Amyloid precursor protein is a ubiquitously expressed single-transmembrane protein with a 590–680 amino acids (aa.) long extracellular amino terminal domain and a short cytoplasmic tail, which contains intracellular trafficking signals (Fig. 1) (6). APP transcripts undergo alternative splicing to yield eight possible isoforms of which APP-695, 751, and 770 isoforms predominate in the brain, with APP-695 isoform being mainly produced by neurons (7). Thus, most of the transgenic mouse and cellular models are based on APP-695 isoform, while some of the constructs also utilize APP-770 isoform.

While APP is normally processed by α- and γ-secretases to release a secreted APPα fragment (APPsα), a 3 kD p3 peptide and a 7 kD APP intracellular domain (AICD), the alternative processing of APP by β- and γ-secretases results in full-length Aβ, an N-terminal soluble APPsβ fragment, and the C-terminal AICD tail (8). The α- and β-secretases have recently been identified as members of the ADAM and BACE family of enzymes, ­respectively, and γ-site processing was found to be dependent on a pentameric complex in which presenilin 1 and 2 (PS) act as aspartyl transmembrane proteases (9,10). It is important to note that while these pathways were called amyloidogenic and nonamyloidogenic pathways, recent studies suggest that p3 or other N-terminally cleaved Aβ fragments are also deposited as amyloid (11) (see later). APP is processed in both the constitutive and alternative secretory pathways in the trans-Golgi network (6) and via the early endosomal-lysosomal-exosomal pathway (6,12). In neurons, APP and its catabolites are also trafficked along the microtubules within the axons via a kinesin-mediated mechanism and APP is observed to be present in synapses (13). This fits well with the arguable physiological function of APP in neurite growth and memory (14). In support of this proposed function, APP deficient mice show a moderate atrophy of corpus callosum together with astrogliosis but without any increase in morbidity or mortality (15). The anterograde transport of APP also might explain why some of the Aβ deposits in the dentate gyrus appear to originate from nerve terminals whose axons traverse the perforant pathway, and why lesions of perforant pathway in animal models of AD result in a reduction in hippocampal amyloid burden (16,17).

Aβ has two major C-terminal isoforms, Aβ40 and Aβ42. Aβ40 comprises 90–95% of secreted Aβ and is the predominant species found in the cerebrospinal fluid while less than 10% of the secreted Aβ is Aβ42 (18). The secreted fraction of Aβ presents as monomers/dimers and soluble Aβ oligomers, which are passively redistributed in brain parenchyma and cleared (19). Several Aβ clearance mechanisms have been described. First, Aβ is taken up and degraded by microglial and astroglial cells (20). In addition, Aβ is also degraded by proteases including neprilysin (NEP), insulin-degrading enzyme (IDE), endothelin-converting enzymes-1 and -2 (ECE), matrix metalloproteinase-2 and -9 (MMPs), and tissue plasminogen activator (tPA) (21). Of these, NEP, IDE, and ECE1 are active intracellularly while IDE, tPA, and MMPs act at the cell surface or are secreted and activated in the interstitial fluid (ISF) (21). Knockout models for many of these proteases show a significant elevation of murine brain Aβ, while their overexpression counterparts show a reduction in Aβ levels and retardation of plaque formation (22). Human studies also show that a local reduction in expression of some of these proteases correlates with a high plaque burden (23) (Fig. 3, upper panel). Lastly, Aβ is also drained out from the brain parenchyma by two mechanisms involving blood vessels. One of these mechanisms is a direct Aβ transport across the blood-brain barrier (BBB) mediated by LDL receptor-related proteins-1 and -2 (LRP-1, LRP-2), and regulated by α2-macroglobulin, apolipoprotein E (ApoE), and apolipoprotein J (ApoJ), respectively (24) (see later). This direct vascular clearance mechanism is highly efficient in clearing not only physiological Aβ but also pathological amounts of Aβ in transgenic mouse models (25). It is postulated that if the influx of circulating Aβ would be completely stopped, LRP-mediated transport could remove the entire Aβ pool from brain ISF within less than 1 min under physiological conditions, or in approximately 40 min at pathological levels (26). A reduction of LRP-1 expression has indeed been observed in AD patients as well as in mouse models (27). The second vessel-related Aβ clearance route utilizes the periarterial ISF drainage pathway where Aβ in ISF is collected around periarterial space (also known as the Virchow-Robbin space) that reaches the cerebrospinal fluid (CSF) in the subarachnoid space. From here, along the perivascular spaces of the circle of Willis, the fluid reaches the olfactory bulbs, then passes with perforant lymphatic channels through the cribriform plate to nasal submucosa and finally to the cervical lymph nodes (28). Tracer experiments in animal brain demonstrate that this is equivalent to lymphatics of brain (29). The ISF drainage pathway constitutes a major route of Aβ elimination in situations when BBB becomes less efficient (as in aging) or is overwhelmed (as in mouse AD models) (24,30). All mechanisms taken together, Aβ clearance from human brain normally exceeds its production with Aβ fractional production and clearance rates measured as 7.6% and 8.3% per hour, respectively (31).

