Noradrenaline in the aging brain: Promoting cognitive reserve or accelerating Alzheimer’s disease?

Abstract

Many believe that engaging in novel and mentally challenging activities promotes brain health and prevents Alzheimer’s disease in later life. However, mental stimulation may also have risks as well as benefits. As neurons release neurotransmitters, they often also release amyloid peptides and tau proteins into the extracellular space. These by-products of neural activity can aggregate into the tau tangle and amyloid plaque signatures of Alzheimer’s disease. Over time, more active brain regions accumulate more pathology. Thus, increasing brain activity can have a cost. But the neuromodulator noradrenaline, released during novel and mentally stimulating events, may have some protective effects—as well as some negative effects. Via its inhibitory and excitatory effects on neurons and microglia, noradrenaline sometimes prevents and sometimes accelerates the production and accumulation of amyloid-β and tau in various brain regions. Both α2A- and β-adrenergic receptors influence amyloid-β production and tau hyperphosphorylation. Adrenergic activity also influences clearance of amyloid-β and tau. Furthermore, some findings suggest that Alzheimer’s disease increases noradrenergic activity, at least in its early phases. Because older brains clear the by-products of synaptic activity less effectively, increased synaptic activity in the older brain risks accelerating the accumulation of Alzheimer’s pathology more than it does in the younger brain.

1. Introduction

1.1. Associations between the locus coeruleus and Alzheimer’s disease

The locus coeruleus (LC), a small brain nucleus in the brainstem, integrates arousal-related signals and releases noradrenaline throughout much of the brain. Recent findings indicate that the integrity of the LC is associated with cognition in late life [for reviews see [1, 2, 3]. For instance, greater neuronal density in the LC at the time of death is associated with less cognitive decline in the years before death even when controlling for neuronal density in the dorsal raphe nucleus, substantia nigra and ventral tegmental area [4]. Post-mortem analyses show fewer LC cells in patients with Alzheimer’s disease compared with controls [5-8], In postmortem human brains, the LC appears to be the earliest site of Alzheimer’s related pathology [9, 10].

In vivo estimates of LC integrity have been challenging due to its small size and lack of clear structural boundaries on typical magnetic resonance images. However, due to the LC’s different magnetic properties than surrounding tissue, specialized magnetic resonance imaging (MRI) sequences show high contrast at the site of the LC compared with surrounding brainstem regions [11]. Consistent with postmortem findings of correlations between LC structure and pre-mortem cognition or disease status, measures of LC contrast are lower in patients with mild cognitive impairment and Alzheimer’s disease than in cognitively healthy adults [12-16] and among cognitively healthy adults, higher LC contrast is associated with better memory and cognition [17-20]. Furthermore, recent findings indicate that even standard T1-weighted structural MRI sequences provide some information about LC decline. In the large Alzheimer’s Disease Neuroimaging Initiative (ADNI) sample, voxel-based morphometry from standard structural scans revealed that, within the brainstem, lower volume estimates within voxels overlapping the LC predicted future diagnoses of Alzheimer’s disease [21] and were associated with lower cognition [22].

These findings indicate that a person’s LC integrity in late life relates to their likelihood of avoiding cognitive decline and Alzheimer’s disease. What is the causal relationship? Does LC degeneration accelerate cognitive deficits and neuropathology in Alzheimer’s disease, as some have argued [23]? Indeed, some have made the case that the LC is the brain substrate of cognitive reserve [3, 24, 25]. Cognitive reserve is a concept evoked to explain the phenomenon that those with more years of education tend to have a delayed onset of the behavioral and cognitive symptoms of Alzheimer’s disease [26, 27]. The LC-as-reserve theory posits that the increased LC activity evoked during cognitively stimulating activities delays onset of the behavioral and cognitive symptoms of Alzheimer’s disease in those who have had more education [3].

This perspective suggests that stimulating noradrenergic activity via arousing, engaging activities would be a helpful strategy throughout life to stave off Alzheimer’s disease [28, 29]. Indeed, cognitive training has been one of the most popular clinical trial interventions, with hundreds of randomized controlled trials published [30-36]. Researchers have also been interested in pharmacological interventions targeting noradrenergic activity. To date, clinical trials have targeted different components of the noradrenergic system, including α2-adrenergic agonists [37-39], β-adrenergic agonists and antagonists [40, 41], an α1-adrenergic antagonist [42], and a noradrenaline transport inhibitor [43]. Yet evidence that either cognitive training or noradrenergic drugs can do anything against the progression of Alzheimer’s disease in older adults is lacking. Most cognitive training studies do not include measures of Alzheimer’s pathology and most registered trials of noradrenergic drugs have not reported results, had samples too small to be interpretable, or had null findings [37].

But as reviewed in this paper, the lack of success in establishing a causal link between interventions stimulating the noradrenergic system and the progression of Alzheimer’s disease may also be explained by findings in animal models that noradrenergic activity can either speed up or slow down Alzheimer’s-related pathophysiological processes, depending on various factors, including the type of adrenergic receptors activated.

To date, it has been assumed that cognitive training should benefit older adults’ brain structure and function, or at worst have no significant impact on the brain [44-46]. But if noradrenergic activity accelerates Alzheimer’s-related pathophysiological processes in the older brain, cognitive training in older adults may have costs beyond the time and effort involved [28]. Better understanding of how behavioral interventions interact with the noradrenergic system to slow or accelerate accumulation of Alzheimer’s-related pathology in the brain may improve future intervention attempts. This review delves into the literature to address the question of when engaging noradrenergic activity via cognitively stimulating activity helps avoid Alzheimer’s pathology and when it may instead accelerate its accumulation.

1.2. The noradrenergic system’s receptor types

To understand the mechanisms of noradrenaline action, it is important to distinguish among the different noradrenergic receptor types. There are three noradrenergic receptor classes (α2, α1, and β), with α2-adrenergic receptors having the highest affinity for noradrenaline and therefore a low threshold for activation. In contrast, β-adrenergic receptors have the lowest affinity for noradrenaline and therefore a high threshold for activation [47-49].

All three types of receptors are G protein-coupled receptors that translate signals from extracellular ligands into intracellular responses by activating specific heterotrimeric G proteins [50, 51]. Each class of noradrenergic receptors typically depends on a different family of heterotrimeric G proteins; G_i/0, G_q/11, and G_s for α2, α1, and β respectively [52]. These G protein families in turn stimulate different sets of intracellular responses [50]. However, there are exceptions to these canonical couplings of receptor types to G protein families. For example, there is evidence of β2 receptors switching from G_s to G_i [53] or existing in a G_icoupled state [54].

β-adrenergic receptors are generally excitatory, stimulating the enzyme adenylate cyclase which synthesizes cyclic adenosine monophosphate (cAMP) from adenosine triphosphate (ATP) [55]. cAMP is a second messenger that relays information into the cell and triggers chains of biochemical events, often including activation of protein-kinase A (PKA) [56] and enhanced long-term potentiation [57, 58]. While responsible for the fight-or-flight response conveyed by catecholamines, cAMP also mediates the effects of many other neurotransmitters and hormones [59].

In contrast, α2-adrenergic receptors generally inhibit neurotransmitter release and decrease neuronal excitability [60]. Positively charged calcium ions are among the ions that can flow into a cell as it undergoes depolarization, increasing neuron excitation and neurotransmitter release [61]. The majority of α2-adrenergic receptors in the human brain are of the α2A subtype [62], which decrease calcium currents [60, 63]. G_i/0 protein activity also inhibits adenylyl cyclase, leading α2-adrenergic receptors to have opposing effects on cAMP as compared with β-adrenergic receptors [50]. α2A receptors are often found at presynaptic sites, serving as autoreceptors on noradrenergic neurons that provide feedback to diminish further noradrenaline release when activated by extracellular noradrenaline [60]. They also serve as heteroreceptors on non-adrenergic neurons and cells. In fact, in mice, most effects of α2 agonist drugs, including on pain, body temperature, sedation and lowering the required dose of anesthetics are mediated by α2 receptors on non-adrenergic neurons and cells [64]. In contrast with other brain regions, in the human and monkey prefrontal cortex α2A receptors appear to be mainly postsynaptic [65, 66].

