, the core tenant of the amyloid cascade hypothesis – that cerebral Aβ is both necessary and sufficient for development of AD – remains controversial. Because rats are 4–5 million years closer to humans than mice (Yang et al., 2004), we hypothesized that they may provide a better animal model for AD. Here, we report the generation and characterization of transgenic rats bearing mutant human APP and PS1 (line TgF344-AD) that manifest the full spectrum of age-dependent AD pathologies in conjunction with cognitive disturbance.
TgF344-AD rats were generated on a Fischer 344 background by co-injecting rat pronuclei with two human genes driven by the mouse prion promoter: ‘Swedish’ mutant human APP (APPsw) and Δ exon 9 mutant human presenilin-1 (PS1ΔE9). Both constructs used have previously been reported (Jankowsky et al., 2001). Transgene integration was confirmed by genotyping and expression levels were evaluated by Western blot (WB) of brain homogenates. Line 19 was selected for colony expansion, aging, and detailed analyses based on highest overall transgene expression levels. TgF344-AD rats and wild-type (WT) littermates were housed and maintained at the Cedars-Sinai Medical Center Department of Comparative Medicine vivarium, and all experiments were conducted in compliance with protocols approved by the Institutional Animal Care and Use Committee. We did not observe gender differences on any of the measures reported, and therefore males and females were combined for all analyses. TgF344-AD rats are available upon request via material transfer agreement.
Rats were initially behaviorally evaluated by neurological screen, which consisted of a battery of neurological reflex tests: righting response after being placed on the dorsal side; eye blink (response to light touch with a small camel hair brush), ear twitch and limb withdrawal in response to tactile stimuli (light touch with a gloved finger); orienting response to olfactory (exposure to orange extract) and visual stimuli (flashlight); and startle response following an auditory stimulus (a metal clicker). Responses were scored as being present or absent. Visual and tactile responses were also assessed. For visual and tactile placing, the rat was loosely held in a gloved hand. For visual placing, the animal was allowed to view the surface of a table while the head and forelimbs were slowly brought down to the surface. For tactile placing, the rat was prevented from viewing the surface of a table by holding its chin up, and the forelimbs were slowly brought down to the surface. Placing was assessed by counting the number of correct placements of the limbs across 10 trials. Approximately 5 min/rat was required to complete the neurological screen. All behavioral tests were of a cross-sectional design and were conducted by a blinded examiner. Furthermore, code was not broken until all analyses were completed.
Locomotor activity was evaluated using a standard open field test. The open field apparatus (San Diego Instruments, San Diego, CA) consisted of an open topped, clear Plexiglas box measuring 40.6 × 40.6 × 38 cm. A ring of photo beams and optical sensors surrounded the box. The optical sensors were connected to a computer by way of an input matrix, and breaks in the photo beams were automatically recorded and used as a measure of locomotor activity. The observation cages were cleaned before the first run of the day, between subjects, and after the last run of the day using tap water followed by ethanol. On the day of testing, rats in their home cages were brought into the experimental room. The level of illumination in the room during experimental testing was set at 325 lux. Rats remained in the experimental room for 30 min, after which each rat was placed into the center of the observation cage, and counting began immediately. Beam breaks (locomotion) were recorded in 5 min bins for 60 min. The apparatus was programmed to record both central (defined as a 30.5 × 30.5 cm region in the center of the box) as well as total beam breaks. Data were plotted as beam breaks during each 5 min interval. For statistical analysis, the scores of beam breaks across the 60 min session were summed for each subject.
An object recognition test was performed to assess learning and memory. During the sampling phase, rats were removed from their home cages and placed in the object recognition box, which contained two identical objects (A1 and A2) fixed to the floor. After 3 min, the rat was removed from the Object Recognition Box and returned to its home cage. Objects A1 and A2 were also then removed from the box. Twenty min were allowed to elapse prior to the start of the 3 min choice phase. During the choice phase, rats were singly removed from their home cages and returned to the object recognition box, where two new objects (A3 and B) were fixed to the floor. Object A3 was identical to objects A1 and A2, whereas object B differed from objects A1, A2, and A3. A videotape recording was made of the sample and choice phases, and two raters independently rated the time each rat explored each of the objects during the sample and choice phases. The percent time spent exploring the new object (C) during the choice phase corrected for any location preference during the sample phase was taken as a measure of memory. Exploration of an object was defined as when the rat directed its nose toward the object at a distance of less than 2 cm.
