Shiyong Peng, Diego J. Garzon, Monica Marchese, William Klein, Stephen D. Ginsberg, Beverly M. Francis, Howard T. J. Mount, Elliott J. Mufson, Ahmad Salehi and Margaret Fahnestock
Journal of Neuroscience 22 July 2009, 29 (29) 9321-9329; DOI: https://doi.org/10.1523/JNEUROSCI.4736-08.2009
Downregulation of brain-derived neurotrophic factor (BDNF) in the cortex occurs early in the progression of Alzheimer’s disease (AD). Since BDNF plays a critical role in neuronal survival, synaptic plasticity, and memory, BDNF reduction may contribute to synaptic and cellular loss and memory deficits characteristic of AD. In vitro evidence suggests that amyloid-β (Aβ) contributes to BDNF downregulation in AD, but the specific Aβ aggregation state responsible for this downregulation in vivo is unknown. In the present study, we examined cortical levels of BDNF mRNA in three different transgenic AD mouse models harboring mutations in APP resulting in Aβ overproduction, and in a genetic mouse model of Down syndrome. Two of the three Aβ transgenic strains (APPNLh and TgCRND8) exhibited significantly decreased cortical BDNF mRNA levels compared with wild-type mice, whereas neither the other strain (APPswe/PS-1) nor the Down syndrome mouse model (Ts65Dn) was affected. Only APPNLh and TgCRND8 mice expressed high Aβ42/Aβ40 ratios and larger SDS-stable Aβ oligomers (∼115 kDa). TgCRND8 mice exhibited downregulation of BDNF transcripts III and IV; transcript IV is also downregulated in AD. Furthermore, in all transgenic mouse strains, there was a correlation between levels of large oligomers, Aβ42/Aβ40, and severity of BDNF decrease. These data show that the amount and species of Aβ vary among transgenic mouse models of AD and are negatively correlated with BDNF levels. These findings also suggest that the effect of Aβ on decreased BDNF expression is specific to the aggregation state of Aβ and is dependent on large oligomers.
Alzheimer’s disease (AD) is the predominant form of dementia in the elderly. Pathological features of AD include the presence of amyloid plaques, soluble amyloid-β (Aβ) oligomers, neurofibrillary tangles, neuritic dystrophy, synaptic loss, and eventual neurodegeneration (Mirra et al., 1991; Hyman, 1997). Mutations in the amyloid precursor protein (APP) and presenilin 1 and 2 (PS-1and PS-2) genes cause familial AD (FAD), and all identified pathogenetic mutations lead to overproduction of amyloid-β (Aβ) or its most fibrillogenic form (Aβ42) (Selkoe, 1994). Assemblies of the Aβ fragment are neurotoxic in vitro (Yankner, 2000), cause synaptic degeneration (Lacor et al., 2007; Walsh and Selkoe, 2007), and interfere with long term potentiation, a form of memory consolidation (Lambert et al., 1998; Walsh et al., 2002). However, the magnitude of Aβ toxicity in vivo remains unclear. In part, this controversy appears to be explained by the observations that distinct aggregation states of Aβ display differential toxic properties. In fact, there exist different soluble Aβ aggregates exhibiting a broad range of sizes (Klein, 2002; LeVine, 2004), and specifically which soluble oligomeric aggregates of Aβ could be the most toxic forms and what is the downstream mechanism of Aβ neurotoxicity are still under debate (Gong et al., 2003; Lesné et al., 2006; Townsend et al., 2006).
