Beta/A4-Amyloid increases nerve growth factor production in rat primary hippocampal astrocyte cultures.

Nerve growth factor (NGF), a member of the neurotrophin family, is an essential mediator of neuronal activity and synaptic plasticity of basal forebrain cholinergic neurons (BFCN). In processes of chronic degeneration of BFCN like in Alzheimer's disease (AD), characterized among others by amyloid containing plaques, NGF has been shown to improve cognitive decline and rescue BFCN but also to reduce survival of hippocampal neurons via p75 neurotrophin receptor (p75). Little is known about the mechanisms of NGF regulation in glial cells under pathological conditions in AD. This study investigates the influence of amyloid administration on the NGF protein secretion in rat primary hippocampal astrocytes. Astrocytes were stimulated with "aged" beta/A4-Amyloid (1-40), and NGF was measured in different fractions, such as supernatant, vesicles, and cytosol fraction. Treatment with amyloid at a final concentration of 10 microM for 72 h led to increased NGF protein levels up to 30-fold increase compared to unstimulated controls. This observation may be an endogenous neuroprotective mechanism possibly contributing to a delay of amyloid-dependent loss of cholinergic neurons or contribute to accelerated neuronal death by activation of p75 within Alzheimer pathology.

Cholesterol, statins and tau.

Many of the known risk factors for Alzheimer's disease (AD) are associated with cholesterol metabolism. Interestingly, it seems as if higher doses of statins, i.e. inhibitors of the cholesterol biosynthesis by blocking formation of mevalonate, might lower the progression of AD. The mechanisms, however, by which statins or cholesterol levels exert their influence are unknown. A hereditary cholesterol-storage disorder, Niemann Pick C, shows Alzheimer-like tau-pathology in youth or adolescence but with no amyloid plaques. This gives rise to the possibility that disturbances in cholesterol metabolism induce changes in tau without interposition of Abeta-protein aggregates. Experimental data suggest that manipulation of cholesterol levels may lead to changes in tau phosphorylation. These changes vary depending on how cholesterol metabolism is manipulated. Effects seem to be either mild and transient, or drastic and related to neurodegeneration, or independent of the mevalonate pathway.

Pathological cholesterol metabolism fails to modify electrophysiological properties of afflicted neurones in Niemann-Pick disease type C.

Niemann-Pick disease type C (NPC) is a recessive inherited neurovisceral lipid storage disease characterized by progressive motor impairment and a loss of neurones including those integrated into the motor system. One of the key neuropathological findings is the intracellular accumulation of lysosomes enriched with free cholesterol. This accumulation is due to impaired transport proteins named NPC1 (approx. 95% of the cases) or NPC2 (approx. 5%) responsible for the transport of endocytosed cholesterol from lysomes to plasma membranes. The perturbed lipid-transport in NPC cells leads to an altered lipid composition of the plasma membrane. Available evidence suggests that the lipid matrix influences the electrophysical properties of ion channels in membranes. We therefore evaluated whether electrophysiological properties of NPC neurones differ from healthy neurones. Both, acute brain slices and primary neuronal cell cultures from wildtype and NPC mice, a well-established mouse model for the Niemann-Pick type C disease, were used for a comparison of electrophysiological properties like resting membrane potential, input resistance, action potential amplitudes and synaptic properties of the neurones. In addition we optically recorded the changes of intraneuronal calcium levels elicited by depolarization. Our results show that the characteristics of ion channels in NPC neurones do not differ significantly from wildtype neurones. We therefore conclude that gross alterations of the electrophysiological properties of neurones will probably not initiate or substantially contribute to the development of the motor impairment or other neurological signs of NPC.

Long-lasting transneuronal changes in rat dentate granule cell dendrites after entorhinal cortex lesion. A combined intracellular injection and electron microscopy study.

