Research ArticleAlzheimer’s Disease

The amyloid-β oligomer Aβ*56 induces specific alterations in neuronal signaling that lead to tau phosphorylation and aggregation

See allHide authors and affiliations

Sci. Signal.  09 May 2017:
Vol. 10, Issue 478, eaal2021
DOI: 10.1126/scisignal.aal2021

A more pathological amyloid-β oligomer

Amyloid-β (Aβ) is implicated in the pathology of Alzheimer’s disease (AD), and various types of Aβ oligomers emerge at different stages of AD. Aβ promotes the modification and aggregation of the microtubule-associated protein tau, which is associated with neuronal toxicity and impaired cognition in various neurodegenerative disorders. Using several transgenic mouse models of AD and cultured cortical neurons, Amar et al. found that the 56-kDa oligomer Aβ*56, but not Aβ dimers or trimers, stimulated an influx in intracellular Ca2+ that triggered phosphorylation of tau at a site that promoted its aggregation. The findings link a specific amyloid form to tau pathology and suggest that dissecting the molecular and stage-specific roles of Aβ oligomers may lead to improved therapies.

Abstract

Oligomeric forms of amyloid-forming proteins are believed to be the principal initiating bioactive species in many neurodegenerative disorders, including Alzheimer’s disease (AD). Amyloid-β (Aβ) oligomers are implicated in AD-associated phosphorylation and aggregation of the microtubule-associated protein tau. To investigate the specific molecular pathways activated by different assemblies, we isolated various forms of Aβ from Tg2576 mice, which are a model for AD. We found that Aβ*56, a 56-kDa oligomer that is detected before patients develop overt signs of AD, induced specific changes in neuronal signaling. In primary cortical neurons, Aβ*56 interacted with N-methyl-d-aspartate receptors (NMDARs), increased NMDAR-dependent Ca2+ influx, and consequently increased intracellular calcium concentrations and the activation of Ca2+-dependent calmodulin kinase IIα (CaMKIIα). In cultured neurons and in the brains of Tg2576 mice, activated CaMKIIα was associated with increased site-specific phosphorylation and missorting of tau, both of which are associated with AD pathology. In contrast, exposure of cultured primary cortical neurons to other oligomeric Aβ forms (dimers and trimers) did not trigger these effects. Our results indicate that distinct Aβ assemblies activate neuronal signaling pathways in a selective manner and that dissecting the molecular events caused by each oligomer may inform more effective therapeutic strategies.

INTRODUCTION

According to our current understanding, pathological changes in the microtubule-associated protein tau in Alzheimer’s disease (AD) may be elicited by soluble oligomeric forms of amyloid-β (Aβ) (14), which in turn leads to the dysfunction and degeneration of the elements subserving cognition, including neuronal cells and their synapses in the brain (5). In the past decade, several groups have documented the effects of various putative endogenous soluble forms of Aβ oligomers, most notably dimers, trimers, and Aβ*56, on memory (68) and on its presumed physiological substrate long-term potentiation (911).

The Aβ assembly, called Aβ*56, was originally identified in brain tissue of young amnestic Tg2576 mice overexpressing a mutant form of the human amyloid precursor protein (APP) used as a model of AD (8). Similar observations from several independent groups validated the existence of this Aβ species in other cognitively impaired APP transgenic mouse models (1215). In addition, brain infusion of Aβ*56 purified from APP mouse brain tissue caused transient memory deficits in young healthy rodents, demonstrating the memory impairing capability of this Aβ oligomer (8). Recent studies further confirmed the presence of Aβ*56 in postmortem human brain tissue and cerebrospinal fluid (16). In these cross-sectional studies, an abnormal increase in abundance of this Aβ oligomer in the brain was seen in samples from subjects in their fifth decade of life, preceding increases in Aβ dimers and trimers by two decades and coinciding with the age at which subtle cognitive deficits first appear (17). Notably, this increase in brain Aβ*56 was associated with aberrantly increased phosphorylation and missorting of tau typically seen in the early stages of the symptomatic phase of AD (16). Overall, these findings indicate that the Aβ oligomer Aβ*56 might alter memory and neuronal function during the presymptomatic phase of AD despite recent independent reports of a potential link between a putative Aβ dodecamer and AD vulnerability in the temporal cortex (18). To further understand the role of Aβ*56 in AD and to develop potential strategies aiming at countering its deleterious effects on cognition, we sought to identify the molecular mechanism by which Aβ*56 disrupts tau biology and neuronal physiology.

RESULTS

Aβ*56 interacts with N-methyl-d-aspartate receptors in an age-dependent manner

We previously reported that Aβ*56 can be detected in brain lysates (enriched in extracellular proteins) from Tg2576 mice starting at 6 months of age (8). This oligomeric Aβ assembly can also be found in membrane-associated lysates containing postsynaptic density (PSD) proteins, including PSD-95 (2, 13, 19). The presence of Aβ*56 in this compartment implies a possible binding of this Aβ molecule to a putative receptor, thereby transducing a deleterious intracellular biological signal. Because Aβ*56 causes memory impairment, we sought to determine whether Aβ*56 could interact with glutamatergic receptors, because they are critically involved in the molecular substrate of memory at a cellular level. We therefore performed coimmunoprecipitation experiments on membrane extracts using various antibodies against receptors previously described as interacting with synthetic Aβ oligomers as well as distinct subunits of the glutamatergic receptors. These included the following: N-methyl-d-aspartate receptor (NMDAR) subunits GluN1, GluN2A, and GluN2B; α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) subunits GluA1 and GluA2; α7-acetylcholine receptor subunit (α7); metabotropic glutamate receptor 5 (mGluR5); the receptor tyrosine kinase Ephrin B2; cellular prion protein (PrPC); and the receptor for advanced glycation end-products (RAGE) (Fig. 1A). We observed that only antibodies against GluN1, GluN2A, and GluN2B pulled down a 6E10-immunoreactive species consistent with Aβ*56 (Fig. 1A). The relative amount of Aβ*56 detected varied across immunoprecipitations with different GluN antibodies and appeared to reflect the abundance of the respective receptor subunits. In addition, no other Aβ species were immunoprecipitated as detected by 6E10. Notably, under our experimental conditions, antibodies targeting other glutamatergic receptors were unable to capture soluble Aβ species (Fig. 1A). By contrast, dimeric Aβ was pulled down using PrPC antibodies, as previously reported by our group (2), whereas monomeric Aβ coimmunoprecipitated with RAGE. Using either membrane-associated or extracellular-enriched protein fractions known to contain Aβ*56 but different amounts of the NMDAR subunit GluN1, we confirmed that Aβ*56 coimmunoprecipitated with GluN1 in membrane-associated extracts of 15-month-old Tg2576 mice (fig. S1A). Reverse coimmunoprecipitations with two Aβ antibodies, 6E10 or 4G8, further supported our initial results (fig. S1, B and C). Similar findings were observed using the A11 antibody raised against nonfibrillar amyloid oligomers (Fig. 1B), confirming the oligomeric nature of the Aβ assembly pulled down by GluN1. In addition, the relative abundance of putative Aβ*56-GluN1 complexes in membrane-associated lysates of Tg2576 forebrain tissue increased with aging, comparing transgenic mice from 2 to 24 months of age (Fig. 1C and fig. S1D). To demonstrate that these observations are not restricted to the Tg2576 line, we performed immunoprecipitations of GluN1 complexes in a different APP transgenic mouse line with lower APP expression [J20 line; (20)]. We selected animals at 3 and 6 months of age because they were previously described to display cognitive deficits and to express Aβ*56 in their brain tissue (13, 21). As predicted, we detected Aβ*56-GluN1 complexes in membrane-associated lysates from young J20 animals (fig. S1E). To validate these observations and assess their specificity, we measured the relative abundance of APP and/or Aβ, A11-positive type I Aβ oligomers, and OC (amyloid fibril antibody)–positive type II oligomers in membrane-associated lysates by nondenaturing dot blotting (22). We found that A11-immunoreactive species rose with age starting at 7 months. This profile contrasted with that observed for both APP/Aβ and type II oligomers (fig. S2). Together, these biochemical analyses of Tg2576 and J20 mice suggested the existence of an Aβ*56-NMDAR complex.

Fig. 1 Coimmunoprecipitation and colocalization of Aβ*56 with the NMDAR subunit GluN1.

(A) Coimmunoprecipitation of Aβ*56 with NMDAR subunits (GluN1, GluN2A, and GluN2B), AMPAR subunits (GluA1 and GluA2), α7-nicotinic acetylcholine receptor subunit (α7), mGluR5 (R5), or Ephrin B2 (B2) in membrane extracts from the forebrain of Tg2576 mice. Aβ was detected with 6E10. Blot is representative of three experiments (n = 6 mice per group). IP, immunoprecipitation; WB, Western blot; IgG, immunoglobulin G. (B and C) Western blots (B) and quantitation (C) of coimmunoprecipitation of Aβ*56 with GluN1 in membrane extracts from Tg2576 mice and a wild-type (WT) control. Antibody A11 was used to detect oligomeric Aβ. Data are means ± SD from n = 5 mice per group. P < 0.05 versus 5-month-old WT mice, P < 0.05 versus 7-month-old Tg2576 mice, by two-way analysis of variance (ANOVA) (F4,30 = 86.9203, P < 0.0001) followed by Student’s t test. Tg, transgenic; m, month. (D) Representative confocal images of Aβ*56 (green) binding to GluN1 (red) on WT or Prnp-null (Prnp−/−) primary cortical neurons. Neurons were also labeled for the dendritic neuronal marker microtubule-associated protein 2 (MAP2) (blue). n = 6 dishes per group. Veh., vehicle. (E) Software-assisted colocalization analysis of Aβ*56 and GluN1 on WT and Prnp-null neurons [six regions of interest (ROIs) per dish; n = 6 dishes per group]. (F) Western blots comparing GluN1 protein amounts in primary neurons and in HEK293 cells expressing GluN1. (G) Representative confocal images of Aβ*56 (A11; magenta) binding to HEK293 cells transfected with GluN1 (red) and/or GluN2B–enhanced green fluorescent protein (green). Arrowheads indicate colocalization between Aβ*56 and GluN1. n = 6 to 8 dishes per condition. Scale bars, 15 μm.