Via G_q-protein-mediated activation of phospholipase Cβ, α1 adrenergic receptors can trigger release of calcium from intracellular stores and potentiate cAMP responses [47]. Thus, unlike α2-adrenergic receptors, they increase calcium excitability, and like β-adrenergic receptors, they mobilize cAMP. Thus, at a cellular level, α1-adrenergic receptors typically have an excitatory effect. However, α1 and β-adrenergic receptors appear to have opposing effects on long-term plasticity of synapses, with α1-adrenergic receptors facilitating long-term depression while β-adrenergic receptors facilitate long-term potentiation [47].

1.3. Benefits of environmental enrichment depend on noradrenergic activity

Education and other cognitively challenging activities involve novelty, working memory and motivated behavior which each tend to activate the LC [67]. Consistent with the link between cognitive stimulation and noradrenergic activity, research with animals shows that being housed in an enriched environment increases noradrenaline levels while not changing serotonin or dopamine levels in the olfactory bulb [68] or in parieto-tempero-occipital cortex, cerebellum or pons/medulla oblongata [69]. Presumably the enriched environment increases various aspects of arousal and thereby stimulates the LC, which in turn leads to release of noradrenaline elsewhere in the brain, as well as to release of dopamine in hippocampus [70].

Noradrenaline release in brain regions involved in cognitive processing during novel arousing situations appears to be critical for the beneficial effects of environmental enrichment, as indicated by findings that rearing rodents with diminished forebrain noradrenaline in a complex environment did not improve their maze learning, unlike their controls with intact noradrenaline neurons [71, 72].

The ability of environmental enrichment to increase neurogenesis also depends on noradrenaline. In mice exposed to a novel odor each day for 47 days, neurogenesis increased in the granular cell layer of the olfactory bulb compared with mice exposed to a stable mix of those same odors for 47 days [68]. However, the enhanced neurogenesis in the enriched odor environment was eliminated if the mice received a β-adrenergic antagonist during the 47-day period. These findings that novelty stimulates neurogenesis via β-adrenergic activity in the olfactory bulb are consistent with findings that hippocampal dentate gyrus neurogenesis depends on noradrenaline [73, 74] and that β-adrenergic receptor activity promotes dentate gyrus cell proliferation [75, 76] whereas α2-adrenergic receptor activity diminishes cell proliferation in the hippocampus [77]. As reviewed previously in Section 1.2, relatively high concentrations of noradrenaline are necessary to activate β-adrenergic receptors whereas low concentrations can activate α2-adrenergic receptors. Thus, high levels of noradrenergic activity in the hippocampus can stimulate neurogenesis. The resulting new neurons may then enhance the ability to retain new information. Thus, β-adrenergic enhanced neurogenesis is one mechanism that can explain why environmental enrichment leads to better memory.

However, increased neurogenesis is likely only one of multiple noradrenergic pathways contributing to the positive effects of environmental enrichment. Another relevant factor is that noradrenaline can promote expression of plasticity-related genes. Long-lasting adaptive changes in the brain require changes in gene transcription in cell nuclei [78]. In the brain, multiple signaling cascades transmit signals from synapses to cell nuclei. Many of these are modulated by adrenergic activity.

In particular, β-adrenergic receptors promote neuronal and synaptic plasticity during environmental enrichment [47, 79]. By initiating the cAMP/PKA/CREB pathway which promotes activity-dependent changes in gene expression [59, 80-82], β-adrenergic receptors modulate cellular processes involved in establishing and maintaining memories over the long term. Destroying locus coeruleus noradrenaline neurons in adult rats reduced expression of plasticity-related genes (e.g., CREB, Arc and BDNF) across the cerebral cortex by about 50% or to the low levels seen during sleep [83]. In particular, noradrenaline appears to be critical for CREB-induced transcription in cortical astrocytes [84].

In humans, neuroimaging demonstrates that daily immersion in cognitively challenging activities such as studying a language to become a foreign language interpreter [85], studying London street routes to become a licensed taxi driver [86], practicing a video game daily [87], or learning a new motor skill [88] leads to structural brain changes. While links between these structural changes and noradrenaline have yet to be established for humans, animal studies suggest that noradrenergic activity plays a key role in this plasticity.

1.4. Local levels of glutamate modulate noradrenergic activity

During moments of high arousal, people attend to whatever matters or stands out most right then while neglecting processing of other stimuli [89]. This increased selectivity under arousal has been linked with noradrenergic processes, in particular β-adrenergic receptors [90, 91]. But how can the LC differentially target the neural processing of stimuli depending on their current priority?

To guide attention towards high priority information, the LC must interact with cortical networks representing stimuli. Stimuli are represented in the brain in neural networks defined by their synaptic connections [92-95]. Neurons representing sensory stimuli tend to form synapses with other neurons representing the same type of stimuli [e.g., 96]. Thus, to selectively enhance the processing of a particular stimulus, the LC needs to interact differently with neural networks that currently have priority than with those that do not. To make this possible, a local signal of priority that can influence noradrenaline release at that local site seems necessary.

As the brain’s most common excitatory neurotransmitter [97, 98], glutamate is well positioned to provide a local signal of priority. Glutamate is released from synaptic vesicles in nerve terminals to the extracellular space via exocytosis [as well as via non-exocytosis mechanisms 98]. However extracellular glutamate concentrations are kept quite low (in the low micromolar range compared with the intracellular concentrations which are in the millimolar range) by glutamate transporters that import it into astrocytes and neurons [97]. While a primary role of glutamate release at synaptic terminals is to excite the post-synaptic neuron, glutamate can also leak out of the synapse [99], where it can activate extrasynaptic N-methyl d-aspartate (NMDA) receptors [100], including those located on nearby axons of LC neurons [Figure 1; 101].

Figure 1.

Aspects of noradrenergic hot spots that may contribute to modulating Aβ and tau release and their impact: 1) Increased vesicle trafficking due to synaptic activity is associated with increased release of Aβ and tau; 2) Increased glutamate release is associated with increased tau release; and 3) β-adrenergic microglial activation modulates toxic effects of Aβ. Figure created with BioRender.com.

The ‘glutamate amplifies noradrenergic effects” (GANE) model posits that glutamate provides a signal of where high priority processing is occurring [101]. According to GANE, when the LC is active, glutamate can modulate levels of local noradrenaline release which in turn activate excitatory β-adrenergic receptors and thereby stimulate hot spots of high synaptic activity in local regions where there is currently high glutamatergic activity (see Figure 1). In contrast, in regions without high glutamate levels, LC activity will activate lower threshold α2-adrenergic receptors. The net effect is an increase in ‘neural gain,’ in which excited neurons become even more excited and inhibited neurons even more inhibited [102, 103]. Supporting this model, prior research indicates that glutamate increases noradrenaline release via NMDA and non-NMDA receptors on LC axons, in various rodent brain regions including prefrontal cortex, hippocampus and hypothalamus [e.g., 104, 105-108]. There is also evidence for NMDA-receptor-evoked noradrenaline release in human neocortex [109-111]. As presynaptic β2-adrenergic receptors can stimulate glutamate release [112] and increase excitatory synaptic transmission [113, 114], there is the potential for a self-amplifying feedback loop.

However, an important feature of NMDA receptors is that they conduct currents only when glutamate is bound to the NMDA receptor and the neuron that receptor is on is simultaneously depolarized [115]. Thus, NMDA receptors detect the coincident activity of pre- and postsynaptic neurons. One interesting property of LC neurons is that the timing and factors that influence release of noradrenaline from the soma versus from varicosities or terminal regions differ, such that release from the soma takes longer (with latencies as long as seconds, compared with synaptic transmission in milliseconds) and requires a higher frequency of action potentials [116] and may occur via dense core vesicles that also contain neuropeptides [117]. This slow release at the LC cell body in response to high frequency action potentials activates local α2A autoreceptors, which should inhibit activity in the LC cell body and thereby decrease noradrenaline release in neocortex [118]. Thus, this provides a mechanism to shut down intense LC activity triggered by an arousal-inducing stimulus after a few seconds, and this should in turn limit the duration of glutamate-noradrenaline hot spots elsewhere in the brain.

