Systemic and Central Immunity in Alzheimer's Disease: Therapeutic Implications

SUMMARY

Clinical pharmaceutical trials aimed at modulating the immune system in Alzheimer's Disease have largely focused on either dampening down central proinflammatory innate immunity or have manipulated adaptive immunity to facilitate the removal of centrally deposited beta amyloid. To date, these trials have had mixed clinical therapeutic effects. However, a number of clinical studies have demonstrated disturbances of both systemic and central innate immunity in Alzheimer's Disease and attention has been drawn to the close communication pathways between central and systemic immunity. This paper highlights the need to take into account the potential systemic effects of drugs aimed at modulating central immunity and the possibility of developing novel therapeutic approaches based on the manipulation of systemic immunity and its communication with the central nervous system.

Introduction

In the past the parenchyma of the central nervous system (CNS) was considered an “immunologically privileged site” because the blood brain barrier (BBB) was thought to prevent the entry, or exit, of many molecules including antibodies from the periphery and to be devoid of macrophages and lymphocytes. However, it is clear that even in the presence of an intact BBB the CNS is capable of mounting inflammatory responses, albeit atypical, in response to tissue injury and infection [1].

Many aspects of innate immunity have been detected in the CNS and many cell types specific to the CNS, including microglial cells and astrocytes, are capable of performing these roles.

In the brain, microglia cells are considered “the CNS professional macrophages”[2]. Indeed microglial cells are, in essence, brain macrophages that entered the brain during embryogenesis [3] and like macrophages, they appear to survey their local environment looking for tissue damage or evidence of infection [4]. In the CNS microglial cells are largely downregulated with low or undetectable expression of cell surface antigens such as Major Histocompatability (MHC) class I and II molecules [5]. However, following an acute insult, such as a head injury or a CNS infection, resident microglia, like tissue macrophages, transform from their normal quiescent state to a morphologically different activation state that is characterized by the production of pro‐inflammatory cytokines, such as Interleukin‐1 (IL‐1), Interleukin‐6 (IL‐6) and Tumour necrosis factor α (TNFα) and an upregulation or de novo synthesis of cell surface receptors or cytoplasmic antigens [6]. This involves a wide variety of receptors[7] but includes advanced glycoslyated end products (RAGE), the peripheral benzodiazepine receptor [8, 9] and toll‐like receptors (TLRs) with emerging data highlighting the importance of the TLRs in the regulation of the innate immune response [10]. This upregulation is tightly regulated at the translational level by antiinflammatory molecules such as Transforming growth factor 1β (TGF‐1β) and Interleukin‐10 (IL‐10) [11] and also by interactions with neuronal cells. Thus, neurons are known to express ligands, for example CD200, that interact with receptors, for example CD200R, on the surface of microglia to generate a downregulated phenotype [3, 12].

Astrocytes, unlike macrophages, are ectodermally derived. Their role in mediating CNS inflammation has been relatively neglected but they also have an important role in innate immunity, including cytokine production, complement and antigen presenting cell properties. In addition, because of their location in close contact with CNS resident cells and blood vessels they can also act to modify BBB permeability and thus support an adaptive immune response [13].

Adaptive immunity is not thought to be as important as innate immunity in the CNS. Thus, microglial cells are poor antigen presenting cells [14]. Lymphocytes are also not found in large numbers in the normal CNS and although activated T helper cells (T_H) are able to enter the CNS those that fail to encounter antigen leave within 1–2 days of entry [15]. There are also few reports of B lymphocytes entering the normal CNS [16].

The Immune System in Alzheimer's Disease

The Systemic Immune System

Cross sectional studies have been variable in terms of establishing differences in serum or plasma markers of innate immunity between Alzheimer's Disease (AD) populations and age matched control groups. Some studies have found increases in plasma markers of pro‐inflammatory cytokines (principally TNFα or IL6) in AD compared with controls [17, 18, 19, 20, 21, 22, 23, 24] others have found no or mixed differences [25, 26] and yet others reduced levels [27, 28].

More consistent have been a number of studies that have suggested an association between peripheral blood indicators of systemic inflammation and the subsequent development of AD. Thus, inflammatory proteins in plasma, notably C reactive protein (CRP) and IL‐6, have been found to be elevated 5 years before the clinical onset of dementia in a number of studies [29, 30, 31]. Indeed, one long‐term follow‐up study has suggested that a raised CRP is associated with a 3‐fold increased risk of developing AD up to 25 years later [32]. However, care needs to be taken when interpreting systemic immune markers and their relationship with AD. Differences between plasma or serum markers of inflammation (e.g., serum CRP or cytokine concentrations) between an AD and a control group is likely to be subject to a plethora of other factors that will either exaggerate or conceal differences between these groups. Thus, there is a need to correct for established confounders, for example, medications (e.g., cholinesterase inhibitors) as well as other factors (time of sampling) that may alter peripheral markers of inflammation but which cannot be considered to be risk factors for the development of AD. However, and equally important, correcting for some established risk factors for AD (e.g., diabetes, obesity, atherosclerosis, vascular disease) may underestimate the role of systemic inflammation as a risk factor for the development of AD. An additional complication is that a number of comorbid inflammatory conditions may be difficult to detect or be largely asymptomatic (e.g., periodontitis; soft tissue injury).