Following entorhinal cortex lesion, inhibitory hippocampal neurons show a persistent rarefication of those dendrites formally receiving entorhinal input. Physiological data indicate a long lasting disequilibrium of inhibition and excitation in the de-entorhinated hippocampus. We analyzed the intracellularly-stained dendritic tree of de-entorhinated excitatory rat granule cells. Granule cells of controls and animals surviving 2, 8, 60 and 270 days after unilateral entorhinal cortex lesion were impaled. Dendrites of control cells were of typical shape, traced to the hippocampal fissure and a complete dye filling of dendrites was ascertained by EM-analysis. Conversely, 60 and 270 days following lesioning, dendrites were only rarely seen to extend into the outer portions of the molecular layer and the dendritic architecture became significantly rarefied. Sixty days post-lesion, intracellularly filled dendrites extending to the middle molecular layer were surrounded by cell clusters resembling glia. Some of these contained the neuronally applied dye, suggesting a close association of the cytosolic compartments with the altered dendrites. These observed alterations exceed the process of sprouting and de novo synaptogenesis of remaining afference for long periods of time. The dendritic morphology of both inhibitory and excitatory neurons seems to require specific input from the entorhinal cortex. Moreover, sprouting of remaining afferents is apparently not sufficient to compensate for this loss of input.

Apolipoprotein E isoforms increase intracellular Ca2+ differentially through a omega-agatoxin IVa-sensitive Ca2+-channel.

Apolipoprotein E (apoE) is the major apolipoprotein in the brain and is known for its important role in plasticity and neurodegeneration. We show that apoE dose-dependently increases intracellular free Ca2+ in rat hippocampal astrocytes and neurons. This effect varies with isoforms in the order E4 > E3 > E2. It is insensitive to blockade of action potentials by tetrodotoxin or inhibition of binding of apoE by heparinase, by the LRP ligand lactoferrin and by low density lipoprotein. ApoE evoked Ca2+-increases are blocked in zero [Ca]o and by the Ca-channel antagonists nickel and omega-Agatoxin-IVa but not by nifedipine and omega-Conotoxin-GVIa, demonstrating an isoform-specific activation of P/Q type Ca2+-channels. This novel mechanism is discussed with respect to Alzheimer's disease, that is linked for most cases to the apoE epsilon-allelic variation (epsilon4 > epsilon3 > epsilon2).

Unbiased estimation of neuronal numbers in the human nucleus coeruleus during aging.

The total number of the neuromelanin-containing neurons of the nucleus coeruleus was determined by means of a newly developed unbiased stereological counting scheme and a low-cost apparative set-up. The individuals (n = 20, age from 49-98 years) included in this study were carefully examined for absence of neurological or psychiatric disorders. However, minor Alzheimer's disease-related neurofibrillary changes occurred in some of the individuals of higher age and these changes were staged. The mean number of neurons per side of the nucleus coeruleus was 15,731 +/- 3,408 SD with a range from 11,737 to 25,319. In three individuals, we compared the left and right nuclei and did not observe significant side differences between the numbers of neurons. There was no correlation between the age of the individuals and the cell number. Also, no correlation was detected between the cell number and the staged occurrence of minor neurofibrillary changes of the Alzheimer type. Due to the novel counting method, the determination of the total cell number took less than 2 h per case.

Spatial, temporal and numeric analysis of Alzheimer changes in the nucleus coeruleus.

The distribution of neurofibrillary tangles in the nucleus coeruleus was topographically and quantitatively analyzed. The topographical analysis showed statistically significant differences with regard to the distribution of neurofibrillary tangles in the dorsal-ventral and medial-lateral axes. More neurofibrillary tangles were found to be located in the dorsal and medial regions than in ventral and lateral areas. No significant difference in neurofibrillary tangle content was found between the rostral and the caudal areas of the nucleus coeruleus. Neurofibrillary tangle formation begins in the central parts of the nucleus coeruleus. The total number of neuromelanized neurons in the nucleus coeruleus was determined using a modern, unbiased sampling scheme and related to the cortical stage of Alzheimer's disease-related neurofibrillary changes present. A statistically significant reduction (50%) in nucleus coeruleus neurons was evident only in cases meeting the histopathological criteria for Alzheimer's disease. The extent of reduction in the total number of neurons in the nucleus coeruleus did not correlate with the number of neurofibrillary tangles observed. Our data suggest that despite the relatively early susceptibility of the nucleus coeruleus to neurofibrillary tangle formation, significant neuronal loss appears to occur much later, with an estimated average delay time of at least 25 years. Nonetheless, comparison of the topographical pattern of neurofibrillary tangle formation and cell loss indicates that neuronal loss is tangle-related.

Decrease in adenylate cyclase activity antecedes neurofibrillary tangle formation.