The Barnes maze is a widely accepted test of hippocampus-dependent spatial reference learning and memory in rats. The maze (San Diego Instruments, San Diego, CA) consisted of a circular platform (122 cm in diameter) with 20 holes evenly spaced around the perimeter, and an aversive stimulus (bright light) located overhead. A dark, recessed chamber (escape box) was located under one of the perimeter holes of the platform. ‘False’ escape boxes, too small for the rat to enter but otherwise identical to the escape box, were oriented under the other holes. To create different learning paradigms, the escape box can be moved to a different location or the maze can be rotated. Photobeams and optical sensors surrounded each hole, and the optical sensors were connected to a computer by way of an input matrix, and breaks in the photo beams were automatically recorded. To begin the experiment, the rat was placed into a start tube (30.5 cm high and 20.3 cm in diameter) in the center of the maze. After a 30 s acclimation period, a bright spotlight was illuminated and the start tube was lifted off of the rat. The session ended when the rat entered the escape box, or after 3 min elapsed. Once the rat entered the escape box, the spotlight was turned off and the rat was allowed to remain in the box for 2 min. Training was repeated 4 times daily. The maze was rotated to change the position of the escape box by at least 90° each day, and the position of the small hole alignment remained consistent (i.e., one hole was always centered on the north edge). Rotation of the escape box forces the rat to use external spatial cues from the room rather than olfactory or local cues. During testing, the number of errors, the location of the first hole searched, and the latency to enter the goal box used to solve the maze were recorded. Errors were defined as searches of any hole that did not have the goal box beneath it. Searches included nose pokes and head deflections into a hole. However, successive pokes into the same hole were not counted as repeated errors. After 4 d of testing, 72 h were allowed to relapse before the animals are tested again for retention. After a single retention trial, the location of the box was shifted, and animals were retrained as above on the next day and the following day.
Antibodies against PS1 (clone NT1) and Aβ/APP (clones 6E10, 4G8, 12F4, 11A50, and 22C11), were obtained from Covance and were used at dilutions of 1:200 for IHC and 1:500 for WB. Antibodies against abnormally phosphorylated tau (clones CP13 and PHF1, 1:50 for IHC and 1:100 for WB) were generously supplied by P. Davies; the oligomeric/conformational Aβ antibody (clone OC, 1:2,500 for IHC) was gifted by D.H.C. and C.G.G. and clone A11 (1:500 for IHC) was obtained from Biosource. The pTau-PADRE pS396/404 antibody was made by immunizing rats with synthetic tau peptide (382–418) conjugated with the prototypical T cell epitope, PADRE (Think Peptides, Sarasota, FL). An antibody against total tau (clone Tau-5, 1:500 for WB), phalloidin (AlexaFluor 647 conjugated, 1:40 for IHC) and ProLong Gold anti-fade mounting media with DAPI were obtained from Invitrogen. Antibody against Tuj1 (1:1000 for IHC and WB) was purchased from Sigma, and actin antibody (1:1000 for WB) was obtained from Millipore. An antibody against Iba1 (1:200 for IHC) was obtained from Wako Ltd. TdT-mediated dUTP nick-end labeling (TUNEL) stain was purchased from Roche and used in accordance with the manufacturer’s instructions. Antibodies against total and cleaved Caspase-3 were obtained from Cell Signaling Technology (1:200 for IHC; 1:1000 for WB).
Following cardiac perfusion with sterile ice-cold phosphate-buffered saline, rat brains were rapidly isolated and bisected into hemispheres. One hemisphere was fixed in 4% paraformaldehyde (PFA) prior to routine processing and paraffin embedding for histochemical analyses. The remaining hemisphere was weighed, snap frozen, and homogenized in 2 mL of ice-cold lysis buffer (Cell Signaling Technology) supplemented with 1 mM phenylmethylsulfonyl fluoride for biochemical analysis. Briefly, brains were mechanically dissociated using an Ika disperser for 2 min and were allowed to stand for 15 min at 4 C. Following homogenization of snap-frozen hemispheres, samples were separated into 2 equal aliquots, which were then centrifuged at 10,000 g for 15 min at 4 °C and stored at −80 °C. One aliquot was subjected to biochemical analysis for APP, Aβ, and Caspase-3, while the remaining aliquot was used for tau biochemical analysis. For EM, rats were perfused for 5 min at 120 mm Hg through the transcardial access using 4% buffered formalin freshly prepared from PFA mixed with 0.1% glutaraldehyde. Freshly isolated brains were post-fixed in the same solution overnight, infiltrated in a graded series of glycerols, and then sectioned in the coronal plane at 100 μm thickness using a vibratome. Regions of interest were micro-dissected, routinely processed with osmium tetroxide followed by uranyl acetate, and embedded in Epon for transmission EM analysis.
