Targeting norepinephrine in mild cognitive impairment and Alzheimer's disease

The Alzheimer's disease (AD) epidemic is a looming crisis, with an urgent need for new therapies to delay or prevent symptom onset and progression. There is growing awareness that clinical trials must target stage-appropriate pathophysiological mechanisms to effectively develop disease-modifying treatments. Advances in AD biomarker research have demonstrated changes in amyloid-beta (Aβ), brain metabolism and other pathophysiologies prior to the onset of memory loss, with some markers possibly changing one or two decades earlier. These findings suggest that amyloid-based therapies would optimally be targeted at the earliest clinically detectable stage (such as mild cognitive impairment (MCI)) or before. Postmortem data indicate that tau lesions in the locus coeruleus (LC), the primary source of subcortical norepinephrine (NE), may be the first identifiable pathology of AD, and recent data from basic research in animal models of AD indicate that loss of NE incites a neurotoxic proinflammatory condition, reduces Aβ clearance and negatively impacts cognition - recapitulating key aspects of AD. In addition, evidence linking NE deficiency to neuroinflammation in AD also exists. By promoting proinflammatory responses, suppressing anti-inflammatory responses and impairing Aβ degradation and clearance, LC degeneration and NE loss can be considered a triple threat to AD pathogenesis. Remarkably, restoration of NE reverses these effects and slows neurodegeneration in animal models, raising the possibility that treatments which increase NE transmission may have the potential to delay or reverse AD-related pathology. This review describes the evidence supporting a key role for noradrenergic-based therapies to slow or prevent progressive neurodegeneration in AD. Specifically, since MCI coincides with the onset of clinical symptoms and brain atrophy, and LC pathology is already present at this early stage of AD pathogenesis, MCI may offer a critical window of time to initiate novel noradrenergic-based therapies aimed at the secondary wave of events that lead to progressive neurodegeneration. Because of the widespread clinical use of drugs with a NE-based mechanism of action, there are immediate opportunities to repurpose existing medications. For example, NE transport inhibitors and NE-precursor therapies that are used for treatment of neurologic and psychiatric disorders have shown promise in animal models of AD, and are now prime candidates for early-phase clinical trials in humans.

and neuropsychiatric symptoms [6,[19][20][21][22][23][24]. A number of studies have also demonstrated signifi cant correlations between LC cell death (or decreased cortical NE levels) and severity and duration of dementia in AD [25,26]. Neurofi brillary changes in the LC occur in prodromal stages of AD (that is, mild cognitive impairment (MCI)), and even in some young, cognitively normal individuals [16][17][18], preceding amyloid-beta (Aβ) deposition. However, whether the LC represents the initial site of pathology or refl ects a nonspecifi c response to brain insults is still under debate [27].
An additional complication is that compensatory changes in the degenerating noradrenergic system appear to occur in AD; despite decreases in tissue forebrain NE in AD, surviving LC neurons show increased abundance of mRNA for tyrosine hydroxylase, the rate-limiting NE biosynthetic enzyme, sprouting of dendrites and axonal projections [28], and increased cerebrospinal fl uid levels of NE are observed in AD patients [29][30][31][32]. Th e knowledge gaps present in these areas highlight the need for additional investigations into the mechanism by which LC loss contributes to AD.

Locus coeruleus and norepinephrine in AD pathogenesis: preclinical studies
Th e strong correlation between LC degeneration, NE depletion and severity of AD in patients has prompted multiple studies of the contribution of LC dysfunction to AD progression through the use of animal models. Th e primary tool for studying the eff ects of LC degeneration and NE depletion in vivo is the neurotoxin N-(2chloroethyl)-N-ethyl-2-bromobenzylamine (dsp-4), which reliably lesions the LC while leaving other aminergic systems intact. Transgenic mice that overexpress human amyloid precursor protein (APP) with familial Alzheimer mutations recapitulate many aspects of AD neuropathology and cognitive defi cits, and have been used extensively to study AD. However, most of these mouse lines do not show the frank LC degeneration that occurs in human AD. To determine the functional consequences of LC loss in AD, therefore, several laboratories have used dsp-4 to lesion LC neurons in these transgenic mice.
