Research Article

A vicious cycle of β amyloid–dependent neuronal hyperactivation

See allHide authors and affiliations

Science  09 Aug 2019:
Vol. 365, Issue 6453, pp. 559-565
DOI: 10.1126/science.aay0198

Dissecting hyperactivation in AD

Progressive accumulation of amyloid β (Aβ) in the brain is a defining feature of Alzheimer's disease (AD). At late stages of AD, pathological Aβ accumulations cause neurodegeneration and cell death. However, neuronal dysfunction, consisting of an excessively increased activity in a fraction of brain neurons, already occurs in early stages of the disease. Zott et al. explored the cellular basis of this hyperactivity in mouse models of AD (see the Perspective by Selkoe). Aβ-mediated hyperactivation was linked to a defect in synaptic transmission exclusively in active neurons, with the most-active neurons having the highest risk of hyperactivation. Aβ-containing brain extracts from human AD patients sustained this vicious cycle, underscoring the potential relevance of this pathological mechanism in humans.

Science, this issue p. 559; see also p. 540

Abstract

β-amyloid (Aβ)–dependent neuronal hyperactivity is believed to contribute to the circuit dysfunction that characterizes the early stages of Alzheimer’s disease (AD). Although experimental evidence in support of this hypothesis continues to accrue, the underlying pathological mechanisms are not well understood. In this experiment, we used mouse models of Aβ-amyloidosis to show that hyperactivation is initiated by the suppression of glutamate reuptake. Hyperactivity occurred in neurons with preexisting baseline activity, whereas inactive neurons were generally resistant to Aβ-mediated hyperactivation. Aβ-containing AD brain extracts and purified Aβ dimers were able to sustain this vicious cycle. Our findings suggest a cellular mechanism of Aβ-dependent neuronal dysfunction that can be active before plaque formation.

The progressive buildup of β-amyloid (Aβ) in the brains of Alzheimer’s disease (AD) patients is a firmly established experimental observation (1, 2). The consequences of this buildup are manifold and include synaptic dysfunction, inflammation, and, ultimately, cell death (3, 4). On the systems level, functional brain changes such as impaired neuronal activity and disturbed brain metabolism have been associated with Aβ amyloidosis (57). Several lines of evidence indicate that neuronal hyperactivity is a potential key feature of early stages of AD. Both in mice and man, there is strong evidence of excessive neuronal activation that, under certain conditions, can induce epileptic seizures (810). Functional imaging studies in individuals with prodromal AD reveal increased neuronal activity in the hippocampus and some neocortical areas (6). The cellular correlates of this hyperactivity have been studied in mouse models of Aβ amyloidosis by using two-photon calcium imaging (11, 12) and implicate an essential role for soluble Aβ (11).

An open question for understanding AD pathology is how soluble Aβ mediates cellular dysfunction such as hyperactivity. A large number of possible Aβ “receptors” have been suggested (13), but their roles in neuronal dysfunction in vivo have not been elucidated. Ample evidence indicates an Aβ-dependent impairment at both inhibitory (5, 8, 10) and excitatory (1417) synapses. Specifically, an impairment of glutamate homeostasis is evident in rodents (1821) and humans (22, 23) and might underlie the disturbed plasticity of hippocampal synapses (18, 19). However, the link between impaired glutamate homeostasis and neuronal function in vivo is unclear. Here, we explored the mechanism of Aβ-dependent neuronal hyperactivation and the forms of soluble Aβ that mediate this cellular dysfunction.

