Beatrix Grubeck-Loebenstein, MD, Institute for Biomedical Aging Research, Austrian Academy of Sciences, Rennweg 10, A-6020 Innsbruck, Austria. Tel.: +43 512583919–0; fax: +43 512583919–8; e-mail: Beatrix.Grubeck-Loebenstein@oeaw.ac.at
A huge amount of evidence has implicated amyloid beta (Aβ) peptides and other derivatives of the amyloid precursor protein (βAPP) as central to the pathogenesis of Alzheimer's disease (AD). It is also widely recognized that age is the most important risk factor for AD and that the innate immune system plays a role in the development of neurodegeneration. Little is known, however, about the molecular mechanisms that underlie age-related changes of innate immunity and how they affect brain pathology. Aging is characteristically accompanied by a shift within innate immunity towards a pro-inflammatory status. Pro-inflammatory mediators such as tumour necrosis factor-α or interleukin-1β can then in combination with interferon-γ be toxic on neurons and affect the metabolism of βAPP such that increased concentrations of amyloidogenic peptides are produced by neuronal cells as well as by astrocytes. A disturbed balance between the production and the degradation of Aβ can trigger chronic inflammatory processes in microglial cells and astrocytes and thus initiate a vicious circle. This leads to a perpetuation of the disease.
Current state of knowledge on amyloid precursor protein, its metabolism and development of Alzheimer's disease
Alzheimer's disease (AD) is the most common dementia form of old age. Currently, approximately 13 million people world-wide suffer from this disorder. AD is the most comprehensively studied neurodegenerative disease and in the last 15 years it has become increasingly clear that the degeneration of neurons is connected with a dysregulation in the metabolism of beta amyloid precursor protein (βAPP) with a consequent transient overproduction or decreased degradation of beta amyloid (Aβ) in the brain.
The pathological hallmarks of AD include profound neuronal loss in the hippocampus, entorhinal and temporoparietal cortex, the presence of intraneuronal neurofibrillary tangles (NFTs), astrogliosis and deposition of Aβ. NFTs are composed of abnormally folded and hyperphosphorylated Tau, a protein involved in microtubule formation. The presence of neurofibrillary pathology (NFT and/or neuritic plaques) is mandatory for the diagnosis of AD (Braak et al., 1993; Jellinger, 1998). Although the pathophysiological association of Aβ and Tau is not completely understood, it seems that in the pathogenesis of AD the dysregulation in the βAPP metabolism precedes the Tau hyperphosphorylation. Genetic studies support this evidence. Mutations in three genes, βAPP, Presenilin 1 (PS1) and Presenilin 2 (PS2), all lead to increased production of Aβ42. A dysregulation of the βAPP metabolism leading to overproduction of Aβ is currently regarded as crucial for the development of AD (Selkoe, 2000). βAPP is a transmembrane glycoprotein belonging to the super-family of amyloid precursor-like proteins, the function of which is (a) adhesion to extracellular matrix proteins and preservation of synaptic plasticity (Coulson et al., 2000), (b) mediation of intracellular sorting mechanisms as a receptor molecule (Kamal et al., 2001) and (c) degradation to cytoplasmic cleavage fragments that might modulate gene expression (Ebinu & Yankner, 2002). βAPP is cleaved by two major metabolic pathways. Secretory βAPP (βAPPs) is cleaved from βAPP by a protease referred to as α-secretase. It participates in the formation of synapses and in the integrity of memory (Sisodia & Gallagher, 1998). Intact Aβ is cleaved from βAPP by the sequential actions of two proteases referred to as β- and γ-secretase (Haass, 2004). Aβ occurs as a 40 and 42 amino acid peptide (Aβ40, Aβ42). Very recent observations suggest that Aβ may play a role in normal synaptic physiology as well as in pathological processes leading to AD, as Aβ depresses synaptic transmission and the activity of neurons in a negative feedback loop (Kamenetz et al., 2003). It is now well understood that an early and invariant step in the pathogenesis of AD is the overproduction and/or decreased degradation of Aβ, which leads to the oligomerization of Aβ in protofibrils and oligomers, which are regarded as pathogenetic hallmark of AD. The development of neuritic plaques precedes the deposition of amyloid in non-fibrillar deposits of Ab42 in the neurophil (diffuse plaques). Although the density of NFTs correlates better with dementia than the frequency of Aβ plaques (Braak et al., 1993; Jellinger, 1998), recent studies suggest that accumulation of Aβ in the brain might precede the development of both neurofibrillary tangles and cognitive impairment (Naslund et al., 2000). Aggregated Aβ is toxic in vitro and in vivo and causes DNA damage and apoptosis (Paradis et al., 1996). Similar to other systemic amyloidoses, masses of Aβ in the extracellular space might also inflict physical or mechanical stress to passing axons and cause diffuse axonal injury leading to neuronal death (Vickers et al., 2000). Decreasing production and accelerating degradation of Aβ peptides seem crucial to avoid or postpone clinically overt AD.
