ALZHEIMER'S DISEASE 187 Unresolved Issues CAA Insoluble A~ plaque Soluble AI3 4 X~ Microvascular damage ~d flow Disturbances Cell death ,S x, NFT Inflammation ' FIG. 4. Many unresolved issues continue to plague our understanding of the pathogene- sis of AD. Although it is known that neuronal death can occur due to NFT formation and cerebrovascular disease, such as cerebral amyloid angiopathy (CAA), altered perfusion, or mi- crovascular pathology (pale arrows), the exact role of AB and inflammation are unknown (dark arrows). A/3 deposition and inflammation are both universal findings in the brain of AD, and the former is necessary for a pathological diagnosis of AD. However, whether either results in neuronal death, and if so by what mechanism, is yet to be determined. major mechanism of neuronal degeneration occurs in AD. Unfortunately, the lack of a readily identifiable marker for this neuronal loss has made the identification of its cause extremely difficult. A study has shown that the prolyl isomerase, Pinl, is sequestered into the NFT and depleted in the brains of AD patients (Lu et al., 1999). 188 JILLIAN J. KRIL AND GLENDA M. HALLIDAY Depletion of Pinl may induce neuron death via mitotic arrest and apopto- sis prior to the development of NFTs (Lu et al., 1996). The neuron-specific activator for cell proteins involved in the mitotic cycle is p35, which is pro- teolytically cleaved to produce p25, a fragment found to accumulate in the brains of patients with AD (Patrick et al., 1999). Application of Aft 1-42 in- duces the conversion of p35 to p25 (Lee et al., 2000). p25 links with cell cycle-dependent kinase 5 to hyperphosphorylate tau and promote apopto- sis (Lee et al., 2000). Degenerating neurons in APP V717F Aft-producing transgenic mice show chromatin segmentation and condensation, as well as increased TUNEL staining, suggestive of apoptosis (Nijhawan et aL, 2000). This supports a link between Aft deposition and apoptosis (Fig. 4). Increased TUNEL staining (Druganow et al., 1995; Lassmann et al., 1995; Smale et al., 1995; Bancher et al., 1997), as well as cleaved caspase 3 (Selznick et al., 1999; Stadelmann et al., 1999), an enzymatic marker of apoptosis, are found in vulnerable brain regions in AD. APP has been identified as a specific sub- strate for caspase 3 with the resultant peptides (including Aft), inducing apoptosis (Gervais et al., 1999). Other apoptotic-specific caspases can also cleave APP (Pellegrini et al., 1999), and the resultant C-terminal fragment from such cleavage has been called C31 (Lu et aL, 2000). C31 is also a potent inducer of apoptosis and was found in the brain of patients with AD (Lu et al., 2000), whereas caspase deficient mice are resistant to this form of cell death (Nakagawa et al., 2000). Despite these studies that suggest apoptosis occurs in AD, apoptotic bod- ies and blebbing are not features of AD neuronal degeneration. In addition, the time sequence of such events remains to be determined. The chronic nature of the neurodegeneration in AD does not fit well with the more rapid time course of apoptosis, which is believed to take only weeks or months at most (Stadelmann et al., 1999). Other mechanisms of neuronal death, such as necrosis, were also demonstrated in AD (Wolozin and Behl, 2000b). In- deed, the same triggers may cause either apoptosis or necrosis, including Aft toxicity, oxidative stress, excitotoxicity, ischemia, and removal of trophic fac- tors. The distinction between apoptotic and necrotic mechanisms, however, may be somewhat false given that neurons may begin with necrosis and then convert to apoptosis or alternatively begin with apoptosis and then undergo necrosis (Wolozin and Behl, 2000b). Although Aft plaques are necessary for a diagnosis of AD, like NFTs, they are poorly related to the degree of neuronal loss. However, studies suggest the intracellular accumulation of Aft may be neurotoxic (Fig. 4). There is an additional site of APP cleavage within the endoplasmic reticulum that gives rise to intracellular Aft 1-42/43, which over time reaches the concentration necessary for fibril formation (Hartmann, 1999; Wilson et al., 1999). Cell rupture would release this intracellular Aft into the surrounding extracellu- lar milieu, which could stimulate further amyloid deposition. Although most ALZHEIMER'S DISEASE 189 cell types express APE neurons produce the highest amount and preferen- tially use the intracellular pathways for Aft production (Hartmann, 1999). It is difficult to know how to prove or refute this model of AD neuronal vulnera- bility, although it is of interest that Aft is not deposited within the vulnerable hippocampal formation or entorhinal cortex (Arnold et al., 1991) and no neuronal loss occurs in elderly APP-transgenic mice who show considerable A/3 deposits (Irizarry et al., 1997a, 1997b). Interestingly, a study identified nonpyramidal neurons containing Aft 1-42 around amyloid plaques in AD patients (Mochizuki et al., 2000), suggesting preserved neurons may con- centrate these peptides intracellularly. In