Whether, when and how chronic inflammation increases the risk of developing late-onset Alzheimer's disease

Neuropathological studies have revealed the presence of a broad variety of inflammation-related proteins (complement factors, acute-phase proteins, pro-inflammatory cytokines) in Alzheimer's disease (AD) brains. These constituents of innate immunity are involved in several crucial pathogenic events of the underlying pathological cascade in AD, and recent studies have shown that innate immunity is involved in the etiology of late-onset AD. Genome-wide association studies have demonstrated gene loci that are linked to the complement system. Neuropathological and experimental studies indicate that fibrillar amyloid-β (Aβ) can activate the innate immunity-related CD14 and Toll-like receptor signaling pathways of glial cells for pro-inflammatory cytokine production. The production capacity of this pathway is under genetic control and offspring with a parental history of late-onset AD have a higher production capacity for pro-inflammatory cytokines. The activation of microglia by fibrillar Aβ deposits in the early preclinical stages of AD can make the brain susceptible later on for a second immune challenge leading to enhanced production of pro-inflammatory cytokines. An example of a second immune challenge could be systemic inflammation in patients with preclinical AD. Prospective epidemiological studies show that elevated serum levels of acute phase reactants can be considered as a risk factor for AD. Clinical studies suggest that peripheral inflammation increases the risk of dementia, especially in patients with preexistent cognitive impairment, and accelerates further deterioration in demented patients. The view that peripheral inflammation can increase the risk of dementia in older people provides scope for prevention.


Introduction
A role for infl ammation in the pathogenesis of Alzheimer's disease (AD) had been discussed even in the earliest days of AD research. A hundred years ago Oskar Fischer wrote that cerebral senile plaque formation could be considered as the result of an extracellular deposition of abnormal substance in the cortex that induces a local infl ammatory reaction, followed by an aberrant regenerative response of the surrounding nerve fi bers. However, he was unsuccessful in his attempts to show the morphological characteristics of an infl ammatory response around plaques and to detect complement proteins by performing complement-binding studies. Seventy years later, with the advent of monoclonal antibodies for immuno histochemistry, complement factors and clustering of activated microglia could be demonstrated within plaques [1]. After the discovery of amyloid-β (Aβ) as the main constituent of senile plaques, the concept was formed that the Aβ peptide itself can induce a local infl ammatory response, which was supported by in vitro fi ndings showing that fi brillar Aβ can bind complement factor C1 and activate the classical complement pathway without involvement of antibodies [2]. Th e infl ammatory process in AD brains is not restricted to just a single step of the pathological process; infl ammation-related proteins are involved in several crucial pathogenic events of the underlying pathological cascade, such as Aβ generation and clearance, gliosis and increased phosphorylation of tau with accelerated tangle formation [3,4]. It is important to keep in mind that infl ammation itself has both benefi cial eff ects, such as the phagocytosis of the toxic Aβ fi brils, and detrimental eff ects on neighboring cells by prolonged elevation of pro-infl ammatory mediators. Clinicopathological studies show that the presence of activated microglia and infl ammation-related mediators in the cerebral neocortex of patients with a low Braak stage for AD pathology precedes extensive tau-related neurofi brillary pathology [5] (Figure 1). Clinical research using positron emission tomography with the peripheral benzodiazepine receptor ligand PK-11195 as a marker for activated microglia indicates that activation of microglia precedes cerebral atrophy in AD patients [6]. A positron emission tomography study using the Pittsburg compound B for visualization of fi brillar amyloid and the PK-11195 ligand for microglia activation showed that amyloid deposition with microglia activation can be detected in vivo in around 50% of patients with mild cognitive impairment [7]. Th us, neuropathological and neuroradiological studies indicate that infl ammatory changes in AD brains are a relatively early pathogenic event that precedes the process of neuropil destruction. Th e primary focus of the present paper is to review human studies for genetic, epidemiological and clinical evidence for whether, when and how infl ammation could increase the risk of developing AD.