In previous studies, the formation of cAMP was seen to be significantly reduced in various brain regions in Alzheimer patients. Recently, a staging system was introduced in which Alzheimer-related. histopathological changes were classified according to the pattern of alterations rather than their density. In this study, we examined the degree of correlation between these stages and the postmortem activity of the CAMP-generating enzyme adenylate cyclase. Our findings suggest that forskolin-stimulated adenylate cyclase activity is significantly decreased before major neurofibrillary changes develop. Early impairment of postreceptor signalling could have potentially important consequences for drug treatment and/or for the development of neurofibrillary changes.

Calretinin immunoreactive structures in the human hippocampal formation.

The calcium-binding protein calretinin is present in an intrinsic GABAergic and an extrinsic non-GABAergic system in the rat and monkey hippocampal formation. Important species differences have been noted in hippocampal cell types immunostained for calretinin and the termination pattern of calretinin containing hypothalamic afferents in the hippocampus. In the present study, calretinin-containing neurons were visualized using immunocytochemistry in the human hippocampal formation of individuals which showed no significant neuropathological alterations. Calretinin-immunoreactivity was present exclusively in non-granule cells of the dentate gyrus and in non-pyramidal cells of Ammon's horn. Calretinin-positive neurons were found most frequently in the hilus of the fascia dentata and in strata radiatum and lacunosum-moleculare of CA1, whereas neurons in CA2 and CA3 were rarely immunostained. The majority of calretinin-immunoreactive neurons were small, bipolar or fusiform neurons. The dendritic trees of the calretinin-positive neurons were, for the most part, parallel to the dendrites of the principal cells. In the hilus, however, we observed cells with dendrites restricted to the hilar area. These dendrites were parallel to the granule cell layer. In the stratum lacunosum-moleculare, neurons with dendrites oriented parallel to the hippocampal fissure were frequently detected. In general, dendrites were smooth or sparsely spiny, displaying small conventional spines. The axons usually emerged from the proximal dendrite and could be followed over long distances. Axons were thin, had small varicosities and displayed only few collaterals which branched relatively far away from the cell body. Distinct bands of darkly stained calretinin-positive fibers occupied the innermost portion of the dentate molecular layer and the pyramidal cell layer of CA2. This distribution of calretinin-immunoreactive structures in the human hippocampus is similar to that observed in other primates but differs from that described in lower mammals, i.e., the rat. Our findings suggest that primates may share a common hippocampal calretinin-containing system, presumably both the intrinsic GABAergic and the extrinsic hypothalamic non-GABAergic components.

Glutamic-acid-decarboxylase-and parvalbumin-like-immunoreactive structures in the olfactory bulb of the human adult.

This study examines the distribution and morphological characteristics of glutamic-acid-decarboxylase-like (GAD)- and parvalbumin-like (PA)-immunoreactive structures in the olfactory bulb of the human adult. GAD-immunoreactive somata occurred in the glomerular layer, the external granule cell layer, the more superficial portion of the external plexiform layer, and the internal granule cell layer. The cells were small- to medium-sized. Demonstration of lipofuscin pigment revealed the presence of unpigmented as well as pigmented neurons, thus suggesting the existence of two subpopulations of GAD-positive neurons. GAD-immunoreactive puncta and/or fibers were mainly seen in the periglomerular region and the internal granule cell layer. All other layers of the bulb, as well as the intrabulbar portion of the anterior olfactory nucleus, displayed considerably less of these puncta and/or fibers. The olfactory nerve layer remained practically clear of immunoreactive material. PA-immunoreactive somata occurred in the glomerular layer and both the external and internal granule cell layer. Only a small number of immunoreactive nerve cells were encountered within the white matter or the olfactory tract. Most PA-positive neurons displayed characteristics of short axon cells whereas a few others resembled van Gehuchten cells. All of the PA-immunoreactive neurons were devoid of lipofuscin pigment. Immunoreactive puncta and fibers were present in all layers though predominating in the periglomerular region, the olfactory nerve layer, and the internal granule cell layer. The intrabulbar portions of the anterior olfactory nucleus did not show any immunoreactive structures.

Receptor-G-protein signalling in Alzheimer's disease.