Recent evidence suggests Aβ-associated neurotoxicity may be a consequence of brain-derived neurotrophic factor (BDNF) deficiency. Several studies indicate that the cortex and hippocampus, areas of the brain associated with learning and memory, exhibit both extensive amyloid pathology and decreased levels of BDNF in AD (Phillips et al., 1991; Connor et al., 1997; Ferrer et al., 1999; Hock et al., 2000; Holsinger et al., 2000; Garzon et al., 2002; Peng et al., 2005). Interestingly, BDNF protein levels are significantly decreased in preclinical and early stages of AD, and this reduction correlates with clinical neuropsychological scores (Peng et al., 2005). Since BDNF is critical for neuronal survival and function (Siegel and Chauhan, 2000; Mufson et al., 2007) and for synaptic plasticity and learning and memory (Korte et al., 1995; Patterson et al., 1996; Lu, 2003; Bramham and Messaoudi, 2005; Nagahara et al., 2009), which are compromised in AD, it is important to understand which Aβ species drive the reduction of BDNF in AD. In vitro data demonstrate that soluble forms of Aβ decrease BDNF mRNA expression and compromise BDNF intracellular signaling in both primary rat neurons and human neuroblastoma cells (Tong et al., 2001, 2004; Garzon and Fahnestock, 2007). Therefore, amyloid-induced neurodegeneration may be a consequence of reduced BDNF. However, whether Aβ assemblies downregulate BDNF in vivo, and which Aβ assembly state is responsible for BDNF downregulation, have not been elucidated. In this study, we measured levels of BDNF mRNA and Aβ42/Aβ40 ratios and characterized the state of Aβ in three different transgenic mouse models of AD (Table 1/See link) containing mutations in APP, two of these in combination with PS-1 mutations, and in a mouse model of Down syndrome (segmental trisomy 16) containing an additional copy of App.
Materials and Methods
All animal experiments were performed in accordance with the Canadian Council on Animal Care Guide for the Care and Use of Laboratory Animals. Table 1 (see link) summarizes the characteristics of each transgenic mouse line. Construction of APPNLh, PS-1P264L, and APPNLh/PS-1P264L mice (Cephalon) has been described previously (Reaume et al., 1996; Flood et al., 2002). Briefly, the Aβ portion of the App gene has been “humanized” and the Swedish mutation (K670N/M671L) inserted (Reaume et al., 1996). A single amino acid substitution in PS-1 (P264L) has been knocked in. Unlike the other mice in this study, these mice express APP and PS-1 mRNA at normal levels under control of their endogenous promoters. APPNLh/PS-1P264L mice demonstrate elevated Aβ levels and plaque deposition at 6 months of age (Flood et al., 2002), whereas plaque deposition was not detected in either the APPNLh or the PS-1P264L single mutant mice up to 22 months (Flood et al., 2002). Mice were killed at 15 and 18 months of age. After removal of the cortical hemispheres, tissue was flash frozen in liquid nitrogen and stored at −80°C.
The construction of the APPswe/PS-1M146V mice is well documented (Kurt et al., 2001; Sadowski et al., 2004). These double transgenic mice are a cross between Tg2576 mice expressing the Swedish APP mutation K670N and M671L (Hsiao et al., 1996) and H8.9 mice expressing the mutant PS-1 M146V (McGowan et al., 1999). These mice overexpress mutated APP and PS-1 under control of the hamster prion protein (PrP) and PDGF promoters, respectively, and have extensive plaque deposition throughout the cortex and hippocampus by 6 months of age (McGowan et al., 1999). The age of the mice at killing was between 19 and 24 months.
TgCRND8 mice contain the human APP695 cDNA cassette with a double APP mutation (Swedish and Indiana V717F) governed by the Syrian hamster PrP promoter. Detailed construction of this mouse strain has been described previously (Chishti et al., 2001). These mice overexpress APP and produce elevated Aβ42 levels at 4 weeks, with plaque deposition by 3 months (Chishti et al., 2001). The age of the mice at killing was 11.4 months.
Ts65Dn mice display triplication of a number of genes orthologous to the human Down syndrome locus, including the App gene, on an additional copy of a segment of chromosome 16 (MMU16) translocated close to the centromere of chromosome 17 (Reeves et al., 1995). These mice serve as a genetic mouse model for Down syndrome and exhibit increased levels of App. The age at killing was 18–21 months.
RNA ISOLATION, DNASE TREATMENT, REVERSE TRANSCRIPTION, AND ABSOLUTE QUANTITATIVE PCR.
Frozen cortical samples (100 mg wet weight) were sonicated with a Sonic dismembrator model 100 (Fisher Scientific) in 1 ml of Trizol (Invitrogen), and RNA was isolated with RNeasy spin columns (Qiagen). The procedure for RNA isolation was followed as specified by Qiagen. RNA purity was confirmed by spectrophotometry (A260/A280 > 1.7), and RNA integrity was visualized by agarose gel electrophoresis.