Ten μm para-median sagittal sections were sliced at 50 μm intervals using a microtome and mounted on glass slides. Sections were routinely dewaxed, hydrated in a graded series of ethanol, and boiled for 30 min in antigen retrieval buffer prior to serum-free protein block (Dako Cytomation) application. Sections were then hybridized with various primary antibodies followed by incubation with the appropriate horseradish peroxidase (HRP)- or fluorophore-conjugated secondary antibodies. For HRP-conjugated secondary antibodies, sections were developed with an HRP-labeled polymer-based kit (Dako EnVision) coupled with the 3′-3′ diaminobenzidine substrate followed by routine dehydration in a graded series of ethanols and xylene. For amyloid burden, sections were directly stained with ThioS according to standard practice. All sections were coverslipped with the appropriate mounting media (Prolong Gold or Permount) prior to imaging. Bright-field and structured illumination fluorescent images were obtained using a Zeiss AxioImager Z1 with attached ApoTome and CCD camera (Carl Zeiss Microimaging). Confocal images were taken using a Eclipse C1 instrument with four independent laser lines (Nikon Instruments). Images were digitized into a PC running Windows XP, and image analysis of micrographs was conducted using ImageJ software (NIH).
We carried out biochemical analysis of Aβ peptides according to a two-step extraction method (Johnson-wood et al., 1997; Tan et al., 2002). Briefly, detergent-soluble Aβ1-40, 42 species were separately detected in rat brain homogenates prepared with lysis buffer described above at a 1:25 dilution. Detergent-insoluble Aβ1-40, 42 species were detected by extraction of homogenate pellets in the chaotropic agent, 5 M guanidine-HCl, followed by a 1:12,500 dilution in lysis buffer. Protein levels were normalized by BCA protein assay (Pierce Biotechnology). Aβ species were separately quantified in detergent-soluble and -insoluble (5M guanidine-HCl-extracted) fractions using Aβ1-40, 42 ELISA kits (Invitrogen) and (N) 82E1 Aβ oligomers (IBL Laboratories) in accordance with the manufacturer’s instructions, except that standards included guanidine-HCl in some cases. For tau analysis, pellets were re-homogenized in a 10% salt-sucrose solution to obtain various soluble and insoluble fractions for WB (Greenberg and Davies, 1990). Crude pellet tau was obtained by re-homogenization of pellets with Tris-buffered saline (pH 7.4), followed by gentle centrifugation at 1,000 g for 5 min at 4 °C. Protein levels of homogenate samples were determined by BCA protein assay prior to electrophoresis. Aliquots of protein were electrophoretically separated using 10% Bis-Tris gels. Electrophoresed proteins were then transferred to nitrocellulose membranes, blocked in Tris-buffered saline (TBS) containing 5% (w/v) non-fat dry milk, and subsequently hybridized with various primary antibodies. Membranes were then incubated with the appropriate HRP-conjugated secondary antibody prior to development with chemiluminescent substrates. Densitometric analysis of blots was conducted using ImageJ software.
Whole number neuronal estimates were done using the optical fractionator method of stereological counting with stereological software (Stereo Investigator; MBF Bioscience). Para-median sagittal serial sections spaced 50 μm apart were stained with NeuN. Anatomical regions of interest (ROIs; including the entire cerebral cortex, cingulate and retrospenial cortex, CA1, CA2, CA3, dentate gyurs, hilus and granule cell layer; illustrated in Figure 1) were defined according to the Paxinos and Watson (2005) rat brain atlas. A grid was placed randomly over the ROI slated for counting. At random positions within the grid, as determined by the software, cells were counted within three-dimensional optical dissectors (50 μm × 50 μm × 10 μm) with a 100X objective. Within each dissector, 1 μm guard zones at the top and bottom of the section surface were excluded. Section thickness was measured regularly and averaged 12 μm for all sections analyzed, allowing for uniform antibody penetration. The average sum of the optical dissectors used was 185 for the cingulate cortex (CC) and 72 for the hippocampus (HC). Estimated totals by number weighted section thickness were obtained with StereoInvesitgator yielding a coefficient of error < 0.10. Neuronal densities were calculated by adjusting these totals by the tissue volume of the ROI investigated.
All data were normally distributed; therefore, in instances of single mean comparisons, Levene’s test for equality of variances followed by t-test for independent samples was used. In instances of multiple mean comparisons, analysis of variance (ANOVA) was used, followed by post-hoc comparison using Bonferonni’s method. Wherever possible, we used a hierarchical analysis strategy, where overall multi-way ANOVAs were run prior to one-way ANOVA followed by post-hoc testing. Alpha levels were set to 0.05, and all analyses were conducted using SPSS™ software, release 15.0 (SPSS Inc., Chicago, IL) or Statictica® (Statsoft, Inc., Tulsa, OK). All analyses were conducted by a blinded examiner, and code was not broken until analyses were completed.