Although the relatively rare familial early onset form of AD is associated with a strongly increased Aβ generation, defective Aβ clearance combined with an increased production is important in late-onset AD (LOAD, age > 70 years). Approximately 90% of all cases are of the LOAD type (Selkoe, 2001). The concept of a defective Aβ clearance in LOAD is also supported by genetic evidence (Myers et al., 2000).
How innate immunity can support the development of AD: a general overview
Aging of adaptive immunity has been profoundly studied in recent years (Grubeck-Loebenstein & Wick, 2002). Considerably less information is available on how aging affects the innate immune system. However, there are clear indications that age-related changes within the innate immune system may influence the development of age-related neurodegenerative disorders such as AD and vice versa that age-related degenerative changes can modulate innate immunity.
Aβ and other proteins found in the senile plaques of AD patients are frequently potent activators of the innate immune response. This is of interest because chronic stimulation of the innate immune system may lead to disadvantages outweighing the advantages initiated by the efforts of an organism to degrade toxic metabolites. Chronically activated microglia and astrocytes can kill adjacent neurons by the release of highly toxic products such as reactive oxygen intermediates, nitric oxide, inflammatory cytokines, proteolytic enzymes, complement factors or excitatory amino acids (for a review see Rogers et al., 2002). Activation of the complement system leads to the production of complement activation fragments, including anaphylatoxins and opsonins (Bradt et al., 1998). The continuous presence of complement activators such as Aβ might vice versa stimulate the chronic state of complement activation and chronic inflammation in the AD brain.
Cytokines such as tumour necrosis factor-α (TNFα), interleukin-1β (IL-1β) and IL-6 can be directly cytotoxic, when chronically produced at high concentrations (Jeohn et al., 1998). They can then also stimulate the synthesis of βAPP (Goldgaber et al., 1989) or may, in combinations such as that of TNFα or IL-1β with interferon-γ (IFN-γ), stimulate the production of Aβ peptides (Blasko et al., 1999). Inflammatory cytokines additionally decrease the secretion of the neuroprotective APPsα (Blasko et al., 1999). Interestingly, βAPP expression is regulated by the same cytokine-sensitive transcription factors that are involved in the expression of most acute phase proteins (Ge & Lahiri, 2002). Several of these proteins such as α1-antichymotrypsin, apolipoprotein E, serum amyloid A, transthyretin, certain proteoglycans and complement factors bind to Aβ and are regarded as pathological chaperones promoting the fibrillar conformation of Aβ (Castillo et al., 1996). The oligomer and fibrillar conformation of Aβ is important for its neurotoxicity as well as its inflammation-inducing properties. It is of interest that glial products such as IL-1β or TNFα can induce the synthesis of most of these Aβ-binding proteins in astrocytes, microglia or neurons (Lieb et al., 1996). Therefore, the chronic release of pro-inflammatory cytokines in the brain is likely to maintain a chronic acute phase protein secretion favouring the formation of Aβ fibrils.