One microgram of RNA was treated with 2 U of Turbo DNA-free (Ambion). For reverse transcription (RT), Invitrogen’s protocol and reagents for Superscript II were used. The final volume of 20 μl contained 250 ng of random primers, 0.5 mm deoxynucleotide triphosphates (0.5 mm each of dATP, dTTP, dCTP, and dGTP), 1× first-strand buffer, 0.05 mm dithiothreitol, 2 U of RNaseOUT, and 200 U of Superscript II RT (Moloney murine leukemia virus reverse transcriptase). As a control, 1 μg of RNA was treated according to the same protocol with addition of water instead of the RT enzyme (“no-RT” control).
Real-time PCR was performed in a Stratagene MX3000P using the DNA binding dye SYBR Green (Platinum SYBR Green qPCR SuperMix UDG, Invitrogen). The 20 μl PCR mix contained 1× qPCR SuperMix, forward and reverse primers, 30 nm ROX reference dye (Stratagene), and cDNA from 50 ng of RNA or reference standard for absolute quantification. Forward and reverse primers were used at 300 nm for all targets except for β-actin, for which forward and reverse primers were used at 150 nm. Forward and reverse primers were as follows: total mouse BDNF, 5′ CAG CGG CAG ATA AAA AGA and 5′ TCA GTT GGC CTT TGG ATA; exon I, 5′ AGT CTC CAG GAC AGC AAA GC and 5′ TCG TCA GAC CTC TCG AAC CT; exon III, 5′ CTT CCA TCC CTC CCT CAT TT and 5′ CTT CCC TTG AGA AGC AGG AG; exon IV, 5′ AGA GCA GCT GCC TTG ATG TT and 5′ TCG TCA GAC CTC TCG AAC CT; exon VI, 5′ GCT TTG TGT GGA CCC TGA GT and 5′ TTC GAT GAC GTG CTC AAA AG; and β-actin, 5′ CTG ACA GGA TGC AGA AGG and 5′ GAG TAC TTG CGC TCA GGA. A “no-template” control was added, which consisted of all the reagents listed above for real-time PCR, except the cDNA template was replaced with water. Levels of BDNF and β-actin were determined using absolute quantification (Garzon and Fahnestock, 2007). Standards for total BDNF, BDNF transcripts, and β-actin were obtained from purified PCR products using the primers listed above. Only experiments with an R2 of >0.997 and a PCR efficiency of between 90 and 100% were used for analysis. All unknowns, no-RT, and no-template controls were run in triplicate. The following thermal profile was used for all measurements: 2 min at 50°C, 2 min at 95°C followed by 40 cycles of 95°C for 15 s, 58°C for 30 s, and 72°C for 30 s. After PCR, a dissociation curve verified that no secondary products had formed.
TISSUE HOMOGENATES FOR WESTERN BLOTTING.
A small amount of tissue (from 0.012 to 0.05 g) was homogenized with a Sonic dismembrator model 100 (Fisher Scientific) directly from the frozen state in 10 μl per mg tissue of homogenization buffer (HB) (0.05 M Tris, pH 7.5, 10 mm EDTA, 0.5% Tween 20, 1 μg/ml leupeptin, 2 μg/ml aprotinin and pepstatin, and 100 μg/ml PMSF) on ice. The homogenates were incubated for 5–10 min on ice and then centrifuged at 12,000 × g for 15 min at 4°C. Supernatants were transferred into autoclaved prechilled 1.5 ml tubes and stored at −80°C for use. Total protein concentrations for all homogenates were assayed with a DC protein assay kit (Bio-Rad).
Calcium treatment can partially separate different forms of Aβ assemblies and provides useful information about species that are present (Isaacs et al., 2006). A slight modification of the published method was applied to homogenates from an APPNLh/ PS-1P264L mouse. Briefly, 20 μl of 20 mmcalcium solution (CaCl2 in HB without EDTA) was added to 20 μl (total 220 μg) of homogenized protein. A parallel sample of homogenate was diluted with 20 μl of HB instead. After a 24 h incubation at 37°C, pellets were spun down at 1500 × g for 3 min and resuspended in 1× loading buffer (0.06 M Tris–HCl, pH 6.8, 8% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.001% bromophenol blue), and both pellets and supernatants were examined by Western blotting.
