The road to restoring neural circuits for the treatment of Alzheimer's disease

Journal name:
Date published:
Published online


Alzheimer's disease is a progressive loss of memory and cognition, for which there is no cure. Although genetic studies initially suggested a primary role for amyloid-in Alzheimer's disease, treatment strategies targeted at reducing amyloid-have failed to reverse cognitive symptoms. These clinical findings suggest that cognitive decline is the result of a complex pathophysiology and that targeting amyloid-alone may not be sufficient to treat Alzheimer's disease. Instead, a broad outlook on neural-circuit-damaging processes may yield insights into new therapeutic strategies for curing memory loss in the disease.

At a glance


  1. Neural circuits and synapses during the progression of AD.
    Figure 1: Neural circuits and synapses during the progression of AD.

    a, Subpopulations of neurons and glial cells form functional circuits through synaptic connections. b, In prodromal AD, amyloid-β fibrils begin to form in the extracellular space, possibly contributing to early circuit dysfunction that stimulates inhibitory sprouting and the initiation of inflammatory processes. c, In late-stage AD, amyloid-β plaques grow as production of the peptide outpaces its clearance. Intracellular neurofibrillary tangles consisting of paired helical filaments of hyperphosphorylated tau protein also form. Deficits in autophagy and other quality-control pathways contribute to the dysfunction of neurons and glia. The activation and proliferation of glial cells promotes inflammation and can affect circuit function in numerous ways. Several mechanisms affect GABAergic signalling and contribute to the loss of inhibitory tone. Despite the loss of synapses, excitatory neurons become hyperexcitable, compromising the fidelity of synaptic network connections for long-range communication.

  2. Hallmarks of AD in the brain and corresponding therapeutic strategies.
    Figure 2: Hallmarks of AD in the brain and corresponding therapeutic strategies.

    Distinctive features of AD include amyloid-β plaques, neurofibrillary tangles and cognitive dysfunction2. Network disruption and inflammation are also important mediators of the AD state49, 146. Each type of alteration seems to start in a specific functional subregion of the brain and spreads through networks to encompass most of the organ in late-stage AD. The most appropriate current strategies for the treatment of each feature are listed.

  3. Network-level treatments for AD.
    Figure 3: Network-level treatments for AD.

    Drugs that target cholinergic and glutamatergic circuits have minimal efficacy yet are the only approved treatments for AD119. Preliminary studies using deep-brain stimulation electrodes in the basal forebrain cholinergic circuit (green) and hypothalamus–hippocampal network (red) have shown promise in slowing disease progression133, 144. Direct stimulation of amygdala–hippocampal circuits (purple) in rodents also improves memory, which suggests that targeting these limbic circuits directly might be another route to restoring cognition. The implementation of non-invasive technologies such as transcranial magnetic stimulation in network modification makes such treatments safer and scalable.

Age-related dementias will affect almost 10% of people in the United States, and these conditions place a tremendous burden on such individuals and their families1. The most prevalent type of dementia, Alzheimer's disease (AD), causes a devastating and progressive loss of cognition, for which there is no effective treatment or cure. Analyses of the brains of people with AD suggest that the presence of extracellular aggregates of amyloid-β peptides, intracellular inclusions of neurofibrillary tangles rich in microtubule-associated protein tau and neuritic plaques are pathological hallmarks of the disease, yet there is no conclusive link between these observations and the cognitive symptoms2. The inability to definitively connect progressive memory loss to biomarkers greatly impedes the quest for effective therapeutic interventions for AD, but enhanced efforts to understand mechanisms of cognitive decline are revealing new avenues for intervention.

The insidious onset of AD-related memory loss has hampered progress towards effective therapies because cognitive symptoms emerge late in the progression of the disease. At this stage, there is already extensive deposition of amyloid-β, formation of neurofibrillary tangles and cell death, each of which might contribute to memory loss2. The first genetic studies performed in the 1990s indicated that amyloid-β was the main causative factor in AD because mutations or duplications in the gene APP3, which encodes the amyloid-β precursor protein (APP, also known as amyloid-β A4 protein), and in the genes PSEN1 and PSEN2, which encode amyloid-β processing-pathway components presenilin-1 and presenilin-2 (refs 4 and 5), lead to inherited, early onset AD. This breakthrough in understanding laid the foundation for the amyloid cascade hypothesis, which postulates that the accumulation of amyloid-β peptides initiates the pathophysiology of AD, leading to neurofibrillary tangles and neurodegeneration that cause memory loss6. As a corollary to this hypothesis, increased levels of amyloid-β in the brain — a result of the excessive generation of peptides or an inability to clear them — underlie cognitive dysfunction, and treatments to reduce the load of amyloid-β should therefore improve cognition7. Despite extensive evidence to support amyloid-β-driven neurodegeneration, drugs targeted at amyloid-β have failed to reverse deficits in memory or to halt cognitive decline, even alongside a considerable reduction in amyloid-β. As treatment strategies that are based on amyloid-β modification shift to preventing the accumulation of amyloid-β peptides, the refinement of drug targets and improving compound design, advances in our understanding of how brain functions are disrupted in AD, coupled with data from omics technologies and brain-imaging platforms, are highlighting other promising routes towards ameliorating cognitive decline.

As well as the genes in the amyloid-β pathway, other human genomic loci have been identified that increase the risk of developing AD considerably. Although some of these genes have been linked to amyloid metabolism, many appear to function in cell signalling pathways and in the immune system of the brain8, which indicates the importance of atypical cellular responses to changing brain states in AD. Emerging functional data are also shifting the aetiological focus of the disease from a neuron-centric view to an integrated outlook that acknowledges the synergistic functions of the different cell types of the brain9. Consistent with this holistic view, numerous disrupted processes are interconnected in AD and interact to provoke a cycle of dysfunction as the disease progresses9. This array of altered cellular processes ultimately disrupts neural circuits and network connectivity10, which can trigger further coping mechanisms in cells and might also propagate the spread of AD-related protein or peptide aggregates.

The complexity of these interactions makes disentangling the causes of AD a daunting task. Although systematic examination of the genetic and cellular changes can provide mechanistic insights into the cascade of events that occur during AD9, only understanding how these alterations contribute to the progressive dysfunction of brain networks and circuits will connect direct observations in cells to biologically abstract cognitive impairment. The identification of vulnerable networks and susceptibility nodes within them might also provide clues to the origins and progression of AD. This Review will explore briefly the role of amyloid-β in the development of AD. Rooted in newly identified genetic risk factors, it will then examine growing evidence that suggests that genetic, cellular and circuit dysregulation results from, and can lead to, the accumulation of amyloid-β peptides, the aggregation of tau proteins and cognitive impairment. With such a perspective, the AD field can move from cognitively unsuccessful clinical trials targeted at amyloid-β towards multipronged approaches to treatment that incorporate therapies that target cells and circuits.

AD through the lens of the amyloid cascade hypothesis

The accumulation of amyloid-β plays an important part in the pathophysiology of AD. Although the physiological functions of amyloid-β remain unknown, decades of research founded in genetic studies of AD continue to suggest a causative role for the peptide. The build-up of amyloid-β in the brain parenchyma probably contributes to the loss of synapses, neurodegeneration and alterations in neuronal activity. Each of these changes disrupts neural circuits, which can lead to widespread network dysfunction and cognitive decline.

Familial genetics implicate amyloid-β in AD

The most compelling evidence to support a central role for amyloid-β in AD comes from studies of the familial form of the disease. Cases of familial AD account for 1–5% (ref. 11) of patients with AD and they are the result of disease-causing, autosomal dominant mutations or duplications in APP, or mutations in PSEN1 and PSEN2 (refs 3,4,5). Mutations in APP, PSEN1 and PSEN2 shift the processing of APP (Box 1) to the amyloidogenic pathway and bias APP cleavage towards the longer, toxic amyloid-β peptides12. Although the mechanism by which the duplication of APP leads to toxicity is less clear, familial-AD-related APP duplications exist13, and people with Down's syndrome, who have three copies of APP, also develop AD14. Therefore, despite differences in their type, location in a particular gene or in the genes that they affect, heritable changes that result in the production of longer amyloid-β species cause AD. (Mutations in the gene MAPT, which encodes tau protein, usually lead to frontotemporal dementia — a similar yet distinct disease15.) Amyloid-β is implicated further in AD by mutations in APP that do not increase levels of amyloid-β and do not seem to be pathogenic; instead, these mutations might confer protection from AD. For example, an APP variant in the population of Iceland reduces levels of amyloidogenic processing and protects against AD and normal age-related cognitive decline16.

Box 1: The production of amyloid-β

Amyloid-β peptides result from the sequential cleavage of APP, a type I integral membrane glycoprotein. The structure of APP includes an amyloid-β domain (purple) with cleavage sites for secretase enzymes. Two pathways compete for the APP substrate, which leads to either amyloidogenic or non-amyloidogenic processing of the protein147 (Box Fig.).

In the non-amyloidogenic pathway (Box. Fig. a), the cleavage of APP in the amyloid-β domain by α-secretase, a complex that contains ADAM metalloproteases, releases the soluble ectodomain sAPPα and the C-terminal fragment CTFα. The subsequent cleavage of CTFα by γ-secretase produces a soluble extracellular p3 peptide and the APP intracellular domain (AICD)148.

The amyloidogenic pathway (Box Fig. b) involves APP cleavage by β-secretase (also known as BACE1), which releases the soluble ectodomain sAPPβ and CTFβ. Cleavage of CTFβ by γ-secretase yields amyloid-β peptides of varying lengths as well as the AICD fragment. The pathogenic impact of amyloid-β peptides varies with length. The longer amyloid-β(1–42) and amyloid-β(1–43) species are more prone to aggregation and prion-like seeding, and because these structures seem to be more toxic than shorter amyloid-β peptides, the ratio of amyloid-β(1–42) and amyloid-β(1–40) might predict the severity of AD149.

New pathways for APP processing are emerging. For example, the η-secretase pathway yields amyloid-η fragments that inhibit neuronal function. N-APP, a mediator of axonal degeneration, arises from sAPPβ following cleavage by β-secretase. And the caspase-mediated cleavage of the AICD can result in transcriptionally active products147, 150.