The observation that inflammation is integrally involved in the progression of dementia might have clinical implications. The inducible form of cyclooxygenase, COX-2, is elevated in the brain of early AD, and IL-6 and transforming growth factor-β1 (TGF-β1) were found to be elevated in severely demented AD patients (Luterman et al., 2000; Ho et al., 2001; Pasinetti, 2001). Therefore, future anti-inflammatory therapeutic strategies will possibly have to depend on the clinical stage of the disease. Inflammation in the AD brain is thus most likely the result of amyloid overproduction, but there is abundant evidence that it can be an additional cause of neurodegeneration and could ultimately be as important as the conditions that gave rise to it.
The specific role of microglia
Microglial cells are the most important cells of the innate immune system in the brain. They play the role of cerebral macrophages and recruit and stimulate astrocytes. There is considerable debate as to whether activated microglia are beneficial or harmful. This may, however, depend upon the degree of activation. In recent years it has become obvious that microglial cells can be activated by factors such as brain trauma, ischaemia or neurodegeneration. Activated microglia express markers, such as MHC class I and class II molecules, together with integrins and Fc receptors (Eikelenboom et al., 1994; McGeer & McGeer, 1995). In the AD brain, activation of microglia by Aβ is associated with chemotactic responses to it, consistent with the extensive clustering of activated microglia at sites of Aβ deposition (Terry & Wisniewski, 1975). Opsonization of Aβ oligomers and fibrils with carrier proteins in serum facilitate its endocytosis and degradation (Ard et al., 1996). Upon activation with Aβ and the above-mentioned acute phase proteins, microglial cells produce the pro-inflammatory cytokines IL-1, IL-6 and TNFα, the chemokines IL-8, macrophage inflammatory protein-1 (MIP-1) and monocyte chemoattractant peptide-1 (MCP-1), and the growth factor macrophage-colony stimulating factor (M-CSF) (Lue et al., 2001). The mRNAs for virtually all these proteins, as well as for complement cascade proteins C1qB, C3 and C4, the IL-1 receptor and receptor antagonist, and TGFβ have been observed in AD microglia (Walker et al., 1995; Shen et al., 1997; Strohmeyer et al., 2000). In peripheral monocytes a similar pro-inflammatory response has been demonstrated following stimulation with Aβ (Akiyama et al., 2000; Fiala et al., 1998).
How old age and cellular senescence change the response of microglia to activation-inducing stimuli such as Aβ is not yet known. The first morphological studies on the effect of healthy aging on microglial cells have recently appeared. Microglial cells from cognitively normal elderly donors exhibited several abnormalities within their cytoplasmic structure, including deramification, spheroid formation, gnarling and fragmentation of processes (Streit et al., 2004). These changes were judged to be different from morphological alterations that typically occur during microglial activation in neurodegenerative disorders. They were referred to as microglial dystrophy. Age-related microglial senescence may still easily lead to functional defects, which may then trigger an intracerebral pro-inflammatory response that would support the development of neurodegenerative disorders such as AD in old age.
The specific role of astrocytes
General information on the role of astrocytes
Astrocytes are the most frequent cells of the brain. In the last decade the view of neurobiologists on astrocytes has shifted from the concept of a mainly supporting cell type to a multifunctional housekeeping cell that allows neurons to become progressively specialized for the tasks of information processing. Astrocytes not only organize the structural architecture of the brain but also organize its communication pathways and plasticity. Neurons co-cultured with astrocytes develop approximately seven-fold more synapses, and have a seven-fold increase in synaptic efficacy, compared with neurons raised in the absence of astrocytes (Pfrieger & Barres, 1997). Astrocytes cannot only modulate synaptic strength and activity, but in neurogenic regions of the adult human brain they positively and negatively regulate neurogenesis (Goldman, 2003). Thus a series of recent studies has clearly demonstrated that reciprocal paracrine interactions between astrocytes, endothelial cells and ependymal cells can regulate neurogenesis and gliogenesis from resident precursor cells (Song et al., 2002). At the same time it has become obvious that astrocytes can play an important role in inflammatory processes (Wyss-Coray et al., 2003).
