Molecular mechanisms for Alzheimer's disease: implications for neuroimaging and therapeutics

Authors

  • Colin L. Masters,

    1. Department of Pathology, The University of Melbourne, VIC, Australia
    2. Centre for Neuroscience, The University of Melbourne, VIC, Australia
    Search for more papers by this author
  • Roberto Cappai,

    1. Department of Pathology, The University of Melbourne, VIC, Australia
    2. Centre for Neuroscience, The University of Melbourne, VIC, Australia
    3. The Mental Health Research Institute of Victoria, VIC, Australia
    Search for more papers by this author
  • Kevin J. Barnham,

    1. Department of Pathology, The University of Melbourne, VIC, Australia
    2. The Mental Health Research Institute of Victoria, VIC, Australia
    Search for more papers by this author
  • Victor L. Villemagne

    1. Department of Pathology, The University of Melbourne, VIC, Australia
    2. The Mental Health Research Institute of Victoria, VIC, Australia
    3. Centre for PET, Department of Nuclear Medicine, Austin Health, Melbourne, VIC, Australia
    Search for more papers by this author

Address correspondence and reprint requests to Prof. Colin Masters, Department of Pathology, The University of Melbourne, Victoria, 3010, Australia. E-mail: c.masters@unimelb.edu.au

Abstract

Alzheimer's disease is a progressive neurodegenerative disorder characterised by the gradual onset of dementia. The pathological hallmarks of the disease are β-amyloid (Aβ) plaques, neurofibrillary tangles, synaptic loss and reactive gliosis. The current therapeutic effort is directed towards developing drugs that reduce Aβ burden or toxicity by inhibiting secretase cleavage, Aβ aggregation, Aβ toxicity, Aβ metal interactions or by promoting Aβ clearance. A number of clinical trials are currently in progress based on these different therapeutic strategies and they should indicate which, if any, of these approaches will be efficacious. Current diagnosis of Alzheimer's disease is made by clinical, neuropsychologic and neuroimaging assessments. Routine structural neuroimaging evaluation with computed tomography and magnetic resonance imaging is based on non-specific features such as atrophy, a late feature in the progression of the disease, hence the crucial importance of developing new approaches for early and specific recognition at the prodromal stages of Alzheimer's disease. Functional neuroimaging techniques such as functional magnetic resonance imaging, magnetic resonance spectroscopy, positron emission tomography and single photon emission computed tomography, possibly in conjunction with other related Aβ biomarkers in plasma and CSF, could prove to be valuable in the differential diagnosis of Alzheimer's disease, as well as in assessing prognosis. With the advent of new therapeutic strategies there is increasing interest in the development of magnetic resonance imaging contrast agents and positron emission tomography and single photon emission computed tomography radioligands that will permit the assessment of Aβ burden in vivo.

Abbreviations used

β-amyloid

APP

amyloid precursor protein

FDG

fluorodeoxyglucose

MPAC

metal–protein attenuated compounds

NFT

neurofibrillary tangles

PS

presenilin

ROS

reactive oxygen species

Alzheimer's disease, the leading cause of dementia in the elderly, is an irreversible, progressive neurodegenerative disorder clinically characterised by memory loss and cognitive decline, leading invariably to death, usually within 7–10 years after diagnosis.

Age is the dominant risk factor in Alzheimer's disease. The progressive nature of neurodegeneration suggests an age-dependent process that ultimately leads to synaptic failure and neuronal damage (Masters and Beyreuther 1998) in cortical areas of the brain essential for memory and higher mental functions. The increase in the number of new cases of Alzheimer's disease is the direct consequence of an improvement in life expectancy. Despite the tremendous corpus of knowledge of genetics, epidemiology, risk factors and neuropathological mechanisms, there is still no cure for Alzheimer's disease.

Clinical features and diagnostic criteria

At present, clinical diagnosis of Alzheimer's disease is based on progressive impairment of memory and decline in at least one other cognitive domain, and exclusion of other diseases that might also present with dementia, such as frontotemporal dementia, dementia with Lewy-bodies, stroke, brain tumour, normal pressure hydrocephalus or depression.

The clinical diagnostic accuracy for Alzheimer's disease depends on the stage of disease and can exceed 90% in academic settings in mid or late stages. Diagnostic criteria for Alzheimer's disease have been proposed within both the DSM and ICD classification systems. However, the criteria followed in most research studies are those proposed by the National Institute of Neurological and Communicative Disorders and Stroke – Alzheimer's Disease and Related Disorders Association (NINCDS-ARDA) for Defining Probable Alzheimer's disease (McKhann et al. 1984).

A variable period of up to 5 years of prodromal decline in cognition characterised by a relatively isolated impairment in long-term memory that may also be accompanied by impairments of working memory, known as mild cognitive impairment, usually precedes the formal diagnosis of Alzheimer's disease (Petersen et al. 1999) These deficits presumably relate to damage to the medial temporal lobe and/or specific prefrontal-temporal lobe circuits. About 40–60% of carefully characterised subjects with mild cognitive impairment will subsequently progress to meet criteria for Alzheimer's disease over a 3–4-year period (Petersen et al. 1999).

Neuropathology of Alzheimer's disease

In the absence of biologic markers, direct pathologic examination of brain tissue derived from either biopsy or autopsy remains the only definitive method for establishing a diagnosis of Alzheimer's disease (Selkoe 2001; Masters and Beyreuther 2005). The typical macroscopic picture is gross cortical atrophy. Microscopically, there is widespread cellular degeneration and neuronal loss that affects primarily the outer three layers of the cerebral cortex, initially affecting more the temporal and frontal cortical regions subserving cognition than the parietal and occipital cortices. These changes are accompanied by reactive gliosis, diffuse synaptic and neuronal loss, and by the presence of the pathological hallmarks of the disease, intracellular neurofibrillary tangles (neurofibrillary tangles) and extracellular amyloid plaques (Selkoe 2001; Masters and Beyreuther 2005).

Neurofibrillary tangles are intraneuronal bundles of paired helical filaments. The main structural component of neurofibrillary tangles is a normal constituent of cellular microtubules but when present in Alzheimer's disease is an abnormally phosphorylated form, known as tau protein. They are most easily identified in the hippocampus. Neurofibrillary tangles are not specific to Alzheimer's disease, and are found in a variety of other neurodegenerative conditions such as frontotemporal dementia, subacute sclerosing panencephalitis, Hallervorden-Spatz disease, Parkinson dementia complex and dementia pugilistica (Perl 2000). Tau is a widely expressed phosphoprotein from the microtubule associated family, the main function of which is to maintain microtubule stability. In Alzheimer's disease, hyperphosphorylated tau aggregates, reducing its ability to bind microtubules and leading to cytoskeletal degeneration and neuronal death (Lovestone and Reynolds 1997).

The plaques consist of extracellular aggregates of amyloid β-peptide (Aβ). Aβ is a 4-kDa self-aggregating, 39–43-amino acid metalloprotein product derived from the proteolytic cleavage of the amyloid precursor protein by β-and γ-secretases (Cappai and White 1999; Selkoe 2001) (Fig. 1). The plaques are intimately surrounded by dystrophic axons and dendrites, reactive astrocytes and activated microglia (Masters and Beyreuther 2005). Aβ is not only found within senile plaques, but is also present around cortical arterioles as a congophilic angiopathy. It can also be assessed in CSF, plasma and even neuronal cultures (Seubert et al. 1992). A number of in vitro and in vivo studies have shown Aβ protein to be directly toxic to neurons, leading to the aggregation and secondary phosphorylation of the tau protein.

Figure 1.

 Schematic diagram of amyloidogenic and nonamyloidogenic proteolytic pathways of amyloid precursor protein and production of Aβ. Amyloid precursor protein is cleaved by either α-secretase (α-sec) or β-secretase (β-sec) yielding α-amyloid precursor proteins or β-amyloid precursor proteins, respectively. The C-terminal C83 fragment produced by α-secretase, and the C-terminal C99 fragment produced by β-secretase, are then further cleaved by γ-secretase (γ-sec) into P3 or Aβ40/42, respectively. APP, amyloid precursor proteins; SP, signal peptide; HBD-1, heparin binding domain-1; CuBD-1, copper binding domain-1; ZnBD-1, zinc binding domain-1; Zn-CuBD-2, zinc copper binding domain-2; NH2, N-terminal end; COOH: C-terminal end.

Aβ was first identified and sequenced from meningeal blood vessels of Alzheimer's disease and Down's syndrome patients 20 years ago (Glenner and Wong 1984; Masters et al. 1985). The aggregation process that converts soluble Aβ into amyloid fibrils is thought to be a nucleation-dependent process (Harper and Lansbury 1997) requiring structural transitions of Aβ.

On electron microscopy, amyloid fibrils are composed of multiple protofibrils wrapped around each other, forming a crossed β-pleated sheet.

Aβ centric theory of Alzheimer's disease

Through the years, several theories have been postulated to explain the molecular mechanisms leading to Alzheimer's disease (Masters and Beyreuther 1998; Selkoe 2001). The Aβ centric theory is the dominant etiologic paradigm at this time, because it is the only one that best or most comprehensively articulates the current available knowledge regarding the cellular, molecular and functional alterations observed in Alzheimer's disease. Not only is there a wealth of histopathological, biochemical, genetic and animal model data that support the key role of Aβ in the pathogenesis of Alzheimer's disease, but no alternative hypothesis has emerged in the last two decades of intensive Alzheimer's disease research. The Aβ centric theory states that an imbalance between the production and removal of Aβ leads to its progressive accumulation, triggering a series of reactions leading to synaptic dysfunction, microgliosis and neuronal loss, clinically manifested with memory loss and impaired cognitive functions (Selkoe 2001) (Fig. 2). The loss of synaptic function seems to be the critical factor in cognitive decline (Selkoe 2002). Much of the controversy derives from the use of the term amyloid. The broad term can be applied not only to Aβ, but to several unrelated extracellular deposits of fibrillar protein, such as β2-microglobulin, amylin or serum amyloid A, each one of them associated with a specific disease.

