|Home | About | Journals | Submit | Contact Us | Français|
Although Aβ peptides are causative agents in Alzheimer's disease (AD), the underlying mechanisms are still elusive. We report that Aβ42 induces a translational block by activating AMPK, thereby inhibiting the mTOR pathway. This translational block leads to widespread ER stress, which activates JNK3. JNK3 in turn phosphorylates APP at T668, thereby facilitating its endocytosis and subsequent processing. In support, pharmacologically blocking translation results in a significant increase in Aβ42 in a JNK3-dependent manner. Thus, JNK3 activation, which is increased in human AD cases and a familial AD (FAD) mouse model, is integral to perpetuating Aβ42 production. Concomitantly, deletion of JNK3 from FAD mice results in a dramatic reduction in Aβ42 levels and overall plaque loads and increased neuronal number and improved cognition. This reveals AD as a metabolic disease that is under tight control by JNK3.
Aβ peptide accumulation is a hallmark of Alzheimer's disease (AD), being released from neurons via extracellular and subsequent intramembrane cleavage reactions of the amyloid precursor protein (APP) (Tanzi and Bertram, 2005). Recent findings suggest that soluble oligomeric are the pathogenic forms, eliciting neurotoxic effects that culminate in synaptic dysfunction and neuronal loss (Haass and Selkoe, 2007). Discovery of Prp and EphB2 as receptors for oligomeric Aβ42 (Cisse et al., 2011; Lauren et al., 2009) provides support for the view that oligomeric Aβ peptides could function as neurotoxic ligands, initiating diverse cellular signaling events that range widely, including inflammation, mitochondrial dysfunction, oxidative stress, apoptosis/autophagy, intracellular calcium imbalance, and a block in LTP (Koo and Kopan, 2004), any of which could contribute to AD pathology. The mechanism by which oligomeric Aβ peptides elicit such diverse cellular outcomes, however, has remained elusive.
Here, we report that oligomeric Aβ42 exerts such diverse effects in part by inducing a translational block, which is accompanied by ER stress as indicated by increased phosphorylation of Eif2α in hippocampal neurons. Increased Eif2α phosphorylation was reported to inhibit the late phase of LTP and memory acquisition (Costa-Mattioli et al., 2007; Costa-Mattioli et al., 2009). Once induced, ER stress activates Unfolded Protein Response (UPR), inducing a widespread secondary reactions, some of which include changes in inflammatory responses as well as cell survival programs (Ron and Walter, 2007), the often reported phenotypes in AD.
As part of UPR, ER stress activates the JNK pathway (Urano et al., 2000). JNK proteins, especially JNK3, a brain-specific JNK isoform, have been reported to play roles under neurodegenerative conditions, such as Parkinson's disease: Deletion of JNK3 in combination with JNK2 prevented loss of dopaminergic neurons after MPTP administration (Hunot et al., 2004). Deleting JNK3 also resulted in a significant increase in neuronal and oliogodendrocyte survival after traumatic injuries in the CNS (Beffert et al., 2006; Li et al., 2007). Although JNK activation has been reported in human AD brains, its role in AD pathology development remained unclear. To address this question, we analyzed an AD mouse model with and without JNK3. Our results indicate that JNK3 activation is integral to AD pathology, where JNK3 deletion restores the translational block induced by oligomeric Aβ42 and the effect of UPR.
Oligomeric Aβ42 inhibits LTP and impairs memory formation in vivo (Cleary et al., 2005; Walsh et al., 2002), suggesting that Aβ peptides are pathogenic species that disrupt normal synaptic function and cognition. Disrupting translational control by disabling eif2α phosphorylation or deleting its kinase, GCN2, also resulted in inhibition of LTP and memory acquisition (Costa-Mattioli et al., 2005; Costa-Mattioli et al., 2007). Considering these parallel findings, we decided to ask whether Aβ42 could induce a translational block. To address the question, we measured the amount of 35S-methionine incorporation in rat hippocampal neurons after treatment with 5 μM Aβ42 overnight. It should be noted that the actual concentration of oligomeric Aβ42 in 5 μM Aβ42 was estimated to be 250 nM (Figure 1A). As controls, parallel cultures were treated with Cycloheximide, a protein synthesis inhibitor, and Rapamycin and Thapsigargin, agents whose actions impinge on the translational machinery. Oligomeric Aβ42 treatment at 250nM inhibited 35S-methionine incorporation by 44% (n=3-5, p≤0.0001), while 10 nM Rapamycin and 0.5 μM Thapsigargin reduced 35S-methionine incorporation by 70-72% (n=3-5, p≤0.01 and 0.001, respectively, Figure 1B and C). The effect of 20 μM Cycloheximide was virtually complete, blocking translation by 99% (n=3-5, p≤0.00001). The reduction in 35S-methionine incorporation was not due to cell death induced by Aβ42, since there were a very few MAP2+ neurons that incorporated propidium iodide when alive (Figure 1D). We therefore conclude that Aβ42 induces a translational block in cultured neurons.