As well as mutations that cause familial AD, other heritable genetic risk factors contribute to an individual's susceptibility to late-onset AD and also to amyloid-β accumulation. The most important mutation in this context, and the first to be identified, is the ε4 allelic variant of the gene APOE (APOE4), which encodes apolipoprotein E (APOE). The presence of APOE4 leads to a three- or four-fold increase in the likelihood of developing AD17, and people with AD who carry APOE4 show more profound evidence of amyloid-β aggregates than do non-carriers with AD18. Studies in mice suggest that ApoE regulates amyloid-β levels in an ApoE-isoform-dependent manner, such that the ε4 isoform promotes amyloid build-up whereas the ε2 isoform seems to enhance its clearance19. Because variants of both APP and APOE can either increase amyloid-β accumulation, which increases susceptibility to AD, or reduce its accumulation, which reduces susceptibility, the genetic data strongly support an important aetiological role for amyloid-β accumulation in AD-related cognitive decline.

Amyloid-β impairs synapses and destabilizes circuits

Amyloid-β peptides can exert numerous detrimental effects on neurons and the other cell types of the brain9 (Fig. 1). As a peptide species with acute toxicity, oligomeric amyloid-β directly incites neuronal apoptosis through interactions with cell-surface receptors. Moreover, longer-term accumulation of toxic amyloid-β species in the parenchyma also leads to oxidative damage of DNA and proteins, to physical injury of cellular organelles and to dysregulation of intracellular calcium levels — each of which can provoke cell death20. Under in vitro conditions, exposure to amyloid-β induces neuronal dysfunction and can cause cell death within hours21. Although exposure to amyloid-β under in vitro conditions can induce neuronal dysfunction and cause cell death within hours, it takes years for amyloid accumulation to have detectable consequences in vivo. Despite the difference in timescale, the brains of people with AD show alterations similar to those seen in vitro. Most notably, considerable neurodegeneration accompanies the onset of cognitive decline22. However, amyloid-β-induced deficits in synaptic plasticity, circuit function and cognition might develop before this cell loss occurs10. Indeed, this possibility is supported by the observation that the levels of amyloid-β in people with AD seem to plateau before the onset of rapid neurodegeneration and cognitive symptoms22, and ex vivo studies in rats demonstrate that synapse loss occurs in response to small elevations in amyloid-β levels23. In people with AD, it is therefore possible that synapse loss precedes the formation of overt aggregates, and the loss of these synapses has been shown to contribute to circuit dysfunction and cognitive decline in mouse models of AD24, 25.

Figure 1: Neural circuits and synapses during the progression of AD.
Neural circuits and synapses during the progression of AD.

a, Subpopulations of neurons and glial cells form functional circuits through synaptic connections. b, In prodromal AD, amyloid-β fibrils begin to form in the extracellular space, possibly contributing to early circuit dysfunction that stimulates inhibitory sprouting and the initiation of inflammatory processes. c, In late-stage AD, amyloid-β plaques grow as production of the peptide outpaces its clearance. Intracellular neurofibrillary tangles consisting of paired helical filaments of hyperphosphorylated tau protein also form. Deficits in autophagy and other quality-control pathways contribute to the dysfunction of neurons and glia. The activation and proliferation of glial cells promotes inflammation and can affect circuit function in numerous ways. Several mechanisms affect GABAergic signalling and contribute to the loss of inhibitory tone. Despite the loss of synapses, excitatory neurons become hyperexcitable, compromising the fidelity of synaptic network connections for long-range communication.

There is abundant evidence to support a physiological role for amyloid-β at the synapse, and the data also suggest that synaptic activity is an important source of amyloid-β in the parenchyma26. Excitatory activity promotes the proteolysis of APP and its release into the extracellular space27, 28, 29. The effects of amyloid-β on synapse activity vary with its extracellular concentration: low levels of amyloid-β promote excitatory activity and higher levels depress it. Small increases in amyloid-β levels promote activity through presynaptic acetylcholine receptors, which elevate internal calcium concentrations to increase the probability of glutamate release30, 31. Although the postsynaptic excitation that results could lead to positive feedback in which the further release of amyloid-β increases synaptic excitability, increasingly high levels of amyloid-β actually depress synapse activity through several mechanisms that modify synapse strength, including the internalization of glutamate receptors32, 33. Whereas acute increases in synaptic amyloid-β can impair the long-term potentiation of synaptic strength and promote the depression of synaptic activity (known as long-term depression)34, 35, 36, chronic elevations can weaken connectivity, alter the dynamics of dendritic spines, promote synapse loss and impair circuit-shaping processes that underlie learning and memory27, 34. Amyloid-β-induced loss of dendritic spines can lead to hyperexcitable neurons that fire more easily24, and amyloid-β also alters the balance between excitatory and inhibitory activity by influencing inhibitory interneurons. Loss of synaptic inhibition occurs through numerous pathways, including the downregulation of cell-surface voltage-gated sodium channels. Indeed, reduced expression of the Nav1.1 sodium channel subunit (SCN1A) in a mouse model of AD hinders the propagation of action potentials through inhibitory parvalbumin-expressing interneurons, which results in a reduction in the release of the primary inhibitory neurotransmitter, GABA (γ-aminobutyric acid) and the loss of inhibition on excitatory neurons37.

Long-term amyloid-β accumulation, disinhibition of excitatory cells and synaptic loss lead to neuronal hyperactivity, which occurs in brain regions associated with learning and memory (such as the hippocampus) in some presymptomatic individuals38. In time, this can lead to epileptiform activity. In a mouse model of AD-like amyloid-β accumulation, such activity induces the compensatory sprouting of inhibitory axons that can impair learning processes25. People with AD who have seizures exhibit worse cognitive outcomes39, and the cognition of individuals with mild cognitive impairment is restored temporarily when they are treated with the anti-epileptic drug levetiracetam to reduce abnormal activity40. The correlation between altered activity and cognitive performance suggests that these aberrant patterns of activity might be directly related to memory impairment. Counteracting hyperactivity in mouse models of AD not only rescues local circuit dynamics41 but also reinstates long-range network coherence to repair learning42.

Network susceptibility is crucial for amyloid-β driven AD

The deposition of amyloids affects circuit connectivity and network activity. Therefore, identification of the brain networks that are susceptible to amyloid-β-induced dysfunction should reveal the pattern of disease propagation through the brain and could explain how the spread of amyloid-β contributes to the symptoms of AD. Initial cross-sectional studies of post-mortem tissue were used to create a hierarchical map of the advance of amyloid-β through the brain43. This led to the observation that amyloid-β deposition that begins in neocortical regions involved in cognition spreads to the neural hubs that underlie learning and memory, and finally progresses to the motor and sensory structures. It also provided a framework for understanding the successive deterioration of cognitive and sensory impairments43. Although the sequential involvement of interconnected brain regions was supported by these findings, postmortem studies of amyloids in the brain could not provide insights into the network pattern or mechanisms of its spread, or correlate low levels of amyloid deposition with a definitive diagnosis of AD in people who died before developing memory impairment. The need to understand the processes that underlie AD progression, and to connect the spread of amyloid to the development of AD, therefore paved the way for longitudinal studies of amyloid-β accumulation in people with the disease.

The development of tracers for positron emission tomography (PET) that bind amyloid-β, such as 11C-Pittsburgh compound B (PIB) and 18F-florbetapir, has enabled the pattern of amyloid-β deposition to be assessed across the course of AD44, 45 (Fig. 2). Cross-sectional PET studies of the brains of people with AD confirm neuropathological findings that show considerable deposits of amyloid-β throughout the neocortex and also demonstrate that the amount of cortical retention is predictive of cognitive decline46, 47. These findings support the genetic data and indicate that amyloid-β levels might be a good biomarker for AD48. Consistency in the locations of tracer retention in PET studies suggests that the accumulation of amyloid-β in an AD-specific pattern is a fundamental change that affects cognitive outcomes, and memory decline could be related to the deposition of amyloid-β in regions of the default mode network (DMN)49. The DMN is a group of functionally connected regions of the brain that are co-activated during passive thinking, remembering and planning. Initial investigations using PIB–PET show an overlap between DMN regions and areas with high levels of amyloid-β, which suggests that the DMN might be vulnerable to pathophysiological processes in AD50. DMN susceptibility has been demonstrated by functional connectivity magnetic resonance imaging (MRI) studies that show reduced connectivity within the network49. About 25% of people with good cognitive health show a high retention of PIB, which correlates with reductions in DMN connectivity and with worse episodic memory and predicts the diagnosis of dementia in follow-up studies51, 52. Together, this evidence suggests that alterations in the DMN might underlie cognitive impairment. Surprisingly, despite cognitive deterioration, increases in amyloid-β levels over the conversion period from the condition amnestic mild cognitive impairment to AD are relatively modest. This is consistent with observations that people with AD show few differences in the retention of PIB as the disease progresses22, 53. Instead, it is possible that progressive DMN dysfunction contributes to clinical deterioration54.

Figure 2: Hallmarks of AD in the brain and corresponding therapeutic strategies.
Hallmarks of AD in the brain and corresponding therapeutic strategies.

Distinctive features of AD include amyloid-β plaques, neurofibrillary tangles and cognitive dysfunction2. Network disruption and inflammation are also important mediators of the AD state49, 146. Each type of alteration seems to start in a specific functional subregion of the brain and spreads through networks to encompass most of the organ in late-stage AD. The most appropriate current strategies for the treatment of each feature are listed.

Although imaging studies in humans show that the localization of amyloid-β correlates with cortical network dysfunction, these techniques have yet to reveal initiating patterns of amyloid-β accumulation or give insights into the mechanisms of amyloid-β propagation. Some preclinical studies suggest that amyloid-β might propagate in a prion-like manner and undergo cell-to-cell transmission55, 56. However, other studies propose that amyloid-β-induced circuit dysfunction affects network connectivity42 and that local aberrant activity could lead to the accumulation of amyloid-β at downstream projection structures, contributing to the sequential appearance of amyloid-β in regions of connected networks. Whether functional or physical mechanisms contribute to amyloid-β progression, the large temporal dissociation between the accumulation of amyloids and the emergence of overt memory loss underscores the urgency of identifying regions that show susceptibility to amyloid-β deposition for the early detection of and intervention in AD.

Amyloid-β-based treatments have yet to restore cognition

Despite various complexities, it is clear that amyloid-β contributes to the progression of AD, which makes it a prime target for therapeutic intervention (Fig. 2). Because elevated levels of amyloid-β probably underlie its pathogenicity, treatment strategies have emerged that target the two determinants of amyloid-β load: the aberrant generation of amyloid-β and faulty clearance mechanisms.