Figure 2.

 Schematics showing the role of Aβ in Alzheimer's disease pathogenesis along with traditional and novel therapeutic strategies. Increased production or reduced clearance of Aβ leads to aggregation, deposition and neuronal injury through a variety of neurotoxic mechanisms, such as generation of oxygen and nitrogen radicals (H2O2, OH·, NO), transition metal ion interactions, excitotoxicity, tau hyperphosphorylation into neurofibrillary tangles, inflammatory response via microglia and astrocytic activation leading to synaptic deficits and cell death. The therapeutic interventions are bolded in grey and boxed and the dotted arrows indicate the target(s). AD, Alzheimer's disease; APP, amyloid precursor preoteins; MPAC, metal-protein attenuating compound.

The earliest structural, microscopically visible pathological changes in Alzheimer's disease are diffuse Aβ deposits. These deposits are also observed in normal ageing individuals, but the density is lower than in Alzheimer's disease patients (Perry et al. 1978) indicative of an immature or not yet toxic form of Aβ. The presence of extracellular Aβ in highly specialised cortical brain regions implicated in memory and cognition precede the other pathognomonic pathological features of Alzheimer's disease, indicating that increases in Aβ are involved in the early presymptomatic stages of the disease. As the earliest phenotypical marker of disease, this has crucial implications for neuroimaging and treatment. The increase in Aβ deposition is accompanied by decreases in Aβ in CSF (Sunderland et al. 2003).

Though extracellular amyloid plaques are the hallmark brain lesions of sporadic Alzheimer's disease, the distribution and density of both diffuse and Aβ plaques at the light microscopic level (McLean et al. 1999) have not been consistently shown to correlate with the degree of cognitive impairment. The best correlation occurs with soluble levels of Aβ, measured biochemically (McLean et al. 1999). Soluble Aβ is in equilibrium with insoluble Aβ in the plaques. The significance of the aggregated amyloid plaques can be interpreted as they either are a reservoir for the soluble oligomers, or represent the sequestered pool of soluble and now precipitated Aβ, therefore fulfilling a ‘protective’ function, or just the end stage or final product of the Aβ cascade.

One of the criticisms raised against the amyloid hypothesis has come from some of the interpretations of the work of Braak and Braak (1991), who stated that neurofibrillary degeneration of cell bodies and their neurites not only predate morphologically detectable amyloid plaques but that they also increase gradually with age. However, as Hardy and Selkoe (2002) point out, the postmortem cases used to establish the Braak Stage I neuropathology criteria were non-demented older individuals, in whom it is impossible to distinguish whether their neurofibrillary changes represent early stages of Alzheimer's disease or a different process altogether (Price and Morris 1999). It has been well established in patients with Down's syndrome that Aβ deposition predates the formation of neurofibrillary tangles (Lemere et al. 1996).

The Aβ theory is strongly supported by compelling genetic data (Hardy and Selkoe 2002). Though it is highly probable than additional genes are associated with Alzheimer's disease, to date only three different genes, all associated with Aβ production, are implicated in the pathophysiology of Alzheimer's disease, and have been described in patients with the rare early onset familial Alzheimer's disease (St George-Hyslop 2000), mutations of the amyloid precursor protein gene (Citron et al. 1992; Rovelet-Lecrux et al. 2006) on chromosome 21, mutations in the presenilin 1 and 2 genes on chromosome 14 and 1, respectively (Rogaev et al. 1995; Miklossy et al. 2003). Presenilin 1, presenilin 2 and amyloid precursor protein have a clear-cut autosomal dominant pattern with a penetrance above 85%. Moreover, a polymorphism in the apolipoprotein E gene on chromosome 19 (Strittmatter et al. 1993) is the most prevalent of these risk factors for Alzheimer's disease and acts as a weaker susceptibility factor. The main feature of the mutations in amyloid precursor protein, presenilin 1 and 2 is their involvement in the different steps of amyloid precursor protein processing pathway, which leads to increased production and elevated plasma levels of Aβ, specially Aβ42 (Scheuner et al. 1996). These various genetic mutations, all manifesting as a similar clinical entity and all leading to increased levels of Aβ, and Aβ build-up in the brain before Alzheimer's disease symptoms arise, further support the Aβ theory of Alzheimer's disease.

Amyloid precursor protein

The amyloid precursor protein gene was cloned following the purification and sequencing of the Aβ peptide from Alzheimer's disease brains and is localised to chromosome 21 (Kang et al. 1987), the chromosome involved in Down's syndrome, a condition that invariably develops the typical Alzheimer's disease neuropathology by age 50, though they start developing amyloid plaques as early as age 12, long before they get neurofibrillary tangles and other Alzheimer's disease lesions (Lemere et al. 1996).

Amyloid precursor protein is a 695–770-residue ubiquitously expressed glycosylated transmembrane protein with a large hydrophilic aminoterminal extracellular domain, a single hydrophobic transmembrane domain consisting of 23 residues and a small carboxy-terminal cytoplasmic domain (Kang et al. 1987). Structural studies support the model that the extracellular domain is composed of a multidomain structure of seven defined regions. The N-terminal cysteine-rich domain is composed of two domains, the growth factor domain (Rossjohn et al. 1999) and the copper binding domain (Multhaup et al. 1996; Barnham et al. 2003a). The growth factor domain and copper binding domains are joined by a short hinge sequence. The crystal structure of the growth factor domain has a large positively charged electrostatic patch on the surface that would allow heparin binding, consistent with its heparin binding activity (Small et al. 1994). The NMR structure of the copper binding domain showed the copper binding site is surface exposed, unlike other copper binding proteins. It has homology to copper chaperones, in agreement with amyloid precursor protein being a modulator of copper homeostasis (White et al. 1999). The cysteine-rich region is followed by an acidic region and no tertiary structural data has been reported. The alternativley spliced Kunitz-protease inhibitor and OX-2 domains follow the acidic domain. The Kunitz-protease inhibitor domain has high sequence and structural homology to other Kunitz inhibitor proteins (Hynes et al. 1990). In vitro and in vivo studies have indicated amyloid precursor protein can modulate hemostasis processes in both a Kunitz-protease inhibitor and a non-Kunitz-protease inhibitor manner (Mahdi et al. 1995; Henry et al. 1998; Xu et al. 2006). The role of the OX-2 domain remains unclear. The carbohydrate domain follows either the acidic domain in the APP695 isoform or the Kunitz-protease inhibitor domain in APP751 or OX-2 domain in APP770, respectively. The carbohydrate domain undergoes both N- and O-glycosylation (Weidemann et al. 1989) and appears to be composed of at least two separate regions. The most proximal domain has been termed CAPPD (Dulubova et al. 2004), E2 (Wang and Ha 2004) or D6a (Andersen et al. 2006). This is a structured region composed of tightly packed helices that can interact to form an antiparallel dimer (Wang and Ha 2004). The CAPPD/E2/D6a sequence is highly conserved amongst the amyloid precursor protein-family orthologues and paralogues, suggesting an important function. The D6a domain binds to the neuronal sorting protein sorLA and the interaction modulates amyloid precursor protein processing into Aβ (Scherzer et al. 2004; Andersen et al. 2005). The distal portion of the carbohydrate domain, which contains the α- and β-secretase cleavage sites, appears to be unstructured and is susceptible to proteolytic degradation (Dulubova et al. 2004). The amyloid precursor protein molecule is tethered to the membrane via a single transmembrane domain which contains the hydrophobic C-terminal portion of the Aβ peptide. The cytoplasmic domain is short (51aa) and undergoes phosphorylation and binds to a number of adaptor molecules, including Fe65, X11 and Numb (Kerr and Small 2005). The cytoplasmic domain is released following γ-secretase cleavage and may be transported to the nucleus (Kerr and Small 2005).

The majority of amyloid precursor protein is degraded in the endoplasmic reticulum and only a small fraction enters the secretase cleavage pathway (Selkoe 2001). While amyloid precursor protein is usually proteolytically cleaved by β-secretase, mutations on the amyloid precursor protein gene were shown to be associated with increased Aβ self-aggregation, and Aβ production by the sequential cleavage by β- and γ-secretases (Citron et al. 1992).

The free N-terminus of Aβ, considered the first critical step in amyloid formation (Selkoe 2001), is derived from the amyloid precursor protein by proteolytic cleavage by β- secretase. Several lines of evidence demonstrate that β-secretase cleavage of amyloid precursor protein is required for Aβ generation (Seubert et al. 1993). Generation of the N-terminus is followed by C-terminal cleavage by γ-secretase to release the final Aβ-product from the β-secretase cleavage fragment C99. Cleavage by γ-secretase occurs within the transmembrane region of amyloid precursor protein yielding mainly 40- and 42-amino acid Aβ C-terminal variants, Aβ40 and Aβ42 (Fig. 1).

Amyloid precursor protein can also undergo nonamyloidogenic processing by α-secretase, which cleaves amyloid precursor protein within the Aβ domain to generate α-amyloid precursor proteins (the ectodomain of amyloid precursor protein ending at the α-secretase cleavage site) (Mudher and Lovestone 2002) and C83 (the C-terminal tail of amyloid precursor protein), which can then undergo γ-secretase cleavage leading to the release of p3 (Fig. 1), a shortened, probably non-pathogenic, form of Aβ (Scheuner et al. 1996).