Neuronal hyperactivity requires preexisting baseline activity

We used two-photon Ca2+ imaging of hippocampal CA1 neurons in vivo (11) (fig. S1, A to C) to test the direct action of soluble Aβ. Synthetic Aβ(1 to 40)S26C, in which the naturally occurring serine at position 26 was replaced with cysteine, was used to produce and test the effects of the disulfide cross-linked dimer [AβS26C]2 (24, 25). [AβS26C]2 was pressure-applied near the CA1 hippocampal pyramidal layer of 1- to 2-month-old wild-type (WT) mice. In most neurons, application of 500 nM [AβS26C]2 reversibly induced a massive increase in activity, similar to the hyperactivity seen in amyloid precursor protein (APP) transgenic mice (11) (Fig. 1A, control experiments in fig. S1, D and E). Applications of [AβS26C]2 were ineffective in hippocampal slices (Fig. 1B and fig. S2, A and B). A possible explanation for these apparently contradictory findings was that the neuronal baseline activity was greatly reduced in hippocampal slices compared with that in vivo (fig. S2, C and D). In this experiment, to test the role of baseline activity for Aβ-induced neuronal hyperactivation, we performed loss-of-(dys)function experiments in the hippocampus of WT mice in vivo and gain-of-(dys)function experiments in hippocampal slices. First, we demonstrated that application of [AβS26C]2 was ineffective when blocking in vivo neuronal activity by antagonists ionotropic glutamate receptors (Fig. 1, C and E) or by the sodium-channel antagonist tetrodotoxin (fig. S3). Next, we turned to the analysis of hippocampal slices in vitro and did opposite experiments in which we induced in vivo–like baseline activity through various pharmacological manipulations. Treatments included (i) block of γ-aminobutyric acid–mediated synaptic inhibition by bicuculline, (ii) addition of glutamate to the bath, (iii) elevation of the extracellular K+ concentration, and (iv) combinations of these manipulations. Each of these treatments induced an average baseline activity that was similar to that detected under in vivo conditions (fig. S4A). Baseline activity increased in the presence of bicuculline in five representative neurons, as illustrated in Fig. 1D, left. In these conditions, the application of [AβS26C]2 resulted in a reversible increase of additional activity (Fig. 1, D and F). A similar effect was observed by the addition of low levels of glutamate (Fig. 1G) or by the elevation of the extracellular potassium concentration (Fig. 1H, and controls are shown in fig. S4, B and D). In studies on the Aβ dependence of activity-dependent synaptic plasticity, such as long-term potentiation (LTP) [for example, (19, 24)] or long-term depression [for example, (18, 26)], an increase of baseline activity is probably not necessary, because the induction protocols for synaptic plasticity involve increased levels of activity. In conclusion, our in vitro experiments support the in vivo observations and indicate that increased levels of baseline activity are a prerequisite for [AβS26C]2-induced hyperactivity. Cell-by-cell analyses show that, despite a considerable variance, there is, on average, a positive correlation between baseline activity and the degree of hyperactivation (Fig. 1I and fig. S4E).

Fig. 1 Activity dependence of the Aβ-dependent neuronal hyperactivation.

(A) (Top) Representative two-photon images of the hippocampal CA1 region of a WT mouse in vivo before (left) and during the application of 500 nM [AβS26C]2 (middle) and after 5 to 10 min of washout (right). The colored dots on the neurons indicate the number of Ca2+ transients per minute. (Bottom) Ca2+ traces of the five neurons circled in the top panel. The gray-shaded area indicates the time period of [AβS26C]2 application. (B) (Top) Representative two-photon images of the hippocampal CA1 region of an acute slice preparation before (left) and during [AβS26C]2 application (middle) and after washout (right). (Bottom) Ca2+ traces of the five neurons circled in the top panel. The gray-shaded area indicates the period of [AβS26C]2 application. (C) Same as (A) for a mouse in which glutamatergic transmission was blocked by bath application of D-APV (50 μM) and CNQX (50 μM). (D) Same as (B) for a slice treated with 80 μM bicuculline and an elevated potassium concentration (6.5 mM). The asterisks denote astrocytes. (E) Summary data of the in vivo experiments in (A) (left) and (C) (right). Each dot represents the mean under baseline (BL), [AβS26C]2 application, and washout (WO) conditions. (F) Same as (E) for experiments in (B) and (D). (G) Summary data of the in vitro experiments in which neuronal baseline activity was induced by the superfusion of glutamate (40 to 60 μM). (H) Summary data of in vitro experiments in which neuronal baseline activity was induced by elevating extracellular K+ (to 7.5 to 8.5 mM). (I) Plot of baseline activity versus [AβS26C]2-dependent relative increase in activity in vivo (ΔHyper) for individual neurons. The numbers of neurons for each bin of baseline activity are indicated in the graph. Red line: linear fit. Scale bars: 5 μm. Error bars show SEM. Wilcoxon signed-rank test, *P < 0.05. n.s. not significant.