Drugs that modulate the activity of the enzymes β-secretase or γ-secretase limit amyloidogenic proteolysis and are the main therapeutic strategy for reducing amyloid-β production. Phase II clinical trials have identified several safe first-generation drugs for each enzyme57, 58. However, most such trials have failed because of problems with target specificity, brain permeability or study design without testing cognitive or amyloid-β outcomes59, 60. The few trials that reported main outcomes showed reductions in markers of amyloid-β in cerebrospinal fluid but with conflicting cognitive results, including hastened decline61. Given these pharmacological setbacks and uninformative outcomes with respect to memory, it is difficult to make conclusions about the feasibility of using proteolytic enzymes as targets for AD treatment. However, redesigned clinical trials with updated dosing and outcome design and an improved understanding of the functions of the enzyme targets should help to ensure the safety and effectiveness of the next generation of compounds.

Enhancing amyloid-β clearance from the parenchyma is an alternative approach to modifying amyloid-β levels, and this is being pursued mainly through active and passive strategies of immunization against the peptide. Initial clinical trials that investigated active immunization reduced the amyloid-β burden successfully but also caused severe side effects62. The more positive effects on amyloid-β levels have facilitated the development of a new generation of immunotherapies that have advanced quickly through clinical trials. In general, people with AD tolerate these newer immunotherapies well, and phase I and phase II trials have demonstrated considerable reductions in amyloid-β levels63. These results are tempered by cognitive performance outcomes that suggest that such treatments might slow decline only in prodromal or mild AD. Furthermore, the efficacy of amyloid-β immunization strategies in this population is not consistent across clinical trials64, 65. A more successful study used florbetapir-based PET imaging to identify people with both AD and amyloid-β deposits66. People who lack such deposits are unlikely to respond to amyloid-β targeted therapies. By choosing participants with AD and amyloid-β deposits, the study was therefore able to test an amyloid-β targeted antibody therapy in individuals who are most likely to respond. The positive findings indicated that the stratification of people with AD, and the identification of biomarkers that correlate with those groups, might be crucial for testing treatments with different mechanisms of action66. As clinical trials advance through phase III and phase IV, further insights into the lasting cognitive outcomes could provide an improved understanding of the efficacy of these strategies for long-term prognosis.

The results from numerous amyloid-β-modifying clinical trials herald considerable progress towards reducing the amyloid-β burden. Although they are varied and minor, the cognitive benefits of such treatments suggest that decreasing amyloid-β levels slows the progression of AD. However, because both cognitive effects and the safe removal of amyloid are seen in the initial stages of the disease, and because many patients first seek treatment after they experience memory loss, preventing further deterioration is only the first step in an effective treatment regimen. The identification of additional interventions that can restore cognitive function and treat cellular and circuit dysfunction at later stages of the disease will be crucial for improving the quality of life of people with AD and to avert the looming public-health crisis.

Numerous factors induce network dysfunction in AD

The accumulation of amyloid-β in the brain has deleterious effects, and people with familial AD and rodent models of AD demonstrate considerable pathological and cognitive alterations in response to the peptide's presence. However, most people with AD have considerable accumulation of amyloid-β without causative mutations in genes related to APP processing. With the genetic evidence mounting for the contribution of other disrupted cellular mechanisms to the development of AD, understanding how these processes interact with amyloid-β aggregates, contribute to neurodegeneration and affect circuit function to produce cognitive decline is important for implementing effective treatment strategies to not only slow, but to also halt and reverse, memory loss.

Genetic risk and dysregulation prime the brain for AD

Cases of AD with no identifiable genetic cause account for more than 95% of individuals who develop symptoms in later life67. Similar to those with familial AD, people with late-onset AD experience amnesic memory loss and develop amyloid-β aggregates and neurofibrillary tangles; however, the levels of disease in late-onset AD vary considerably2. Because variation between individuals is high, and because numerous complex and interrelated processes can lead to amyloid-β accumulation and tau aggregation, understanding the mechanisms that drive late-onset AD has proven more difficult than disentangling familial AD9. Although functional studies of the amyloid-β peptide in models of familial AD have yielded insights into the mechanisms of peptide accumulation, the description of further intracellular functions for APP and the presenilins, together with the identification of increasing numbers of genetic risk alleles and regulatory mechanisms, has shed light on other processes disrupted in AD that might lead to amyloid-β accumulation and neurofibrillary tangle formation and contribute to cognitive decline.

Although protein changes related to familial AD increase amyloid-β levels directly, they might also have consequences for processes other than APP proteolysis. Mutations in APP, PSEN1 and PSEN2 could affect other roles of the proteins that these genes encode, including cholesterol binding, cell adhesion, cytoskeletal dynamics, ion homeostasis, endocytosis and synaptic plasticity68, 69, all of which can affect the properties of circuits. Mice that lack APP have deficits in the induction of long-term potentiation and also show age-dependent cognitive decline70, 71. Although familial AD-linked mutations in the presenilins have a clear effect on amyloid-β production, it is possible that the disruption of the other physiological functions of the presenilins also contributes to circuit dysfunction and memory impairment. For example, deletion of the presenilins suggests that they have crucial roles in neural development, synaptic plasticity and memory performance, and mutations in familial AD affect the function of these proteins as calcium-leak channels in the endoplasmic reticulum72, 73. Similarly, the association of APOE with the clearance of amyloid-β drew focus away from its roles in regulating cholesterol and lipids; however, studies of the brains of people with AD suggest that alterations in these processes could be an important mechanism of pathogenesis74. The identification of new genetic risk factors in recent years has underscored the potential relevance of these processes to cognitive decline.

The increasing accessibility of genome-level analyses has enabled large-scale screens of individuals with AD to identify further genetic risk loci that might increase susceptibility to the disease. These studies have implicated genes that are involved in lipid metabolism and cholesterol homeostasis, cell adhesion, cytoskeletal dynamics, ion regulation, vascularization and maintenance, transcriptional regulation, inflammation and endocytosis (for example, CLU, PICALM, SLC24A4, TREM2 and BIN1) (ref. 8). So far, confirmation of a subset of the newly identified loci in functional experiments has demonstrated that lipid processing, endocytosis and inflammation might contribute substantially to the development of AD75, 76, 77. Unlike APOE, APP, PSEN1 and PSEN2, the disease-causing genetic variants and the cellular consequences of alterations at these newly identified risk loci are largely unknown8. Although many of the implicated proteins seem to play parts in amyloid-β processing or clearance8, they might also be important for molecular homeostasis, cell function and synaptic dynamics, all of which influence circuit activity and network connectivity9, 10.

Similar to encoded genetic risk factors, epigenetic mechanisms can also alter gene expression. People with late-onset AD show altered DNA methylation at several AD genetic risk loci78. Association of the methylation status at some of these loci with the risk of AD and with cellular processes that are linked to the disease also indicates a functional significance for such alterations79. Studies in a mouse model of AD confirm the potential for epigenetic contributions to the disease state and suggest that the progression of the disease is associated with widespread shifts in histone modifications in hippocampal cells, which correlates with the transcriptional repression of synaptic genes and an induction of immune-related genes80. Evolutionarily conserved immune-gene regulatory regions with altered histone modification profiles are enriched for genetic variants identified in people with AD, which supports the genetic links between immune processes and the risk of AD80. Further insights into the epigenetic changes that are associated with neuronal genes in the disease comes from both rodent models and studies of people with AD, both of which have elevated levels of the enzyme histone deacetylase 2, an important negative regulator of synaptic plasticity81.

Other factors, such as circuit activity and peripheral signalling, can also modulate genetic risk by acting on transcriptional components, which could lead to fundamental changes in cellular health. An example of a protective transcription factor is RE1-silencing transcription factor (REST), which upregulates cellular protection pathways in ageing82. Higher levels of REST expression are correlated with increased hippocampal volume, improved cognitive outcomes with age and protection against amyloid-β toxicity. Levels of REST are lower in people with AD, and although the mechanism that underlies these reduced levels remains unclear, the time-dependent expression kinetics of REST could provide clues to the influence of circuits in healthy ageing. REST is regulated by non-cell-autonomous signalling, which indicates the vital role of neuronal communication and circuit function in neuroprotection against AD82.

Molecular alterations affect cellular health

Genetic loci identified in the past 5 years through enhanced screening of people with AD have helped to pinpoint cellular processes that might be involved in the development of AD and highlight the importance of both neurons and non-neuronal cells in disease progression9 (Fig. 1). There is considerable evidence to support the idea that dysfunction across different cell types in the brain can lead to circuit dysregulation and cognitive impairment. The cellular phase of AD refers to the time during which processes that normally contribute to feedback and feed-forward loops go awry, resulting in cellular disease states. This progression is aligned with the observation that the deposition of amyloid-β plateaus before the onset of memory symptoms9. Although cognition remains largely intact at the initiation of deposition, it is probable that progressive changes during and following the cellular phase of AD alter network connections and circuit activity to produce cognitive decline10. Because the events that lead to amyloid-β aggregation in the absence of causative mutations remain vague, understanding the convergent cellular factors that might affect circuit activity, drive cognitive deterioration and initiate amyloid-β accumulation is crucial for restoring memory capacity in advanced AD.

A number of cellular processes that are implicated by AD genetic risk loci might contribute to the onset and progression of AD when aberrant. These include mitochondrial function, the oxidative stress response, DNA damage repair and calcium homeostasis. Although each locus has not yet been confirmed, molecular studies imply that many such processes go awry in AD. Mitochondria accumulate age-related damage that can decrease efficiency, increase the release of reactive oxygen species, and contribute to other changes, including the oxidation of proteins and lipids and β-secretase (also known as BACE1) upregulation83, 84. Mitochondria are the main producers of reactive oxygen species, the release of which can cause the relaxation of heterochromatin, a tightly packaged form of DNA that represses certain portions of the genome85. The brains of people with AD show reduced heterochromatin-associated methylation marks, as do animals that overexpress tau protein, and this change is correlated with the expression of genes that are normally repressed in healthy individuals85. Detectable breaks in DNA occur at the initial stages of disease progression in mouse models of AD and in the brains of people with the disease86, 87. These breaks might be the result of physiological processes: DNA breaks arise in response to neuronal activity, although amyloid-β can exacerbate the damage88. Because the breaks that form during learning-related activity initiate the transcription of genes that are important for learning and memory, it is possible that as DNA repair mechanisms fail with age, mutations and breaks accumulate at plasticity-related loci, which stunts cognition89, 90. This suggests a direct link between neural circuit activity, DNA integrity and memory.