contrast, a number of studies suggest that soluble A/3, and particularly Aft 1-40, is synaptotoxic without causing plaque formation or overt cell death (Mucke et al., 2000). Reductions in soluble A/31 40 concentrations correlate with synaptic loss in patients with AD (Lue et al., 1999). Interestingly, in the same patients, soluble A/31-40 levels correlate with cerebrovascular amyloid angiopathy and ApoEe4 allele frequency (Lue et al., 1999), suggesting a greater influence on vascular changes than neuronal degeneration. E SUMMARY Taken together, these studies suggest a multifactorial origin of neuronal loss in AD where a number of primary and secondary factors may cause neuronal death (Fig. 4). More work is needed to link all the potential cellular events that underlie the clinical symptoms of AD. At present, we do not have a good understanding of the association between A/3 deposition (required for a diagnosis of AD) and the degenerative process. The link between soluble A/3 and brain atrophy needs to be clarified, and mechanisms of cell death other than NFT formation (and possibly apoptosis) need to be elucidated. It will be important to determine the time sequence of these events to target appropriate therapeutic measures. IV. Genetic Influences As many as 50% of patients with AD have at least one first-degree relative with dementia (Writing Committee Lancet Conference 1996, 1996), and numerous studies have investigated family history as a risk factor for AD. Nine of the 14 case control studies reviewed byJorm (1990) showed a sig- nificantly increased risk of AD in subjects with a positive family history. The odds ratios ranged from 2.1 to 9.9 and reflect data obtained from prevalence and incidence studies. 190 JILLIAN J. KRIL AND GLENDA M. HALLIDA¥ A. DOMINANT INHERITANCE It is estimated that between 5% and 10% of AD cases have a demon- strable pattern of inheritance. These cases, although rare, provide valuable insights into the pathogenesis of AD. To date, three genes have been iden- tified. These are APP mutations on chromosome 21, presenilin-1 (PS-1) mutations on chromosome 14, and presenilin-2 (PS-2) mutations on chro- mosome 1. Each of these genes have an autosomal dominant pattern of inheritance, although PS-2 does not appear to have complete penetrance (St. George-Hyslop, 2000). These three identified genes do not fully account for all autosomal dominant cases of AD, suggesting other genes are yet to be identified. The APP gene encodes a transmembrane protein of 770 amino acids from which Aft is derived (see section III.C above). The normal function of APP is not known, although it is highly conserved and expressed ubiqui- tously. In addition to AD, mutations in APP can also result in hereditary cere- bral haemorrhage with amyloidosis-Dutch type (HCHWA-D). Mutations in the APP gene are mostly located in or around the amyloidogenic portion of the molecule, especially near the three secretase sites. Mutations in the PS-1 gene are the most common of the early-onset familial AD mutations, accounting for 30-50% of all autosomal dominant cases. PS-1 is a transmembrane protein that is also expressed ubiquitously and has six or eight transmembrane domains (Checler, 1999). There is an increasing body of evidence that suggests the presenilins function as the y-secretase, or in close association with y-secretase, in the production of A/3 (Checler, 1999; Ray et al., 1999; Wolfe et al., 1999) and thus increase the pro- duction Afll-42/43. More than 50 mutations in PS-1 have been identified. The majority of these are missense mutations and are scattered throughout the molecule. In addition, a number of splice acceptor mutations that cause the deletion of the sequence encoded by exon 9 were also described (Kwok et al., 1997; Crook et al., 1998; Smith et al., 2001). A proportion of PS-1 mutations with a deletion of exon 9 have AD with spastic paraparesis (SP; Crook et al., 1998; Verkkoniemi et al., 2000). In AD+SP, there is progressive weakness and wasting of the lower extremities and a later age of onset of dementia has been described in some of these families (Smith et al., 2001). The pathology of exon 9 mutations is also interesting in that very large, noncored, and faintly neuritic plaques are described (Crook et al., 1998; Smith et al., 2001). These have been termed "cotton-wool" plaques because of their size and uniform appearance (Fig. 5). The PS-2 gene encodes a transmembrane protein that is 67% homol- ogous to PS-1 (Checler, 1999). Unlike APP and PS-1, PS-2 is expressed more strongly in peripheral tissues (pancreas, cardiac, and skeletal muscle) ALZHEIMER'S DISEASE 191 FIG. 5. Photomicrographs of the temporal neocortex of a patient with a presenilin-1 (PS-1) mutation. In the upper panel, both neuritic (arrows) and diffuse plaques can be seen. The diffuse plaques (inset) in these patients are unusual because they are large, only faintly neu- ritic, and lack cores. They have been termed "cotton wool" plaques and are tound exclusively in patients with PS-I mutations. than in the brain (St. George-Hyslop, 2000). A small number of families with missense mutations in PS-2 have been identified, indicating they are much rarer than PS-1 mutations. The exact mechanism by which PS-2 mutations cause AD is unclear, although because of its sequence homology with PS-1, it is believe to have a similar function. 