Based on radioligand binding studies, it has long been assumed that the neurochemical pathology of Alzheimer's disease (AD) does not involve widespread changes in post-synaptic neurotransmitter function. However, more recent studies suggest that receptor function in AD may be compromised due to disrupted post-receptor signal transduction, in particular that mediated by the G-protein regulated phosphoinositide hydrolysis and adenylate cyclase (AC) pathways. The phosphoinositide hydrolysis pathway has been shown to be altered at a number of levels in AD post-mortem brains, including impaired agonist and G-protein regulation of phospholipase C, decreased protein kinase C (PKC) levels and activity, and a reduced number of receptor sites for the second messenger, Ins(1,4,5)P3. Of these, loss of Ins(1,4,5)P3 receptors and PKC in the entorhinal cortex and hippocampus correlates with AD-related neurofibrillary changes, as staged according to Braak's protocol. Disregulation of the phosphoinositide hydrolysis pathway may therefore have consequences for the progression of AD pathology. In contrast to the extensive pattern of disruption seen with the phosphoinositide hydrolysis pathway, changes to AC signalling in AD appear more circumscribed. Disruptions include a lesion at the level of Gs-protein stimulation of AC and, at least in the hippocampus, reduced enzyme activities in response to forskolin stimulation. Of these, the latter change has been shown to precede neurofibrillary changes. Apart from a loss of calcium/calmodulin sensitive AC isoforms, other components of this signalling pathway, including G-protein levels, Gi-protein mediated inhibition and protein kinase A levels and activity, remain relatively preserved in the disorder.

Apolipoprotein E and beta A4-amyloid: signals and effects.

In humans, the apolipoprotein E gene (APOE) is polymorphic with the alleles APOE epsilon 2, 3 and 4 coding for apolipoproteins (Apo) E2, 3 and 4. Apart from age, the APOE epsilon 4 allele represents the most important risk factor in sporadic Alzheimer's disease (AD). Compared to APOE epsilon 3 homozygotes, the histopathological onset of tau pathology is found 1-2 decades earlier but progresses with the same speed. ApoE dose-dependently and specifically increases free intraneuronal calcium levels in the order ApoE4 > ApoE3 > ApoE2. This effect is amplified in the presence of beta A4-peptide. The ApoE effects on calcium are not affected by the blockade of action potentials with tetrodotoxin, or by inhibition of common ApoE binding sites. The calcium channel involved has been identified as a P/Q-type-like channel. Brain tissue ApoE levels differ with respect to APOE alleles and Braak-stage for Alzheimer-histopathology. The production of ApoE in astrocytes is controlled by several receptor/effector systems such as adrenoceptors and cAMP. In the presence of beta A4-peptide fragments, astrocytes stop their synthesis of ApoE resulting in a massive reduction in the bioavailability of ApoE. In the periphery, ApoE directs cholesterol transport and thereby influences its cellular concentrations. In neurons, changes in the concentration of cholesterol influence the phosphorylation status of the microtubule-associated protein tau at sites known to be altered in AD.

Dysfunctional intracellular calcium homoeostasis: a central cause of neurodegeneration in Alzheimer's disease.

The clinical symptoms of all forms of Alzheimer's disease (AD) result from a slowly progressive neurodegeneration that is associated with the excessive deposition of beta-amyloid (A beta) in plaques and in the cerebrovasculature, and the formation of intraneuronal neurofibrillary tangles, which are composed primarily of abnormally hyperphosphorylated tau protein. The sequence of cellular events that cause this pathology and neurodegeneration is unknown. It is, however, most probably linked to neuronal signal transduction systems that become misregulated in the brains of certain individuals, causing excessive A beta to be formed and/or deposited, tau to become aggregated and hyperphosphorylated and neurons to degenerate. We hypothesize that a progressive alteration in the ability of neurons to regulate intracellular calcium, particularly at the level of the endoplasmic reticulum, is a crucial signal transduction event that is linked strongly to the initiation and development of AD pathology. In this chapter we will discuss the key findings that lend support to this hypothesis.

Gender and age modify the association between APOE and AD-related neuropathology.