To demonstrate that Aβ*56 might directly interact with NMDAR subunits, we first isolated NMDARs from brain tissue of middle-aged 15-month-old Tg2576 mice by immunoaffinity capture and segregated putative NMDAR complexes by size exclusion chromatography (SEC; fig. S1, F to H). We observed that Aβ*56 coeluted with GluN1 in SEC fractions at molecular weights consistent with those of heteromeric NMDARs. Furthermore, nondenaturing dot blot analysis showed reactivity for A11 and GluN1 in the same fraction, proving the interaction between a type I Aβ oligomer (22) and NMDAR. Next, we applied Aβ*56 purified from APP transgenic mouse brains onto primary cortical neurons and human embryonic kidney (HEK) 293 cells transfected with GluN1 and/or GluN2B. Vehicle Aβ and monomeric Aβ were used as negative controls. Preparations of isolated Aβ species including monomers and Aβ*56 derived from APP transgenic animals were obtained using a modified protocol previously described for purifying endogenous Aβ oligomers from human brain tissue (fig. S3) (2). Purified Aβ*56 was applied to cells for 60 min, as previously reported (2). Quantitative reverse transcription polymerase chain reaction (RT-PCR) analysis of GluN mRNAs and dendritic spine imaging using various fluorescent reporters validated the maturity of cultured cortical neurons (fig. S4), consistent with earlier reports (2326). We observed that Aβ*56 readily colocalized with GluN1 on the surface of primary cortical neurons (Fig. 1D). Because some Aβ species bind to PrPC (2, 2729), we also analyzed Aβ*56 binding to GluN1 in Prnp-null neurons and found that binding occurs in a PrPC-independent manner (Fig. 1, D and E). Finally, we found that Aβ*56 bound to HEK293 cells transfected with GluN subunits, in which GluN expression was comparable to that detected in primary neurons (Fig. 1F). To assess which subunit might be responsible for the interaction, we subsequently transfected cells with individual GluN subunits. After the application of Aβ, Aβ*56 readily colocalized with GluN1 but not with GluN2B, indicating that Aβ*56 is binding to the NMDAR through a direct interaction with GluN1 (Fig. 1G). These results suggest that GluN1 and Aβ*56 might interact directly in brain tissue.

Aβ*56 enhances synaptic NMDAR-dependent calcium influx

Next, we sought to examine the effect of Aβ*56 on synaptic NMDAR-dependent calcium influx. Calcium transients were visualized in mouse cortical neurons transfected with the genetically engineered calcium sensor GCaMP6f (30), and bicuculline/4-aminopyridine (Bic4AP) was applied onto cells for 15 s to stimulate the NMDAR, because this paradigm was previously shown to selectively activate synaptic NMDARs in primary cultured neurons (26, 31). In the absence of Aβ*56, the Bic4AP pulse induced a lasting elevation of GCaMP6f fluorescence (Fig. 2, A and B). In the presence of 2.5 pM Aβ*56 (a 30-min pretreatment followed by a 5-min recording period), application of Bic4AP resulted in an enhanced and sustained influx of calcium mediated by synaptic NMDARs (Fig. 2, A and B). Upon calculating the peak amplitude and the global magnitude of the recorded responses (Fig. 2, C and D), Aβ*56 was found to potentiate synaptic NMDAR-induced Ca2+ influx by ~4.5-fold compared to cells exposed only to vehicle (a respective 4.99 ± 0.16– and 3.94 ± 0.16–fold increase compared to control cells).

Fig. 2 Aβ*56 enhances synaptic NMDAR-dependent calcium transients in primary cultured neurons.

(A) Representative confocal images for GCaMP6f-transfected neurons in the presence or absence of Aβ*56 at rest or after stimulation of synaptic NMDARs with Bic4AP. Scale bars, 20 μm. (B) Fluorescence responses of GCaMP6f-transfected neurons after synaptic NMDAR activation in the presence (red) or absence (black) of Aβ*56. Bold solid lines correspond to the average response; the flanking upper and lower gray shaded areas indicate SD. The black bar indicates the exposure of Bic4AP. A.U., arbitrary units. (C and D) Quantitation of the peak maximum GCaMP6f fluorescence (C) and area under the curve (D) in neurons exposed to Aβ*56 after simulation of synaptic NMDARs. F1,13 = 21.122 and F1,13 = 22.306, respectively. P < 0.05, Student’s t test; n = 6 to 9 cells per group. CTL, control. (E) Longitudinal fluorescence changes within GCaMP6f-transfected neurons in the absence of Aβ*56 (bars 1 to 3) or after a 15-min application of Aβ*56 (bars 5 to 9). Each bar corresponds to sequential bath stimulations. Histograms show means ± SD; one-way ANOVA (F8,51 = 9.4731, P < 0.0001) followed by Student’s t test, P < 0.05 versus Bic4AP (stimulation #2), P < 0.05 versus Bic4AP post-Aβ*56 (stimulation #5); n = 8 cells per group. (F) Mean Ca2+ responses in cortical neurons consecutively exposed to Bic4AP, Bic4AP post-Aβ*56 application, and Bic4AP + MK801. Histograms show means ± SD; one-way ANOVA (F2,40 = 14.7673, P < 0.0001) followed by Student’s t test, P < 0.05 versus Bic4AP, P < 0.05 versus Bic4AP post-Aβ*56; n = 16 responses per group.

To demonstrate that Aβ*56 specifically enhances Ca2+ currents mediated by NMDARs, we exposed the cells to a series of sequential bath stimulations with Bic4AP before or after a 15-min exposure to 2.5 pM Aβ*56; then, we subjected the cells to Bic4AP/MK801 stimulations to block opened synaptic NMDARs. Finally, the same cells were exposed to another Bic4AP bath stimulation to ensure that synaptic NMDARs were blocked (Fig. 2E). Adding the NMDAR antagonist MK801 led to an 82% reduction in NMDAR-mediated Ca2+ influx in cells exposed to Aβ*56 (Fig. 2, E and F). The lack of potentiated Ca2+ influx triggered by the subsequent Bic4AP stimulation indicated that extrasynaptic NMDARs were unlikely to be involved in the potentiation induced by Aβ*56 (Fig. 2, E and F). Counterintuitively, these findings are consistent with a selective activation of synaptic NMDARs by Aβ*56.

Aβ*56-NMDAR complexes selectively activate Ca2+-dependent calmodulin kinase IIα

NMDAR-mediated neuronal responses depend on the composition and subcellular localization of the receptors at the plasma membrane (synaptic or extrasynaptic) and involve distinct, opposing, or overlapping signaling pathways (32). Briefly, the activation of extracellular signal–regulated kinases (ERKs), adenosine 3′,5′-monophosphate (cAMP) response element–binding protein (CREB), and Ca2+-dependent calmodulin kinases (CaMKs) has been traditionally linked to mediating the effect of synaptic NMDARs. By contrast, activation of p38 kinase and Forkhead box protein O (FOXO) as well as inhibition of ERK and CREB are believed to be the downstream effectors of extrasynaptic NMDAR (32). In addition, we included analyses of the Src kinase Fyn, considering its proposed involvement in mediating Aβ-induced toxicity (2, 27, 28, 3336). We therefore assessed key components of these major pathways in Tg2576 mice at an age when Aβ*56 is present (7 months) or absent (4 months). There was no change of phosphorylation in the cell survival–promoting kinases ERKs and CREB at Ser133 or in the cell death–inducing kinases p38 at Thr180/Tyr182 and Fyn at Tyr416 (fig. S5). Because gene expression activity of the FOXO member FOXO-1 relies on nuclear translocation, we examined its relative abundance in extracts enriched in intracellular proteins and in those containing nuclear and membrane-bound proteins (8, 19). Similar to other intracellular messengers, we did not observe any overt changes in the biochemical segregation of FOXO-1 across ages and genotypes (fig. S5E).