WESTERN BLOTTING FOR AΒ.
Protein homogenates (30 μg) from transgenic and Ts65Dn mice were separated on 12% SDS–polyacrylamide gels at 120 V for 70 min. Proteins were transferred onto PVDF membranes in transfer buffer [25 mm Tris, 192 mm glycine, 20% (v/v) methanol] for 2 h at 110 V at 4°C and blocked for 1 h at room temperature in Tris-buffered saline–Tween 20 (TBS-T) [50 mm Tris, pH 8.0, 133 mm NaCl, 0.2% (v/v) Tween 20] with 10% (w/v) Carnation nonfat milk powder. The blots were incubated with a 1:3000 dilution (0.34 μg/ml) of 6E10 (monoclonal anti-Aβ antibody, Signet Laboratories), which recognizes all species of Aβ aggregates (Lesné et al., 2006), overnight at 4°C in TBS-T. After being washed in TBS-T, membranes were incubated in a 1:5000 dilution of HRP-conjugated sheep anti-mouse IgG secondary antibodies (GE Healthcare) in TBS-T with 5% nonfat milk powder for 1 h at room temperature. Finally, a chemiluminescence system (ECL, GE Healthcare), followed by exposure to CL-XPosure x-ray film (Thermo Fisher Scientific), was used to detect immunoreactive protein. The same blot was reprobed with anti-β-actin as the loading control after 1 or 2 d, allowing for the decay of anti-Aβ signals. To characterize Aβ assemblies, Western blots were cut into thirds along the molecular weight marker lanes (Fermentas) and incubated with a 1:2000 dilution of either the amyloid-derived diffusible ligand (ADDL)-selective antibody NU-2 (Lambert et al., 2007), the APP-specific antibody 22C11 (Millipore), or 6E10. Blocking buffer (Li-Cor Biosciences) mixed 1:1 with PBS plus 0.05% Tween 20 was substituted for TBS-T–milk powder. The secondary antibody was IRDye 800CW goat anti-mouse (Li-Cor Biosciences), used at a dilution of 1:8000 and detected using Odyssey infrared system version 1.2 (Li-Cor Biosciences). All Westerns were repeated in three independent experiments except for Ts65Dn mice, which were analyzed in two separate Western blotting experiments.
ELISA FOR AΒ42 AND AΒ40.
Homogenates were appropriately diluted to equal protein concentrations with sample washing buffer (supplied from the kit) and then assayed for total Aβ40and Aβ42 using a commercially available sandwich-type ELISA (CRP). This ELISA had a detection limit of 12 pg/well for Aβ40 and Aβ42. The values for Aβ40 and Aβ42 were calculated as picograms per milliliter.
QUANTITATIVE AND STATISTICAL ANALYSIS.
Results were obtained as copies per nanogram of total RNA by use of Stratagene MxPro software and were normalized as a ratio of BDNF or transcript/β-actin. The volumes of the large-oligomer band and its corresponding β-actin band were determined by densitometry of films using an HP Scanjet Scanner (Hewlett-Packard Development) and Scion Image Beta 4.01 acquisition and analysis software (Scion). Means of triplicates were used for statistical analysis by one-way ANOVA with post hoc Tukey test for pairwise group comparisons or unpaired Student’s t tests where indicated, with 95% confidence interval (SPSS version 14 software, SPSS). Correlations between large oligomers or Aβ42/Aβ40 ratio and BDNF mRNA levels were assessed with Kruskal–Wallis test and Spearman rank correlation. The level of statistical significance was set at 0.05.
BDNF mRNA levels are reduced in some but not all mouse strains
mRNA levels for the housekeeping gene β-actin in all four experimental mouse strains showed no difference from wild-type controls (p > 0.05, data not shown) and served to normalize levels of BDNF mRNA. Specifically, BDNF mRNA levels were expressed as a ratio of copies of BDNF mRNA/copies of β-actin mRNA. BDNF mRNA levels from these four strains of wild-type mice did not differ from one another (p = 0.95, one-way ANOVA).