In general, dsp-4 lesions of the LC exacerbate AD-like neuropathology and cognitive defi cits, suggesting that LC degeneration plays a causal role in AD progression. For example, the fi rst study to use this approach showed that dsp-4 lesions of the LC in APP23 mice resulted in increased Aβ deposition, neurodegeneration, neuronal loss, cognitive defi cits and microglial activation, and reduced cerebral glucose metabolism [33]. Importantly, the eff ects of dsp-4 were confi ned to forebrain areas that received projections directly from the LC, while brain regions that receive noradrenergic innervation from non-LC cell groups were unaff ected. APP/presenilin-1 (PS1) mice treated with dsp-4 displayed severe loss of norepinephrine transporter (NET) in the LC and cortex, along with a loss of noradrenergic innervation [34]. Lesioning of the LC induced accelerated amyloid deposition and neuron death with age, and more severe defi cits in spatial memory compared with vehicle-treated animals [34]. Th e mechanism underlying the increased amyloid deposition appears to be related to reduced clearance, as occurs in sporadic AD [35], due to the inhibition of Aβ 1-42 (Aβ42) phagocytosis by microglia rather than an infl uence on APP production or processing [36]. NE has several strong infl uences on microglial function, and in general suppresses the production of proinfl ammatory cytokines and promotes the production of anti-infl ammatory molecules. Th us, it is not surprising that dsp-4 treatment also exacerbates the neuroinfl am matory response in multiple brain regions of APP/PS1 mice [36,37]. Interestingly, a recent study reported that in addition to increased Aβ deposition, dsp-4 lesions of the LC in APP/PS1 mice also resulted in olfactory defi cits, another common and early pathology seen in AD patients [38].
Among the questions raised by these fi ndings, an important issue with therapeutic implications is whether the eff ects of LC lesions in AD mouse models are due solely to the loss of NE itself, the loss of co-transmitters in LC neurons, collateral damage from the neuro degenera tive process itself, or some combination thereof. To help resolve these issues, we recently crossed APP/ PS1 mice with dopamine β-hydroxylase knockout (DBH -/-) mice that lack the ability to synthesize NE but have intact LC neurons [39]. While APP/PS1 and DBH -/singlemutant mice each displayed moderate hippocampal longterm potentiation (LTP) and spatial memory impairments, the two mutations had an additive eff ect, resulting in double mutants with severely compromised LTP and maze per for mance. Somewhat surprisingly, the genetic loss of NE had no apparent eff ect on AD-like neuropathology in the double mutant. Nondegenerative loss of NE produced by Ear2 knockout, which prevents the development of most LC neurons, also exacerbated LTP and memory defi cits but had no eff ect on plaque deposition in APP/PS1 mice. However, dsp-4 worsened neuropathology in the APP/PS1, DBH -/double mutant. Combined, these results indicate that the LC neuronal loss contributes to distinct aspects of AD; loss of NE itself impairs synaptic plasticity and cognitive performance, while the physical process of LC neuron degeneration exacerbates AD-like neuropathology.
In summary, combining expression of familial AD mutations with LC lesions or NE defi ciency appears to more closely recapitulate the neuropathological and cogni tive symptoms of AD compared with mutant APP expression alone, and implicates LC loss as a crucial component of AD.

Neuroinfl ammation is a key mechanism linking loss of locus coeruleus neurons and norepinephrine innervation with AD
Recent studies provide insights into the mechanisms by which LC dysfunction and NE loss facilitate AD pathogenesis. Th ere is growing evidence suggesting that the infl ammatory response induced and/or augmented by LC degeneration is a key mechanism contributing to the initiation and progression of AD pathogenesis. Microglia, astrocytes and endothelia are among the major targets of NE, and, under normal conditions, these cells control the delicate balance of the infl ammatory response. In general, NE is an anti-infl ammatory molecule; acting via βadrenergic receptors, NE suppresses the expression of multiple proinfl ammatory genes, including major histocompatibility complex class II, TNFα, inducible nitric oxide synthase and IL-1β, while simultaneously promoting the expression of anti-infl ammatory molecules such as NF-κB, inhibitory IκB, heat shock protein-70 and chemokine monocyte chemotactic protein-1 in astrocytes and microglia [7,40]. Th at NE defi ciency results in undesirable proinfl ammatory eff ects is therefore not surprising.
