Research ArticleAlzheimer’s Disease

β-amyloid redirects norepinephrine signaling to activate the pathogenic GSK3β/tau cascade

See allHide authors and affiliations

Science Translational Medicine  15 Jan 2020:
Vol. 12, Issue 526, eaay6931
DOI: 10.1126/scitranslmed.aay6931

The noradrenergic link

Norepinephrine (NE) and the noradrenergic system play a main role in cognition and noradrenergic changes have been documented in patients with Alzheimer’s disease (AD). However, the role of NE in AD has not been completely elucidated. Here, Zhang et al. show that β-amyloid oligomers are allosteric ligands of the α2A adrenergic receptor (α2AAR) and modulate NE signaling, redirecting the pathway toward GSK3β activation and subsequent tau hyperphosphorylation. Blocking α2AAR reduced tau phosphorylation and ameliorated pathological and cognitive abnormalities in AD mouse models.


The brain noradrenergic system is critical for normal cognition and is affected at early stages in Alzheimer’s disease (AD). Here, we reveal a previously unappreciated direct role of norepinephrine signaling in connecting β-amyloid (Aβ) and tau, two key pathological components of AD pathogenesis. Our results show that Aβ oligomers bind to an allosteric site on α2A adrenergic receptor (α2AAR) to redirect norepinephrine-elicited signaling to glycogen synthase kinase 3β (GSK3β) activation and tau hyperphosphorylation. This norepinephrine-dependent mechanism sensitizes pathological GSK3β/tau activation in response to nanomolar accumulations of extracellular Aβ, which is 50- to 100-fold lower than the amount required to activate GSK3β by Aβ alone. The significance of our findings is supported by in vivo evidence in two mouse models, human tissue sample analysis, and longitudinal clinical data. Our study provides translational insights into mechanisms underlying Aβ proteotoxicity, which might have strong implications for the interpretation of Aβ clearance trial results and future drug design and for understanding the selective vulnerability of noradrenergic neurons in AD.


Alzheimer’s disease (AD) and related dementia affect nearly 50 million people globally, and there is currently no effective therapy to cure this devastating disease or to slow its progression. Strong genetic and experimental evidence indicates toxic β-amyloid (Aβ) peptides as a key driving factor of AD pathogenesis (14). However, the failure of multiple clinical trials that directly target Aβ in the brain suggests that simply reducing Aβ burden does not necessarily result in alleviation of cognitive impairment (5). The microtubule-associated protein tau is an essential mediator of Aβ toxicity (6, 7). Hyperphosphorylated and aggregated tau disrupts neuronal functions and plasticity, and spreading of tau pathology positively correlates with cognitive impairment in AD (810). Yet, the molecular pathway from Aβ to tau pathology remains elusive, presenting a major gap in in-depth understanding of the pathological cascade of AD.

Brain locus coeruleus (LC) noradrenergic neurons are highly vulnerable in AD and degenerate at early stages of the disease (1113). Noradrenergic degeneration often leads to compensatory changes (1214) and enhanced responses to norepinephrine (NE) that likely underlie agitation, aggressive behaviors, and sleep disturbance in early AD (1416). Whereas the noradrenergic system is well recognized as a sensitive target of Aβ and tau toxicity, our study reveals an unexpected direct etiological role of NE in AD pathogenesis. We report that Aβ oligomers at nanomolar concentrations hijack NE-elicited signaling through α2A adrenergic receptor (α2AAR) to activate glycogen synthase kinase 3β (GSK3β), resulting in tau hyperphosphorylation. GSK3β is a prominent tau kinase (1720) and serves as an integral regulator in the development of AD pathophysiology and cognitive deficits (2123). Thus, NE/α2AAR directly mediates Aβ toxic effects. This NE-dependent mechanism markedly increases the response sensitivity of GSK3β/tau signaling to Aβ by nearly two orders of magnitude and provides a possible role for NE in failures of clinical trials targeting Aβ clearance. Given the enriched expression of α2AAR in noradrenergic neurons, this mechanism may also render this neuronal population selectively vulnerable in AD. Our data obtained from human tissue samples and longitudinal clinical analysis, and two mouse models collectively support hyperactive noradrenergic signaling in AD as a critical element linking Aβ to the pathogenic GSK3β/tau cascade that ultimately leads to cognitive impairment.