OBJECTIVE: To assess the impact of apolipoprotein E (APOE) polymorphism on AD-related neurofibrillary tangle (NFT) formation and senile plaques (SP). METHODS: A sample of 729 routine autopsy brains (359 men, 370 women; age range, 60 to 99 years) was investigated. All brains were classified neuropathologically according to a procedure permitting differentiation of six NFT stages and three SP stages. APOE genotyping was performed on all cases. RESULTS: The epsilon4 allele of APOE was associated not only with SP (p < 0.0001) but also with NFT formation (p < 0.0001). The effect of the epsilon4 allele on NFT formation was noted at ages > or =80 years (p < 0.0001) but not between ages 60 and 79 years (p = 0.12). An association between the epsilon4 allele and SP for women was found at ages 60 to 79 years (p < 0.0001) but not at > or =80 years of age (p = 0.063). By comparison, men showed an association in both age categories (p = 0.001 and p = 0.001). CONCLUSION: The results confirm the association between the epsilon4 allele and both types of AD-related lesions and show that this association is differentially modified by age and gender.

Hippocampal apolipoprotein D level depends on Braak stage and APOE genotype.

Apolipoprotein (APO, gene; apo, protein) D, a member of the lipocalin family, has been implicated in several, pathological conditions but neither its physiologic function(s) nor ligand(s) has been clearly identified so far. Presuming a role in nerve de- and regeneration, several groups investigated apoD alterations in Alzheimer's disease (AD). Reported data, however, were not unanimous. We determined apoD protein levels in the hippocampus in a large, carefully matched autopsy case sample. ApoD levels were compared with the severity of neuropathological changes as determined by the Braak classification and with APOE genotype, a major risk factor for developing AD. ApoD was found to be related to the severity of AD-related neurofibrillary (NF) changes and not to old age alone. No correlation was found to amyloid deposits. Brain samples with widespread NF changes showed significantly higher apoD than cases with low Braak stages. This increase, however, was restricted to the APOE epsilon3/3 group, whereas the APOE epsilon4 group did not show significant variations in hippocampal apoD.

Genotype-related differences of hippocampal apolipoprotein E levels only in early stages of neuropathological changes in Alzheimer's disease.

Inheritance of the epsilon4 allele of apolipoprotein E (APOE, gene; apoE, protein) represents the most common genetic risk factor for developing Alzheimer's disease (AD), but the role of apoE in AD pathogenesis is yet to be clarified. A number of studies investigating apoE expression and protein levels in AD brain in correlation to its genetic polymorphism has yielded controversial results. We designed our approach based on neuropathological characteristics of AD to investigate apoE levels in relation to the APOE genotype and AD-related neurofibrillary changes, and amyloid deposits. We determined hippocampal apoE levels by reducing sodium dodecylsulfate-polyacrylamide gel electrophoresis and immunoblotting in 70 Braak-staged and APOE-genotyped autopsy brains. In our stage-, age- and gender-matched case sample, we found a significant increase of hippocampal apoE in the APOE epsilon3 homozygotes with beginning AD-related pathology (Braak stages I and II) compared with brain samples free of neurofibrillary changes and amyloid deposits. In the APOE epsilon4 allele carriers no such increase was found. In both genotype groups, severely affected brain samples with widespread neurofibrillary changes (Braak stages V and VI) and amyloid deposits (Braak stage C) showed low apoE levels comparable to those found in unaffected brain samples (Braak stage 0). Our data suggests that the isoform-specific impact of apoE on the development of AD may be of crucial importance only in the early stages of the disease. These stages are believed to represent phases of the disease in which the beginning neurodegeneration can be compensated by plastic reorganization.

Calbindin-D-28k-like immunoreactive structures in the olfactory bulb and anterior olfactory nucleus of the human adult: distribution and cell typology--partial complementarity with parvalbumin.

Calbindin-D-28k and parvalbumin are calcium-binding proteins. The laminar distribution and morphological features of calbindin-D-28k-like immunoreactive structures were studied in 60-microns-thick sections of the human olfactory bulb. Except for the olfactory nerve layer, immunoreactive neurons were present in all layers of the olfactory bulb. They reached highest densities in the external plexiform layer and internal granule cell layer. Considerable numbers of calbindin-like nerve cells were also found in the olfactory tract and in distal portions of the anterior olfactory nucleus. When comparing the distribution of calbindin-positive structures to that of parvalbumin-positive ones a partially complementary distribution pattern was found. Calbindin-like immunoreactive portions of the anterior olfactory nucleus and olfactory tract were mirrored by immunonegative areas in adjacent sections stained for parvalbumin. Using the combined pigment-Nissl procedure we observed the presence of lipofuscin deposits in nearly 80% of all the calbindin-immunoreactive neurons analysed. Moreover, analysis of their lipofuscin deposits rendered the further differentiation of morphologically similar neuronal subpopulations possible. In contrast, all parvalbumin-like immunoreactive neurons remained free of lipofuscin granules.