One of the fi rst pieces of evidence connecting LC degeneration and neuroinfl ammation in an AD model was reported by Heneka and colleagues [41]. Injections of Aβ42 in the cortex of rats induced severe cortical infl ammation and the expression of several proinfl ammatory genes -including inducible nitric oxide synthase/nitric oxide synthase-2, IL-1β and IL-6 -within hours. Th is neuro infl ammation was profoundly exacerbated when LC neurons were lesioned with dsp-4 prior to the cortical injection of Aβ42 . In addition, dsp-4 pretreatment increased inducible nitric oxide synthase expression solely in neurons rather than in microglial cells, more accurately replicating the expression pattern seen in AD patients [41]. Augmented forebrain microglial and astroglial acti va tion and proinfl ammatory gene expres sion that coincide with the development of other AD-like neuro pathologies such as Aβ plaques were also obtained using dsp-4 and the APPV171 and APP/PS1 transgenic mouse models of AD [36]. LC lesions profoundly increased the Aβ plaque load, brain infl ammation and spatial memory defi cits concurrently in APP23 transgenic mice. In addition, dsp-4 treatment was associated with a switch in microglial cytokine expres sion from a neuroprotective anti-infl ammatory profi le to a proinfl ammatory and neurotoxic profi le [33,36,42].
Because NE promotes microglia-mediated degradation and phagocytosis of Aβ in cell culture [43], another deleterious eff ect of LC degeneration on the neuroinfl ammatory response is the dysfunction of cellular machinery involved in Aβ metabolism and clearance. For example, in V717F APP transgenic mice, dsp-4 lesions of the LC produce a fi vefold increase in Aβ plaques that is accompanied by microglial and astroglia activation and decreased expression of the Aβ plaque-degrading enzyme, metallopeptidase neprilysin [42]. Another study showed that NE suppressed Aβ-induced cytokine and chemokine production and increased microglial migration and phago cytosis in cell culture, while dsp-4 lesions prevented the recruitment of microglia to Aβ plaques and impaired Aβ phagocytosis in APP/PS1 transgenic mice [36].
A few epidemiological studies have investigated interactions between NE and neuroinfl ammation in AD. A small pilot study in a Spanish population found that a SNP associated with low DBH activity alone had no eff ect, but signifi cantly increased AD risk in combination with SNPs in the IL-1A or IL-6 genes [44]. Th is result was partially confi rmed and extended in an independent study with a larger sample population and wider patient demographics. Th is follow-up study reported a significant association between the low-activity variant of DBH alone and AD risk that was mostly attributable to males over the age of 75, and also replicated the interaction between DBH and IL-1A polymorphisms [45]. Interestingly, SNPs that are thought to increase adrenergic signaling have also been linked to a risk for developing AD. Individuals homozygous for the C allele of ADRB1 (the β1-adrenergic receptor) and the the T allele of GNB3 (the G protein β3 subunit gene), which are associated with in creased cAMP levels and mitogen-activated protein kinase activation, have an increased risk for AD [46]. A Chinese case-control study found that a β 2adrenergic receptor polymorphism which enhances responsiveness is also associated with the risk of sporadic late-onset AD [47]. Th ese studies highlight the complicated nature of noradrenergic signal ing in AD; activation of some receptor subtypes may suppress neuroinfl ammation and neuropathology, while other receptors may exacerbate aspects of the disease.
Recent biomarker studies in living subjects have also confi rmed a proinfl ammatory state in AD [48][49][50][51]. Of note, increased proinfl ammatory and decreased antiinfl ammatory markers account for the majority of changes detectable in a large panel of cerebrospinal fl uid analytes in MCI and AD [49,50]. By promoting proinfl am matory responses, suppressing anti-infl amma tory responses and impairing Aβ degradation and clearance, LC degeneration and NE loss can therefore be considered a triple threat to AD pathogenesis.