Both soluble and fibrillar aggregates of Aβ are shown to be toxic and mouse models have been very useful in confirming the toxic potential of different Aβ aggregation states, either by ­showing inhibition of hippocampal long-term potentiation (LTP) or by demonstrating memory impairment after injecting ­aggregated Aβ species in rodent brain. For instance, together with human and in vitro studies, mouse models have been instrumental in showing that fibrillar and/or nonfibrillar forms of Aβ deposited in vessel walls are associated with degeneration of endothelial cells, smooth muscle cells (SMCs), and brain pericytes (32–34). Similarly, reduction in synaptophysin levels and disruption of neuronal calcium homeostasis has been shown in mouse models, suggesting a considerable role of soluble Aβ aggregates (35–37). Increasing data on varied models also indicate that Aβ can in turn activate microglia and thus indirectly damage neurons, contributing to the neurodegenerative process and cognitive loss (38,39). It is maybe for this reason that the “neuritic plaques” are more active as these are shown to be in histological proximity with inflammatory markers (40,41). Some of these issues will be discussed in relevant sections of this chapter.

AD also shows tau proteinopathy with aggregated forms of hyperphosphorylated tau deposited in neurons as neurofibrillary tangles (NFTs) and neuropil threads. Hyperphosphorylated tau is also a major constituent of Pick bodies in FTD. Tau is a highly conserved protein that belongs to the family of microtubule associated proteins (MAPs) and is essential for the assembly of tubulin into microtubules (42). In human CNS, tau is abundantly expressed in the axons of mature and growing neurons; however, low levels of tau are also present in oligodendrocytes and astrocytes (43,44). The human tau gene (MAPT) is located on chromosome 17q21 and encodes 15 exons of which three exons (4A, 6, and 8) are never present in any mRNA expressed in human brain (reviewed in (45)). Tau mRNA is spliced in to six possible isoforms with lengths between 352 and 441 aa (46). These isoforms occur by combining two possible inserts (0N, 1N, 2N; encoded by exons 2 and 3) in the N-terminal domain, and 3 or 4 possible repeats in the C-terminal tail (3R or 4R, depending on the alternative splicing of exon 10) (45). Also, the six tau isoforms are not equally expressed in neurons, for example tau mRNA containing exon 10 is not found in granular cells of the dentate gyrus. The four repeat domains represents the microtubule-interacting domain, while the C-terminal acidic region projects between the microtubules, acting as a linker with the other components of the cytoskeleton and as a spacer for the other surrounding microtubules. In fact, tau proteins may ­interconnect microtubules with other cytoskeletal components such as neurofilaments and actin filaments. In mice lacking the tau gene, an increase in the microtubule associated protein 1A is observed, and this may compensate for the loss of tau functions (47). However, in these mice, axonal diameter in some neurons is particularly affected.