Although the function of amyloid precursor protein is unknown, a significant body of evidence suggests it functions in maintaining cellular Cu homeostasis (Barnham et al. 2004a) by possibly delivering Cu and maybe Fe to metallo-enzymes and proteins, such as superoxide dismutase 1, and the Cu ATPase. Amyloid precursor protein knockout mice have increased Cu levels in both brain and liver (White et al. 1999), whilst overexpression of the Aβ containing carboxyl-terminal fragment of amyloid precursor protein in transgenic mouse models results in significantly reduced brain Cu not Fe levels (Maynard et al. 2002). Cu can modulate in vivo amyloid precursor protein processing (Bayer et al. 2003; Phinney et al. 2003) with higher Cu levels resulting in a reduction in Aβ production and a consequential increase in the nonamyloidogenic p3 form of the peptide (Borchardt et al. 1999). Independent Cu-binding sites have been identified on both Aβ and the amyloid precursor protein ectodomain. The Cu-binding domain of amyloid precursor protein shows some structural homology to copper chaperones. It contains a tetrahedral binding site consisting of two histidine residues (positions 147, 151), a tyrosine (position 168) and methionine (position 170) that favours Cu(I) co-ordination (Barnham et al. 2003a). The Aβ Cu-binding site is located near the N-terminal part of the peptide and consists of histidines 6, 13, 14 and an as yet unidentified fourth ligand (Curtain et al. 2001; Smith et al. 2006).

Presenilins

There is significant genetic evidence coming from mutations in the presenilin 1 and presenilin 2 genes (Rogaev et al. 1995; Miklossy et al. 2003) that the presenilin proteins affect γ-secretase activity (De Strooper et al. 1998). The majority of early onset familial Alzheimer's disease cases are linked to mutations within the presenilin genes. More than 40 mutations have been described in the gene for presenilin 1 that can subsequently result in Alzheimer's disease. Mutations in both genes selectively increase the production of Aβ42 in cultured cells and in the brains of transgenic mice and are associated with early onset familial Alzheimer's disease (Selkoe 2001). Some presenilin mutations associated with increases in Aβ metabolism instead of presenting Alzheimer's disease symptoms show large plaques and special symptoms such as spastic paraparesis (Smith et al. 2001). Analysis of a range of presenilin 1 familial Alzheimer's disease mutations on different γ-secretase substrates (amyloid precursor protein, Notch, N-cadherin and N-syndecan) suggested they affected presenilin- associated γ-secretase activity in diverse ways (Bentahir et al. 2006).

Presenilin 1 and 2 are ubiquitously expressed within the brain, primarily in neurons (Rogaev et al. 1995; Sherrington et al. 1995). The proteins contain multiple transmembrane domains, with an amino and carboxy terminus as well as a large hydrophilic loop located in the cytoplasm (Rogaev et al. 1995; Sherrington et al. 1995). Both proteins, the 46-kDa presenilin 1 and 55-kDa presenilin 2, share 67% amino acid identity. The exact functions associated with presenilin protein have not been fully elucidated yet. Presenilin 1 is involved in normal neurogenesis and formation of the axial skeleton, as well as in γ-secretase activity. Gene deletion of presenilin 1 shows that it is indispensable for the generation of Aβ (De Strooper et al. 1998). Two transmembrane aspartate residues in presenilin 1 are essential for Aβ production, indicating that presenilin 1 is either an essential cofactor for γ-secretase or maybe γ-secretase itself (Kimberly et al. 2000). Presenilin 2 also contains the two transmembrane aspartate residues which appear to be critical for γ-secretase activity (Kimberly et al. 2000).

Apolipoprotein E

Genetic variability in Aβ catabolism and clearance increase the risk for late-onset Alzheimer's disease (Bertram et al. 2000; Olson et al. 2001). In contrast to the rare, early onset autosomal dominant forms, the only consistent marker for both the early onset familial and late-onset non-familial form of dementia is the polymorphism of apolipoprotein E allele on chromosome 19 (Saunders et al. 1993). Encoded on the long arm of chromosome 19, apolipoprotein E is a 34-kDa lipid transport protein considered the major genetic risk factor in the pathogenesis of Alzheimer's disease (Marques and Crutcher 2003). Apolipoprotein E is normally present in oligodendroglia, astrocytes and microglia. A lipid carrier protein involved in the transport of cholesterol and phospholipids, apolipoprotein E is believed to play an important role in synaptic plasticity and neuronal repair mechanisms. It protects neuronal-glial cells cultures against H2O2 oxidative injury by reducing secondary glutamate excitotoxicity in vitro (Lee et al. 2004) and is both directly and indirectly involved in oxidative mechanisms in the brain (Ramassamy et al. 2000). Apolipoprotein E interacts directly with Aβ and with amyloid precursor protein through the carboxy-terminal domain of apolipoprotein E. The association of apolipoprotein E and Aβ inhibits fibril formation (Beffert and Poirier 1998) and also attenuates glial activation by Aβ (Hu et al. 1998). Apolipoprotein E exists in three allelic variants: ε2 (8%), ε3 (77%) and ε4 (15%). The presence of the apolipoprotein E4 allele increases fourfold the risk of Alzheimer's disease and much more if the allelic variant is inherited from both parents. The ε4 allele is absent in approximately 30–40% of patients with Alzheimer's disease and present in about 30% of healthy subjects (Mayeux et al. 1998) as well as in patients with Down's syndrome (St George-Hyslop et al. 1994). In carriers of apolipoprotein E4 allele, Aβ deposition responsible for the congophilic angiopathy (Kalaria 2002) could play an important role in contributing to the chronic cortical hypoperfusion typically observed in neuroimaging studies of patients with Alzheimer's disease (Villemagne et al. 2005a). Whereas the ε4 allele is associated with an increased risk for Alzheimer's disease, the ε2 allele is believed to represent no increased or decreased risk and the ε3 allele may even confer some protection against Aβ-induced toxicity (Corder et al. 1993) through its anti-oxidant and membrane-stabilising properties and via complexation and internalisation of Aβ through apolipoprotein E receptors (Jordan et al. 1998). Furthermore, apolipoprotein E is a metal chelator, and the ε4 allele variant binds more rapidly to Aβ while at the same time displaying the weakest chelator affinity (Moir et al. 1999).

Transgenic mice models

Transgenic mice models of Alzheimer's disease with mutations in amyloid precursor protein and presenilin genes lead to increase production and progressive aggregation of Aβ, reproducing the major features of Alzheimer's disease: Aβ plaques, associated with neuronal and microglial damage (Games et al. 1995; Hsiao et al. 1996). The difference in tau sequence between mouse and humans may explain why despite the progressive Aβ deposition there are no neurofibrillary tangles and very little neuronal loss (Games et al. 1995; Hsiao et al. 1996; Irizarry et al. 1997). Other reasons to be considered are species differences in neuronal vulnerability, the relatively short duration of exposure to Aβ, and the lack of certain cytokines necessary for a full complement inflammatory response.

Mutations in tau protein leading to large tau deposits in intracellular neurofibrillary tangles are not associated with amyloid deposits, and are clinically manifested as frontotemporal dementia with parkinsonism (Hutton et al. 1998; Spillantini et al. 1998), indicating that the neurofibrillary tangles in Alzheimer's disease are secondary to Aβ production (Hardy et al. 1998) and probably triggered by Aβ (Rapoport et al. 2002).

Although the density of neurofibrillary tangles correlates better than Aβ aggregates with the degree of dementia (Terry et al. 1994), in patients with the rare presenilin 1 mutations or individuals with Down's syndrome who died prematurely from other diseases, the presence of Aβ (either as diffuse deposits or typical plaques) precedes the appearance of neurofibrillary tangles (Lippa et al. 1998). This has been established experimentally with transgenic mice overexpressing both mutant human tau and mutant human amyloid precursor protein which have the same number and structure of amyloid plaques but a significantly higher number of tau-positive neurofibrillary tangles than transgenic mice overexpressing only mutant human tau (Lewis et al. 2001), indicating that the mutant amyloid precursor protein and the consequent Aβ production precede and promote the formation of neurofibrillary tangles (Götz et al. 2001). Apolipoprotein E-deficient mice crossed with amyloid precursor protein transgenic mice showed a significant reduction in Aβ deposition (Bales et al. 1997), supporting the role played by apolipoprotein E in the metabolism of Aβ.

Mechanism of Aβ toxicity

Because of its high lipid content and high oxygen consumption, the brain is particularly susceptible to oxidative stress. Several mechanisms have been proposed to explain Aβ neurotoxicity: production of reactive oxygen species such as hydrogen peroxide, nitric oxide, superoxide, highly reactive hydroxyl radicals and nitric oxide (NO), excitotoxicity with intracellular calcium accumulation, decreased membrane fluidity, energy depletion, alteration of the cytoskeleton, inflammatory processes and alteration of metal homoestasis (Hardy and Selkoe 2002; Bush 2003; Barnham et al. 2004a). All of these events converge into similar pathways of synaptic disruption, necrosis or apoptosis, leading to progressive loss of specific neuronal cell populations.

A common factor in the postulated mechanisms of Aβ toxicity is the oligomerisation of Aβ, whether as fully formed fibrils, dimers or trimers (Roher et al. 1996; Walsh et al. 2002) or as protofibrils (Harper et al. 1999). Despite several attempts, the main obstacle to the full validation of the Aβ theory remains the identification in vivo of the specific neurotoxic Aβ soluble oligomer. There is an inverse relationship between amyloid burden and oxidative damage in vivo as assessed by 8-OH guanosine levels in Alzheimer's disease-affected tissue. Several lines of evidence demonstrate that diffusible soluble Aβ oligomers, but not monomers or insoluble amyloid fibrils, are toxic to cultured neurons and responsible for the neurotoxicity and synaptic dysfunction present in Alzheimer's disease. Micro-injection into rats of culture medium containing soluble oligomers of human Aβ (in the absence of monomers and amyloid fibrils) inhibits long-term potentiation in the hippocampus (Walsh et al. 2002). Changes are observed in young amyloid precursor protein transgenic mice before plaque formation (Hsia et al. 1999; Mucke et al. 2000), though the diversity and unstable nature of Aβ intermediates, from monomers to mature fibrils, makes it difficult to identify the specific species responsible for the neurotoxic effects.