α2AAR activity is elevated in patients with AD and mouse models

α2AAR is broadly expressed in both noradrenergic and non-noradrenergic neurons in the brain and controls both NE input and its resulting responses (24, 25). As a member of the G protein–coupled receptor (GPCR) superfamily, α2AAR activates heterotrimeric G proteins to trigger signal transduction. Our pharmacological characterization of α2AAR in postmortem prefrontal cortex (table S1) revealed a significant increase (P < 0.01) in α2AAR activity [Emax/Bmax, reflecting maximum G protein activation in response to NE per receptor (26)] in AD cases compared to nondemented, low pathology control subjects (Fig. 1A). Furthermore, our analysis of cases from the National Alzheimer’s Coordinating Center (NACC) database (table S2) revealed that usage of clonidine, an α2AAR activator, was associated with worsened cognitive function in patients with cognitive deficits, whereas it had no effect in subjects with normal cognition (Fig. 1B). The adverse effect of clonidine was stronger in patients with more severe dementia (Fig. 1C). Combined, these data suggest that α2AAR signaling is hyperactive in patients with AD, and activation of this receptor is detrimental to cognitive function.

Fig. 1 α2AAR function is enhanced in patients with AD and animal models.

(A) Membrane homogenates were prepared from postmortem prefrontal cortical tissues of patients with AD and control subjects. Bmax reflects α2AAR density. Emax reflects maximum α2AAR-mediated G protein activation in response to NE (applied with propranolol and prazosin to selectively activate α2AAR). For each set of experiments, an AD subject and a control subject were analyzed in parallel. **P < 0.01 by paired t test. (B) Changes in the adjusted z score for the mini-mental state (ZMMS) examination during the time period with or without clonidine usage were analyzed. ΔZMMS reflects average change in ZMMS in a year. **P < 0.01; ****P < 0.0001, by post hoc Sidak’s multiple comparisons test. (C) Average changes in the adjusted ZMMS score in patients with cognitive deficits during the time period with clonidine usage. Cognitive status code: 2, impaired not mild cognitive impairment (MCI); 3, MCI; and 4, dementia. ***P < 0.001 by one-way ANOVA. (D) Brain homogenates were prepared from nontransgenic (nTg) or APP/PS1 littermates at 7.5 months of age. G protein activation was measured in response to NE (with prazosin and propranolol to selectively activate α2AAR). **P < 0.01 by two-way ANOVA. (E) Sedation was measured by rotarod test in response to an α2AAR activator, UK14,304, in nTg, or APP/PS1 littermates at 7.5 to 8 months of age. **P < 0.01 by two-way ANOVA. (F) G protein activation in response to an A1R-selective activator, (R)-N6-(1-Methyl-2-phenylethyl)adenosine (R-PIA), in brain homogenates prepared from nTg or APP/PS1 littermates at 7.5 months of age. (G) α2AAR-mediated G protein activation in WT mouse brain homogenates in the presence or absence of AβO (100 nM, monomer equivalent). **P < 0.01 by two-way ANOVA. (H) α2AAR-mediated G protein activation in WT mouse brain homogenates in the presence of human TBS extracts with or without Aβ depletion. **P < 0.01, TBS extracts versus control by two-way ANOVA. All data are shown as means ± SEM. a.u., arbitrary units; GTP-γ-S, guanosine 5′-O-(3′-thiotriphosphate).

We recapitulated the AD-related increase in α2AAR activity in mouse models. We first compared APP(Swe)/PS1ΔE9 (APP/PS1) transgenic mice (27) and their age-matched nontransgenic littermates at 7.5 months of age, well after initiation of Aβ deposition. There was a leftward shift of the ex vivo NE dose-response curve that indicates enhanced efficiency of α2AAR-mediated G protein activation in the brains of APP/PS1 mice compared to their age-matched nontransgenic littermates (Fig. 1D). In addition, the α2AAR-elicited sedation response was potentiated in 7.5-month-old APP/PS1 mice (Fig. 1E), suggesting that the increase in G protein activation efficiency by α2AAR indeed results in enhancement of in vivo α2AAR function in experimental AD. G protein activation by another Gi/o-coupled receptor, the adenosine A1 receptor, is not altered in APP/PS1 mice (Fig. 1F), suggesting specific enhancement of α2AAR activity in these mice. Similarly, we observed enhanced efficiency of α2AAR-mediated G protein activation in an independent AD mouse model, AppNL-G-F/NL-G-F knock-in (APP-KI) (28), at 7.5 months of age compared to age-matched wild-type (WT) controls (fig. S1). There was no change in α2AAR density in transgenic or KI mice compared to their respective controls (fig. S2).

We found no difference in α2AAR-mediated G protein activation nor α2AAR density between APP/PS1 and nontransgenic mice at 5 weeks of age before the development of Aβ-related pathology (fig. S3), suggesting that the change in α2AAR response efficiency in 7.5-month-old transgenic mice is likely attributed to Aβ accumulation in the brain. In support of this notion, addition of Aβ42 oligomers (AβO; 100 nM, monomer equivalent; fig. S4), which are the primary toxic species in AD, sufficiently increased the efficiency of NE to induce G protein activation through α2AAR (Fig. 1G). Furthermore, tris-buffered saline (TBS) extracts from AD prefrontal cortex (AD-TBS extracts), which contain soluble Aβ oligomers (29), also increased the efficiency of α2AAR-mediated G protein activation (Fig. 1H and fig. S5).

