Potential novel targets for Alzheimer pharmacotherapy: I. Secretases


Robert B. Raffa, PhD Professor of Pharmacology, Temple University School of Pharmacy, 3307N, Broad Street, Philadelphia, PA 19140, USA. Tel.: +1215 7074976; fax: +1215 7075228; e-mail: rraffa@nimbus.temple.edu


The prevailing major theory of Alzheimer's disease (AD) is that insoluble amyloid β-peptide (Aβ) found in the cerebral plaques characteristic of the disease is causative or is at least a contributing factor. According to this theory, inhibition of aberrant Aβ production should prevent or at least limit the extent of AD pathophysiology. As three ‘secretase’ enzymes (α, β and γ) catalyse the proteolytic cleavage of amyloid precursor protein (APP) (the precursor protein of Aβ), one or more secretases have become targets for potential novel AD pharmacotherapy. Secretase inhibitors have been designed and are in various stages of development. The clinical trials of these compounds will, if positive, result in drugs with dramatically better clinical efficacy or, if negative, will force a reassessment of the theory about the role of Aβ in AD.


Alzheimer's disease

Alzheimer's disease (AD) is a progressive neurodegenerative disorder having diagnostic criteria according to DSM-IV (1) of memory impairment plus decrements in language (aphasia), voluntary motor activities (apraxia), object recognition (agnosia), or planning and organizational skills – in the absence of other causative factors. It was first described by Alois Alzheimer's in 1907 (2). Patients with AD score about 3–4 points less each year on standardized cognition tests such as the Mini-Mental State Examination (1). Alzheimer's disease is a progressive and eventually fatal disease. It worsens to include impaired motor function, and eventually AD patients become mute and bedridden (1). Patients do not die of AD directly, but rather of related sepsis, pneumonia, or malnutrition (3). Survival after onset of AD can range from 3 to 20 years (4). However, diagnosis typically follows onset by several years, so life expectancies after diagnosis typically range from only 4 to 8 years (4). Generally, survival time decreases as the age of onset increases (5).

Alzheimer's disease has gradual onset of unknown etiology (6). It differs from senility [L. senilis, old; technically in the senium (60 or older)] (a general term for a variety of mental or physical weakness that can occur in old age) (7) and delirium (a temporary state of mental confusion of rapid onset arising from high fever, intoxication, or shock) (6). Although a definitive diagnosis of AD is only possible by biopsy or autopsy (4), probable AD is diagnosed by ruling out other disorders known to cause similar symptoms, such as cerebrovascular diseases, Parkinson disease, brain tumor, hypothyroidism, neurosyphilis and HIV infection (1).

Possible causes of AD would be more limited if AD could be shown to be a purely modern disease. Unfortunately, it is difficult to determine if AD was around in ancient times, because people generally did not live past their forties and therefore would have died before showing AD symptoms. For those who lived longer than average, it was presumed that forgetfulness and ‘senility’ were a normal part of aging and that there was little hope of improvement or successful treatment (8). Currently, with life expectancy around 75 years, AD has a point prevalence of approximately 4 million people and an incidence of about 250 000 new cases each year (4).

Present understanding implicates both genetic and environmental factors in the causation of AD. A genetic component is fairly well established in certain cases. For example, most early onset (less than about 60 years old) ‘familial’ AD have been traced to alterations on chromosomes 1, 4, and 21 (9) and first-degree relatives of patients who have early onset familial AD are more likely than are unrelated individuals to develop the disease (1). People who have Down syndrome have an extra copy of chromosome 21 and have a high likelihood of developing AD by age 50 (10). The more common late-onset (sporadic) cases (age 60 or over) might have an inheritable trait related to the gene for cholesterol transporter Apolipoprotein E (Apo E), which is found on chromosome 19 (4, 9). There are three types of Apo E: E2, E3, and E4 (9). People inherit one Apo E type from their mother and one from their father. Inheriting one E4 type increases the risk of developing late-onset AD and an E4/E4 combination increases that risk even more (8). However, the presence of Apo E4 in people with Down syndrome seems to be unrelated to the development of their AD (11). Females (post-menopausal) are more at risk for developing late-onset AD than are males (1). Females with Down syndrome also have an increased risk of developing AD compared with males with Down syndrome (11). Environmental factors speculated or implicated to contribute to the development of AD include (8): level of education [in twins, longer schooling is associated with reduced risk of developing AD (12)]; history of head trauma (the inflammation and death of nerve cells might lead to AD later in life); vascular changes (because of impaired blood flow and oxygen supply to the brain); decreased acetylcholine levels (associated with memory loss); and exposure to aluminium {however, although Al(OH)3 is neurotoxic, it destroys brain cells in a different pattern than seen in AD and therefore several experts discount it as a major cause of AD [e.g. (4)]}.