Generation of reactive oxygen species

Extra- and intracellular production of reactive oxygen species initiates and promotes neurodegeneration in Alzheimer's disease (Schippling et al. 2000). Evidence of oxidative stress in Alzheimer's disease is manifested through higher levels of oxidised proteins (Schippling et al. 2000), advanced glycation (Smith et al. 1995), lipid peroxidation products (Ramassamy et al. 2001), formation of toxic species, such as peroxides, alcohols, aldehydes, ketones, cholesterol oxide – toxic to microglial cells (Chang et al. 1998), cholestenone (Bernheimer et al. 1987), altered gene expression (Allen and Tresini 2000) and damaged DNA (Dizdaroglu 1992). Aβ induces lipoperoxidation of membranes and lipid peroxidation products (Mark et al. 1999). Lipids are modified by reactive oxygen species and there is a high correlation between lipid peroxides, anti-oxidant enzymes, amyloid plaques and neurofibrillary tangles in Alzheimer's disease brain (Lovell et al. 1995). Markers of oxidative DNA damage have been localised in plaques and neurofibrillary tangles (Mecocci et al. 1994; Good et al. 1996).

Several breakdown products of oxidative stress including 4-hydroxy-2,3-nonenal (HNE) (Selley et al. 2002) acrolein, malondialdehyde and F2-isoprostanes have been observed in Alzheimer's disease brains when compared to age-matched controls (Arlt et al. 2002). HNE modifies proteins, resulting in a multitude of effects, including inhibition of neuronal glucose and glutamate transporters (Keller et al. 1997) and Na-K ATPases (Mark et al. 1995) plus activation of kinases and dysregulation of intracellular calcium signaling that ultimately induce an apoptotic cascade (Mattson and Chan 2003; Tamagno et al. 2003).

Catalase, superoxide dismutase, glutathione peroxidase and glutathione reductase, indicators of cellular defence mechanisms against oxidative stress, are increased in the hippocampus and amygdala of Alzheimer's disease patients (Pappolla et al. 1992). DNA bases are vulnerable to oxidative stress damage involving hydroxylation (Gabbita et al. 1998), protein carbonylation and nitration. Reactive oxygen species-induced calcium influx, via activation of glutamate receptors, triggering an excitotoxic response leading to cell death, has also been observed in Alzheimer's disease brains (Yamamoto et al. 1998; Mattson and Chan 2003). Reactive oxygen species are generated when oxygen reacts with unregulated redox-active metals. Metalloproteins such as Aβ in Alzheimer's disease that might abnormally present Cu or Fe for inappropriate reaction with O2 are implicated in several age-dependent neurodegenerative disorders (Barnham et al. 2004a).

Generation of reactive nitrogen species

NO-induced neurotoxicity has been extensively studied. NO is synthesised by NO synthases (NOS), and the three isoforms of NOS – endothelial (eNOS), neuronal (nNOS) and inducible (iNOS) – are present in the brain (Law et al. 2001). NO synthesis is activated by glutamate release accompanied by excess Ca2+ influx through activation of the NMDA (Parks et al. 2001) and AMPA receptors (Blanchard et al. 2004). Aβ induces NO production by interacting with glial cells or by disrupting Ca2+ homeostasis through NMDA receptor (Parks et al. 2001). NO combines with superoxide anion forming peroxynitrite, and the resultant reactive nitrogen species can induce significant oxidative stress leading to lipid peroxidation, damaged DNA, and neuronal death.

Activation of inflammatory processes

Aβ oligomers/fibrils are toxic for cultured neurons and activate microglia. Blocking Aβ oligomers/fibrils formation prevents this toxicity (Meda et al. 1995). Astrocytes and microglial cells are involved in the chronic inflammatory responses in Alzheimer's disease through the up-regulated expression of phospholipase A2, leading to increased arachidonic acid/prostaglandin inflammatory pathway activity by secreting interleukin-1 (Griffin et al. 1989) and activation of complement pathways (Rogers et al. 1992), and producing a variety of potentially neurotoxic compounds, including superoxides, glutamate and NO (Brown and Bal-Price 2003).

Altered energy metabolism

Intermediate metabolism is essential to maintain signaling activities and depends on mitochondrial function. Disturbed energy metabolism and the appearance of degenerating mitochondria in axonal terminals is an early feature of Alzheimer's disease (Byrne 2002).

Reactive oxygen species production, Ca2+ uptake and mitochondrial membrane depolarisation have been linked to neuronal apoptosis (Kruman and Mattson 1999) by disrupting the normal mitochondrial functioning, through the uncoupling of oxidative phosphorylation and impairment of cellular respiration, compromising energy production (Cadenas and Davies 2000). The mitochondrial electron transport chain specifically, cytochrome C oxidase or complex IV, is altered in Alzheimer's disease (Parker et al. 1994), perhaps secondary to mutated and oxidatively damaged mitochondrial DNA (Mecocci et al. 1994). This is supported by results with cytoplasmic hybrid or cybrid cells (Swerdlow et al. 1997) which resemble electron transport chain defects observed in Alzheimer's disease (Parker et al. 1994).

Aβ metal interactions

The evidence supports brain metal homeostasis, specially Zn and Cu, as significantly altered in Alzheimer's disease (Bush 2003). The progressive synaptic disruption and ultimately neuronal loss observed in Alzheimer's disease might be secondary to toxic oxidative stress from excessive free-radical generation favoured by redox active transition metals bound to Aβ (Smith et al. 1997; Sayre et al. 2000; Barnham et al. 2004a). The generation of reactive oxygen species usually requires the reaction of O2 with a redox metal ion such as Cu or Fe. Aβ is a metalloprotein with high in vitro affinity for Cu (highest), and Fe and Zn (lowest) (Atwood et al. 1998; Bush et al. 2003). Aβ has been shown to co-ordinate transition metal ions through bridging histidine residues at positions 6, 13 and 14 (Miura et al. 2000; Curtain et al. 2001; Smith et al. 2006), similar to the metal co-ordination sphere found in the active site of superoxide dismutase (Barnham et al. 2004a). When Aβ binds Cu and Fe, extensive redox chemical reactions take place (Opazo et al. 2002; Barnham et al. 2004a). Isolated senile plaques generate reactive oxygen species in a manner dependent upon Cu and Fe (Sayre et al. 2000; Opazo et al. 2002).

Several lines of evidence point to the participation of transition metals in Aβ neurotoxicity. Brain copper and iron concentrations increase with age (Takahashi et al. 2001; Maynard et al. 2002). Very high concentrations of Cu (400 μm), Zn (1 mm) and Fe (1 mm) have been found in plaques of Alzheimer's disease-affected brains (Lovell et al. 1998). Genetic ablation of the zinc transporter 3 protein, required for zinc transport into synaptic vesicles reduced plaque formation in Tg2576 transgenic mice (Lee et al. 2002). Two methods of inducing aggregation of Aβ are (i) metal induced cross-linking leading to amorphous aggregates and fibril formation or (ii) lowering the pH (Yoshiike et al. 2001). Zn, Cu and Fe induce Aβ aggregation in vitro (Huang et al. 1997; Atwood et al. 1998). Aβ accumulation within the synaptic cleft, at which high concentrations of Zn (300 μm) and Cu (30 μm) are released during neurotransmission, would lead to high concentrations of soluble metallated Aβ, thereby promoting its toxicity and perhaps explaining the synaptic loss observed in Alzheimer's disease (Lee et al. 2002). The high Zn concentrations also promote the aggregation of the Cu/Fe-metallated Aβ, creating a reservoir of potentially toxic Aβ that is in equilibrium with the soluble pool. The large polymeric deposits of misfolded proteins not only represent the end result of the aggregation process but may act as inactive reservoirs in equilibrium with the small diffusible oligomeric toxic species responsible for the neurodegenerative pathology. Paradoxically, some emerging data suggest that Aβ might have a role as an anti-oxidant, a function that may wane with aging (Bush et al. 2003).

In a lipid membrane environment the addition of Cu or Zn to Aβ can induce a conformational change from β-sheet to α-helix, generating an allosterically ordered membrane-penetrating oligomer (Curtain et al. 2001). The extensive oxidative damage associated with Aβ (Martins et al. 1986; Butterfield et al. 2001) may involve calcium dysregulation, caused by either the formation of membrane calcium channels (Arispe et al. 1993) or modulation of an existing channel (Mattson et al. 1993).

Reactions of Cu with Aβ lead to oxidative modifications of the peptide. Among these modifications is the formation of a methionine sulfoxide at residue 35 (Met(O)Aβ) (Barnham et al. 2003b) and Raman spectroscopic studies have shown that Zn and Cu are co-ordinated to the histidine residues of the deposited Aβ in the senile plaque and that the Met35 of Aβ is oxidised (Dong et al. 2003). Met(O)Aβ has been isolated from Alzheimer's disease amyloid brain deposits (Naslund et al. 1994). Like the reduced form of Aβ, synthetic Met(O)Ab is neurotoxic, and this toxicity can be rescued by catalase and clioquinol, suggesting that the reduced and oxidised peptides have similar mechanisms of action.

Another potential oxidative modification is the covalent crosslinking of Aβ through dityrosine moieties at residue 10, i.e. oxidative modifications to Aβ generate soluble oligomeric forms of Aβ. Soluble oligomeric forms of Aβ have been associated with Aβ toxicity and are correlated with cognitive and memory decline (McLean et al. 1999). Covalently crosslinked Ab, such as that generated by the formation of dityrosine moieties, would resist clearance (Barnham et al. 2004a).