BDNF mRNA levels were determined in transgenic mice at ages consistent with heavy cortical plaque load, comparable with late-stage AD, in all three strains. RT-PCR analysis of frontoparietal cortex obtained from homozygous APPNLh mice containing only the Swedish APP mutation exhibited a significant 55% decrease in BDNF mRNA levels compared with wild-type littermate mice (p = 0.020) (Fig. 1). APPNLh/PS-1P264L mice demonstrated a significant 60% decrease in BDNF mRNA levels (p = 0.006) (Fig. 1). PS-1P264L mice with the PS-1 mutation alone exhibited intermediate BDNF mRNA levels, although there was no significant difference between BDNF mRNA levels in PS-1P264L and either APPNLh/PS-1P264L (p = 0.192) or control (p = 0.125) (Fig. 1) mice. Consistent with these results, TgCRND8 mice expressing an APP double mutation (Swedish and Indiana) exhibited a significant 35% decrease in BDNF mRNA levels compared with wild-type littermates (p = 0.018) (Fig. 1). In contrast, neither APPswe/PS-1M146V mice nor PS-1M146V mice showed any change in BDNF mRNA levels compared with control mice (p = 0.364) (Fig. 1). However, as in the PS-1P264L strain, BDNF mRNA levels in the PS-1M146V mice showed a trend toward reduction.
Ts65Dn mice contain an additional copy of ∼130 genes including the Appgene. They show significant atrophy and degeneration of basal forebrain cholinergic neurons that are rescued by deleting the extra copy of App(Salehi et al., 2006). These mice exhibit elevated levels of App in the absence of any significant Aβ increase (Reeves et al., 1995; Holtzman et al., 1996). Ts65Dn mice showed no difference in BDNF mRNA levels compared with 2N littermates (p = 0.177) (Fig. 1), demonstrating that overproduction of Aβ, but not App, is involved in BDNF gene downregulation.
Characterization of Aβ assemblies in mouse brain homogenates
To further examine which species of Aβ plays a major role in BDNF downregulation in these transgenic mouse models of AD, we used Western blotting to examine cortical homogenates. We found a high-molecular-weight (∼115 kDa) Aβ assembly in APPNLh/PS-1P264L and TgCRND8 mice (Fig. 2A) that totally disappeared from the soluble fraction after incubation in 10 mmCa2+, as previously described for protofibrils (Isaacs et al., 2006). The insoluble fraction (pellet) contained increased levels of species with molecular weights of >150 kDa, possibly fibrillar Aβ, and several distinct bands at lower molecular weights, including small amounts of the ∼115 kDa assembly.
To further characterize the ∼115 kDa Aβ species recognized by antibody 6E10 (Fig. 2B), brain tissue homogenates from each strain were tested by Western blotting with an antibody raised against ADDLs, NU-2 (Lambert et al., 1998, 2007). Extremely strong Aβ* signals (Lesné et al., 2006) (56 kDa 12-mers, as indicated in Fig. 2) and other low-molecular-weight Aβ soluble oligomers (<80 kDa) were found in all strains, but the higher-molecular-weight soluble assemblies at a molecular weight of ∼115 kDa were strongly evident only in APPNLh and TgCRND8 strains (Fig. 2B, NU-2). NU-2 recognized two different bands in APPNLh and TgCRND8 transgenic mice (>110 kDa) (Fig. 2B), including the high-molecular-weight band detected by 6E10. To demonstrate that the high-molecular-weight signal did not originate from soluble APP, the APP-specific antibody 22C11 detected soluble APP in all strains of mice at a molecular weight of ∼110 kDa (Fig. 2B, 22C11).
Quantification of the ∼115 kDa oligomer in transgenic and trisomic mice
Semiquantitative Western blotting in these mouse models showed that the amounts of the ∼115 kDa oligomer were different in TgCRND8 and APPNLh/ PS-1P264L mice compared with APPswe/PS-1M146V and Ts65Dn mice. These experiments revealed virtually none of this larger oligomer in APPswe/PS-1M146V and Ts65Dn mice, but high amounts in both TgCRND8 and APPNLh/PS-1P264L mice (Fig. 3A). Interestingly, strong ∼115 kDa oligomer bands were apparent for three of the five APPswe/PS-1M146V mice (data not shown). There were no large oligomers in Ts65Dn mice (Fig. 3D) or in any of the corresponding wild-type control mice examined (Fig. 3C). β-Actin did not vary between strains (Fig. 3B).