Genetic evidence
In this section we evaluate the relationship between genetic risk factors for AD and two major components of amyloid plaques in AD brains, namely the presence of complement proteins and clusters of activated microglia, which are a source for the production of pro-infl ammatory cytokines.

Aβ-associated proteins
Complement proteins were the fi rst molecules detected in senile plaques in AD brains [8], two years before the identifi cation of Aβ as the core protein of the senile plaques in 1984. In the following years a growing list of other proteins, mostly acute phase proteins, were demon strated to be associated with Aβ deposits. Th ese so-called Aβ-associated proteins include, next to the complement proteins, α1-antichymotrypsin (ACT), apolipoprotein E (ApoE), clusterin, intercellular adhesion molecule-1, α2-macroglobulin, serum amyloid P component (SAP) and heparan sulfate proteoglycans [9][10][11][12][13][14][15]. Th ese proteins play a role in the transport, fi brillogenesis and deposition of Aβ and they are also important for the sequestration of neurotoxic Aβ species in plaques [16]. Th e presence of certain Aβ-associated proteins within plaques depends on the plaque type (see below) [17,18], and the accu mu lation of most depends on a certain degree of Aβ fi brillization; for example, SAP is found especially in plaques with fi brillar Aβ deposits but not in diff use plaques [18]. In vitro studies also indicate that a certain degree of fi bril formation is necessary for SAP to bind to Aβ, as SAP was found to bind to mature fi brils but not to protofi brils of Aβ1-42 [19].
Neuropathological studies show that diff use Aβ deposits, characterized by the presence of non-fi brillar (noncongophilic) Aβ and without neuritic changes or reactive glia, are the predominant plaque types in non-demented controls, and that the amount of fi brillar (congophilic) Aβ deposits increases with progression of the disease [20]. In contrast to the classic plaques, characterized by highly fi brillar Aβ deposits, the list of Aβ-associated proteins present in diff use plaques is much shorter. Immunohistochemical studies have demonstrated that ApoE, clusterin, complement proteins and ACT are present in diff use plaques [17]. Studies in transgenic mice expressing the causal AD mutations crossed with transgenic mice overexpressing or genetically depleted for amyloid-associated proteins such as complement factors, ApoE, clusterin and ACT have shown that these proteins have an important role in the dynamic balance The occurrence of amyloid-β deposits, glial response and tau-neurofi brillary pathology in the mid-temporal cortex compared to the neuropathological staging of Alzheimer's disease (modifi ed after [5]).
between Aβ deposition and removal [21][22][23][24][25]. Four recent large genome-wide association studies in late-onset AD have documented, in addition to the ApoE ε4 allele genotype, nine novel loci as risk factors for developing AD, includ ing genes encoding proteins implicated in infl ammatory processes (clusterin (CLU, chromosome 8, rs11136000) and complement receptor-1 (CR1, chromosome 1, rs6656401)) [26,27]. Clusterin is a multifunctional protein involved in both complement attack inhibition and cholesterol metabolism, and complement receptor-1 is the receptor for the cleavage products of complement proteins 3 and 4, which are present in plaques. All nine new genes map onto three pathways leading to late-onset AD: immune system function, cholesterol metabolism, and synaptic dysfunction [28]. Diff use plaques, the initial pathological lesion in AD, contain a limited number of proteins, which are now all linked with a genetic risk for AD, with ACT as the exception. Abnormalities in the generation of Aβ are considered as the causal factor for the familial forms of early-onset AD, whereas polymorphisms of the genes encoding for ApoE, clusterin and complement proteins are the genetic risk factors for lateonset AD. Genome-wide association studies for lateonset AD have not revealed genetic factors related to Aβ generation but instead to the Aβ-associated proteins already present in the initial neuropathological lesion. Taken together, the data on the protein composition of diff use plaques and the recent genetic fi ndings strongly suggest an interaction between Aβ, cholesterol metabolism and complement activation in the initial steps of the pathological cascade in AD.