The participation of astrocytes in the neurodegeneration of AD: potential beneficial effects
In the last decade considerable attention has been paid to understanding the role of astrocytes in neurodegenerative disorders like AD. In contrast to microglia, astrocytes are able to remove and to degrade Aβ without mediators or stimuli such as opsonins or cytokines (Bard et al., 2000; Wyss-Coray et al., 2003). Following activation, astrocytes can release cytokines and growth factors similar to those produced by microglia (McGeer & McGeer, 1995), but can also produce trophic substances such as nerve growth factor (NGF) (Aguado et al., 1998), the neurotrophic signalling molecule S100β (Mrak & Griffin, 2001), the brain-derived neurotrophic factor (BDNF) (Hock et al., 2000), neurotrophin 3 (NT-3) (Blondel et al., 2000) and neurotrophin 4/5 (NT-4/5), which have all been shown to have a trophic effect on neurons. Astrocytes regulate extracellular levels of excitatory amino acids such as glutamate and contribute to homeostasis within the central nervous system. Neurotrophic factors, transporter molecules and enzymes involved in the metabolism of excitatory amino acids or in the antioxidant pathway may help to protect neurons and other brain cells from damage by controlling the production of potentially toxic substances. The high flexibility of astrocytes to express different molecules in dependence of age or in dependence of pathological stimuli is documented by the example of experimental AD. Studies on transgenic animals overexpressing human Aβ peptides show that in young animals astrocytes produce small amounts of the Aβ-degrading enzyme neprilysine. However, as the animals get older and the amounts of Aβ-peptides increase, the expression of neprilysine is stimulated (Blondel et al., 2000; Apelt et al., 2003).
The influence of aging on the ability to degrade Aβ-peptides is of interest in this context. In contrast to neonatal mouse and rat astrocytes (Shaffer et al., 1995), adult mouse astrocytes efficiently clear exogenously added and surface-bound Aβ and are capable of removing Aβ deposited in brain slices from mice overexpressing human βAPP (Wyss-Coray et al., 2003). Astrocytosis is a typical morphological feature of the AD brain and represents either proliferation of astrocytes in an effort to replace dying neurons or a reaction to degrade the increasing amounts of toxic Aβ peptides. The chemokine MCP-1 is believed to play an important role in astrocytosis, as its levels were found to be increased after brain injury in contrast to the levels of the pro-inflammatory cytokines TNFα, IL-6 and IL-1β (Little et al., 2002). In the context of aging it is also of interest that adult mouse astrocytes do not respond to stimulation with Aβ by increasing their release of MCP-1 in the same way as young control cells (Wyss-Coray et al., 2003).
The participation of astrocytes in the neurodegeneration of AD
How Aβ affects astrocytes to produce inflammatory products
Because astrocytes greatly outnumber microglia in the brain (Savchenko et al., 2000), these cells could have a more critical role in the development of AD pathology than previously thought. Depending on their degree of aggregation and deposition status, Aβ peptides can activate astrocytes. Thus, reactive IL-1β and IL-6 positive astrocytes were found in close proximity to both fibrillary and diffuse Aβ deposits detectable at very early stages of plaque development in APP SW mice, whereas activated microglia appeared in and around fibrillary Aβ plaques only (Benzing et al., 1999). Reactive astrocytes can produce TGF-β1, TGF-β3 and IL-10 in AD-transgenic mice (Apelt & Schliebs, 2001), but have also been shown to secrete pro-inflammatory mediators such as MCP-1, RANTES, TNF-α and IL-1 in response to stimulation with Aβ42 (Smits et al., 2002). Using primary human astrocytes isolated post mortem from AD patients, we recently demonstrated a high constitutive production of TGFβ-1. However, this potentially anti-inflammatory mechanism could be inhibited by the addition of TNFα (Table 1). This might indicate that pro-inflammatory factors can down-regulate anti-inflammatory responses and neutralize protective mechanisms. A vicious circle could thus be initiated.