Calcium is a crucial mediator of many intracellular events and calcium levels regulate synaptic plasticity and signal propagation91. Brain cells regulate cytosolic calcium levels tightly via calcium influx through cell-surface channels and some neurotransmitter receptors and storing calcium in the endoplasmic reticulum91. In AD, a number of processes increase cytosolic calcium, including damage to mitochondrial or plasma membranes, circuit hyperactivity and amyloid-β-induced calcium influx across the plasma membrane. These events then trigger the release of calcium from the endoplasmic reticulum. Interestingly, PSEN1 has been shown to be an important regulator of such calcium release73. The AD risk gene SLC24A4 encodes a potassium-dependent sodium–calcium exchanger and mutations could alter cellular calcium handling and increase the risk of developing late-onset AD8. Furthermore, calcium can induce signalling cascades; for example, the calcium-mediated cleavage of the cyclin-dependent kinase 5 (Cdk5) activator p35 to p25 by calpain regulates synaptic depression in neurons under physiological conditions92. However, under pathogenic conditions, enhanced p25–Cdk5 signalling results in increases in histone deacetylase 2, synaptic depression, dendritic spine loss, neurodegeneration, inflammation and aberrant tau phosphorylation, which correlates with the presence of neurofibrillary tangle-like aggregates in mouse models of AD80, 81, 92. Elevated calcium levels in the cell can also produce ionic imbalance, facilitate the aberrant release of presynaptic neurotransmitters and dysregulate postsynaptic signal transduction that ultimately alters gene regulation, cellular dynamics and neuronal activity31, 91, 92.

Intracellular pathophysiology impairs circuits

Neuronal activity and circuit connectivity rely on connections between cells that can span long distances. This specialized morphology necessitates highly coordinated intracellular functions, however, AD disrupts processes that maintain neuronal health and support circuit function. The hallmark intracellular deposits of AD, neurofibrillary tangles, are the result of aberrant aggregation of tau protein, which usually associates with microtubules. Microtubules are the primary cytoskeletal component that gives neurons structure and enables trafficking. The stability and functionality of microtubule assemblies depends partially on the dynamic binding of tau93, 94. Aggregates of tau form when hyperphosphorylated tau proteins assemble into paired helical filaments and the presence of these structures is a requisite for the diagnosis of AD2. Furthermore, the level of neurofibrillary tangles correlates strongly with the progression of AD and cognitive symptoms2. Tau binds microtubules in its unphosphorylated state, and site-specific phosphorylation weakens tau's microtubule-binding capacity, which releases tau into the cytosol and reduces its solubility93, 94. Dephosphorylation by protein phosphatase 2A (PP2A) maintains appropriate tau dynamics, which enables tau to actively bind and be released from the cytoskeletal network95, 96. But in AD, circuit hyperactivity and calcium influx trigger aberrant tau phosphorylation by kinases such as Cdk5, mitogen-activated protein kinase (MAPK) and protein kinase A (PKA) and induce concomitant dysfunction of PP2A that leads to tau hyperphosphorylation, insolubility and aggregation95, 97. Although the pathogenicity of tau was assumed to be related to its aggregation and apparent mislocalization, tau also has an important role at the synapse, where it may coordinate the postsynaptic density and, similar to amyloid-β, may be released in an activity-dependent manner98, 99, 100. Further observations of the aberrant acetylation of tau and the accumulation of tau-rich autophagosomes in the brains of people with AD suggest that pathways involving tau and certain genetic risk factors (for example, the gene PICALM) converge, indicating that tau acts in complex cellular systems, the dysfunction of which is only beginning to be understood101, 102.

Endocytosis is a process by which cells transport molecules into the cytosol and is crucial for regulating intracellular and extracellular environments to influence cellular and circuit homeostasis. Large cytosolic protein assemblies and aged organelles undergo lysosomal degradation through macroautophagy processes that use the clathrin-mediated endocytosis of the donor membrane to engulf particles into autophagosomes102. These structures guide the degradation of proteins, and alterations in genetic risk loci associated with endocytosis (for example, the gene BIN1) might impair this ability8. Furthermore, synaptic activity and plasticity also rely on clathrin-mediated endocytosis at both presynaptic and postsynaptic terminals. At the presynapse, the process regulates the recycling of neurotransmitter vesicles to maintain a pool of ready-to-release synaptic vesicles that enables fast signalling103. At the postsynapse, clathrin-mediated endocytosis mediates plasticity through receptor trafficking and signal transduction cascades36. Synaptic APP and β-secretase undergo internalization from the plasma membrane through clathrin-mediated endocytosis, and despite separate secretory pathways, the rapid fusion of vesicles containing APP and β-secretase with recycling endosomes regulates the activity-dependent cleavage of APP into amyloid-β (ref. 29). The activity-dependent trafficking of late endosomes also mediates the proximity of γ-secretase to APP28. Interestingly, landmark studies in yeast that investigated amyloid-β-induced dysfunction identified several genetic modifiers of toxicity in the endocytic pathway104. As well as genetic risk factors that can impair plasticity and enhance amyloid-β production, amyloid-β itself can disrupt clathrin-mediated endocytosis by altering the pattern of clathrin aggregation at the plasma membrane, which exemplifies the complexity of the feedback mechanisms that can go awry in AD104.

As with intracellular degradation and clearance, maintenance of the extracellular environment also requires the removal of debris to protect connectivity and signalling capacities. The cells that hold the main responsibility for this elimination in the brain are microglia, which survey their surroundings for unwanted and potentially disease-causing debris105. Microglia can remove such waste from the parenchyma through specialized endocytic processes called phagocytosis and pinocytosis, and the uptake of amyloid-β by microglia seems to be an important means of amyloid-β-aggregate degradation. Loss of the late-onset AD risk gene TREM2 reduced phagocytic cell migration towards amyloid-β plaques and exacerbated amyloid-β deposition106, and phagocytic genes were downregulated in microglia following amyloid-β accumulation107. It is therefore possible that pathogenic feed-forward processes are involved in which deficits in phagocytosis, induced either genetically or by amyloid-β, lead to increased levels of amyloid-β and further dysfunction. Microglia have also been shown to take up amyloid-β oligomers through pinocytosis, a related endocytic process108. Because these oligomers induce long-term depression and have other deleterious effects on the synapse36, the loss of phagocytic functions might contribute to inefficient communication and circuit dysfunction even before the formation of overt plaques.

Emerging functions of microglia at synapses suggest that these cells also participate directly in synaptic remodelling77, 105. Although considerable data implicate microglia in the development of AD, the exact role of these cells in disease progression remains uncertain. Microglia can both protect and damage circuits, which further complicates understanding of their contribution to AD77, 109. For example, lipopolysaccharide-activated microglia shear inhibitory presynaptic terminals from the soma of excitatory neurons to enhance coherent firing and upregulate protective pathways109. But in response to oligomeric amyloid-β, microglia eliminate synapses through a complement-mediated pathway in a mouse model of AD77. Although both findings indicate that microglia have a considerable role in shaping synaptic communication, it remains unclear whether they initiate neurodegeneration, confer protection or affect both outcomes. These contrasting results might reflect the distinct responses of microglia to the different phases of AD. More generally, microglia-mediated immune responses probably affect the progression of AD as well; however, these responses are also poorly understood in AD. Future studies that use omics-based screens and live-cell imaging will be necessary to disentangle the complex functions of microglia in AD.

Microglia are not the only non-neuronal cells that affect circuits: astrocytes, oligodendrocytes and endothelial cells also have considerable roles in maintaining circuit function. In AD, each of these cell types undergoes changes that probably affect neuronal activity, a topic that is thoroughly covered elsewhere9. Building on the conclusion that mapping alterations in cells can provide insights into incipient changes that underlie memory loss, it will also be crucial to understand how such changes alter circuit dynamics to affect cognition. For example, a striking finding in both amyloid-β and tau models of AD is that the presence of aggregates increases the size of hippocampal place fields and decreases their stability110, 111. The ensemble firing of place cells in normal brains generates a map that promotes spatial navigation and location recognition, and alterations could explain the spatial deficits that are seen in people with AD112, 113. It is unknown whether such deficits are the result of the intrinsic inability of place cells to encode information or altered network dynamics; however, reducing hyperexcitability in the dentate gyrus and CA3 subfields of the hippocampus in a mouse model of AD and in people with AD who use levetiracetam restores cognition40, 41. Although the restoration was observed in an amyloid-β mouse model of AD, it seems to be independent of any effects on amyloid-β and correlates instead with increases in synaptic markers41. This suggests that complex alterations at both the circuit and network levels might contribute independently to cognitive decline.

Network dysfunction predicts cognitive symptoms

The first evidence for network dysfunction in AD was the observation of cholinergic neuron loss in postmortem brains114. Cholinergic neurons are found in distinct subcortical regions of the brain and use acetylcholine as a neurotransmitter. The regions are connected to each other, and the cholinergic neurons have processes that extend as a diffuse network through the brain, which enable them to modulate the excitatory tone of circuits and to facilitate states such as sleep, wakefulness and attentiveness115. Identification of neurofibrillary-tangle formation and degeneration in cholinergic neurons116 supports the cholinergic hypothesis, which states that the loss of cholinergic neurons is an event that leads to AD117. These cells affect cognitive processing through the modulation of circuit activity and coherent firing118, and drugs that target cholinergic signalling account for 75% of treatments for AD that have been approved by the US Food and Drug Administration. However, the efficacy of therapies that target cholinergic pathways is short and they might temporarily slow, but do not halt or reverse, memory decline119. Although enhancing cholinergic signalling cannot reverse the progression of AD, the limited efficacy of such treatments suggests that the loss of regulatory tone might exacerbate dysfunction and accelerate memory loss. Accumulating evidence indicates that the long-range structure of the network could make these cells particularly susceptible to deficits in axonal transport and metabolic insults that induce neurofibrillary tangles and other pathologies120. The identification of further subcortical neuromodulatory systems that show neurofibrillary tangles at the earliest stages of disease indicates that studying these networks could pinpoint specific vulnerabilities in cellular processes that change in the initial stages of AD121.

The first symptoms that people with AD report are a difficulty in remembering new information and episodic memory loss67. Neuropathological examinations show that the accumulation of neurofibrillary tangles occurs initially in the entorhinal cortex (EC) of the brain then spreads through the hippocampal formation122. Advances in PET tracers have enabled tau localization to be imaged with 18F-T807; this revealed similar patterns to those determined by pathological studies of AD brains, which suggests that these tools might be useful for diagnosing AD123. In support of this, alterations in entorhinal cortex–hippocampal regions of people with neurological disorders but not AD, as well as in rats with neural lesions in the entorhinal cortex, impair learning and the retrieval of episodic memories, which shows that the anatomical findings correspond with the initial clinical symptoms of AD124, 125. This indicates that processes in these regions that are disrupted in AD can alter circuits that impair learning and memory. Also, because the accumulation of neurofibrillary tangles develops subsequently in connected brain regions that control planning, emotion and navigation, the temporal sequence of increasing burden might explain the deterioration of cognitive function as AD advances126, 127.