Rapamycin inhibits translation by blocking recruitment of mTOR to the translational initiation complex (Ma and Blenis, 2009), and Thapsigargin by inducing ER stress, which results in phosphorylation of Eif2α (Costa-Mattioli et al., 2009; Ron and Walter, 2007). In order to understand whether the mechanism by which oligomeric Aβ42 causes a translational block resembles that of Thapsigargin or Rapamycin, we examined the temporal changes in the phosphorylation status of various proteins that are known to be involved in the mTOR pathway and UPR in hippocampal neurons. Oligomeric Aβ42 induced a rapid increase in AMP-activated protein kinase α (AMPKα) phosphorylation (Figure 1E). Rapamycin and Thapsigargin also activated AMPK, but the kinetics of its activation differed from that by oligomeric Aβ42. Monomeric and fibrillar forms of Aβ42 did not activate AMPK in hippocampal neurons (Figure 1F). AMPK was shown to phosphorylate TSC2 and Raptor at S1387 and S792, respectively, thereby inhibiting the mTOR pathway (Gwinn et al., 2008; Inoki et al., 2003). Indeed, both TSC2 and Raptor were phosphorylated by oligomeric Aβ42 in kinetics similar to that of AMPK phosphorylation, and phosphorylation of S6 kinase decreased gradually over time (Figure 1D). We hypothesized that oligomeric Aβ42 inhibits the mTOR pathway in part by activating AMPK. Indeed, when AMPK activation was inhibited with 5 μM Compound C as indicated by a reduction in phospho-RaptorS792 levels, S6K phosphorylation was increased significantly with oligomeric Aβ42 at all the time points tested (Figure 1F). These results together suggest that oligomeric Aβ42 initiates a translational block by inhibiting the mTOR pathway via a mechanism that involves AMPK activation. It should also be noted that oligomeric Aβ42 increased phosphorylation of Eif2α, although the extent of its increase is much less than that by Thapsigargin (Figure 1E). We interpret this result as suggesting that oligomeric Aβ42-mediated translational block leads to UPR.
UPR was shown to activate the JNK pathway (Urano et al., 2000). We hypothesized that once UPR is induced, it activates JNK3, which in turn promotes further APP processing by phosphorylating it at T668P in neurons. Indeed, Thapsigargin and Rapamycin induced JNK3 activity in rat hippocampal neurons: JNK3 activity increased 5-30 min after treatments, which was sustained for up to 3-6 hrs (Figure 2A, B).
APP has a MAP kinase phosphorylation site at T668P, which becomes phosphorylated only in neurons (Iijima et al., 2000). We hypothesized that JNK3-mediated phosphorylation of APP results in internalization of APP into endosomal compartments, wherein it undergoes amyloidogenic processing. In support, Aβ42 has been detected in endosomes via immuno-EM in AD cases (Takahashi et al., 2002). We first asked whether JNK3 phosphorylated T668P in vitro. Active JNK3 indeed phosphorylated APP at T668P in vitro, as JNK3 failed to phosphorylate APP when T668 was mutated to A668 (Figure 2C). Active CDK5/p35 complex, on the other hand, still phosphorylated the mutant APPA668, suggesting that JNK3 phosphorylates T668P, while CDK5/p35 phosphorylates sites other than T668P.