Correlation between BDNF mRNA and ∼115 kDa oligomer levels
Levels of the ∼115 kDa oligomer in each sample were quantified by densitometry and compared with the corresponding BDNF mRNA levels determined by real-time quantitative RT-PCR. Results showed a strong relationship of this species to lower BDNF mRNA levels (p = 0.005), although there were large variations of BDNF mRNA levels in APPswe/PS-1M146V mice. These four strains of mice were divided into two groups: (1) “decreased BDNF” (TgCRND8 and APPNLh/PS-1P264L mice) and (2) “unchanged BDNF” (APPswe/PS-1M146V and Ts65Dn). Comparing these two groups for ∼115 kDa oligomer levels, we found that the “decreased BDNF” group had significantly higher levels of this oligomer than did the “unchanged BDNF” group (p < 0.01), with no difference in β-actin levels between these two groups. Interestingly, among the five APPswe/PS-1M146V mice examined, mutant mice with a strong oligomer signal showed significantly lower BDNF expression than did the mice with fewer large oligomers (p = 0.049), suggesting that formation of this ∼115 kDa oligomer plays a major role in BDNF downregulation in these transgenic mice.
Correlation between BDNF mRNA and Aβ40 and Aβ42
Since the assembly state of Aβ is dependent on the amount and composition of Aβ (Aβ40 and/or Aβ42) (Bitan et al., 2003) and the ratio of Aβ42/Aβ40 is a better indication of AD pathology than either Aβ40 or Aβ42 alone (Lewczuk et al., 2004), we performed ELISAs for Aβ40 and Aβ42 in cortical homogenates from each mouse. This analysis revealed that Aβ42, but not Aβ40, was elevated in mice with high levels of ∼115 kDa oligomer expression compared with those with low levels. Aβ42 levels were threefold higher in mice with the larger oligomer than in mice not expressing them (p = 0.04), but Aβ40 was found at comparable levels in all of the transgenic mice examined (p = 0.46). Furthermore, Aβ42, but not Aβ40, was significantly higher in “decreased BDNF” than in “unchanged BDNF” mice (p = 0.03, data not shown). Notably, the ratio of Aβ42/Aβ40 was highly correlated with decreased BDNF mRNA levels in these mice (r = −0.54, p = 0.045) (Fig. 4).
Transcripts III and IV are responsible for total BDNF mRNA decrease in mouse brain
BDNF gene expression is controlled by multiple promoters regulating production of >10 different transcripts (Pruunsild et al., 2007). Transcripts I, II, III, and V (now known as transcripts I, II, IV, and VI) are downregulated in AD cortical tissue (Garzon et al., 2002). Transcripts I, III, IV, and VI are expressed in cortical tissues of mouse brain (Aid et al., 2007). These transcripts are highly homologous to their counterparts in human beings and are regulated by similar mechanisms; for example, transcript IV is the most highly expressed transcript in both mouse and human cortical tissue (Aid et al., 2007; Garzon and Fahnestock, 2007). The exception is mouse transcript III, which exhibits only 60% homology to human transcript III; the latter is not expressed in human cortex (Garzon and Fahnestock, 2007). In TgCRND8 mice, transcript III was decreased by 33% (p = 0.02) and transcript IV was decreased by 29% (p = 0.04) compared with wild-type controls (Fig. 5). Transcript II is not expressed in cortical tissues of mouse brain (Aid et al., 2007). Although transcripts I and VI are downregulated in AD, the corresponding counterparts in mouse showed no significant decrease compared with their age-matched wild-type littermates (p > 0.05) (Fig. 5.).