Both neurofibrillary tangles and amyloid-β plaques propagate through the limbic system, a network of distinct brain regions that control memory and emotional behaviours122, 128. The sequential involvement of limbic brain regions, and the loss of their proper function, could explain the progression of clinical symptoms, and MRI studies in people with AD demonstrate that specific atrophy in nodes of the network correlates with cognitive complaints129. Limbic regions receive regulatory inputs from subcortical structures, in particular from those that accumulate tau. Therefore, such accumulations could contribute to dysregulation of the limbic regions130. PET imaging of glucose metabolism shows a considerable reduction in glucose use in the hippocampal formation and other limbic temporal regions131. Furthermore, disruption of limbic white matter is correlated with cognitive decline in people with amnestic mild cognitive impairment132. Evidence from clinical trials supports an underlying role for limbic network dysfunction in memory loss; preliminary data from a small cohort of people with AD suggest that electrical stimulation of the fornix, a main white-matter bundle that connects limbic regions, leads to enhanced cognition133. Stimulation also induces DMN activation, which indicates that the DMN is highly interconnected with many limbic nodes133. This observation could also reconcile the temporal sequence of cognitive symptoms and point to the earliest hubs of disease generation. In the future, longitudinal, high-resolution functional MRI or PET-based investigations should examine specifically the subcortical limbic nodes to ascertain whether amyloid-β deposits and functional disconnection can be detected there before deficits are seen in DMN function. If so, innovative methods for assessing limbic function could be useful for identifying people with the earliest stage of AD.

As well as the uncertainty about the core networks that are affected by AD, the mechanisms that underlie the spread of neurofibrillary tangles and amyloid-β are unknown. Amyloid-β peptides and tau proteins released at the synapse could propagate in a prion-like manner, but the biological pathways that underlie this potential movement are unclear and it remains technically difficult to demonstrate seed propagation directly. An alternative explanation for the spread of aggregates is that local circuit dysfunction causes alterations in network activity, which drives increases in amyloid-β levels and tau aggregation through abnormal activity at downstream targets. Regional vulnerability might lead to the propagation of disrupted cascades, which could develop into the hallmarks of AD, if left unchecked. Although increasing evidence suggests that amyloid-β and tau spread within networks, region-specific responses to experiential factors such as ageing, diet, emotional stress and mental health might also explain the distribution of disease hallmarks11, 120. The limbic system is particularly responsive to stress and to the diet, and many of its main regions, including the hippocampal formation, express high levels of receptors for stress-related glucocorticoids and gut hormones134, 135. Indeed, high-fat diets and chronic stress can damage the hippocampal formation, and these environmentally responsive limbic structures might be important points at which experience, genetic risk and the onset of AD converge120, 136. In a mouse model of AD, emotional stressors exacerbated the loss of dendritic spines in the hippocampal formation, and repeated exposure to a stressor in normal mice led to a neurodegenerative-like phenotype in the same brain region, as well as cognitive impairment137, 138.

Network treatments might restore memory

Therapeutic drugs that target amyloid-β have been largely unsuccessful at restoring cognition, and most treatments for AD that have the potential to improve cognition target networks to improve or protect circuit integrity and activity40, 64, 119, 133, 139, 140, 141 (Fig. 3). The main drugs prescribed to people with early stage memory loss — donepezil, rivastigmine and galantamine — promote cholinergic signalling by inhibiting acetylcholine degradation to maintain its levels in the extracellular space119, which is thought to promote neuronal activity and improve cholinergic tone. These compounds also probably affect microglia and astrocytes, both of which express acetylcholine receptors and respond to cholinergic signalling141, 142. Another therapeutic drug approved for the treatment of AD is memantine, an NMDA (N-methyl-D-aspartate) receptor antagonist119. It is the only compound that has been approved for use in late-stage AD, and it could promote circuit connectivity by preserving synapse integrity. Memantine probably counteracts the effects of hyperactive excitatory circuits and prevents high levels of glutamate from weakening synapse strength141. Similar to molecular strategies that target amyloid-β, these network-level treatments are mildly effective. However, they have yet to stop cognitive deterioration or to restore memory function.

Figure 3: Network-level treatments for AD.
Network-level treatments for AD.

Drugs that target cholinergic and glutamatergic circuits have minimal efficacy yet are the only approved treatments for AD119. Preliminary studies using deep-brain stimulation electrodes in the basal forebrain cholinergic circuit (green) and hypothalamus–hippocampal network (red) have shown promise in slowing disease progression133, 144. Direct stimulation of amygdala–hippocampal circuits (purple) in rodents also improves memory, which suggests that targeting these limbic circuits directly might be another route to restoring cognition. The implementation of non-invasive technologies such as transcranial magnetic stimulation in network modification makes such treatments safer and scalable.

Other molecular treatments to restore cellular health and to repair circuit and network functions are under development. Alongside those for amyloid-β, antibody-based treatments that target tau aggregation and APOE have also emerged, although evaluating their efficacy using data from clinical trials is not yet possible143. These antibody-based strategies probably influence immune functions; however, the increasing evidence for non-inflammatory roles of microglia in circuit maintenance suggests that targeting other microglia processes could be a better route to mitigating cognitive impairment.

Directly targeting the activity of brain networks might also help to restore memory. Phase I clinical trials that used deep-brain stimulation techniques to directly manipulate network activity in individuals with AD reported positive memory outcomes133, 144. Stimulation of the fornix also changed protein expression in animal models, which suggests that the development of non-invasive brain-stimulation strategies could be a scalable and safe route to restoring cellular health and network function145. Furthermore, using optogenetic techniques to excite cells in the hippocampus to increase the number of dendritic spines restored learning and memory139. Because similar effects were found using histone deacetylase inhibitors140, it is probable that diverse treatment strategies can leverage these effects to repair circuits. The optogenetics139 and histone deacetylase140 studies are particularly interesting because they used inducible models as a control for learning. They showed that, despite considerable memory impairment following neuronal degeneration, the restoration of synapses could restore recall performance. This suggests that memories in AD might be inaccessible rather than lost. Together, the efficacy of these experiments demonstrates that intervention at both the network and circuit levels can restore cellular health and circuit integrity and could provide new directions for restoring memory in AD.


A growing understanding of cognitive impairment in AD suggests that alterations at the genetic and cellular levels contribute to circuit dysfunction, which affects long-range network connectivity. By restoring these connections and the circuit-level and cellular-level processes that support them, it might be possible to reverse memory loss. Because no single strategy or target has been fully effective in promoting cognition, it is probable that a multitiered approach to treating AD will be necessary. Treatments that reduce amyloid-β levels stop the cascade that triggers cellular dysfunction, and these will be especially important for people with familial AD, in which genetic mutations induce the overproduction of amyloid-β directly. Together with a reduction in amyloid-β to halt disease progression, restoring cognition and brain health will probably need therapeutic drugs that regulate circuit activity and stimulate neuronal communication to improve the function of long-range networks. The conventional path to these treatments requires an understanding of which disrupted processes contribute to amyloid-β accumulation and circuit disruption. However, because circuit activity also affects cellular processes, it may be possible to circumvent this pipeline by restoring network-level and circuit-level functions in people with AD. These treatments might be able to directly improve memory function and feedback on molecular processes to re-establish cellular health, paving the way to a healthy ageing brain.