We generated transgenic (Tg) AD rats on a Fischer 344 background by co-injection with APPsw and PS1ΔE9 (Figure 2A, B), each independent autosomal dominant causes of familial early-onset AD. TgF344-AD rats expressed 2.6-fold higher human holo- and secreted APPsw proteins than endogenous rat APP by N-terminal APP antibody 22C11 and 6.2-fold increased human PS1ΔE9 protein abundance vs. endogenous rat N-terminal PS1 fragment using human-selective PS1 antibody NT1 (Figure 2C, D). We began by evaluating behavioral impairment in three cohorts of WT and TgF344-AD rats at 6, 15, and 24 months of age. Neurologic screening did not reveal between-genotype differences with respect to righting response, eye blink, ear twitch, limb withdrawal in response to tactile stimuli, orienting response to olfactory and visual stimuli, startle response, or visual and tactile placing (data not shown). Yet, TgF344-AD rats displayed hyperactivity, which may result from disinhibition associated with hippocampal or cortical injury. This phenotype was operationalized as increased numbers of beam breaks and rears in an open field behavioral assay. Specifically, analysis of variance (ANOVA) revealed significant (P < 0.01 for beam breaks; P < 0.001 for rears) genotype x time effects that were confirmed by post-hoc analyses in 15 month-old TgF344-AD rats (Figure 2E). Hyperactivity was age-dependent, because overall ANOVA did not reveal significance (P > 0.05) in younger (6 month-old) TgF344-AD rats for beam breaks (data not shown) or rearing activity, although 6 month-old Tg rats reared significantly more (P < 0.05) during the first 5 min of open field testing (Figure 2F). Novel object recognition, a hippocampus-dependent measure of working memory in the rat (Wan et al., 1999), was significantly (P < .001) and almost completely impaired in older (24 month-old) Tg animals (Figure 2G).
We also assessed cognitive performance in the Barnes maze, a widely accepted test of hippocampus-dependent spatial reference learning and memory in rats (Barnes, 1979; Barnes et al., 1994). While 6 month-old WT and TgF344-AD rats learned the initial location of the escape box equally well (Figure 2H), 15 month-old Tg animals made significantly (P < 0.01) more errors during the learning phase by overall ANOVA and by post-hoc testing (*P < 0.05; Figure 2I). In the memory probe trial, this older cohort of Tg rats performed significantly (P < 0.01) worse than WT animals by overall ANOVA and by post-hoc analyses, indicating impaired spatial reference memory (*P < 0.05; Figure 2I). Further evidence of impairment came from the reversal phase of the task, where the escape box was randomly rotated to another location. Specifically, 15 month-old Tg rats made significantly (P < 0.01) more errors than WT littermates by ANOVA and by post-hoc analyses (*P < 0.05; **P < 0.01; Figure 2I). A similar pattern of significant (P < 0.05) results was observed in 24 month-old Tg rats (Figure 2I), and even the 6 month-old Tg rats trended toward significance during the reversal phase (Figure 2H). Thus, TgF344-AD rats exhibit progressive, age-dependent abnormalities in open field activity and spatial learning and memory.
We next examined cerebral β-amyloid pathology and Aβ abundance in cohorts of Tg animals at various ages (6, 15–17, and 25–27 months-old) by multiple independent strategies (see MATERIALS AND METHODS). Quantitative histological analysis of Tg rats using Aβ antibody (4G8) and Thioflavin S (ThioS) disclosed significant (P < 0.001) age-dependent β-amyloid deposition (~10–82 fold increased vs. 6 month-old animals) in the CC and HC of Tg rats (Figure 3A–D), brain areas classically associated with AD-type pathology in humans (Selkoe, 2001). A similar pattern of results was also observed in the striatum and cerebellum (data not shown), additional brain regions that commonly succomb to amyloid plaques in AD and Down’s syndrome (Braak and Braak, 1990; Braak et al., 1989; Dickson et al., 1990; Joachim et al., 1989). Deposition of Aβ in cerebral vessels, a pathological feature known as cerebral amyloid angiopathy (CAA), occurs in 86% of AD patients (Ellis et al., 1996). CAA analyses were conducted as previously described (Wyss-Coray et al., 2001; Wyss-Coray et al., 1997), and revealed copious (Figure 3A, see arrows) and significant (*P < 0.05; **P < 0.01) age-dependent pathology in Tg rat CC and HC (Figure 3E), while WT littermates were devoid of both brain parenchymal β-amyloid and CAA at any age (data not shown). CAA-like pathology also progressively accumulated in the striatum and cerebellum of Tg rats (mean CAA score ± SEM of Tg rats at 6 vs. 16 vs. 26 months-old: striatum, 0.05 ± 0.05 vs. 0.88 ± 0.07 vs. 1.47 ± 0.16; cerebellum, 0.05 ±. 0.05 vs. 1.93 ± 0.55 vs. 3.18 ± 0.43). To determine if β-amyloid deposits could be detected in TgF344-AD rats by molecular imaging, we injected them with the radiotracer [18F] 2-(1-(6-[(2-[F-18]fluoroethyl) (methyl)amino]-2-naphthyl)ethylidene) malononitrile propene (18F-FDDNP) (Agdeppa et al., 2001; Small et al., 2006) for dynamic MicroPositron Emission Tomography (MicroPET). As early as 15 months of age, TgF344-AD rats showed higher 18F-FDDNP distribution volume ratios (DVRs) in the frontal cortex (FC) region vs. WT littermate rats (Figure 3F). Fluorescence microscopy showed abundant extracellular FDDNP+ hippocampal deposits as well as apparent intraneuronal signals (Figure 3G).