An approach that helps to reduce variability due to uncontrollable environmental inflammatory trigger factors is to examine the ability of whole blood or peripheral blood mononuclear cells (PBMCs) to produce cytokines following a nonspecific mitogen, such as lipopolysaccharide (LPS) and phytohaemagglutinin (PHA) challenge. This, in effect, is a measure of an individual's intrinsic cytokine producing ability following a controlled stimulus. Using this approach a number of cross sectional studies [33, 34, 35] have shown an increase in the intrinsic production of proinflammatory cytokines in AD subjects compared with controls; although not all [36]. More recently, a prospective study [37] suggested that cognitively intact individuals in the top tertile of PBMC TNFα (or IL‐1β) production have an approximately three times increased risk of developing AD compared with those in the lowest tertile.

The role of systemic adaptive immunity in the development or natural progression of AD has been underexplored and studies so far have produced mixed findings. Thus, whilst some comparisons of plasma samples in AD and control subjects have shown some evidence for an increase in antibody titres to beta amyloid (Aβ) in AD compared to control subjects [38, 39] other studies have no significant increases or even decreases [40, 41] with other studies [42] suggesting that serum antibodies for oligomeric preparations of Aβ1–42 decline with advancing AD.

The Central Immune System

Cerebrospinal Fluid

In theory because cerebrospinal fluid (CSF) bathes the brain and exchanges with the extracellular fluid it should contain molecules produced by neurones, astrocytes and microglia, and could therefore provide an indication of how these cells are altered in AD. However, peripheral biomarkers of inflammation may also gain access from the bloodstream and so they may not be specific for neuroinflammation. In addition, the action of inflammatory signals in the brain may be very localized and short lived without extensive diffusion into the CSF and so their absence cannot be equated with a lack of effect. Many studies have measured CSF levels of cytokines and other inflammatory markers in AD. Some studies [43, 44] have shown evidence of increases in proinflammatory cytokines including IL‐1β; IL‐6 and TNFα in AD compared with control subjects. However, other studies [45, 46] have found no significant differences.

Post Mortem Brain Studies

Disturbances of brain innate immunity in AD has been extensively reviewed elsewhere [10, 47]. Direct antibody‐independent activation of the alternative complement pathway by fibrillar Aβ[48] and decoration of dystrophic neurites with membrane attack complexes (MAC) [49] has been shown. ApoJ (clusterin), a complement defense protein, shows increases in the AD brain and is associated with senile plaques, possibly reflecting the need to protect against ongoing complement activation [50]. Furthermore, expression of class II MHC is increased on the surface of microglia cells in AD compared with control brain tissue [51]. Microglia aggregate more around amyloid‐containing neuritic plaques than diffuse plaques in AD [52] and many different laboratories have shown that microglia, both in vivo and in culture, phagocytose exogenous fibrillar Aβ[53]. These interactions are modulated in part by TLR2, TLR4 and TLR9 [54, 55, 56] suggesting that the expression of these innate immune receptors might be a mechanism to prevent the accumulation of Aβ in the CNS that may be impaired in AD [10]. Indeed, it has been suggested that Aβ may be interpreted as an invading pathogen since bacteria produce similar amyloidogenic aggregates on their cell surface [7, 57]. However, although phagocytosis of Aβ has generally been considered beneficial, there is also the possibility that this process may be harmful. Phagocytosis by peripheral macrophages is accompanied by the release of cytotoxic compounds. Moreover, phagocytosis by brain‐derived macrophages in culture results in the release of potentially destructive reactive oxygen species [58], reactive nitrogen species [59], and TNF‐α[60]. Interestingly, recent research has suggested that Aβ, rather than being an inadvertent instigator of the innate immune reponse, might be acting as a bacteriocidal agent [61]. Thus, the accumulation of Aβ with age might represent a physiological defense reaction to infectious organisms.

Evidence for increased proinflammatory cytokines in AD brain has largely focused on studies of IL‐1 with relatively few studies examining or showing elevation of IL‐6 or TNFα levels [62]. IL‐1 has been to shown to be over‐expressed within cortical regions of the AD brain, as shown by elevated brain tissue IL‐1 concentrations and by increased numbers of IL‐1 immunoreactive microglia associated with AD plaques [63, 64]. IL‐1 overexpression also seems to occur early in plaque evolution. It is, for example, already evident in diffuse, non‐neuritic Aβ deposits, and can be observed in autopsied brain samples from children with Down's syndrome [64]. Notably, IL‐1 promotes the synthesis [65] of amyloid precursor protein (APP) and may therefore promote further amyloid production and deposition in plaques. Thus, IL‐1 has been proposed to be the initiator of a cascade of self‐perpetuating events resulting in the genesis and progression of neurodegeneration in AD [66]. However, it is important to note that the elevations of proinflammatory cytokines, where found in AD, are small compared with what might occur following a direct microbial challenge. As stated earlier this is likely to reflect the tight control of inflammation within the CNS by antiinflammatory molecules such as TGF‐β1 [67].

The role of adaptive immunity within the CNS of AD subjects has received even less attention than the peripheral immune system. However, one recent study [68] of AD brain tissue suggests that the majority of neuritic plaques in AD brain tissue are decorated by IgG and have a corresponding increase in associated phagocytic microglia.