Close-meshed prevalence rates of different stages as a tool to uncover the rate of Alzheimer's disease-related neurofibrillary changes.

The speed of progression of Alzheimer's disease-related neurofibrillary changes is unknown. One reason for this is the impossibility to histopathologically follow-up one and the same individual over decades of their life. The present approach takes advantage of a recently introduced classification system which allows for a ranking of Alzheimer's disease-related neurofibrillary changes into six stages [Braak and Braak Acta Neuropath (1991) 82, 239-259] and analyses a staged sample of 887 brains obtained from routine autopsy. It sets out to interpret these cross-sectional data in dynamic longitudinal terms, in order to estimate the rate of passing through the various stages. The time needed to attain respective stages of pathology for 5% of a given cumulative sample is determined. The resulting fifth centiles are a measure of the average rate by which the disease-related changes progress assuming that the underlying stages represent a sequence of events and do not independently emerge. Advancing age and the prevalence of Alzheimer's disease-related changes of a given stage show a nonlinear positive correlation with only slight acceleration above the age of 65 years. Statistically, it takes at least 16 years from stage I to stage II, about 14 years pass by from stage II to III, 13 years from stage III to IV and five years from stage IV to V (= Alzheimer's disease) for 5% of a given cumulative sample. Thus, the deep roots of Alzheimer's disease-related neurofibrillary changes can be traced about 50 years back and may even extend into adolescence.(ABSTRACT TRUNCATED AT 250 WORDS)

The effects of beta/A4-amyloid and its fragments on calcium homeostasis, glial fibrillary acidic protein and S100beta staining, morphology and survival of cultured hippocampal astrocytes.

Aggregated beta/A4-amyloid is known to increase intraneuronal calcium by various mechanisms and to lead eventually to the death of the cultured neuron. This study deals with the role of beta/A4-amyloid and several of its fragments in calcium homeostasis, glial fibrillary acid protein and S100beta staining, morphology and survival of cultured rat hippocampal astrocytes as determined by Fura imaging, indirect immunofluorescence and life/death assays. In contrast to cultured neurons, none of the 12 different beta/A4 fragments tested caused an increase in intra-astrocytic free calcium. However, among the compounds evaluated, the fragments 10-20mer, 25-35mer and the full-length peptides (1-40, 1-42 and 1-43mer), at 5 and 10 microM, decreased free intra-astrocytic calcium statistically significantly after the cells had been incubated for 48 and 72 h. This occurred both for astrocytes treated with vehicle alone or the reversed sequence of the 1-40mer, i.e. the 40-1mer. However, survival was not altered under the conditions examined, even when there was a change in free intracellular calcium. Concomitant with the decrease in intracellular free calcium, the shape of the astrocytes became more spider-like, normally an indication of activated astrocytes, and markedly more intense anti-S100beta and anti-glial fibrillary acidic protein staining was seen. The functional relevance of altered calcium homeostasis for apolipoprotein E secretion, potentially relevant for neuronal plasticity in general and in Alzheimer's disease, is discussed.

Apolipoprotein E polymorphism influences not only cerebral senile plaque load but also Alzheimer-type neurofibrillary tangle formation.

Only recently, evidence was provided that apolipoprotein E allele epsilon 4 located on Chromosome 19 is associated with late onset (i.e. senile) sporadic Alzheimer's disease. Histologically, Alzheimer's disease is associated with intraneuronal neurofibrillary changes and extraneuronal A4/beta-amyloid deposition. We set out with a histological staging system which considers the gradual development of Alzheimer's disease-related histological changes over time and correlates highly with the cognitive decline ante mortem. Our analysis revealed that both the mean stage for A4/beta-amyloid deposits and the mean stage for neurofibrillary tangles get significantly shifted upwards in epsilon 4-carriers. This represents an earlier onset of the histopathological process of about one decade. The fact that both types of Alzheimer's disease-related changes correlate positively with the prevalence of the epsilon 4-allele suggests for a causal relationship between the apolipoprotein E polymorphism and the development of Alzheimer's disease.