ALZHEIMER'S DISEASE 177 as compared with subjects with higher education levels. The level of in- creased risk varies between studies (Katzman, 1993; The Canadian Study of Health and Aging Study Center, 1994; Stern et al., 1994; Letenneur et al., 1999; Hall et al., 2000) but is generally found to be between 1.5 and 2 times that of the higher-educated reference groups. Nevertheless, the finding of an association between education and AD is by no means universal because a number of studies have found no relationship (Beard et al., 1992; Cobb et al., 1995). In particular, no association between autopsy-confirmed AD and either education or occupation was found in a study of 115 patients with AD, although the authors suggest this may reflect different attitudes to consent to autopsy among groups with different education levels (Munoz et al., 2000). The mechanism that links low educational attainment and AD is unclear. Some authors suggested education provides increased functional capacity or "brain reserve," which requires the brain to undergo a greater period of de- generation before the critical threshold for dementia is reached. Conversely, low education may reflect other factors such as lower socioeconomic status, increased likelihood of exposure to adverse events, or childhood deprivation (Hall et al., 2000). This latter hypothesis, referred to as "brain battering," proposes that subjects with higher education have higher socioeconomic status and enjoy healthier lives with fewer coexisting brain diseases (Del Ser et al., 1999). This hypothesis is supported by the work of Del Ser and col- leagues (1999), who, in an autopsy study, found patients with low education had more cerebrovascular disease than those with a high level of education. Gaining a better understanding of this association between low education and AD is of great importance because education is a modifiable factor and, unlike increasing age or genotype, amenable to intervention and possible correction. A. BRAIN ATROPHY It is well established that the brains of older individuals are, on average, smaller than their younger counterparts (Dekaban, 1978). Although this may be interpreted as a loss of brain tissue with age, it may also represent cohort differences in body size as a result of improvements in nutrition and health standards (Miller and Corsellis, 1977). Cross-sectional in viva studies have demonstrated atrophy of all brain compartments (Murphy et al., 1996; Yue et aL, 1997), prefrontal grey matter (Raz et al., 1997), and hippocampus (Convit et aL, 1995). In several in vivo studies, age-associated atrophy was found to be greater in men than in women (Matsumae et al., 1996; Murphy et al., 1996; Yue et al., 1997; Coffey et al., 1998). However, postmortem analy- sis of normal subjects ages 46 to 92 years find no decrease in cortical volume, 178 JILLIAN J. KRIL AND GLENDA M. HALLIDAY but significant white matter atrophy (Double et al., 1996), suggesting that marked loss of cortical neurons is not a feature of normal aging. This is sup- ported by studies in older primates (Peters et al., 1996) and by longitudinal MRI studies (Mueller et al., 1998; Fox et al., 2000). Because brain atrophy oc- curs in a number of conditions other than neurodegenerative disease (Kril and Halliday, 1999) and factors such as hypertension, smoking, and high alcohol consumption contribute to atrophy (Akiyama et al., 1997), rigorous exclusion criteria are necessary in cross-sectional samples investigating true age-related changes. A loss of cerebral white matter may underlie the slow- ing of mental processing identified in many elderly subjects (Howieson et al., 1993; Ylikoski et al., 1993). Overall, the data are consistent with the clinical finding that, at least in a proportion of the elderly, there is no substantial deficit over time. In contrast to normal aging, cross-sectional studies show that there is marked cortical atrophy in AD (Fig. 2) and that the degree of atrophy cor- relates with the severity of dementia (Double et al., 1996; Mouton et al., 1998; Regeur, 2000). Atrophy in AD is most severe in the temporal lobe, particularly in the medial temporal lobe (Double et al., 1996; Convit et al., 1997; Detoledo-Morrell et al., 1997; Jack et al., 1998; Frisoni et al., 1999; Visser et al., 1999). More important, longitudinal analyses of brain volume confirmed that marked temporal lobe atrophy distinguishes AD (Fox et al., 1996, 2000; Smith and Jobst, 1996; Kaye