As a way to address whether JNK activation affects APP internalization, we asked whether JNK activation alters the amount of the full-length APP on the cell surface. In addition, since Anisomycin activates p38 besides JNK, cortical neurons were pre-incubated for 1 hr with 1 μM cell-permeable JNK-Interacting-Protein (JIP) peptides that inhibit JNK specifically (Barr et al., 2002; Farias et al., 2009), while control cultures were treated with cell-permeable control peptides. JIP peptides, TAT-TI-JIP, contain 22-amino acids, which were shown to be necessary for JIP to inhibit JNK activation (Dickens et al., 1997). The JIP peptides were also modified to make it permeable to mammalian cells by adding 10-amino acid sequences of HIV-TAT protein (Bonny et al., 2001; Schwarze et al., 1999). After treatments with 50 ng/ml Anisomycin at 37°C for 0, 5, 15, and 30 min., cells were subjected to cell surface biotinylation on ice for 30 min, and the amount of the cell surface-located, biotinylated, full-length APP was determined following Neutravidin pulldown and subsequent Western blotting for APP. In control cultures, Anisomycin treatment resulted in a rapid increase in biotinylated APP on the cell surface, which was followed by a gradual reduction in the amount of cell surface biotinylated APP, reaching 52% of the original APP level by 30 min (Figure 2D and E). In contrast, when the endogenous JNK was inhibited by the JIP peptides as evidenced by the reduction in phospho-cjun levels, cell surface biotinylated APP levels remained unchanged (Figure 2D and E). Anisomycin treatment also increased the extent of APP phosphorylation at T668 in control cultures, while it did not in cultures treated with JIP peptides. These results suggest that JNK activation induces rapid trafficking of APP to the cell surface and subsequent internalization in part by phosphorylating APP at T668.
In order to test whether a JNK-mediated increase in internalization results in greater APP processing, cortical neurons were subjected to cell surface biotinylation using a reversible biotin crosslinker, Sulfo-NHS-SS-Biotin, prior to treating them for 2 hrs with Anisomycin and also inducing internalization at 37°C. At the end of the incubation time, remaining biotins on cell surface proteins were removed by treating cells with 50 mM DTT on ice, thus allowing selective detection of the internalized, cell-surface biotinylated proteins via Neutravidin pulldown/APP blotting or Streptavidin-conjugated secondary antibody after immunoprecipitation with 6E10 (Figure 2F, (Yu et al., 2011)). Anisomycin treatment increased the amount of biotinylated CTF production significantly, which correlated with increased T668 phosphorylation on CTF (Figure 2F). These results together suggest that JNK activation rapidly induces APP internalization/endocytosis, thereby facilitates APP cleavage reactions.
We next determined whether T668P phosphorylation by JNK is required for the internalization and processing of APP. For this, the full-length APP and a point mutant, A668, were transfected into 293T cells, and subjected to cell surface biotinylation and internalization assays as described above. The amount of the full-length APP that was biotinylated on the cell surface decreased with 30 min Anisomycin treatment in the wild type, but not in the A668P mutant, suggesting that phosphorylation of T668P facilitates the internalization of the full-length receptor (Figure 2G, H). Upon internalization, a greater amount of biotinylated CTF was detected with the wild type APP after Anisomycin treatment, but not with the A668P mutant (Figure 2G). These results together suggest that T668P phosphorylation by JNK is necessary for APP to be internalized into endosomes and processed to generate Aβ peptides.
Since our in vitro data suggested that JNK-dependent phosphorylation of T668P in APP is necessary for APP internalization and processing, we wished to determine the role of JNK in AD pathology by deleting JNK3 from an AD mouse model. We focused on JNK3, since unlike JNK1 and 2, JNK3 is enriched in the nervous system and plays a role under pathological conditions but has little effect on normal development (Kuan et al., 1999; Yang et al., 1997). As our first step toward addressing the question, we asked whether JNK3 activity increased in human AD cases as well as in 5XFAD (henceforth called “FAD”) mice in comparison to normal and mice cases, respectively, by performing immunoprecipitation/kinase assays using JNK3-specific antibody. The specificity of JNK3 antibody has been demonstrated (Li et al., 2007). FAD mice express mutant human APP (Swe/Fl/Lon) and PS1 (M146L/L286V) genes, each under a neuronal Thy1 promoter, producing more Aβ42 than Aβ40 (Oakley et al., 2006). Indeed, JNK3 activity increased by 34% in human AD compared to normal cases (n=9-13, p≤0.05), and by 27% in FAD:JNK3+/+ mice compared to normal JNK3+/+ mice at 5-6 months (n=4, p≤0.05; Figure 3A-D). These results suggest that JNK3 activity correlates with AD pathology.