Phosphorylation is by far the most common posttranslational modification of tau. At least 25 different tau phosphorylation sites have been identified. Tau phosphorylation might have a physiological role, as during development fetal tau is more heavily phosphorylated than adult tau (48). Tau hyperphosphorylation and aggregation is also a major molecular change leading to accumulation of insoluble intracellular paired helical or straight filaments (PHF/SF) in a number of tau proteopathies such as AD, Pick’s disease (PiD), progressive supranuclear palsy (PSP), and corticobasal degeneration (CBD) (49). In the following sections, we will discuss how Aβ and tau interact in relevant mouse models.

Characterization of vascular amyloid in AD and Down’s syndrome patients led to the identification of the Aβ peptide and APP gene (52), which was followed by the identification of Aβ in parenchymal plaques (53). The first pathogenic mutation (E693Q) was also identified in a form of vascular amyloidosis called hereditary cerebral hemorrhages with amyloidosis – Dutch-type (HCHWA-D) (54,55). HCHWA-D patients are characterized by recurrent hemorrhages and extensive Aβ deposition in cerebral blood vessel walls, in the absence of senile plaques and tau-related pathology (56–58). Besides the occurrence of large lobar intracerebral hemorrhage, HCHWA-D patients may show cognitive deterioration, frequently associated with white matter abnormalities on MRI and small ischemic and hemorrhagic infarctions on pathological examination (59). At the level of APP processing, this mutation drastically increases the Aβ40/Aβ42 ratio, with decreased Aβ42 levels, both in vivo and in vitro (60,61). Moreover, the mutated Aβ Dutch peptides show an enhanced fibrillogenesis and toxicity towards vascular cells in various in vitro experimental setups, explaining to some extent its affinity to blood vessels (62,63).

The Flemish (A692G) APP mutation, identified subsequently, leads to both cerebral hemorrhages and AD (64). Pathologically, Flemish APP mutation carriers show senile plaques with the ­largest central dense-cores observed in AD and a severe degree of cerebral amyloid angiopathy (CAA) (41,65,66). Also, in contrast to HCHWA-D and consistent with a clinical diagnosis of AD, Flemish APP patients have considerable neurofibrillary pathology (41,65). Besides Dutch and Flemish APP mutations, other mutations that have been identified near the α-secretase processing site are Arctic (E693G), Italian (E693K), and Iowa (D694N) mutations (Fig. 1). These APP mutations lead to very similar ­phenotypes, dominated by amyloid angiopathy, diffuse amyloid deposits, and occasionally ischemic infarctions (67–69). This is explained by the fact that these α-secretase site APP mutations change the primary Aβ sequence and increase its vascular affinity and/or fibrillogenic potential including formation of Aβ oligomers.

The α-secretase site APP pathology contrasts sharply with pathology caused by mutations identified near the β- or γ-secretase sites. The only APP mutation near the β-secretase site is the double Swedish (K670N/M671L) mutation (70) that promotes β-secretase activity and increases the total Aβ production without altering the Aβ40/Aβ42 ratio (71). This is consistent with the pathological data in APP Swedish carriers, which deposit more Aβ40 compared to carriers of APP mutations near the γ-secretase site (72). The first γ-secretase site APP mutation identified was the London (V717I) APP mutation that results in an increased Aβ42/Aβ40 ratio, without altering total secreted Aβ (73,74). The enhanced secretion of more aggregatable Aβ42 directs the pathology towards a diffuse amyloid distribution predominantly constituted of Aβ42, and exceeding the burden observed in sporadic AD patients (72). Many γ-secretase site APP mutations have been identified and while almost all of these mutations decrease the total Aβ production and increase Aβ42/Aβ40 ratios (75), the pathology caused by Austrian (T714I) APP mutation is the most remarkable one. This mutation leads to a very aggressive form of AD with a very young age of onset and a diffuse, nonneuritic amyloid pathology (76). In vitro cell culture experiments in neuronal and nonneuronal cells showed that APP Austrian mutation led to one of the most drastic increases in the Aβ42/Aβ40 ratio in a panel of γ-secretase site mutations studied (76,77). Lastly, several independent duplications of the APP locus have been recently identified in French and Dutch families with autosomal dominant AD and/or lobar cerebral hemorrhage with prominent CAA (78,79). Similarly, Down’s syndrome patients, also carrying three copies of APP, develop CAA as young as 30 years and the severity of CAA and vessel-related risks increase with age (80). APP gene dosage should be considered as a situation similar to APP Swedish mutation because it also increases total Aβ without changing the Aβ42/Aβ40 ratio. These data underpin the ­growing consensus that increase in absolute levels of Aβ40 leads to a predominant vascular amyloidosis (81).