O acts as an allosteric ligand of α2AAR

The direct effect of AβO on α2AAR function motivated us to test whether AβO could bind to α2AAR directly. Flow cytometry assays showed binding of AβO, but not Aβ42 monomers, to cells expressing α2AAR (Fig. 2, A to C). Bound AβO was colocalized with α2AAR on the cell surface (Fig. 2D). Binding of AβO to immunopurified α2AAR showed a saturable curve with a dissociation constant (Kd) less than 30 nM (monomer equivalent concentration of total Aβ42 peptide, Fig. 2E). We could not detect substantial binding of AβO to two other GPCRs, α2CAR (another α2AR subtype), and adenosine A1 receptor (Fig. 2F and fig. S6), demonstrating the specificity of the AβO2AAR interaction.

Fig. 2 O binds to an allosteric site of α2AAR with nanomolar affinity.

(A and B) Human embryonic kidney (HEK) cells transfected with the empty vector or hemagglutinin (HA)–tagged α2AAR were incubated with AβO for 30 min. AβO bound to the surface of cells was detected by flow cytometry assays. ****P < 0.0001 by one-way ANOVA in (B). (C) Flow cytometry assays were performed with cells expressing HA-α2AAR after incubation with vehicle, monomer, or oligomer Aβ. (D) AβO and HA-α2AAR were detected by immunocytochemistry. Scale bars, 5 μm. (E) HA-α2AAR was immunopurified from HEK cells and incubated with increasing amounts of 5-Carboxyfluorescein (FAM)–labeled AβO or scrambled (scbd) Aβ42 peptide. (F) Saturation binding curves of FAM-AβO to different receptors expressed on the surface of intact HEK cells. (G) The docked Aβo-α2AAR complex model. Green, Aβ pentamer with hydrophobic C termini of monomers indicated in orange. Purple, the 3eL of α2AAR. Dashed black lines and orange lines indicate hydrogen bonds and hydrophobic contacts, respectively. (H) Binding of FAM-AβO to WT or α2AAR mutants, as indicated, expressed on the cell surface. (I) Binding of FAM-AβO (20 nM, monomer equivalent) to immunoisolated α2AAR in the presence of increasing concentrations of NE. All data are shown as means ± SEM. (J) Total brain lysates prepared from APP/PS1 or APP/PS1,α2AARHA/HA mice were subjected to coimmunoprecipitation assays using an HA antibody. The α2AARHA/HA allele harbors an HA tag at the N terminus of the endogenous α2AAR locus. APP/PS1 mouse brains were used as a negative control. Representative blots from multiple experiments are shown. IgG, immunoglobulin G; IP, immunoprecipitation.

We next determined the nature of AβO binding to α2AAR. Our in silico docking result indicated that AβO likely binds to the third extracellular loop (3eL) of α2AAR (Fig. 2G). Alanine mutations of nine amino acids (amino acids 399 to 406) in the 3eL (3eL-9A) abolished AβO binding to the receptor (Fig. 2H and fig. S7) without affecting binding of an orthosteric ligand, RX821002 (fig. S8). Conversely, a mutation at D113 of α2AAR (D113A) that eliminates binding of orthosteric ligands (fig. S8) (30) did not alter binding of AβO to α2AAR (Fig. 2H). These data demonstrate that AβO binds to an allosteric site of α2AAR involving the 3eL. Furthermore, NE enhanced the binding affinity of AβO to α2AAR in a dose-dependent and saturable manner (Fig. 2I). These results, along with the fact shown above that AβO enhanced the potency of NE (reflecting its binding affinity to the receptor) to elicit α2AAR-mediated G protein activation, clearly demonstrate the reciprocal nature of the regulation between a GPCR orthosteric ligand (NE) and an allosteric ligand (AβO). Moreover, our data suggest that AβO actions can be sensitized by an endogenous neurotransmitter (NE). To examine the endogenous interaction between AβO and α2AAR, we crossbred the APP/PS1 line with the HA-tagged α2AAR KI line (α2AARHA/HA) that we generated previously (31) to acquire APP/PS1,α2AARHA/HA mice. Using these mice, we detected stable complex formation between the endogenous α2AAR and AβO in the brain (Fig. 2J).