Treatments that increase norepinephrine in AD animal models ameliorate AD-like pathology and cognitive decline
In vitro and animal studies have provided the most compelling evidence that increasing NE could have benefi cial eff ects on both AD neuropathology and cognitive symptoms. In vitro challenge of human acute monocytic leukemia cells (THP-1) with Aβ42 induced cytotoxicity and provoked a neuroinfl ammatory response that was dose-dependently attenuated by NE [52]. Treatment with cAMP or forskolin, a protein kinase A activator, had similar eff ects, suggesting that NE's protective eff ects were regulated, at least in part, via stimulation of β-adrenergic receptors and the corresponding activation of the cAMP/protein kinase A signaling pathway [52]. Another in vitro study in hNT neuronal and primary hippocampal cultures revealed a neuroprotective eff ect of NE against both Aβ42-and Aβ 25-35 -induced increases in oxidative stress, mitochondrial dysfunction and cell death [53]. Th e neuroprotective eff ects were mediated by activation of β-adrenoceptor/cAMP signaling and also required the brain-derived neurotrophic factor/tropo myosin-related kinase B pathway, although some β-receptor-independent eff ects of NE persisted [53].
Th e earliest in vivo animal studies using noradrenergic pharmacotherapies focused on the α 2 -adrenergic autoreceptor. Th e α 2 -antagonists that enhance NE release, such as piperoxane, reversed memory defi cits in aged mice as assessed by performance in a step-down inhibitory avoidance response task [54]. Another α 2 -antagonist, fl uparoxan, prevented age-related decline in the spontaneous alternation task (a test of spatial working memory) in APP/PS1 mice, although it had no eff ect in other memory tasks such as object recognition or the Morris water maze, and occurred in the absence of obvious concomitant change in pathology [55]. Drugs targeting other NE receptors and transporters have also been tested in animal models of AD. Desipramine, a tricyclic antidepressant that inhibits endogenous NE reuptake, induced the production of the anti-infl ammatory cytokine monocyte chemotactic protein-1 [56]. CL316243, a selective β 3 -adrenergic receptor agonist, rescued perfor mance in a learning paradigm by chicks given intracranial injections of Aβ42 [57]. Recently, β-adrenoceptor activation of cAMP/protein kinase A signaling was found to reverse the synaptotoxic eff ects of human Aβ oligomers on LTP and behavior [58].
Compelling evidence in favor of noradrenergic treatments for AD has also been observed using the NE precursor, l-threo-3,4-dihydroxyphenylserine (L-DOPS). For example, L-DOPS restored the balance of the brain infl ammatory system, facilitated microglial migration and Aβ phagocytosis, and reversed learning defi cits in dsp-4 lesioned APP transgenic mice [36], and also partially rescued spatial memory defi cits in the DBH -/-, APP/PS1 double-mutant mice [39]. Treatment of 5xFAD mice, which have robust and early development of ADlike neuropathology, with a combination of L-DOPS and the NET inhibitor, atomoxetine, elevated brain NE levels, increased expression of Aβ clearance enzymes and brainderived neurotrophic factor, reduced infl ammatory changes and Aβ burden, and improved spatial memory [59].
To generate further proof-of-principle for the effi cacy of NET inhibitors in AD, we took advantage of norepinephrine transporter knockout mice (NET KO) that lack the NET completely, and have elevated basal extracellular NE levels, similar to what might be observed with chronic NET inhibitor treatment [60]. We crossed the NET KO mice to APP/PS1 transgenic mice that overexpress mutant human APP and PS1 and develop age-dependent Aβ plaques, and examined AD-like neuropathology by western blot assay at 6 months of age and by immunocytochemistry at 1 year of age. As shown in Figure 1a, APP/PS1 mice that carry wildtype copies of NET (NET WT, APP/PS1) contain heavy plaque load in the hippocampus and cortex, as detected by immunohistochemistry using antiserum 2964 against fi brillar Aβ42 [61]. Th e Aβ levels were much higher in female NET WT, APP/PS1 mice compared with males (Figure 1b), as reported previously for APP/PS1 and other lines of APP transgenic mice (for example, [62]). Remarkably, plaques were almost completely abolished in littermate APP/PS1 mice that lack the NET (NET KO, APP/PS1). Similar results were obtained with western blots of brain homogenates (Figure 1b).