Table 1. The in vitro production of TGF-β1 (ng mL−1) by human astrocytes is suppressed following incubation with TNFα (1000 U mL−1 for 24 h). Post mortem astrocytes removed from the grey matter were used. They were purified and cultured, as previously described (De Groot et al., 1997; Veerhuis et al., 1998)
Astrocyte line no.
Brain bank no.
How inflammatory factors might affect the processing of βAPP in astrocytes
Based on the evidence that inflammatory proteins were found in brains of AD patients, several groups were stimulated to investigate the influence of cytokines on the processing of βAPP by astrocytes. βAPP holoprotein synthesis by astrocytes can be stimulated by cytokines such as IL-1β, TNFα or TGF-β1 (Amara et al., 1999; Rogers et al., 1999) through regulation of gene transcription at the promoter level (Ge & Lahiri, 2002). Combinations of cytokines such as IL-1β or TNFα and IFN-γ are able to affect the metabolism of βAPP and to stimulate the production of Aβ40 and Aβ42 peptides (Blasko et al., 1999; Sastre et al., 2003). The mechanism by which cytokines stimulate Aβ production is complex, but it seems likely that the maturation (i.e. proper glycosylation) of the βAPP holoprotein is disturbed (Blasko et al., 1999, 2000). At the same time, β-secretase activity is stimulated (Sastre et al., 2003).
Can astrocyte senescence affect the response to inflammation?
There is presently no definitive answer to the question of why aging is still the biggest risk factor for AD. It seems likely that LOAD is orchestrated by a whole variety of different factors, which are each subtle in influence, but highly detrimental in their combination. LOAD is, for instance, not solely induced by the overproduction of toxic metabolites but may coincide with the decreased capacity of aged glial cells to degrade and/or store toxic products. Little is known about age-related changes in astrocytes (Flanary & Streit, 2004). As astrocytes can be easily passaged in cell culture (Evans et al., 2003), we studied the replicative senescence of primary human astrocytes (Rozovsky et al., 1998; Xie et al., 2003).
We measured the proliferative capacity and population-doublings of post mortem isolated astrocytes obtained from seven donors with AD. The cells were isolated from white matter obtained either from corona radiata or parietal white matter as well as from cortical grey brain matter from the frontal cortex. The patients’ characteristics are listed in Table 2. The cells were obtained from The Netherlands brain bank. Their isolation and culture as well as their phenotype have recently been described (De Groot et al., 1997; Veerhuis et al., 1998; Blasko et al., 2000). In analogy to the well-known ‘in vitro senescence’ models on fibroblasts and endothelial cells (Wagner et al., 2001; Unterluggauer et al., 2003), astrocytes from AD patients proliferated rigorously in culture and underwent 6–8 population doublings (PDs) before proliferation decreased and astrocytes stopped dividing. Growth arrest was reached after 9–10 PDs. Astrocytes are therefore referred to as ‘late passage astrocytes’ after 8 PDs. Throughout the duration of the cultures, astrocytes constantly expressed the glial fibrillary acidic protein (GFAP) monitored by immunoblotting and immunofluorescence (data not shown). The quantification of GFAP cellular content by loading the same amount of cellular protein on gels revealed that late passage astrocytes had a tendency (nonsignificant) to express more GFAP in comparison with early passage astrocytes (Fig. 1A; (Nichols et al., 1993, 1995). In agreement with this observation, a higher GFAP in vivo content was found in aged rats and mice in comparison with their young littermates (O’Callaghan & Miller, 1991; Yoshida et al., 1996; Kyrkanides et al., 2001). Additionally, late passage astrocytes had a significantly increased expression of senescence markers such as p16, p21 and cyclin D1 (Fig. 1A; Stein et al., 1999; Morisaki et al., 1999; Wainwright et al., 2001).