  1. Hurd, M. D., Martorell, P., Delavande, A., Mullen, K. J. & Langa, K. M. Monetary costs of dementia in the United States. N. Engl. J. Med. 368, 13261334 (2013).
  2. Hyman, B. T. et al. National Institute on Aging—Alzheimer's Association guidelines for the neuropathologic assessment of Alzheimer's disease. Alzheimers Dement. 8, 113 (2012).
  3. Goate, A. et al. Segregation of a missense mutation in the amyloid precursor protein gene with familial Alzheimer's disease. Nature 349, 704706 (1991).
    One of the first demonstrations that mutations in the gene APP correlate with familial AD, suggesting a prominent role for amyloid-β processing in the aetiology of AD.
  4. Schellenberg, G. et al. Genetic linkage evidence for a familial Alzheimer's disease locus on chromosome 14. Science 258, 668671 (1992).
  5. Levy-Lahad, E. et al. Candidate gene for the chromosome 1 familial Alzheimer's disease locus. Science 269, 973977 (1995).
  6. Hardy, J. & Higgins, G. Alzheimer's disease: the amyloid cascade hypothesis. Science 256, 184185 (1992).
    First article to synthesize the available data to formally propose that amyloid-β leads to varied neuronal disruption and cognitive impairment in AD.
  7. Hardy, J. & Selkoe, D. J. The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics. Science 297, 353356 (2002).
  8. Lambert, J. C. et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nature Genet. 45, 14521458 (2013).
    By aggregating data, this paper confirmed previous genetic risk factors associated with late-onset AD and also identified new loci that might increase susceptibility to the disease.
  9. De Strooper, B. & Karran, E. The cellular phase of Alzheimer's disease. Cell 164, 603615 (2016).
  10. Palop, J. J. & Mucke, L. Amyloid-β-induced neuronal dysfunction in Alzheimer's disease: from synapses toward neural networks. Nature Neurosci. 13, 812818 (2010).
  11. Reitz, C. & Mayeux, R. Alzheimer disease: epidemiology, diagnostic criteria, risk factors and biomarkers. Biochem. Pharmacol. 88, 640651 (2014).
  12. Hardy, J. The Alzheimer family of diseases: many etiologies, one pathogenesis? Proc. Natl Acad. Sci. USA 94, 20952097 (1997).
  13. Rovelet-Lecrux, A. et al. APP locus duplication causes autosomal dominant early-onset Alzheimer disease with cerebral amyloid angiopathy. Nature Genet. 38, 2426 (2006).
  14. Glenner, G. G. & Wong, C. W. Alzheimer's disease and Down's syndrome: sharing of a unique cerebrovascular amyloid fibril protein. Biochem. Biophys. Res. Commun. 122, 11311135 (1984).
  15. Hutton, M. et al. Association of missense and 5′-splice-site mutations in tau with the inherited dementia FTDP-17. Nature 393, 702705 (1998).
  16. Jonsson, T. et al. A mutation in APP protects against Alzheimer's disease and age-related cognitive decline. Nature 488, 9699 (2012).
  17. Corder, E. H. et al. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer's disease in late onset families. Science 261, 921923 (1993).
  18. Morris, J. C. et al. APOE predicts amyloid-beta but not tau Alzheimer pathology in cognitively normal aging. Ann. Neurol. 67, 122131 (2010).
  19. Verghese, P. B. et al. ApoE influences amyloid-β (Aβ) clearance despite minimal apoE/Aβ association in physiological conditions. Proc. Natl Acad. Sci. USA 110, E1807E1816 (2013).
  20. Kayed, R. & Lasagna-Reeves, C. Molecular mechanisms of amyloid oligomers toxicity. J. Alzheimers Dis. 33, S67S78 (2013).
  21. Yankner, B. A., Duffy, L. K. & Kirschner, D. A. Neurotrophic and neurotoxic effects of amyloid beta protein: reversal by tachykinin neuropeptides. Science 250, 279282 (1990).
  22. Jack, C. R. et al. Serial PIB and MRI in normal, mild cognitive impairment and Alzheimer's disease: implications for sequence of pathological events in Alzheimer's disease. Brain 132, 13551365 (2009).
    Showed how certain events unfold in people with AD and contributed to an understanding of the temporal disconnection between amyloid-β deposition and cognitive impairment.
  23. Shankar, G. M. et al. Natural oligomers of the Alzheimer amyloid-β protein induce reversible synapse loss by modulating an NMDA-type glutamate receptor-dependent signaling pathway. J. Neurosci. 27, 28662875 (2007).
  24. Šišková, Z. et al. Dendritic structural degeneration is functionally linked to cellular hyperexcitability in a mouse model of Alzheimer's disease. Neuron 84, 10231033 (2014).
  25. Palop, J. J. et al. Aberrant excitatory neuronal activity and compensatory remodeling of inhibitory hippocampal circuits in mouse models of Alzheimer's disease. Neuron 55, 697711 (2007).
    Demonstrated that amyloid-β and AD-like features could induce hyperactivity in regions of the brain, challenging the idea that neurodegeneration leads to reduced neuronal activity and highlighting the complexity of changes seen in AD.
  26. Cirrito, J. R. et al. Synaptic activity regulates interstitial fluid amyloid-β levels in vivo. Neuron 48, 913922 (2005).
    First to indicate a considerable physiological role for amyloid-β in the brain, suggesting it could have important functions that influence AD phenotypes.
  27. Wei, W. et al. Amyloid beta from axons and dendrites reduces local spine number and plasticity. Nature Neurosci. 13, 190196 (2010).
  28. Wu, J. et al. Arc/Arg3.1 regulates an endosomal pathway essential for activity-dependent β-amyloid generation. Cell 147, 615628 (2011).
  29. Das, U. et al. Activity-induced convergence of APP and BACE-1 in acidic microdomains via an endocytosis-dependent pathway. Neuron 79, 447460 (2013).
  30. Dougherty, J. J., Wu, J. & Nichols, R. A. β-Amyloid regulation of presynaptic nicotinic receptors in rat hippocampus and neocortex. J. Neurosci. 23, 67406747 (2003).
  31. Abramov, E. et al. Amyloid-β as a positive endogenous regulator of release probability at hippocampal synapses. Nature Neurosci. 12, 15671576 (2009).
  32. Snyder, E. M. et al. Regulation of NMDA receptor trafficking by amyloid-β. Nature Neurosci. 8, 10511058 (2005).
  33. Roselli, F. et al. Soluble β-amyloid1–40 induces NMDA-dependent degradation of postsynaptic density-95 at glutamatergic synapses. J. Neurosci. 25, 1106111070 (2005).
  34. Walsh, D. M. et al. Naturally secreted oligomers of amyloid β protein potently inhibit hippocampal long-term potentiation in vivo. Nature 416, 535539 (2002).
  35. Li, S. et al. Soluble oligomers of amyloid β protein facilitate hippocampal long-term depression by disrupting neuronal glutamate uptake. Neuron 62, 788801 (2009).
  36. Hsieh, H. et al. AMPAR removal underlies Aβ-induced synaptic depression and dendritic spine loss. Neuron 52, 831843 (2006).
  37. Verret, L. et al. Inhibitory interneuron deficit links altered network activity and cognitive dysfunction in Alzheimer model. Cell 149, 708721 (2012).
  38. Vossel, K. A. et al. Seizures and epileptiform activity in the early stages of Alzheimer disease. JAMA Neurol. 70, 11581166 (2013).
    Provided an early demonstration of neuronal hyperactivity in people with AD and confirmed findings made originally in rodent models of AD, suggesting that such models can recapitulate facets of AD accurately, despite limitations.
  39. Volicer, L., Smith, S. & Volicer, B. J. Effect of seizures on progression of dementia of the Alzheimer type. Dementia 6, 258263 (1995).
  40. Bakker, A., Albert, M. S., Krauss, G., Speck, C. L. & Gallagher, M. Response of the medial temporal lobe network in amnestic mild cognitive impairment to therapeutic intervention assessed by fMRI and memory task performance. Neuroimage Clin. 7, 688698 (2015).
  41. Sanchez, P. E. et al. Levetiracetam suppresses neuronal network dysfunction and reverses synaptic and cognitive deficits in an Alzheimer's disease model. Proc. Natl Acad. Sci. USA 109, E2895E2903 (2012).
  42. Busche, M. A. et al. Rescue of long-range circuit dysfunction in Alzheimer's disease models. Nature Neurosci. 18, 16231630 (2015).
  43. Thal, D. R., Rüb, U., Orantes, M. & Braak, H. Phases of Aβ-deposition in the human brain and its relevance for the development of AD. Neurology 58, 17911800 (2002).
  44. Klunk, W. E. et al. Imaging brain amyloid in Alzheimer's disease with Pittsburgh compound-B. Ann. Neurol. 55, 306319 (2004).
    This important paper was the first to describe the PET imaging of amyloids in people with AD.
  45. Wong, D. F. et al. In vivo imaging of amyloid deposition in Alzheimer disease using the radioligand 18F-AV-45 (florbetapir F 18). J. Nucl. Med. 51, 913920 (2010).
  46. Jack, C. R. et al. Brain beta-amyloid measures and magnetic resonance imaging atrophy both predict time-to-progression from mild cognitive impairment to Alzheimer's disease. Brain 133, 33363348 (2010).
  47. Forsberg, A. et al. PET imaging of amyloid deposition in patients with mild cognitive impairment. Neurobiol. Aging 29, 14561465 (2008).
  48. Albert, M. S. et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging–Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 7, 270279 (2011).
  49. Greicius, M. D., Srivastava, G., Reiss, A. L. & Menon, V. Default-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional MRI. Proc. Natl Acad. Sci. USA 101, 46374642 (2004).
  50. Buckner, R. L. Molecular, structural, and functional characterization of Alzheimer's disease: evidence for a relationship between default activity, amyloid, and memory. J. Neurosci. 25, 77097717 (2005).
  51. Sperling, R. A., LaViolette, P. & O'Keefe, K. Amyloid deposition is associated with impaired default network function in older persons without dementia. Neuron 63, 178188 (2009).
  52. Mormino, E. C. et al. Synergistic effect of β-amyloid and neurodegeneration on cognitive decline in clinically normal individuals. JAMA Neurol. 71, 13791385 (2014).
  53. Villemagne, V. L. et al. Longitudinal assessment of Aβ and cognition in aging and Alzheimer disease. Ann. Neurol. 69, 181192 (2011).
  54. Damoiseaux, J. S., Prater, K. E., Miller, B. L. & Greicius, M. D. Functional connectivity tracks clinical deterioration in Alzheimer's disease. Neurobiol. Aging 33, 828.e19828.e30 (2012).
  55. Harris, J. A. et al. Transsynaptic progression of amyloid-β-induced neuronal dysfunction within the entorhinal–hippocampal network. Neuron 68, 428441 (2010).
  56. Jucker, M. & Walker, L. C. Self-propagation of pathogenic protein aggregates in neurodegenerative diseases. Nature 501, 4551 (2013).
  57. Fleisher, A. S. et al. Phase 2 safety trial targeting amyloid β production with a γ-secretase inhibitor in Alzheimer disease. Arch. Neurol. 65, 10311038 (2008).
  58. May, P. C. et al. The Potent BACE1 inhibitor LY2886721 elicits robust central Aβ pharmacodynamic responses in mice, dogs, and humans. J. Neurosci. 35, 11991210 (2015).
  59. De Strooper, B. Lessons from a failed γ-secretase Alzheimer trial. Cell 159, 721726 (2014).
  60. Vassar, R. BACE1 inhibitor drugs in clinical trials for Alzheimer's disease. Alzheimers Res. Ther. 6, 89 (2014).
  61. Doody, R. S. et al. A phase 3 trial of semagacestat for treatment of Alzheimer's disease. N. Engl. J. Med. 369, 341350 (2013).
  62. Orgogozo, J. M. et al. Subacute meningoencephalitis in a subset of patients with AD after Aβ42 immunization. Neurology 61, 4654 (2003).
  63. Farlow, M. R. et al. Long-term treatment with active Aβ immunotherapy with CAD106 in mild Alzheimer's disease. Alzheimers Res. Ther. 7, 23 (2015).
  64. Holmes, C. et al. Long-term effects of Aβ42 immunisation in Alzheimer's disease: follow-up of a randomised, placebo-controlled phase I trial. Lancet 372, 216223 (2008).
  65. Siemers, E. R. et al. Phase 3 solanezumab trials: secondary outcomes in mild Alzheimer's disease patients. Alzheimers Dement. 12, 110120 (2016).
  66. J. Sevigny et al. The antibody aducanumab reduces Aβ plaques in Alzheimer's disease. Nature 537, 5056 (2016).
  67. McKhann, G. M. et al. The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging–Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 7, 263269 (2011).
  68. van der Kant, R. & Goldstein, L. S. B. Cellular functions of the amyloid precursor protein from development to dementia. Dev. Cell 32, 502515 (2015).
  69. Duggan, S. P. & McCarthy, J. V. Beyond γ-secretase activity: the multifunctional nature of presenilins in cell signalling pathways. Cell. Signal. 28, 111 (2016).
  70. Müller, U. et al. Behavioral and anatomical deficits in mice homozygous for a modified β-amyloid precursor protein gene. Cell 79, 755765 (1994).
  71. Dawson, G. R. et al. Age-related cognitive deficits, impaired long-term potentiation and reduction in synaptic marker density in mice lacking the β-amyloid precursor protein. Neuroscience 90, 113 (1999).
  72. Xia, D. et al. Presenilin-1 knockin mice reveal loss-of-function mechanism for familial Alzheimer's disease. Neuron 85, 967981 (2015).
  73. Nelson, O. et al. Familial Alzheimer disease-linked mutations specifically disrupt Ca2+ leak function of presenilin 1. J. Clin. Invest. 117, 12301239 (2007).