To further investigate the possibility of intraneuronal Aβ, we immunostained brain sections with C-terminal Aβ isoform-specific antibodies. While Aβ1-40 was not detected in neuronal somata, TgF344-AD rats exhibited intraneuronal Aβ1-42 by confocal microscopy at 16 months of age (data not shown). Staining with A11 antibody, which has been reported to recognize Aβ oligomers (Necula et al., 2007; Yoshiike et al., 2008), revealed hippocampal pyramidal neurons that were double-positive for FDDNP (data not shown), supporting the existence of intraneuronal Aβ in Tg rat brains. Furthermore, biochemical analysis of brain homogenates from TgF344-AD rats confirmed age-dependent Aβ accumulation. Separate sandwich ELISAs for Aβ1-40 and Aβ1-42 in both detergent-insoluble and –soluble brain extracts (Johnson-wood et al., 1997; Tan et al., 2002; Town et al., 2008) revealed significant (P < 0.001) progressive increases for both fractions of Aβ1-40 and Aβ1-42 in Tg rats (Figure 4A, B). Notably, Tg rats had striking overabundance of soluble Aβ1-40 as early as 6 months of age (Figure 4B), prior to appreciable β-amyloid plaque formation (Tg rats at this age only have 0–3 plaques per brain section). Intriguingly, ratios of the more pathogenic detergent-soluble Aβ1-42 (Selkoe, 2001; Walsh et al., 2002a; Walsh et al., 2002b; Walsh et al., 2000) to Aβ1-40 progressively increased in TgF344-AD rat brains, while a similar ratio considering detergent-insoluble Aβ species revealed a corresponding decrease with age (Figure 4C). These findings raise the intriguing possibility that neurotoxic Aβ1-42 oligomers (Shankar et al., 2008; Walsh et al., 2002a), which reside in the detergent-soluble brain homogenate fraction and preferentially increase over time relative to insoluble Aβ peptides in Tg animals.
To further examine this possibilty, we performed immunohistochemical analysis using OC, a conformational antibody that recognizes soluble oligomeric fibrillar Aβ (Kayed and Glabe, 2006; Kayed et al., 2003). Consistent with this hypothesis, TgF344-AD rat brains showed age-dependent accumulation of OC immunoreactivity in both the CC and the HC (Figure 4D). Quantification of soluble (N) 82E1-oligomers by ELISA revealed strikingly greater abundance of these Aβ species in TgF344-AD rats vs. PSAPP mice (both Tg rodents were generated using the same constructs) that was statistically significant (** P < 0.01; *** P < 0.001; Figure 4E). As independent validation, WB with both Aβ carboxyl-terminal and amino-terminal specific antibodies revealed presence of putative soluble oligomeric Aβ (likely a ~22 kDa pentamer) in Tg rat brain homogenates (Figure 4F). Antibody 6E10 revealed monomeric Aβ, a putative Aβ oligomer, holo-APP, and an ~18 kDa band that was likely the β-carboxyl-terminal fragment of APP (β-CTF, which contains the Aβ domain), all of which were increased in Tg vs. WT rat brains (Figure 4G). The β-CTF identity of this ~18 kDa band was confirmed by WB with Aβ middle region-specific 4G8 antibody (Figure 4H). Finally, electron microscope (EM) analysis of β-amyloid plaques revealed fibrillar ultrastructure and numerous dystrophic neurites in close proximity to plaques (Figure 4I).
APP and PS1 mutations increase brain Aβ abundance but do not precipitate frank neurofibrillary tangles (NFTs) in Tg mouse models of cerebral amyloidosis (Duff et al., 1996; Games et al., 1995; Holcomb et al., 1998; Hsiao et al., 1996; Jankowsky et al., 2001; Mucke et al., 2000; Sturchler-pierrat et al., 1997). Some of the Tg AD mouse models do manifest hyperphosphorylated tau (Sturchler-pierrat et al., 1997; Tan et al., 1999), which may represent ‘pre-tangles’ that do not progress to NFTs. We probed for tau pathology in Tg rats by silver-impregnating brain sections using the Gallyas method (Lamy et al., 1989; Rosenwald et al., 1993). Strikingly, numerous structures were detected in close proximity to β-amyloid plaques in aged Tg rats that were reminiscent of NFTs found in AD patient brains (Figure 5A). In addition, CP13 immunostaining revealed teardrop-shaped structures morphologically consistent with NFTs in 16 month-old Tg rats (Figure 5A). Importantly, NFT-like structures were frequently observed in non-plaqued areas of CC and HC, much akin to human AD (Figure 5B).
