Neuropathology of Alzheimer's Disease, as Seen in Fixed Tissues Charles Duyckaerts1,2, Dominique Langui1,2, Nadege Girardot1,2, Jean-Jacques Hauw1,3, Benoft Delatour4 Summary The progress of in vivo imaging will in the future determine if deductions made from fixed tissue concerning the pathology of Alzheimer's disease (AD) were cor rect. It will mainly enable us to apprehend directly the chronological sequence of the lesions and their duration. This paper reviews some of the data concerning the classical pathology of AD. Lesions may be categorized into A~ peptide deposition, tau accumulation and loss of neurons and of synapses. A~ peptide originates from the amyloid precursor protein, a transmembrane protein that is found in lipid rafts of the cellular membrane. Accumulation of flotillin-l, a raft marker, in AD indicates a disturbance of this membrane transport system in AD. Diffuse depo sits of A~ peptide within the cerebral cortex have little clinical consequence. Focal deposits are generally amyloid, i.e., Congo red positive and highly insoluble. They are usually associated with microglial activation and low-grade inflammation. A~ peptide is initially embedded in a cell membrane and partly hydrophobic. It may actually be linked to cholesterol in the extracellular milieu, as recently suggested. Tau accumulation takes place in the cell body, the dendrites and the axons of the neurons, forming, respectively, neurofibrillary tangles, neuropil threads and co ronae of senile plaques. Heiko and Eva Braak have shown that the progression of neurofibrillary pathology takes place in a stereotyped manner that appears to be correlated with the clinical symptoms. Contrary to the prediction of the amyloid cascade hypothesis, neurofibrillary pathology in the rhinal cortex and pyramidal sectors of the hippocampus most often precedes the first morphological evidence of A~ peptide accumulation. This observation suggests that tau and amyloid pa thologies are, at the start, independent processes that secondarily interact. Data from transgenic mice support this view: neurofibrillary pathology is not observed in APP transgenic mice, except if a human mutated tau transgene has also been incorporated. Progression of the lesions is different in the entorhinal-hippocam pal region and in the isocortex. In the former, tau pathology may be observed in the absence of A~ deposition, which appears to be a relatively late phenomenon. In the isocortex, by contrast, A~ peptide deposition is the first observable event. 1 Laboratoire de Neuropathologie Raymond Escourolle, Association Claude Bernard, CHU Pitie,-Salpetriere, Paris, France 2 Inserm U 106 3 Inserm U 360 4 Laboratoire NAMC CNRS UMR 8620, Orsay, France Hyman eta!. The Living Brain and Alzheimer's Disease ©Springer-Verlag Berlin Heidelberg 2 C. Duyckaerts et al. The sequence of the lesions in the isocortex appears to be the following: diffuse, focal, then amyloid deposits, neuropil threads, neuritic corona of the plaque and finally neurofibrillary tangles. Neuronal loss, in this sequence of events, occurs probably at a late stage, shortly preceding or occurring at the same time as tangle formation. Synaptic loss is a more complex process than previously thought: vesi cular markers of the synapse drop sharply in advanced cases, whereas membrane markers are relatively spared. The decrease in the number of synapses is associa ted with enlargement of the ones that survive. Up to now, observation of fixed brain tissue has dealt almost exclusively with proteins immobilized by formalin or other fixatives. Recent data show that mem brane lipids play an essential role in the pathogenesis of AD. One of the future challenges of neuropathology will be to visualize not only the proteins but also the membrane domains that bear them. Introduction In 1902 and 1903, Max Bielschowsky described a new silver method that allowed him to visualize the "neurofibrils" (Bielschowsky 1902, 1903). Just a few years later, Alzheimer (Alzheimer 1907) and his pupil Perusini (1910) found what they Fig.!. Neurofibrillary tangles; A a neurofibrillary tangle, as stained by a silver impregnation (Bodian stain). The nucleus of a normal neuron (small arrow) and its nucleolus (arrowhead) is seen close to a tangle-bearing neuron. The tangle is indicated by large arrows. Bar = 10 }lm; B a "ghost tangle," i.e., an extracellular neurofibrillary tangle (arrows). Ghost tangles are direct evidence of the death of tangle-bearing neurons. Neuropathology of Alzheimer's Disease, as Seen in Fxed Tissues 3 initially thought was the pathology of the neurofibrils, i.e., the "neurofibrillary tangle." We know today that tangles are not made of neurofilaments, the protein