If the Tyr10 of Aβ is replaced by an alanine residue, this renders the peptide non-toxic. This modified form of Aβ is still able to form oligomers, driven by non-specific hydrophobic interactions but cannot form dityrosine crosslinked oligomers, suggesting that a specific subset of Aβ oligomers is responsible for toxicity and not the general formation of oligomers per se (Barnham et al. 2004a).

How the soluble oligomers of Aβ induce neurotoxicity is still the subject of ongoing research but a general theme has emerged whereby the interaction of Aβ with lipid membranes is a necessary step in neurotoxicity (Kagan et al. 2002; Butterfield and Boyd-Kimball 2004; Kawahara 2004; Eckert et al. 2005; Puglielli et al. 2005). These interactions cause changes in membrane fluidity, resulting in depolarisation and disorder (Eckert et al. 2005), pore/channel formation leading to ion channel formation and disrupted calcium homeostasis (Kagan et al. 2002; Kawahara 2004), lipid peroxidation via membrane-associated free radical formation (Butterfield and Boyd-Kimball 2004) and cholesterol oxidation (Puglielli et al. 2005). We have recently shown that the toxicity of Aβ/Cu complexes correlates with lipid peroxidation and that this can be inhibited by preventing the Aβ/Cu interaction (Smith et al. 2006).

The road to therapeutics

To date, no therapy has been shown to halt or reverse the underlying disease process and therapy remains confined to symptomatic palliative interventions (Barrow 2002).

Given the neuronal degeneration with impairment in cholinergic transmission in hippocampal and basal fore brain, areas associated with memory and cognition, as well as decreased levels of the cholinergic markers, choline acetyltransferase and acetyl cholinesterase, most treatment strategies are based in increasing intrasynaptic acetylcholine (ACh) levels (Auld et al. 2002), The acetyl cholinesterase inhibitors (AchE-I) tacrine, donepezil, rivastigmine and galantamine are now approved for the treatment of Alzheimer's disease (Emilien et al. 2000). However, subjects treated with AChE-I respond with a paradoxical increase in AChE levels and activity. This seems to negate the intended effect of the AChE-I in increasing intrasynaptic ACh levels. At the same time, clinical trials of AChE-Is and their meta-analyses continued to show favorable, albeit mild, effects on cognitive parameters, at least during the first 6–12 months of treatment. Against this background, attention has focused on identifying other possible mechanisms of action of the AChE-Is, especially on the amyloid precursor protein/Aβ pathway, and have begun to ask whether these drugs might have any disease-modifying effects (Caccamo et al. 2006) by attenuating the effects of Aβ-induced neuronal cytoxicity (Kimura et al. 2005), promoting α-secretase or decreasing β-secretase activity (Zimmermann et al. 2005; Caccamo et al. 2006), inhibiting Aβ aggregation (Belluti et al. 2005) or inhibiting GSK 3β activity and tau phosphorylation (Caccamo et al. 2006).

The modulation of glutamatergic transmission in Alzheimer's disease has also received increasing attention with the results of the clinical trials of the non-competitive NMDA antagonist memantine, proposed as a safe and effective symptomatic treatment of Alzheimer's disease patients (Rogawski and Wenk 2003). Assuming the toxic soluble oligomers of Aβ may inhibit long-term potentiation at the presynaptic level and that Aβ promotes the endocytosis of the NMDA receptor (mediated in part through α7 nicotinic ACh receptor [nAChR], protein phosphatase PP2B and tyrosine phosphase STEP (Snyder et al. 2005)), the findings on the beneficial behavioral effects of memantine in both Aβ toxicity models (Yamada et al. 2005) and amyloid precursor protein transgenic mouse models (Van Dam et al. 2005) will require further evaluation.

Compounds with the ability to inactivate reactive oxygen species might have therapeutic potential in the treatment of Alzheimer's disease. Existing knowledge and screens of natural product libraries have thrown up a wide variety of anti-oxidants and ‘neuroprotectants’ which have an effect on the actions of Aβ in some experimental cell culture toxicity assays (Moosmann and Behl 2002). Many of these assays are difficult to control, and there is little agreement in the field as to their validity. Though it has been proposed that cholinergic drugs are more effective in the treatment of Alzheimer's disease if used in association with anti-oxidants than the individual agents alone (Prasad et al. 2000), there has been limited clinical evaluation of the efficacy of anti-oxidants. Nevertheless, an increasing number of papers are appearing reporting efficacy of compounds derived from plants [ferulic acid (Sultana et al. 2005), green tea extracts (Rezai-Zadeh et al. 2005), curcumin (Yang F. et al. 2005), resveratol (Marambaud et al. 2005) fucoidan (Jhamandas et al. 2005) and various other plant materials (Lecanu et al. 2005)], as well as other natural products [docosahexaenoic acid (Lim et al. 2005), vitamin E (Quintanilla et al. 2005), oestrogens (Quintanilla et al. 2005), glutathione (Woltjer et al. 2005), melatonin (Quinn et al. 2005), gelsolin (Qiao et al. 2005) and insulin-like growth factor 1 (Aguado-Llera et al. 2005)] and a variety of small compounds (Caraci et al. 2005; Marrazzo et al. 2005). The classical lipophilic free radical scavenger, α-tocopherol (vitamin E) has been evaluated in both Alzheimer's disease and Parkinson's disease, and though it showed some encouraging results in Alzheimer's disease patients (Sano et al. 1997), especially when combined with ascorbic acid (Bano and Parihar 1997), it was found to have no beneficial effects in Parkinson's disease. Use of coenzyme Q10, l-carnitine and creatine might prevent mitochondrial oxidative damage and mitochondrial mutations (Moreira et al. 2005). Up-regulation of reactive oxygen species-scavenging enzyme capacities through neurotrophins may provide a mechanism for the prevention of neurotoxicity (Spina et al. 1992). There is a growing interest in the use of polyphenolic anti-oxidants to reverse age-related decline in neuronal signal transduction and cognitive and motor behavior deficits (Parihar and Hemnani 2003). Reactive oxygen species generation triggers glutamate-mediated excitotoxicity. Memantine, which targets the NMDA receptor, slows the development of the disease and is of modest benefit to patients in the moderately severe to severe range of the disease (Reisberg et al. 2003) (Fig. 2). From these investigations, a common theme emerges: that a wide variety of anti-oxidants can ameliorate the toxic gain-of-function of Aβ. This is consistent with our argument that Aβ itself is the principal pro-oxidant in Alzheimer's disease.

Other approaches to alter the progression of Alzheimer's disease involve the use of oestrogen, anti-oxidants (alone or in combination with selegiline) or non-steroidal anti-inflammatory drugs (NSAIDs) (Fig. 2).

A controversial area involves the effects of hormones (oestrogens and testosterone especially) and how they may affect amyloid precursor protein metabolism. Oestrogens have been shown not only to modulate neurotransmission, but also act as free radical scavengers, activating nuclear oestrogen receptor in intracellular signalling (Behl and Holsboer 1999) and preventing Aβ formation by promoting the α-secretase amyloid precursor protein-nonamyloidogenic pathway (Xu et al. 1998). Conflicting results in experimental models have appeared, in which oestrogen deficiency exacerbates Aβ in the amyloid precursor protein23 transgenic model (Yue et al. 2005) and neither oestrogen deprivation nor replacement affected Aβ deposition in the PDAPP transgenic model (Green et al. 2005). The mechanisms through which oestrogen/testosterone might act remain obscure, but include oestrogen-dependent regulation of metal homeostasis in the brain through the expression of the neuronal zinc-transporter, ZnT3.

If, as postulated, Alzheimer's disease pathology is the consequence of a chronic imbalance between Aβ production and clearance, the most rational strategy to treat the disease would involve retarding, halting or even reversing the process that leads to increased production of Aβ (Barrow 2002).

The most promising strategy for neuroprotection might be reducing formation of Aβ by partially inhibiting either β- or γ-secretase (Fig. 2) and/or stimulation of α-secretase activity (Xia 2003; Zimmermann et al. 2004; Caccamo et al. 2006). Total inhibition of either β- or γ-secretase should block Aβ production completely. There are already potent γ-secretase inhibitors (Lanz et al. 2004; Wong et al. 2004) undergoing human trials. During 2005, the first publications of in vivoγ-secretase inhibition/modulation of Aβ42 biogenesis appeared. One of the first known inhibitors (DAPT) was shown to be effective in acute experiments in behavioral tests (contextual fear conditioning) in transgenic mice (Comery et al. 2005). Chemical modifications to the structure of DAPT has improved its delivery to the brain (Quélever et al. 2005) as well as with other compounds (Laras et al. 2005) to achieve lower effective dosages while minimising the risk of peripheral adverse effects. Several classes of inhibitors and modulators are showing favourable acute pharmacokinetics, with rapid lowering of plasma and CSF Aβ levels (Anderson et al. 2005; Barten et al. 2005). The first in-human Phase I results have shown that LY450139 (Lilly Indianapolis, USA) achieved a significant lowering of plasma Aβ, but not CSF Aβ, in normal volunteers (up to 50 mg/day for 14 days) or subjects with Alzheimer's disease (up to 40 mg/day for 6 weeks) (Siemers et al. 2005; Siemers et al. 2006). The drug was well tolerated. Further explorations of the properties of γ-secretase inhibitors are revealing unanticipated effects on synaptic function (Dash et al. 2005). New classes of γ-secretase inhibitors/modulators continue to be disclosed, as part of the attempt to develop compounds that are devoid of side-effects (Gundersen et al. 2005; Lewis et al. 2005). The major concern is the inhibition of signalling in the Notch pathway, which affects cellular differentiation (van Es et al. 2005). Further research on the mechanistic operations of the γ-secretase complex (Sato et al. 2005) may lead to new paths of drug discovery, as might gene targeting of presenillin, PEN-2, APH-1 and nicastrin lead to selective regulation of γ-secretase activity (Saura et al. 2005; Xie et al. 2005). The development of β-secretase inhibitors has been focused on the discovery and design of compounds which target the active site of BACE-1. Improved assays (Pietrak et al. 2005) and structural-based in silico designs (Hanessian et al. 2005; Huang et al. 2005) have added to the existing pipe-line of drugs in early preclinical development (Kornacker et al. 2005; Lefranc-Jullien et al. 2005) or early discovery programs (Lee et al. 2005b). Other proteins interacting with BACE-1 may become drug targets (Xie and Guo 2005) and gene targeting of BACE-1 mRNA using siRNA is also producing encouraging preliminary results (Singer et al. 2005).