In the GANE hot spot model, the release of noradrenaline from LC varicosities increases excitatory synaptic activity in nearby glutamatergic neurons that in turn stimulates further noradrenaline release (as illustrated in Figure 1). This noradrenergic amplification of neural gain is supported by various mechanisms at play in and around the hot spot synapses. Astrocytes likely play a key role. One of the main functions of astrocytes is to take up unused synaptically released glutamate to prevent glutamate excitotoxicity [119]. Astrocytes with immunoreactivity for β-adrenergic receptors are often located around glutamatergic synaptic junctions, thus β-adrenergic activity on these astrocytes may modulate uptake of glutamate and other amino acids [120]. β-adrenergic signaling increased astrocytic lactate release during a startle paradigm [121]. Lactate provides additional fuel to neurons when glucose levels might not be sufficient [122]. Thus, noradrenergic signaling in astrocytes provides a mechanism to support momentarily locally high levels of metabolism. Astrocytes also express α1-adrenergic receptors that modulate their Ca2+ signaling [123].

β-adrenergic signaling increases dendrite excitability, which not only promotes the spread of depolarization [124] but may promote spike timing-dependent synaptic plasticity, which is linked with associative learning [125]. α2-adrenergic autoreceptors decrease further release of noradrenaline when they sense low levels of NE, thus suppressing activity in non-hot-spot regions. In some regions of the brain, β-adrenergic receptors may also serve as autoreceptors, increasing release of noradrenaline in the presence of high levels of noradrenaline [126]. This hot-spot glutamate-noradrenaline interaction may be most prominent in sensory regions where it is important to amplify the activity associated with specific perceptual representations under arousal. In contrast, noradrenergic activity may have different effects in cortical regions involved in coordinating broad-scale activity and tracking which perceptual representations currently have the highest priority. In particular, there is evidence that noradrenergic modulatory activity plays out differently in the prefrontal cortex than elsewhere in the brain [127]. In the prefrontal cortex, α2-adrenergic postsynaptic receptors tend to enhance prefrontal function [127, 128], while there are fewer inhibitory presynaptic α2-adrenergic receptors compared with most other brain regions [66]. The opposing effects of phasic noradrenergic activity in sensory vs. control networks was seen in a recent analysis of functional MRI resting-state brain activity [129]. LC phasic bursts were associated with more within- and across-network correlated activity in and between ventral attention network, executive and default mode networks but less correlated activity in and between visual and somatomotor networks [129]. This is consistent with phasic LC activation promoting focused hot spots that are not correlated with surrounding activity in sensory regions but simultaneously promoting coordinated frontoparietal activity that helps to coordinate competition based on relative priority.

In summary, the current level of glutamatergic activity in a particular brain region has the potential to modulate local noradrenaline release when the LC is activated. Thus, during moments of high arousal, LC activity can elicit greater noradrenaline release in cortical and hippocampal regions where there are already high levels of excitatory activity, leading to hot spots of high noradrenaline. These hotspots in turn can produce β-adrenergic activation in those specific regions, leading to qualitatively different effects than activation of α2A receptors by low levels of noradrenaline.

If, as the GANE model outlines, glutamate and β-adrenergic receptor interactions can create hot spots of high synaptic activity during arousal states involving LC activation, how does this interact with the production of Alzheimer’s-related pathology? As outlined in the following sections, increased synaptic activity has been associated with increased production of both amyloid peptides and tau proteins. Does this mean that LC activity accelerates the production of amyloid and tau and increases the risk of accumulation of amyloid plaques and tau tangles in ‘hot spot’ regions where there are high levels of glutamate? Also, does the LC pathology observed in aging and/or Alzheimer’s disease change the impact of LC activity? Subsequent sections address these questions.

2. Relationship between neuronal activity and amyloid-β and tau

2.1. Synaptic activity affects production of amyloid-β

One way that noradrenaline can affect Alzheimer’s-related pathology is simply by amplifying or suppressing synaptic activity. Like the body’s other metabolic activities, brain activity produces waste that can become problematic if not disposed of (Sections 3.2 and 3.6 review how noradrenaline influences waste clearance). For instance, synaptic activity stimulates the production of amyloid-β (Aβ) peptides, which are cleavage products of amyloid precursor protein (APP). Aβ plaques are a post-mortem hallmark of Alzheimer’s disease, and although these Aβ plaques are unlikely to be the cause of the disease [130], smaller aggregate forms (oligomers) of Aβ have toxic properties, disrupting synaptic activity and memory processes [131-133]. Why does synaptic activity produce Aβ? During synaptic activity, the process of releasing synaptic vesicles containing neurotransmitters involves fusing the vesicle with the cell membrane and creating an opening in both the vesicle and the cell membrane to release the contents of the vesicle. APP covers parts of the cell membrane [134] and can end up being moved inside the cell as part of the process of vesicle recycling. Once inside the cell, APP is processed through the cell’s endosome system, which sorts through cell membrane molecules to either recycle them or dispose of them. While in the endosome, APP is sequentially cleaved by β-secretase and then γ-secretase to produce Aβ [135], which is then released into the fluid-filled interstitial spaces between brain cells [136]. Because synaptic activity increases vesicle recycling, it increases the likelihood that APP will be cleaved to form Aβ [136-138].

There is also an alternative metabolic pathway for breaking down APP that does not produce Aβ. In this alternative pathway, before it can be cleaved by β-secretase and then γ-secretase inside the neuron, APP is cleaved by α-secretase and then γ-secretase while it is still on the surface of the neuron [139]. The factors determining whether APP goes through the Aβ-promoting β-secretase or the non-Aβ-forming α-secretase pathway are not yet fully understood, but in addition to synaptic activity increasing the likelihood of the Aβ-promoting β-secretase pathway, activation of muscarinic acetylcholine receptors stimulates α-secretase and release of non-Aβ-forming APP fragments [140, 141], thus tipping the scales against the formation of amyloid plaques.

Aβ itself is not necessarily problematic. It may even have some physiologically beneficial effects. One suggestion is that it provides feedback to prevent overexcitation of neural activity where there is high synaptic activity, as elevated extracellular Aβ (at high nanomolar and low micromolar levels) locally inhibits further synaptic activity [142, 143]. However, in healthy brains, extracellular concentrations of Aβ are in the low picomolar range, far lower than the range used in studies showing Aβ-induced synaptic depression [144]. Studies investigating the effect of either endogenous Aβ or extracellular application of picomolar concentrations of Aβ have found that it enhances hippocampal long-term potentiation and memory performance [145-148]. Thus, in the healthy brain there may be some benefits to Aβ.

However, if Aβ is not cleared and instead aggregates, it can damage neurons. The cleavage of APP by β-secretase and then γ-secretase produces Aβ monomers. These singlet forms of Aβ are attracted to each other and can eventually assemble into oligomers, and strings of oligomers known as protofibrils [149]. Over time Aβ protofibrils and fibrils can aggregate into the much larger amyloid plaques that are a hallmark of Alzheimer’s disease. However, it is the smaller Aβ oligomers and protofibrils that are most toxic [149-151].

Consistent with the role of synaptic activity in the formation of Aβ, there is evidence that more active brain regions accumulate more Aβ. In positron emission tomography (PET) imaging studies, the most metabolically active regions of the brain (in particular, those comprising the brain’s ‘default mode network’) show the most uptake of Aβ markers [152, 153]. In transgenic mice expressing a mutated form of APP to mimic aspects of human Alzheimer’s disease, 30 minutes of unilateral whisker stimulation increased interstitial Aβ in the contralateral barrel cortex [154], demonstrating that an episode of high neural activity in a particular brain region can increase Aβ levels in that region. Furthermore, depriving mice of whisker sensation on one side for about a month reduced amyloid plaque growth by 78% in the contralateral hemisphere compared to the control hemisphere [154].

2.2. Neuronal activity also increases extracellular tau

Aβ is not the only potentially toxic waste product of neuronal activity. Another one is the microtubule-associated protein tau which, among its multiple roles, provides a protein scaffold on microtubules [155]. When hyperphosphorylated, the tau protein no longer binds as effectively to microtubules and more readily misfolds, creating insoluble fibrils that are the main constituent of the neurofibrillary tangles that, together with Aβ plaques, are the key pathological features of Alzheimer’s disease [10, 156].