FAD mutations in presenilin 1 and 2 are more common than APP mutations (4,5,50). Structurally, PS are multiple transmembrane domain proteins localized in the endoplasmic reticulum, Golgi apparatus, and lysosomes as a part of the γ-secretase complex (6,82). The functions of PS are not entirely clear; however, PS1 knock-out mice die ante partum or immediately after birth and show lissencephaly, severe altered somitogenesis, skeletal, and vascular disturbances resembling a Notch phenotype (83,84). PS mutation carriers present with an earlier age of onset than APP mutation carriers, and with abundant diffuse-like amyloid pathology, perhaps because these mutations increase the production of the more insoluble and readily depositable Aβ42 isoform (75). Interestingly, most of the PS mutations tested in vitro not only increase Aβ42, but also lower Aβ40 production drastically with an overall reduction of total Aβ production (75,85). This might have important implications for amyloidogenic mouse models based on PS mutations (see further). Also, interestingly, select AD-linked PS1 mutations and the Volga-German PS2 N141I mutation associate with a marked CAA (86,87). While the precise cause is again unknown, it is plausible that PS mutations associated with predominant CAA favor the production of more soluble Aβ40 that is drained and deposited along the perivascular space in association with Aβ42 nidi (81).

As previously mentioned, the deposition of hyperphosphorylated tau as insoluble filaments in the brain is a pathological hallmark of several neurodegenerative disorders, defined as tauopathies. An autosomal-dominant inherited form of frontotemporal dementia with parkinsonism (FTDP) was initially linked to chromosome 17q21-22 in 1994 (88). In the following years, 13 additional families with FTD and parkinsonism with linkage to 17q21-22 were identified (45). These patients presented with disinhibition, loss of initiative, obsessive-compulsive behavior, and/or psychosis, followed by cognitive decline. In most patients, extrapyramidal symptoms occurred only late in the clinical course, but considerable heterogeneity is observed both between mutations and within families with the same mutation (89). In 1997 at a consensus conference, the term of frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17) was introduced to describe these patients (89) and in the following year mutations in MAPT were reported in the majority of these families (90,91). At present, 39 mutations in MAPT have been identified in 115 families, including missense mutations, silent mutations, in-frame single codon deletions, and intronic mutations (45). Most FTDP-17 MAPT mutations result in increased 4R/3R tau ratios, and modify tau splicing pattern or cause single aa changes, which probably result in impaired microtubule binding capacity or increased aggregability (45). On gross neuropathology, FTDP-17 patients present with a severe atrophy of the frontal and temporal lobes (or blade-edge atrophy described in classical Pick’s disease), although the superior part of the precentral gyrus is commonly spared. On microscopy, astrocytic gliosis and a massive neuronal loss are found in the superficial cortical layers and sometimes also in basal ganglia (92). Filamentous tau intracytoplasmic inclusions in neuronal and glial cells are the characteristic finding. Extensive mutation analysis revealed many other MAPT mutations in several additional FTDP-17 families, as well as in unrelated FTD, CBD, or PSP patients (for a complete update, visit www.molgen.ua.ac.be/FTDMutations).