Brain Imaging Studies

The demonstration that the peripheral benzodiazepine receptor is upregulated in activated microglia [9] has led to the development of a ligand [11C](R)PK11195 that can be used to label these cells in the living brain using positron emission tomography (PET) [8]. In patients with AD, initial studies suggest that the signal from [11C](R)PK11195 binding to activated microglia correlates inversely with cognitive function [69]. Furthermore, as cognitive function declines over time the [11C](R)PK11195 signal increases without a change in the signal from a ligand detecting amyloid ([11C]PIB), consistent with the view that microglia are contributing to neuronal dysfunction.

Interactions Between Systemic and CNS Inflammation in AD

It is important to recognize that even in the face of an intact BBB the periphery and the CNS communicate inflammatory signals to one another. During a systemic infection a range of behaviors occur including lethargy, apathy, decreased social interaction, and poor concentration. These centrally derived behaviors are known collectively as “sickness behavior” and are not merely unpleasant side effects of infection; together they form an important, evolutionary conserved, homeostatic mechanism that allows the body to adapt to the infection or injury [70]. Four major routes of communication from the periphery have been proposed, all of which lead to the synthesis of cytokines and inflammatory mediators in the brain. First, inflammatory events in the thoracic and abdominal cavities are signaled to the brain through vagal‐nerve sensory afferents, and in turn the vagal efferent outflow modifies these inflammatory events through acetylcholine secretion [71]. Second, circulating cytokines and other inflammatory mediators, for example, pathogen‐associated molecular patterns (PAMPS) enter the blood and communicate directly with perivascular macrophages and other cells in the circumventricular organs, that lack a BBB, initiating the transcription of proinflammatory cytokines across the brain parenchyma [10]. Third, cytokines and other inflammatory mediators [72] interact with the induction of lipid mediators (e.g., prostaglandin E2) that communicate directly across the BBB [73]. Fourth, there may be direct entry of immune cells (monocytes and possibly bone marrow‐derived microglial cells) from the periphery into the brain [74, 75].

While communication between the systemic and central immune system can occur in the presence of an intact BBB, in AD there is also evidence for a breakdown in BBB permeability [76]. Increased permeability may be because of associated vascular pathology but may also be due to the degeneration of neurones and activation of glial cells causing changes to endothelial cell phenotype and transport properties [77].

Aetiological Considerations for Defective Immunity in AD

Genetics

In a small number of individuals, less than 0.1% of the total AD population, mutations in one of three genes, Presenilin 1, Presenilin 2 and APP or a duplication of the APP gene [78, 79] are directly responsible for the development of largely early onset AD by altering APP processing so that Aβ deposition is greatly enhanced [80]. However, a majority (95%) of the cases of AD have an exponential growth in incidence after 65 years of age and are of late onset [81]. Until recently, the only established genetic risk factor for the development of late onset AD was Apolipoprotein (ApoE) e4. ApoE e4 influences the degree of activation and neurotoxin production of microglia cells to Aβ[82] and is also a risk factor in other chronic neurodegenerative conditions including Parkinson's disease that do not have Aβ plaques as a feature [83]. ApoE e4 phenotype may influence outcomes for a number of CNS and systemic infections [84, 85, 86] and, more recently, has been shown to be associated with an attenuation of the peripheral blood levels of CRP in nondemented subjects [87, 88]. In addition, a large number of other genetic polymorphisms in proinflammatory genes, either alone or in association with ApoE e4, have also been implicated as risk factors for the development of AD. These include IL‐1 [89]; IL‐6 [90]; TNF‐α[91, 92, 93] and α1‐antichymotrypsin [94]. These findings have not, however, always been replicated, possibly due to the small individual effect size of these polymorphisms. However, more recently, two large‐scale genome‐wide association studies have been performed that have put the direct role of genetic variation in innate immunity on a clearer footing. These two studies [95, 96] both suggest that genes with important roles in immunity are genetic risk factors for the development of late onset AD. Both studies thus identified ApoJ and complement receptor 1 (CR1) the gene that encodes for the main receptor of the complement C3b protein. Further support for the increased expression of innate pro‐inflammatory cytokine profiles in relatives of AD subjects has also been shown elsewhere [97].

Environmental Factors

In late onset AD, twin studies suggest a heritability of approximately 60%[98] with non‐shared environmental risk factors playing an increasingly important role in the etiology of the disease with increasing age [99]. Research on the role of environmental infectious agents in the etiology of AD has largely followed the hypothesis that there are specific CNS pathogens (in an analogous fashion to established pathogens in other chronic neurodegenerative diseases, for example, Human Immunodeficiency Virus dementia or Creutzeld Jacob Disease) that have a direct influence on the pathogenesis of AD. Thus, Herpes simplex virus type 1, Chlamydophila pneumoniae and Borrelia burgdorferi have all been proposed as potential pathogens contributing to AD development. Each of these different pathogens have their own protagonists but there is a general paucity of consistent experimental evidence [100]. However, a wide variety of common systemic infections are a major cause of delirium in the elderly and delirium has been shown in a number of studies to be associated with an increased risk of developing dementia. Thus, in one study this increased risk was substantial with a cumulative incidence of 55% for the subsequent development of dementia in cognitively intact individuals after 1‐year follow‐up [101]. In addition, the risk of developing AD is also increased following the development of an infection in the absence of an obvious delirium. Thus, in a retrospective general practitioner database the presence of one or more infections over a 5‐year follow‐up period increased the odds of developing AD by around 2‐fold. Risk increased with increasing age a finding that is consistent with the known decline of genetic risk with increasing age [102]. Other chronic inflammatory diseases, for example, depression [103], atherosclerosis [104], and obesity [105] have a clearer epidemiological basis for being proposed to be risk factors in the development of AD. For all of these risk factors the individual attributable risk is likely to be small [81, 103, 106]. However, their combined cumulative effects over time might be considerable. Thus, it is known that the accumulation of acute and chronic inflammatory events bombarding the immune system throughout life is accompanied by an age‐dependent upregulation of the inflammatory response [107]. This suggests that increasing age, the biggest risk factor for the development of AD, could be considered to be a proxy for increased time of exposure to systemic inflammation.