192 JILLIAN J. KRIL AND GLENDA M. HALLIDAY The mechanism common to the known mutations is an increased pro- duction of Aft 1-42/43 and an increased rate of aggregation of Aft plaques (see Wolozin and Behl, 2000a, for commentary). However, it appears that the PS mutations may also be involved in other aspects of the pathology of AD by participating in cell death due to apoptosis and in the phos- phorylation of tau (see Checler, 1999; Czech et al., 2000). Our knowledge of AD has advanced substantially since the identification of the mutations responsible for familial forms of AD and of the presenilins in particular. This rapidly moving field of research provides valuable insights into the disease processes and has the potential for the development of strategies for ther- apeutic intervention. However, it is still unknown whether the knowledge gained from studying these cases is generally applicable to the majority of AD patients. In addition, the knowledge gained has still not elucidated the cause(s) of sporadic AD. B. GENETIC RISK FACTORS Apart from dominant inheritance, the clustering of dementia within families must be viewed as evidence for the role of an individual's genotype in determining their risk of AD. The apolipoprotein E (ApoE) gene, found on chromosome 19, encodes three isoforms of e2, e3, and e4, and the presence of the e4 allele has been found to increase the risk ofAD (Katzman, 1994; Strittmatter and Roses, 1995). ApoE is involved in lipid transport and is present in the serum (Uterman, 1994). An association between ApoE e4 and AD was first described in 1993 in both sporadic (Saunders et al., 1993) and familial (Corder et al., 1993) AD. It has subsequently been confirmed in many other studies of early- and late-onset AD and a variety of other neurological diseases, including other dementias (e.g., Roses, 1996; Stevens et al., 1997; Horsburgh et al., 2000). In addition, an allelic dose dependence has been shown where subjects who are homozygous for e4 have a greater risk of AD at an earlier age than those who are heterozygous (Corder et al., 1993). In this study of families with late-onset AD, subjects with no e4 had a mean age of onset of 84.3 years compared with 75.5 years in those with one s4 allele and 68.4 years with two alleles. In addition to its effect on age of onset, ApoE genotype has also been shown to influence, albeit variably, the response to drug treatment. A poorer response to the cholinesterase inhibitor tacrine has been shown in patients with AD who possess the ApoE e4 allele than those who do not (Poirier et al., 1995), although this effect has not been found in all studies (MacGowan et al., 1998). In addition, only patients with e4 showed improvement in ALZHEIMER'S DISEASE 193 cognitive performance when treated with a drug that facilitiates noradrener- gic and vasopressinergic activity in the brain (Richard et al., 1997). Some de- bate also exists over whether ApoE genotype modifies the type or amount of AD pathology in individuals carrying the e4 allele. Several studies (Schmechel et al., 1993; Nagy et al., 1995; Overmyer et al., 1999), but not all (Morris et al., 1995; Landen et al., 1996), have found an increase in the density of neu- rofibrillary tangles and senile plaques in AD. Moreover, the correlation with brain pathology is further complicated by the finding that normal subjects in their forties and older who possess an e4 allele have smaller right hip- pocampi than those without e4 (Tohgi et al., 1997). It is unclear whether this finding represents a lifelong trait or is an indicator of "preclinical" AD. Longitudinal studies on such groups of subjects are necessary to clarify this issue. In patients with AD, greater brain atrophy (Lehtovirta et al., 1995; Juottonen et al., 1998b) and an increased rate of atrophy has been found in individuals with e4 (Wahlund et al., 1999). However, this association has not been found in all studies (Barber et al., 1999). The mechanism of action of ApoE is not fully elucidated. ApoE is in- volved in the regulation of the transport of cholesterol and phospholipid and has an important role in the distribution of these molecules during peri- ods of membrane remodeling, such as synaptic plasticity and membrane re- pair. In addition, ApoE-lipid complexes are believed to assist in the removal of Aft via the low-density lipoprotein-related receptor (Wolozin and Behl, 2000a). Isoform differences in the behavior of ApoE have been identified (e.g., Strittmatter et al., 1993; Nathan et al., 1994), and these are believed to underlie the susceptibility to AD in individuals with the e4 allele (Horsburgh et al., 2000; Wolozin and Behl, 2000a). C. SUMMARY In addition to these genetic