Aβo binding to α2AAR redirects receptor signaling to activate the GSK3β/tau cascade

Binding of an allosteric ligand to a GPCR often alters signaling through the same receptor (32, 33). We therefore searched for AβO-induced changes in α2AAR signaling using protein kinase arrays. Among the kinases tested, we found a change in GSK3β phosphorylation at Ser9 (indicating an increase in activity) (P < 0.05; Fig. 3A and fig. S9). In cultured primary neurons, we observed a significant reduction (P < 0.01) in GSK3β phosphorylation at Ser9 only in cells cotreated with AβO and NE but not in cells treated with either agent alone (Fig. 3, B and C). When coapplied with clonidine, AβO sufficiently induced GSK3β dephosphorylation/activation at a concentration of 20 nM (monomer equivalent; Fig. 3, B and D). This amount is less than 1% of the concentration required for GSK3β activation in neurons by AβO alone (34, 35). Furthermore, naturally secreted oligomeric Aβ at nanomolar concentrations (36) also induced GSK3β dephosphorylation/activation in neurons in the presence of clonidine (fig. S10). Concurrent with GSK3β activation in cells cotreated with NE and AβO, we observed a significant increase (P < 0.01) in tau hyperphosphorylation at AD-relevant sites, Ser202 and Thr205, detected by AT8 antibody (Figs. 3E and 4, A and C, and fig. S11). This change in tau could not be detected in cells where GSK3β expression was suppressed by small interfering RNAs (siRNAs) (Fig. 3E and fig. S11), suggesting that GSK3β activation is required for tau hyperphosphorylation in response to NE and AβO cotreatment. Together, these data demonstrate that Aβ aberrantly redirects NE-induced α2AAR signaling to GSK3β activation and subsequent tau hyperphosphorylation. This NE/α2AAR-dependent pathway can increase the response sensitivity of GSK3β/tau signaling to Aβ by two orders of magnitude.

Fig. 3 O redirects α2AAR signaling to activation of the GSK3β/tau cascade.

(A) Representative blots and quantitation of protein kinase arrays. Array blots were incubated with lysates from Neuro2A cells expressing WT α2AAR with treatment as indicated. NE (400 nM) was applied with prazosin and propranolol to selectively activate α2AAR. Ctrl, positive controls for array blotting. *P < 0.05 by one-way ANOVA. (B) Primary cortical neurons (14 days in vitro) were stimulated as indicated for 30 min. V, vehicle; Clon, clonidine (1 μM). Representative Western blots of phospho-GSK3β (pGSK3β), total GSK3β, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). (C and D) Quantitation of changes in the ratio of pGSK3β to GSK3β. **P < 0.01 by one-way ANOVA Tukey’s multiple comparisons. *P < 0.05 by unpaired t test. (E) Neuro2A cells were cotransfected with WT α2AAR and a siRNA against GSK3β or scrambled (scbd) siRNA. Representative blots of tau phosphorylation are shown. (F to H) Mice that received bilateral intrahippocampal injection of AβO (100 pmol, monomer equivalent, each side) or vehicle were treated intraperitoneally with saline, idazoxan (3 mg/kg), lithium (300 mg/kg), or idazoxan and lithium. Twenty-four hours later, hippocampal lysates were analyzed by Western blot. Representative blots (F) and quantitation (G and H) of GSK3β and tau phosphorylation are shown. *P < 0.05, **P < 0.01, ***P < 0.001 by one-way ANOVA Tukey’s multiple comparisons. All data are shown as means ± SEM.

Fig. 4 Allosteric binding to the 3eL is required for AβO to induce activation of the GSK3β/tau cascade through α2AAR.

Neuro2A cells expressing WT or 3eL-9A mutant α2AAR were treated with vehicle, NE (400 nM), AβO (100 nM, monomer equivalent), or NE and AβO. (A) Representative Western blots of pGSK3β at Ser9, total GSK3β, phospho-tau (ptau, detected by AT8 antibody), total tau, and GAPDH are shown. (B) Quantitation of changes in the ratio of pGSK3β to GSK3β. ****P < 0.0001 by two-way ANOVA post hoc Sidak’s multiple comparisons test. (C) Quantitation of changes in the ratio of ptau to tau. ****P < 0.0001 by two-way ANOVA Sidak’s multiple comparisons. (D) α2AAR-mediated G protein activation was measured by GTPγS binding assays using membrane homogenates prepared from cells expressing the WT or indicated mutant α2AAR.

We further examined the role of endogenous α2AAR activation by NE in Aβ-induced GSK3β/tau signaling in vivo. AβO (100 pmol, monomer equivalent) or vehicle was microinjected bilaterally into the dorsal hippocampus of WT C57BL/6 mice, which then received treatment with either saline, idazoxan (an α2AAR blocker), or lithium (a GSK3β blocker). AβO injection induced a decrease in GSK3β phosphorylation and a concurrent increase in tau hyperphosphorylation (detected by AT8 antibody) in dorsal hippocampi when compared to vehicle sham controls (Fig. 3, F to H). AβO-induced tau hyperphosphorylation was diminished by lithium treatment (Fig. 3, F and H), indicating an essential role of GSK3β in this process. When α2AAR was blocked by idazoxan, AβO-induced changes in GSK3β and tau were abolished (Fig. 3, F to H). These data suggest that endogenous α2AAR activation is required for Aβ-induced GSK3β/tau signaling in vivo, providing strong evidence for an essential role of noradrenergic signaling in Aβ-induced tau hyperphosphorylation.