Glia and pro-infl ammatory cytokines
Fibrillar Aβ plaques are associated with clustering of activated microglia and astrocytes. Th ese glial cells play an important role in innate immunity and express a family of receptors, the Toll-like receptors (TLRs), as a fi rst line of defense responsible for recognizing specifi c pathogen-associated molecular patterns like fi brillar Aβ. TLRs and CD14, both innate immunity receptors, mediate activation of transcription factors such as NF-κB and subsequently the production of infl ammatory cytokines [29]. Accumulating evidence indicates that CD14 and TLR4 are essential in the interaction of glial cells with Aβ for the production of pro-infl ammatory cytokines [30][31][32]. In vitro studies have shown that Aβ fi brils interact with the TLR2/4 accessory protein CD14 and both TLR2 and TLR4 mediate Aβ-induced production of TNF-α in human monocytes. Mouse microglia show in vitro increased ingestion of Aβ after activation of TLR2, TLR4 or TLR9. Stimulation of the innate immune system via TLR9 is reported to be highly eff ective at reducing the parenchymal and vascular amyloid burden without apparent toxicity in a transgenic AD mouse model [33].
Both Aβ and the oxidized low-density lipoprotein can trigger infl ammatory signaling through a TLR4 and -6 heterodimer. Th is assemblage is regulated by signals from the scavenger receptor CD36 [34]. Th ese data show that CD36-TLR4-TLR6 activation is a common molecular mechanism by which atherogenic lipids and Aβ stimulate an infl ammatory response in arteriosclerosis and AD, respectively. Arteriosclerosis is a major risk factor for late-onset AD and the discussed fi ndings suggest that both dis orders share a common pathogenic mechanism rooted in the innate immune system. Introducing a functionally destructive mutant of TLR4 into a transgenic mouse model (APPswe/PS1) results in increased levels of Aβ deposits, as well as reduced microglia activity [35]. Microglia from CD14 null mice failed to aff ect Aβl-42 damaged neurons and neuronal survival was accompanied by a signifi cant reduction in the production of IL-6, indicative of reduced microglial activation [36]. All these data suggest that activation of TLR4 and CD14 signaling is involved in both the detrimental production of pro-infl ammatory cytokines as well as the benefi cial removal of Aβ in AD, illustrating the two-edged sword aspect of the neuroinfl ammatory response in AD brains [37]. Experimental animal studies indicate that microglia 'acivation' is not simply one phenotypic manifestation but includes heterogeneous, functional phenotypes that range from a pro-infl ammatory, classic activation state to an alternative activation state involved in repair and extracellular matrix remodeling [38,39].