In wild-type mice, tau is released into the extracellular space in the brain under normal conditions [157]. Increasing neuronal activity increases release of tau into the interstitial space in both in vitro and in vivo rodent models [for review see 158], and if the excited neuron contains pathological tau, its release can lead to its spread to other neurons [159]. In a transgenic mouse Alzheimer’s disease model, chemogenetic attenuation of neuronal hyperactivity reduces the spread of pathological tau [160]. Presynaptic glutamate release is sufficient to increase tau release [161], and the release apparently can be triggered by either AMPA [162] or NMDA receptors [163].

Growing evidence suggests that, as with Aβ, vesicle-trafficking stimulates the release of tau. Exosomes are small extracellular vesicles that derive from the fusion of intracellular vesicles and organelles with the cell membrane [164]. Secretion of phosphorylated tau often occurs via exosomal release [165]. Depolarization of neurons promotes release of tau-containing exosomes [166], providing additional evidence of the importance of neuronal activity in stimulating release of the building blocks of Alzheimer’s pathology. Tau secretion has been linked not just with exosomes released from neurons but also with exosomes released from microglia [167]. Microglia likely accumulate tau as part of the process of phagocytosing dying cells, debris and protein aggregates [168].

Thus, both Aβ and tau release can be stimulated by neuronal activity. Therefore, insofar as LC activity increases the gain on neural excitation/inhibition levels [101], it should also influence Aβ and tau release simply by modulating levels of synaptic activity.

2.3. Relationship between environmental enrichment and Alzheimer’s pathology

A question motivating this review is how stimulating noradrenergic activity via arousing, engaging cognitive activities will affect accumulation of Alzheimer’s pathology. The findings reviewed in Sections 2.1 and 2.2 indicate that neuronal activity stimulates release of Aβ and tau into the extracellular space. However, connecting these findings at the level of synaptic activity with broad-scale interventions such as environmental enrichment in rodents or increased years of education in humans is not straightforward. Stimulating environments will not affect activity in all brain regions in the same way. Enriched environments may also increase brain activity in a particular brain region during an initial novel phase but decrease activity in that region in the long run. For instance, as part of learning, brain regions may execute tasks more efficiently [169], and so one effect of environmental stimulation could conceivably be decreased metabolism in the long run in those brain regions initially activated by the environmental challenges.

For these reasons, perhaps it is not surprising that non-specific environmental enrichment in APP transgenic mice can either increase amyloid plaque formation [170, 171] or decrease it [172, 173]. One interesting possibility is that these contradictory findings may relate to whether the mice were placed in environmentally enriched housing immediately after weaning [172, 173] or from when they were two months old [170, 171]. Early in the life, the amyloid produced as a by-product of synaptic activity is regularly cleared out along with other waste and potentially toxic substances. In contrast, through middle-age and in late life, clearance of amyloid declines significantly in mice and rats [174, 175]. Thus, one possibility is that the neuronal activity stimulated by enriched environments leads to more amyloid accumulation later in life than the same stimulation would early in life.

In humans, random assignment to dramatically differing environments for long periods of life is not a feasible research strategy. However, we can examine whether there are correlations between factors that reflect time spent in an enriched environment (such as education levels, job complexity, or being bilingual) and development of Alzheimer’s disease. Across many studies, people with higher education or other indicators of enriched environments are less likely to have been diagnosed with Alzheimer’s disease [176, 177]. But despite this relationship between prior education and likelihood of a diagnosis, those with more education do not tend to show an advantage in terms of brain pathology. In other words, while diagnoses of Alzheimer’s disease based on cognitive performance are lower in those with more education, diagnoses based on post-mortem brain pathology show no indication that education protects against Alzheimer’s pathology. For instance, in a post-mortem sample of 2372 participants whose brain pathology met criteria for an Alzheimer’s diagnosis, there was no correlation between education and level of Alzheimer’s neuropathology [e.g., 178]. Nevertheless, in this same sample, whether they had received a clinical diagnosis before death was related to education, with those with a ante-mortem diagnosis having 2-3 years less education on average than those with no ante-mortem diagnosis. Likewise, examination of data from 835 autopsies from the United States National Alzheimer Coordinating Center found that although higher levels of education decreased the odds of cognitive impairment, this effect was not mediated by extent of pathological changes (including amyloid plaques) [179].

In a large meta-analysis employing participant level data from 55 studies (N = 8694) with individuals with normal cognition, subjective cognitive impairment or mild cognitive impairment, those with above-median education had a higher prevalence of being amyloid positive on the basis of either cerebrospinal fluid (CSF) or PET imaging than those with less education, regardless of cognitive status, age, sex or APOE-ε4 carrier status [180]. However, as this meta-analysis was not a population-based study, it is possible that highly educated people with amyloid positivity were more likely to seek out clinical care or research studies than less educated people with amyloid positivity. Indeed, a large population-based study in Minnesota (N = 942) revealed that neither years of education, degree of intellectual challenge in jobs held, nor midlife weekly cognitive activity (such as reading or crafts) were associated with PET imaging categorization as amyloid positive or negative [181]. In contrast with both of the above larger studies, one study has found that those with higher education showed less PET amyloid binding than those with lower education [182]. However, this study had a small sample (N = 30) in which education level was confounded with participant sex. Thus, consistent with post-mortem data, the best evidence from PET imaging studies indicates no relationship between education and amyloid load. Education also fails to slow the progression of cortical atrophy [183]

This disjunct between the effects of education on Alzheimer’s diagnoses versus brain pathology led to arguments that education provides some sort of reserve that allows people to keep functioning well despite the encroaching brain pathology [26, 177]. If education provides reserve capacity, it should slow cognitive decline in those who can call on that reserve when challenged by encroaching pathology. However, whether higher education attainment slows cognitive decline has been a challenging question to address due to the nonlinear nature of cognitive decline and methodological confounds that might lead to comparing those with higher and lower education at different points in their downward cognitive trajectory [184].

A recent systematic review of the literature that focused especially on studies that had longitudinal data and used appropriate statistical approaches concluded that the association between educational attainment and late-life cognitive decline is small and inconsistent, and most of the effects of education are due to those with higher education achieving a higher peak level of functioning [185]. Alzheimer’s pathology seems to cause cognitive decline at the same general rate regardless of one’s peak cognitive attainment [184]. Thus, if one starts at a higher point, it will take more pathology to bring that person down to the threshold of attaining a diagnosis based on their cognitive functioning.

In summary, the difference in environmental enrichment between completing high school versus college or graduate school fails to provide any protection from cognitive decline or Alzheimer’s disease pathology. But the rodent studies described above suggest that timing of environmental enrichment may matter, and that providing environmental enrichment early in development has the most positive effect on later resistance to Alzheimer’s pathology. Thus, perhaps environmental enrichment that occurred during childhood would have a more noticeable effect on later pathological processes.

One study estimated early-life cognitive enrichment in 813 participants with post-mortem neuropathology results. Participants were recruited from the community when they were older, and, at baseline, answered questions about their early life (about socioeconomic status, availability of cognitive resources, frequency of cognitively stimulating activities and foreign language instruction) from which an early-life cognitive enrichment score was derived [186].

This score was negatively associated with a score assessing global Alzheimer’s pathology, as well as with levels of tau and Aβ measured separately. These results provide the strongest link to date between cognitive enrichment and Alzheimer’s pathology. Comparing these findings to the lack of association between cognitive enrichment in adolescence and adulthood and Alzheimer’s pathology seen in other studies raise the possibility that childhood cognitive enrichment has more of an impact on later Alzheimer’s pathology than factors operating in late adolescence or early adulthood, such as additional years of total education attainment.

However, this study unfortunately did not compare the relationship of early and mid- or late-life cognitive enrichment. Further work is needed to directly compare the effects of environmental enrichment initiated very early in development or at a later stage on amyloid plaque formation to see if indeed the risks of stimulating amyloid plaque formation are greater when the novel mental stimulating environment is first experienced later in life.

3. Relationships between adrenergic activity and the progression of Alzheimer’s disease

3.1. β-adrenergic activation influences microglial activity

The previous sections reviewed how neuronal activity and environmental enrichment affects Alzheimer’s pathology. In the next few sections, I address how noradrenaline modulates these processes.