We next determined where active JNK is localized in FAD brains using active JNK or p-JNK antibody in immunohistochemistry. Beginning at 2-3 months, the time when plaques begin to appear in FAD:JNK3+/+ mice, p-JNK signals were predominantly detected near plaque structures, co-localizing with 8E5 immunoreactivity (Figure 4A, B, and D), a dystrophic neurite marker (Games et al., 1995). p-JNK signals were similarly reported to colocalize with 6E10 immunoreactivity in Tg2576/PS1M146L mice (Braithwaite et al., 2010). Also in aged monkey, p-JNK signals were detected near plaque structures (Figure 4C), suggesting that accumulation of active JNK near plaque structures is a common feature in primates as well. When FAD:JNK3-/- mouse brains were analyzed, on the other hand, p-JNK signals were reduced dramatically, to near background levels, and co-localization of p-JNK with 8E5 was not detected (Figure 4A and B). These results indicate that JNK3 is the principle JNK isoform that is activated in FAD mice.
A closer observation into FAD:JNK3+/+ mice revealed that p-JNK signals were detected predominantly at sites of neuritic damage assessed by 8E5 staining (Figure 4D): At 6 months, p-JNK signals are rarely detected in the soma. This result suggests that JNK3 becomes activated in damaged and degenerating neuritic processes, in agreement with previous reports (Abe et al., 2009; Cavalli et al., 2005; Muresan and Muresan, 2005). It should be noted that JIP and JNK3 have been reported to be transported along the axon under pathological conditions, presumably linking Kinesin-1 to receptor carrying vesicles, such as APP (Cavalli et al., 2005; Taru et al., 2002). APP itself has been known to be transported along the axon via fast axonal transport (Koo et al., 1990). We interpret these results as suggesting that JNK3 can potentially phosphorylate APP at T668P in the axon. Indeed, p-T668P signals were detected from damaged neurites in FAD:JNK3+/+ mice (Figure 4E), similarly to p-JNK signals. Unlike p-JNK signals, p-T668P signals were also prominent in cell bodies (data not shown). Together, these results suggest that JNK3 becomes activated in damaged and degenerating neuritic processes, where it can phosphorylate APP and regulate its processing. It should be noted that active JNK also co-localized with hyper-phosphorylated tau in FAD:JNK3+/+ mice (data not shown).
We next analyzed the effect of deleting JNK3 on overall plaque deposition in FAD mice. In FAD:JNK3-/- mice, insoluble Aβ42 levels were reduced dramatically, by 87% at 6 months (n=8, p=0.0004) and 70% at 12 months (n=8, p=0.005), compared to those in FAD:JNK3+/+ mice, based on Aβ40 and 42-specific sandwich Elisa analyses of the brain samples (Figure 5A). Soluble Aβ40 and 42 levels were also reduced with JNK3 deletion (data not shown), but levels of soluble Aβ peptides were negligible in FAD mice. Similar reductions were observed when the area occupied by plaques was quantified after 6E10 antibody labeling at 6 months (Figure 5B and C): 68% (n=4, p≤0.01), 71% (n=4, p≤0.01), and 65% (n=4, p≤0.05) reductions were found in the frontal cortex, the subiculum and the hippocampus, respectively. As evidenced by Thioflavin S staining (Figure 5C), the size and the number of plaques were also reduced in the frontal cortex and the hippocampus at 6 months by 58% (n=4, p≤0.01) and 47% (n=4, p≤0.01), respectively. Silver staining also indicated that JNK3 deletion resulted in a significant reduction in plaques throughout the brain at 6 months (Figure 5D). More importantly, the number of neurons in layers 5 and 6 of the frontal cortex was 17% higher in FAD:JNK3-/- compared to that in FAD:JNK3+/+ mice at 12-13 months, although it did not reach the levels found in non-FAD mice (n=5-6; Figure 5E). In line with these data, deletion of JNK3 from FAD mice resulted in a significant increase in long-term retention of fear memories at 12-13 months (n=12; Figure 5F). Similarly to NeuN data, the extent of improvement in cognitive function did not reach the normal levels found in non-FAD mice, although the difference was not statistically significant. Since modulation of associative plasticity in the amygdala where fear memories are encoded involves both the hippocampus and prefrontal cortex (Maren and Quirk, 2004), these results suggest that JNK3 activation affects cognitive function in FAD mice. Together, these results indicate that JNK3 plays a critical role in development of AD pathology by not only regulating Aβ peptide production but also impacting neuronal survival and associative learning capacity in FAD mice.