  74. Fabelo, N. et al. Altered lipid composition in cortical lipid rafts occurs at early stages of sporadic Alzheimer's disease and facilitates APP/BACE1 interactions. Neurobiol. Aging 35, 18011812 (2014).
  75. Sakae, N. et al. ABCA7 deficiency accelerates amyloid-β generation and Alzheimer's neuronal pathology. J. Neurosci. 36, 38483859 (2016).
  76. Miller, S. E. et al. CALM regulates clathrin-coated vesicle size and maturation by directly sensing and driving membrane curvature. Dev. Cell 33, 163175 (2015).
  77. Hong, S. et al. Complement and microglia mediate early synapse loss in Alzheimer mouse models. Science 352, 712716 (2016).
  78. De Jager, P. L. et al. Alzheimer's disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci. Nature Neurosci. 17, 11561163 (2014).
  79. Yu, L. et al. Association of brain DNA methylation in SORL1, ABCA7, HLA-DRB5, SLC24A4, and BIN1 with pathological diagnosis of Alzheimer disease. JAMA Neurol. 72, 1524 (2015).
  80. Gjoneska, E. et al. Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer's disease. Nature 518, 365369 (2015).
  81. Gräff, J. et al. An epigenetic blockade of cognitive functions in the neurodegenerating brain. Nature 483, 222226 (2012).
  82. Lu, T. et al. REST and stress resistance in ageing and Alzheimer's disease. Nature 507, 448454 (2014).
  83. Jo, D.-G. et al. Evidence that γ-secretase mediates oxidative stress-induced β-secretase expression in Alzheimer's disease. Neurobiol. Aging 31, 917925 (2010).
  84. Mecocci, P., MacGarvey, U. & Beal, M. F. Oxidative damage to mitochondrial DNA is increased in Alzheimer's disease. Ann. Neurol. 36, 747751 (1994).
  85. Frost, B., Hemberg, M., Lewis, J. & Feany, M. B. Tau promotes neurodegeneration through global chromatin relaxation. Nature Neurosci. 17, 357366 (2014).
  86. Kim, D. et al. Deregulation of HDAC1 by p25/Cdk5 in neurotoxicity. Neuron 60, 803817 (2008).
  87. Anderson, A. J., Stoltzner, S., Lai, F., Su, J. & Nixon, R. A. Morphological and biochemical assessment of DNA damage and apoptosis in Down syndrome and Alzheimer disease, and effect of postmortem tissue archival on TUNEL. Neurobiol. Aging 21, 511524 (2000).
  88. Suberbielle, E. et al. Physiologic brain activity causes DNA double-strand breaks in neurons, with exacerbation by amyloid-β. Nature Neurosci. 16, 613621 (2013).
  89. Lu, T. et al. Gene regulation and DNA damage in the ageing human brain. Nature 429, 883891 (2004).
  90. Madabhushi, R. et al. Activity-induced DNA breaks govern the expression of neuronal early-response genes. Cell 161, 15921605 (2015).
  91. LaFerla, F. M. Calcium dyshomeostasis and intracellular signalling in Alzheimer's disease. Nature Rev. Neurosci. 3, 862872 (2002).
  92. Seo, J. et al. Activity-dependent p25 generation regulates synaptic plasticity and Aβ-induced cognitive impairment. Cell 157, 486498 (2014).
  93. Lindwall, G. & Cole, R. D. Phosphorylation affects the ability of tau protein to promote microtubule assembly. J. Biol. Chem. 259, 53015305 (1984).
  94. Bramblett, G. T. et al. Abnormal tau phosphorylation at Ser396 in Alzheimer's disease recapitulates development and contributes to reduced microtubule binding. Neuron 10, 10891099 (1993).
  95. Sontag, E., Nunbhakdi-Craig, V., Lee, G., Bloom, G. S. & Mumby, M. C. Regulation of the phosphorylation state and microtubule-binding activity of tau by protein phosphatase 2A. Neuron 17, 12011207 (1996).
  96. Vogelsberg-Ragaglia, V., Schuck, T., Trojanowski, J. Q. & Lee, V. M. PP2A mRNA expression is quantitatively decreased in Alzheimer's disease hippocampus. Exp. Neurol. 168, 402412 (2001).
  97. Cruz, J. C., Tseng, H. C., Goldman, J. A., Shih, H. & Tsai, L. H. Aberrant Cdk5 activation by p25 triggers pathological events leading to neurodegeneration and neurofibrillary tangles. Neuron 40, 471483 (2003).
  98. Yamada, K. et al. Neuronal activity regulates extracellular tau in vivo. J. Exp. Med. 211, 387393 (2014).
  99. Pooler, A. M., Phillips, E. C., Lau, D. H. W., Noble, W. & Hanger, D. P. Physiological release of endogenous tau is stimulated by neuronal activity. EMBO Rep. 14, 389394 (2013).
  100. Wu, J. W. et al. Neuronal activity enhances tau propagation and tau pathology in vivo. Nature Neurosci. 19, 10851092 (2016).
  101. Cohen, T. J. et al. The acetylation of tau inhibits its function and promotes pathological tau aggregation. Nature Commun. 2, 252 (2011).
  102. Moreau, K. et al. PICALM modulates autophagy activity and tau accumulation. Nature Commun. 5, 4998 (2014).
  103. Kononenko, N. L. & Haucke, V. Molecular mechanisms of presynaptic membrane retrieval and synaptic vesicle reformation. Neuron 85, 484496 (2015).
  104. Treusch, S. et al. Functional links between Aβ toxicity, endocytic trafficking, and Alzheimer's disease risk factors in yeast. Science 334, 12411245 (2011).
  105. Crotti, A. & Ransohoff, R. M. Microglial physiology and pathophysiology: insights from genome-wide transcriptional profiling. Immunity 44, 505515 (2016).
  106. Wang, Y. et al. TREM2 lipid sensing sustains the microglial response in an Alzheimer's disease model. Cell 160, 10611071 (2015).
    The first demonstration that AD genetic risk loci affect microglia, which play an important part in modulating the AD state.
  107. Orre, M. et al. Isolation of glia from Alzheimer's mice reveals inflammation and dysfunction. Neurobiol. Aging 35, 27462760 (2014).
  108. Mandrekar, S. et al. Microglia mediate the clearance of soluble Aβ through fluid phase macropinocytosis. J. Neurosci. 29, 42524262 (2009).
  109. Chen, Z. et al. Microglial displacement of inhibitory synapses provides neuroprotection in the adult brain. Nature Commun. 5, 4486 (2014).
  110. Cacucci, F., Yi, M., Wills, T. J., Chapman, P. & O'Keefe, J. Place cell firing correlates with memory deficits and amyloid plaque burden in Tg2576 Alzheimer mouse model. Proc. Natl Acad. Sci. USA 105, 78637868 (2008).
  111. Ciupek, S. M., Cheng, J., Ali, Y. O., Lu, H.-C. & Ji, D. Progressive functional impairments of hippocampal neurons in a tauopathy mouse model. J. Neurosci. 35, 81188131 (2015).
  112. Morris, R. G., Garrud, P., Rawlins, J. N. & O'Keefe, J. Place navigation impaired in rats with hippocampal lesions. Nature 297, 681683 (1982).
  113. Wilson, M. A. & McNaughton, B. L. Dynamics of the hippocampal ensemble code for space. Science 261, 10551058 (1993).
  114. Whitehouse, P. J., Price, D. L., Clark, A. W., Coyle, J. T. & DeLong, M. R. Alzheimer disease: evidence for selective loss of cholinergic neurons in the nucleus basalis. Ann. Neurol. 10, 122126 (1981).
  115. Metherate, R., Cox, C. L. & Ashe, J. H. Cellular bases of neocortical activation: modulation of neural oscillations by the nucleus basalis and endogenous acetylcholine. J. Neurosci. 12, 47014711 (1992).
  116. Mesulam, M., Shaw, P., Mash, D. & Weintraub, S. Cholinergic nucleus basalis tauopathy emerges early in the aging–MCI–AD continuum. Ann. Neurol. 55, 815828 (2004).
  117. Bartus, R. T., Dean, R. L., Beer, B. & Lippa, A. The cholinergic hypothesis of geriatric memory dysfunction. Science 217, 408414 (1982).
  118. Hangya, B., Ranade, S. P., Lorenc, M. & Kepecs, A. Central cholinergic neurons are rapidly recruited by reinforcement feedback. Cell 162, 11551168 (2015).
  119. Zhu, C. W. et al. Long-term associations between cholinesterase inhibitors and memantine use and health outcomes among patients with Alzheimer's disease. Alzheimers Dement. 9, 733740 (2013).
  120. Mattson, M. P. Pathways towards and away from Alzheimer's disease. Nature 430, 631639 (2004).
  121. Braak, H. & Del Tredici, K. The pathological process underlying Alzheimer's disease in individuals under thirty. Acta Neuropathol. 121, 171181 (2011).
  122. Braak, H. & Braak, E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 82, 239259 (1991).
    The first comprehensive hierarchical staging of AD, which suggested a stereotypical path of progression and highlighted correlations between pathological observations of the brain and cognitive decline.
  123. Johnson, K. A. et al. Tau positron emission tomographic imaging in aging and early Alzheimer disease. Ann. Neurol. 79, 110119 (2016).
  124. Steffenach, H. A., Witter, M., Moser, M. B. & Moser, E. I. Spatial memory in the rat requires the dorsolateral band of the entorhinal cortex. Neuron 45, 301313 (2005).
  125. Vargha-Khadem, F. et al. Differential effects of early hippocampal pathology on episodic and semantic memory. Science 277, 376380 (1997).
  126. Schöll, M. et al. PET imaging of tau deposition in the aging human brain. Neuron 89, 971982 (2016).
  127. Ossenkoppele, R. et al. Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer's disease. Brain 139, 15511567 (2016).
  128. Ye, L. et al. Progression of seed-induced Aβ deposition within the limbic connectome. Brain Pathol. 25, 743752 (2015).
  129. Bruen, P. D., McGeown, W. J., Shanks, M. F. & Venneri, A. Neuroanatomical correlates of neuropsychiatric symptoms in Alzheimer's disease. Brain 131, 24552463 (2008).
  130. Iba, M. et al. Tau pathology spread in PS19 tau transgenic mice following locus coeruleus (LC) injections of synthetic tau fibrils is determined by the LC's afferent and efferent connections. Acta Neuropathol. 130, 349362 (2015).
  131. Nestor, P. J., Fryer, T. D., Smielewski, P. & Hodges, J. R. Limbic hypometabolism in Alzheimer's disease and mild cognitive impairment. Ann. Neurol. 54, 343351 (2003).
  132. Fletcher, E., Carmichael, O., Pasternak, O., Maier-Hein, K. H. & DeCarli, C. Early brain loss in circuits affected by Alzheimer's disease is predicted by fornix microstructure but may be independent of gray matter. Front. Aging Neurosci. 6, 106 (2014).
  133. Laxton, A. W. et al. A phase I trial of deep brain stimulation of memory circuits in Alzheimer's disease. Ann. Neurol. 68, 521534 (2010).
    This paper describes the feasibility of deep-brain stimulation in people and targets regions that are not typically described in studies of disrupted processes in AD; it suggests that complex brain-wide network effects occur in AD.
  134. McEwen, B. S. Central effects of stress hormones in health and disease: understanding the protective and damaging effects of stress and stress mediators. Eur. J. Pharmacol. 583, 174185 (2008).
  135. Diano, S. et al. Ghrelin controls hippocampal spine synapse density and memory performance. Nature Neurosci. 9, 381388 (2006).
  136. Lupien, S. J. et al. Cortisol levels during human aging predict hippocampal atrophy and memory deficits. Nature Neurosci. 1, 6973 (1998).
  137. Baglietto-Vargas, D. et al. Short-term modern life-like stress exacerbates Aβ-pathology and synapse loss in 3xTg-AD mice. J. Neurochem. 134, 915926 (2015).
  138. Rei, D. et al. Basolateral amygdala bidirectionally modulates stress-induced hippocampal learning and memory deficits through a p25/Cdk5-dependent pathway. Proc. Natl Acad. Sci. USA 112, 72917296 (2015).
  139. Roy, D. S. et al. Memory retrieval by activating engram cells in mouse models of early Alzheimer's disease. Nature 531, 508512 (2016).
  140. Fischer, A., Sananbenesi, F., Wang, X., Dobbin, M. & Tsai, L.-H. Recovery of learning and memory is associated with chromatin remodelling. Nature 447, 178182 (2007).
    One of the first papers to highlight the importance of epigenetic alterations in AD and the first to demonstrate experimentally that memories can be recovered through treatment, even after considerable neurodegeneration has occurred.
  141. Talantova, M. et al. Aβ induces astrocytic glutamate release, extrasynaptic NMDA receptor activation, and synaptic loss. Proc. Natl Acad. Sci. USA 110, E2518E2527 (2013).
  142. Shytle, R. D. et al. Cholinergic modulation of microglial activation by α7 nicotinic receptors. J. Neurochem. 89, 337343 (2004).
  143. Pedersen, J. T. & Sigurdsson, E. M. Tau immunotherapy for Alzheimer's disease. Trends Mol. Med. 21, 394402 (2015).
  144. Kuhn, J. et al. Deep brain stimulation of the nucleus basalis of meynert in early stage of Alzheimer's dementia. Brain Stimul. 8, 838839 (2015).
  145. Sankar, T. et al. Deep brain stimulation influences brain structure in Alzheimer's disease. Brain Stimul. 8, 645654 (2015).
  146. Hamelin, L. et al. Early and protective microglial activation in Alzheimer's disease: a prospective study using 18F-DPA-714 PET imaging. Brain 139, 12521264 (2016).
  147. Nhan, H. S., Chiang, K. & Koo, E. H. The multifaceted nature of amyloid precursor protein and its proteolytic fragments: friends and foes. Acta Neuropathol. 129, 119 (2015).
  148. Lammich, S. et al. Constitutive and regulated α-secretase cleavage of Alzheimer's amyloid precursor protein by a disintegrin metalloprotease. Proc. Natl Acad. Sci. USA 96, 39223927 (1999).
  149. De Jonghe, C. et al. Pathogenic APP mutations near the γ-secretase cleavage site differentially affect Aβ secretion and APP C-terminal fragment stability. Hum. Mol. Genet. 10, 16651671 (2001).
  150. Willem, M. et al. η-Secretase processing of APP inhibits neuronal activity in the hippocampus. Nature 526, 443447 (2015).