One of the major consequences of the activation of NMDARs is the influx of extracellular calcium ions (Ca2+), thereby triggering Ca2+-dependent signaling molecules, including the calmodulin-CaMK axis (32). The finding that the CaMKIIα isoform has previously been linked to neuronal toxicity in neurodegenerative disorders (37, 38), together with our results showing no changes in CREB phosphorylation mediated by CaMKIV in 7-month-old Tg2576 mice, led us to assess CaMKIIα activity by measuring its activating phosphorylation state at Thr286. This phosphorylation (hereafter denoted as pCaMKIIα) increased ~2.5-fold in intracellular-enriched lysates of 7-month-old Tg2576 mice compared to younger 4-month-old Tg2576 mice or age-matched wild-type controls (Fig. 3, A and B). Because Tg5469 animals overexpress human wild-type APP to an extent similar to that of mutant APP in Tg2576 mice (39), we also included Tg5469 mice to our comparative analyses to evaluate potential effects of transgene-derived APP on CaMKIIα. Contrary to that in Tg2576 animals, the relative phosphorylation of CaMKIIα at Thr286 in Tg5469 mice was indistinguishable from that of nontransgenic mice, indicating that overexpression of soluble APP-α was not responsible for increasing the observed pCaMKIIα abundance (Fig. 3, A and B). Aβ*56 abundance correlated with that of pCaMKIIα in 6- to 9-month-old transgenic animals, with no changes observed for total CaMKIIα abundance (fig. S6). Using confocal imaging, pCaMKIIα immunoreactivity was markedly enhanced in the synaptic fields of Tg2576 prefrontal cortex and CA1 pyramidal neurons at 7 months of age (Fig. 3C). In both brain regions, the observed subcellular distribution of pCaMKIIα was consistent with the translocation of CaMKII from the soma to the synapse once phosphorylated (40). On the basis of this observation, we biochemically assessed whether pCaMKIIα was accumulating in PSD-containing lysates (membrane-associated fraction) of Tg2576 mice over time. We found that pCaMKIIα accumulated in an age-dependent manner in this fraction (Fig. 3, D and E), further validating the confocal imaging analyses. Overall, these results suggest a supraphysiological activation of CaMKIIα at synapses associated with the onset of Aβ*56 detection in Tg2576 mice.

Fig. 3 CaMKIIα is abnormally phosphorylated at Thr286 in the brain tissue of 7-month-old Tg2576 mice and in cortical neurons treated with Aβ*56.

(A and B) Representative Western blots (A) and densitometry analysis (B) for pThr286-CaMKIIα and CaMKIIα in intracellular (IC) protein extracts of 4- and 7-month-old Tg2576 or age-matched WT and Tg5469 mice. Data are means ± SD; two-way ANOVA (F7,28 = 38.7825, P < 0.0001) followed by Student’s t test, P < 0.05 versus 4-month-old WT mice; n = 6 to 9 mice per group. (C) Confocal imaging analysis of pThr286-CaMKIIα abundance and subcellular localization in prefrontal cortex (PFC) and CA1 pyramidal neurons of 7-month-old WT and Tg2576 mice (n = 5 mice per group). Scale bars, 20 μm. (D and E) Representative Western blots (D) and quantitation (E) for pT286-CaMKIIα and CaMKIIα in membrane-bound (MB) protein extracts of 4-, 7-, 12-, and 16-month-old Tg2576 mice. Histograms show means ± SD; one-way ANOVA (F3,24 = 30.4023, P < 0.0001) followed by Student’s t test, P < 0.05 versus 4-month-old WT mice, P < 0.05 versus 7-month-old Tg2576 mice; n = 6 per group. (F and G) Western blots (F) and densitometry analysis (G) for pT286-CaMKIIα and total CaMKIIα in 12- to 14-DIV (day in vitro) primary mouse cortical neurons treated with vehicle or increasing concentrations of brain-derived Aβ*56 for 60 min. Histograms show means ± SD; ANOVA (F4,30 = 14.6822, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle, P < 0.05 versus 1 pM condition; n = 6 to 8 per group. (H and I) Western blot images (H) and quantitation (I) for pT286-CaMKIIα and total CaMKIIα in 12- to 14-DIV primary mouse cortical neurons treated with 2.5 pM brain-derived Aβ*56 for 1, 6, 8, 12, or 24 hours. Histograms show means ± SD; ANOVA (F4,34 = 17.4461, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle, P < 0.05 versus 1-hour condition; n = 6 per group. (J and K) Representative Imaris surface images (J) and quantitation (K) of the colocalization of pCaMKIIα with PSD-95 (yellow) with respect to MAP2 (blue) in neurons treated with vehicle or 2.5 pM Aβ*56 for 60 min. Histograms show means ± SD; Student’s t test, F1,14 = 37.339, P < 0.05 versus vehicle; n = 8 ROIs per group. Scale bars, 3 μm.

To demonstrate that Aβ*56 directly caused a selective exacerbation of CaMKIIα activity, we applied soluble Aβ oligomers purified from APP transgenic mouse brains at pathophysiologically relevant concentrations (picomolar to nanomolar range) to primary cultured cortical neurons. Preparations of apparent Aβ monomers, dimers, trimers, and Aβ*56 derived from transgenic animals were obtained using a protocol previously described for purifying endogenous Aβ oligomers from human brain tissue (fig. S3) (2). Because of the low abundance of Aβ*56 in brain tissue (8, 22), Aβ*56 was applied to cells in a dose-dependent manner, ranging from 1 to 25 pM. By contrast, low–molecular weight soluble Aβ species were applied at concentrations of 1 to 5 nM, previously found to be biologically active in our in vitro system (2). To attempt inducing a sustained effect on CaMKIIα that might recapitulate the exposure of neurons to Aβ*56 in vivo, primary cortical neurons were exposed to each purified Aβ preparation for 60 min, as previously described (Fig. 3F and fig. S7) (2). Whereas baseline levels of pCaMKIIα were readily detected in vehicle-treated neurons, application of increasing concentrations of Aβ*56 induced a dose-dependent potentiation of CaMKIIα phosphorylation, which plateaued at 5 pM to an average of 2.89-fold of the baseline (Fig. 3, F and G). On the basis of these data, we chose to use Aβ*56 at a concentration of 2.5 pM for all subsequent experiments. Time course experiments using 2.5 pM Aβ*56 indicated that CaMKIIα activity peaked after 1 hour of treatment and declined to baseline within 8 hours (Fig. 3, H and I). Consistent with the increase in CaMKIIα phosphorylation observed by Western blotting, confocal immunofluorescence analyses revealed both an increase in pCaMKIIa immunoreactivity (fig. S8A) and a translocation of pCaMKIIα to PSDs in neurons treated with Aβ*56 for 60 min (fig. S8, A to C). Z-stack reconstruction analyses confirmed a 2.53-fold increase in pCaMKIIα colocalization with PSD-95 in neurons exposed to Aβ*56 (Fig. 3, J and K). Furthermore, assessment of neuronal cell death using lactate dehydrogenase (LDH) assays indicated that Aβ*56 was not cytotoxic at the concentrations used in our experimental conditions (fig. S9, A to C). Different from Aβ*56, applications of 1 nM Aβ monomers, dimers, or trimers onto cortical neurons did not alter CaMKIIα phosphorylation compared to vehicle-exposed cells (fig. S9, D and E). Finally, in vitro applications of Aβ*56 did not trigger the activation of ERK, CREB, p38, or Fyn kinases (fig. S7), similar to what was observed in 7-month-old Tg2576 animals. These findings indicate that Aβ*56 specifically activates CaMKIIα in vitro.

Aβ*56-induced CaMKII activation leads to tau hyperphosphorylation and missorting

One substrate of CaMKII is the tau protein. CaMKII phosphorylates tau at various serine residues including Ser262, Ser409, and Ser416, the phosphorylation of which is increased in AD brain tissue (41, 42). Of particular interest, phosphorylation at Ser416 has been proposed to induce a conformational change in the tau protein (41). We therefore examined Tg2576 mice for CaMKII-related changes in tau along with other changes commonly associated with AD (Fig. 4A; n = 6 to 9 animals per age per genotype). Across transgene and age groups, no obvious changes in phosphorylation were detected at Tyr18, Ser262, Ser396/Ser404, and Ser409 (Fig. 4, A and B). However, we detected a ~2.7-fold increase in phosphorylation at Ser202 and Ser416 in 7-month-old Tg2576 forebrains compared to nontransgenic littermates (Fig. 4, A and B). In addition, we observed a delayed electrophoretic migration of pSer416-tau molecules, resulting in the detection of two additional bands of ~55 and 60 kDa, as previously documented (Fig. 4A) (41), which tau epitope and dephosphorylation assays confirmed to be putative hyperphosphorylated 0N4R tau conformers (fig. S10) (43).

Fig. 4 Hyperphosphorylation and missorting profile of soluble tau species in young Tg2576 mice.

(A and B) Representative Western blots (A) and quantitation (B) of soluble tau species detected in intracellular (IC)–enriched fractions from 4- and 7-month-old WT and Tg2576 mice. Histograms show means ± SD; two-way ANOVA (F2,21 = 67.6019, P < 0.0001) followed by Student’s t test, P < 0.05 versus age-matched WT mice, P < 0.05 versus 4-month-old Tg2576 mice; n = 6 to 9 mice per group. (C and D) Western blots (C) and densitometry analysis (D) of total soluble tau, PSD-95, and actin in membrane-bound (MB) protein extracts of Tg2576 mice at 4, 7, and 12 months of age. Histograms show means ± SD; one-way ANOVA (F2,18 = 19.7636, P < 0.0001) followed by Student’s t test, P < 0.05 versus 4-month-old Tg2576 mice, P < 0.05 versus 7-month-old Tg2576 mice; n = 6 to 9 mice per group. (E and F) Western blots (E) and quantitation (F) of total soluble tau in intracellular-enriched (I) or membrane extracts (M) of 4- and 7-month-old Tg2576 mice. Histograms show means ± SD; two-way ANOVA (F3,30 = 47.2095, P < 0.0001) followed by Student’s t test, P < 0.05 versus 4-month-old Tg2576 mice; n = 6 to 9 mice per group. (G) Representative confocal images of CA1 hippocampal neurons immunostained for MAP2 (blue), pSer202-tau (CP13; green), and pSer416-tau (red) revealed an aberrant accumulation and differential missorting of soluble tau species in 7-month-old Tg2576 mice. (H) Z-stack reconstruction from confocal images illustrating the colocalization for pSer202-tau and pSer416-tau (yellow), shown with the three-dimensional rendering of MAP2. n = 6 sections per animal; n = 3 to 6 animals per group. Scale bars, 20 μm (top and middle) or 10 μm (bottom) in (G) and 3 μm in (H).