Biochemistry of AD

There are four major anatomical changes that occur in the AD diseased brain: cortical atrophy, degeneration of neurons, accumulation of extracellular amyloid plaques and accumulation of intracellular neurofibrillary tangles. The latter two, plaques and tangles, are the hallmarks of AD and their main histological components are two proteins, amyloid β-peptide (Aβ) and τ, respectively. The Aβ is so named because of the β-pleated sheet secondary structure of the amyloid protein, while τ proteins are critical for polymerization and stabilization of tubulin in axonal microtubules and the cross-linking of microtubules to other cellular structures (13, 14).


Neurofibrillary tangles are comprised primarily of paired helical filaments formed from τ protein (15–18) and are found primarily in the pyramidal regions of the amygdala, hippocampus and neocortex (19). In non-AD patients, τ protein provides support to the cell's framework of microtubules. In AD patients, τ protein is hyper-phosphorylated. The extra phosphate is thought to cause microfilaments to become abnormally paired into helical filaments, which in turn form tangles. The cell then collapses and dies. It is unclear how many or which kinases are responsible for τ hyperphosphorylation in AD (14). A variety of kinases are capable of phosphorylating τin vitro(20–23) and these may represent a target for AD pharmacotherapy (however, τ is hyperphosphorylated in normal human fetal brain without evidence of tangles) (8). The more likely immediate target is the plaques.


Amyloid (also ‘neuritic’ or ‘senile’) plaques are extracellular lesions consisting of a core of protein (Aβ) deposits surrounded by a ‘···corona of dystrophic neurites containing intracytoplasmic tau tangles’ (24) found diffusely throughout the brain, but notably in the cerebral cortex and hippocampus of AD patients (25–27). The centre core contains aggregates of Aβ protein (40- to 42-amino acids long) (28–30) which is derived by the proteolytic cleavage of a 695-residue transmembrane glycoprotein termed amyloid precursor protein (APP) (31). The manner in which proteases cleave APP determines the length of the amino acid segments (32). The majority of the Aβ produced is 40 amino acids long (Aβ40), soluble and presumably not neurotoxic, whereas the more hydrophobic (less soluble) 42 amino acid long Aβ (Aβ42) is less prevalent overall, but is more prevalent in the protein core of plaques (33–35). ‘diffuse plaques’ (36), which are focal diffuse deposits of amyloid without accompanying dystrophic neurites that might be early stage AD plaques, also contain primarily Aβ42 rather than Aβ40 (34, 37–39).

Whether plaques and tangles independently or jointly are the causative agents of neuronal degeneration or, instead, a marker of the disease is still an issue of some debate. Plaques and tangles may be present in other diseases and even normal aging (9), but AD does not develop in the absence of both. Plaques and tangles can occur independently of each other in diseases other than AD (19, 37, 40–43), but there is increasing evidence that Aβ and τ are linked in AD (37, 44, 45). There is some evidence that Aβ might precede τ formation. For example, τ formation in transgenic mice is influenced by Aβ (45, 46) and Aβ exists in the absence of τ in young Down syndrome patients, most notably in areas of the brain most affected by AD (37, 42, 47). There is evidence to support those who adhere to the ‘amyloid hypothesis’, that aggregation of Aβ and fibrils is a causative agent of cell death and ultimately AD progression (34, 48–51). Despite this ‘cause/effect’ debate, there appears to be a clear consensus that Aβ and AD are closely linked, making the understanding of Aβ, its formation, and pathology, important to understanding AD and potential treatment modalities.


The APP is enzymatically cleaved by three different forms of proteases (enzymes that break down protein molecules) named α-, β-, and γ-secretase (10). These proteases are a heterogeneous group located in various cellular regions. The order in which these proteases cleave APP determines whether Aβ is formed (48, 52). For example, cleavage of APP by α-secretase produces fragments that are substrates for γ-secretase, but processing of these fragments by γ-secretase does not lead to Aβ (48, 53). In contrast, cleavage of APP by β-secretase produces a peptide fragment that is a substrate for γ-secretase that is converted by γ-secretase to Aβ40 or Aβ42 (54) (Fig. 1). There is likely to be an additional step that determines which form of Aβ is produced (55).

Figure 1.