We have hypothesized that the exposure of partially activated or “primed” microglial cells (arising from their chronic exposure to amyloid or degenerating neurones) to recurrent acute and chronic proinflammatory systemic signals, through one of the proposed communicating channels between the systemic and central innate immune system, may lead to a switch in phenotype and the production of a central proinflammatory cytokine profile that may act as a potent driver of neuronal degeneration [108]. A number of animal studies support this hypothesis [109, 110]. In addition, we have recently shown that in AD subjects high serum levels of TNFα at baseline (top three quartiles compared with bottom quartile) and increases in serum TNFα associated with intermittent systemic infections are associated with a marked increase in the rate of cognitive decline in AD subjects over a 6‐month period that is independent of the acute effects of delirium [111].

Therapeutic Strategies

Clinical pharmaceutical trials aimed at modulating the immune system in AD have largely focused on either dampening down central proinflammatory innate immunity or have manipulated adaptive immunity to facilitate the removal of centrally deposited beta amyloid. More recently, the possibility of developing novel therapeutic approaches based on the manipulation of systemic immunity and its communication with the CNS have been proposed.

Modulation of Central Innate Immunity

A number of existing drugs thought to have central modulatory effects have been examined as possible therapeutic targets in the treatment or development of AD. These include nonsteroidal antiinflammatory drugs (NSAIDS), PPAR‐γ agonists, statins, thalidomide, TNFα inhibitors, and other agents (Table 1).

Table 1.

Antiinflammatory agents for AD

Potential mechanisms of action	Clinical studies
NSAIDS COX inhibition causing reduced prostaglandin synthesis In vitro/animal models: Effects on amyloid pathways [47] PPAR‐γ activation [195]	Prevention: Case‐control/cohort studies: Modest reduction in risk [114, 115], greater risk reduction in ApoE e4 carriers [117] Clinical prevention trials: Overall no significant risk reduction[122] Treatment: Overall no significant benefit [118] COX2 inhibitors: No benefit [112] Some evidence of ibuprofen benefit in ApoE e4 carriers [120]
PPAR‐γ agonists Decreased microglial activation in mouse model [128] Reduced COX‐2 and iNOS expression [129]	Treatment: Overall no significant benefit [132]
RAGE Inhibitors Reduced generation of reactive oxygen species in response to Aβ[134, 135] Reduced Aβ load in a mouse model [136]	Treatment: Phase II Study (n = 67): well tolerated, not powered to show treatment effect [137]
Statins Reduced Aβ‐induced inflammation [140] Decreased lymphocyte traffic between CNS and periphery [141] Modulation of APP processing [139]	Prevention: Cochrane review of prevention trials revealed no significant risk reduction [142] Treatment: No evidence of benefit in meta‐analysis of clinical treatment trials [143]
Thalidomide Reduced TNF‐α levels and neurodegeneration in response to Aβ (mouse model) [148] Reduced reactive gliosis and inflammatory vascular pathology in response to Aβ (mouse model) [149]	Treatment: Phase II pilot study (n = 12): Too small to show changes in behaviour or cognition [150]
TNF‐α Antagonists Reduced CNS inflammation [152] Intra‐cerebral injection: reduced amyloid‐associated neuropathology in 3xTgAD mice [152] Peripheral injection: Reduced behavioural deficits in PDAPP mouse model [190]	Treatment: Peripheral injection: Phase II pilot study (n = 9): Well tolerated, too small to show clinical benefit [189] Perispinal injection: Open‐label, no control group, n = 15: improved cognitive scores in some subjects [153]
Omega‐3 Fatty Acids Found in fish oils Reduced arachidonic acid metabolites Prevents oxidative damage in mouse model [196]. Several other putative neuro‐protective and anti‐amyloid effects [197].	Prevention: Epidemiological evidence that low levels of Omega‐3 increase risk of AD [197]. Treatment: A preliminary study (n = 35) found some improvement in Clinician's Interview‐Based Impression of Change Scale (CIBIC‐plus) over the 24 week follow‐up (P= 0.008) but no change in formal cognitive scores [198]. On‐going trials
Curcumin The yellow pigment in turmeric. Possible effects on innate immunity resulting in less oxidative damage and increased amyloid clearance [199]. Decreases LPS‐stimulated IL‐1β and iNOS in a mouse model of AD [200].	Treatment: Exploratory trial (n = 30) showed no effect on cognition over 6 months [201]. On‐going trials
Sodium valproate Short chain branched fatty acid, commonly used as an anticonvulsant, with antiinflammatory and neuroprotective effects [202]. Decreases LPS‐induced microglia activation [203], although chronic use causes increased microglia activation [204].	Treatment: Anecdotal use for agitation in advanced dementia, but no clear evidence of benefit, and one Canadian trial showed worsening [205]. National Institute for Ageing (NIA) clinical trial is on‐going (VALID study)
Antibiotics Mutliple studies showing evidence for possible role of infective agents in AD [100].	Treatment: A trial of doxycycline and rifampicin in AD patients showed improved cognitive scores, despite no effect on Chlamydia burden [206]. A trial of Helicobacter pylori eradication showed improved cognitive scores at 2 years [207].
Cholinesterase inhibitors Commonly used symptomatic drugs in AD – may have antiinflammatory modes of action in addition to effects on cholinergic neurotransmission. Decrease Aβ‐induced microglia activation [208]. Alpha‐7 nicotinic receptor agonists prevent LPS‐induced microglia activation [209].	Treatment: Good evidence of symptomatic benefit [210]. Possibility that some of the efficacy is due to anti‐inflammatory effects regulated by alpha‐7 nicotinic receptors [209].