One important way that noradrenaline modulates the progression of Alzheimer’s disease is via its influence on microglia. Microglia are the main immune cells in the brain that dynamically survey their environment, ready to extend their processes towards disease or injury sites to clear damage [187, 188]. Microglia can either protect against Alzheimer’s disease or exacerbate the progression of the disease [189]. Postmortem analyses and Alzheimer’s mouse models indicate that in Alzheimer’s disease, microglial inflammatory activity increases while microglial clearance activity decreases [190].

Microglia express more β2-adrenergic receptors than most other cortical cell types [191, 192]. Recent advances now allow tracking of microglia dynamics in awake mice using two-photon microscopy. These in vivo methods have revealed that, during wakefulness, microglia are less active with diminished responsiveness to injury compared to during anesthesia [193, 194]. The LC is much more active during wakefulness than during sleep [195] and β2-adrenergic receptor activity decreases microglial activity during wakefulness [193, 194]. A β2-adrenergic agonist also suppresses proliferation of microglia in culture [196]. However, stress (a potent noradrenergic stimulus) activates microglia and a β-adrenergic blocker inhibits this stress-induced microglial activation [192]. Thus, effects of noradrenaline on microglial activity appear to depend on stress levels.

As would be predicted by the GANE model (see Section 1.4), levels of local neuronal activity modulate microglial β2-adrenergic receptor activity [193]. For instance, trimming a mouse’s whiskers on one side (a sensory-deprivation method that suppresses neuronal activity in the contralateral barrel cortex), led microglia in the contralateral barrel cortex to increase their process area, an effect similar to that seen under anesthesia [193]. Interestingly, whereas resting microglia primarily express β2-adrenergic receptors, they switch expression to α2A receptors under proinflammatory conditions [197], which should lower the threshold for noradrenergic influence (due to the different affinities of β and α2 receptors; see Section 1.2) and may also change its effect.

3.2. β-adrenergic activation influences Aβ accumulation and clearance

As already discussed, cortical regions with high levels of neural activity tend to produce more Aβ. An intriguing possibility is that β-adrenergic activation during high levels of neuronal activity may provide a countervailing influence that reduces the toxic influences of Aβ. Even though, as reviewed in the previous section, local β-adrenergic activation decreases microglial process surveillance, it may stimulate microglial cells to dispose of Aβ₄₂. When isolated in a cell culture, mouse microglia cells show increased Aβ₄₂ uptake when exposed to either noradrenaline or a β-adrenergic agonist, apparently driven by upregulation of the microglia receptor for Aβ₄₂ [198]. One study consistent with the possibility that β-adrenergic activation helps combat the negative effects of Aβ examined the effects of environmental enrichment in wild-type mice exposed to human Aβ oligomers [79]. Mice living in an enriched environment for two months starting just two weeks after birth had higher expression of long-term potentiation (LTP) related signaling proteins in the hippocampus than did control mice. Furthermore, exposing hippocampal slices from these mice to soluble Aβ oligomers revealed the time in the enriched environment made hippocampal LTP less vulnerable to Aβ and that this protection could be up- and down-regulated by β-adrenergic agonists and antagonists, respectively. In addition, feeding normally housed mice a β-adrenergic agonist protected LTP from the toxic effects of Aβ. Exposing mice at midlife to an enriched environment also enhanced their LTP and protected it from Aβ, although it took 8 weeks to reach similar benefits as the 4-week exposure used in younger mice. Consistent with the protective effects of β-adrenergic receptors, other studies demonstrate that injection of a β-adrenergic receptor agonist before LTP induction prevents Aβ-induced LTP attenuation [199, 200]. Rats injected with human Aβ₄₂ a week after toxin-induced degeneration of the LC showed greater inflammatory responses to Aβ₄₂, effects that were attenuated by co-injection with noradrenaline or a specific β-adrenergic receptor agonist [201]. Furthermore, in studies with APP-transgenic mice, toxin-induced degeneration of the LC increased Aβ₄₂ levels [202-204] and reduced the colocalization of Aβ deposits and a microglial activation marker by 70% [205]. Providing the noradrenaline-depleted mice with the noradrenaline precursor L-threo-DOPS restored microglial phagocytosis of Aβ [205]. These findings suggest that without noradrenaline, the microglia failed to take up Aβ as part of the phagocytosis process. β-adrenergic activity has been identified as the mechanism underlying the benefits of environmental enrichment on microglia exposed to Aβ [206]. In wild-type mice housed in standard cages, exposure to soluble oligomers of Aβ (isolated directly from human brains with Alzheimer’s disease) increases local microglial inflammation, an effect that is prevented by housing mice in enriched environments [207]. The positive effects of environmental enrichment can be mimicked by giving the mice in the standard cages a β-adrenergic receptor agonist or blocked by giving the environmentally enriched mice a β-adrenergic antagonist [206]. In mice genetically altered to lack β-adrenergic receptors, environmental enrichment failed to prevent microglia inflammatory responses to Aβ [206]. Also demonstrating the beneficial effects of β-adrenoceptor activity, giving amyloid precursor protein/presenilin 1 (APP/PS1) double transgenic mice (who exhibit Aβ accumulation, synaptic deficits and memory impairment at 8 months) a β-adrenergic agonist for two months starting when they were 8 months old led to decreased Aβ plaques, greater hippocampal neurogenesis, less pronounced memory deficits and increased dendritic branching and density of dendritic spines than giving them a vehicle control [208].

In contrast with the positive effects reviewed above of β-adrenergic agonists in wild-type and transgenic mice [198, 206, 208], other studies with transgenic mice show negative effects. For instance, a study using APP/PS1 double-transgenic mice found that 30 days of β-adrenergic agonists increased amyloid plaque formation [209]. Another study using Tg2576 mice which at 6 months start showing exponential increases in Aβ peptide content in the brain found that giving these mice a β-adrenergic antagonist reduced Aβ₄₂ accumulation in the hippocampus [210]. It is not clear how to account for the different results across studies.

Data in humans also indicates that β-adrenergic activity influences the course of Alzheimer’s disease. In a study of Han Chinese, there were significant differences between a group of 109 Alzheimer’s disease patients vs. 109 controls in the Glu27Gln (rs1042714) and the Arg16Gly (rs1042713) polymorphisms of the β2-adrenergic receptor gene [211]. Both Gly16 and Glu27 polymorphisms were more likely to occur in patients with Alzheimer’s disease. It is not clear yet what mechanism leads to this association, but these polymorphisms are associated with greater vascular responsiveness to a β2-adrenergic agonist [212] and with higher low-density lipoprotein cholesterol [213], either of which might be a relevant factor in addition to potential brain effects. An observational longitudinal study of older adults with dementia found that functional decline was slower in those taking β-adrenergic antagonists for cardiovascular conditions than the other participants not taking this type of medication [214]. Likewise, a study comparing medication histories of patients newly diagnosed with dementia versus age- and sex-matched controls found that the dementia group was significantly less likely to have been continuously taking a β-adrenergic antagonist in the previous three years [215]. These findings suggest that, at least once Alzheimer’s disease is relatively advanced (as amyloid pathology is already typically evident at least a couple of years before typical diagnosis [216]), β-adrenergic receptor activity may accelerate disease progression. Currently, the human literature lacks analyses of the effects of β-adrenergic medications taken earlier in the life on later timing and progression of Alzheimer’s disease. The fact that β-adrenergic activity can have either beneficial or harmful effects in rodents suggests the effects depend on the presence of some other factor such as the level of existing pathology.

In summary, β-adrenergic receptor activity has significant effects on the accumulation of Aβ pathology, especially via its modulatory effects on microglial activity. However, whether these effects are positive or negative depends on factors that are not yet clear.

3.3. α2A-adrenergic activation influences Aβ accumulation

There is also evidence that α2A-adrenergic receptors modulate Aβ accumulation. APP/PS1 transgenic mice genetically modified to lack α2A-adrenergic receptors show reduced Aβ generation and Alzheimer’s-related pathology whereas activating the α2A-adrenergic receptor with noradrenaline (while blocking β-adrenergic receptors) enhances Aβ generation and Alzheimer’s-related pathology [217]. As reviewed in Section 1.2, α2A-adrenergic receptors are found in the presynaptic membrane of noradrenergic terminals as well as on other types of neurons. They also are widely expressed in astroglia. Thus, the findings that α2A-adrenergic receptor activity is associated with Aβ generation and Alzheimer’s-related pathology [217] could be due to a variety of possible mechanisms.