Amyloid β-peptide, the core protein of the characteristic amyloid plaques of AD, is formed from the proteolytic cleavage of its transmembrane precursor protein (APP) by ‘secretase’ enzymes. If APP is cleaved by α-secretase, Aβ is not formed and cannot be further processed to Aβ. In contrast, sequential cleavage of APP by β- and γ-secretase results in the production of the 40-amino acid long Aβ40 or the 42-amino acid long Aβ42 (depending on where the cut is made).


The cleavage of APP by α-secretase does not result in the formation of Aβ. Moreover, once cleaved by α-secretase, APP can no longer be processed to form Aβ by other enzymatic pathways. This suggests that processing of APP through this pathway not only does not lead to AD, but also may actually be neuroprotective. α-Secretase cleaves the APP protein on the C-terminal side of residue 16 of the Aβ sequence to form an 83-residue fragment called C83 (56). The C83 is a substrate for γ-secretase that produces as a final product a short peptide termed p3 (9). The function of p3 has not yet been determined, but to date has not been implicated in the formation of Aβ or in the progression of AD. Thus α-secretase cleavage of APP essentially precludes the formation of Aβ (57). Evidence for this derives from experiments in which mutations in APP near the site of α-cleavage lead to formation of greater amounts of Aβ(57–59). In addition, some research suggests that some product(s) of APP cleavage by α-secretase may block the neurotoxic effects of Aβ(60–62).

Several enzymes have been reported to have α-secretase activity (63, 64). Two enzymes in a disintegrin and metalloprotease (ADAM) family, ADAM-10 and ADAM-12 [or TACE, tumor necrosis factor-α (TNF-α) converting enzyme] either have inherent α-secretase activity or are thought to be somehow involved in the regulation of α-secretase (64, 65). Experimental cell lines deficient in only one of these enzymes (e.g. TACE) retain α-cleavage products of APP (64).


In order for Aβ to be formed, β-secretase must be the first enzyme to cleave APP. The product of β-secretase cleavage results in protein fragments that are processed by γ-secretase to Aβ. Thus an increase in β-secretase activity results in increased production of Aβ and may be responsible for progression of AD.

The product of β-secretase cleavage of APP is a membrane bound fragment (C99) of the original APP. The C99 fragment is a substrate for γ-secretase and is converted to the two versions of Aβ (Aβ40 and Aβ42) (10, 54). Thus, processing of APP by β-secretase increases the amount of substrate for γ-secretase that can lead to insoluble Aβ formation (presumably neurotoxic) and simultaneously decreases the amount of substrate available for processing by α-secretase to non-toxic products (55). The presence of Aβ in most cells of the body indicates that there is β-secretase activity in most cells (66). This suggests that β-secretase might have some role in the periphery, but studies in which mice are genetically altered to lack β-secretase activity do not display obvious negative developmental conditions despite less Aβ formation (67, 68). There is evidence that the β-secretase pathway is more active in brain tissues than in peripheral tissues (69).

Recently, a β-site APP-cleaving enzyme termed β-site APP-cleaving enzyme (BACE1) (also termed Asp2 or memapsin 2) that demonstrates all of the properties of β-secretase has been identified and cloned (55). Thus, BACE1 has been postulated to be the enzyme responsible for the first step in a process that ultimately results in Aβ formation (55) and thus inhibition of BACE1 is a target for drug-discovery efforts.


The final step in Aβ formation occurs when membrane-bound γ-secretase cleaves the membrane bound portion of APP (C99) that remains after β-secretase cleavage. The products of γ-secretase cleavage are two forms of Aβ which vary in length by two amino acid residues (Aβ40 and Aβ42). The insoluble Aβ42 is associated with AD pathogenesis (35).

The peptide products of APP catalytically cleaved by α or β-secretase (C83 and C99, respectively) are substrates for further proteolysis by γ-secretase enzymes (10; reviewed in 70). The γ-secretase cleaves C83 to non-toxic products. The γ-Secretase cleaves C99 to form either Aβ40 or Aβ42 (54). There might even be multiple γ-secretase enzymes with varying degrees of selectivity for producing Aβ40 or Aβ42 (9). Whether it is Aβ40 or Aβ42 that is produced also appears to be influenced by two membrane-associated proteins known as the presenilins (PS1 and PS2) (53, 71). Mutations in the genes for PS1 cause increases in Aβ42 in humans (38, 71, 72) and are correlated with the familial form of AD (73–75). PS1 or PS2 ‘knock-out’ mice (genetically altered to lack PS1 or PS2) also lack γ-secretase activity (76, 77), suggesting that presenilins are required for γ-secretase activity. Although the exact role of the presenilins in γ-secretase cleavage remains uncertain at this time, recent biochemical studies suggest that presenilins assist in γ-secretase activity (as chaperones or in the APP cleavage reaction) (70). Presenilins are also required for the proteolytic processing of ‘Notch’ protein, which like APP is membrane-associated and undergoes intramembrane proteolysis (78, 79). The Notch signalling pathway is critical during cell development (80). However, it appears that inhibition of γ-secretase sufficient to reduce Aβ production can be achieved without abolishing Notch activity (81, 82).