NSAIDS

The incidence of AD in cohorts of older people taking NSAIDS has been examined in several large prospective studies [112]. The largest of these studies, the Baltimore Longitudinal study of Ageing, found a relative risk of AD of 0.35 for 10 years of NSAID use [113]. Case‐control studies also suggest a reduced odds of developing AD in people taking NSAIDS regularly for long periods. Thus, a meta‐analysis of the case‐control studies carried out up to 1996 found that regular NSAID use was associated with a 2‐fold reduction in the odds of developing AD (OR = 0.5; P= 0.0002) [114] and, furthermore, the largest case control study to date (49,349 cases and 196,850 matched controls) [115] showed a significant effect of NSAIDS after 5 years of regular use, with a combined OR of 0.76 (0.68–0.85).

There is notably some epidemiological evidence that the protective effect of NSAIDS is only found in ApoE e4 allele carriers. In the Cache County cohort NSAIDS were found to slow rates of cognitive decline in ApoE e4 allele carriers only [116] and a large prospective trial showed that NSAIDS reduced the risk of developing AD, but again only in e4 allele carriers [117].

Several clinical trials of NSAIDS have now been completed. A meta‐analysis of three early trials of NSAIDS in patients with AD revealed no significant improvement in rates of cognitive decline [118]. Of these studies only one small study of indomethacin showed any significant benefit [119]. More recently a clinical trial of ibuprofen, while showing no significant effect overall, found some improvement in the rate of cognitive decline in a subgroup analysis of patients that were ApoE e4 allele carriers [120]. A potentially pivotal primary prevention study (ADAPT) using either naproxen, celecoxib or placebo in nondemented people with a family history of AD trial was unfortunately stopped at 2 years because of fears about the cardiovascular risks of celecoxib [109]. Early reports suggested no benefit for naproxen after 2 years of treatment but more recently a 4‐year follow‐up study of the ADAPT cohort showed a significant protection effect for naproxen [121].

Trials using COX‐2 specific NSAIDS have not shown any benefit in AD patients or in mild cognitive impairment [112] and primary prevention studies have also been negative. Indeed, in the ADAPT study the evidence gathered prior to the premature cessation of the trial revealed a worsening risk for the celecoxib arm [122] and another primary prevention study with rofecoxib was also negative [123].

A number of explanations have been put forward to explain these mixed clinical findings.

First, any central antiinflammatory effect of NSAIDS assumes that NSAIDS cross the BBB. However, penetration of NSAIDS into the brain is generally low; levels in the CSF are only 1–2% of the plasma levels required for a therapeutic effect in humans [124]. Ibuprofen and indomethacin, which interestingly show the most evidence for benefit in epidemiological studies, are the most lipophilic NSAIDS and are therefore likely to cross the blood‐brain barrier more easily than other NSAIDS. Second, while the major antiinflammatory effects of NSAIDS is thought to be mediated through the inhibition of prostaglandins, a number of studies show that cerebral inflammation in AD is characterized by microglia activation, IL‐1 and complement, rather than by elevated prostaglandin levels or increased COX expression [125]. Thus, while, elevated levels of prostaglandin E₂ (PGE₂) are found in the CSF of patients with AD [126], these elevated CSF PGE₂ levels decrease in AD patients as the severity of dementia increases [127]. Thus, any benefit that NSAIDS may have in AD because of a reduction in cerebral prostaglandin levels would be anticipated to reduce as the disease progressed and PGE₂ levels fell. Third, it may be that NSAIDS will only be effective if started early enough in the disease course. Hence the positive findings in the epidemiological, case control and long‐term intervention studies may be a result of early disease intervention.

PPAR‐γ Agonists

In addition to the use of NSAIDS other more specific agents aimed at stimulating PPAR‐γ have been investigated. Rosiglitazone, a PPAR‐γ agonist, is commonly used in diabetes and has been shown to reduce microglia activation in mouse models and reduce expression of COX‐2 and iNOS [128, 129]. In animal models rosiglitazone has been shown to improve learning and memory [130]. These encouraging findings led to a preliminary study in AD subjects (N = 30) that showed suggestions of improvements in memory and cognition [131]. This was followed by a large scale (N = 511) 6‐month clinical trial in patients with mild‐to‐moderate AD patients [132]. No statistically significant differences on primary end points of cognition and global impression of change were detected overall between placebo and rosiglitazone although exploratory analyses suggested that ApoE e4 noncarriers exhibited cognitive and functional improvement in response to rosiglitazone, whereas ApoE e4 carriers showed no improvement and even some decline. These genetic findings are contrary to the findings with NSAIDS and clearly require confirmation.