et al., 1997; Yamada et al., 1998) from the relatively constant brain volumes during healthy aging (Shear et al., 1995; Mueller et al., 1998). The greatly accelerated atrophy of the temporal neocortex, not the hippocampus, in AD patients is associated with the symptomatic onset of dementia (Fox et al., 1996, 2000; Smith andJobst, 1996; Convit et al., 1997, 2000; Detoledo-Morrell et al., 1997; Kaye et al., 1997; Juottonen et al., 1998a, 1998b; Yamada et al., 1998), whereas atrophy of the hippocampus occurs 1 to 2 years before dementia onset (Fox et al., 1996; Convit et al., 1997). These data show that significant cortical atrophy occurs in AD and distinguishes it from normal aging. The degeneration begins in the hippocampus and spreads to involve first the temporal lobe and then other cortical association areas (Fig. 2). B. NEURONAL LOSS Controversy exists over whether neuronal loss is a normal consequence of aging or is only related to disease processes. Many earlier studies using measures of neuronal density found widespread degeneration in older sub- jects (Brody, 1955; Henderson et al., 1980; Anderson et al., 1983; Terry et al., 1987), although this finding was not universal (Haug and Eggers, 1991). ALZHEIMER'S DISEASE 179 Hippocampal atro )hy AD diagnosis ~ Hippocampal and temporal atrophy , Global atrophy l (end-stage) ~t i / FIG. 2. At autopsy, AD is characterized macroscopically by generalized atrophy of the cere- bral hemispheres (left panel), which results in widening of the sucli (upper), ventricular dilata- tion (V, lower), and atrophy of the hippocampal formation, causing dilatation of the temporal horn (TH, lower) of the lateral ventricles. Atrophy of the hippocampal formation can be de- tected in susceptible patients prior to the diagnosis of dementia (right upper). The atrophy progresses to involve the adjacent temporal lobe (right center) and then, uhimately, spreads to involve most regions of the brain (right lower). The introduction of unbiased quantitative techniques has revolutionized quantitative neuropathology; however, in many instances, there is still un- certainty as to whether neuronal loss with aging occurs. Pakkenberg and Gundersen (1997) found a 10% decline in total estimated neuron number between 20 and 90 years of age. This study was performed on samples from the entire neocortex, regardless of anatomical or functional location, but has yet to be confirmed by others. Interestingly, they also demonstrated a large (16%) difference in neuron number with gender, which is not as a result of differences in body height. Studies in which specific functional regions of the brain have been exam- ined using unbiased techniques have reported variable results with regard 180 JILLIAN J. KRIL AND GLENDA M. HALLIDAY to an age-associated loss of neurons. No loss of neurons was found in the su- perior temporal (Gomez-Isla et al., 1997) or entorhinal cortices (Gomez-Isla et al., 1996) of nondemented controls between the sixth and ninth decades, or from the locus coeruleus (Ohm et al., 1997). In the hippocampal for- mation, a loss of CA1 (West and Gundersen, 1990; Simic et al., 1997), CA4 (West et al., 1994), and subicular (West, 1993; West et al., 1994) neurons was reported. However, this is in contrast to the finding that the apparent reduc- tion in CA1 neuron number with age can be accounted for by differences in cerebrum volume between younger and older adults (Harding et al., 1998). This relationship between premorbid brain size and hippocampal neuron number highlights some of the difficulties with cross-sectional cohort stud- ies and suggests multiple factors need to be analyzed to determine potential cause and effect. The most consistent finding in AD is substantial neuronal loss from the entorhinal cortex and hippocampus (Fig. 3). This reflects the pattern of neurofibrillary pathology, which is a cardinal feature of AD and ap- pears to occur very early in the disease process (Braak and Braak, 1997). A 32% loss of neurons from the entorhinal cortex was found in AD patients Control AD CA1 Ch4 FIG. 3. Marked neuronal loss and NFT formation is seen in AD (right panels) compared with controls (left panels) in both the CA1 region of the hippocampus (upper panels) and cholinergic basal forebrain (Ch4, lower panels). Eventually, neuronal loss exceeds NFT forma- tion in the hippocampus but is equivalent in the basal forebrain. Nickel peroxidase with cresyl violet counterstain. ALZHEIMER'S