Download references


We thank the US National Institutes of Health for grants R01 NS051874, R01 NS078839, RF1 AG042978 and RF1 AG047661 in support of L.-H.T. We thank the Barbara J. Weedon Fellowship and Norman B. Leventhal Fellowship for supporting R.G.C. and the Human Frontier Science Program for supporting J.P. We also thank the JPB Foundation, the Belfer Neurodegeneration Consortium, the Glenn Foundation for Medical Research, the Cure Alzheimer's Fund and the Alana Foundation for support of L.-H.T. and for continued championship of ageing and neurodegenerative disease research. We thank C. Yao for contributions to the original figure artwork. Last, we express profound gratitude to A. Watson, H. Meharena, W. Raja and N. Dedic for insightful comments on the manuscript.

Author information


  1. The Picower Institute for Learning and Memory, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA.

    • Rebecca G. Canter,
    • Jay Penney &
    • Li-Huei Tsai
  2. The Broad Institute of MIT and Harvard, Cambridge, Massachusetts 02142, USA.

    • Li-Huei Tsai

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Reprints and permissions information is available at

Reviewer Information Nature thanks the anonymous reviewer(s) for their contribution to the peer review of this work.

Author details


  1. Report this comment #69125

    Majid Ali said:

    Integration of Empirical Therapies for Brain Rejuvenation

    Majid Ali, M.D.

    Writing about the restoration of neural circuits for the treatment of Alzheimer's disease (AD), Canter and colleagues (ref. 1) write: Emerging functional data are also shifting the aetiological focus of the disease from a neuron-centric view to an integrated outlook that acknowledges the synergistic functions of the different cell types of the brain. Consistent with this holistic view, numerous disrupted processes are interconnected in Alzheimer's disease and interact to provoke a cycle of dysfunction as the disease progresses.

    In light of such an integrative outlook and a holistic view of brain rejuvenation in treating neurodegenerative disorders, the writer offers the following integrative clinical perspective of neurorejuvenation. He began Oxygen and Aging (2000) with the following words: Oxygen is the organizing principle of the human body and governs the aging process (ref. 2). In this book, drawing upon the work of some ancient writers, he offered a clinical simplicity: The guardian angel of the brain is the liver and that of the liver is the bowel. In 2004, he published evidence for respiratory-to-fermentative shift in 236 patients with chronic fatigue associated with immune-inflammatory disorders (ref. 3). He followed that with Darwin and Dysox Triology, the 10th, 11th, and 12th volumes of The Principles and Practice of Integrative Medicine, in which he presented a large body of clinical, morphologic, and biochemical data to support his perspective of the ecologic relationships among the bowel, liver, and brain (ref. 4-6).
    In the context of neurorejuvenation, the writer illustrates the core point of holistic-integrative philosophy with the following brief notes of two patients: (1) a woman in early nineties lost her husband after a brief illness and developed severe confusion and memory loss; and (2) a man with malignant melanoma who also developed acute-onset mental confusion and memory loss. His brain MRI scan revealed the presence of new brain lesions. The neurologic symptoms in both instances responded dramatically with twice weekly intramuscular injections of the following: (1) glutathione (400 mg); (2) Taurine (200 mg); (3) methylcbalamine (5,000 mcg); (4) magnesium sulfate (1500 mg); and (5) vitamin B complex (30 to 50 mg of each) for four weeks, followed by weekly injections of the same for another eight weeks. Additionally, the male patient received injectable testosterone (200 mg, slow release). This parenteral treatment was administered as a part of a broader integrative dietary and detox plan, which includes oral supplementation (ref. 7), focusing on the issues of bowel ecology (ref. 8) and hepatic detox pathways (ref. 9).

    Both individuals reported dramatic improvement in their symptoms lasting for months. The author also reports that he and his colleagues often observe moderate benefits with the above protocol in some patients with early stages of AD as well.

    In the broader context of aging populations worldwide, the important hallmarks of aging are mitochondrial dysfunction, telomere shortening, stem-cell ageing and dysfunction of intercellular communication. (ref. 10) It seems likely that dysfunctions of cellular cross-talks among neuronal populations will eventually be established as primarily rooted in disruptions of oxygen signaling, insulin homeostasis, and immune-inflammatory dynamics (11,12). This forms the scientific rationale for our focus on the foundational aspects of cellular energetics in designing treatments protocols for nutrient supplementation and dietary guidelines that address the core challenges to brain health.

    1. Canter RG, Penney J, Tsai L-H. The road to restoring neural circuits for the treatment of Alzheimer's disease. Nature. 2016; 539, 187-196.
    2. Ali M. Oxygen and Aging. (2nd ed.) New York, Canary 21 Press. 2000.
    3. Ali M. Respiratory-to-Fermentative (RTF) Shift in ATP Production in Chronic Energy Deficit States. Townsend Letter for Doctors and Patients. 2004. 253: 64-65.
    4. Ali M. Darwin, Oxygen Homeostasis, and Oxystatic Therapies. Volume X, 3 rd. Edi The Principles and Practice of Integrative Medicine (2009) New York. Institute of Integrative Medicine Press.
    5. Ali M. The Principles and Practice of Integrative Medicine Volume XI: 3rd. Edi. Darwin, Dysox, and Disease. 2000. 3rd. Edi. 2008. New York. (2009) Institute of Integrative Medicine Press.
    6. Ali M. The Principles and Practice of Integrative Medicine Volume XII: Darwin, Dysox, and Integrative Protocols. New York (2009). Institute of Integrative Medicine Press.
    8. Ali M. Ali M. Altered States of Bowel Ecology. (monograph). Teaneck, NJ, 1980.
    9. Ali M. Oxygen, Inflammation, and Castor-Cise Liver Detox. Hormones. Townsend Letter-The examiner of Alternative Medicine. 2007.
    10. Ali M. Lopez-Otin C, Blasco MA, Partridge L, et al. The hallmarks of aging. Cell. 2013;153:1194-1217.
    11. Ali M. Insulin Homeostasis in a General Population ? Shifting the Focus. Townsend Letter for Doctors and Patients. (in press).
    12. Ali M. Beyond insulin resistance and syndrome X: The oxidative-dysoxygenative insulin dysfunction (ODID) model. J Capital University of Integrative Medicine. 2001;1:101-141.

Subscribe to comments

Additional data