RAGE Inhibitors

This area has been recently extensively reviewed elsewhere [133]. RAGE acts as a receptor for Aβ on neurons, microglia and astrocytes and increased expression of RAGE is observed in regions of the brain affected by Alzheimer's disease (AD)[134]. Interaction between Aβ and RAGE leads to cell stress with the generation of reactive oxygen species [135] and so therapeutic strategies aimed at reducing this interaction are an attractive therapeutic target. Interception of Aβ interaction with RAGE, by infusion of soluble RAGE, decreases Aβ content and amyloid load, as well as improving learning/memory and synaptic function, in a murine transgenic model of Aβ accumulation [136]. Thus, this data suggests that RAGE may be a therapeutic target for AD. A phase II study of a RAGE antagonist in 67 AD subjects has been reported as being well tolerated [137].

Statins

There has been considerable interest in statin therapy as a potential preventative or symptomatic treatment for AD [138]. Cholesterol modulates APP processing in cell culture and animal models and has been implicated in the production of Aβ[139]. However, statins also have various effects on the immune system that appear independent of any effect on cholesterol metabolism. Thus, inflammation stimulated by Aβ is reduced by statin therapy [140] and therapy also results in decreased lymphocyte trafficking from the periphery to the CNS [141].

Unfortunately, the clinical effects of statins are largely negative. Thus, while retrospective case‐control studies indicated large reductions in AD risk among statin users, prospective studies have failed to find evidence of benefit. A Cochrane review of randomized, double‐blind, placebo‐controlled studies aimed at preventing AD concluded that there was no significant risk reduction [142]. A meta‐analysis of trials that used statins as a treatment for existing AD also found no evidence of benefit [143].

Statin therapy does not therefore appear to be of any benefit in AD. This may in part be due to the short half life and high liver metabolism of most statins [144] or to the differing solubility of statins in lipids or water. Lipophilic statins (lovastatin, cerivistatin, simvastatin) cross the BBB and penetrate cell membranes more effectively and thus may be more efficient in the treatment of AD than the hydrophilic statins (atorvastatin, fluvastatin, pravastatin) [138]. However, cholesterol is necessary for synaptogenesis and neural homeostasis and so cholesterol reduction may have detrimental effects on neuronal survival during chronic neurodegenerative stress [145]. The antiinflammatory effect of statins may simply not be robust enough to prevent damaging inflammation. In this regard it is notable that statins may have too small an effect on Aβ burden to prevent inflammation in the brain [146]. It is also notable that there are conflicting reports about the effects of statins on the production of pro‐inflammatory cytokines. Thus, one group has found reduced expression of inflammatory markers, including TNFα and Nitric Oxide [141], while another group has found evidence of activated microglia and increased TNFα expression [147].

Thalidomide

Thalidomide was first used in the late 1950s as an antiemetic. Despite its teratogenic effects there has been renewed interest in the powerful antiinflammatory effects of thalidomide in recent years and it is now used to treat inflammation in specific conditions such as multiple myeloma and erythema nodosum leprosum. One action of thalidomide is to decrease the stability of TNF‐α mRNA [148]. Thalidomide has therefore been used in animal models to assess the effect of reduced TNF‐α on neurodegeneration.

Intracerebral injection of Aβ causes up‐regulation of mRNA for TNF‐α and inducible nitric oxide synthase (iNOS), and subsequent neurodegeneration and behavioral change in mice. Thalidomide treatment in this animal model reduced TNF‐α and prevented neurodegeneration [148]. Furthermore, Aβ‐induced neurodegeneration is not seen in TNF‐α (–/–) knock‐out mice, supporting the hypothesis that neurodegeneration is dependent on TNF‐α[148]. Thalidomide was also neuroprotective in another animal model of the inflamed AD brain, with reduced gliosis and vascular pathology and reduced levels of TNF‐α[149].

There is only one small pilot study (n = 12) of thalidomide in patients with AD [150]. The trial was too small to show any change in behavior or cognition, although there was a nonsignificant trend toward lower serum levels of TNF‐α. Thus, thalidomide reduces the activity of TNF‐α and this may be neuroprotective in AD, but this hypothesis has not been properly tested in patients.

TNF‐α Inhibition

Work in animal models has lent support to the hypothesis that reducing TNF‐α levels will reduce neurodegeneration in AD. In an acute model of neuroinflammation, initiated by intracerebral injection of Aβ, an intracerebral injection of an anti‐TNF‐α antibody prevented the nitration of proteins in the hippocampus and the impairment of recognition memory induced by Aβ[148]. In a chronic systemic inflammation model in the 3xTgAD mouse model [151] chronic inhibition of soluble TNF signaling, using specifically engineered antibodies delivered intracerebrally, prevented amyloid‐associated neuropathology in 3xTgAD mice and reduced the deposition of intraneuronal amyloid species [152]. Etanercept has also been given as a perispinal injection to a small number of patients with AD [153, 154]. Cognitive scores has been reported as improved after administration of etanercept, but no control group was used in this open‐label study and the number of subjects was low (n = 15).