DISEASE 181 with a CDR score of 0.5, whereas a 48% loss was found in all AD patients (Gomez-Isla et al., 1996). When specific laminae were examined, the loss was more dramatic with a 60% loss of layer II neurons in mild AD and a 90% loss in severe AD (CDR = 3; Gomez-Isla et al., 1996). Marked neuronal loss from the hippocampus has also been described. Simic and colleagues (1997) found a 23% loss of neurons from the dentate gyrus and subiculum, whereas West and colleagues (1994) found a 25% loss from the CA4, 47% from the subiculum, and 68% from the CA1. The dramatic loss of neurons from the CA1 and subiculum has been confirmed in other studies (Bobinski et al., 1995) and has been found to occur early in the disease process. Thus, the early atrophy noted clinically in medial temporal lobe structures (see above) is a result of marked neuronal loss in this region (Bobinski et al., 2000). Other consistently affected regions in AD are the cholinergic nucleus basalis (Vogels et al., 1990; Cullen et al., 1997; Fig. 3), the serotoninergic raphe nuclei (Aletrino et al., 1992; Halliday et al., 1992), and the noradren- ergic locus coeruleus (Busch et al., 1997). These subcortical nuclei innervate cortical pyramidal neurons, capillaries, and arterioles, and play an impor- tant role in cortical synaptic neurotransmission and the neurogenic control of blood flow through the capillary bed. The early loss of cortical choliner- gic transmission is believed to lead to hyperactivity of acetylcholinesterase and a loss of cholinergic neurogenic control, thus significantly contributing to the cognitive deterioration seen in AD (Bartus et al., 1982; Francis et al., 1999; Tong and Hamel, 1999). Hyperactivity of acetylcholinesterase under- lies the currently recommended treatments for AD, which use cholinesterase inhibitors such as tacrine, donepezil, or rivastigmine (Francis et al., 1999; Ladner and Lee, 1998). Despite mixed success with such treatments, there is a great deal of evidence supporting the cholinergic hypothesis of AD. Choline acetyltransferase levels were found to correlate with cognitive im- pairment in AD (Baskin et al., 1999), whereas degeneration in cholinergic basal forebrain neurons correlates with MMSE score (Iraizoz et al., 1999), cortical atrophy (Cullen et al., 1997), the stage of cortical pathology (Cullen and Halliday, 1998; Iraizoz et al., 1999; Beach et al., 2000), and the ear- liest depositions of A/~ (Beach et al., 2000). A/3 potently inhibits various cholinergic neurotransmitter functions (Auld et al., 1998) by killing corti- cally projecting cholinergic neurons (Harkany et al., 2000). Furthermore, cortical cholinergic denervation elicits vascular Aft deposition (Roher et al., 2000), suggesting a link between Aft deposition, small vessel disease, and cholinergic cell loss in AD. In addition, it has been shown that the action of tacrine is through improving cerebral blood flow rather than due its effects on neuronal cholinergic neurotransmission (Peruzzi et al., 2000). Although cortical atrophy is a consistent feature of AD (see above), whether this atrophy represents neuronal loss is not universally agreed upon. 182 JILLIAN j. KRIL AND GLENDA M. HALLIDAY Earlier studies of neuron density found a widespread and marked loss of neu- rons in AD (Colon, 1973; Shefer, 1973; Ball, 1977). However, using unbiased techniques, Reguer and colleagues (1994) found no overall loss of cortical neurons in AD. This study, which was conducted on entire lobes of the brain, generated much debate (see commentaries in Neurobiology of Aging (1994) 15(3) :353-380), the consensus of which was that regional and population differences do exist in AD and that they were masked by the quantitative technique used. Using unbiased techniques, total neuron number was found to decrease by 53% in the superior temporal gyrus (Gomez-Isla et al., 1997) and 30% in visual areas 17 and 18 (Leuba and Kraftsik, 1994). In addi- tion, a study described the loss of microcolumnar ensemble organization in AD (Buldyrev et al., 2000), although the relationship between neuronal patterning and cell loss remains to be determined. A considerable amount of research is still required to evaluate the specificity of the disease process for cortical regions and neuron type, and to correlate these findings with atrophy, clinical indices, and the temporal sequence of events. As long as research remains concentrated on individual brain regions affected