Lipopolysaccharide (LPS), a bacterial coat component, is widely used as a potent stimulator of the innate immune system and it is recognized by a receptor complex containing fully functional TLR4 and CD14. Chronic neuroinfl ammation induced by LPS in rats reproduced components of the neurobiology of AD, such as increased activation of microglia and astrogliosis, increased tissue levels of IL-1 and TNF-α, elevated expression of the amyloid precursor protein, and a working memory defi cit [40]. Innate immunity responsive ness can be investigated by the incubation of whole blood samples with LPS, followed by the determination of levels of various infl ammatory cytokines. Twin studies demonstrate that heritability for serum levels of circulating infl ammatory mediators is modest (about 20%). In contrast to circulating infl ammatory mediators, however, cytokine production capacity is under strong genetic control. In the non-diseased population, estimates for the heritability of the production of the various cytokines ranges from 53 to 86% [41]. We have studied cytokine production capacity in ex vivo stimulated full blood samples from middle aged off spring with and without a parental history of late-onset AD [42]. We found that the production capacity of the pro-infl ammatory cytokines IL-1β, IL-6, TNFα and interferon-γ was signifi cantly higher in off spring with a parental history of AD upon stimulation of whole blood with LPS. Similar results were found for the IL-1β to IL-1 receptor antagonist (IL-1ra) ratio, which was calculated because IL-1ra is the natural antagonist of the pro-infl ammatory cytokine IL-1β. Th is higher ratio refl ects a pro-infl ammatory genotype. While the ApoE ε4 allele genotype was more frequent among the off spring with compared to those without a parental history of AD, these fi ndings were independent of ApoE4 genotype. In contrast to the stimulated whole blood samples, the evaluation of the unstimulated blood samples does not show signifi cant diff erences between the off spring with versus without a parental history of lateonset AD. Th e aim of this study was not to identify the genetic variability of a particular receptor or cytokine but to investigate the genetic contribution of the whole pathway that mediates the production of pro-infl ammatory cytokines after activation by LPS of the innate immune receptors CD14 and TLRs by LPS. Similar to LPS, the Aβ-induced cytokine production capacity is also under genetic control, as shown by results from twin studies in which whole blood samples were ex vivo stimulated with Aβ [43].
In conclusion, neuropathological and experimental studies indicate that fi brillar Aβ can activate the innate immunity-related CD14 and TLR signaling pathways for pro-infl ammatory cytokine production. Th e production capacity of this pathway is under genetic control and off spring with a parental history of late-onset AD have a higher production capacity for pro-infl ammatory cytokines.

Epidemiological evidence
Prospective case cohort studies have shown that high serum levels of the acute-phase proteins ACT, C-reactive protein and IL-6 could predict cognitive decline or dementia [44][45][46]. Yaff e and colleagues [47] reported that elderly subjects with a metabolic syndrome and high serum level of IL-6 and C-reactive protein were more likely to experience cognitive decline in the next four years, compared with those with a metabolic syndrome and low levels of these infl ammatory proteins. In another population study the metabolic syndrome was also negatively associated with cognition, especially in subjects with high levels of infl ammation [48]. In the Framingham study a higher spontaneous production of IL-1β or TNF-α by peripheral blood mononuclear cells was associated with future risk of AD in older individuals [49]. Th e epidemiological fi ndings from several case cohort studies indicate that non-demented subjects with increased serum levels of acute-phase reactants, indicating a low-grade peripheral systemic infl ammation, are at risk for developing a sporadic late-onset form of AD.
Th e acute phase response is initiated and orchestrated by cytokines, most notably IL-1. AD brains are charac ter ized by overexpression of IL-l and there are strong arguments for an important role for IL-1 in amyloid plaque formation [50]. Aβ deposition and neuronal injury may trigger a self-propagating cytokine cycle, which initiates a vicious feedback of continuing IL-1 elevation when chronically induced. In this way further neuronal dysfunction and Aβ plaque accumulation could be promoted. Taken together, the epidemiological studies suggest that elevated serum levels of acute phase reactants can be considered as a risk factor for AD and neuropathological data demonstrate the presence of acute phase reactants already in human brains with preclinical stages of AD pathology (low Braak score). Studies suggesting that immune blood markers can be used as a clinical test to identify those patients with mild cognitive impairment who progress to clinical AD are consistent with the view that peripheral immune and infl ammatory mechanisms contribute to the pathogenesis of AD [51,52].

Clinical evidence
Th e question of whether systemic infl ammation or peripheral chronic infl ammation could contribute to AD pathology was a neglected research topic until recently. In particular, the dogmatic belief that the blood-brain barrier excludes cross-talk between both systems hampered studies in this fi eld for a long time. Th is view has changed dramatically, however, as it became clear that morphological 'delegates' of the immune system, the microglial cells, are present in the brain and that the peripheral lymphoid organs are innervated. A further fi nding was that cytokines and neurotransmitters, as well as their receptors, are endogenous to both the brain and the immune system. Th ese fi ndings have led to the view that the immune system and brain share a common biochemical language and that their functions are intertwined [53]. Pro-infl ammatory cytokines such as IL-1β and TNF-α, which are generated in the periphery, communicate with the brain. Several mechanisms exist by which an initial, exclusively peripheral cytokine signal can be transmitted to the brain, including direct neural pathways (via primary autonomic aff erents) that transport it across the blood-brain barrier, or entry via the cirvumventricular region, where the blood-brain barrier is non-existent or discontinuous [54].