In brain parenchyma, Aβ exist as monomers, dimers, low molecular weight oligomers, mycelia, protofibrils, and finally, mature densely compact fibrils (Fig. 1). While most of the Aβ assembly is extracellular, recent data also suggest that Aβ dimers could be constitutively secreted from neurons (93). Except the insoluble mature fibrils, all other precursor forms can be regarded as the soluble Aβ fraction, and most of these states including dimers but excluding monomers are toxic (94). While the soluble monomers and early aggregation states do not cross the threshold for immunohistochemical detection, densely packed aggregates are also less visible by immunohistochemistry due to epitope masking (see further). Thus, immunohistochemistry is most robust in identifying higher-order Aβ aggregation states including loosely arranged fibrils. Densely packed fibrils also acquire a β-pleated sheet conformation (95,96) and are therefore readily identified by silver impregnation techniques and Congo red or thioflavin (Th)-S or -T (95,96). Based on these different staining protocols, various types of extracellular and intracellular amyloid can be identified in both humans and mouse AD models.

Diffuse deposits are the first extracellularly visible Aβ deposits observed in AD and Down’s syndrome patients (11,80,97), and are also termed preamyloid deposits or preplaques (98). In AD brains, diffuse plaques are mainly present in the entorhinal cortex, allocortex (hippocampus and olfactory cortex), followed by the superficial layers of the neocortices, especially association neocortices (99). Occasionally, a cloud of diffuse plaques are observed under the cortical pia mater or the subependymal periventricular zones (99). These plaques also involve neostriatum and the molecular layer of cerebellar cortex (100). Various morphological forms of diffuse plaques have been described. However, in most frequent forms they appear as clusters of dots and rods with irregular borders merging diffusely with the parenchyma, and ranging in size from a few microns to many hundreds of microns (Fig. 2). Ultrastructural studies of such deposits have revealed scattered bundles of amyloid fibrils as well as amorphous, nonfibrillar material (101). Diffuse plaques frequently surround neuronal cell bodies and neurites without implicating vascular elements (101), suggesting that such deposits might be neuronally derived. Although it was shown that the progression to the end-stage of AD is associated with a proportional increase in fibrillar plaque isoforms (102), recent studies suggest independent evolution paths for fibrillar-dense and diffuse plaques (103). Biochemically, diffuse plaques seem to be composed of full-length and N-truncated Aβ42 (100,104). Moreover, diffuse plaques neither seem to associate with any cytoskeletal neuritic alterations nor display any significant astrogliosis or microglial activation (100,105,106). Only exceptionally, in end-stage diseases or in very aggressive familial forms of AD, such as Austrian T714I pathology, diffuse plaques are observed to be neuritic and associated with abundant glial and inflammatory pathology (76,100,107).

AD mouse models based on APP and PS mutations also develop diffuse plaques that resemble human diffuse plaques biochemically and morphologically, in that they are also ThS-negative and predominantly composed of Aβ42. In mice, diffuse deposits can exist as small patches of amyloid, but given the supraphysiological levels of Aβ production, the number of plaques increases rather rapidly to occupy large areas of brain parenchyma. All brain regions can be affected, although the first region to be involved largely depends on the specific promoter utilized. Certain brain regions such as cerebellum and basal ganglia seem to have a paucity of diffuse deposits (further discussed under section Mouse Models).