Other Agents

There are several other anti‐inflammatory agents currently under investigation (see Table 1).

Modulation of Adaptive Immunity

In the amyloid cascade hypothesis for the pathogenesis of AD amyloid deposition in the brain is thought to be the initiating step, with subsequent inflammation, tau hyperphosphorylation and eventually neurodegeneration [80]. It follows that removal of amyloid and prevention of further amyloid deposition would be beneficial. Schenk and colleagues developed an anti‐amyloid vaccine to test the hypothesis that immune‐mediated removal of amyloid from the brain would reduce neurodegeneration. Systemic vaccination of transgenic mice that overexpressed human APP with Aβ_1–42 produced high titres of antibodies directed against Aβ and reduced central Aβ deposition [155]. In a separate experiment, passive immunization with monoclonal antibodies to Aβ similarly reduced cerebral amyloid deposits implying that the beneficial effects of the vaccine were due to the generation of Aβ‐specific antibodies [156].

Aβ‐specific antibodies could reduce cerebral amyloid load in several ways. Opsonization of amyloid deposits by specific IgG antibodies allows phagocytosis by microglia [156]. Antiamyloid antibodies may directly bind to and dissolve amyloid deposits [157], with the resulting soluble oligomers being removed via the blood stream [158]. Binding and sequestration of soluble amyloid species in the blood may provide a “peripheral sink” that draws soluble amyloid out of the brain parenchyma [159]. Antibodies may also prevent the toxic effects of soluble oligomers by binding them and preventing toxic interactions [160].

The success of the initial mouse vaccine studies led to human trials of AN1792, an antiamyloid vaccine, in 2001. Eighty patients were enrolled in the initial phase I trial. Fifty‐three percent of the immunized participants developed detectable antibodies to Aβ[161]. The vaccine was altered slightly by the addition of the emulsifier polysorbate 80 and the subsequent larger phase II trial enrolled 372 patients [162]. This trial was halted after 18 out of 298 (6%) immunized patients developed symptoms of meningo‐encephalitis [163]. In the full analysis of the trial data, including the placebo group, there was no therapeutic effect on cognitive decline [162].

Post mortem examination of the vaccine‐treated patients in the initial phase I trial revealed extensive plaque clearance from the cerebral cortex [164, 165]. Microglia contained Aβ particles, implying phagocytosis as the method of clearance [165] and furthermore, the degree of plaque removal was associated with mean antibody response [166]. However, although postmortem examination of AN1792‐treated patients showed sustained and significant reductions in amyloid deposits within the brain, there were, however, no long‐term beneficial therapeutic effects. Thus, even immunized patients with almost complete plaque removal had severe end stage dementia at death, with no reduction in the time until severe dementia or survival time [166].

Antibody production by B‐cell lymphocytes in response to vaccination is facilitated by Th2 helper T‐cells. Animal models show that vaccination with Aβ_1–42 produces a Th2 response that encourages B‐cells to produce anti‐Aβ antibodies [167]. However, postmortem examination of the patients enrolled in the AN1792 trial showed that some patients had evidence of a pro‐inflammatory Th1 T‐cell reaction around some cerebral blood vessels [164, 165]. The T‐cell response to the vaccine in humans was further examined in experiments using peripheral blood mononuclear cells (PBMCs) taken from immunized patients [167]. PBMCs from many participants produced IL‐2 and IFNγ in response to challenge with Aβ, indicating a Th1 response. This excessive Th1‐mediated response in some patients, possibly associated with the choice of adjuvant in the AN1792 trial, may have been the cause of the severe meningo‐encephalitis seen in some patients [168].

The failure of the AN1792 vaccine has led to further work in mouse models with redesigned vaccines. These vaccines have been designed to reduce the risk of stimulating Th1 lymphocytes and to encourage a purely humoral immune response mediated by Th2 cells. Aβ_1–42 is thought to have one major antibody binding site at the N‐terminus and two major T‐cell epitopes located at the central and C‐terminal hydrophobic regions [160]. Newer vaccines, therefore, consist of short Aβ species containing the N‐terminus region of Aβ_1–42 with the T‐cell epitopes either altered or deleted altogether [169]. Other groups are working on vaccines using adjuvants less likely to elicit a cytotoxic Th2 response. A predominantly humoral antibody response can also be achieved by altering the route of administration of a vaccine, for example, mucosal or transdermal [170, 171, 172, 173]. DNA vaccines, which allow precise control of the immune reaction to the vaccine and have the advantage of not requiring adjuvants with unpredictable immune consequences, are also in development in animal models [174]. A DNA vaccine containing three copies of the Aβ B‐cell epitope, a chemokine (MDC/CCL22) and a Th2‐cell epitope to help drive a Th2 response, reduces AD pathology in a transgenic mouse model [175].

Even where the primary response to vaccination is antibody production rather than Th1‐mediated inflammation there may be less desirable innate immune consequences. In a microglial cell culture model opsonization of amyloid with anti‐Aβ IgG increased microglial chemotaxis and phagocytosis of the Aβ[176]. However, the phagocytosis was also associated with secretion of TNFα and IL‐6 from microglial cells. Interestingly the authors found that the NSAID indomethacin reduced TNFα and IL‐6 production in this model, without impairing Aβ clearance by the microglial cells [176]. It has therefore been suggested that NSAIDs might be a useful adjunct in clinical trials of Aβ vaccines [177].