by AD, the entire disease process will not be fully understood. C Aft DEPOSITION Aft is a hydrophobic peptide, 39-43 residues long, which tends to form insoluble aggregates. There has been considerable debate about the toxicity of this peptide, with its neurotoxic activity believed to depend on its abil- ity to form fibrils (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 2000; Gandy and Petanceska, 2000). The peptide is derived by the proteolytic processing of its high molec- ular weight precursor, the amyloid precursor protein (APP). APP is a trans- membrane protein with a small C-terminal cytoplasmic domain, one trans- membrane domain, and a large N-terminal extracellular domain (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 2000; Gandy and Petanceska, 2000). The Aft domain is partially embedded within the phospholipid bilayer. APP is cleaved via two proteolytic pathways, with only one pathway gener- ating Aft peptide (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 2000; Gandy and Petanceska, 2000). During transport to the cell surface, APP is cleaved at the membrane by an unknown protease called 0t-secretase into its soluble extracellular domain (sAPP) and a membrane-bound 10-kD C-terminal fragment. The membrane-bound fragment is further processed by, the as yet unidentified, y-secretase at the C-terminal end of the Aft domain into a small rapidly ALZHEIMER'S DISEASE 183 released peptide called p3. This pathway is the major processing pathway for APP and does not involve the production of Aft. p3 is found in abun- dance in the plaques associated with aging (Dickson, 1997). Uncleaved APP that is reinternalized is processed in the endosome/lysosome system by two hypothetical enzymes called/3- and y-secretases./3-secretase cleaves APP at the N-terminus of the A/3 domain, creating a 12-kD intermediate peptide, which recycles back to the cell surface, y-Secretase (s) cleave this intermedi- ate peptide at the C-terminal end of the A/3 domain, releasing A/3 into the extracellular space. y-Secretase cleavage occurs at one of two main sites producing mainly A/31-39/40 or sometimes A/31-42/43 (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 2000; Gandy and Petanceska, 2000). These peptides concentrate in the plaques found in AD (Iwatsubo et al., 1996; Dickson, 1997), although there is a general age-related increase in A/3 generation by neural cells (Turner et al., 1996), with the longer A/3 peptide being more amyloidogenic. The develop- ment of specific antisera for A/31-40 and A/31-42/43 has enabled the evo- lution and composition of plaques to be systematically studied (Iwatsubo et al., 1996; Dickson, 1997). The results suggest that A/31-42/43 initially forms the nucleus of a plaque, enabling the subsequent deposition of the more soluble Afll-40 and other protein fragments. This is consistent with the identification of mainly A/31-42/43 in plaque cores of both demented and nondemented individuals (Fukumoto et al., 1996). Evidence suggests that protofibrils of A/3 may also be toxic and that fibril formation is concen- tration dependent (Hartley et al., 1999), with A/3 peptides changing from soluble forms in control brain to insoluble forms in AD brain (Wang et al., 1999). Although we know a lot about the production of A/3, we know much less about its clearance from brain tissue. Evidence suggests that A/3 deposition is regulated by a specific protease that degrades extracellular A/3 (Iwata et al., 2000). Infusions of the protease inhibitor thiorphan into rat brain cause extracellular deposits of endogenous A/3 as diffuse plaques. The enzyme responsible for the clearance of Aft peptides is neutral endoprotease or neprilysin (enkephalinase; Iwata et al., 2000). Cross-sectional analysis of cases at different stages of AD suggests the A/3 plaque deposition occurs only early in AD with resorption surpassing deposition at end-stage disease (Thal et al., 1998). This suggests that A/3 clearance mechanisms are largely intact throughout the disease process and that the disease starts with early excessive A/3 production and deposition. Cross-sectional studies suggest the progressive deposition of A/3 in the brain and microvasculature appears to precede the onset of dementia by many years. Examination of a large unselected autopsy series shows a small 184 JILLIAN J. KRIL AND GLENDA M. HALLIDAY proportion of people in their 40s begin to deposit Aft plaques in the basal cortex (Braak and Braak, 