To further understand the nature of tauopathy in Tg rats, brain homogenates from TgF344-AD and WT rats were extensively analyzed by tau biochemistry using previously described methods (Greenberg and Davies, 1990). We began by raising our own antibodies in rats against a peptide fragment of pathogenic tau containing the pS396/404 epitope (designated pTau-PADRE, see MATERIALS AND METHODS). WB for total tau in rat brain homogenates using Tau-PADRE antibody revealed six bands ~48–62 kD in size (data not shown), which likely contain six isoforms of rat tau that are similar to those present in the human (Hanes et al., 2009), and pTau-PADRE WB revealed increased immunoreactivity in aged TgF344-AD vs. WT rats (Figure 5C). Quantitative analyses disclosed elevated abundance of insoluble (crude pellet-extracted) tau in aged TgF344-AD rats, and abnormal rat tau precipitation from sarkosyl soluble-to-insoluble fractions in 6 and, more strikingly, 16 month-old Tg vs. WT rat brains (†P < 0.10, *P < 0.05, **P < 0.01; Figure 5D, E). To ensure that the pTau-PADRE antibody was specific for human pTau, a peptide neutralization experiment was performed using sections from AD patient cortex. As shown in Figure 5F, a 10-fold molar excess of blocking peptide completely abolished NFT labeling. To further cofirm tauopathy in TgF344-AD rats, a panel of well-validated pathogenic tau antibodies was utilized, and revealed a similar pattern of results (Figure 6A–F). It is noteworthy that initial tau changes at 6 months of age preceed frank β-amyloid plaque formation, and may therefore represent an early response to abnormally high concentrations of cerebral soluble Aβ species.
We further evaluated neuroinflammation in multiple age cohorts of Tg and WT rats. Immunohistochemistry for activated microglia (using Iba1 antibody) and astrocyes (by GFAP immunostaining) disclosed progressive and statistically significant (**P < 0.01; ***P< 0.005) increased microgliosis and astrogliosis burden in the CC and HC of TgF344-AD rats vs. age-matched WT littermates (Figure 7A–D). Interestingly, reactive microglia and astrocytes were elevated as early as 6 months of age in Tg vs. WT rats, prior to appreciable Aβ deposition but concurrent with elevated soluble and oligomeric Aβ species. Furthermore, microglia from aged Tg rats were hyperplasic and hypertrophied (Figure 7A, C) in close vicinity of β-amyloid plaques (Figure 7E), and often contained ThioS+ deposits near ThioS-decorated neurons (Figure 7F). We reasoned that, if microglia were actually phagocytosing ThioS+ neurons, then the microglia should also show evidence of engulfment of a bona fide neuronal marker. Indeed, IHC revealed neuronal nuclei (NeuN) deposits within the cytoplasm of Iba1+ cells (Figure 7G). If microglia were in fact clearing neuronal debris, then we may also expect neuronal loss in TgF344-AD rats.
The above results raised the possibility that neurons harboring pathogenic tau may be rendered susceptible to death. Two independent stategies were adopted to assess putative neuronal loss in Tg animals. Following immunohistochemical analysis with NeuN antibody, we 1) conducted exhaustive manual subfield counting and 2) performed stereological whole number estimates of neurons in the cerebral cortex and HC. Strikingly, there was statistically significant (***P < 0.005) cortical and hippocampal neuronal loss in Tg rats that was both progressive and frank (ranging from 23–45%) (Figure 8A–C). Progressive neuronal loss was also observed in specific subfields of the HC. At 6 months of age, cell counts (NeuN+ cells/mm3) were similar between WT and Tg rats in the granule cell layer of the dentate gyrus (mean ± SD; WT: 18810 +/− 1143, Tg: 19031 +/− 2192), the hilus of the dentate gyrus (WT: 900 +/− 135, Tg: 947 +/− 184) and the CA2+CA3 regions of the HC (WT: 3025 +/−212, Tg: 3116 +/− 363). However, at 16 and 26 months of age, there were respective 33% and 37% decreases of NeuN+ cells in the dentate gyrus (16 months; WT: 18235 +/− 2310, Tg: 12178 +/− 1220; 26 months; WT: 17175 +/− 2283, Tg: 10750 +/− 1976), 63% and 66% decreases in the hilus (16 months; WT: 820 +/− 67, Tg: 305 +/− 73; 26 months; WT: 794 +/− 109, Tg: 267 +/− 62), and 36% and 45% decreases in CA2+CA3 (16 months; WT: 2790 +/− 284, Tg: 1778 +/− 310; 26 months; WT: 2740 +/− 421, Tg: 1490 +/− 350).