To validate that AβO acts through allosteric binding to α2AAR to activate the GSK3β/tau signaling, we performed experiments with cells expressing the 3eL-9A mutant α2AAR, which cannot interact with AβO (Fig. 4A). Mutations in the 3eL did not affect binding of orthosteric ligands to α2AAR (fig. S8), and this mutant receptor was still able to mediate G protein signaling in response to NE (Fig. 4D). However, in cells expressing the 3eL-9A mutant receptor, cotreatment with AβO and NE failed to alter either GSK3β or tau phosphorylation (Fig. 4, A to C). These data suggest that the allosteric binding of AβO to the 3eL of α2AAR is required for activation of the pathogenic GSK3β/tau cascade. Collectively, our data reveal a previously unappreciated molecular mechanism, namely, hijacking NE/α2AAR signaling, that enables nanomolar concentrations of extracellular Aβ to activate the pathogenic GSK3β/tau cascade and that this pathway can be effectively blocked by α2AAR inhibitors (fig. S12).

Blockade of α2AAR reduces GSK3β activation and tau hyperphosphorylation and ameliorates AD-related pathological and cognitive deficits

Our results reveal an NE/α2AAR-dependent mechanism connecting Aβ to the pathogenic GSK3β/tau cascade. We therefore hypothesized that blocking α2AAR in the presence of Aβ pathology would have therapeutic potential. To test this, we treated APP/PS1 mice with idazoxan for 8 weeks starting at 7.5 to 8 months of age when Aβ plaques were present, and α2AAR activity was enhanced. Compared to saline treatment, idazoxan reversed hyperactivation of GSK3β in APP/PS1 mouse brains (Fig. 5, A and B), providing additional support for the critical role of α2AAR in mediating Aβ-induced GSK3β activation in vivo. In the cerebral cortex of idazoxan-treated APP/PS1 mice, the extent of Aβ load was lower than that in saline-treated mice (Fig. 5, C and D), indicating that blocking α2AAR slows the progression of Aβ pathology. This effect likely results both from the reduction in GSK3β activity, given that GSK3β can promote Aβ generation (37), and from regulation of sorting-related receptor with A repeats (SorLA)–mediated APP trafficking, as we previously reported (38). Similarly, idazoxan treatment of APP-KI mice also reduced Aβ plaque load (fig. S13) and GSK3β activation (fig. S14) in the brain. We further examined another important feature of AD pathology, microglial activation (39). Idazoxan treatment decreased the density of Iba-1–positive microglial cells (Fig. 5, E and F), suggesting reduction of neuroinflammation.

Fig. 5 Blocking α2AAR in AD model mice with profound Aβ pathology reduces GSK3β activity, amyloid pathology, and tau hyperphosphorylation.

Eight-month-old APP/PS1 and nontransgenic (nTg) littermate mice were treated with saline or idazoxan for 8 weeks, followed by a 1-week washout period. (A) Representative Western blots and (B) quantitation of phospho-GSK3β (pGSK3β) at Ser9 and total GSK3β in total cortical lysates. **P < 0.01 by unpaired Student’s t test. (C) Representative images and (D) quantitation of Aβ plaques (detected by 6E10 antibody) in the cerebral cortex and hippocampus of APP/PS1 mice subjected to treatments indicated. Scale bars, 500 μm. *P < 0.05 by unpaired t test. (E) Representative images and (F) quantitation of microglial cells (detected by Iba-1 antibody) in the cerebral cortex of APP/PS1 mice subjected to treatments were indicated. Scale bars, 100 μm. **P < 0.01 by unpaired t test. (G) Representative images of AT8 (for hyperphosphorylated tau) and an Aβ antibody staining in the cortex of nTg and APP/PS1 mice after the indicated treatment. Scale bars, 20 μm. (H) Quantitation of the intensity of AT8 signals plotted against the area of Aβ accumulations in the cortex. r2 = 0.84. Slope values for saline and idazoxan groups are 34.29 (±1.519) and 22.98 (±1.017), respectively. (I) Relative AT8 intensity normalized against the corresponding area of Aβ depositions. **P < 0.01 by unpaired t test.