1997). Few people at these ages have dementia, and the low frequency of pathology is believed to represent very early "pre- clinical" disease. By the age of 74, 50 % of the population will have Aft plaque deposits (Duyckaerts and Hauw, 1997), although few people will have overt dementia at this age (Jorm, 1990). At these and older ages, a subset of cog- nitively intact individuals have extensive neocortical Aft plaque deposition (Price and Morris, 1999), reinforcing the concept of "preclinical" disease. Furthermore, an accumulation of AD-type pathology was shown to nega- tively correlate with the change in MMSE score in nondemented subjects, indicating that burden of pathology does reflect functional performance (Morris et al., 1996; Green et al., 2000) and thus may represent "preclinical" AD. However, the concept that normal aging is synonymous with preclini- cal AD, which then proceeds to clinical AD, requires close scrutiny prior to being universally accepted. Several sets of data are difficult to reconcile with this model ofa contiuum between aging and AD. NFTs are present in all autopsy samples from people ages 91-95 years, whereas approximately 20% of these subjects are free from plaques (Braak and Braak, 1997). This suggests that Aft plaque accumulation may not be an inevitable component of aging. Alternatively, as discussed above, plaque-dominant AD has been proposed as a developmental stage of the disease only (Berg et al., 1998; Thal et al., 1998), with longitudinal data of cerebrospinal fluid showing changes in Aft levels are greatest within the first 2 years of diagnosis (Tapiola et al., 2000). Although much research has concentrated on determining the cellular biology of Aft production, there is only limited information on the relationship between Aft deposition and measures of degeneration. Large cross-sectional studies incorporating volumetric, neuronal, Aft deposition, and functional indices are necessary to determine the time sequence and relationship between these measures, particularly the role that Aft may play in the neurodegeneration of AD. D. NYI" FORMATION NFTs were first identified by Alzheimer in 1907. They consist of paired helical filaments of the microtubule-associated protein tau. In the normal brain, tau is bound to axonal microtubules where it stabilizes the micro- tubles, promotes their assembly, and allows fast axonal transport to occur (Goedert et al., 1991). In AD, tau becomes hyperphosphorylated and no longer binds to the microtubules, impairing their stability, and consequently impairing much of the normal function of the neuron. The hyperphosphor- ylated tau aggregates into paired helical filaments and ultimately NFTs. The ALZHEIMER'S DISEASE 185 gene for tau is on chromosome 17 and contains 15 exons. Mternative splic- ing of these leads to six isoforms of tau, ranging from 352 to 441 amino acids and with either three or four tandem repeats at the C-terminus end (Goedert et aL, 1991; Tolnay and Probst, 1999). In normal brain, three and four repeat tau is expressed in approximately equal amounts, and these same isoforms are present, in a hyperphosphorylated form, in AD (Tolnay and Probst, 1999). NFTs progressively accumulate in the cell body and processes of neu- rons until the cell dies (Bancher et al., 1989; Braak et al., 1994). The earli- est feature of NFT formation is the accumulation of hyperphosphorylated tau, which aggregates into insoluble granules (Bancher et al., 1989). This is called the "pretangle" stage and precedes the formation of the classi- cal fibrillar NFTs ("mature tangles"). Once the neuron dies, the largely insoluble NET remains in the neuropil as a "ghost" or "tombstone" tan- gle (Bondareff et al., 1994). The time taken for an NFT to form and ma- ture is unknown. Several estimates have been made based on extrapolation from relationships with disease duration. Bobinski and colleagues (1998) calculated it takes 3.4 years in the CA1 and 5.4 years in the subiculum for a mature NFT to become a ghost tangle. This, together with the find- ing of Morsch and colleagues (1999) that CA1 neurons with NFTs can survive for 15-25 years, suggests that NFTs are slow to develop and that the onset of pathology is many decades before the onset of clinical dis- ease. This hypothesis is supported by the findings that the calculated time taken to progress from NFT stage I to 1V is nearly 50 years (Ohm et al., 1995) and that lower scores on neuropsychological