We next evaluated whether similar changes in CaMKIIα and tau phosphorylation occurred in a second APP transgenic mouse model of AD, the J20 line (fig. S11) (20). Because J20 mice were previously shown to generate Aβ*56 at 3 to 6 months of age when spatial cognitive deficits are first detected (13), we analyzed the CaMKIIα-tau axis in these mice at 3 months of age. After confirming the presence of Aβ*56 in the forebrain tissue of APP transgenic animals, we observed that the phosphorylation of Thr286-CaMKIIα, Ser202-tau, and Ser416-tau was selectively increased in a similar fashion as that of Tg2576 mice. These findings therefore support the notion that the specific alteration of the CaMKIIα-tau axis by Aβ*56 may be a general feature of APP transgenic mouse models of AD.

Consistent with this apparent specificity in the observed pattern of tau hyperphosphorylation, we observed that neither Cdk5 (cyclin-dependent kinase 5) nor GSK3β (glycogen synthase kinase 3β), two major tau kinases, was abnormally activated in 7-month-old Tg2576 mice or in primary cortical neurons exposed to Aβ*56 (fig. S12). Accordingly, none of the additional tau sites linked to the aforementioned kinases (Ser396, Ser409, and Ser404) were hyperphosphorylated in 7-month-old Tg2576 mice (Fig. 4, A and B), arguing against their potential involvement in mediating the initial signaling response induced by Aβ*56.

Because tau missorting to the PSD alters synaptic function (23, 44, 45), we measured total tau levels in membrane-associated extracts obtained from forebrains of Tg2576 mice at 4, 7, and 12 months of age containing PSD-95, as described elsewhere (2). First, we observed that both total tau and pSer416-tau accumulated with age in membrane-associated lysates from Tg2576 mice. Second, we found that PSD-95 protein levels were decreasing with aging in Tg2576, paralleling the elevation of tau in this compartment (Fig. 4, C and D). To assess whether tau abnormally cosegregated with PSD-95, we measured the membrane-associated/intracellular extract ratio of total tau species at 4 and 7 months, which revealed a nearly fivefold rise of tau in the membrane-bound protein fraction of 7-month-old Tg2576 animals (Fig. 4, E and F). To support these biochemical changes, we performed immunofluorescent labeling followed by confocal imaging analyses (Fig. 4G). In CA1 pyramidal neurons of 7-month-old Tg2576 mice, there was a marked increase of pSer202-tau in the soma and dendrites of the stratum radiatum (Fig. 4G). By contrast, pSer416-tau was nearly exclusively detected in dendrites (Fig. 4G), matching the distribution of pCaMKIIα (Fig. 3C). Because pSer202-tau and pSer416-tau were both present in the synaptic fields of CA1 neurons (Fig. 4G), we wondered whether both tau modifications were colocalized or spatially cosegregated. Using software-based analysis of z stacks, a colocalization channel between pSer202-tau and pSer416-tau was created. Under these settings, it was apparent that pSer202-tau and pSer416-tau colocalized within dendritic spines in transgenic mice (Fig. 4G). These results indicate that the ~2-fold elevation in CaMKIIα activity observed in Tg2576 at ages when Aβ*56 starts forming is associated with ~2.5-fold increase in tau hyperphosphorylation at Ser202/Ser416 and missorting of distinct tau species into spines.

To demonstrate that Aβ*56 is triggering these pathological changes in tau, we exposed primary cortical neurons to increasing concentrations of Aβ*56 previously shown to activate CaMKIIα and examined the phosphorylation and missorting status of tau (Fig. 5). Mirroring the data observed in 7-month-old Tg2576 mice, tau phosphorylation was unaltered at Tyr18, Ser262, Ser396/Ser404, and Ser409 in the presence of Aβ*56 (Fig. 5, A and B). In contrast, neuronal levels of soluble pSer202-tau and pSer416-tau rose sharply in a dose-dependent manner in the presence of increasing amounts of Aβ*56 (Fig. 5, A and B). It is worth noting that although purified Aβ dimers and trimers can trigger a Fyn-mediated phosphorylation of tau at Tyr18 (2), neither species induced hyperphosphorylation of tau at Ser202 and Ser416 (fig. S13). Furthermore, we recently reported that Aβ trimers could trigger the formation of Alz50-tau conformers in cultured mouse primary cortical neurons in a selective fashion (46). We therefore measured Alz50-tau abundance in the absence or presence of increasing concentrations of Aβ*56 (0 to 25 pM), and we failed to observe any changes in Alz50-positive tau conformers. In parallel, an Aβ*56-mediated ~2- to 2.5-fold accumulation of tau was observed in membrane-associated lysates containing PSD-95 (Fig. 5, C and D). Last, we assessed whether these tau changes induced by Aβ*56 could also apply to human tau by using Htau primary neurons overexpressing human tau isoforms in the absence of mouse tau. Htau neurons were exposed to Aβ*56 for 60 min, and biochemical analysis of tau species present in the intracellular protein–enriched extracts revealed a ~3.5-fold elevation of pSer416-tau (Fig. 5, E and F). These in vitro findings demonstrate that, unlike Aβ dimers or Aβ trimers, Aβ*56 causes highly selective pathological changes in the tau protein at Ser202 and Ser416 residues.

Fig. 5 Selective tau hyperphosphorylation in primary neurons exposed to Aβ*56.

(A and B) Western blots (A) and quantitation (B) of soluble tau species detected in mouse cortical neurons exposed to increasing concentrations of Aβ*56 for 60 min. Histograms show means ± SD; one-way ANOVA (F2,21 = 67.6019, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle-treated neurons, P < 0.05 versus 1 pM Aβ*56 condition; n = 6 to 8 dishes per treatment. (C and D) Western blots (C) and densitometry analysis (D) for total soluble tau detected with the antibody tau5, PSD-95, and actin in membrane-associated extracts from vehicle or Aβ*56-treated neurons. Histograms show means ± SD; one-way ANOVA (F2,21 = 67.6019, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle-treated neurons, P < 0.05 versus 1 pM Aβ*56 condition; n = 6 to 8 per group. (E and F) Western blots (E) and quantitation (F) for pS416-tau, total soluble tau detected with the antibody tau5, and actin in intracellular-enriched lysates of vehicle- or Aβ*56 (2.5 pM)–treated neurons. Histograms show means ± SD, P < 0.05 versus vehicle-treated neurons by t test; n = 6 dishes per group.

Aβ*56-induced tau hyperphosphorylation at Ser416 is dependent on CaMKIIα activity

To demonstrate that NMDAR activity is required to mediate the effects of Aβ*56, primary neurons were pretreated with the NMDAR-PSD uncoupling peptide tatNR2B9c (47, 48) in the presence or absence of Aβ*56. We found that disrupting the interaction between NMDAR and PSD-95 prevented the downstream phosphorylation of CaMKIIα (Fig. 6, A and B, n = 4 to 6 dishes per group, and fig. S14) and inhibited the hyperphosphorylation of tau at Ser202 and Ser416 induced by Aβ*56 (Fig. 6, C and E), reminiscent of the protection conferred by the peptide from Aβ-induced toxicity (49).

Fig. 6 Inhibiting CaMKII prevents Aβ*56-induced tau hyperphosphorylation at S416.

(A and B) Western blots (A) and quantitation (B) for pCaMKIIα and total CaMKIIα in primary cortical neurons pretreated with the NMDAR uncoupling peptide tatNR2B9c for 15 min in the presence or absence of 2.5 pM Aβ*56. Histograms show means ± SD; one-way ANOVA (F3,14 = 252.0481, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle, P < 0.05 versus Aβ*56-treated neurons; n = 4 to 6 dishes per group. (C to E) Western blots (C) and densitometry analysis (D and E) for pSer202-tau, pSer416-tau, total tau, and actin in primary cortical neurons pretreated with the NMDAR uncoupling peptide tatNR2B9c for 15 min in the presence or absence of 2.5 pM Aβ*56. Total tau was detected with the antibody tau5. Histograms show means ± SD; one-way ANOVA (F3,14 = 22.6029, P < 0.0001, and F3,12 = 16.1364, P = 0.0009, respectively) followed by Student’s t test, P < 0.05 versus vehicle, P < 0.05 versus Aβ*56-treated neurons; n = 4 to 6 dishes per group. (F and G) Western blots (F) and quantitation (G) for pThr286-CaMKIIα and total CaMKII in primary cortical neurons pretreated with the CaMKII inhibitor tatCN21 in presence or absence of 2.5 pM Aβ*56. Histograms show means ± SD; two-way ANOVA (F3,35 = 25.0063, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle-treated neurons, P < 0.05 versus Aβ*56-treated neurons; n = 6 to 9 dishes per group. ANOVA results: Aβ*56 (F = 26.7966, P < 0.0001), tatCN21 (F = 16.2025, P = 0.0003), and Aβ*56 × tatCN21 interaction (F = 27.4058, P < 0.0001). (H to J) Western blots (H) and quantitation (I and J) for soluble pSer416-tau and total tau (as measured with the tau5 antibody) in mouse primary neurons pretreated with the CaMKII inhibitor tatCN21 in the presence or absence of 2.5 pM Aβ*56. Histograms show means ± SD; two-way ANOVA (F3,35 = 28.4569, P < 0.0001, and F3,35 = 24.8972, P < 0.0001, respectively) followed by Student’s t test, P < 0.05 versus vehicle-treated neurons, P < 0.05 versus Aβ*56-treated neurons; n = 6 to 9 dishes per group.