Hyperactivation through an Aβ-dependent block of glutamate reuptake

Under in vivo conditions, neuronal activity generated by glutamatergic excitation was required for [AβS26C]2-induced hyperactivity (Fig. 1C). In our search for a cause underlying a synaptic potentiation, we considered pre- and postsynaptic mechanisms. To investigate whether [AβS26C]2 acted presynaptically, we performed paired-pulse facilitation experiments tested for possible changes of presynaptic release probability of glutamate (27). Application of [AβS26C]2 had no detectable impact on paired-pulse facilitation (fig. S5), as expected (18, 24, 28). An alternative hypothesis was inspired by reports on an Aβ-dependent defect of the glutamate homeostasis, possibly involving an impairment of glutamate reuptake (1820). In a first step, we tested whether pharmacologically blocking glutamate uptake in vivo had any detectable effect on neuronal activity. For this, we used the unspecific glutamate uptake blocker DL-threo-β-benzyloxyaspartic acid (TBOA), which can mimic some effects of Aβ on activity-dependent synaptic plasticity in vitro (18, 19). The local application of TBOA to hippocampal CA1 neurons in WT mice induced neuronal hyperactivity (Fig. 2, A and B), an effect that was similar to that observed with [AβS26C]2 applications (Figs. 1, A and E, and 2C). Nevertheless, [AβS26C]2 and TBOA may have exerted their actions through different mechanisms. To address this issue, we repeated the experiments in the transgenic APP23 × PS45 mouse model of Aβ-amyloidosis (12). We used young mice with no obvious amyloid plaques but with high levels of soluble Aβ (12) and pronounced hippocampal hyperactivity (11) (Fig. 2, D and E, and fig. S6). Application of TBOA had strong hyperactivating action in WT mice (Fig. 2F) but almost no effect in APP23 × PS45 mice (Fig. 2, G and I). Applications of [AβS26C]2 were also ineffective in APP23 × PS45 mice (Fig. 2, H and J). Thus, endogenous Aβ largely occluded both TBOA- and [AβS26C]2-induced hyperactivation.

Fig. 2

[AβS26C]2-dependent suppression of glutamate reuptake. (A) Same experimental arrangement as in Fig. 1A but with application of 250 μM DL-TBOA. (B) Summary data for the experiment in (A). Each dot represents the mean number of Ca2+ transients per minute for all neurons in one mouse under baseline, TBOA application, and washout conditions. (C) Bar graph showing the normalized number of Ca2+ transients. Each point represents the mean number of Ca2+ transients in one mouse during application of 500 nM [AβS26C]2 (left, n = 7 mice) or TBOA (right, n = 7), normalized to baseline. (D and E) Pie chart depicting the proportion of silent (blue), normal (white), and hyperactive (orange) neurons in WT (D) (n = 275 cells from seven mice) and APP23 × PS45 transgenic (TG) mice (E) (n = 299 cells from six mice). (F and G) Overlaid Ca2+ traces from all neurons in one WT mouse (F) and one APP23 × PS45 mouse (G) for baseline (left), TBOA application (middle), and washout (right) conditions. The blue-shaded area corresponds to the time of TBOA application. (H) Overlaid Ca2+ traces from all neurons (n = 20 cells) in one APP23 × PS45 mouse for baseline (left), [AβS26C]2 application (middle), and washout (right) conditions. The gray-shaded area indicates the period of [AβS26C]2 application. (I) Bar graph of the normalized activity during the application of TBOA in WT (left, solid bars, n = 7) or APP23 × PS45 transgenic (right, open bars, n = 5) mice. Each point represents the mean number of Ca2+ transients in one mouse during the application of TBOA, normalized to baseline. (J) Same as (I) for the application of [AβS26C]2 in WT (n = 6) or transgenic (n = 6) mice. Error bars show SEM. Wilcoxon signed-rank test (D, E) or Wilcoxon rank sum test (F), **P < 0.005, *P < 0.05.