that made the neurofibrils, but of tau protein. Neurofibrillary Pathology Neurofibrillary tangles kill the neurons (Fig. 1). A direct proof of this killing is the so-called ghost tangle, i.e., a tangle that is found outside the cell body of the neuron. Neurofibrillary tangles are quite common in the brain of aged people, but they are often confined to the entorhinal cortex and adjoining hippocampus (Fig. 2). They may constitute the only lesion visible in an aged brain. In many cas es, according to the prevalence curves (Duyckaerts et al. 1997) drawn from Braak and Braak (1997) data concerning nearly 3,000 brains, the amyloid pathology fol lows tau accumulation in the neurofibrillary tangles. The mean delay between the two may be as long as 25 years (Duyckaerts et al. 1997). Fig. 2. Sensitivity of the entorhinal cortex to neurofibrillary pathology. A Macroscopic aspect of the entorhinal cortex in a normal case. The rhinal sulcus is shown by arrows. An asterisk indi cates the external surface of the entorhinal area. B Atrophy of the entorhinal cortex. An asterisk indicates the surface of the entorhinal area. C Numerous ghost tangles (arrows) are seen in layer II of a case with atrophy of the entorhinal cortex. Bodian silver stain. Bar = 10 }lm. 4 C. Duyckaerts et al. The Amyloid Pathology The term amyloid has been used too extensively: in the strict sense, it means a Congo red- positive deposit that appears to be made of fibrils around 10 nm in diameter at electron microscopy (Fig. 3B, C). The Ap peptide that precipitates in the cerebral parenchyma is initially not amyloid, and is shown only by immu nohistochemistry (Fig. 3A). The peptide is the product of the cleavage of a large transmembrane precursor protein called amyloid precursor protein (APP). The first cleavage takes place in the exodomain and is due to the p-secretase activ ity attributed to the enzyme called BACE. The second cleavage occurs within the membrane domain and is related to an enzyme activity called gamma-secretase, linked to a multi-protein complex including presenilin 1, nicastrin, APH-l and PEN-2 (Edbauer et al. 2003; Takasugi et al. 2003). Ap Peptide and APP The cell(s) in which the Ap peptide is produced, is (are) not fully identified. The neuron is, no doubt, involved, but the presence of Ap peptide within the neuron has been documented only in a few instances in man (Chui et al. 2001), particu- Fig. 3. Types of Ap peptide deposits. A Diffuse deposits. One deposit is indicated by arrows. Immunohistochemistry of AI3 peptide. Avidin-Biotine-Peroxidase method. Chromogen: dia minobenzidine. Bar = 20 j.lm. B A focal deposit (arrow) is surrounded by a clear halo and an outer rim, less densely stained. C Congo red staining: the focal deposit shown in B is intensely congophilic. Neuropathology of Alzheimer's Disease, as Seen in Fxed Tissues 5 larly in Down's syndrome (Gyure et al. 2001; Mori et al. 2002). Vesicles are also found in young transgenic APPxPS 1 mice, before or at the time of the first depos its of A~ peptide being seen (Wirths et al. 2001). These vesicles harbor markers of lysosomes and have their ultrastructural characteristics (Langui et aI., manuscript in preparation). The stages of A~ peptide production in the neuron are still in completely known. APP, a transmembrane protein, has to be synthesized in the rough endoplasmic reticulum, the cellular organelle that allows the making of hydrophobic proteins or hydrophobic protein domains. The topography of A~ production is still being discussed. It has been established that BACE acts before the gamma-secretase (Cupers et al. 2001; Maltese et al. 2001). It produces a frag ment of APP (C99 or CIOO) that still contains the transmembrane domain and may then be subjected to gamma-secretase activity. This last cleavage takes place after the endoplasmic reticulum (Iwata et al. 2001; Maltese et al. 2001). A~ peptide could then be liberated in the extracellular space by exocytosis or other mecha nisms (Lam et al. 2001). It should be stressed, however, that a third of the length of the A~ 42 peptide (namely its 14 last C terminal amino-acids) is probably still located in the membrane after the gamma-cleavage and is, anyway, hydrophobic. The presence of the pure peptide in the extracellular space is therefore improb able, and indeed a high content of cholesterol has been found in the core of the senile plaque (Mori et al. 2001; Girardot et al. 2003). The origin of the cholesterol in the plaque is still unknown. Apolipoprotein E, a transporter of cholesterol, is also present in the core of the senile plaque (Uchihara et al. 1995, 1996). APP and the Lipid Rafts APP is mobile in the cell membrane; it moves along the axon before