Several clinical studies suggest that systemic infl ammation can be involved in the pathogenesis of AD. In a twin study it was found that AD cases with a history of severe systemic infection tended to have earlier onset than their corresponding twin [55]. A case-control case study reported a positive association between episodes of infection during the four years preceding the diagnosis and an increased likelihood of a diagnosis of AD in older individuals [56]. In a prospective cohort study of community-dwelling subjects with mild to severe AD it was found that acute episodes of systemic infl ammation with increased serum levels of TNF-α were associated with a two-fold increase in the rate of cognitive decline over a 6-month period. High baseline levels of TNF-α were associated with a four-fold increase in the rate of cognitive decline while subjects who had low levels of TNF-α throughout the study showed no cognitive decline over the 6-month period [57]. Recent studies indicate that chronic periodontitis has been associated with AD. Periodontitis is a prevalent, persisting peripheral infection associated with Gram-negative, anaerobic bacteria that are capable of exhibiting localized and systemic infection in the host. Emerging evidence suggests that tooth loss and periodontal disease predict cognitive decline in community-dwelling older adults [58]. In addition, patients with AD had higher levels of plasma TNF-α and antibodies against periodontal bacteria compared with cognitively normal subjects [59]. Although several case-control and prospective studies indicate periodontitis as a risk factor for cognitive decline, it is important to be cautious to draw any conclusion about a causal relationship. Premorbid cognition is important both for oral health and a risk factor for dementia. Furthermore, factors other than infl ammatory mediators could be responsible for the positive association, such as changes in life style and dietary factors, such as a poor nutritional status, especially in relationship to B vitamins.
Most interesting with regard to understanding the contribution of systemic infl ammation to the patho genesis of AD seems to be current research into the role of infl ammation in delirium. Delirium is defi ned as an acute disturbance of consciousness with signs of attention, a typically fl uctuating course and a change in cognition (that is, disorientation, disturbed memory). Delirium is the most prevalent neuropsychiatric syndrome that can be observed in the general hospital, especially in older patients with preexisting cognitive impairment. It is independently associated with increased mortality, institutionali zation and dementia [60]. Many clinical conditions that are accompanied by systemic infl ammatory reactions can induce delirium. A case-controlled neuropatho logical study has shown an association between severe systemic infection (sepsis) and microglia activation [61]. A recent postmortem study found that delirium is associated with higher immunoreactivity for microglial and astroglial activity and IL-6 compared with age-matched controls without delirium [62]. In healthy persons a severe systemic infl ammation, such as sepsis, can lead to delirium, but in patients with preexisting brain pathology even just a mild urinary tract infection can cause it [63]. Experimental animal studies have shown that microglia respond diff erently to a stimulus if other stimuli precede, coexist or follow it. Microglia can become primed by an initial factor, which prepares them for an enhanced proinfl ammatory cytokine response, even if the subsequent challenges are only mild [64]. Signifi cantly increased serum levels of IL-6 were found in acutely admitted older patients with delirium compared to those without delirium after adjusting for infection, age and cognitive decline [65]. Th is study indicates that acute phase reactants could contribute to the pathogenesis of delirium. A high incidence of delirium is seen in older patients undergoing surgery for hip fracture. Contrary to popular belief, there is little evidence that general anesthesia is associated with delirium after surgery [66]. Th e most important predisposing factor for delirium in these patients is preexisting cognitive decline or dementia. Th e precipitating factors could be related to the release of pro-infl ammatory cytokines as a consequence of the fracture and tissue destruction resulting from surgery. In a time course study of cytokines during delirium in older patients admitted for surgery after hip fracture, significant diff erences in serum levels of IL-6 were found between patients with and without delirium [67].