We next asked whether JNK3 phosphorylates APP in the brain, thereby regulating β and/or γ cleavage processes by preparing whole cell lysates as well as membrane fractions from 6 month-old mice. JNK3 indeed phosphorylates APP at T668P in vitro and in FAD brains, without affecting the total APP protein levels (Figure 5G): While the human APP levels in whole cell lysates were not very different between FAD:JNK3+/+ and FAD:JNK3-/- mice, p-T668P signals as well as human APP protein levels were significantly reduced when membrane fractions were analyzed (Figure 5G). It should be noted that sw192 antibody is specific to Swedish mutation in human APP (Haass et al., 1995), thus it was used as a marker for FAD mice. In particular, p-T668P levels in membrane fractions were reduced to a much greater extent in α and β CTF than in the full-length APP with JNK3 deletion. This finding closely parallels the observation in human AD brains, wherein increased T668P phosphorylation mainly associated with α and β CTF and not the full-length APP (Lee et al., 2003). In addition, total protein levels of α and β CTF were also reduced to a much greater extent than those in the full-length APP in the membrane fraction (Figure 5G). These results correlate faithfully with our Aβ42 Elisa results at 6 months. We therefore interpret these results as suggesting that JNK3 phosphorylates APP preferentially in membranous compartments, such as vesicles/endosomes, thereby promoting APP processing. It should be noted that although BACE1 and PS1 levels were increased in FAD mice compared to those in normal mice as reported (O'Connor et al., 2008), JNK3 deletion did not affect their levels greatly (Figure 5H). Similarly, neither the levels nor extent of tau phosphorylation was altered by JNK3 deletion in FAD mice (data not shown).
In a preliminary RNAseq-based transcriptome analysis of 3 month-old FAD mice with and without JNK3 and the control cortices from JNK3+/+ and JNK3-/- mice, we obtained the results that suggest that there is a general translational block in FAD:JNK3+/+mice; genes involved in translation, such as ribosomes and translation-initiation factors were dramatically reduced in FAD:JNK3+/+ compared to JNK3+/+ mice and JNK3 deletion restored the effect on these genes to nearly normal levels (data not shown). We therefore tested whether there is indeed a global translational block in FAD mice by Western blotting cortical lysates with an antibody against phospho-S6235/236 ribosomal protein, a marker for active translation. Indeed, the p-S6 signal was reduced by 48% in FAD: JNK3+/+ mice, compared to that in the normal mice and FAD:JNK3-/- mice (Figure 6A, B). Immunohistochemistry with p-S6 antibody also revealed similar findings: both the number of cells that are positive for p-S6 signals and the intensity of its signals decreased significantly in the cortex of FAD:JNK3+/+ mice, compared to those in other genotypes (Figure 6C). It should also be pointed out that p-RaptorS792 levels were increased by 4-fold in FAD:JNK3+/+ compared to those in FAD:JNK3-/- (Figure 6A, B). Since S792 in Raptor is phosphorylated by AMPK (Gwinn et al., 2008), these results suggest that AMPK activity is also reduced in FAD:JNK3-/- compared to FAD:JNK3+/+ mice. These results agree also with our culture data in Figure 1.