being re distributed in the dendrites by transcytosis (Koo et al. 1990). The speed of its transport is not compatible with a passive diffusion. Cellular membrane is made of three types of lipids : glycerophospholipids, sphingolipids and cholesterol. At physiological temperature, the glycerophospholipids are above their melting point: they are "liquid,", actually in a "liquid crystalline" state. Sphingolipids, by contrast, are below their melting point in a "solid" state. Sphingolipids do not mix with glycerophospholipids. They form "microdomains" that are enriched in cholesterol. These "solid" micro domains float on the "liquid" glycerophospho lipids of the membrane as rafts on water (for review see Fantini et al. 2002). Lipid rafts are of low density and sediment in the upper part of saccharose gradient. They are insoluble in Triton X-IOO (Octyl phenol ethoxylate) at 4°C, a property that is used to isolate them. Five to 10% of neuronal APP are located at the cell membrane in lipid rafts (Bouillot et al. 1996). Recent data have suggested that the cytoplasmic domain of APP contains a kinesin receptor (Kamal et al. 2000, 2001). It links the raft to kinesin, which plays the role of a molecular engine mobile on the neurotubules. BACE (Riddell et al. 2001; Ehehalt et al. 2003) is also present in the raft fraction of the cell membrane. We have studied flotillin-1, a marker of rafts, in AD and found the accumulation of this marker in lysosomal vesicles of the neurons, a possible consequence of the disorganization of rafts (Girardot et al. 2003). 6 C. Duyckaerts et al. ~ t") CI.l o§ ~ I ~ ~ ~ ~ ~ fZl :G::Il = ...C.-.I.J. 0 e~ QCI J - .~" Q., =~ "~C ';..; = - .G.Il ...... ~ V) • ~ C=ol J:J < Fig. 4. Relationship between dementia and the density of Ab peptide deposits in Brodmann area 17. Intellectual status was evaluated in 26 cases (women aged over 75 years of age) by the test score of Blessed et al. (1968). Each point of the curve is a running mean offour successive cases (ranked in decreasing order of Blessed test score). The density of the Ab peptide deposits and of the Congo red-positive deposits, initially calculated as a proportion of volume, has been standardized [standardized value = (the original value -the minimal value over the whole co hort)/standard deviation over the whole cohort]. Standardization allows comparison of values, which may differ by several degrees of magnitude. All the Ab deposits, whatever their shapes, Neuropathology of Alzheimer's Disease, as Seen in Fxed Tissues 7 Diffuse and Amyloid Deposits The distribution of A~ peptide deposits in the isocortex is not as selective as the distribution of tau pathology. In most cases, A~ deposits are, indeed, found in all the isocortical areas that are examined, an observation that suggests that the de posits occur almost simultaneously in the whole cerebral cortex. The deposits may be only diffuse without neuritic components. In those cases, the cognitive deficit is absent or minimal (Delaere et al. 1990; Dickson et al.l991; Dickson 1997). In the study of a cohort of aged people, we used a cognitive index, the Blessed Test Score, as a time scale along which we plotted the density of the lesions (Metsaars et al. 2003). We showed that the presence of Congo red-positive plaques was detectable in cases that were only slightly more affected than those in which diffuse deposits of A~ peptide were observed (Fig. 4). We could also show that the presence of Congo red-positive (amyloid) material was tightly associated with the presence of macrophages (Arends et al. 2000), that in turn initiate the inflammatory response. A receptor complex is involved in the detection of fibrillar A~ by the microglial cell (Bamberger et al. 2003). The Neuritic Plaque Only at a more advanced stage is the central core of the amyloid plaque surround ed by a corona of degenerating neurites (Fig. 5); still later neurofibrillary tangles are detectable in the isocortex (Fig. 6). The neurites, which make up the corona of the senile plaque, could be axons or dendrites. Since they contain synaptic vesi cles (Terry et al. 1964), are immunostained mainly by antibodies directed against phosphorylated neurofilaments (present in the axons), and are negative with an tibodies against MAPs (present in dendrites; Schmidt et al. 1991), the conclusion has been drawn that they are mostly axonal terminals. However, the neurons from which these axons stemmed remained unclear, although indirect evidences point ed to the involvement of cortico-cortical connections (Duyckaerts et al. 1986). The Innervation of the Neuritic Plaque Transgenic mice gave us the opportunity to test this possibility in vivo. Bioti nylated dextran amine (BDA; Brandt and Apkarian 1992; Veenman et al. 1992; Reiner et al. 1993) was used as an anterograde tracer to visualize axonal tracts and normal or dystrophic terminal boutons in transgenic mice with numerous A~ peptide deposits. Ten aged (between 14 and 24 months) double transgenic mice, carrying the Swedish and London APP mutations and the PSI M146L mutation (collaboration with Aventis Pharma), were used. The tracer was injected either in .. were taken into account. Notice that the curve describing amyloid formation is close to the one describing Ab deposition. This observation suggests that amyloidogenesis occurs shortly after the accumulation of Ab peptide in the extracellular space. 