For a long time it has been a common observation in medicine that, when older patients become delirious while suff ering from an acute urinary tract or other common infection, treatment of the infection may go well but the patients emerge with dementia, even when they had appeared cognitively intact or only mildly impaired prior to hospitalization. Th ese patients often fail to recover to their initial level of functioning and some never resume independent life at home. Similar clinical observations have been made after postoperative delirium in elderly hip fracture patients free from preexisting dementia [68][69][70]. Th ere is now increasing evidence that postoperative delirium after hip surgery is an important predictor of incident dementia in elderly patients living independently at home [70]. In a prospective study it was found that, after a follow-up of 2.5 years, the risk of dementia or mild cognitive impairment is almost doubled in elderly hip surgery patients with postoperative delirium compared with at-risk patients without delirium [68]. It has recently been reported that delirious episodes in a cohort of AD patients accelerate cognitive decline [71].
In conclusion, recent studies suggest that delirium and AD share a neuroinfl ammatory response as a common pathogenic mechanism that could explain the vulnerability of AD patients for further cognitive worsening after an episode of delirium associated with a systemic infl ammatory reaction.

Discussion
Th e etiology of AD may be heterogeneous, but the underlying pathological cascade has distinct common themes. In the autosomal dominant form of familial AD the etiology is related to causal mutations leading to higher production of Aβ1-42. Th e subsequent deposition of fi brillar Aβ elicits a brain infl ammatory response as a secondary event in the pathological process. In contrast to the monocausal etiology of this rare form of AD, the etiology of the highly prevalent sporadic late-onset form is considered to be multifactorial. Genome-wide association studies of late-onset AD strongly suggest a role for lipoproteins and immune-associated proteins in its etiology and pathogenesis. In addition to the role of ApoE and clusterin in the process of Aβ deposition and drainage, these proteins can also attenuate the activity of NF-κB signaling and the production of pro-infl ammatory cytokines [72,73]. Th e fi brillar Aβ-induced infl ammatory response is a relatively early event in the pathological cascade and it is already present in the brain during stages of the disease that precede the stages characterized by tau-related neurofi brillary changes, which are most closely related to a clinical dementia syndrome (Figure 1).
Animal and neuropathological studies in humans show that systemic infl ammation can induce an infl ammatory response within the brain. Th is response can lead to acute cognitive disturbance and behavior changes (delirium) [74]. In healthy adults only a severe systemic infl am mation can induce delirium, but in older people, especially those with mild cognitive impairment or dementia, even just a mild systemic infl ammation can lead to it [75]. Infl ammation-induced delirious episodes in adults with preexisting normal cognitive functions are frequently followed by a period of cognitive impairment that can last for months, although there is no evidence in these cases for a further progression of the cognitive symptoms to a clinical AD syndrome. However, infl ammationinduced delirium in older patients with preexisting mild Figure 2. Relationship between infl ammation and the etiology and clinical syndrome of Alzheimer's disease. Schematic diagram showing that interactions between innate immunity-related genetic risk factors and infl ammation-inducing events (brain trauma, ischemia and infection) can contribute to the multifactorial etiology of the sporadic late-onset form of AD. The diagram illustrates also that delirium and AD share a neuroinfl ammatory response as a common pathogenic mechanism that could explain the vulnerability of AD patients to further cognitive worsening after an episode of delirium associated with a systemic infl ammatory response. Aβ, amyloid-β peptide; AD, Alzheimer's disease; APOE4, apolipoprotein E4; APP, amyloid precursor protein; CLU, clusterin; CR1, complement receptor-1; PS1, presenilin-1; PS2, presenilin-2. cognitive impairment can lead to further cognitive deterioration and dementia ( Figure 2). Th ese clinical fi nd ings suggest that systemic infl ammation does not initiate the pathological AD cascade but can accelerate the underlying cascade.