Strong enhancement of synaptic stimulation–evoked glutamate transients through Aβ

To further test this glutamate accumulation hypothesis, we used two-photon glutamate imaging involving the viral expression of the fluorescent glutamate sensor iGluSnFr (29). For this purpose, an iGluSnFr viral construct was injected unilaterally into the hippocampal CA1 region in vivo (Fig. 3A), leading after 3 weeks to a strong and dense neuronal expression of iGluSnFr (Fig. 3B). In parallel, we also performed as controls sparse-labeling experiments of CA1 pyramidal neurons (Fig. 3C). To induce synaptic glutamate release, we electrically stimulated a bundle of afferent Schaffer collateral axons in hippocampal slices. We performed two-photon glutamate imaging and collected the bulk response in a region of interest (Fig. 3D, inset), which covered a substantial part of the glutamate sensor–expressing dendrites of CA1 neurons (Fig. 3C). Single-shock stimulation evoked large transient increases in extracellular glutamate concentration (fig. S7A, and control experiments are shown in fig. S7, B and C). Local application of Aβ produced a strong and reversible potentiation of the glutamate transients (Fig. 3, D and E, and fig. S7D). Similar glutamate transients were induced by applications of TBOA (Fig. 3, F and G, and fig. S7D). Thus, perisynaptic glutamate accumulations, through impaired uptake of synaptically released glutamate, may drive Aβ-dependent hyperactivity. In line with this conclusion, whole-cell recordings of N-methyl-D-aspartate (NMDA) receptor–dependent excitatory postsynaptic currents in CA1 pyramidal cells of hippocampal slices (30) were similarly affected by both TBOA and [AβS26C]2 (fig. S8). Together, these results demonstrate that TBOA and [AβS26C]2 act through a similar, yet unknown, molecular mechanism.

Fig. 3 [AβS26C]2-dependent potentiation of synaptic stimulation–evoked glutamate transients.

(A) Scheme of the injection of SF-iGluSnFr A184S into the mouse hippocampal CA1 region. (B) Confocal image of a hippocampal slice 21 days post-injection with SF-iGluSnFr (green). Cell bodies are stained with Neurotrace (blue). Scale bar: 100 μm. DG, dentate gyrus. (C) Sparse labeling of the hippocampal CA1 neurons with SF-iGluSnFr. The dashed lines indicate the pyramidal layer (P) of the hippocampal CA1 region. O, stratum oriens; R, stratum radiale. Scale bar: 50 μm. (D) Individual (gray) mean (color) glutamate transients collected in a rectangular region of interest in the stratum radiatum (inset left) after electrical stimulation (arrowhead, 100 μs/40 V) before (left) and during the application of 500 nM [AβS26C]2 (middle) and after washout (right). The inset indicates the positions of the stimulation and the [AβS26C]2-application pipettes, respectively. Scale bar: 50 μm. (E) Overlay of the average glutamate transients elicited by synaptic stimulation under baseline (black solid), [AβS26C]2 application (500 nM, red), and washout conditions (black dashed). (F) Overlay of the average glutamate transients elicited by synaptic stimulation under baseline (black solid), TBOA application (10 μM, blue), and washout conditions (black dashed). (G) Box plot of the amplitude of the glutamate transient after the injection of artificial cerebrospinal fluid (left), [AβS26C]2 (middle), or TBOA (right). N numbers are indicated next to the boxes. (H) (Top) representative two-photon images of hippocampal CA1 in vivo under baseline conditions (left), during the application of anti–GLT-1 polyclonal AB (middle), and after washout (right). The colored dots on the neurons indicate the number of Ca2+ transients per minute. (Bottom) Ca2+ traces of the five neurons circled in the top panel. The shaded area represents the time of AB application. Scale bar: 5 μm. (I) Summary data of the experiment in (H) for n = 6 mice. Error bars show SEM. Kruskal-Wallis test with Dunn-Sidak post hoc comparison (G) or Wilcoxon signed-rank test (I). **P < 0.005, *P < 0.05.