In contrast to the inhibition of Aβ biogenesis, therapeutic strategies which directly target Aβ itself should have a lower risk of unanticipated side-effects, as the accumulated Aβ molecule is a disease-specific trait of Alzheimer's disease. If the Aβ fragment (or its domain within amyloid precursor protein) does, however, subserve some critical normal function, then targeting Aβ itself might interfere with this function and thereby lead to adverse side-effects. To date, however, a normal function for Aβ has not been identified. Amyloid precursor protein knockout mice are viable and healthy, providing some support for this idea.

Given the evidence that levels of soluble Aβ correlate with disease severity (McLean et al. 1999) and that the Aβ is the main neurotoxic factor in the development of Alzheimer's disease, the design of drugs directly targeting Aβ and its varied conformations as well as agents inhibiting Aβ oligomerisation should be more effective than those that merely block Aβ deposition (Wolfe 2002). Two basic strategies have been proposed to reduce or remove Aβ from the brain: immunisation (McLaurin et al. 2002; Schenk 2002), breaking the pathway that leads to Aβ deposition by inducing an active immune response against the Aβ (Janus et al. 2000; Weiner et al. 2000), passive administration of specific anti-Aβ antibodies (Bard et al. 2000; DeMattos et al. 2001; Wilcock et al. 2003), promoting microglial clearance (Bard et al. 2000), and/or by redistribution of Aβ into the systemic circulation (DeMattos et al. 2001) (Fig. 2). Since 1999, increasing evidence has accumulated to make a compelling antibody-mediated Aβ clearance/neutralisation strategy. Experiments in in amyloid precursor protein transgenic mice models continue to demonstrate efficacy without detectable toxicity (Brendza et al. 2005; Klyubin et al. 2005). The aborted clinical trial with the Elan Ab42 antigen (AN1792) has provided a wealth of clinical information (Gilman et al. 2005; Lee et al. 2005d; Masliah et al. 2005) which will assist further development of strategies designed to avoid the auto-immune adverse events (Lee et al. 2005a; Racke et al. 2005). Chief among these will be avoidance of T-cell-mediated responses (Agadjanyan et al. 2005) and the development of passive immunisation protocols (Hartman et al. 2005). Passive immunisation clinical trials are currently underway. In the meantime, novel methods of antigen presentation (Frenkel et al. 2005; Youm et al. 2005) and the use of neo-epitopes (Arbel et al. 2005; Yamamoto et al. 2005) are under investigation. Neo-epitopes generated post-transationally by oxidative modification of Aβ should have inherently less potential to generate an auto-immune adverse reaction.

Trials of anti-inflammatories in Alzheimer's disease have been conducted, and considerable research efforts undertaken to examine the effects of anti-inflammatories in a variety of experimental models. These include the non-steroidal anti-inflammatories (Farias et al. 2005; Morihara et al. 2005), peroxisome proliferator-activated receptor-γ agonists (Echeverria et al. 2005; Shie et al. 2005), cannabinoids (Ramírez et al. 2005) and glucocorticoids (Boedker et al. 2005). To date, no prospective clinical trial with an anti-inflammatory has shown a convincing beneficial outcome. Anti-inflammatory medication has also been shown to have direct effects on the cleavage of amyloid precursor protein by γ-secretase, an effect that is independent of the drugs’inhibition of cyclooxygenase and other inflammatory mediators (Beher et al. 2004) (Fig. 2). Some such drugs reduce cytopathology in amyloid precursor protein transgenic mice (Lim et al. 2001; Jantzen et al. 2002).

While high-cholesterol diets increase Aβ pathology in experimental animals (Refolo et al. 2000) general caloric restriction has often been associated with longevity in rodent models of aging, and recent studies in transgenic Alzheimer's disease models (Patel et al. 2005; Wang et al. 2005) and normal rodents (Tang 2005), suggesting an effect on Aβ plaque load or α-secretase processing of amyloid precursor protein. This led to the assessment of the effects of modulating cholesterol homeostasis over Alzheimer's disease pathology. Cholesterol and inhibitors of cholesterol synthesis (statins) have been shown to significantly alter amyloid precursor protein processing in vitro, with a reduction in β-secretase cleavage and a decreased Aβ production (Fassbender et al. 2001), reducing pathology in amyloid precursor protein transgenic mice (Refolo et al. 2001). Statins have been associated with a lowered incidence of Alzheimer's disease (Wolozin et al. 2000). While some early phase clinical trials with statins have shown encouraging results (Masse et al. 2005), others have not (Höglund et al. 2005a, b). Cholesterol-independent effects have also been noted for statins acting on isoprenyl intermediates in the cholesterol biosynthetic pathways, with a putative anti-inflammatory effect induced by reactive microglia (Cole et al. 2005; Cordle and Landreth 2005). Statins have also been implicated in the toxic gain-of-function of Aβ interacting with a7-nAChR (Si et al. 2005), although the mechanism for this remains unclear.

Based on the role that metal ions such as Cu, Fe and Zn play in the biochemical processes associated with Aβ deposition and neurotoxicity (Cherny et al. 1999; Barnham et al. 2004b), a further therapeutic strategy based on inhibiting Aβ/metal interactions led to the design and development of molecules, known as metal-protein attenuating compounds (MPACs) (Barnham et al. 2004b) (Fig. 2), that inhibit the deleterious effects of aberrant metal interactions through competition with the target protein for the metal ions, leading to a normalisation of metal homeostasis. MPACs not only inhibit the in vitro generation of hydrogen peroxide but also have been shown to reverse the precipitation of Aβin vitro and in postmortem human brain specimens (Bush 2002), reducing Aβ amyloid burden by a direct solubilisation and by reducing toxic oxidative stress (Cherny et al. 2001). Clioquinol, 5-chloro-7-iodo-8-hydroxyquinoline, is a hydrophobic Zn and Cu chelator that freely crosses the blood–brain barrier (Padmanabhan et al. 1989). Preclinical studies showed that clioquinol increased soluble phase Aβ by more than 200% in a concentration-dependent fashion in homogenised postmortem human brain samples, and its efficacy tested in transgenic Tg2576 mice showed a dramatic 49% decrease in brain Aβ deposition after 9 weeks of oral treatment (Cherny et al. 2001; Raman et al. 2005). Clioquinol was chosen to be tested as an Aβ amyloid solubilising and antitoxic agent in a randomised, double blind, placebo-controlled pilot Phase II clinical trial (Ritchie et al. 2003; Ibach et al. 2005). Oral clioquinol treatment was statistically significant in preventing cognitive deterioration in the moderately severe Alzheimer's disease patient group, with no evidence of toxicity (Ritchie et al. 2003). Other groups have considered chelators (Gaeta and Hider 2005; Liu et al. 2005) or other novel compounds (Cui et al. 2005). Our own studies have progressed with a new chemical entity based around the 8-OH quinoline structure. This compound (PBT2, Prana Biotechnology, Melbourne, Australia) has passed phase I and will soon commence phase II clinical development.

Additional binding sites on Αβ, such as the glycosaminoglycan (GAG) site [HHQK (13–16)], have been targeted with compounds such as 3-amino-1-propanesulfonic acid (3-APS: Alzhemed, Neurochem Inc., Quebec, Canada). The results of early clinical trials released by the company have shown some effects on CSF Aβ42, but none on ADAS-cog or MMSE.

The pharmaceutical industry has for a long time interrogated their libraries for compounds that are anti-aggregants and/or antifibrillogenic. Many hits with compounds that look similar to Congo Red have never been developed. Similarly, compounds capable of disaggregating or defibrillating Aβ have been sought, but not with the intensity of the search for anti-aggregants. Many peptidyl/protein-like designs have been examined (Gibson and Murphy 2005; Lee et al. 2005e; Schuster et al. 2005), but other small molecules have also been discovered which hold some promise (Hennessy and Buchwald 2005; Kanapathipillai et al. 2005; Lee et al. 2005c). We have identified other structural changes or mechanisms of toxicity for Aβ which include the oxidative modifications of Tyr10 and Met35, the interaction of Aβ with the polar head groups of the lipid bilayer, and the interaction of Aβ with other proteins (Mettenburg et al. 2005; Yang S. P. et al. 2005).

There are many new potential therapeutic strategies under evaluation. With the advent of RNA interference silencing, it is to be expected that attempts at direct amyloid precursor protein gene regulation will emerge. As a forerunner to this, models in which the overexpressed human amyloid precursor protein transgene in mice can be down-regulated with doxycline provide a proof-of-principle that rapid control over Aβ expression and deposition can be obtained without gross adverse side-effects (Jankowsky et al. 2005). Unexpectedly, Aβ deposits formed before the onset of down-regulation of APP seemed to be remarkably stable, indicating that any treatment of this type in isolation might have to be administered early in the natural history of Alzheimer's disease.

