Abnormal Intracellular Accumulation and Extracellular Aβ Deposition in Idiopathic and Dup15q11.2-q13 Autism Spectrum Disorders

Background It has been shown that amyloid ß (Aβ), a product of proteolytic cleavage of the amyloid β precursor protein (APP), accumulates in neuronal cytoplasm in non-affected individuals in a cell type–specific amount. Methodology/Principal Findings In the present study, we found that the percentage of amyloid-positive neurons increases in subjects diagnosed with idiopathic autism and subjects diagnosed with duplication 15q11.2-q13 (dup15) and autism spectrum disorder (ASD). In spite of interindividual differences within each examined group, levels of intraneuronal Aβ load were significantly greater in the dup(15) autism group than in either the control or the idiopathic autism group in 11 of 12 examined regions (p<0.0001 for all comparisons; Kruskall-Wallis test). In eight regions, intraneuronal Aβ load differed significantly between idiopathic autism and control groups (p<0.0001). The intraneuronal Aβ was mainly N-terminally truncated. Increased intraneuronal accumulation of Aβ17–40/42 in children and adults suggests a life-long enhancement of APP processing with α-secretase in autistic subjects. Aβ accumulation in neuronal endosomes, autophagic vacuoles, Lamp1-positive lysosomes and lipofuscin, as revealed by confocal microscopy, indicates that products of enhanced α-secretase processing accumulate in organelles involved in proteolysis and storage of metabolic remnants. Diffuse plaques containing Aβ1–40/42 detected in three subjects with ASD, 39 to 52 years of age, suggest that there is an age-associated risk of alterations of APP processing with an intraneuronal accumulation of a short form of Aβ and an extracellular deposition of full-length Aβ in nonfibrillar plaques. Conclusions/Significance The higher prevalence of excessive Aβ accumulation in neurons in individuals with early onset of intractable seizures, and with a high risk of sudden unexpected death in epilepsy in autistic subjects with dup(15) compared to subjects with idiopathic ASD, supports the concept of mechanistic and functional links between autism, epilepsy and alterations of APP processing leading to neuronal and astrocytic Aβ accumulation and diffuse plaque formation.


Introduction
Autism is a developmental disorder characterized by qualitative impairments in reciprocal social interactions, verbal and nonverbal communication, and restricted, repetitive and stereotyped patterns of behavior [1]. Autism is often diagnosed in subjects with genetic disorders, including maternal origin duplications 15q11. 2-q13

Results
The Difference between Intraneuronal Ab Accumulation in dup (15)