The astroglial excitatory amino-acid transporter 2 (EAAT2) (also termed GLT-1 in mice) is the predominant EAAT in the hippocampal CA1 region. Therefore, we tested whether Aβ interferes with EAAT2-mediated glutamate uptake. First, we tested the GLT-1 antagonist Dihydrokainic acid (DHK) in WT mice. Similar to TBOA and [AβS26C]2, DHK caused robust neuronal hyperactivity (fig. S9, A and B). Furthermore, the cross-linking GLT-1 antibody (GLT-1 AB) also induced hyperactivity (Fig. 3, H and I, and controls are shown in fig. S9, C and D). Cross-linking GLT-1 ABs can impair glutamate uptake by obstructing lateral membrane diffusion of glutamate–GLT-1 complexes along astrocytic protrusions out of the synaptic cleft, a process suggested to be essential for clearing synaptically released glutamate (31). Our results using DHK, GLT-1, and Aβ suggest that the Aβ-dependent block of glutamate reuptake may involve not Aβ binding to transporter proteins but rather perturbation of astrocytic membrane dynamics and obstruction of GLT-1 diffusion (31).

Effectiveness of human Aβ species derived from Alzheimer’s patients

To further explore the relevance of our findings to the human disease, we employed forms of Aβ derived from AD brain. First, we used Aβ-containing AD brain extracts (32). When examined by immunoprecipitation and immunoblotting, the Aβ in such brain extracts migrates on denaturing SDS–polyacrylamide gel electrophoresis with molecular weights indicative of monomers and SDS-stable dimers. Aβ-containing AD extracts (Fig. 4A), but not those immunodepleted of Aβ (ID extract, Fig. 4B), are capable of inducing a variety of disease-relevant effects (24, 3335). In vivo local applications of AD extract to CA1 neurons of WT mice produced a marked neuronal hyperactivity (Fig. 4, C, D, and F), whereas the ID extract did not induce hyperactivity (fig. S10, A and B). Similarly, when tested in vitro, AD extract induced hyperactivity in active hippocampal CA1 neurons treated with bicuculline (Fig. 4, E and G), but ID extract had no effect (fig. S10, C and D). Moreover, AD extract failed to cause hyperactivation in vivo in the presence of D-aminophosphovalerate (D-APV) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (fig. S10, E and F) or in unmanipulated hippocampal slices (fig. S10, G and H).

Fig. 4 Aβ derived from human AD patients induces neuronal hyperactivity.