As a presumptive cell surface receptor, amyloid precursor protein probably has ligands and effector mechanisms for signal transduction. Nearly 200 proteins have been reported as having direct interactions with amyloid precursor protein. Suspected ligands in the extracellular domain include growth factors (nerve growth factor in particular), heparin-containing extracellular matrix, metals (through the extracellular Cu/Zn binding domain) and amyloid precursor protein itself through hetero- and homo-dimerisation. Small compounds such as propentofyline (Chauhan et al. 2005) can affect nerve grwoth factor release, and through this modulate the amyloidogenic pathway.

The re-uptake, clearance and degradation of Aβ is still subject to considerable uncertainties. If sporadic Alzheimer's disease is the result of a low level shift (< 10%, for example) in the efficiency in any of these mechanisms, then a therapeutic strategy aimed at restoring or by-passing this faulty mechanism could be very useful. Each of the different pools of Aβ probably has slightly different mechanisms of elimination, varying with the cellular compartment in which Aβ resides over the course of its catabolic cycle. Several pieces of evidence point towards the enzymes neprilysin and insulin degrading enzyme as key players (Saito et al. 2005) but the highly sought evidence from gene linkage studies remains elusive (Eckman and Eckman 2005). A new candidate, angiotensin-converting enzyme (ACE), has emerged (Hemming and Selkoe 2005) and it will be of great interest to learn whether the ACE-inhibitors could be having an adverse influence over the natural history of Alzheimer's disease.

While Aβ has captured the imagination of most Alzheimer's disease researchers, studies of the neurofibrillary tangle and its constituent, the tau microtubule associated protein, have progressed to a point where clear therapeutic strategies are emerging. The exact form of tau which causes neuronal degeneration is now being re-examined (Duff and Planel 2005), with data emerging that the soluble aggregated species, akin to soluble Aβ oligomers, might represent the best target. The binding sites on tau (Mukrasch et al. 2005) for a variety of interactors are potential targets. Down-regulation of expression of the tau gene (Santacruz et al. 2005) or changing the alternative splicing (Rodriguez-Martin et al. 2005) also offer some new strategies. As the molecular basis for the accumulation of tau in the Alzheimer's disease brain becomes clearer, so will more precise therapeutic targets. If tau accumulation is closely linked to Aβ toxicity, then oxidative modifications of tau become understandable (Santa-María et al. 2005) and subject to anti-oxidative classes of drugs. Metal ions might also affect this pathway (Ma et al. 2006). Looking at the normal function and processing of tau has raised the possibility of using microtubule-stabilising agents such as paclitaxel (Taxol) (Michaelis et al. 2005). Great controversy still persists about the role of normal and abnormal phosphorylation of tau in its passage from a highly soluble cytoskeletal-associated protein into an aggregated neurofibrillary tangle. If phosphorylation of specific amino acids by specific kinases such as c-Abl (Derkinderen et al. 2005), Cdk5 (Sakaue et al. 2005), GSK-3 (Noble et al. 2005) or MAPK (Lambourne et al. 2005) proves to be pathogenic, then specific kinase inhibitors, including lithium (Noble et al. 2005), might be developed for Alzheimer's disease. However, if phosphorylation proves to be a secondary event, following aggregation and accumulation of intracellular tau, then this approach would not be expected to be useful. Other post-translational modifications including proteolytic cleavages have been proposed (Cotman et al. 2005) – all amenable to therapeutic drug discoveries. As with Aβ, small compounds capable of inhibiting aggregation and fibrillisation of tau are now being examined in vitro (Necula et al. 2005; Taniguchi et al. 2005) but require much more work in animal models.

It is extremely unlikely that a single class of compound or targeting a single mechanism of action will be sufficient to treat Alzheimer's disease. For this complex disease, it is far more likely that a combination of drugs targeting various aspects of the greater amyloid precursor protein/Aβ pathway will evolve into some form of rational therapy.

Neuroimaging

The insights into the molecular mechanism of Alzheimer's disease pathogenesis have opened new opportunities not only for the successful development of neuroprotective treatment strategies aimed at the prevention of Aβ generation but also into new structural and functional neuroimaging approaches.

Structural neuroimaging techniques, such as computed tomography and magnetic resonance imaging, are routinely used in the clinical evaluation of Alzheimer's disease patients, and are mainly used to exclude other treatable causes of dementia (Scheltens 2001). Widespread cortical atrophy with a thinning of medial temporal lobe structures are the most consistent structural neuroimaging findings associated with Alzheimer's disease, though not pathognomonic of the disease because there is overlap with ‘normal’ aging (Jobst et al. 1992). To account for the cognitive impairment commonly observed in Alzheimer's disease, magnetic resonance imaging has been used to examine atrophy of the entorhinal, perirhinal and temporal cortices in patients with early Alzheimer's disease (Chetelat and Baron 2003; Thompson et al. 2004). Volumetric changes on magnetic resonance imaging are entirely consistent with the patterns of neuropathological progression in Alzheimer's disease, and the severity of volume loss is correlated with disease severity. The same regions were found to be reduced in both mild cognitive impairment and Alzheimer's disease compared to controls (Du et al. 2001; de Leon et al. 2006). Both conditions were also significantly associated with cortical grey matter loss and ventricular enlargement. Substantial neuronal loss has occurred by the time atrophy is detectable by MRI (Killiany et al. 2002). Furthermore, the absence of cortical atrophy or medial temporal lobe changes is not sufficient to exclude a diagnosis of Alzheimer's disease. The fact that structural changes at visual inspection are not evident until late in the course of the disease has prompted the development and refinement of more sophisticated quantitative techniques, capable of revealing subtle changes over time (Petrella et al. 2003).

Modern functional neuroimaging techniques such as positron emission tomography (PET), single photon emission tomography (SPECT), magnetic resonance spectroscopy (MRS), functional MRI, MR Diffusion weighted imaging and magnetoencephalography (MEG) have been developing new approaches not only to determine if an individual suffers from a particular form of dementia, but also delving into the molecular mechanisms of synaptic failure and neurodegeneration (Schuff et al. 2002; Petrella et al. 2003). More sensitive than structural imaging modalities, functional neuroimaging approaches have the capability to identify subtle pathophysiologic changes in the brain before structural changes are present (Xu et al. 2000; Dickerson et al. 2001), therefore possessing greater potential for accurate and early diagnosis, monitoring disease progression, and better treatment follow-up (Silverman and Phelps 2001; Villemagne et al. 2005a).

Perfusion assessed by functional MRI has been shown to have about 85–95% sensitivity for the diagnosis of patients with mild or moderate Alzheimer's disease with 88–95% specificity (Bozzao et al. 2001). Proton MRS studies (Doraiswamy et al. 2000) have shown regional decreases in N-acetylaspartate in patients with Alzheimer's disease in temporal and parietal cortices (Jessen et al. 2000) demonstrating a positive correlation between the degree of N-acetylaspartate reductions and disease severity by neuropathologic criteria (Mohanakrishnan et al. 1995). MRS has also been applied to monitor response to therapeutic interventions in Alzheimer's disease (Satlin et al. 1997). Patients with Alzheimer's disease showed significantly higher mean diffusibility in hippocampus, cingulate, temporal and parietal white matter than a control group using diffusion weighted MR imaging (Kantarci et al. 2001), while diffusion tensor MR imaging has shown diffuse reduction in white matter integrity in patients with Alzheimer's disease (Rose et al. 2000).

PET is a sensitive molecular imaging technique that allows in vivo quantification of radiotracer concentrations in the picomolar range, permitting detection of disease processes at asymptomatic stages when there is no evidence of anatomic changes on CT and MRI (Phelps 2000).

Several studies have evaluated regional cerebral glucose metabolism with fluorodeoxyglucose (FDG) and PET. A typical pattern of reduced temporoparietal FDG uptake with sparing of the basal ganglia, thalamus, cerebellum and primary sensorimotor cortex is typical of Alzheimer's disease (Coleman 2005). Though FDG-PET is mainly used in the differential diagnosis of Alzheimer's disease, it is the neuroimaging technique that has been shown to yield the highest prognostic value for providing a diagnosis of presymptomatic Alzheimer's disease 2 years or more before the full dementia picture is manifested (Silverman et al. 2001; Chang and Silverman 2004). In a multicentre study the prognostic value of FDG-PET showed a high degree of sensitivity (93%) and moderate specificity (73%) for prediction of progressive dementia (Silverman et al. 2001).

SPECT studies evaluating regional cerebral blood flow (rCBF) have shown a similar pattern as the one described for PET-FDG studies, with relative rCBF paucity in the temporoparietal regions (Camargo 2001). In a prospective study with histologic confirmation of over 200 dementia cases and 119 control cases (Jobst et al. 1998), SPECT rCBF evaluation allowed differentiation of patients with Alzheimer's disease from control subjects with high sensitivity and specificity (89% and 80%, respectively).

PET and SPECT can also assess neurotransmitter/neuroreceptor systems in vivo. Abnormally low densities of nAChRs have been measured in vitro in autopsy brain tissue of Alzheimer's disease patients. PET studies revealed a reduced uptake and binding of 11C-nicotine in the temporal and frontal cortices of Alzheimer's disease patients (Nordberg 1993). Tacrine treatment increased cerebral blood flow, cerebral glucose utilisation, and uptake of 11C-nicotine to the brain paralleled by improvement in neuropsychological performance (Nordberg et al. 1998). Though the main focus of neuroceptor studies in Alzheimer's disease has been the study of nAChRs, several other neurotransmitter/neuroreceptor systems were also evaluated in dementing neurodegenerative diseases (Higuchi et al. 2000; Walker et al. 2002; Piggott et al. 2003; Kepe et al. 2006).