Oligomeric Aβ42 treatment in hippocampal neurons also resulted in phosphorylation of Eif2α, albeit at a lower extent than that by Thapsigargin, suggesting that Aβ42 induces ER stress (Figure 1E). A reduction in the global rate of translation is one of the earliest events in UPR, which leads to induction of a widespread secondary response that includes transcriptional activation of the target genes for UPR, namely apoptotic as well as survival promoting genes (Ron and Walter, 2007). Western blotting of a selected set of ER stress transducers, such as p-Eif2α ATF4, and PERK, supported the finding that expression of these proteins was increased by approximately 2-fold in 3 month-old FAD:JNK3+/+ compared to those in normal and FAD:JNK3-/- mice (p<0.05; Figure 7A, B). More importantly, cortical samples from five AD patients demonstrated a significant increase in ER stress markers as well as translational block, including p-S6, p-Eif2α, ATF4, CHOP, and PERK compared to five age-matched control cases (Figure 7C). Together, these suggest that the ER stress response occurs in human AD, and JNK3 activation may contribute to it.
We hypothesized that once UPR is induced, it activates JNK3, which in turn promotes further APP processing by phosphorylating it at T668P. T668P phosphorylation facilitates increased internalization of the receptor into endosomal vesicles wherein APP undergoes processing to generate more Aβ42. As the cycle repeats itself, more Aβ42 accumulates, exacerbating the pathology (Figure 8A). To test the hypothesis that translational block and/or ER stress increase APP processing in a JNK3-dependent manner, organotypic slices were prepared from 8-10 week-old FAD:JNK3+/+ and FAD:JNK3-/- mice and treated with the vehicle, 0.5 μM Thapsigargin or 10 nM Rapamycin for the indicated amounts of time. Both CTF and Aβ peptide levels increased gradually over the 9 hr period in FAD:JNK3+/+ slices with Thapsigargin as well as Rapamycin treatments (Figure 8B). These results suggest that ER stress or inhibiting the mTOR pathway is sufficient to induce Aβ peptide production. Importantly, this increase in APP processing was dependent on JNK3, since the overall CTF and Aβ peptide levels were significantly reduced in FAD:JNK3-/- slices (Figure 8B). These results together indicate that JNK3 activation is necessary to perpetuate the cycle of translational block via mTOR, ER stress, JNK3 activation, and further production of Aβ42.
Here, we report that JNK3 activation is critical for maintaining a positive feedback loop that culminates in continued production of Aβ42. This conclusion is supported by a dramatic reduction in overall Aβ42 levels upon deleting JNK3 from FAD mice. Such a dramatic reduction is not only due to the ability of JNK3 to phosphorylate APP directly, but also due to the cyclic activation of JNK3 by a translational block and subsequent UPR that had been induced by oligomeric Aβ42 in the first place.
JNK3 phosphorylates APP at the T668P site in its cytoplasmic domain both in vitro and in vivo. The fact that JNK or JNK3 phosphorylates APP is also supported by a study, where double deletion of putative upstream JNK kinases, MKK4 and MKK7, from another FAD mouse line resulted in a reduction in T668P phosphorylation (Mazzitelli et al., 2011). T668P phosphorylation was increased in AD brains, wherein the β-CTF rather than the full-length APP exhibited increased T668P phosphorylation compared to the control (Lee et al., 2003). Supporting the notion that T668 phosphorylation contributes to APP processing, T668P to A668P mutation reduced Aβ peptide generation in vitro (Lee et al., 2003). Our data also support this view: When JNK phosphorylated APP at T668P, the amount of CTF increased. This was in part due to the fact that JNK phosphorylation of T668P in APP facilitated rapid internalization of the receptor as indicated by a reduction in the amount of the full-length APP on the cell surface. Since JNK3 is not the only kinase that phosphorylates APP at T668P in vivo as indicated by our data, and JNK3 can also be activated in the axon under pathological conditions (Falzone et al., 2009; Morfini et al., 2009), we hypothesize that JNK3 is the predominant kinase that phosphorylates APP within particular endosomal compartments in the axon where APP encounters BACE1 (Abe et al., 2009; Cavalli et al., 2005).