(A) AD brain extracts were immunoprecipitated with anti-Aβ polyclonal AB AW7 or preimmune serum (PI) and Immunoprecipitates analyzed by immunoblot by using a combination of 2G3 and 21F12. Molecular weight markers are indicated on the left. At least two different Aβ species are in AD brain extract: monomers (single arrow) and SDS-stable Aβ dimers (double arrow). Nonspecific bands detected are indicated by a solid black line. (B) Mock immunodepleted (AD-ex) and AW7 immunodepleted (ID-ex) brain extracts were analyzed by an MSD-based Aβx-42 immunoassay. To assess the levels of monomeric and soluble aggregated Aβ, samples were pretreated with or without incubation in 5 M GuHCl. The AD extract contained much higher amounts of aggregates than monomer, and both were effectively removed by AW7 immunodepletion. (C) (Top) representative two-photon images of hippocampal CA1 in a WT mouse in vivo under baseline conditions (left), during the application of AD-ex (diluted 1:10), and after washout (right). The colored dots on the neurons indicate the number of Ca2+ transients per minute. (Bottom) Ca2+ traces of the five neurons circled in the top panel. The green-shaded area represents the time of AD extract application. Scale bar: 5 μm. (D) Overlaid Ca2+ traces from five representative neurons recorded in vivo under baseline (left), AD-ex application (middle), and washout conditions (right). The green-shaded area corresponds to the time of AD extract application. (E) Overlaid Ca2+ traces from five neurons recorded in vitro in a slice treated with bicuculline under baseline (left), AD-ex application (middle), and washout conditions (right). The green-shaded area corresponds to the time of AD-ex application. (F) Summary data for the experiment in (D). Each dot represents the mean number of Ca2+ transients per minute for all neurons in one mouse under baseline, AD-ex application, and washout conditions. (G) Summary data for the experiment in (E). Error bars show SEM. Wilcoxon signed-rank test. **P < 0.005, *P < 0.05.

Purified AD brain–derived cross-linked dimers can block LTP and impair neuritic integrity (36). We thus investigated whether such material (fig. S11A) could also induce hyperactivity. As a control, we isolated Aβ monomer from the same AD brain (fig. S11A). Aβ dimers, but not equimolar Aβ monomers (36, 37) (fig. S11, B to E), reduced neurite length (fig. S11, B and C) and the number of branch points (fig. S11, D and E). Similarly, the application of brain-derived Aβ dimers effectively but reversibly induced hyperactivity in WT mice in vivo (Fig. 5, A and B). Human Aβ dimers induced similar levels of hyperactivity at substantially lower concentrations than the synthetic ones (0.2 μg/ml human versus 4.3 μg/ml synthetic Aβ dimers). The application of human Aβ monomers had little or no effect (Fig. 5E). The activity of dimers was highly dose-dependent, with an apparent median effective concentration of 27.5 ng/ml (Fig. 5C). Application of AD brain–derived Aβ dimers to bicuculline-treated mouse hippocampal slices reliably induced hyperactivity in neurons with high baseline activity (fig. S12) and produced an activity-dependent hyperactivation in vivo (Fig. 5D). Finally, when Aβ monomers were applied to hippocampal CA1 neurons, they had little or no ability to induce both in vivo and in vitro (fig. S13).

Fig. 5 Role of human Aβ dimers and vicious cycle of hyperactivation.

(A) (Top) representative two-photon images of hippocampal CA1 region of a WT mouse in vivo under baseline conditions (left), during the application of 200 ng/ml human Aβ dimer (hAβ-dim, middle) and after washout (right). The colored dots on the neurons indicate the number of Ca2+ transients per minute. (Bottom) Ca2+ traces of the five neurons circled in the top panel. The gray-shaded area represents the time of human Aβ dimer application. Scale bar: 5 μm. (B) Summary data for the experiment in (A). Each dot represents the mean number of Ca2+ transients per minute for all neurons in one mouse under baseline, human Aβ dimer application, and washout conditions. (C) Dose-dependency curve of the action of human Aβ dimer. The activity during human Aβ dimer application, normalized to baseline (hyper ratio), for different dilution steps of the human Aβ dimer for 5 ng/ml (n = 5), 20 ng/ml (n = 5), 50 ng/ml (n = 5), and 200 ng/ml (n = 6) are plotted. (D) Plot of baseline activity versus human Aβ dimer-dependent relative increase in activity in vivo for individual neurons. The numbers of neurons for each bin of BL activity is indicated in the graph. Red line: linear fit. (E) Bar graph of the normalized activity during the application of 200 ng/ml human Aβ dimer (left, n = 6) or human Aβ monomer (hAβ-mon, right, n = 6). Each point represents the mean number of Ca2+ transients in one mouse during the application, normalized to baseline. Error bars show SEM. Wilcoxon signed-rank test (B) or Wilcoxon rank sum test (D), **P < 0.005, *P < 0.05. (F) Scheme of the vicious cycle of Aβ-dependent neuronal hyperactivation.