One possibility is that α2A-adrenergic receptors have an amyloid plaque promoting effect because their activation reduces APP localization to the Golgi where it could be recycled and increases APP sorting to endosomal compartments. There it is colocalized and interacts with β-secretase [217], biasing processing towards the Aβ-promoting metabolic pathway for breaking down APP. In addition to α2A-adrenergic receptors influencing APP, APP also influences α2A-adrenergic receptors, enhancing the surface retention of α2A-adrenergic receptors and their signaling intensity and duration [218]. As reviewed in a later section (4.2), Aβ oligomers also potentiate the effects of α2A-adrenergic receptors [219]. Together, these findings suggest positive feedback between α2A-adrenergic receptor activity and APP and Aβ oligomers that may accelerate Alzheimer’s pathology in brain regions where there already is some pathology — a possibility consistent with findings that taking an α2-adrenergic agonist accelerates cognitive decline in those who already are diagnosed with Alzheimer’s disease but not in healthy older adults (see Section 4.2) [219].

3.4. The locus coeruleus and pathological tau

In addition to its role in influencing the progression of Alzheimer’s disease pathology elsewhere in the brain, the LC is significantly affected itself in the course of the disease. In fact, in a large study with over 2000 post-mortem brains from age 1 to 100, it was the first site in which the pretangle stage of tau pathology was detected [9]. This pretangle stage was detected using an antibody (AT8) to tau phosphorylated on amino acids Ser202 and Thr205 [for more details see 10]. Another study using a different Ser202 antibody (CP-13) estimated that around 8% of LC neurons showed a positive antibody response in those at Braak stage 0 of Alzheimer’s disease (when there is no cortical indication of the disease, in this particular study N=13 cases with a mean age of 58) [220].

In addition to this early vulnerability of the LC to abnormal tau, linkages between adrenergic activity and tau hyperphosphorylation have been detected. Dexmedetomidine is a selective α2-adrenergic agonist that is commonly used as an anesthetic to induce a sedative state [221] because it reduces LC noradrenaline release via α2-adrenergic autoreceptor activation [222]. Dexmedetomidine induces tau hyperphosphorylation in vivo and in vitro in both transgenic human Tau mice and in non-transgenic mice, with an α2A receptor antagonist blocking this dexmedetomidine-induced tau hyperphosphorylation [223]. The hyperphosphorylation levels were significantly elevated half an hour after dexmedetomidine administration in cortex and hippocampus, with elevated rates still detectable in the hippocampus but not the cortex 6 hours afterwards.

Links with tau hyperphosphorylation have also been demonstrated with β2-adrenergicreceptors. Removing the gene encoding β2-adrenergic receptors from a mouse model overexpressing mutant human tau led to mice that lived longer and showed reduced hyperphosphorylated tau in the basal ganglia and hippocampus (but not in the cortex) than the tau model mice with β2-adrenergic receptors [224].

Interestingly, in mice, toxin-induced degeneration of the LC led to significant impairments in spatial learning and memory in P301S tau transgenic mice (who show progressive accumulation of neurofibrillary tangles) but not in wild-type mice [225], suggesting that loss of LC neurons is especially problematic within the context of tau pathology.

3.5. Interactions of Aβ and tau can be influenced by noradrenergic activity

Both human and animal research have documented that Aβ and hyperphosphorylated tau aggregates interact to accelerate Alzheimer’s disease degeneration [226, 227]. Some aspects of these interactions are mediated by noradrenergic activity. For instance, Aβ oligomers bind to an allosteric site on α2A adrenergic receptors, thereby redirecting noradrenergic signaling to glycogen synthase kinase 3β (GSK3β) activation and tau hyperphosphorylation [219]. GSK3 contributes to the abnormal phosphorylation of tau proteins in Alzheimer’s disease, as well as being involved in many other disorders [228].

3.6. Noradrenergic activity influences glymphatic clearance

The accumulation of Alzheimer’s disease pathology depends not only on the rates of production of Aβ and tau but also on how effectively these by-products of metabolic activity are cleared out of the brain. There are a number of ways that waste products are eliminated from the brain [229-231]; one of these is via the glymphatic pathway [232, 233]. Glymphatic clearance of waste occurs when CSF traveling in compartments wrapped around brain arteries exchanges with the interstitial fluid between brain cells. Eventually, the CSF exits the brain in compartments wrapped around veins, taking substances such as Aβ [234] and tau [235] that were in the interstitial space with it.

Glymphatic function is active in both naturally sleeping and anesthetized mice but suppressed during wakeful states [236, 237]. As already discussed, noradrenaline is a key regulator of arousal states, including wakefulness [67]. Indeed, it turns out that elevated noradrenaline levels help drive suppression of glymphatic function during wakefulness. Administering a cocktail of adrenergic antagonists, including β, α1, and α2-adrenergic antagonists, significantly increased the influx of cerebral spinal fluid into the cortex and the volume of the fluid-filled interstitial space between brain cells [237]. Dexmedetomidine, the commonly used anesthetic that reduces LC noradrenaline release via α2-adrenergic autoreceptor activation [222] increases glymphatic CSF influx [236, 238, 239]. One way that noradrenaline modulates glymphatic function is by lowering production of CSF at the choroid plexus [240-242]. Another potential pathway is via noradrenaline’s modulation of vasoconstriction and vasodilation [243-245]. In awake mice, slow oscillations in vasoconstriction appear to facilitate clearance of particles from interstitial fluid [246].

Thus, in summary, noradrenaline impedes glymphatic clearance both by decreasing production of CSF and by decreasing the interstitial space volume, restricting movement of interstitial fluid. This is an important effect that could accelerate the accumulation and eventual aggregation of Aβ and tau, in particular in those individuals with high levels of noradrenergic activity. As reviewed below, both age and Alzheimer’s disease have been associated with increased CSF and blood levels of noradrenaline. It is possible that disease-related increases in noradrenaline in turn lead to decreased glymphatic clearance of Aβ and tau [247].

4. Influences of Alzheimer’s disease and aging on locus coeruleus function

4.1. Locus coeruleus vulnerability in Alzheimer’s disease and other neurodegenerative diseases

Thus far, this review has focused on how noradrenergic activity affects the production and accumulation of Aβ and tau. However, we should not ignore that these effects of noradrenergic activity occur in a context in which the LC is changing with age. In particular, the LC is a key node in the age-related progression of Alzheimer’s disease pathology. As already mentioned, the LC appears to be the earliest site of pathological tau [9, 10], with post-mortem abnormally phosphorylated tau seen in the LC in most non-selected brains by age 30 [9]. In cognitively normal older adults, decreased structural volume in the LC predicts conversion to an Alzheimer’s disease diagnosis [21].

Previously, we had speculated that the high energetic demands of LC neurons contribute to their vulnerability [2]. Recent findings indicate that metabolism in the LC carries a particularly high risk due to a specific intermediate metabolite of noradrenaline [248]. This metabolite, 3,4-dihydroxyphenylglycolaldehyde (DOPEGAL), is produced by monoamine oxidase A (MAO-A) oxidation of noradrenaline and triggers tau pathology in the LC that can spread to other brain regions [248]. DOPEGAL activates a lysosomal endopeptidase (asparagine endopeptidase) that cleaves tau at specific locations (N255 and N368) and the resulting fragments increase tau hyperphosphorylation and aggregation [249]. Postmortem LC samples from subjects with Alzheimer’s disease showed more asparagine endopeptidase along with aggregated tau and more tau cleaved at N368 than LC samples from control brains [248].

In addition to suffering from the DOPEGAL metabolic risk, the LC may also be vulnerable for a number of other reasons. Its extensive contact with central nervous system capillaries and its proximity to the fourth ventricle increases its exposure to circulating toxicants [2]. Also, viruses or other noxious substances may be transported to the LC from the gut via the vagus nerve and nucleus of the solitary tract [250]. Herpes infections may cause a particular risk, because after people recover from a herpes infection, the common herpes simplex type 1 virus lies dormant in the trigeminal ganglion, which projects to the LC [251]. Another possibility is that the narrow nature of the cerebral aqueduct connecting the third and fourth ventricles induces CSF flow (especially in those with hardened arteries that fail to absorb cardiac pulsations and so pass along greater pulsatility to upstream vessels) that alters the molecular structure of tau proteins being carried by the CSF in ways that eventually promote their aggregation [252].