Present status

Basic science

In a general manner (10), α-, β- and γ-secretases all provide potential therapeutic targets for the treatment of AD because Aβ formation is a leading candidate for causation and progression of AD. The cleavage of APP by α-secretase prevents Aβ formation, whereas cleavage of APP by β-secretase followed by γ-secretase leads to the formation of the putatively toxic Aβ fragment (10). Hence, either enhancement of α-secretase activity or inhibition of β- and/or γ-secretase activity represent logical approaches. Enhancement of α-secretase activity has received less attention, although some work has been carried out. For example, Zhu et al. (83) report on protein kinase C (PKC) isozymes that promote α-secretase activity (thus reducing β-secretase mediated cleavage of APP) and Wolfe (9) cites earlier studies that evaluated the ability of compounds such as metabotropic glutamate receptors and muscarinic agonists (M1 and M3) to shift APP processing towards α-secretase mediated cleavage. Muscarinic agonists, however, have proven ineffective in early clinical trials (52) and, as with the other secretases, α-secretases may have multiple substrates (9), potentiality complicating the usefulness of non-selectively enhancing α-secretase activity. In addition, Nunan and Small (52) caution that the role of P3, a product of α-secretase cleavage, is unknown, and that α-secretase mediated APP cleavage has not been definitively proven to mitigate Aβ formation.

For several reasons, β- and γ-secretases are the subject of most intensive interest: (a) β-secretase is involved in the initial catalytic step in the formation of Aβ and its inhibition might result in a shift to more α-secretase processing of APP (9), (b) the structure of the enzyme-inhibitor complex has been elucidated and (c) it is easier to design enzyme inhibitors than activators. γ-Secretase may provide an equally attractive target. The APP mutations near the β-secretase site result in an increase of both Aβ40 and Aβ42, whereas mutations near the γ-secretase site lead only to enhanced levels of Aβ42 (10). The identity and structure of γ-secretase is still unknown, but mounting evidence points to a macromolecular complex involving presenilins (70). Presenilins may be the catalytic component of γ-secretase (10, 70) and mutations in PS1 and PS2 result in enhanced Aβ42 levels.

The ideal clinical candidate, in addition to being able to work in vitro, cross the blood brain barrier, and be taken up by cells, would be selective so that the functions of aspartyl proteases other than secretases are not adversely affected. One possible problem would be inhibition of Notch processing. Fortunately, there is some evidence that although β-secretase shows a high degree of structural similarity with other aspartyl proteases, there are some differences which might allow identification and development of a selective therapeutic agent. For example, unlike other human aspartyl proteases, the S2 and S4 regions are hydrophilic and open to solvent and the β-secretase backbone has a slightly different conformation than other aspartyl proteases (10).

Until recently, BACE was thought to be a relatively uncomplicated target. For example, Walter et al. (84) mention two studies that had not detected any deleterious phenotypes in BACE-deficient mice. At the time, APP was the only known substrate of BACE. A very recent study, however, suggests that decreased levels of BACE may lead to abnormal B cells by disruption of ST6Gal I processing via a putative second substrate (85).

A γ-secretase inhibitor may be more selective than a β-secretase inhibitor regarding Aβ formation. For example, Esler and Wolfe (10) summarized work that indicates APP mutations near the β-secretase site result in an increase in both Aβ40 and Aβ42, whereas mutations near the γ-secretase site lead to an increase in only Aβ42. There might also be more precise targets regarding γ-secretase: presenilins. Once thought to be the same, γ-secretase and presenilins are distinct entities (86), but mounting evidence points to an important role for presenilins in γ-secretase action: (a) presenilin-1 and -2 (PS1 and PS2) mutations lead to enhanced levels of Aβ42 (9, 10), (b) presenilin knockout mice have reduced amounts of Aβ40 and Aβ42 in neuronal tissue, but large amounts of the γ-secretase substrates C83 and C99 (9, 14, 87), and (c) PS1/PS2 double-knockout mice have no γ-secretase activity (9, 52). Although the mechanism of presenilin action is unknown, they may form the catalytic components of γ-secretase, they may act as a cofactor, or they may regulate the trafficking of the catalytic components (14, 87). Hence, PS1 and PS2 or presenilin co-factors provide potential therapeutic targets (10). A caveat is that there does not appear to be a correlation between high levels of γ-secretase activity and high levels of presenilins (88). Namely, over-expression of presenilins does not necessarily result in enhanced levels of γ-secretase mediated cleavage (9). Another possible contributor to γ-secretase action – and thus possible therapeutic target – may be the transmembrane protein nicastrin, which has been postulated to be a functional component of PS1 and PS2 complexes and appears to affect APP processing (14, 87, 89). Remaining hurdles include separation from Notch, or other critical enzyme, inhibition and possible deleterious effects of complete γ-secretase inhibition (9).