Th ese results suggest that attenuating NET activity can reduce Aβ levels, perhaps by increasing phagocytosis or another NE-mediated mechanism described in this review. Interestingly, full-length APP and the C-terminal fragment of APP were also reduced. Th e reasons for this are not clear, but raise the possibility that a change in APP production or turnover contributes to the decrease in Aβ levels. Consistent with this fi nding, selective lesion of the ascending noradrenergic bundle with 6-hydroxydopamine in rats increased cortical APP [63]. Com bined with the results that atomoxetine + L-DOPS reduces ADlike neuropathology and cognitive defi cits in 5xFAD mice [59], these data support the use of NET inhibitors in AD patient populations.
While studies using NE pharmacotherapy in AD models show promise for disease treatment, these studies must be interpreted with caution because the eff ects of noradrenergic drugs are complicated by multiple adrener gic receptor subtypes with diff erent distributions and signaling capabilities. Th ere are a number of studies that suggest noradrenergic stimulation actually increases certain proinfl ammatory markers, and that some adrenergic receptor blockade can be therapeutic. Pharmacological activation of β-adrenergic receptors (especially β 2 -adrenergic) increases mRNA and protein levels for IL-1B and 1L-6 in macrophages, microglia and brain parenchyma [64][65][66]. Administration of adrenergic receptor antagonists in vivo can protect against the infl am matory response induced by a foot shock [67], peripheral bacterial challenge [68] or ischemia [69,70]. Nevibolol, a β 1 -blocker, can also reduce amyloid production in TG2576 mice that have established amyloid and cognitive impairment, although it does not improve cognition [71]. One potential explanation for the dual benefi cial and harmful eff ects of adrenergic receptor stimulation is that the loss of LC neurons coupled with the compensatory sprouting by surviving cells probably creates a situation where NE transmission is compromised in some brain regions, and overactive in others [6,[19][20][21][22][23][24]28].

Clinical studies of pharmacotherapies that modulate norepinephrine in AD
Most clinical studies using noradrenergic pharmacotherapy to date have been primarily focused on treating the aggression and other behavioral disturbances that occur in many late-stage AD patients. β-adrenergic receptor antagonists (that is, propranolol) are somewhat eff ective in the treatment of aggression and agitation, which may be caused by NE overstimulation [72,73], while antidepressants inhibiting NE reuptake, such as the tricyclic imipramine, have been used to treat depression, which may be caused by NE defi ciency [74]. Tantalizing pieces of evidence continue to support the idea of increasing NE to treat cognitive impairment in AD. For example, clonidine -which suppresses NE release by activat ing the α 2 -adrenergic autoreceptor -impairs short-term recognition memory in patients [75], suggesting that facilitating NE release may be benefi cial. Th e same group determined that clonidine could also enhance spatial working memory in AD patients [76], however, highlighting the complexity of these processes. Several clinical studies examining hypertension suggest that β-blockers may have therapeutic eff ects on infl amma tion and dementia. Dementia incidence and annual rate of cognitive decline tend to be lower in older patients that take β-blockers for hypertension [77][78][79]. Th e β 1antagonists nevibolol and metoprolol have been shown to attenuate the release of atherosclerotic infl ammatory markers such as soluble intercellular adhesion molecule-1 in humans after 1 year of treatment [80]. Since hypertension itself is a risk factor for AD, however, it is diffi cult to know whether the benefi ts of β-blockade are mediated by direct eff ects on neuro infl ammation or are indirect eff ects mediated by control of hypertension.
Overall, the strong links between LC/NE loss in AD and disease progression in AD animal models combined with human clinical and preclinical data demonstrate the exciting disease-modifying potential of drugs that modulate NE levels. Th e urgent and essential next step is to translate these discoveries to humans. Although NE pharmacotherapies are widely used in medicine, drugs that regulate NE transmission in the brain could have com plicated eff ects in AD. Th e integrity of the LC and pharmacological responsiveness in prodromal stages of AD are poorly understood. While preclinical studies suggest potential for NE-enhancing therapies to reduce neuroinfl ammation and amyloid burden and to ameliorate cognitive impairment, clinical observations in AD patients also suggest the potential to impact noncognitive symptoms of AD including mood, apathy, disinhibition, sleep, agitation and aggression [81,82].