The amyloid cascade hypothesis postulates Aβ overproduction as the initial insult in AD (Selkoe, 1994). However, although soluble Aβ is now thought to be the toxic species (Hardy and Selkoe, 2002), which of the soluble Aβ aggregates from dimers to high-molecular-weight oligomers and protofibrils is responsible is a matter of debate (Caughey and Lansbury, 2003). One downstream effect of Aβ overexpression is decreased BDNF levels, which may lead to neuronal and synaptic dysfunction and eventual neurodegeneration. BDNF is required for survival and function of hippocampal, cortical, basal forebrain, and entorhinal cortex neurons, all areas of the brain affected by AD (Knusel et al., 1992; Ghosh et al., 1994; Lowenstein and Arsenault, 1996; Ando et al., 2002). BDNF is also an important mediator of synaptic plasticity (Kang and Schuman, 1996; McAllister et al., 1999; Lu, 2003; Bramham and Messaoudi, 2005); heterozygous BDNF knock-out mice, which exhibit decreases in BDNF comparable to those of subjects with AD, demonstrate defective long-term potentiation which can be rescued by BDNF administration (Korte et al., 1995; Patterson et al., 1996). Learning and memory deficits exhibited by transgenic mouse models of AD can also be rescued by BDNF delivery (Nagahara et al., 2009). In this study, we found a positive correlation between decreased levels of BDNF mRNA, the ratio of Aβ42/Aβ40, and the concentration of ∼115 kDa oligomers produced in three transgenic AD mouse strains. Only transgenic mice expressing high Aβ42/Aβ40 ratios and this larger oligomeric Aβ species demonstrated significantly decreased total BDNF mRNA compared with wild-type controls.
We compared the reaction profiles of three monoclonal antibodies, NU-2 (ADDL specific), 6E10 (reacts with Aβ in all conformations), and 22C11 (APP specific), with mouse cortical homogenates. NU-2 strongly recognized 12-mers (Aβ*) in all four strains of AD and Down syndrome mice, demonstrating that these small Aβ oligomers are not responsible for the difference in BDNF levels between these mouse strains. 22C11 strongly recognized a 100 kDa band (soluble APP) in all four strains of AD and Down syndrome mice, demonstrating that the high-molecular-weight band specific to APPNLh and TgCRND8 transgenic mice is not APP. However, NU-2 and 6E10 detected strong signals for the ∼115 kDa oligomer in APPNLh and TgCRND8 transgenic mice but very weak signals in both APPswe/PS-1 and Ts65Dn mice, implicating this high-molecular-weight species in BDNF downregulation. Precipitation experiments (Isaacs et al., 2006) showed that these species were unstable in high calcium, a property described earlier for preparations enriched in synthetic protofibrils. The structure of the ∼115 kDa oligomer has not yet been determined as protofibrillar, however, and nonfibrillar Aβ oligomers of this size have been described (Klein, 2002; LeVine, 2004). Similar high-molecular-weight Aβ assemblies have been shown to induce cognitive and memory defects in TgCRND8 mice (McLaurin et al., 2006), suggesting that these larger oligomeric forms of Aβ are toxic to the CNS.
Aβ42/Aβ40 ratios are strongly correlated with decreased BDNF levels in the present report. The rate of amyloid deposition, as well as the levels of both Aβ40 and Aβ42, vary considerably depending on the APP mutation(s) and the promoters thereof (Rockenstein et al., 2003). For example, the Swedish mutation, located at the N terminus of APP, favors β-secretase action resulting in a six- to eightfold increase in Aβ, consisting of both the Aβ40 and Aβ42 fragments (Citron et al., 1992; Cai et al., 1993; Haass et al., 1995). However, the Indiana mutation flanking the γ-secretase site does not increase levels of total Aβ but shifts generation of amyloid to the longer, fibrillogenic Aβ42 fragment (Clark and Goate, 1993; Suzuki et al., 1994). The Aβ42fragment nucleates rapidly, is more fibrillogenic than Aβ40, and is the initiation component of amyloid deposits (Burdick et al., 1992; Jarrett et al., 1993; Roher et al., 1993). Consistent with this literature, we showed that Aβ42/Aβ40ratios are higher in TgCRND8 and APPNLh mice than in the other strains. Although Aβ40 levels were detectable but not significantly different in all three strains of AD mice, Aβ42 was detectable only in those mice expressing ∼115 kDa oligomers. These results further support Aβ42 as a key triggering factor in the pathogenesis of FAD (Siman et al., 2000).