8 C. Duyckaerts et al. - .- (IS o I: QeI QI "C C!) C .~ . (IS -f / ' u -- I: 0 N ... ~ = .C..J to ~ N ... ~ ~ "'CS -~ ~ ~ 0 =~ M Fig. 5. Relationship between the A~ peptide and the amyloid deposits and the corona of the senile plaque. Same methods as in Figure 4. Notice that the deposits surrounded by tau-positive processes occur at a later stage than the amyloid (Congo red-positive) deposits. Neuropathology of Alzheimer's Disease, as Seen in Fxed Tissues 9 V) 0 V) -en (1) aOIl 0 .... N --~ .I NV ) ) ·c .- ,{.I-) ~ 0 t+e:: en ~ 0{I ) (1) p. ;:::s ;(:1::s) -r0o'" ..e.-.n. r~/J ""G=)' Z 0. en CO N -Gt) ...u.--.. 00 . ""G=)' S "('10) ~ ""e=' Q) Q) Z ;> 0 + ..e.-.n. dbO "'0 en 0 Q) U "'" 0 0. 0 (I) OIl "'0 ~ 0 « U • Fig. 6. Relationship between neuritic plaques and neurofibrillary tangles. Same methods as in Figure 4. The neurofibrillary tangles occur shortly after neuritic plaques, but there are samples with neuritic plaques without tangles. 10 C. Duyckaerts et al. the thalamus or in cortical areas (cingulate cortex, hippocampus). A high density of labeled neurites surrounding amyloid plaques was observed close to the in jection site. These swollen, distended neurites were often observed in grape-like clusters around plaques. Not only short cortico-cortical axons (such as CAl-su bicular connections) but also long-distance fiber association pathways (such as cingulo-frontal fibers that course the whole caudo-rostral extent of the brain to reach their targets) showed marked morphological alterations in relation to amy loid deposits. In other animals, the tracer was iontophoresed in subcortical areas (septal and thalamic nuclei). In all brains examined, only a very small number of degenerating neurites was observed at the level of cortical axonal terminal fields. For example, despite very dense projections from the mediodorsal thalamic nu cleus to the plaque-enriched prefrontal pole, no clear evidence of pathological boutons was seen close to the amyloid deposits in this cortical area. We concluded that the senile plaques were mainly innervated by cortico-cortical connections (Delatour et al. 2003). Pathological Grades As there is a sequence in the order of involvement of the cortical areas, there is a sequence of lesions affecting a given isocortical area: diffuse deposits of Ap peptide precede amyloid formation, neuritic plaques and finally neurofibrillary tangles. This sequence can be used to grade the lesions within a given area (Met saars et al. 2003; Fig. 7). Synaptic and Neuronal Loss Synaptic loss has been considered the best correlate of dementia (Terry et al. 1991), although this has been questioned (Dickson et al. 1995). The vesicular markers of the synapse, such as synaptophysin, are much more severely affected than the membrane markers (Shimohama et al. 1997). Neuronal loss is often considered a major correlate of dementia. Its evaluation is difficult and subjected to numerous methodological problems. No neuronal loss could be observed when the total number of neurons was evaluated in the ce rebral cortex (Regeur et al. 1994). On the contrary, neuronal loss was found to be severe when the analysis was restricted to well-circumscribed areas, such as layer II of the entorhinal cortex or the cortex bordering the superior temporal sulcus (Gomez-Isla et al. 1996, 1997). We tried to evaluate the laminar topography of the neuronal loss by using a method that assesses the area around each neuron rather than the number of neurons per area. When the neighbors of a neuron are far apart, the free area around it is large and the cellular density is low (Duyckaerts et al. 1994,2000). The measure of the free space surrounding each neuron is a local measurement of the density. The "numerical density of one neuron" is the ratio lI(the area left free around that neuron). This area is the region of space that is closer to that neuron than to any other. It has the shape of a polygon, called a Vor onoi (or Dirichlet) polygon, the sides of which are located at mid-distance from