Autism, Idiopathic Autism and Control Groups
In all subjects with dup15/autism spectrum disorder (ASD) and the majority of individuals with idiopathic ASD, intraneuronal Ab immunoreactivity was observed in more neurons, and the amount of immunoreactive material was increased in comparison to the control subjects (Fig. 1). The morphology of the intracellular deposits of Ab-positive material was cell type-specific. Cortical pyramidal neurons showed significant heterogeneity of intraneuronal deposits with a mixture of fine granular material and several times larger 4G8-positive granules. In Purkinje cells, fine granular deposits were accumulated in the cell body. In the dentate nucleus, large neurons accumulated fine granular material, whereas small neurons accumulated a few much larger Ab-positive granules. Neurons in the reticulate nucleus in the thalamus contained a mixture of fine granular material and large 4G8-positive granules.
Immunocytochemistry with monoclonal antibodies (mAbs) 4G8 (17-24 aa of Ab) and 6E10 (4-13 aa of Ab) revealed that almost all intraneuronal Ab is 4G8-positive, but only a very small proportion is labeled with 6E10.
Quantitative evaluation of 12 brain subregions/cell types (frontal, temporal and occipital cortex, Purkinje cells, amygdala, thalamus, lateral geniculate body (LGB), dentate gyrus, CA1 and CA4 sectors and dentate nucleus) revealed that in 11 subregions intraneuronal Ab load was significantly greater in the dup (15) autism group than in the control and idiopathic autism cohorts (p,0.0001 for all comparisons). In eight regions (all three cortical subregions, Purkinje cells, amygdala, thalamus, LGB, and dentate gyrus), intraneuronal Ab load differed significantly between the idiopathic autism and control groups (p,0.0001). In structures with almost all neurons positive for Ab-the dentate nucleus and the inferior olive-the amyloid load was insignificantly higher in control subjects than in subjects with idiopathic autism.
Quantitative study revealed different patterns of immunoreactivity in brain subregions (Fig. 2, and Supporting Information, Fig.  S1). The characteristic feature distinguishing the amygdala, thalamus and Purkinje cells of subjects with dup (15) autism was the very high percentage of neurons with strong cytoplasmic Ab immunoreactivity (46%, 46% and 35%, respectively); the percentage was significantly lower in the idiopathic autism group (32%, 38% and 19%, respectively), and very low in control subjects (6%, 6% and 12%, respectively). However, in pyramidal neurons in the frontal, temporal and occipital cortex, the percentage of neurons with strong Ab immunoreactivity was low (3-10%), whereas the total percentage of Ab-positive neurons was significantly higher in the dup (15) group (81-83%) than in the idiopathic autism group (56-71%) and in control subjects (45-51%).
The percentage of Ab-positive neurons and neuronal amyloid load was smaller in the hippocampal formation, especially in the CA1 sector and dentate gyrus of control subjects. The amyloid load was significantly higher in the dup (15) autism group than in control subjects, but the difference in amyloid load between the idiopathic autism and control groups was insignificant (Fig. S1).
The feature distinguishing the LGB, inferior olive and dentate nucleus from other brain structures is the childhood onset of lipofuscin accumulation. In LGB, strong Ab immunoreactivity was observed in 73% of neurons in dup (15) autism and in 62% in idiopathic autism but only 16% of LGB neurons were strongly Abpositive in control subjects. In the dentate nucleus, the percentage of strongly positive neurons was comparable in all three groups (41%, 35% and 41%, respectively), but overall amyloid load was statistically higher in dup (15) autism. The percentage of strongly Ab-positive neurons in the inferior olive was the same in the idiopathic autism and in the dup(15) (32%) group, and there was no difference in overall amyloid load between autistic and control subjects (Fig. S1).

Ab in Glial Cells
Astrocytes and microglia in the control brains were usually Abnegative or contained only traces of Ab immunoreactivity. Enhanced neuronal Ab accumulation in the brains of individuals with autism was associated with Ab accumulation in the astrocytes' cytoplasm and in some microglial cells (Fig. 3). Two patterns of Ab immunoreactivity were observed in astroglia. The most common form was a condensed aggregate of Ab in one pole of the astrocyte soma typical for CA4 sector, some cortical areas but without clear anatomical predilection, and the cerebellar cortex border zone between granule and molecular layers. The less common form was deposition of Ab-immunoreactive granular material in the entire astrocyte body and in a proximal portion of processes radiating from the cell body (frequent in the molecular layer of the cerebral cortex). The increase in the amount of cytoplasmic Ab was often paralleled by (a) a several-fold increase in the number of astrocytes, all of which were Ab-positive (Fig. 3a), (b) clustering of astrocytes in groups of 3-10 cells (Fig. 3b), (c) numerous mitoses as a sign of astrocyte proliferation (Fig. 3c,d) and (d) astrocyte death resulting in deposition of extracellular remnants of Ab aggregates (Fig. 3e) similar to those seen in astrocyte cytoplasm. Extracellular Ab deposits were found in neuropil, but larger aggregates (more than 10) were more often in the perivascular space. Confocal microscopy confirmed Ab accumulation in GFAP-positive astrocytes (Fig. 3, lower panel).