Passive immunization reduces the risk of a Th1 cell‐mediated inflammatory response. Animal models have demonstrated similar effects on amyloid burden as with antigenic vaccination [156] and several clinical trials are underway [160]. A phase II trial of bapineuzumab, a monoclonal antibody against Aβ, has now been completed [178]. Two hundred thirty‐four patients were enrolled in the study and treated with infusions over 18 months. No significant differences were found in the primary efficacy tests of cognitive function and disability. However, a subgroup analysis showed some possible benefit in ApoE e4 noncarriers. A phase III trial is now underway that will examine the effect of ApoE status in more detail.

Amyloid vaccines are designed to break up amyloid plaques, and it is hoped that this will lead to reduced neurodegeneration. However, despite the promising results in animal studies human trials of both active and passive immunization have not yet demonstrated convincing clinical results for pre‐determined clinical outcomes. Failure of these studies may be due to a number of factors. Once dementia is diagnosed neurodegeneration in AD is already well established. Interventions that reduce amyloid burden at this stage may be too late to significantly improve the disease. Breaking plaques up into soluble amyloid species may be more harmful than beneficial. Soluble amyloid oligomers are potentially more neurotoxic than amyloid sequestered in plaques. Increased cerebral amyloid angiopathy and consequent microhemorrhages have been observed in postmortem examination of subjects from the AN1792 trial [179]. This finding may be explained by a movement of amyloid from plaques to cerebral blood vessels in vaccinated patients, with a failure of adequate perivascular drainage [180]. The side effects of meningo‐encephalitis in the AN1792 trial, and of vasogenic edema in the bapineuzumab trial, demonstrate the potential harm of unintended activation of proinflammatory pathways by vaccination.

More recently another approach has been to utilize intravenous immunoglobulin (IVIg) that is obtained from the pooled plasma of healthy human blood donors, and contains natural anti‐amyloid antibodies. In a phase I safety and preliminary efficacy clinical trial, eight AD patients were treated with IVIg (Gammagard, Baxter Healthcare Corporation, Westlake Village, CA) for 6 months of therapy. Cognitive function stopped declining in seven patients and improved in six of the eight patients [181]. A randomized phase III trial is now under way. In another phase 1 safety and preliminary efficacy clinical trial five “clinically probable or possible” AD patients were treated with IVIg. A slight improvement was observed on neuropsychological testing at 6 months in all patients except one where the score did not change between baseline and at 6 months [182].

Modulation of Systemic Immunity

It follows from the earlier discussion around the close relationship between systemic and central innate immunity that a lack of consideration of the deleterious effects of systemic inflammation on cognitive outcomes might greatly influence treatment outcomes in a number of drug studies. Thus, the exclusion of co‐morbid inflammatory disease in randomized placebo controlled trials of antiinflammatory agents will tend to reduce naturalistic cognitive decline in the placebo arm and mitigate against any beneficial effects in the treatment arm. Comorbid exclusion criteria in such studies might further explain why observational studies in populations with chronic inflammatory conditions might see more benefit to NSAIDS than randomized placebo control trials [7, 108]. Furthermore, interventions aimed primarily at central immunity might have indirect consequences on systemic immunity that might lead to modulation of the desired central effects. Thus, for example, NSAID dosage used in clinical trials for AD subjects may have little effect on, [183] or may paradoxically cause an increased production of, proinflammatory cytokines such as TNF‐α in the periphery [184, 185, 186]. This paradoxical increase in systemic TNF‐α in response to NSAID treatment might thus mitigate any central effects. Similar considerations apply to statin therapy [187]. Conversely, the beneficial effects of some agents, for example, cholinesterase inhibitors and antibiotics, might be primarily due to their systemic effects on innate immunity rather than their central effects [100, 188].

Direct modulation of systemic immunity as a way of modulating central immunity has largely focused on the modulation of adaptive immunity to remove central amyloid deposits and has already been discussed in detail. Direct manipulation of systemic innate immunity, as a therapeutic target in its own right, has been underexplored. Thus, to date, there has only been exploratory use of systemic anti‐TNF in human subjects [189], although closer examination of the effects of systemic anti‐TNF agents is warranted in view of the effective use of peripheral anti‐TNF treatment in the PDAPP mouse model of AD [190] and the documented improvements in sickness behavior in humans taking anti‐TNF agents [191]. Other systemic approaches will depend on increasing our understanding of the communicating pathways between systemic and central innate immune systems. Thus, manipulation of the inflammatory reflex has a number of potential applications in AD but no direct interventions have, as yet, been realized [192]. The finding that peripheral immune cells can be mobilized to the CNS [74] has raised the possibility that stimulation of bone marrow‐derived microglia by systemic macrophage colony‐stimulating factor might be beneficial in AD [193] as well other approaches including genetic engineering of bone marrow‐derived microglial cells to favor a more neuroprotective phenotype [194]. What is clear is that our knowledge of the molecular communication between systemic and central immunity is in its infancy and there are many uncertainties [194].

Conflict of Interest

The authors have no conflict of interest.