Table 2. Clinical and pathological characteristics of AD patients, from which brain specimens were obtained for isolation of astrocytes. The Braak score represents histopathological classification of AD, in which stages 5–6 represent the destruction of virtually all isocortical association areas (Braak & Braak, 1991). The global deterioration scale (GDS) represents seven stages of cognitive, functional and behavioural impairment in AD (Reisberg et al., 1982). Apolipoprotein E (apoE) ɛ4 allele is a genetic risk factor for late-onset familial and sporadic AD. The presence of at least one ɛ4 allele increases the risk of AD 3–5 times in comparison with non-Apo ɛ4 carriers
Brain bank no.
Braak score (1–6)
GDS stage (1–7)
The grey matter astrocytes proliferated better than their white matter counterparts. The number of PDs reached by grey matter astrocytes in different culture periods was always higher than by white matter astrocytes (Fig. 1B). Whether this increased growth rate depends on in vivo degenerative or inflammatory processes in AD patients is not yet known. To define whether astrocytes died of apoptosis after having reached the end of their replicative lifespan, the number of apoptotic nuclei was analysed at the beginning and the end of the in vitro culture. Figure 1(C) demonstrates that in contrast to fibroblasts (Wagner et al., 2001), but in accordance with endothelial cells (Unterluggauer et al., 2003), astrocytes are prone to undergo apoptosis after they have lost their capacity to replicate.
To analyse whether cellular senescence changes the capacity of astrocytes to respond to cytokines and to produce Aβ we stimulated early as well as late passage astrocytes with TNFα and IFN-γ. Late passage astrocytes were still capable of producing Aβ peptides following stimulation, demonstrating that stimulation of Aβ production by inflammatory products does not decrease with aging (Fig. 1D).
Perspectives and conclusions: evidence that anti-inflammatory agents might protect against neurodegeneration in AD
Based on the epidemiological evidence that nonsteroidal anti-inflammatory drugs (NSAIDs) decrease the risk of AD (Broe et al., 2000; in t’Veld et al., 2001), many anti-inflammatory drugs have been investigated for their potential influence on βAPP processing and the production of Aβ peptides. The common feature of most NSAIDs is their capacity to inhibit COX and consequently to decrease the production of prostaglandins (PGs), in particular PG-E2. Ibuprofen, flurbiprofen, indometacine and sulindac sulphide are, however, able to decrease Aβ42 production by influencing γ-secretase cleavage without affecting COX activity (Weggen et al., 2001; Eriksen et al., 2003). Ibuprofen, as a commonly over-the-counter used NSAID, also decreases cytokine-stimulated Aβ production in human neuronal cells and astrocytes (Blasko et al., 2001). It also reduces neuritic plaque pathology and inflammation in a mouse model of AD (Lim et al., 2000). Recently, ibuprofen at lower concentrations than used in previous studies has been shown to reduce Aβ40 and Aβ42 by decreasing COX-mediated PG-E2 production (Qin et al., 2003). The binding of ibuprofen to the peroxisome proliferator-activated receptor-γ (PPARγ) receptor, which is known to mediate anti-inflammatory activities, is thought to be responsible for this effect (Sastre et al., 2003).
In summary, senescence of the innate immune system can be associated with a pro-inflammatory status of glial cells. This shifted background reactivity of cerebral defence responses in old age may precipitate the perpetual progression of AD. The overproduction or decreased degradation of potentially toxic peptides such as Aβ may just be initial steps in a long cascade of detrimental changes. A better understanding of how innate immunity and chronic inflammation affect neurodegeneration will help to develop new diagnostic and therapeutic concepts.
We thank the brain bank at The Netherlands Institute for Brain Research (coordinator R. Ravid) for providing brain specimens. This work was funded by the European Community (Project No. QLRT-1999-02004 ‘MANAD’) and by the Austrian Science Fund (Project No. P15347).