In the cerebral cortex of APP/PS1 mice, we detected accumulations of pretangle phospho-tau clusters (detected by AT8 antibody) in areas with Aβ accumulation (detected by an Aβ antibody) (Fig. 5G), and the intensity of AT8 staining positively correlated with the size of Aβ depositions (Fig. 5H), consistent with the idea that the Aβ plaque environment promotes pathological changes in tau (40). In idazoxan-treated APP/PS1 mice, the appearance and intensity of AT8-positive signals were markedly reduced compared to those in saline-treated littermates (Fig. 5, G to I). Furthermore, idazoxan treatment reduced tau hyperphosphorylation induced by the same amount of Aβ deposition (Fig. 5, H and I), suggesting that blockade of α2AAR effectively alleviates Aβ-induced tau pathology. In APP-KI mice, idazoxan treatment also reduced tau hyperphosphorylation in the brain (fig. S14).

Reduced Aβ pathology and tau hyperphosphorylation would result in enhanced preservation of cognitive function. Saline-treated APP/PS1 mice showed clear deficits in the Morris water maze task when compared to their age-matched nontransgenic littermate controls, whereas idazoxan-treated APP/PS1 mice performed significantly better (P < 0.01; Fig. 6, A and B). Despite the presence of a substantial Aβ burden (Fig. 5C), idazoxan-treated APP/PS1 mice behaved similarly to their nontransgenic littermates on the last day of training (Fig. 6, A and B). Idazoxan treatment did not alter the baseline activity or anxiety (figs. S15 and S16). Cognitive changes in APP-KI mice were also mitigated by idazoxan treatment. Compared to age-matched WT controls, APP-KI mice showed reduced latency to the dark on day 2 in passive avoidance task (Fig. 6C), which was normalized by idazoxan treatment (Fig. 6D). These behavioral changes were not due to acute drug effects, but are interpreted as a result of the reduction in GSK3β/tau signaling by idazoxan treatment because a 1-week drug washout period was incorporated before testing the mice. These data collectively demonstrate that blocking NE signaling through α2AAR is an effective strategy to ameliorate pathological and cognitive deficits associated with Aβ.

Fig. 6 Blocking α2AAR in AD model mice with profound Aβ pathology ameliorates cognitive deficits.

(A) Measurement of escape latency on each day in Morris water maze tests in APP/PS1 and nTg mice. ***P < 0.001, saline-treated APP/PS1 versus nTg mice; **P < 0.01, saline-treated versus idazoxan-treated APP/PS1 mice by two-way ANOVA. (B) Quantitation of the number of crosses of the target quadrant in probe trial. *P < 0.05 by one-way ANOVA post hoc Tukey’s multiple comparisons test. (C) Measurement of escape latency to the dark side in passive avoidance tests in 8-month-old WT and APP-KI mice. ***P < 0.001 by two-way ANOVA Tukey’s multiple comparisons test. (D) Measurement of escape latency to the dark side in passive avoidance tests in 8-month-old APP-KI mice treated with saline or idazoxan. **P < 0.01 and ****P < 0.0001 by two-way ANOVA Tukey’s multiple comparisons test. All data are shown as means ± SEM.


Our current study provides strong evidence for an essential role of noradrenergic signaling in Aβ proteotoxicity. We show that AβO can hijack NE-elicited signaling through α2AAR to induce activation of the pathogenic GSK3β/tau cascade (fig. S12), resulting in tau hyperphosphorylation and accelerated cognitive decline. This NE/α2AAR-dependent mechanism allows AβO to induce GSK3β activation at a concentration as low as 1% of that required for GSK3β activation by AβO alone and thus could be engaged in early stages of AD when Aβ concentrations are low. Normal physiological concentrations of Aβ are in picomolar ranges (41, 42), and oligomers start to form when Aβ concentrations reach a critical aggregation concentration of 90 nM (43). At this concentration, AβO can sufficiently induce the pathogenic GSK3β/tau cascade through the mechanism identified here, providing a route for Aβ to initiate the disease process. We thus speculate that the noradrenergic system, α2AAR in particular, could play a critical role in establishing the Aβ-dependent tipping point at which hyperphosphorylation of tau accelerates. Interventions targeting this NE/α2AAR-dependent mechanism of Aβ action would be helpful in slowing or even halting the transition from normal physiology to the earliest stages of disease. Our study may also inform interpretation of failed clinical trials targeting Aβ. Considering that Aβ in human AD brains can reach micromolar concentrations (44), it would be difficult to reduce Aβ to below nanomolar concentrations to prevent activation of the α2AAR/GSK3β/tau cascade.

We show that α2AAR activity is enhanced in AD, which would further sensitize neurons to Aβ-induced tau pathology and cognitive dysfunction. This notion aligns with results reported here from unbiased epidemiological analysis of the NACC database and preclinical studies using two mouse models of experimental AD. Given that α2AAR is highly expressed in LC noradrenergic neurons, the Aβ/α2AAR/GSK3β/tau cascade also provides a possible mechanism for these neurons to be exquisitely vulnerable in AD.