As mentioned before, because new treatment strategies to prevent or slow disease progression through early intervention are being developed and implemented, there is an urgent need for early disease recognition, which is reflected in the necessity of developing sensitive and specific biomarkers, specific for a particular trait underlying the pathological process, as adjuncts to clinical and neuropsychological tests. Clinical criteria together with current structural neuroimaging techniques (CT or MRI) are sensitive and specific enough for the diagnosis of Alzheimer's disease at the mid or late stages of the disease; however, they focus on non-specific features derived mainly from neuronal loss and atrophy, which are late features in the progression of the disease, and are secondary to the basic functional alteration. The development of a reliable method of assessing disease-specific biomarkers, such as Aβ amyloid burden in vivo, may permit early diagnosis at presymptomatic stages and more accurate differential diagnosis, while also allowing treatment follow-up.

In the same way neuropathology was boosted by the techniques and dyes introduced by visionary pioneers like Cajal and Nissl, we are now seeing some derivatives of those histological dyes finding their way into emission tomography (Sair et al. 2004; Villemagne et al. 2005a) and magnetic resonance imaging (Sato et al. 2004; Zhang et al. 2004; Higuchi et al. 2005).

As Aβ is at the centre of pathogenesis of Alzheimer's disease, most efforts were focussed on developing radiotracers or agents that allow Aβ imaging in vivo (Sair et al. 2004; Villemagne et al. 2005a). Tau imaging is still in its early stages of development (Okamura et al. 2005).

Several compounds have been evaluated as potential Aβ probes: derivatives of histopathological dyes such as Congo red, Chrysamine-G, Thioflavin S and T, and acridine orange (Mathis et al. 2002; Klunk et al. 2003; Shimadzu et al. 2003; Zhang et al. 2005), NSAID derivatives (Shoghi-Jadid et al. 2002) as well as self-associating Aβ amyloid fragments (Marshall et al. 2002), anti-Ab monoclonal antibodies (Walker et al. 1994), serum amyloid p, and basic fibroblast growth factor (Shi et al. 2002). Some of these PET radiotracers are being evaluated in transgenic mice as potential MR contrast agents (Sato et al. 2004; Higuchi et al. 2005; Vanhoutte et al. 2005; Wadghiri et al. 2005).

Quantitative imaging of Aβ burden in vivo is allowing us to define the relationship between Aβ burden and clinical and neuropsychological characteristics in Alzheimer's disease. 11C-PIB, a derivative of Thioflavin T, has been shown to possess high affinity and high specificity for amyloid fibrils and binds to amyloid plaque but not neurofibrillary tangles in postmortem human brain homogenates in vitro (Lockhart et al. 2005; Ye et al. 2005). PET studies in human subjects have shown a robust difference between the retention pattern in Alzheimer's disease patients and healthy controls, with Alzheimer's disease cases showing significantly higher retention of 11C-PIB in neocortical areas of the brain affected by Aβ deposition (Klunk et al. 2004; Price et al. 2005; Villemagne et al. 2005b). Aβ burden is significantly elevated in Alzheimer's disease, dementia with Lewy-bodies, and about 50% of mild cognitive impairment subjects compared to healthy controls, while frontotemporal dementia and non-demented Parkinson's disease subjects show no cortical 11C-PIB binding (Fig. 3). About 25% of the healthy controls showed cortical binding, predominantly in the prefrontal cortex, though to a lesser degree than Alzheimer's disease patients. The demonstration of 11C-PIB binding in a proportion of healthy control subjects supports in vitro observations that Aβ aggregation predominantly occurs before onset of dementia (Price and Morris 1999; Morris and Price 2001).

Figure 3.

 Parametric PIB PET distribution volume ratio images (DVR) of a spectrum of neurodegenerative diseases secondary to misfolded proteins (α-syn, Aβ, tau). Representative PET images of a 73-year-old healthy control (NC) subject (MMSE 30), a 61-year-old Parkinson's disease (PD) patient (MMSE 27), a 78-year-old dementia with Lewy-body dementia (DLB) patient (MMSE 19), a 70-year-old mild cognitive impairment (MCI) patient (MMSE 26), an 82-year-old Alzheimer's disease (AD) patient (MMSE 22) and a 78-year-old frontotemporal dementia (FTD) patient (MMSE 26). DVR PET images show no cortical PIB retention in NC, PD or FTD with a clearly different pattern from DLB, MCI or AD patients, and significant PIB retention in the frontal and temporal cortices.

PIB PET shows promise in the differential diagnosis of Alzheimer's disease from frontotemporal dementia (Fig. 3) but the emphasis on amyloid imaging should not be limited to its capability for differential diagnosis. With new treatments to prevent or slow Alzheimer's disease progression by either preventing Aβ deposition or increasing its clearance entering clinical trials, agents that could delay the onset of dementia, and the role of imaging and quantifying Aβ burden in vivo are becoming crucial (Ritchie et al. 2003; Schenk et al. 2004). The ability to detect preclinical or early stage disease through clinical, laboratory and neuroimaging tests, combined with anti-Aβ amyloid in the at-risk patient, or the patient with mild cognitive impairment, may prevent or delay functional and irreversible cognitive losses, making it possible at the same time to customise and monitor treatment.

Conclusions

Alzheimer's disease is a neurodegenerative disorder characterised by a slow but relentless progressive cognitive decline and memory loss. It has a devastating effect not only on the sufferer but also on their caregivers, with a tremendous socio-economic impact not only on families but also on the health system, which will only increase in the upcoming years.

The neuropathologic hallmarks of the disease are extracellular deposits of Aβ in senile plaques, neurofibrillary tangles, with selective neuronal and synaptic loss in cortical areas of the brain associated with cognitive and memory functions.

Aβ is the main component of the amyloid plaques. All available evidence points to the breakdown of Aβ homeostasis as the key role in Alzheimer's disease pathogenesis. Genetic studies have shed light on the pathogenesis and progression of Alzheimer's disease. To date, four genes have been linked to autosomal dominant, early onset familial Alzheimer's disease: amyloid precursor protein, presenilin 1, presenilin 2 and apolipoprotein E. All mutations linked to amyloid precursor protein and presenilin proteins lead to an increase in Aβ production. Aβ not only aggregates into amyloid plaques but is toxic per se, while having an effect on intracellular tangle formation and other factors (e.g. cytokines, neurotoxins, etc.) that also play an important role in the neurotoxic progression of Alzheimer's disease.

Aβ is neurotoxic through a number of possible mechanisms including oxidative stress, excitotoxicity, energy depletion, inflammatory response and apoptosis, and whereas the exact mechanism by which Aβ might produce synaptic loss and neuronal death is controversial, it is believed that a toxic oxidative interaction between various metal species and Aβ triggers an oxidative response with free radical production, progressive disruption of synaptic and neuronal function leading ultimately to cell death.

At this point there is no cure for Alzheimer's disease. A deeper understanding of the molecular mechanism of Aβ formation, degradation and neurotoxicity is being translated into new neuroimaging and therapeutic approaches. Most of the approved palliative treatments regimens involve the use of acetylcholinesterase inhibitors, glutamatergic agents, non-steroidal anti-inflammatory drugs and anti-oxidants. The clinical development of drugs directly targeting the Aβ pathway is at an early stage. The most promising approaches focus on reducing Aβ formation, increasing its removal or blocking the formation of Aβ oligomers and fibrils, therefore inhibiting neurotoxicity. The γ-secretase inhibitors trials are of immense theoretical interest, as they are likely to provide the most compelling support for the Aβ theory of Alzheimer's disease. The trials around the Aβ metal binding site or the CAG binding sites also have the potential to address this aspect. Immunisation/immunomodulation of Aβ holds great promise for elucidating the Aβ clearance/neutralisation strategies on which there is currently a dearth of information. A variety of prospective statin-mediated approaches will also test the hypothesis that cholesterol has an important role in the biogenesis of Alzheimer's disease. The anti-oxidant trials have the disadvantage of lacking specificity for Aβ, but nonetheless will continue to provide much needed guidance for the general theory of the Alzheimer's disease brain being under oxidative stress.

Currently, clinical diagnosis of Alzheimer's disease is based on progressive impairment of memory and decline in at least one other cognitive domain, and by excluding other diseases using structural neuroimaging techniques (CT or MRI). This approach is only sensitive and specific enough for the diagnosis of Alzheimer's disease at the mid or late stages of the disease. Because new treatment strategies to prevent or slow disease progression through early intervention are being developed and implemented, there is an urgent need for early disease recognition, which is reflected in the necessity of developing sensitive and specific biomarkers, specific for a particular trait underlying the pathological process, as adjuncts to clinical and neuropsychological tests.

The development of a reliable method of assessing Aβ amyloid burden in vivo may permit early diagnosis at presymptomatic stages and more accurate differential diagnosis, while also allowing treatment follow-up. In vivo amyloid imaging with PET is allowing new insights into Aβ deposition in the brain, facilitating research into the causes, diagnosis and future treatment of dementias, where Aβ may play a role.

Following Alois Alzheimer's groundbreaking presentation in Tübingen 100 years ago (Alzheimer 1907) it has been a long night's journey into the day. We are now entering a new dawn that promises the delivery of revolutionary developments for the control of dementias.

Acknowledgements

We thank Christopher Rowe, Catriona McLean, Steven Ng, Michelle Fodero-Tavoletti, Tiffany Cowie, Lisa Foster, Laura Leone, Fairlie Hinton and Emma Mitchell for their crucial role in our ongoing research projects.

We apologise to our colleagues for omitting references due to space constraints.

We acknowledge the funding support of the National Health and Medical Research Council of Australia, Neurosciences Victoria, Austin Hospital Medical Research Foundation, Schering AG and Prana Biotechnology Ltd.

Ancillary