What makes older patients with peripheral infl ammation so vulnerable to dementia? Several mechanisms can play a role. First, neuropathological studies show infl ammatory changes in early stages of AD pathology, such as fi brillar Aβ-induced microglia activation. Microglia cells are already 'primed' in preclinical stages of AD for increased production of pro-infl ammatory cytokines later on by systemic infl ammation as a second challenge. Second, neurotransmitters such as acetylcholine play an active role in controlling glia activation and can inhibit the production of proinfl ammatory cytokines [76]. AD brains are characterized by cholinergic defi cits that can contribute to an uncontrolled neuroinfl ammatory response when the brain is challenged by peripheral infl ammation. Th ird, arteriosclerosis and other vascular risk factors common in older people lead to blood-brain barrier dysfunction with brain endothelial cell activation resulting in the secretion of multiple neurotoxic and infl ammatory factors. Th ese factors could be responsible for the increased susceptibility of the brain to systemic infl ammatory mediators [77,78]. Th is could also explain why the prevalence of delirium in demented patients in late-onset AD and vascular dementia is higher than in those with early onset AD [79].
Th e pathological, epidemiological and clinical fi ndings reviewed in this paper suggest that several infl ammationrelated events can contribute to the pathogenesis of AD and accelerate the rate of progression of the clinical course of AD. However, this view does not necessarily indicate that treatment with anti-infl ammatory drugs will be eff ective in AD patients. As discussed earlier, infl ammation itself has both benefi cial and detrimental eff ects and every infl ammation-based therapeutic strategy infl uences the delicate balance between both eff ects. Inhibition of complement activation or blocking of IL-1β have been considered as therapeutic options for treatment of AD patients. However, it appeared that complement C3 defi ciency in transgenic AD mice led to accelerated amyloid plaque formation and overexpression of IL-1β to a reduction of amyloid pathology [21,22,80]. Th ese fi ndings indicate potential negative eff ects of anti-infl ammatory drugs in AD patients if these drugs inhibit the primary function of the innate immune system: removal of the pathological agents (fi brillar Aβ) that induce the infl ammatory response. Epidemiological and observational studies in humans have found evidence that the use of non-steroidal anti-infl ammatory drugs (NSAIDs) is associated with a lower risk of developing AD. In contrast, randomized trials have reported no eff ect of NSAIDs on clinical progression in patients with clinically established AD [81]. An explanation for these apparently divergent conclusions could be the fact that expression levels of cyclooxygenase-1 and -2, as the main targets of NSAIDs, are dependent on the stage of AD pathology [82].

Further directions
Th e view that peripheral infl ammation can increase the risk of dementia in older people off ers scope for prevention. In particular, clinical trials are warranted to investigate whether anti-infl ammatory drugs can prevent further cognitive decline in patients with mild cognitive impairment during the periods that peripheral infl ammation exerts increased infl ammatory pressure on the brain. Th ese studies could be performed using available drugs that inhibit microglia activation, such as minocycline, or that restore the cholinergic control of microglia activation by cholinomimetics [63]. Minocy cline, a tetracycline derivative with anti-infl ammatory properties, attenuates the production of pro-infl amma tory cytokines by human microglia without aff ecting their benefi cial activities, such as phagocytosis of Aβ fi brils [83].

Conclusion
Genetic, pathological and epidemiological studies show that innate immunity is involved in the early stages of the pathological cascade of AD and can also contribute to the etiology of late-onset AD. Clinical studies suggest that peripheral infl ammation increases the risk of dementia, especially in patients with preexisting cognitive impairment, and accelerates further deterioration in demented patients.