Further evidence indicating the LC is especially vulnerable to pathological tau is that degeneration of the LC is a common feature of tauopathies such as progressive supranuclear palsy [253, 254]. Furthermore, the LC is a specific target of tauopathies rather than of all proteinopathies [e.g., 255].

Degeneration of the LC is also seen early in Parkinson’s disease [256]. Like other α-synucleinopathies such as dementia with Lewy bodies, Parkinson’s disease is characterized by the abnormal accumulation of aggregates of α-synuclein protein in the brain, α-synuclein promotes tau hyperphosphorylation and can trigger tauopathy [257]. Abnormal tau and α-synuclein appear to coaggregate in the locus coeruleus and elsewhere in the brain [258], which may help explain why the LC is also targeted in Parkinson’s disease and other α-synucleinopathies [259].

How do the increases in abnormal tau in the LC seen in aging and Alzheimer’s disease relate to LC function? This is still a mystery as, to date, we have only hints as to how increases in hyperphosphorylated tau affect the functioning of LC neurons [e.g., 260]. Although not documented for the LC itself, there are indications that tau hyperphosphorylation decreases excitation within the neuronal networks involved [261]. However, we have yet to have definitive information regarding even the simple question of whether LC tonic activity increases or decreases in aging. The following section reviews some of the existing relevant evidence.

4.2. Changes in noradrenergic activity in aging

Does LC activity increase or decrease with age? This is a challenging question to answer given the currently available data [for reviews see 262, 263, 264]. Postmortem measures of noradrenaline extracted from brain tissue tend to be lower in older adults than younger adults and lower in those with Alzheimer’s disease compared with cognitive normal older adults [265-268]. However, noradrenaline levels reflect production, release and re-uptake and so are not a direct measure of LC activity. Furthermore, these measures use homogenized brain tissue and so estimates of noradrenaline depend both on the number of noradrenergic neurons and their corresponding internal store of neurotransmitter as well as on how much of the neurotransmitter was circulating in the interstitial spaces between cells.

In contrast with the results from homogenized brain tissue, plasma noradrenaline concentrations are estimated to increase 10-15% per decade during adulthood [269]. Under resting conditions, noradrenaline in the bloodstream mostly comes from noradrenaline released by sympathetic nerves, with only a small proportion coming from the adrenal gland [270]. Consistent with the plasma findings, levels of noradrenaline in CSF are greater in cognitively normal older adults than in younger adults [271, 272]. Advanced Alzheimer’s disease is associated with even greater levels of noradrenaline in CSF and plasma than in cognitively normal older adults [271, 273].

Thus, despite the substantial decline in noradrenaline neurons in the LC in Alzheimer’s disease [e.g., 5], noradrenaline levels in CSF and plasma still remain high, and in many cases even higher than in normal aging or in younger adults. This opposing effect of Alzheimer’s disease on noradrenaline neurons versus on circulating levels of noradrenaline suggests that remaining noradrenaline neurons may compensate by increasing their activity levels (this could include non-LC sources of noradrenaline such as A5 and A7 cell groups and sympathetic ganglion neurons).

Research with rodents supports the possibility that intact noradrenergic neurons compensate for damaged ones by increasing their activity rates. For instance, using intraventricular administration of 6-hydroxydopamine to destroy 80-90% of noradrenergic terminals in the rat forebrain led the still-intact LC cell bodies to increase their firing rates by about four times normal levels [274]. Likewise, injection of 6-hydroxydopamine directly into mouse LC at different doses to destroy between 30-70% of the neurons led surviving LC neurons to fire more frequently and with a more irregular pattern three weeks later [275]. Compensation may also occur via increased sprouting of axonal projections. Normal-to-elevated presynaptic α2A binding site density in hippocampus and prefrontal cortex in Alzheimer’s disease patients compared with controls suggest greater numbers of axonal projections from each LC neuron into these regions in patients [276, 277].

But the human data indicating high circulating levels of noradrenaline could also result from impaired clearance processes, such as declines in noradrenaline transporter availability [278]. To address this, additional insight can be gained by examining metabolites that precede versus follow noradrenaline in its metabolic process. Studies assessing the noradrenaline metabolite methoxyhydroxyphenylglycol (MHPG) levels indicate sustained noradrenergic metabolic activity in Alzheimer’s disease patients despite a declining noradrenergic system: In studies quantifying both post-mortem brain noradrenaline and MHPG, while noradrenaline tended to be lower in patients with Alzheimer’s disease compared with controls, MHPG tended to be at least as high in the patients as in the controls [266, 279]. Other studies also show similar CSF MHPG levels between patients with Alzheimer’s disease and age-matched controls, with lower levels in younger adults [280, 281]. Of particular interest, greater CSF MHPG levels in memory clinic patients predicted a steeper decline in their memory in the next few years [282]. Alzheimer’s disease pathology in the context of low MHPG levels was not associated with significant memory decline, raising the possibility that noradrenergic hypermetabolism plays an essential role in memory declines in Alzheimer’s disease.

Another indication that, in Alzheimer’s disease, the surviving LC neurons produce more noradrenaline is that patients with Alzheimer’s disease had more tyrosine hydroxylase mRNA expression in those LC neuron cell bodies that survived the profound neuronal loss associated with the disease than in the LC cell bodies of controls [276]. The greater the neuronal loss, the more each remaining cell expressed tyrosine hydroxylase mRNA. Tyrosine hydroxylase is the rate limiting factor in catecholamine biosynthesis. The decrease seen in noradrenaline transporter density in the brain with aging [283] and with Alzheimer’s disease [284, 285] may also lead to increases in extracellular levels of noradrenaline in the brain which would be reflected in CSF levels.

A study that administered yohimbine, an α2A antagonist that increases release of noradrenaline, clonidine, an α2A agonist that decreases release of noradrenaline, or a placebo to participants revealed that CSF and blood levels of noradrenaline were modulated similarly by the adrenergic drugs in older and Alzheimer’s disease patients as in younger adults [286]. The three groups of participants also showed similar increases in dihydroxyphenylglycol (DHPG), a metabolite of noradrenaline that is an indicator of its clearance, in response to yohimbine. However, in the yohimbine condition, the older and Alzheimer’s group showed greater increases than did younger adults in CSF levels of the noradrenaline precursor dihydroxyphenylacetic acid (DOPA), an indicator of noradrenaline synthesis. These findings suggest that aging and Alzheimer’s disease increase synthesis of noradrenaline.

Alzheimer’s patients have Aβ oligomers in the LC that, based on a mouse model of increased Aβ production (APP-PSEN1), appear to contribute to its hyperexcitability [287]. Like Alzheimer’s patients (but unlike wild-type mice), APP-PSEN1 mice have LC neurons that express Aβ oligomers. Electrophysiological recordings in 2-to 4-month-old mice brain slices revealed that APP-PSEN1 LC neurons had greater firing rates than wild-type LC neurons [287]. Both GABA and glycine provide inhibitory modulation of LC [288]. APP-PSEN1 LC neurons decreased their firing significantly less in response to GABA than did wild-type LC neurons, whereas there was no significant difference in response to glycine [287]. There also were no detectable differences in levels of expression of glutamatergic synaptic markers across the two mice groups. These findings suggest that Aβ oligomers in the LC impair the activity of GABA receptors that reduce activity rates.