testing can be found as much as 10 years prior to onset of dementia (Elias et al., 2000; Small et al., 2OOO). NFTs and other abnormalities of tau are not unique to AD. Several other neurodegenerative diseases such as Down syndrome, progressive supranu- clear palsy, corticobasal degeneration, and parkinsonism-dementia com- plex of Guam and Pick disease also have tau-positive inclusions (Tolnay and Probst, 1999). This has led to the collective name of tauopathies, and much effort has been expended to understand the commonality of these disorders. To date, a number of differences were found in the cellular pop- ulations affected and the tau isoforms expressed (Brion, 1998). However, similarities in types of tau deposited and clinical expression of the diseases were also described. Cross-sectional studies suggest that progressive NFT formation in the brain precedes the onset of dementia by many years. Examination of a large unselected autopsy series shows that a small proportion of people in their 20s begin to form NFT in the entorhinal cortex (Braak and Braak, 1997). Few people at these ages have dementia and the low frequency of pathology 186 JILLIAN J. KRIL AND GLENDA M. HALLIDAY is not believed to affect cognitive function. By the age of 47, 50% of the population will have NFTs (Duyckaerts and Hauw, 1997), although few peo- ple will have overt dementia at this age (Jorm, 1990). As mentioned above, all subjects ages 91-95 years have NFT formation (Braak and Braak, 1997), and although dementia is more prevalent at these ages, it is not inevitable (Jorm, 1990). By 86 years of age, 50% of the population have sufficient accu- mulation of NFTs to suspect a pathological diagnosis of AD, particularly in the presence of Aft plaques (Duyckaerts and Hauw, 1997). At the age of 86 and older, approximately 20% of people meet NFT criteria for AD (Braak and Braak, 1997). This is consistent with the prevalence of clinical AD at these ages (Jorm, 1990). In contrast to the Aft deposits, NFTs accumulate in regions of neuron loss (Braak and Braak, 1997; Cullen and Halliday, 1998; Duyckaerts et al., 1998; Iraizoz et al., 1999), and their accumulation correlates with measures of functional decline (McKee et al., 1991; Arriagada et al., 1992; Bancher et al., 1993; Grober et al., 1999) and the degree of hippocampal atrophy (Bobinski et al., 1995; Nagy et al., 1996, 1999; Smith andJobst, 1996). However, as de- scribed above, NFTs take many years to evolve and, therefore, the temporal relationship between the formation of NFTs and the rapid neuronal loss and brain atrophy in AD is difficult to reconcile. In addition, as dementia is present only when NFTs occur in the neocortex and the extent of neo- cortical neuron loss is unclear in AD (see above), the association between this cortical degeneration and NFT and Aft deposition needs to be further examined. E. MECHANISMS OF DEGENERATION Studying the mechanism(s) of neuronal death in AD is difficult because of the extended interval between the onset of symptoms and associated cell death, and investigation at autopsy. NFT formation is considered to be the major cause of neuron death in AD (Fig. 4), and cells dying as a result of NFT formation can be identified by the presence of ghost NFTs. However, reports show NFTs are not responsible for all the neuron loss seen in AD. Studies on the temporal (Gomez-Isla et al., 1997) and occipital (Leuba and Kraftsik, 1994) cortices, and hippocampus (Kril et al., 2000) have shown that neuronal loss exceeds the degree of NFT formation. This is in contrast to studies of the cholinergic basal forebrain in AD (Cullen and Halliday, 1998) and the parkinsonism-dementia complex of Guam (Schwab et al., 1998, 1999), where NFT formation does account for all the neuron loss. In the CA1 region of the hippocampus, NFTs were found to account for less than 20% of the neuron loss (Kril et al., 2000) suggesting that another . A/3 1-3 9/40 or sometimes A/3 1-4 2/ 43 (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 20 00; Gandy and Petanceska, 20 00). These peptides concentrate. and a large N-terminal extracellular domain (Haas, 1996; Neve and Robakis, 1998; Storey and Cappai, 1999; Wilson et al., 1999; Coughlan and Breen, 20 00; Gandy and Petanceska, 20 00). The Aft. enzymes called/ 3- and y-secretases./3-secretase cleaves APP at the N-terminus of the A/3 domain, creating a 1 2- kD intermediate peptide, which recycles back to the cell surface, y-Secretase (s)