Discussion and conclusions

In this study, we characterized rapid actions of synthetic and AD brain–derived Aβ on the activity of mouse hippocampal neurons in vitro and in vivo. Our findings suggest that Aβ can induce hyperexcitation in sensitive neurons and that this drives a vicious cycle of hyperactivation (Fig. 5F). To explain the scheme, we begin with the insight that a simple solution exists for the puzzle that Aβ-dependent hyperactivity is readily observed in vivo but not in vitro. We were able to show that there is an absolute need of ongoing activity for the induction of Aβ-dependent synaptic hyperactivation. Next, the block of synaptically released glutamate at active excitatory synapses is an important element of the vicious cycle. The third component of the cycle is excessive perisynaptic accumulation of glutamate. The final element of the cycle is revealed by the dependence of the increase in hyperactivation on the level of baseline activity, both for synthetic and human brain–derived Aβ dimers. This process of amplification appears to be self-limited at high levels of hyperactivity, as indicated by the results of the occlusion experiments. Thus, multiple lines of evidence underscore the role of all four elements of the cycle.

Although the dependence of hyperactivation on impaired excitatory synaptic transmission involving defective glutamate reuptake had not previously been predicted, there is prior evidence for impaired glutamate homeostasis in both rodents (1821) and AD patients (22, 23). Furthermore, there is evidence for beneficial effects of certain antiglutamatergic drugs, such as memantine, against AD (20, 3840). It is suggested that these drugs may act perisynaptically on extrasynaptic NMDA receptors (40). Moreover, this process may be aggravated by pathologically reduced expression levels of glutamate transporters, such as EAAT2 in AD patients (22), or by reduced levels of synaptic inhibition (5, 8, 10). Finally, Aβ-dependent hyperactivity precedes plaque formation and is present at early stages, long before overt clinical symptoms of AD (6). A gradual neuronal “silencing” occurs after plaques are formed and may be the prelude to neurodegeneration (41). Although functional deficits of circuits caused by massive neuronal degeneration are nearly impossible to be repaired with current approaches, it may be possible to therapeutically target hyperactivation at early stages of the disease by lowering Aβ levels, reducing neuronal activity by enhancing synaptic inhibition, or by pharmacologically manipulating EAATs.

Supplementary Materials

science.sciencemag.org/content/365/6453/559/suppl/DC1

Materials and Methods

Figs. S1 to S13

References (4249)

References and Notes

Acknowledgments: We thank C. Karrer, C. Obermayer, F. Beyer, and R. M. Karl for technical support. We are grateful to L. Looger for providing iGlu-SnFr constructs. Funding: This work was supported by the Deutsche Forschungsgemeinschaft (SFB 870) and a European Research Council Advanced Grant to A.K. and by grants to D.M.W. from the National Institutes of Health (AG046275), Bright Focus, and the Massachusetts Alzheimer’s Disease Research Center (AG05134). D.M.W. is an Alzheimer Association Zenith Fellow. A.K. is a Hertie-Senior-Professor for Neuroscience. Author contributions: B.Z. and A.K. conceived the study. All authors planned experiments. B.Z., M.M.S., W.H., F.U., and H.-J. C.-E. performed the experiments. All authors interpreted the data. B.Z. and A.K. wrote the first draft of the manuscript. All authors read and commented on the manuscript. Competing interests: None of the authors have biomedical financial interests or potential conflicts of interest related to the work performed in the present study. Unrelated to the current study, D.M.W. is an advisor to CogRx and Regeneron and has active collaborations with Medimmune, Sanofi, Gen2, and Roche. D.M.W. is also affiliated with The Institute of Neurology, University College London. Data and materials availability: All data are available in the main text or the supplementary materials.
View Abstract

Stay Connected to Science

Subjects

Navigate This Article