Dense-cored plaques were first observed in 1892 and named “senile plaques” in 1911 by Simchowicz (108). These plaques consist of a dense core of amyloid fibrils arranged radially, surrounded by an immediate clear halo and an external belt of amyloid. Sometimes, long arms of amyloid connect the surrounding corona with the inner core of the plaque. The central core regions are one of the most compact forms of amyloid observed in human brain, and are readily stained with Congo-red or ThS, appearing round, star, or cross-shaped (“Malthese cross”). With Aβ immunohistochemistry, the dense-core regions stain faintly perhaps due to the masking of Aβ epitopes. This type of dense aggregated cored plaques is the most characteristic finding in the Flemish APP pathology, with the “largest” dense-cored plaques encountered in AD (41). Dense-core plaques are also called neuritic plaques as they are frequently surrounded by dystrophic neurites with or without NFT pathology or neuropil threads and also seem to incite both astrocytosis and microgliosis (99). Occasionally, dense-core plaques seem to lack a coronal plaques region as well as dystrophic neurites and these compact cores are sometimes referred to as “burnt-out” or “end-stage” plaques with the belief that the surrounding neuronal elements are destroyed in these plaques. Biochemically, while the corona of the plaques has predominantly, if not solely, Aβ42, the central dense-core regions are rich in Aβ40. Thus, Aβ42 is present in both the core and the corona of senile plaques although in different proportions, but it seems that increased Aβ42 is a common feature for both normal aging and senile AD cored-type of deposits (109). It is possible that with increasing compactness and aggregation state of the plaque cores, an increasing ratio of Aβ40/Aβ42 occurs for the deposited amyloid.

The highest number of dense-cored plaques is found in the superior hippocampal regions (CA1–CA2), subiculum, layers II, III, and V of the entorhinal cortex, and the plaques become fewer and larger towards CA4 (110,111). Dense plaques together with reticular plaques (see later) are also observed in lower cortical layers of the association cortices, such as the temporal association cortex. Primary motor and visual cortices also deposit mostly dense-core plaques. In the basal ganglia, pallidum is usually the only striatal nucleus where dense plaques appear. In cerebellum, senile plaques are abundant in the Purkinje and granular cell layers (100) and, interestingly, the highest incidence of amyloid plaques in the cerebellum is found in presenile patients (112). Substantia nigra and other brainstem nuclei also deposit dense plaques and the deposition is proportional to the amyloid burden found in cerebellum (112).

Dense-cored plaques are also observed in mouse models, especially in the environment of a high Aβ40/Aβ42 ratio (81). Importantly, dense-cored plaques lack the characteristic surrounded corona of diffuse amyloid, and frequently show presence of multiple small cores, especially in PSAPP mice where Aβ42 levels are high. Biochemically, the dense cores resemble human dense senile cores being ThS positive and rich in Aβ40. As mentioned, peri-plaque corona is mostly absent, but when present, is predominantly of Aβ42-type as shown in AD. Interestingly, dense-core plaques in mouse models most resemble similar plaques observed in the Flemish APP pathology (34). Dense-cored plaques are observed in all regions of the mouse brain with a preference for the hippocampal region and the hippocampal sulcus (34,107).

Compact deposits lacking a discernable amyloid core are also called “primitive plaques” due to an arguable viewpoint that they progress in time to form dense-core plaques. Such plaques are the most common type of plaques in the cerebral cortex in most AD patients and consist of agglomerated wisps of amyloid deposits arranged in a reticular pattern (99). Typically, these plaques contain almost exclusively Aβ42. When the faint reticular pattern has no readily demarcated edge, these are called “coarse” reticular plaques, and seem to predominate in AD, especially in FAD patients such as those harboring PS1 G384A mutation (113). When the reticular pattern is clearly demarcated by a strong ­amyloid edge on immunohistochemistry, these are called “fine” reticular dense plaque. The latter type of plaques is the most frequent compact, noncored plaque found in the Flemish APP pathology (41) (see further).

In mice models, diffuse plaques are seen especially in very early ages, and dense-diffuse plaques are more common (34). The compacted region contains more Aβ40, and diffuse deposits are predominantly Aβ42-positive. Dense-diffuse deposits can appear throughout the mouse brain, with a relative sparing of the basal nuclei and cerebellum.