APP-PSEN1 mice lack tau pathology, the other main feature of Alzheimer’s disease. But recent work has created a rat model of the spread of pathological tau from the LC to cortical regions by expressing pseudohyperphosphorylated human tau in rat LC [260]. At 6 weeks post-infusion of tau/control in the LC, the rats with pathological human tau showed spread of that tau along LC axons toward forebrain targets, and by 12 weeks, it had reached the olfactory bulb and other cortical areas such as the hippocampus and piriform cortex. By 7-8 months post-infusion, the rats with pathological tau showed impaired olfactory discrimination and degeneration of LC axons in the piriform cortex. At the same time, they showed upregulation of β1-adrenoceptors in the piriform cortex. Previous studies in rats have also shown compensatory increase of cortical β-adrenoceptors after damage to noradrenergic neurons [289, 290]. Interestingly, in humans, depressed suicide victims show fewer rostral LC neurons [291] and greater prefrontal β-adrenergic receptor binding [292-294], consistent with the notion that β-adrenoceptors upregulate in response to LC degeneration. Data from Alzheimer’s patients are limited, but one post-mortem study showed increased β-adrenoreceptors in the hippocampus with a mixed profile in other brain regions (e.g., in frontal cortex β2-adrenoceptors but not β1-adrenoceptors were greater in Alzheimer’s patients than in controls) [295]. While increased cortical β-adrenoceptors may help compensate for decreased LC neurons, they may also lead to hyperexcitability, such as seen in older adults’ hippocampus [296, 297].

Along with these indications of increased noradrenergic activity and upregulation of excitatory β-adrenergic receptors, there also appears to be upregulation of α2A noradrenergic receptors in Alzheimer’s disease. Postmortem prefrontal brain samples from patients with Alzheimer’s disease show greater α2A-receptor-mediated G protein activation in response to noradrenaline even when adjusting for α2A receptor density compared with controls [219]. How could α2A receptors show a more potent response to noradrenaline in patients with Alzheimer’s disease?

G protein-coupled receptors not only have their primary binding sites for their target neurotransmitter or other ligand (known as orthosteric binding sites), but also have other (allosteric) binding sites that can modulate the effects of ligand binding at the orthosteric sites [298]. Experiments with an Alzheimer’s disease mouse model suggest that Aβ oligomers’ ability to bind to an allosteric site on α2A receptors allows this Alzheimer’s related peptide to potentiate the effects of noradrenaline binding [219].

Thus, the emerging data so far suggest a pattern of increased noradrenergic activity in Alzheimer’s disease that is accompanied by increased effectiveness of typically inhibitory α2A receptors. It is not clear yet whether one or both of these phenomena accelerate or slow down the course of the disease. However, one relevant observation comes from 267 Alzheimer’s patients who happened to be taking clonidine at some but not all clinic visits [219]. Scores on the mini-mental state examination declined more during time periods with clonidine than without clonidine, an effect not seen in normal controls. Furthermore, this decrease in cognitive function when on clonidine was more pronounced for those diagnosed with Alzheimer’s disease than those with mild cognitive impairment or who were impaired without mild cognitive impairment. As clonidine decreases LC noradrenergic activity [299], these correlational findings suggest that, at least once the Alzheimer’s disease process is fairly advanced, reducing noradrenergic activity accelerates decline.

4.3. Alzheimer’s disease is associated with an increase in extracellular glutamate

Yet another factor that might produce higher extracellular noradrenaline levels in patients with Alzheimer’s disease is the increase in extracellular glutamate that is associated with the disease [97]. In the brain, intracellular glutamate concentrations are in the millimolar range, but transporters that import glutamate into astrocytes and neurons keep the extracellular concentrations in the low micromolar range [300]. In patients with Alzheimer’s disease, post-mortem examination reveals decreased expression and activity of transporters that remove glutamate from the extracellular space [301-304]. In some studies, patients with Alzheimer’s disease have higher levels of glutamate in CSF compared with either healthy controls [305-307] or those with mild cognitive impairment [308], although a reduced level of CSF glutamate in patients with Alzheimer’s compared to controls has also been observed [309].

One reason for decreased glutamate clearance in Alzheimer’s disease may be increased levels of Aβ, as soluble Aβ peptides and oligomers reduce glutamate uptake and decrease glutamatergic transporter activity [310-312].

As outlined earlier, the GANE model posits that locally high levels of extrasynaptic glutamate can stimulate greater noradrenaline release via NMDA receptors if the LC neuron is concurrently depolarized. Thus, effects of noradrenergic activity on neuronal activity and on Aβ and tau dynamics could be amplified by Alzheimer’s-related increases in glutamate concentrations in the extrasynaptic space. Furthermore, as outlined by GANE, higher levels of local noradrenaline release can in turn lead to higher glutamate release in those regions. These interactive effects in turn may accelerate the accumulation of Alzheimer’s pathology, as in vitro work indicates that glutamate activity upregulates the tau mRNA translation involved in the transport of non-axonal tau across synapses [313] and increases the missorting of tau into dendrites [314].

5.1. Conclusions

On the one hand, researchers have posited that the LC may be the main brain mechanism underlying cognitive reserve [3, 24, 25]. On the other hand, the argument has been made that drugs that lower noradrenergic signaling should protect against a broad range of diseases, including Alzheimer’s disease [315]. This paper reviews findings suggesting that the story is not as simple as either of these opposing perspectives, as noradrenergic activity can be either beneficial or problematic when it comes to Alzheimer’s disease.

Although existing findings leave many unanswered questions, one hypothesis suggested by the existing literature is that increases in tonic LC activity due to age-related LC deterioration or stress are detrimental whereas phasic LC activity associated with cognitive stimulation may be more beneficial. However, even the phasic noradrenergic activity induced by cognitive stimulation is likely to have a higher risk-to-benefit ratio for brain health as the brain’s ability to clear out waste products associated with neural activity diminishes in aging. This set of findings suggest that it may be more challenging for cognitive training trials to have a net positive impact in older brains than in younger brains not just because of age-related decline in the processes underlying synaptic plasticity, but also because the side effects of synaptic plasticity are likely to be higher in older brains, where metabolic waste products linger longer and so can have more toxic effects.

5.2. Future directions

How does LC activity and structural integrity over the life course accelerate or slow down Alzheimer’s disease neuropathology? To address this question, in this review I had to draw mainly on research using animal models, as relevant findings in living humans are still quite limited. However, I expect that we will soon see significant advances in answering this question in living humans due to the recent and on-going methodological advances in both imaging the LC and quantifying Aβ and tau.

A first critical step along these lines would be to improve measures of LC function and structure in large longitudinal studies. Most large publicly available longitudinal imaging studies do not include any of the recently developed scan sequences that are optimal for localizing and characterizing the LC [316, 317]. We have found that, given a sufficiently large number of participants, even general-purpose magnetic resonance images identify the LC as being a brain region where decreased structural volume predicts later conversion to Alzheimer’s disease [318]. However, scan sequences specialized to detect the LC should provide more sensitive measures and should be included in longitudinal studies, especially in studies that provide measures of Alzheimer’s disease or pathology.

We also need to characterize the typical age trajectory of noradrenergic activity in the brain and how this corresponds to structural change in the LC. For instance, one hypothesis is that in aging and early Alzheimer’s disease, LC hyperactivity exacerbates Alzheimer’s pathology but that in later stages of the disease, loss of LC neurons and diminished noradrenergic activity lead to cognitive impairment and neuroinflammation [264]. As outlined in Section 4.2, we still have only scant human data regarding age differences in the levels of production and clearance of noradrenaline. Although we cannot yet measure LC neuron activity non-invasively in the human brain, we do have indirect measures such as pupil dilation [319] and the p300 event-related potential [320]. Frontoparietal attention networks provide another indirect window into LC activity and integrity [321-323]. Given the documented age differences in these measures, the degree to which resting-state LC activity is coordinated with frontoparietal attention networks [324, 325] and the variability of frontoparietal network activity at rest [326] may prove to be indicators of age differences in LC activity. Noradrenaline release may also lead to EEG α-β desynchronization [327] as well as increase stimulus-induced gamma oscillations [328]. At this point, mapping how these different measures relate to each other and to structural MRI LC contrast might help clarify which measures would be most relevant. In particular, animal work that includes functional MRI measures alongside direct measures of LC activity would be very helpful for advancing LC imaging in humans. Developing well-validated measures will be essential to guide future interventions and to characterize the trajectory of change in LC activity during aging and Alzheimer’s disease.

Indeed, the ultimate goal of this line of inquiry is to slow down the accumulation of Alzheimer’s pathology. Pharmacological, behavioral and lifestyle interventions all have some potential; however, gaining more information regarding how noradrenergic activity changes in aging and how these changes affect Aβ and tau will be essential to design the best targeted interventions.