To further demonstrate that the downstream signaling cascade induced by Aβ*56 is dependent on CaMKIIα, we applied Aβ*56 to primary cultured neurons pretreated with tatCN21, a selective inhibitor of CaMKII (50), for 15 min (Fig. 6). On the basis of toxicity and target engagement assays, we chose to use this potent inhibitor at 1 μM, a concentration 2- to 10-fold lower than previous reports (51, 52). As previously described in independent experiments (Fig. 3 ), a 60-min exposure to Aβ*56 triggered a ~2.5-fold increase in CaMKIIα phosphorylation at Thr286 compared to vehicle-treated cells (Fig. 6, F and G). Pretreatment with 1 μM tatCN21 prevented the elevation in pThr286-CaMKIIα induced by Aβ*56 and, notably, restored CaMKIIα activity to that measured at baseline. As predicted, tatCN21 lowered the baseline levels of active CaMKIIα in the primary neurons.

Once the efficacy and the extent of the inhibition of CaMKIIα were established in the presence of Aβ*56, we evaluated a potential rescue of the tau phenotype (Fig. 6, H to J). In the presence of Aβ*56 alone, pS416-tau amounts increased by ~2-fold (2.03 ± 9.79) compared to the vehicle control, as shown in Fig. 5. In addition, a putative tau conformer was readily observed (Fig. 6H), reminiscent of those observed in vivo (Fig. 4). Inhibiting CaMKIIα with tatCN21 blocked the aberrant hyperphosphorylation of tau at Ser416 and prevented the observed electrophoretic shift of tau proteins (Fig. 6, H and I). Similar observations were obtained for tau phosphorylation at Ser202 (Fig. 6J).

We also addressed whether inhibiting CaMKIIα could block the apparent missorting of tau triggered by Aβ*56. Pretreating cortical neurons with tatCN21 completely abolished the Aβ*56-induced translocation of tau to dendrites and into dendritic spines (Fig. 7, A to C). Moreover, the biochemical segregation of tau into membrane-enriched compartments (2) further supports these findings (Fig. 7, D and E), indicating that tau hyperphosphorylation at Ser416 might be necessary to redistribute tau.

Fig. 7 CN21 pretreatment prevents the missorting of tau in cortical primary neurons exposed to Aβ*56.

(A) Representative confocal images of primary mouse cortical neurons immunostained for MAP2 (blue), pThr286-CaMKIIα (green), and pSer416-tau (magenta) after treatment with 2.5 pM Aβ*56 for 60 min. n = 6 dishes per group. Scale bars, 30 μm. (B) Surface rendering of dendrites labeled with MAP2, pSer416-tau, and PSD-95, illustrating the cellular distribution of the pS416-tau/PSD-95 colocalization channel (yellow) with respect to MAP2 (blue) in neurons treated with vehicle, 2.5 pM Aβ*56, or tatCN21 pretreatment (15 min), followed by 2.5 pM Aβ*56 for 60 min. Scale bars, 3 μm. (C) Quantitation of the colocalization of pS416-tau with PSD-95 in mouse primary neurons exposed to vehicle, 2.5 pM Aβ*56, or tatCN21 pretreatment, followed by 2.5 pM Aβ*56. Histograms show means ± SD; one-way ANOVA (F2,20 = 67.7832, P < 0.0001) followed by Student’s t test, P < 0.05 versus vehicle, P < 0.05 versus Aβ*56-treated neurons; n = 8 ROIs per group. (D and E) Western blots (D) and quantitation (E) for soluble tau in membrane-associated lysates from neurons exposed to vehicle, Aβ*56, CN21, or Aβ*56 + CN21 using the pan tau-specific antibody tau5. Actin was used as internal standard. Histograms show means ± SD; one-way ANOVA (F3,12 = 197.3191, P < 0.0001) followed by Student’s t test; P < 0.05 versus vehicle, P < 0.05 versus Aβ*56-treated neurons; n = 4 dishes per group.

DISCUSSION

Here, we found that Aβ*56, but not Aβ monomers, dimers, or trimers, coimmunoprecipitates with the NMDAR subunits GluN1, GluN2A, and GluN2B in Tg2576 mouse brain tissue in an age-dependent manner. Furthermore, we demonstrated that Aβ*56 directly bound to GluN1 and accentuated synaptic NMDAR-mediated Ca2+ influx, providing a potential mechanism by which the Aβ*56-NMDAR complex is functionally relevant. The detection of this Aβ*56-NMDAR complex was associated with abnormally increased activation of CaMKIIα kinase in young Tg2576 mice and in cultured neurons exposed to Aβ*56. It is interesting to note that the elevation in CaMKIIα activity measured in vitro (ranging from 1.74- to 3.01-fold increase) was consistent with the 3.4-fold increase in pT286-CaMKIIα detected in 7-month-old Tg2576 mice. Consistent with our current knowledge of CaMKII physiology, the activated kinase translocated to postsynaptic sites in both experimental systems (37). We also showed that CaMKIIα coupled Aβ*56 to the selective phosphorylation and missorting of tau, both in vivo and in vitro, and that this process could be blocked by inhibiting CaMKIIα directly or by uncoupling NMDAR from PSD-95.

Four important concepts emerge from this work. First, the memory-impairing effects of Aβ*56 do not appear to require cell death. We previously highlighted the role of Aβ*56 in causing memory deficits in rodents where neurodegeneration is absent (8). Here, we were able to recapitulate in vitro the phenotypic changes in CaMKIIα and tau observed in vivo, suggesting that the concentrations of Aβ*56 applied to cells were comparable to those likely to exist in brain tissue. We have not observed evidence of Aβ*56 inducing cell death in vitro or in vivo. This conclusion is supported by the absence of (i) an elevation of LDH release in neurons exposed to Aβ*56; (ii) modulations in intracellular messengers classically linked to neuronal cell death, such as the activation of p38 or reductions in ERK/CREB activity (32); and (iii) caspase-3 activation. Together, these findings are consistent with the accumulation of Aβ*56 in brain tissue of individuals in their 40s (16), in which subtle age-related memory decline begins in mid-to-late 30s (17).

Second, not all Aβ oligomers purified from brain tissue alter the same neuronal signaling pathways. We reported that the detection onset of Aβ*56 in the AD mouse model Tg2576 mice at 6 to 7 months of age is associated with a selective activation of CaMKIIα and not other kinases, as well as specific alterations in tau phosphorylation. Furthermore, we went on to demonstrate that exposure to picomolar concentrations of Aβ*56 was sufficient to mimic these changes in vitro without modulating intracellular messengers, such as the Src kinase Fyn, ERK, p38, Cdk5, GSK3β, FOXO-1, and CREB. By comparison, purified Aβ monomers, dimers, and trimers failed to activate CaMKIIα within the confines of our experimental settings. By contrast, we previously reported that both Aβ dimers and trimers purified from AD brain tissue activated Fyn in vitro, whereas Aβ*56 did not activate this Src kinase (2). On the other hand, synthetic mixtures of Aβ oligomers activate a combination of several of these kinases (28, 53, 54). Future studies are needed to determine how CaMKIIα is regulated in the presence of Aβ*56 and whether the Aβ*56-GluN1 complex is stabilized at the neuronal plasma membrane.

These observations also extend to tau phosphorylation. We previously showed that tau was hyperphosphorylated at Tyr18 in neurons treated with these low-n Aβ oligomers, whereas pY18-tau abundance was unchanged in the presence of Aβ*56 (2). More recently, we reported that Aβ trimers selectively induced a pathological conformation change of tau detected by the Alz50 antibody in vitro (46). Here, we found that Aβ dimers and trimers were not affecting tau phosphorylation at Ser416, whereas Aβ*56 triggered a nearly two- to threefold elevation in pS416-tau amounts. Similarly, Aβ*56 induced a dose-dependent hyperphosphorylation of tau at Ser202, whereas neither Aβ dimers nor trimers triggered this change. These results support the view that distinct oligomeric Aβ species exert different effects on neuronal signaling and tau biology. Recent work (3) suggested the involvement of CaMK kinase 2 (CaMKK2) in mediating the toxicity induced by synthetic oligomeric Aβ of unknown molecular size, culminating in the hyperphosphorylation of tau at Ser262. We did not observe a change of tau phosphorylation at Ser262 either in Tg2576 and J20 mice or in primary neurons exposed to endogenous oligomeric Aβ. Our results are inexplicably inconsistent with the aforementioned published data (3). However, our results are consistent with the current knowledge that CaMKK2 preferentially regulates the activity of CaMKI and CaMKIV, but not CaMKII (55), and with recent studies assessing the posttranslational modification of tau (56). In the latter report, the authors compared 32 different tau modifications in two independent experiments using wild-type and J20 APP transgenic mice. None differed significantly and consistently between young J20 and wild-type mice (including at S262) apart from tau phosphorylation at Ser416 in the PSD (56). There, pS416-tau was detected at higher frequency (more than twofold). Instead, we consistently found that Aβ*56 was inducing the selective hyperphosphorylation of tau at Ser202 and Ser416. The observed hyperphosphorylation of tau at Ser202 might appear as counterintuitive because it is not a specific substrate of CaMKIIα. In combination with other tau residues, Ser202 has classically been considered to be a substrate for many kinases including GSK3β, protein kinase A (PKA), Cdk5, and DYRKs (42). However, we demonstrated in cortical neurons that the hyperphosphorylation of tau at Ser202 required active CaMKIIα, indicating a hierarchical phosphorylation of tau induced by CaMKIIα, as previously reported for PKA (57). Functionally, the consequence of tau hyperphosphorylation at Ser202 by the Aβ*56/GluN/CaMKIIα axis leads to a well-established disruption of microtubule dynamics (58). Despite the recognition of the apparent specificity of tau posttranslational changes induced by Aβ*56 in tau, we also acknowledge that tau is abnormally phosphorylated at many other sites, which may be just as relevant to AD (42).