Studies of the PS-1 mutation report a 1.5- to 2-fold elevation of the Aβ42/Aβ40ratio (Citron et al., 1992; Cai et al., 1993; Haass et al., 1995; Borchelt et al., 1997). In addition, the nature of the amyloid-β produced in mutated PS-1mice differs from that in transgenic APP mice. Transgenic APP mice contain the human Aβ sequence, which differs from mouse Aβ by three amino acids and is more fibrillogenic than mouse Aβ (Dyrks et al., 1993; Reaume et al., 1996). Mice expressing only the PS-1 mutation, however, contain the mouse App gene, resulting in production of mouse Aβ. Both PS-1 P264L and PS-1M146V transgenic mice demonstrate a trend toward decreased BDNF mRNA levels but no significant change in BDNF mRNA levels compared with controls. Therefore, the levels and species of mouse Aβ42 produced by mutated PS-1 are neither sufficient nor fibrillogenic enough to significantly compromise BDNF mRNA expression.
Overproduction of App in Ts65Dn mice caused no change in BDNF mRNA expression compared with normal disomic mice. Although the Ts65Dn mice contain an extra copy of the App gene, they do not display age-related amyloid plaque deposition (Reeves et al., 1995). Moreover, these Down syndrome mice exhibit no increase in Aβ compared with controls (Holtzman et al., 1996) and less formation of the ∼115 kDa oligomer. In contrast, APPNLhand TgCRND8 mice contain mutations in APP resulting in Aβ overproduction and aggregation, as well as early learning and memory deficits (Chishti et al., 2001; Chang et al., 2006). Furthermore, TgCRND8 mice exhibit neurodegenerative changes as well (Bellucci et al., 2006). Both these transgenic mouse strains exhibit extremely high levels of ∼115 kDa oligomers and downregulation of cortical BDNF mRNA, supporting a role for Aβ aggregation in compromised BDNF levels and deficits in neuronal function in AD.
The BDNF gene is complex, consisting of unique 5′-untranslated exons controlled by multiple promoters, each individually spliced to a common 3′ BDNF protein coding region (Aid et al., 2007; Pruunsild et al., 2007). These transcripts are differentially regulated in a tissue-specific, developmentally specific, and insult-specific manner (Metsis et al., 1993; Timmusk et al., 1995). Information on regulation of specific BDNF transcripts can provide clues to the mechanisms of regulation. Our previous study (Garzon et al., 2002) determined that in AD cortex, BDNF transcription is decreased specifically via downregulation of transcripts I, II, III, and V (transcripts III and V are now transcripts IV and VI in the new nomenclature) (Pruunsild et al., 2007). Furthermore, we demonstrated that oligomeric Aβ downregulates transcript IV in human neuroblastoma cells in culture (Garzon and Fahnestock, 2007). In that study, the aggregated Aβ consisted of a mixture of oligomers including dimers, trimers, and an assortment of material of from ∼50 kDa to >100 kDa. We did not attempt to determine which of those species was responsible for BDNF downregulation in vitro. In the current study, we show that high levels of soluble ∼115 kDa oligomeric Aβ are responsible for downregulation of BDNF transcripts III and IV in TgCRND8 mice compared with wild-type controls. Transcript III, which is regulated by methylation in mouse brain (Aid et al., 2007), is not expressed in human cortex (Garzon and Fahnestock, 2007) and exhibits only 60% homology between mouse and human. Transcript IV, however, is highly conserved between mouse and human, is the most highly expressed transcript in human cortex, and is regulated by activity-dependent CREB phosphorylation (Shieh et al., 1998; Tao et al., 1998; Tong et al., 2001, 2004; Garzon and Fahnestock, 2007). Decreased CREB phosphorylation is a well documented phenomenon in AD and its mouse models (Silva et al., 1998; Yamamoto-Sasaki et al., 1999) and provides clues to further understanding the mechanism of large Aβ oligomer-mediated toxicity.
In summary, we demonstrate that transgenic mouse strains expressing high Aβ42/Aβ40 ratios and ∼115 kDa SDS-stable Aβ oligomers exhibit significantly decreased cortical BDNF mRNA levels. These findings suggest that the effect of Aβ on decreased BDNF expression is specific to the aggregation state of Aβ and is dependent on large oligomer and/or protofibril formation.