Western blots of rat brain homogenates showed age-dependent decreases in NeuN abundance, supporting evidence of neuronal loss (†P < 0.10, P < 0.05; ***P < 0.005; Figure 8D). The appearance of vacuolar pathology in older Tg rats further bolstered these conclusions (Figure 8E, F). To determine if a relationship existed between cerebral Aβ species and neuronal loss in TgF344-AD rats, Pearson product-moment correlation analyses were conducted and revealed statistically significant (***P < 0.001) inverse correlations between Aβ1-42 and (N) 82E1 Aβ oligomers with numbers of NeuN+ neurons present in the CC and HC (***P < 0.005; Figure 8G). Strikingly though, strongest significance was detected when inversely correlating triton-soluble Aβ1-42 or (N) 82E1-oligomers with NeuN+ cells in both the cerebral cortex (data not shown) and HC of transgenic rats (Figure 8G). Moreover, Pearson product-moment correlation analyses also revealed statistically significant (P < 0.05) inverse correlations between numbers of NeuN+ neurons present in the CC and HC and either 4G8 (CC: r = −0.677; HC: r = −0.874) or ThioS burden (CC: r = −0.593; HC: r = −0.739) in these brain regions (data not shown).
There were additional indices of neuronal degeneration and death in TgF344-AD rats. For example, quantitative microscopy for nicked DNA by TUNEL assay (Figure 9A, B) and apoptosis by total and cleaved (active) caspase-3 (Figure 9C, D) prompted the conclusion that neurons were significantly (†, a trend of P < 0.10; *P < 0.05; ***P < 0.005) dying via apoptosis in older Tg rats. In addition, we observed Hirano bodies at both light microscopic and ultrastructural levels in the cortex and HC (Figure 9E, F). Immunohistochemical analyses of cleaved (active) caspase-3 in the dentate gyrus revealed pyknotic nuclei, further evidencing age-dependent neuronal apoptosis in Tg vs. WT rats (data not shown). Finally, numerous dystrophic neurites were observed by EM (Figure 9E). Each of these indicators of neurodegeneration were age-dependent, and tracked with progressive tauopathy and cognitive impairment in Tg animals.
Together, our data demonstrate that TgF344-AD rats manifest a complete repertoire of AD pathological features. Expression of only two mutant human transgenes, each independent causative factors for early-onset familial AD, is sufficient to precipitate the full array of AD pathological features in the rat. We interpret these results as rodent model evidence in support of the over two decades-old amyloid cascade hypothesis of AD (Hardy and Allsop, 1991; Rozemuller et al., 1989). Specifically, TgF344-AD rats develop age-dependent accumulation of cerebral Aβ that preempts tauopathy, cognitive disturbance, apoptosis, and neuronal loss. While Aβ-driven rodent models of AD also exhibit brain amyloidosis (Games et al., 1995; Holcomb et al., 1998; Hsiao et al., 1996; Lawlor et al., 2007; Leon et al., 2010; Liu et al., 2008; Oakley et al., 2006; Oddo et al., 2003), cerebral Aβ abundance in TgF344-AD rats (~60 μg/wet g of brain in aged animals) is high enough to be within the clinicopathological range of the human syndrome [~50–1000 μg/wet g of brain (Delacourte et al., 2002; Ingelsson et al., 2004; Lue et al., 1999; Steinerman et al., 2008)]. Notably, abundance of soluble oligomeric Aβ species is markedly greater in TgF344-AD rats compared with PSAPP mice harboring the same mutant human transgenes under mouse prion protein regulatory control. Further, while frank NFT pathology has been found in rodent models of AD, these models rely on the presence of mutated human tau (Filipcik et al., 2010; Oddo et al., 2003), which is not causative of AD in humans. It is noteworthy that TgF344-AD rats manifest NFT pathology independent of human tau mutations, relying solely on endogenous rat tau protein. This important aspect of the model will allow for more physiologic investigation into Aβ-mediated tauopathy. Finally, neuronal loss in other rodent models of AD is either absent entirely or quantitatively minor, or only present in a limited subset of brain areas classically affected in human AD (Calhoun et al., 1998; Oakley et al., 2006; Oddo et al., 2003; Rebeck et al., 2010). By contrast, TgF344-AD rats show consistent and extensive neuronal loss in cortical and hippocampal regions to the degree that we even observe an age-dependent trend toward decreased hemispheric brain weight accompanied by spongiform-like vacuolar pathology.
But why do TgF344-AD rats present with the full array of AD pathology while mice expressing the same mutant human transgenes do not? The answer to this question is likely owed to two elements: 1) the magnitude and nature of cerebral amyloidosis in Tg rats and 2) a rat tau proteome that is more akin to humans than to mice. Specifically, TgF344-AD rats have progressively elevated abundance of Aβ1-40, Aβ1-42, and particularly of soluble oligomeric Aβ species that are directly neurotoxic in rats (Shankar et al., 2008). Furthermore, rats have the full complement of six tau isoforms present in humans, whereas mice only harbor three of the human tau isoforms (Hanes et al., 2009; McMillan et al., 2008). These two key factors seem to create a cerebral milieu that allows for the development of Aβ-directed tauopathy and accompanying neuronal loss. However, given the complexities inherent to comparative biology across multiple species, further study is needed to definitively answer this important question.