Third, the abnormal subcellular distribution of soluble tau species appears to depend on the presence of a distinct hyperphosphorylated epitope. Our dual in vivo labeling approach demonstrated that pS202-tau differed markedly in its cellular distribution from pS416-tau. In particular, the latter did not seem to accumulate in the soma where pS202-tau was abundant. Further studies are needed to determine why some hyperphosphorylated forms of tau are present in the soma and dendrites when other forms are mainly detected in dendrites in vivo.

Fourth, we previously reported that the abundance of Aβ*56 in human brain tissue declines with disease progression in a cross-sectional study (16). Accordingly, previous studies have indicated that CaMKIIα immunoreactivity or pCaMKIIα levels are drastically reduced in AD compared to age-matched controls (59, 60) and that pCaMKIIα is redistributed from the dendrites to the soma in mild cognitive impairment (MCI) and AD hippocampi (61). Although an exhaustive analysis of CaMKIIα activity in AD is needed to better understand the role of this kinase in pathogenesis, our studies are consistent with the notion that CaMKIIα may be overactive when Aβ*56 increases in preclinical AD and suppressed when Aβ*56 is lowered in MCI and AD.

Together, our in vivo and in vitro results indicate that different endogenous Aβ oligomers alter tau biology in a highly selective manner. It will be interesting in future studies to determine whether homeostatic and pathophysiological processes regulating CaMKIIα in the presence of Aβ*56 may contribute to the cellular phase before reaching end-stage AD.

Despite intensive efforts to isolate distinct soluble Aβ species and reporting differential effects of Aβ oligomers purified from human AD brains on neuronal signaling and tau (2, 46), it is also fair to indicate intrinsic limitations of our studies: (i) We cannot exclude the possibility that the relative abundance of these species might be altered ex vivo compared to their endogenous state in vivo because of the absence of Aβ assembly–specific reagents. (ii) For the same reason, we can also not rule out the possibility that the conformational state of the purified Aβ species applied onto neurons will be preserved for the duration of the exposure. (iii) We cannot assume that the concentrations of Aβ species used in vitro exactly match those found in vivo; related to this point, it is still unclear whether the current enzyme-linked immunosorbent assay–based approaches can adequately inform on this issue, because qualitative differences in Aβ species might be more critical for AD pathophysiology than quantitative differences in total Aβ oligomers (62, 63). (iv) The longitudinal profile of distinct Aβ oligomers in aging and in AD is unknown; consequently, the pathophysiological relevance of each separate Aβ assembly, including Aβ*56, in AD requires further confirmation despite independent reports of a potential link between a putative Aβ dodecamer and AD vulnerability in the temporal cortex (18). (v) We cannot exclude the possibility that other forms of Aβ oligomers, not captured by our approach, could affect the Aβ*56-activated NMDAR-CaMKIIα-tau pathway (64). Further work is needed to fill these technical and knowledge gaps, but these studies also provide unprecedented insights into differential molecular mechanisms between Aβ oligomers purified from brain tissue.

To conclude, not only are these findings reminiscent of a role of CaMKIIα dysfunction in brain disorders, such as Angelman syndrome (65), attention deficit hyperactivity disorder (66), and cerebral ischemia (37), but also they resonate with the emerging concept that CaMKIIα is altering neuronal physiology and cognition more generally when aberrantly overactivated in the brain. In combination with earlier studies (2, 46), these results establish that distinct endogenous Aβ oligomers activate specific neuronal signaling pathways and that mapping the specific tau changes induced by each of these Aβ toxins might provide a general template for monitoring AD progression. In this context, the recent failures of immunotherapies targeting Aβ in AD might be due to the fact that these antibodies do not bind the correct type of Aβ oligomers or bind specific Aβ oligomers with sufficiently high affinities (67).

Beyond the scope of AD research, we would like to argue that our proof-of-principle approach, that is, identifying the functional and mechanistic properties attributable to a distinct entity of amyloid oligomers, is also relevant for other neurodegenerative disorders including Parkinson’s and Huntington’s diseases, amyotrophic lateral sclerosis, frontotemporal dementias, and chronic traumatic encephalopathy, where oligomeric forms of amyloid proteins have been proposed to cause both synaptic and cellular toxicity. We should therefore strive to not consider all amyloid oligomers equally toxic but, instead, rigorously assess the role of each molecular assembly separately.

MATERIALS AND METHODS

Transgenic animals

Mice from the APP line Tg2576, which expresses the human APP with the Swedish mutation (APPKM670/671NL), directed by the hamster prion promoter (68), were purchased from Taconic Farms Inc. and bred to obtain wild-type and hemizygous animals. Mice from the J20, MAPT-null, and Htau lines (20, 69) were purchased from the Jackson Laboratory. J20 animals bred following the guidelines provided by the Mucke Laboratory. Both male and female mice were used in all experiments. All mice were group-housed by gender (aggressive animals were singly housed), given food and water ad libitum, and maintained on a 12-hour light/dark cycle (7:00 a.m./7:00 p.m.). None of the animals analyzed were excluded. All animal procedures and studies were reviewed and approved by the University of Minnesota Institutional Animal Care and Use Committee and Institutional Review Board.

Primary cell cultures

Mouse cortical cultures of neurons were prepared from 14- to 15-day-old embryos, as described previously (2, 70, 71), using 5 × 105 cells per dish. After 3 DIV, neurons were treated with 10 μM arabinosylcytosine to inhibit proliferation of nonneuronal cells. All experiments were performed on near-pure neuronal cultures [>98% of MAP2-immunoreactive cells] after 12 to 14 DIV. Three to nine 35-mm dishes per culture per condition were used across three independent experiments.

The concentration and duration of the pretreatments with tatCN21 or with tatNR2B9c were determined by dose response (1, 5, and 10 μM) and time course (15, 30, and 60 min) experiments using CaMKII and neuronal cell death estimated by LDH assay. Accordingly, pretreatments were set to 15 min at 1 μM for both peptides because longer durations and higher concentrations prove toxic to cells.

After treatment(s), cells were harvested in an ice-cold lysis solution containing 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.1% Triton X-100 (Sigma) with 1 mM phenylmethylsulfonyl fluoride (PMSF), 2 mM 1,10-phenanthroline monohydrate (1,10-PTH), 1% (v/v) mammalian protease inhibitor cocktail (Sigma), and 0.1% (v/v) phosphatase inhibitor cocktails A [Santa Cruz Biotechnology Inc. (SCBT)] and 2 (Sigma-Aldrich). Cell lysates were centrifuged for 10 min at 13,000g, supernatants were isolated, and corresponding pellets were resuspended with the protease/phosphatase inhibitor–containing lysis buffer to extract membrane-bound proteins. Plasma membranes were solubilized in radioimmunoprecipitation assay (RIPA) lysis buffer [50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Triton X-100, 1 mM EGTA, 3% SDS, 1% deoxycholate, 1 mM PMSF, 2 mM 1,10-PTH, 1% (v/v) mammalian protease inhibitor cocktail (Sigma), and 0.1% (v/v) phosphatase inhibitor cocktails A (SCBT) and 2 (Sigma-Aldrich)]. Membrane lysates were then subjected to centrifugation for 10 min at 16,000g, and the soluble fraction was removed and stored for analysis.

Neuron transfection

Primary cortical neurons cultures (13 DIV) were transfected with GCaMP6f (gift from L. Looger; Addgene #40755), actin-mCherry (#632589, Clontech Laboratories Inc.), and PSD-95 tagged with yellow fluorescent protein (PSD95-TS:YFP was a gift from R. Tsien; Addgene plasmid #42225) plasmids using the phosphate calcium technique. Cells were exposed to DNA (25 μg/ml) for 30 min. Cells were washed and returned to preconditioning medium supplemented with 5% fetal bovine serum, 5% horse serum, and 1% glutamine. Experiments were performed on 14-DIV cultures. Protrusions with a head localized at less than 5 μm from the dendritic shaft were considered as spines. Our cortical cell cultures display an average of six spines per 10-μm dendrites.