Several NE pharmacotherapies are already used in clinical practice for a variety of neurological and psychiatric disorders, including attention-defi cit disorder, depression and orthostatic hypotension. NET inhibitors such as atomoxetine, a US Food and Drug Adminis tration-approved drug that is a widely prescribed treatment for children and adults with attention-defi cit hyperactive disorder, and reboxetine, approved in many countries around the world for depression, have been used safely in older subjects. Th e NE prodrug L-DOPS crosses the blood-brain barrier and has been used safely in Asia for several decades to treat hypotension. As mentioned above, treatment of 5xFAD transgenic mice (which accumulate amyloid burden at early ages) with a combination of L-DOPS and atomoxetine elevated brain NE levels, increased expression of Aβ clearance enzymes and brain-derived neurotrophic factor, reduced infl ammatory changes and Aβ burden, and improved spatial memory [59].
In clinical studies, atomoxetine has also been shown to improve working memory, response inhibition and other executive functions in patients with attention-defi cit hyperactivity disorder [83][84][85][86]. Several small studies have examined atomoxetine treatment in older patients with neurodegenerative disease to assess safety, tolerability and symptomatic eff ects. Marsh and colleagues studied 12 patients with Parkinson's disease with doses up to 100 mg daily (mean tolerated dose 89.6 mg), with excellent safety, tolerability and improved executive function [82]. Weintraub and colleagues found that 80 mg once daily was well tolerated by Parkinson's disease subjects as a treatment for depression; only four of 29 patients withdrew because of adverse eff ects [87]. Although atomoxetine was ineff ective for the treatment of depression in the study, atomoxetine was associated with improvement of global cognition. A 6-month phase II trial in mild to moderate AD tested up to 80 mg atomoxetine once daily in 47 subjects [88]. Although atomoxetine was well tolerated (only fi ve subjects withdrew because of adverse eff ects), there were no signifi cant improvements in cognitive function, global clinical impression or neuropsychiatric symptoms. However, this study was not powered for clinical effi cacy and, more importantly, did not investigate the potential anti-infl ammatory neuroprotective role of NE pharmacotherapy. Moreover, since patients with mild to moderate AD already have extensive neurodegeneration, most investigators now realize the best chance for neuroprotection will come from earlier intervention.
Logical next steps would therefore be to test NE pharmacotherapies for their potential anti-infl ammatory and other neuroprotective mechanisms in phase II trials with individuals with preclinical or early clinical (that is, MCI) stages of AD. For example, it would be important to evaluate the eff ect of NE-based treatments such as atomoxetine and L-DOPS on biomarkers of AD pathology and infl ammation [49,50,89,90]. A potential target would be cerebrospinal fl uid infl ammatory markers, which have been used successfully as surrogate markers of drug response in multiple sclerosis [91,92] and are among novel biomarkers that distinguish MCI and AD from other neurodegenerative diseases and correlate with both baseline cognitive impairment and subsequent cognitive decline [50].
In sum, there is a growing body of evidence linking LC neurodegeneration and altered NE neurotransmission to the pathogenesis of AD, in addition to the longestablished links with cognitive and behavioral symptoms. Preclinical studies demonstrate that restoration of NE function has great potential to slow neurodegeneration by enhancing anti-infl ammatory and suppressing proinfl ammatory responses, facilitating amyloid clearance and via other protective mechanisms. However, the complexities of NE signaling and multiplicity of eff ects of adrenergic receptor subtypes, together with the limitations of animal studies, underscore the importance of translating these studies to humans. Th e availability of clinically approved drugs that enhance central nor adrenergic function provides a timely opportunity to repurpose their use to determine their potential as a novel disease-modifying therapeutic strategy.

Competing interests
TC, BK, MPK, TH, MTH, DW and AIL declare that they have no competing interests. WTH received one compensated meal from Eli Lilly as part of the Alzheimer's Association International Conference (under $100). WTH has patents pending on cerebrospinal fl uid biomarkers for f rontotemporal lobar degeneration and plasma biomarkers for AD. Some markers in these panels overlap with cerebrospinal fl uid biomarkers to be measured in the atomoxetine trial.