Several studies have established oligomeric Aβ as the principle neurotoxic species (Shankar et al., 2008; Walsh et al., 2002a), and TgF344-AD rats exhibit intraneuronal Aβ and high abundance of soluble Aβ oligomers. Strikingly, soluble oligomeric Aβ species were most strongly associated with neuronal loss in Tg rats, supporting this notion. It is also worth noting that we observe apoptotic cells not only in close vicinity to amyloid plaques, but also in brain areas devoid of amyloid deposits, but where putative soluble oligomeric species are likely present. That TgF344-AD rats show clear evidence of age-dependent neuronal loss is interesting given the current debate surrounding mechanisms of neuronal loss in AD (Behl, 2000; Cotman and Anderson, 1995; Gorman, 2008; Zhu et al., 2006). Importantly, there have been reports in postmortem brain samples suggesting that apoptotic mechanisms may contribute to the disease. For example, Stadelmann and colleagues reported DNA fragmentation in AD patient brains, and observed a small population of neurons that displayed morphological characteristics of apoptosis (Stadelmann et al., 1998). Apoptosis is a dynamic and short-lived process, and it is likely that endogenous brain clearance machinery actively removes dead cells and consequently reduces the detection threshold. While TgF344-AD rats evidence neuronal apoptosis, it remains possible that apoptosis is not the sole mechanism for neuronal loss in these animals or in human AD.
Similar to the findings of Su and colleagues in human AD patients, co-localization of pathologic tau (as indicated by CP13 or PHF1 positivity) with TUNEL+ cells in close proximity to amyloid plaques suggests temporal and spatial proximity of tauopathy and neuronal death (Su et al., 2001). These observations fall in line with recent data suggesting a link amongst caspase-3 activation, tauopathy, and neuronal apoptosis (de Calignon et al., 2010; Ittner et al., 2010). Such findings raise the possibility that neuronal death may be consequent upon tau pathology driven by overly abundant cerebral Aβ. It is also interesting that increased phosphorylation of tau occurs as early as 6 months of age in Tg rats, when soluble Aβ is elevated, but prior to appreciable Aβ deposition. When taken together with histologic and behavioral observations, these results suggest that early elevation in soluble Aβ species promotes abnormal tau phosphorylation, and then continued bombardment with multiple forms of Aβ further perpetrates tauopathy and neuronal loss, culminating in cognitive impairment.
In conclusion, TgF344-AD rats manifest age-dependent cerebral amyloidosis, tauopathy, gliosis, apoptotic loss of neurons, and cognitive disturbance. Cerebral amyloidosis seems to be the driving force for tauopathy, neuronal loss, and interruption of learning and memory in this model, supporting the 20 year-old amyloid cascade hypothesis. Given that TgF344-AD rats capture the full array of AD pathology, this new model will be a critical tool for the neuroscience community to enable future studies in basic and translational AD research.
This work was supported in part by the NIH/NIA (5R00AG029726-03, 5R00AG029726-04, and 3R00AG029726-04S1, to T.T.) and the NIH/NINDS (1R01NS076794-01), an Investigator Initiated Research Grant (IIRG-05-14993, to T.T.) from the Alzheimer’s Association, intramural funds from the NIMH (to R.C.), and an NIH/NIA grant (1R21AG033394-01, to R.C.). T.T. is a recipient of an Alzheimer’s Association Zenith Fellows Award (ZEN-10-174633), an Ellison Medical Foundation/American Federation for Aging Research Mid-Career Award in Aging Research (M11472), and is the inaugural holder of the Ben Winters Endowed Chair in Regenerative Medicine. R.C. holds the Steven C. Gordon Family Foundation Chair in Memory Disorders. The authors would like to thank Chad Dickey (University of South Florida), Stephen Johnson (Cedars-Sinai Medical Center), and Dominique Jodry (Cedars-Sinai Medical Center), for technical assistance and advice. We are also grateful to Jun Tan (University of South Florida) and Dennis Dickson (Mayo Clinic Jacksonville) for helpful discussion and for assistance with interpreting data. We also thank the Department of Pathology at Cedars-Sinai Medical Center for assistance with EM acquisition and interpretation.
AUTHOR CONTRIBUTIONSR.M.C., K.R-Z., and T.T. designed all experiments; K.R-Z., T.M.W., A.R., D.G., I.S., Y.B., J.J.B., V.K., J.B., S.B., R.N.P., and T.T. carried out experiments. V.V., C.G.G., H.D., and M.G.A. contributed new analytical tools and P.R. assisted with microscopy interpretation; R.M.C., K.R-Z., C.A.S., and T.T. conducted statistical analysis of the data; R.M.C., K.R-Z., and T.T. supervised the project; K.R-Z. and T.T. wrote the manuscript and R.M.C., K.R-Z., T.M.W., and T.T. edited the manuscript.