factors, other modifying influences have been identified (e.g., HLA, butyrylcholinesterase K, ~ 1 antichymotrypsin) ; however, the exact nature of the relationship between genotype and disease susceptibility remains obscure. Although there is strong evidence for an as- sociation between ApoE e4 and AD, the presence of e4 is not causative or is it necessary to develop AD. For these reasons, it is recommended that ApoE not be used for predictive testing (American College of Medical Genetics/American Society of Human Genetics Working Group on ApoE and Alzhemer's Disease, 1995). Similar results are likely for other genetic risk factors. Nevertheless, such genotypes are important variables to be considered in research studies examining aspects of the pathogenesis 194 JILLIAN J. KRIL AND GLENDA M. HALLIDAY and progression of AD, especially as reports of monozygotic twins that are discordant for AD (Creasey et al., 1989) suggest that inheritability is not solely responsible for one's risk of AD. Few studies are integrating the multi- ple genotype analyses required to understand genetic versus environmental influences. V. Inflammation and Anti-inflammatory Drugs Numerous lines of evidence suggest a link between brain inflammation and AD (see Gahtan and Overmier, 1999; Halliday et al., 2000a). Initial ev- idence from clinical studies for a role of anti-inflammatory drugs in the prevention of AD came from case control studies that examined arthritis as a risk factor and found a reduced risk of dementia in patients who con- sumed anti-inflammatory drugs (Broe et al., 1990; Breitner, 1996). However, a number of similar studies were unable to identify a significant reduction in risk (e.g., Heyman et al., 1984). This inconsistency may reflect the relatively small samples examined in each study individually because a meta-analysis of 17 studies showed a reduced risk of AD dementia in patients taking both steroidal and nonsteroidal anti-inflammatory drugs (NSAIDs; McGeer et al., 1996). It should be noted, however, that the majority of these studies were of cross-sectional design where significant biases exist in selection of cases for study and the reporting of drug use (Stewart et al., 1997). Antigens of the major histocompatibility complex are intimately associ- ated with inflammation and polymorphisms of the genes encoding these proteins have been associated with an increased risk of disease. In partic- ular, CNS and peripheral diseases with an inflammatory basis occur more commonly in subjects who have a particular HLA genotype; notable among these is the association between rheumatoid arthritis and HLA-DR4 (Khan et al., 1983; Stastny et al., 1988). A number of different associations were de- scribed between AD and HLA alleles. In late-onset patients who do not have ApoE e4 alleles, an increased risk of AD was found in patients with HLA- DR1, 2, or 3, and a reduced risk was found in patients with HLA-DR4 or 6 (Curran et al., 1997). However, these findings were not replicated by others (Middleton et al., 1999b), or only partly replicated (Neill et al., 1999), and the converse relationship (HLA-DR3 is protective) was found in a study of autopsy-confirmed cases ofAD (Culpin et al., 1999). In addition, an earlier age of onset by 3 years has been reported in subjects with HLA-A2 com- pared with other alleles (Payami et al., 1997; Combarros et al., 1998), and when the patient's ApoE status was examined, the effect of HLA-A2 and ALZHEIMER'S DISEASE 195 ApoE e4 appeared to be additive (Payami et al., 1997). A similar additive effect of HLA-A2 and ApoE e4 has been found in early-onset familial AD (Ballerini et al., 1999). Other associations with HLA alleles were reported (Small et al., 1991; Middleton et al., 1999a), but these studies are yet to be replicated. It is therefore unclear whether the initial studies implicat- ing anti-inflammatory medications as protective for AD are due to a direct effect on brain inflammation or are associated with genotype and disease susceptibility. To date, there have been only three longitudinal studies analyzing the question of drug protection in AD. Two of these studies (Stewart et al., 1997; Prince et aL, 1998) found a beneficial effect of NSMDs. The Baltimore Longitudinal Study of Aging found a reduced risk of AD among users of NSMDs and aspirin, which was increased the longer the drugs were used (Stewart et al., 1997). Prince and colleagues (1998) showed less decline in some tests of cognitive function in NSMD users, although the benefit was reduced in older subjects. In contrast, a study of Australians ages 70 years or older (mean age of 80) found that NSMDs or aspirin provided no protection against cognitive decline or incidence of dementia over a 3- to 4-year period (Henderson et al., 1997). Taken together, these studies suggest that some protection is conferred at ages when susceptibility is relatively low. It may be that sufficient protection occurs only with long-term drug usage. It is therefore not surprising that clinical trials aimed at assessing