As noted by Nunan and Small (52), all of the current secretase research is based on the belief that AD and Aβ formation are intimately connected. α-Secretase enhancers and β- and γ-secretase inhibitors may provide a definitive test of this hypothesis.

Compounds in development

From information available publicly, some of the companies working on secretase inhibitors are targeting β-secretase whereas others are focusing on γ-secretase. Merck, DuPont, Elan, Amgen, Eli-Lilly, GlaxoSmithKline, and Bristol-Myers Squibb each have indicated that they are testing secretase inhibitors; Bristol-Myers Squibb is the only one presently known to be conducting human clinical trials with a γ-secretase inhibitor (90). The challenge for these companies is the development of secretase inhibitors that are highly selective for the targeted enzyme, can cross the blood–brain barrier, and can be readily taken up into neurons (10). It is also important that these compounds do not interfere with other signaling pathways or intracellular proteases. The compounds presently in development are reversible inhibitors and thus would attenuate, but not completely eliminate, enzymatic activity (14, 87).

β-Secretase inhibitors.

All of the β-secretase (BACE) inhibitors reported to date contain a similar peptidomimetic structural backbone (peptide analogs) (Fig. 2). The first of these compounds was based on the so-called ‘Swedish mutant’ of APP (a double mutation around the cleavage site of β-secretase), but it had only relatively poor β-secretase inhibitory activity (9). Replacement of one of the diastereomers (S- with R-statine) led to loss of inhibitory activity, demonstrating that stereoselective binding is involved (9). Two newer compounds added an amino acid (alanine) and a hydroxyethylene transition state analog isostere, but selectivity for β-secretase vs. other enzymes declined with these larger compounds (91). Additional structural refinement may make these leads more viable. Toward this end, the crystal structure of the complex of one of these compounds bound to β-secretase has been determined and should aid in the structure-based design of new inhibitors (91, 92). Similarly, β-secretase shares a common backbone structure with pepsin, but characteristics unique to β-secretase may help in designing inhibitors (9).

Figure 2.

Examples of β-secretase inhibitors, with analogs indicated by shaded spheres. The compounds were designed to mimic transition-state intermediates in the reaction catalyzed by β-secretase.

Selectivity is an important consideration. For example, blocking the β-secretase homolog BACE2, an enzyme that is expressed in highly vascularized tissues and believed to have an essential physiologic role (52), might have toxic consequences, so it will be necessary to develop drugs that selectively block BACE1 and not BACE2 (10). The issue of possible side-effects of β-secretase inhibitors is considered later in this review.

γ-Secretase inhibitors.

γ-Secretase is suspected of being ‘···central to AD pathogenesis, because of its importance in determining the ratio of Aβ40 to Aβ42’ (10). Hence, inhibition of γ-secretase activity is an enticing target and many different approaches have been taken to develop γ-secretase inhibitors. Some of these efforts are summarized below.

Higaki et al. (93) reported the first γ-secretase inhibitors, peptide aldehydes originally designed to be calpain inhibitors (Fig. 3a). These compounds relatively selectively blocked Aβ40 vs. Aβ42 production, demonstrating the selectivity of γ-secretase (94).

Figure 3.

Examples of γ-secretase inhibitors, with analogs indicated by shaded spheres (see text for details)

The first substrate-based inhibitor (Fig. 3b) was reported by Wolfe et al. (95). This compound, known as MW167, had poor potency and did not have selectivity for Aβ42 (9). Moore et al. (96) developed a series of related compounds using varying substituents and found that bulky substituents, which increased steric hindrance, also increased potency. However, these compounds did not have selectivity for Aβ42. Murphy et al. (97) reported a substrate-based peptide aldehyde, known as cbz-GVV-CHO, that selectively inhibits Aβ40 production.