Our preclinical studies demonstrate that pharmacologically blocking endogenous NE/α2AAR signaling can effectively reduce activation of the GSK3β/tau cascade, resulting in mitigation of behavioral deficits. α2AAR blockers such as idazoxan have been developed for use in other disorders, and repurposing these drugs could be a potentially effective, readily available strategy for AD treatment. In addition, our data suggest that the Aβ-α2AAR interaction represents an attractive disease-specific therapeutic target for AD because the α2AAR/GSK3β/tau cascade can only be activated in the presence of Aβ oligomers. Directly targeting the Aβ-α2AAR interface would not interfere with normal α2AAR functions and therefore be less likely to result in complications associated with an extended dosing period necessary for AD treatment. Furthermore, from a pharmacological perspective, simultaneously targeting both the orthosteric and allosteric binding sites would create a synergistic effect on receptor-mediated responses (45). Thus, the combined use of α2AAR blockers (targeting the orthosteric site) and drugs that reduce Aβ load (decreasing the allosteric binding of Aβ to α2AAR) may lead to an enhanced therapeutic effect.

We are aware of limitations of our study. Our analysis of the clonidine effect on cognition in human patients is not a controlled study, and the sample size is relatively small. Nonetheless, our unbiased epidemiological analysis of the longitudinal clinical data supports the notion that chronic α2AAR activation exacerbates AD disease progression in human patients. It would also be useful to analyze the effect of α2AAR blockers in patients. Unfortunately, the sample size of subjects using other α2AAR agonists or antagonists in the NACC database is too small for statistical tests. Another limitation of the study is that a homology model of α2AAR was used in the in silico docking study, as the α2AAR crystal structure is not available. A precise view of the structural base of the Aβ-α2AAR interaction would be necessary for the design of compounds to disrupt the interaction interface.

In summary, our current study demonstrates that Aβ rewires NE signaling to induce activation of the pathogenic GSK3β/tau pathway, providing new insights into mechanisms underlying Aβ proteotoxicity, which have strong implications for the interpretation of Aβ clearance trial results and future drug design.


Study design

The overall goal of our study is to address the potential role of the brain noradrenergic system in AD pathogenesis. We focused on α2AAR, a key component of the noradrenergic system. We first examined the potential disease relevance of this receptor to AD using human tissue samples and longitudinal clinical data and then recapitulated the AD-related increase in α2AAR activity in two independent AD mouse models. Next, we determined properties of Aβ oligomer (AβO) binding to α2AAR using combined biochemical, cell biological, pharmacological, and computational methods. We then investigated the biological consequence of the AβO2AAR interaction on intracellular signaling in neuronal cells and in the brain. Last, we performed preclinical studies to explore the therapeutic potential of blocking α2AAR in ameliorating AD-related pathological and cognitive deficits. Throughout the study, we exploited an interdisciplinary approach, used multiple technical controls, and included both technical replicates and biological repeats in our assays. Littermate mice were randomly assigned to different treatment groups. Experimenters were blinded with phenotypes or treatments in animal studies. Sample sizes were determined on the basis of previous experience with similar studies. The number of samples indicated in the figure legends reflects independent biological repeats. Conclusions were drawn on the basis of careful statistical analyses.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 7.0 software. In general, Student’s t test was performed to determine differences between two groups, and one-way or two-way analysis of variance (ANOVA) was performed to determine variations in multiple groups with one or two variances. For Fig. 1A, age- and gender-matched controls and AD samples were paired and analyzed in parallel, and paired t test was performed to determine the difference between the two groups. For Fig. 1C, one-way ANOVA was performed, and for the rest of the panels in Fig. 1, two-way ANOVA was used. For Fig. 3 (C, G, and H), one-way ANOVA was performed, and post hoc Tukey’s multiple comparisons were used to determine the difference between two groups. For Fig. 4 (B and C), two-way ANOVA and post hoc Sidak’s multiple comparisons were performed. For Fig. 6 (A, C, and D), two-way ANOVA and post hoc Tukey’s multiple comparisons were performed. For all statistical tests, P < 0.05 was considered statistically significant. Data are presented as means ± SEM. Nonlinear regression curve fit for saturation binding and dose-response curves was also performed using GraphPad Prism.


Materials and Methods

Fig. S1. α2AAR-mediated G protein activation in brain homogenates prepared from WT or APP-KI mice at 7.5 months of age.

Fig. S2. α2AAR density measured by radioligand-binding assays in AD models.

Fig. S3. α2AAR-mediated G protein activation and receptor density tested in brain homogenates prepared from nTg or APP/PS1 mice at 5 weeks of age.

Fig. S4. Profiling of Aβ42 peptide oligomerization by fluorescent size-exclusion chromatography.

Fig. S5. α2AAR-mediated G protein activation in WT mouse brain homogenates in the presence of human TBS extracts with or without Aβ depletion.