the role of NSMDs in preventing AD produced conflicting results. Rogers and col- leagues (1993) performed a study ofindomethacin in 28 patients and found a small but significant slowing of cognitive decline in the treated patients. Conversely, Scharf and colleagues (1999) used an NSMD in combination with a gastroprotective agent and found no difference between groups in measures of cognitive performance. Drop-out rates in both studies were considerable (up to 50%) and follow-up times short (around 6 months), so neither study can be considered conclusive. Nevertheless, on cross-sectional analysis cognitive performance is improved in AD patients taking NSMDs and aspirin (Broe et al., 2000) compared with their nontreated counter- parts. Interestingly, this effect was present at low doses of aspirin, which are not considered to be anti-inflammatory suggesting the effect of these drug is not through reducing inflammation but through some other, possibly peripheral mechanism (Broe et al., 2000). Neuropathological studies demonstrated a close relationship between Aft plaques and both reactive astrocytes and microglia (Rozemuller et al., 1992; McGeer and McGeer, 1995; Halliday et al., 2000b). Mthough a glial response might be expected to occur secondary to the degeneration in AD, evidence suggests the inflammatory response itself may contribute to the 196 JILLIAN J. KRIL AND GLENDA M. HALLIDAY pathology of AD. Many of the proteins of the complement pathway, together with acute phase proteins, are found in Aft plaques (see Walker, 1998) and are believed to be synthesized by microglia. In addition, activated microglia synthesize and excrete a number of inflammation-related substances that have been shown to be neurotoxic in rats (Weldon et al., 1998), and it has been suggested that microglia might facilitate Aft deposition (see Gahtan and Overmier, 1999). Overall, the data show that patients with AD have an active immune response in the brain. An age-related increase in inflammatory microglia has also been found (Mattiace et al., 1990; Mackenzie and Munoz, 1998) which may reflect the brain's reaction to the increased AD-type pathology in aging or, alternatively, indicate changes to the immune status of the elderly brain. Interestingly, this age-associated increase in activated microglia is ameliorated by NSAID use (Mackenzie and Munoz, 1998), unlike AD patients where NSAID use does not decrease inflammation (Halliday et al., 2000b). This suggests the disease process itself stimulates an immune response. Whether inflammation is a primary cause for the neurodegeneration in AD or a secondary event to aid in its clearance is still unclear because the sequence of these events is still poorly understood (Fig. 4). Although some epidemiological and clin- ical evidence suggests a beneficial effect of treatment with NSAIDs, other research suggests any such benefit is mediated through a noninflammatory mechanism (Broe et al., 2000; Halliday et al., 2000b). A clearer picture of the sequence of the early and subsequent cellular events in patients with AD would help clarify any direct role of inflammation in the disease process. The enhanced immune response in AD patients is now being used for a new type of treatment, Aft peptide immunization (Schenk et al., 2000). Immunization trials are about to commence following the dramatic find- ings that transgenic mice that overproduce APP and deposit Aft can recover following immunization (Schenk et al., 2000). Specifically, when the mice were immunized at a young age, they developed little if any Aft depositions with advancing age. Moreover, the progression of both neuritic dystrophy and astrogliosis were significantly reduced in the treated animals, suggest- ing the immunization had benefits beyond simply reducing Aft deposition. When immunization was begun at later ages when the mice exhibit Aft de- position, further Aft deposition was blocked and somewhat reversed, as was the neuritic dystrophy and astrogliosis. In addition, remaining Aft deposits were often actively metabolized by microglia cells, questioning the premise that reduction of the activity of these cells by anti-inflammatory medica- tions would be of benefit in AD. These studies support the concept that the immune system may be harnessed into an appropriately targeted therapy for AD. If the trials of Aft immunization are effective in AD, it will provide compelling evidence for its causative role in AD. . understanding of the association between A /3 deposition (required for a diagnosis of AD) and the degenerative process. The link between soluble A /3 and brain atrophy needs to be clarified, and. into the pathogenesis of AD. To date, three genes have been iden- tified. These are APP mutations on chromosome 21, presenilin-1 (PS-1) mutations on chromosome 14, and presenilin-2 (PS-2) mutations. association with y-secretase, in the production of A /3 (Checler, 1999; Ray et al., 1999; Wolfe et al., 1999) and thus increase the pro- duction Afll-42/ 43. More than 50 mutations in PS-1 have been