Compounds developed by Seiffert et al. (98) (Fig. 3c) contain a 2,3-dialkylated succinamide. Another potent compound (Fig. 3d) resulted from replacing the lactam ring with a benzodiazepene moiety (9). Replacement of the succinamide moiety with difluorophenylacetyalanine retains inhibitor action in whole cells (9).

Transition state analogs can serve as useful molecular templates for designing enzyme inhibitors. Li et al. (99) have utilized this approach to develop compounds (Fig. 3e) designed to function as transition state analog inhibitors directed to the active site of an aspartyl protease. More recently, Moore et al. (100) developed difluoro ketone transition state mimics that effectively block Aβ40 and Aβ42 production. However, subinhibitory concentrations of these compounds actually increased Aβ42 production (100) and to date all other non-selective γ-secretase inhibitors also demonstrate this phenomenon.

Murphy et al. (97) have developed a dipeptide epoxide inhibitor, known as cbz-IL-epoxide, based on the Aβ40-selective inhibitor originally reported by Higaki et al. (101). Cbz-IL-epoxide is a potent inhibitor of Aβ.

Dovey et al. (102) have developed a N-[N-(3, 5-difluorophenylacetyl)-l-anayl]-S-phenylglycine t-butyl ester, known as DAPT. DAPT reduces brain Aβ production in transgenic mice within 3 h after oral administration. No signs of toxicity in vivo or in vitro were noted (102).

A concern is that γ-secretase inhibitors may also block Notch protein, which is essential for mammalian development. At present, it is not clear what the consequences of interfering with Notch signalling would be, but in vivo studies of viable γ-secretase inhibitor candidates may help answer this question (10). It seems important to try to develop compounds that are selective for APP over Notch.

Other approaches

Inhibition of secretase (s) might not be sufficient, by itself, to prevent or delay the onset or progression of AD. A wide variety of other therapeutic approaches might be used in conjunction with secretase inhibitors. Therapies which may be useful include acetylcholine esterase (AChE) inhibitors, non-steroidal anti-inflammatory drugs (NSAIDs), estrogens, antioxidants, N-methyl-d-aspartate (NMDA) antagonists, ampakines, nerve growth factor (NGF) agonists, cholinergic M1 or M3 subtype-selective agonists, serotonin 5-HT6 subtype-selective antagonists, a vaccine, calcium channel blockers, angiotensin converting enzyme (ACE) inhibitors, acetyl-l–carnitine, Ginkgo biloba, and others (4, 8, 10, 14, 19, 51, 87, 103–110) (Table 1).

Table 1.  Present and possible future pharmacotherapy for Alzheimer's disease (AD)
?Acetylcholine esterase (AChE) inhibitors increase synaptic levels of acetylcholine and counteract the disproportionate ACh decline in brain regions associated with memory (such as the nu. basalis of Meynert).
?There might be an important inflammatory component to AD, since neuritic plaques contain acute phase inflammatory proteins (103) and individuals who use non-steroidal anti-inflammatory drugs (NSAIDs) for at least 2 years are less likely to develop AD then those who do not (104). It has recently been reported that a subset of NSAIDs may selectively reduce the production of Aβ42 (105). Likewise, COX-2 inhibitors were recommended for clinical trials for use in AD (106).
?Estrogen can improve cerebral blood flow, act as an antioxidant, and help repair damaged neurons. It has been reported that oestrogen reduces neuronal generation of Aβ (107). An oestrogen agonist, Neurestrol® (Endocin), is currently in development to treat female AD patients (106).
?Al least two antioxidants are in development: ARL-16556 acts as a free radical scavenger and modulates the effects of nitric oxide (NO); MDL-7418DDA acts as a vitamin E (α-tocoppherol) analogue that inhibits lipid oxidation (106). However, the effectiveness of antioxidants in the treatment of AD remains unsubstantiated.
?The decline in cognitive functioning in AD patients may be due to changes in excitatory neurotransmitter, particularly glutamate, activity (108). Under investigation are NMDA antagonists such as memantadine, dextromethorphan, nitroglycerine, L-701252, LY235959 (a competitive inhibitor) and WIN-63480-2 (a non-competitive inhibitor) (106), and ‘ampakines’ (Cortex Pharmaceuticals and The University of California Irvine) that have been shown to stimulate AMPA receptors and to reverse memory deficit in old rats (8).
?Nerve growth factor (NGF) has neuroprotective properties and is required for maintaining neuronal connections. Potential drug therapies in development include AIT-082 Neotrofin® (enhances NGF), AK-30-NGF (monoclonal antibody delivery system), NBI-106 (immune stimulator and memory enhancer), and rbNGF (recombinant protein) (106). These approaches would presumably continue to operate during new injury from cytotoxic effects of Aβ (14, 87).
?Selective M1 or M3 muscarinic cholinergic receptor agonists have been shown to decrease Aβ production (10) and drugs such as arecoline, xanomeline, and oxotremorine have shown some promising results in clinical trials (109).
?A 5-HT6 receptor antagonist, SB-271046 (GlaxoSmithKline) developed to help improve cognitive impairment, is currently in Phase I clinical trials.
?Excess intracellular calcium is cytotoxic and calcium enters cells more readily as people age therefore blocking calcium channels may be beneficial.
?Angiotension II is increased in AD patients and is known to impair memory. ACE inhibitors may serve to improve cognitive function.
?Two studies in 1991 demonstrated that long-term use of acetyl-l-carnitine improves cognitive function and may slow disease progression (19), suggesting that drugs that enhance acetyl-l-carnitine levels or action might be beneficial.
?An extract of Ginkgo biloba, Egb 761, which might have antioxidant (free-radical scavenging), anti-inflammatory, and neuromodulatory properties (4) was reported to improve cognitive function vs. placebo in a double blind, randomized trial of 309 patients (110).
?A vaccine, AN-1792 (Elan), prevented the development of amyloid plaques in PDAPP mice (mice genetically engineered to develop AD) and reduced the progression of AD-like pathology in older mice (51). Human trials are underway.