Fig. S6. Cell-surface expression of HA-tagged receptors tested by fluorescence-activated cell sorting.

Fig. S7. AβO was detected on the surface of cells expressing WT but not 3eL-9A mutant α2AAR.

Fig. S8. Binding of an orthosteric ligand to WT or mutant α2AARs.

Fig. S9. Full blots of the AKT pathway phosphorylation arrays.

Fig. S10. Naturally secreted oligomeric Aβ induced GSK3β dephosphorylation/activation in neurons in the presence of clonidine.

Fig. S11. Quantitation of tau phosphorylation and GSK3β expression.

Fig. S12. Proposed model of Aβ hijacking NE signaling through α2AAR to induce activation of GSK3β/tau cascade.

Fig. S13. Idazoxan treatment reduces Aβ pathology in APP-KI mouse brains.

Fig. S14. Idazoxan treatment reduces GSK3β activity and tau hyperphosphorylation in APP-KI mouse brains.

Fig. S15. Open-field and elevated zero maze tests in nTg and APP/PS1 mice.

Fig. S16. Open-field and elevated zero maze tests in APP-KI mice.

Table S1. Information of human samples used in Fig. 1A.

Table S2. Extracted data used in Fig. 1 (B and C).

Table S3. Information of antibodies used in this study.

Data file S1. Raw data.

References (2729, 31, 38, 4665)


Acknowledgments: We thank Y. Peng, I. Mendoza, and U. Nur for technical support. Funding: This work is funded by NIA/NIH grants AG056815 (to Q.W.) and 064664 (to Q.W.) and NIMH/NIH grant MH081917 (to Q.W.). Autopsy materials used in this study were obtained from the University of Washington Neuropathology Core, which is supported by the Alzheimer’s Disease Research Center (AG05136), the Adult Changes in Thought Study (AG006781), and the Morris K. Udall Center of Excellence for Parkinson’s Disease Research (NS062684). The NACC database is funded by NIA/NIH Grant U01 AG016976. NACC data are contributed by the NIA-funded ADCs: P30 AG019610 (E. Reiman), P30 AG013846 (N. Kowall), P50 AG008702 (S. Small), P50 AG025688 (A. Levey), P50 AG047266 (T. Golde), P30 AG010133 (A. Saykin), P50 AG005146 (M. Albert), P50 AG005134 (B. Hyman), P50 AG016574 (R. Petersen), P50 AG005138 (M. Sano), P30 AG008051 (T. Wisniewski), P30 AG013854 (M. M. Mesulam), P30 AG008017 (J. Kaye), P30 AG010161 (D. Bennett), P50 AG047366 (V. Henderson), P30 AG010129 (C. DeCarli), P50 AG016573 (F. LaFerla), P50 AG005131 (J. Brewer), P50 AG023501 (B. Miller), P30 AG035982 (R. Swerdlow), P30 AG028383 (L. Van Eldik), P30 AG053760 (H. Paulson), P30 AG010124 (J. Trojanowski), P50 AG005133 (O. Lopez), P50 AG005142 (H. Chui), P30 AG012300 (R. Rosenberg), P30 AG049638 (S. Craft), P50 AG005136 (T. Grabowski), P50 AG033514 (S. Asthana), P50 AG005681 (J. Morris), and P50 AG047270 (S. Strittmatter). Author contributions: Q.W., K.J., E.D.R., H.X., and B.S. conceived, designed, and/or planned the experiments. M.G. analyzed receptor densities and activities in human and animal samples. F.Z. performed amyloid binding and signaling experiments. Y.C. analyzed the clinical data. M.G. treated the mice and performed behavioral tests. M.G., F.Z., S.Y., and W.F. performed pathological analyses of the mice. S.Z. performed in silico modeling. J.T. analyzed amyloid oligomerization. Z.L. performed flow cytometry assays. Q.W., K.J., B.S., F.Z., M.G., Y.C., S.Z., and J.T. analyzed the data. C.D.K. provided human tissue samples. T. Saito and T. Saido generated the APP-KI line and provided suggestions. Q.W., K.J., F.Z., M.G., Y.C., S.Z., J.T., E.D.R., H.X., and C.D.K. prepared the manuscript. Competing interests: E.D.R. has served on scientific advisory boards for Biogen, AVROBIO, AGTC, and Novartis. Q.W., M.G., F.Z., Y.C., K.J., S.Y., W.F., B.S., J.T., Z.L., S.Z., C.D.K., H.X., T. Saito, and T. Saido declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials. The APP-KI line is available from the RIKEN Center for Brain Science under a material transfer agreement with the institute, and the human postmortem tissues are available from University of Washington under a material transfer agreement with the university.

Stay Connected to Science Translational Medicine

Navigate This Article