Structurally, phosphorylation at T668P induces propyl isomerization, converting the p-T668P peptide from trans to cis configuration (Ramelot and Nicholson, 2001). Pin1, a phosphorylation-dependent propyl isomerase, indeed binds p-T668P in vitro, thereby facilitating cis to trans conversion (Pastorino et al., 2006). Since Pin1 deletion from an AD mouse line resulted in a 46% increase in Aβ peptide production, cis configuration induced by T668 phosphorylation is believed to render APP vulnerable to amyloidogenic processing (Pastorino et al., 2006). Our data and those of Lee et al. (Lee et al., 2003) also support that T668P phosphorylation is critical for Aβ peptide generation in vitro. Whether T668P phosphorylation causes greater Aβ peptide generation in vivo is, however, still unresolved. In normal aged mice, A668P knock-in mutation did not affect β CTF generation (Sano et al., 2006), leading the authors to conclude that T668P phosphorylation plays no role in APP processing. Such a conclusion is premature especially with gain-of-function mutations such as phosphorylation, until the role of T668P phosphorylation is assessed in AD mouse models. A case in point is that although hyper-phosphorylation of tau is clearly indicated as pathologic, deleting tau alone showed a relatively minor defect in axon degeneration (Dawson et al., 2001; Gomez de Barreda et al., 2010; Harada et al., 1994). It was only when tau was deleted under an AD background that improvement in cognitive behavior was observed (Roberson et al., 2007). Although critical experiments are still needed to address whether T668P phosphorylation causes APP processing in vivo, our study provides additional support to the idea that T668P phosphorylation significantly contributes to APP processing in vivo.
We provide compelling evidence that a translational block is a prominent feature in FAD mice and to some extent in human AD cases. Since oligomeric Aβ42 induced a translational block in hippocampal neurons in culture, it is highly likely that oligomeric Aβ42 has a similar effect in vivo. Oligomeric Aβ42 is widely believed to be the central pathologic species that is responsible for inhibiting LTP and memory formation in vivo (Cleary et al., 2005; Walsh et al., 2002). Since inhibiting normal translational processes by disabling eif2α phosphorylation or deleting its kinase, GCN2, resulted in inhibition of LTP (Costa-Mattioli et al., 2005; Costa-Mattioli et al., 2007), it is tempting to speculate that such synaptotoxicity observed with oligomeric Aβ42 is likely be due to its inhibitory effect on translation.
Our data indicate that oligomeric Aβ42 inhibits translation in part by blocking the mTOR pathway. Dysregulation of the mTOR pathway or loss of energy balance has been identified as causative in normal aging as well as type 2-diabetes and obesity (Cohen et al., 2009; Demontis and Perrimon, 2010; Koo et al., 2005; Mair et al., 2011; Song et al., 2010). Our findings that widespread disruption of normal energy balance is prominent in FAD mice and to some extent in human AD cases suggest that in progressive diseases whose symptoms develop over a long period time, chronic metabolic imbalance becomes a pervasive phenotype. Our data clearly illustrate that oligomeric Aβ42 perturbs energy homeostasis, as indicated by activation of AMPK, a kinase that responds to energy imbalance in the cell (Steinberg and Kemp, 2009). AMPK was shown to play a critical role in aging in yeast and C. elegans, although the loss of snf1p, the yeast homolog of AMPK, increased the life span (Lin et al., 2001), while mutation in aak-2, the worm homolog of AMPK, decreased the life extension induced by stress (Apfeld et al., 2004). Besides this apparent species-related difference in homolog roles, the role of AMPK itself in aging appears clear. It is of special interest in this regard that oligomeric Aβ42 activates AMPK, thereby inhibiting the mTOR pathway. Aβ peptides are normally produced and cleared rapidly in human brains (Bateman et al., 2006). It is plausible that normal production of Aβ peptides contributes to the aging process in part by activating AMPK.