In relation to cure


Since the first description by Alzheimer of the characteristic histopathological features of the disease that has come to bear his name, AD has been associated with amyloid plaques and neurofibrillary tangles. The evidence that links Aβ peptide with AD includes the following (summarized in 8): (a) Aβ is a defining characteristic of AD at autopsy and is the definitive diagnostic criterion – the absence of Aβ excludes a confirmatory diagnosis of AD; (b) Aβ-containing neuritic plaques can occupy as much as 50% of brain volume in AD patients, possibly sufficient by itself to contribute to AD symptomatology; (c) in AD patients Aβ plaques are found aggregated in brain regions in which neuronal degeneration is observed and in regions known to be associated with the cognitive functions that are impaired in AD (such as memory and organizational skills); (d) cultured neuronal cells die when exposed to excess Aβ; (e) a plot of incidence of AD as a function of age of persons with Down syndrome is parallel to, and precedes by about 30–40 years, a similar plot of persons who do not have Down syndrome (virtually all Down syndrome patients over the age of 35–40 have massive deposits of Aβ in brain regions affected in AD) and (f) ‘familial’ AD has been traced to mutations of genes that encode for Aβ (e.g. on chromosomes 1, 14, and 21).

Once formed, amyloid plaques and their core of Aβ appear to be relatively resistant to reversal or removal. The biochemical processes that normally perform this function, if they exist, are not known. This leaves the production of Aβ as the major present target for consideration as therapeutic intervention. As Aβ is formed from the cleavage of its precursor protein (APP) by secretases, the secretase family of enzymes, and the inhibition of their action, has become a major focus of attention. Unfortunately, in the majority of cases (but not all), the symptoms of AD do not appear until Aβ deposition in brain is extensive. Hence, in the present circumstances in which no pre-mortem diagnostic tool exists, substantial quantities of Aβ have already formed by the time AD is suspected or a diagnosis is made. By this time, the amount of Aβ already deposited may be sufficient to lead to an inexorable decline in neuronal viability and ultimate demise. In such situations the reduction, or even complete elimination, of Aβ production would not yield a ‘cure’, or possibly even much therapeutic benefit.


The scenario for the prevention, or early intervention, of Aβ deposition in persons who are pre-disposed to AD or suspected of having early signs of the disease is quite different from that presented above for reversal or cure. If the Aβ deposition does not per se lead to irreversible changes in other regulatory mechanisms, and if quantity of Aβ is in any way correlated to disease progression, then progression might be halted – or at least slowed – by inhibition of further Aβ production. In this regard, it is consequential that the product (s) that are formed by the enzymatic cleavage of APP depends on the type of secretase (s) involved. For example, Aβ is not formed by the action of α-secretase (TACE) on APP and, thus, inhibition of α-secretase is not a prime strategy. In contrast, the β-secretase (BACE1) and γ-secretase cleavage pathway of APP does lead to Aβ and inhibitors of these two enzymes are presently being designed and tested by several academic laboratories and pharmaceutical companies.