Calcium imaging

Experiments were performed as described earlier (26), except that the genetically engineered calcium indicator GCaMP6f (gift from L. Looger, Addgene) replaced Fura2/acetoxymethyl.

Protein extractions

To analyze Aβ species, we harvested one dissected hemi-forebrain per animal and used two extraction protocols described elsewhere (8, 9, 21). Extracellular-enriched protein extracts refer to protein lysates obtained after the first step of a serial extraction with a lysis buffer composed of 50 mM tris-HCl (pH 7.6), 150 mM NaCl, 0.01% NP-40, 2 mM EDTA, and 0.1% SDS. Samples were then centrifuged at 800g for 10 min at 4°C to separate extracellular lysates from the remaining protein pools [see (21) for details]. In addition, membrane-enriched protein extracts refer to protein lysates obtained after the third step of a serial extraction with a lysis RIPA buffer composed of 50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Triton X-100, 1 mM EDTA, 3% SDS, and 1% deoxycholate. Samples were then centrifuged at 16,100g for 90 min. Supernatants were collected, and pellets were further extracted with formic acid to analyze fibrillar/deposited proteins. It is possible that the use of the RIPA lysis buffer might strip loosely bound Aβ from plaques.

Protein amounts were determined by the Bradford protein assay (BCA Protein Assay, Pierce). All supernatants were ultracentrifuged for 60 min at 100,000g. Finally, before analysis, endogenous Igs were removed from the protein fractions by sequentially incubating extracts for 1 hour at room temperature with 50 μl of Protein A Sepharose Fast Flow followed by 50 μl of Protein G Sepharose Fast Flow (GE Healthcare Life Sciences).

Tau dephosphorylation

Tau (50 μg of intracellular lysate per reaction) was dephosphorylated by treatment with calf-intestinal alkaline phosphatase (New England BioLabs Inc.) at 20 U/ml for 3 hours at 37°C. The reaction was stopped by adding SDS–polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer and heat denaturation at 95°C.

Affinity purification of human Aβ oligomers

Total brain proteins (1 to 2 mg) from Tg2576 or J20 APP transgenic mice were incubated for 3 hours at 4°C with protein G–coupled magnetic beads (MagG beads, GE Life Sciences) previously cross-linked with 200 μg of purified 6E10 antibody (Covance), with 200 μg of Mab13.1.1 and Mab2.1.3 (100 μg of each), or with 200 μg of GluN1 antibody (Millipore). Immunocaptured proteins were eluted from the immune complexes using 1% n-octyl β-d-thioglucopyranoside (Sigma-Aldrich) in 100 mM glycine (pH 2.8) for 1 min (three to five rounds).

Relative amounts of purified oligomeric Aβ were calculated on the basis of synthetic Aβ1–42 standards (0.001, 0.025, 0.05, 0.1, 0.25, 0.5, 1, and 2.5 ng) ran alongside the samples used for experiments. Because there are no reagents selective to each Aβ oligomer to date, we opted to determine the relative abundance of purified Aβ oligomers compared to eight Aβ standards. The mass of a given Aβ oligomer was estimated on the basis of these standards, and molar concentration was calculated on the basis of the empirical molecular weight of each given Aβ oligomer (9, 14, and 56 kDa for putative Aβ dimers, trimers, and Aβ*56, respectively).

Considering that the relative abundance of a given Aβ oligomer varies with aging, amyloid deposition, and protein segregation (8, 72, 73), protein lysates of 15- to 18-month-old APP Tg mice were screened by Western blotting to measure the abundance of apparent Aβ dimers, trimers, and Aβ*56 before they were subjected to the purification steps consisting of sequential immunoaffinity captures and SEC. Similar segregations were obtained regardless of the line used, although the relative yields for a given oligomeric Aβ differed between lines.

Consistent with our previous findings (2, 8, 16, 72, 73), Aβ dimers are far more abundant in membrane-enriched protein lysates compared to extracellular-enriched lysates, whereas this pattern is reversed for putative Aβ trimers. Similar biochemical segregation was also observed using postmortem human brain tissue (16). The example provided in fig. S3 using extracellular lysates reflects this segregation following our four-step extraction protocol.

Size exclusion chromatography

Immunoaffinity-purified protein extracts were loaded on Tricorn Superdex 75 or 200 Increase columns (GE Healthcare Life Sciences) and run at a flow rate of ~0.3 ml/min. Fractions of 250 μl of eluate in 50 mM tris-HCl, 150 mM NaCl, and 0.01% Triton X-100 (pH 7.4) were collected using BioLogic DuoFlow QuadTec 40 System (Bio-Rad) coupled to a microplate-format fraction collector. A280 (absorbance at 280 nm) was determined live during the experiments and confirmed after each run on a DTX800 Multimode microplate reader (Beckman Coulter).

Western blotting and quantification

SDS-PAGE was done on precast 10 to 20% SDS-polyacrylamide tris-tricine gels and 10.5 to 14% and 4 to 10.5% tris-HCl gels (Bio-Rad). Protein levels were normalized to 2 to 100 μg of protein per sample (depending on targeted protein) and resuspended with 4× loading buffer. Thereafter, proteins were transferred onto 0.2-μm nitrocellulose membrane (Bio-Rad) in 5 to 10% methanol-containing transfer buffer for 2 to 3 hours at 4°C. To then detect Aβ molecules, nitrocellulose membranes were boiled in 50 ml of phosphate-buffered saline (PBS) by microwaving for 25 and 15 s with a 3-min rest interval in between. Membranes were blocked in tris-buffered saline (TBS)–0.1%Tween 20 (TTBS) containing 5% bovine serum albumin (BSA) (Sigma) for 1 to 2 hours at room temperature and probed with appropriate antisera/antibodies diluted in 5% BSA-TTBS overnight at 4°C. Primary antibodies were probed with either anti–IgGs conjugated with biotin or infrared dyes (LI-COR Biosciences). When biotin-conjugated secondary antibodies were used, infrared-conjugated NeutrAvidin (Thermo Scientific) was added to amplify the signal. Blots were revealed on an Odyssey platform (LI-COR Biosciences). For the detection of A11-reactive Aβ species, all blotting steps were performed in total absence of detergent in the buffers used. When required, membranes were stripped using Restore PLUS Stripping buffer (Pierce) for 30 to 180 min at room temperature depending on antibody affinity. Densitometry was performed using either OptiQuant software (Packard BioScience) or Odyssey software (LI-COR). Each protein of interest was probed in three individual experiments under the same conditions and quantified by software analysis, after determination of experimental conditions ascertaining linearity in the detection of the signal. The method used allows for a dynamic range of ~100-fold above background. Respective averages were then determined across the triplicate Western blots. Normalization was performed against the actin or the total form of the studied protein in the case of phosphorylated proteins. Because of the large number of samples analyzed, specimens were processed in two separate ways to compare possible effects induced by aging and by the transgene across groups.

Dot blotting

Two micrograms of membrane-bound protein extracts was mixed with sterile-filtered, deionized water in a total volume of 2.5 μl. Each sample was adsorbed onto a nitrocellulose membrane until dry for 30 min. After a brief activation in 10% methanol-containing TBS, membranes were boiled in PBS to enhance antigen detection (8). All steps were performed without detergent to enhance A11 and OC binding of oligomeric species (8).

Immunoprecipitation

Aliquots (100 to 250 μg) of protein extracts were diluted to 1 ml with dilution buffer [50 mM tris-HCl (pH 7.4) and 150 mM NaCl] and incubated with appropriate antibodies (5 μg) overnight at 4°C, and 50 μl of Protein G Sepharose Fast Flow (GE Life Sciences) or protein G–coupled magnetic beads (MagG beads, GE Life Sciences) at 1:1 (v/v) slurry solution with dilution buffer [50 mM tris-HCl (pH 7.4) and 150 mM NaCl (pH 7.4)] was added for 2 to 16 hours. The beads were washed twice in 1 ml of dilution buffer, and proteins were eluted in 25 μl of loading SDS-PAGE buffer by boiling.