AMPK activation was rapid but transient by oligomeric Aβ42, detectable at 10 min, but greatly reduced by 3 hr after Aβ42 addition. Although transiently activated, AMPK substrates, Raptor and TSC2 remain phosphorylated up to 16 hrs, providing an explanation for a prolonged translational inhibition. What will be the mechanism by which oligomeric Aβ42 activates AMPK? AMPK was shown to be activated by LKB1 (Hawley et al., 2003; Shaw et al., 2004), CaMKK (Anderson et al., 2008; Hawley et al., 2005; Hurley et al., 2005; Woods et al., 2005), and TAK1 (Momcilovic et al., 2006). In the nervous system, however, LKB1 does not seem to serve as a kinase for AMPK, since AMPKα phosphorylation at T172 was not changed in LKB1 null mice (Barnes et al., 2007). A most likely candidate is CaMKK, since it is highly expressed in the brain and it associates with AMPK α and β subunits (Anderson et al., 2008). If that is the case, it is possible that oligomeric Aβ42 itself modulates intracellular calcium levels, thereby activating CaMKK. It will be of interest to test whether intracellular calcium levels change with oligomeric Aβ42 addition in neurons.
Can mTOR activity modulation be developed into a therapy for AD? This is an attractive idea, since there are already many FDA approved drugs that were designed to target the mTOR pathways for treating other progressive metabolic diseases. Although attractive, the idea appears too premature at the present time mainly because the role of the mTOR pathway in AD is not fully understood. For instance, some reported improvement in cognitive function and neuronal toxicity with Rapamycin administration (Berger et al., 2006; Bove et al., 2011; Caccamo et al., 2010; Khurana et al., 2006; Spilman et al., 2010), while others reported the opposite (Lafay-Chebassier et al., 2005). Similarly, the reports vary as to whether there is an inhibition or activation of the mTOR pathway in AD mouse models and/or human cases (Caccamo et al., 2011; Ma et al., 2010). Our data indicate that there is a significant translational block early in FAD mice. This notion was also supported by a global transcriptome analysis via RNAseq, which demonstrated a dramatic reduction in transcripts for ribosomes and elongation factors in FAD compared to the wild type mice (data not shown). It is possible that the use of different animal models at different ages in each study contributed to the opposite outcomes. It seems safe to surmise that before one takes further steps to alter the mTOR pathway or AMPK activity in pursuit of a treatment for AD, more systematic and consistent analyses are necessary.
In conclusion, our findings suggest that JNK3 activation is central to the development of AD pathology by exacerbating metabolic stress that is induced by Aβ42 accumulation. This study thus identifies JNK3 as a promising new target of therapeutic intervention for Alzheimer's disease.
Tissues from the frontal cortex were obtained through UCSD Experimental Neuropath Laboratory.
FAD mice in B6/SJL F1 hybrid background were initially crossed with JNK3 in knockout mice in B6 background to obtain FAD:JNK3+/- and control non-transgenic:JNK3+/-.
The hippocampal neurons were isolated from E18 rat embryos, and cultured for 7 days in Neurobasal medium supplemented with B27 and 2 mM glutamine.
500 μm-thick organotypic slices were prepared according to (del Rio and Soriano, 2010) from 12-14 week-old FAD:JNK+/+ and FAD:JNK3-/- mice.
The lysates from hippocampal neurons were subjected to immunoprecipitation with JNK3 antibody, and the immune complexes were used in kinase reactions using GST-c-jun as a substrate as described (Li et al., 2007).
Brain tissues that contain the cortex, the hippocampus, the septum and the striatum were used to extract proteins using 70% formic acid.
Brain tissues were processed to obtain membrane and soluble fractions for biochemical analyses as described (Pastorino et al., 2006).
For the quantification of the areas occupied by plaques, two 60 μm, floating sections from the bregma positions from +0.26 to +0.5 for the frontal cortex were processed for staining with 6E10 (n=4).
Coronal sections of the brains (60 μm) were processed for silver staining using a FD NeuroSilver kit from FD Neurotechnologies as directed by the manufacturer.
We thank Elan pharmaceuticals for the gift of 8E5 and 192sw antibodies, and Dr. Li Huei Tsai for APP-wild type and T668A mutant constructs. We also thank Drs. Gary Landreth, Bruce Carter, and Joachim Herz for valuable comments on the manuscript. This work was funded by a grant from The Alzheimer's Association (IIRG-08-90129) and NINDS (RO1NS050585) to S.O.Y. and The Ohio State Neuroscience Center Core from NINDS (P30NS045758), P30 CA016058-30 National Cancer Institute. RNA sequencing was performed at the OSUCCC Nucleic Acid Shared Resource-Illumina Core.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.