If the concept of giving secretase-inhibitors as preventative proves to be a viable strategy, then questions arise regarding the appropriate time that administration of such drugs should be initiated. For example, should all persons over some pre-specified age (perhaps over 60 years old) be given the drug (s)? Or should they be withheld until some suspicion of AD arises? What about members of families with the ‘familial’ form of the disease, should they be started earlier in their lives, perhaps as prophylactic treatment? Should all persons with Down syndrome be put on the drugs from birth? The answers to these questions await a better understanding of the consequences of inhibiting secretase activity. The secretase enzymes normally perform necessary functions. What will be the consequences of inhibiting these actions? If β- and γ-secretases are only important during development, then their inhibition later in life will not be problematic. If, however, they are needed later in life, inhibition of their actions will be accompanied by adverse effects that might be deemed unacceptable until serious AD symptoms develop.

In relation to present therapy

Novel inhibitors of β- and γ-secretases are being developed and some are in early clinical trials. If they prove to be safe and efficacious throughout the remaining clinical trials, they will enter into widespread clinical use. They will represent a dramatic advancement in the drug treatment of AD. For example, in USA currently only four drugs are approved by the FDA for the treatment of mild to moderate dementia. All are AChE inhibitors: tacrine (Cognex®), the oldest of the four, is a reversible inhibitor not often used as first line therapy because of inconvenient times–times daily dosing and possible association with hepatotoxicity; donepezil (Aricept®) is a ‘second-generation’, reversible inhibitor devoid of hepatotoxic effects and more convenient once-daily dosing; rivastigmine (Exelon®), a ‘pseudoirreversible’ inhibitor for which proper dose-titration is recommended to avoid gastrointestinal upset, has the advantage of being available in both capsule and solution formulations; and galantamine (Reminyl®), the most recent addition to the market, is a selective, reversible, and competitive inhibitor of about equal clinical efficacy to the others (111). Thus, the approval of β- or γ-secretase inhibitors would represent a major breakthrough.

In relation to present theory

The other possible outcome of the clinical trials, that β- or γ-secretase inhibitors are not efficacious in AD, would be just as dramatic. It would mean, at best, that the present understanding of the biochemical pathways leading to the formation of Aβ, or the regulatory mechanisms involved in the process, is incomplete. At worst, it would mean that the present prevailing theory of AD – that Aβ production plays a critical role in its etiology – would come under question. Previous uncertainty about the link between Aβ peptide with AD includes the following (summarized in 8): (a) Aβ deposits are present in some elderly people who have normal cognitive function (although it can be argued that the amounts are insufficient, or in the wrong brain regions, to produce symptoms); (b) certain Aβ deposits (such as in the cerebellum) are not associated with nerve cell damage or decrement in function (such as problems of movement typical of patients with cerebellar damage); (c) a general lack of strong correlation between the number of Aβ deposits and the severity of AD symptoms (although, again, it can be argued that it is the brain region and the patient's baseline cognitive abilities that are the critical factors). In any case, the discovery that β- or γ-secretase inhibitors do not alter the progression of AD would necessitate a significant reassessment of the present view of the disease.

In relation to ADR's

The adverse drug reactions that might result from inhibition of β- or γ-secretases are not well known. The effects of chronic exposure, particularly if drug treatment is to be initiated early in life as a preventative measure, are not easily measured. Fortunately, early results from studies of BACE1 ‘knock-out’ mice (112, 113) do not indicate abnormalities in gross anatomy, tissue histology, or clinical chemistry. Nevertheless, the profile of individual compounds must be determined in clinical trials.


The fundamental principle upon which much of the current theory about AD and its treatment is based is that the Aβ located in plaques is intimately involved and that the prevention of Aβ deposition would prevent, or at least suspend, the inexorable decline in neuronal and clinical functioning. In the absence of much information about biochemical processes that might chemically scavenge plaques already formed, efforts have focused on preventing the formation of plaques and, in particular, its core constituent Aβ. As the primary source of Aβ is believed to be the cleavage of APP by β- and γ-secretase, these enzymes have become major targets for drug development. It will be some time before the results of clinical trials are known, but there are seemingly only two possible outcomes, both equally dramatic. If the results of the clinical trials are positive, the drugs will validate the prevailing theory of AD and will provide a novel strategy for its therapy using drugs. If the results of the clinical trials are negative, not only will it represent a clinical setback, but it will call into question the core concepts upon which much of the current theories of AD are based. Seldom does so much hang in the balance.