Antibodies

The following primary antibodies were used in this study: 6E10 (1:2500), 4G8 (1:2500), biotinylated 6E10 (1:2500) (Covance), 40/42-end–specific Mab2.1.3 and Mab13.1.1 (1:1000) (gift from P. Das, Mayo Clinic), A11 (1:1000) (gift from R. Kayed, University of Texas Medical Branch), tau5 (1:2000) (Covance), pS262-tau (1:2000) (catalog no. AB9656, EMD Millipore), anti–pY18-tau (1:2500) (gift from G. Lee, University of Iowa), anti–pS416-tau (1:2000) (catalog no. ab119391, Abcam), anti–pS409-tau (1:2000) (catalog no. OPA1-03150, Thermo Scientific), pS202-tau (CP13) (1:500), PG5 (1:500), PHF1 (1:500), Alz50 (1:500) (gifts from P. Davis, Albert Einstein College of Medicine, Yeshiva University), 0N4R-tau (1:2000) (42), anti-GluN1 (1:1000) (catalog nos. sc-1467 and sc-9058, SCBT), anti–GluN1-CT and anti-GluN2B (1:1000) (catalog nos. 05-432, AB1557, and 06-600, EMD Millipore, and catalog nos. sc-1469 and sc-9057, SCBT), anti-GluN2A (1:1000) (catalog no. sc-9056, SCBT, and catalog no. 2720, Tocris Bioscience), anti-mGluR5 (1:1000) (catalog no. RA16100, Neuromics, and catalog no. G9915, Sigma-Aldrich), anti-GluN2C (1:1000) (catalog no. sc-1470, SCBT), anti-GluN2D (1:1000) (catalog no. sc-10727, SCBT), anti-AChRα7 (1:1000) (catalog no. sc-1447, SCBT), anti-GluA1 (1:1000) (catalog no. sc-7609, SCBT), anti-GluA2 (1:1000) (catalog no. sc-7610, SCBT), anti-GluA3 (1:500) (catalog no. sc-7613, SCBT), anti-GluA4 (1:500) (catalog no. sc-7614, SCBT), anti-GluK1 (1:500) (catalog no. sc-26475, SCBT), anti-GluK2 (1:500) (catalog no. sc-7618, SCBT), anti-GluK2/3 (1:500) (catalog no. 04-921, EMD Millipore), anti-GluK3 (1:500) (catalog no. sc-7620, SCBT), anti-mGluR1a (1:100) (catalog no. G9665, Sigma-Aldrich, and catalog no. 07-617, Upstate), anti-mGluR2/3 (1:100) (catalog no. G9790, Sigma-Aldrich, and catalog no. 06-676, Upstate), anti–Ephrin B2 (1:2000) (512012, R&D Systems Inc., and catalog no. sc-28980, SCBT), anti-PrPC (1:1000) (catalog no. sc-7693, SCBT), anti-RAGE (1:1000) (catalog no. 250462, Abbiotec Inc., and catalog no. sc-28980, SCBT), anti-actin (1:10,000) (C4, EMD Millipore, and catalog no. A2066, Sigma-Aldrich), anti-CaMKIIα (1:1000) (catalog no. NBP1-20008, Novus Biological; catalog no. PA1-14077, Thermo Scientific; and catalog no. sc-5306, SCBT), pT286-CaMKIIα (1:1000) [catalog no. MA1-047, Thermo Scientific, and catalog no. 3361, Cell Signaling Technology (CST)], anti-MAP2 (1:500) (catalog no. NB300-213, Novus Biologicals), anti-PSD95 (1:200) (catalog no. sc-8575, SCBT, and catalog no. ab18258, Abcam), anti-CREB (1:1000) (catalog no. MAB5432, EMD Millipore), anti–pS133-CREB (1:1000) (catalog no. 06-519, EMD Millipore), anti-p38 (1:1000) (catalog no. 9212, CST), pT180/Y182-p38 (1:2000) (catalog no. 9216, CST), anti-ERK (1:1000) (catalog no. 06-182, EMD Millipore), anti-pERK (1:1000) (12D4, EMD Millipore), anti-GSK3α/β (1:1000) (catalog no. sc-56913, SCBT), pS21/S9-GSK3α/β (1:1000) (catalog no. 9327, CST), pS9-GSK3β (1:2000) (catalog no. 9336, CST), anti-p35 (1:1000) (catalog no. sc-820, SCBT), anti–FOXO-1 (1:1000) (2H8.2, EMD Millipore), anti–calcineurin A (1:1000) (catalog no. 2614, CST), anti-PP1α (catalog no. 2582, CST), anti-PP2AC (1D6, EMD Millipore), and anti–cleaved caspase-3 (catalog no. 9664, CST). Validation of the antibodies was performed comparing the specificity and segregation pattern of the target protein using extracellular-, intracellular-, and membrane-bound–enriched protein extracts as described in the “Protein extractions” section.

Quantitative real-time RT-PCR

Total RNAs were extracted with a NucleoSpin RNA II kit (Macherey Nagel) according to the manufacturer’s protocol. For each sample, 1 μg of total RNAs was reverse-transcribed using Reverse Transcription System (Promega) and Eppendorf Mastercycler Personal (Bioblock), and RT was first performed at 70°C for 5 min, followed by a second step consisting of RT at 37°C over 1 hour. PCR amplification was performed on 5 μl of RT products in a total volume of 25 μl. Forward and reverse primers were designed with Beacon Designer software (Bio-Rad) and used after validation. Assays were made in triplicate using the iCycler iQ real-time PCR detection system (Bio-Rad). The amplification profile was as follows: 95°C for 30 s (twice), 95°C for 15 min (once), and 40 cycles in 15 s at 95°C, 1 min at 60°C, and 30 s at 95°C. PCR was run using the ABsolute QPCR SYBR Green Fluorescein Mix (ABgene) and its associated protocol. The amount of target was given by the following formula: 2 − [(Ct gene of interest − Ct housekeeping gene)T – (Ct gene of interest − Ct housekeeping gene)C], where Ct is the threshold cycle value, T is the treated condition, and C is the control condition. Results are expressed relative to the housekeeping genes cyclophilin and actin.

Confocal imaging

Triple- or quadruple-label immunofluorescence was performed as previously described (2, 19) using Alexa Fluor 488–, Alexa Fluor 555–, and Alexa Fluor 635–conjugated secondary antibodies (Molecular Probes, Invitrogen). Mouse brain sections were treated for autofluorescence with 1% Sudan Black solution and coverslipped with ProLong mounting medium (Molecular Probes). Digital images were obtained using an Olympus IX81 FluoView 1000 microscope with laser intensities ranging from 7 to 11%. Raw image z stacks (0.1- to 0.5-μm intervals) were analyzed using Bitplane’s Imaris 7.x software suite. Frame size was maintained at 1024 × 800, and optical zoom of 1.00 was used to allow for maximum distribution of pixel size to tissue dimensions without oversampling. Six ROIs per brain section (six sections per brain) per animal (four to six animals per group) were used in a randomized fashion. For in vitro analyses, eight ROIs per dish per group were used. Z stacks were reconstructed using the Surpass or Easy3D modules of the Imaris software package (version 7.x, Bitplane Inc.). Experimenters performing image acquisition and analyses were blind to the genotype or treatment conditions.

Statistical analyses

When variables were non-normally distributed, nonparametric statistics were used (Spearman rho correlation coefficients and Kruskal-Wallis nonparametric analysis of variance followed by Bonferroni-corrected two-group post hoc Mann-Whitney U tests). When variables were normally distributed, parametric statistics were used (one- or two-way ANOVA followed by Bonferroni-corrected two-group post hoc Student’s t tests). Sample size was determined by power analysis to be able to detect statistically significant changes within a 20% variation of measured responses. Analyses were performed using JMP 11 (SAS Institute).

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/10/478/eaal2021/DC1

Fig. S1. Reverse coimmunoprecipitation of Aβ*56 with NMDAR subunits in Tg2576 and in human brain tissue.

Fig. S2. Age-dependent accumulation of Aβ oligomers in Tg2576 mice.

Fig. S3. Biochemical characterization of soluble Aβ species present in APP transgenic brain tissues.

Fig. S4. Molecular and morphological characterization of primary cortical neurons.

Fig. S5. The major pathways regulated by extrasynaptic NMDARs are not altered in 7-month-old Tg2576 mice.

Fig. S6. Relationship between CaMKIIα activity and Aβ*56 expression in brain tissue from young Tg2576 mice.

Fig. S7. The major pathways regulated by extrasynaptic NMDARs are not altered by endogenous Aβ oligomers in mouse cortical primary neurons after a 60-min exposure.

Fig. S8. Aβ*56-induced translocation of pCaMKIIα to postsynaptic sites in primary cortical neurons.

Fig. S9. CaMKIIα activation is not induced by low-n Aβ oligomers purified from APP transgenic mice.

Fig. S10. Epitope and dephosphorylation tau assays confirm the presence of hyperphosphorylated tau conformers.

Fig. S11. Aberrant phosphorylation of CaMKIIα and tau in young J20 mice expressing Aβ*56.

Fig. S12. Temporal expression profiles of CDK5 adaptor proteins and GSK3 in young wild-type and Tg2576 mice.

Fig. S13. Brain-derived Aβ dimers and trimers do not induce tau phosphorylation at Ser202.

Fig. S14. Dose-dependent uncoupling of PSD95 and NMDAR GluN1 subunit induced by tatNR2B9c in primary neurons.

REFERENCES AND NOTES

Acknowledgments: We are indebted to P. Davies (Albert Einstein College of Medicine, Yeshiva University) for providing the CP13, PG5, MC1, PHF1, and Alz50 antibodies; R. Kayed (University of Texas Medical Branch) for the A11 antibody; G. Lee (University of Iowa) for the PY18 antibody; H. Orr for Cdk5 and GSK3α/β antibodies; L. Looger (Janelia) for GCaMP6 vectors; A. Aguzzi for Prnp-null mice; M. Kuskowski for biostatistics; K. H. Ashe, J. Jankowsky, M. K. Lee, D. Walsh, E. Newman, and H. Orr for critical discussions; and K. Kanamura, H. Nguyen, H. Schley, and M. LaCroix for technical help. Funding: This work was supported in part by NIH grants R00AG031293 and R01NS33249, the Strom and Moe gifts, and startup funds from the University of Minnesota Foundation (to S.E.L.). Author contributions: F.A., M.A.S., T.R., M.L., G.B., A.B., and S.E.L. performed experiments. J.G., A.B., and S.E.L. conceived, designed, and supervised experiments. L.C. provided reagents and technical help. S.E.L. wrote the manuscript. T.R., A.B., and S.E.L. prepared and organized the figures. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Requests for materials should be addressed to S.E.L. (lesne002{at}umn.edu).
View Abstract

Navigate This Article