When Aβ is deposited in the walls of small-to-medium-sized blood vessels of the brain and of leptomeninges and incites a pathological response, it is referred to as cerebral (congophilic or ThS-positive) amyloid angiopathy of the Aβ type (114). More than 10% of the persons over 60 years of age and ≈80% of the AD patients have at least some degree of CAA (115). CAA affects most frequently the occipital, temporal, and frontal lobes, being less frequently found in hippocampus, cerebellum, and basal ganglia, and almost absent in the white matter and brain stem (116). Leptomeningeal arteries are most frequently involved, followed by cortical arteries, arterioles, and capillaries, while the venous sector is less frequently involved (116). While Aβ deposits seem to appear in the media near the smooth muscle cells in larger vessels, capillary CAA (cap-CAA) has Aβ deposited first on the abluminal side of the basement membrane (117). Interestingly, cap-CAA have frequently small perivascular caps of amyloid or perivascular amyloid fibrils radiating in the surrounding brain tissue, described as dysphoric angiopathy or “drusige Entartung” (114,118,119). In pathologies where CAA is a prevailing feature such as Flemish APP mutation, dysphoric angiopathy is also observed within larger vessels (41). Biochemically, Aβ40 is the major constituent of CAA affecting large leptomeningeal and neocortical vessels, while both Aβ40 and Aβ42 are present in CAA affecting smaller parenchymal vessels.

Vascular Aβ depositions also cause secondary changes in vessels such as hyalinization and degeneration as indicated by increased representation of collagen-IV in the basement membrane. Not infrequently, “double-barreling” or “vessel-within-vessel” configuration is also observed due to deposition of Aβ both in the vessel wall and under glia limitans, separated by the distended perivascular spaces (116,120). Thus, not surprisingly, CAA accounts for approximately 10% of all intracerebral nonfatal hemorrhages in the elderly (121). Interestingly, CAA also seems to be capable of initiating a strong local neuritic and inflammatory pathology (86) and in this respect cap-CAA, but not CAA associated with larger vessels, seems to correlate well with parenchymal pathology (119,122).

Mouse models of amyloidosis also show significant CAA pathology (123–125). As in humans, vascular Aβ in mouse models is also ThS-positive and chiefly constituted of Aβ40. Both CAA and dysphoric angiopathy can be found throughout the mouse brain including thalamus (34).

Recent data also indicate that Aβ is deposited within neurons in brains of AD and Down’s syndrome patients (126,127). While this is a relatively new and understudied area of amyloidosis, it is highly likely that intraneuronal Aβ, similar to a plethora of ­inclusion proteinopathy diseases, may induce cytopathic effects with ­important roles in AD pathogenesis. Several mouse models have been established that deposit intraneuronal Aβ either solely or in early stages of diseases and shall be discussed in relevant sections.

Tau is the major intraneuronal proteopathy in AD inside the body of the neurons (as NFTs) or along dystrophic neurites (as neurophil threads). Early studies on tau deposits recognized phosphorylated tau as the major component of PHF (128–131). In AD, PHF consist of 10–20 nm filaments distributed especially in the entorhinal cortex, hippocampus, amygdala, and association temporal cortex (100). Staging of AD based on the prevalence of NFT pathology is also proposed, which seems to correlate better with the cognitive deficits compared to criterions based on amyloid deposits (132,133).

Tau is also the main pathological finding in FTLD-tau disorders such as Pick’s disease. The histopathological features of Pick’s disease are dominated by argyrophilic tau-positive inclusions (Pick bodies) and swollen achromatic cells (Pick cells). They usually occur in the granular layer of dentate gyrus, pyramidal cells of the CA1, subicular hippocampal sectors, neocortex, and in several subcortical nuclei. In the neocortex, they predominate in the layers II and VI, in contrast to the neurofibrillary tangles in AD that appear especially in layers III and V. By electron microscopy, Pick bodies contain intermediate filaments, 15 nm straight filaments, and some paired helical filaments (134). Biochemical characterization showed that the insoluble tau in Pick bodies consists of 3R-tau isoforms (135,136). However, recent studies have demonstrated much greater biochemical heterogeneity and up to 50% of patients with Pick’s disease having either at least as much 4R as 3R-tau, or even a predominance of 4R-tau (49,137,138). Mouse models expressing various isoforms of tau are discussed hereunder.