Intracellular Distribution of Amino-terminally Truncated Ab in Neurons
Intraneuronal Ab deposits revealed striking neuron typespecific differences in amount, morphology and cytoplasmic Figure 1. Enhanced intraneuronal accumulation of amino-terminally truncated Ab in autism. Mapping of Ab [17][18][19][20][21][22][23][24] in the brain AN09402 reveals brain region-and cell type-specific patterns of abnormal Ab accumulation in the cytoplasm of neurons and glial cells of a male diagnosed with dup(15), autism and intractable epilepsy, whose sudden unexpected death at the age of 11 years was seizure-related. Almost all neurons in the frontal (FC) and temporal cortex (TC) are 4G8-positive, but the reaction intensity varies from weak to strong. Strong immunoreactivity is observed in many neurons in the lateral geniculate body (LGB), thalamus (Th), amygdala (Amy), Purkinje neurons and basket and stellate neurons in the molecular layer in the cerebellar (Crb) cortex, in many neurons and astrocytes in the CA4, and large and small neurons in the dentate nucleus (DN). Some types of neurons (in the reticular nucleus in the thalamus and small neurons in the dentate nucleus) have different types of deposits: fine-granular and 2-to 3-mm in diameter 4G8-positive deposits. No reaction or only traces of a reaction detected with mAb 6E10 in the frontal cortex, thalamus, cerebellum and dentate nucleus indicate that in intraneuronal Ab the amino-terminal portion is missing, and the prevalent form of Ab is a-secretase product. Immunoreactivity with mAb 4G8 is present in the brain of the control subject (14 years of age), but fewer neurons are positive, and immunoreactivity in the frontal cortex, thalamus, cerebellum and dentate nucleus is weaker than in the affected subject. In the control subject, glial cells are usually 4G8-immunonegative. doi:10.1371/journal.pone.0035414.g001 Figure 2. Two major patterns of alterations in intraneuronal Ab accumulation. Graphs show a high percentage of neurons with strong cytoplasmic immunoreactivity (mAb 4G8) in the amygdala, thalamus and Purkinje cells in subjects diagnosed with dup(15) autism (D15), a lower percentage in idiopathic autism (IA) subjects, and a low percentage in control subjects. In contrast, the characteristic feature of pyramidal neurons in the frontal, temporal and occipital cortex is a low percentage of neurons with strong Ab immunoreactivity, whereas the total percentage of Abpositive neurons is significantly higher in the dup (15) group than in the idiopathic autism group or in control subjects. Differences in Ab immunoreactivity in the dup(15) autism vs. control cohort, the idiopathic autism vs. control group, and the dup(15) autism vs. idiopathic autism are significant (p,0.0001). doi:10.1371/journal.pone.0035414.g002 distribution; however, they had the same immunoproperties. They revealed no reaction or traces of reaction with mAb 6E10 (Fig. 1) or 6F3D (not shown). The morphological diversity of Ab deposits suggested that Ab was present in different compartments of the endosomal-lysosomal pathway and in lipofuscin in neuron typespecific amounts. The number and size of Lamp1- (Fig. 4) lysosomes was from 2 to 3 times more than the number of Abpositive deposits; however, only about 10% of Ab was detected in rab5-positive endosomal vesicles and in LC3B-positive autophagic vacuoles. Colocalization of Ab with COXIV-positive mitochondria was observed in only a very few mitochondria.
Immunoreaction for Ab detected with mAb 4G8 was present in some intracellular autofluorescent granules; however, the 4G8-immunoreactive deposits were detected also in neurons with scanty lipofuscin (Fig. 5) and in neurons with abundant autofluorescent granules. On the other hand, some neurons with scanty immunoreaction for Ab contained numerous autofluorescent granules. The autofluorescent granules were not immunostained with mAb 6E10. Immunoreaction with polyclonal antibody (pAb) R226, specific for the C-terminus of Ab42, showed only a fraction of labeling colocalized with autofluorescent granules. These results indicate that the detected intraneuronal immunostaining reflects accumulation of Nterminally truncated Ab in several cellular compartments, including lipofuscin granules.

Specificity of Immunohistochemical Detection of Ab with mAb 4G8 and 6E10
The epitopes of mAbs 6E10 and 4G8 (4-13 aa and 17-24 aa of the Ab sequence, respectively) are present in full-length APP and APP C-terminal fragments. In brain tissue that has been fixed in formalin for several months, embedded in polyethylene glycol (PEG) and pretreated with 70% formic acid for 20 min, the immunostaining with mAb 4G8 (Fig. 6) and with 6E10 and 7F3D (8-17 aa of Ab; not shown) is consistent with the distribution and amount of Ab, but different from the distribution and amount of neuronal APP. In control brains, antibody R57 detects abundant intraneuronal APP immunoreactivity, but mAb 4G8 reveals only a very limited reaction with Ab. In numerous neuronal populations in autistic subjects, the immunoreactivity for Ab increases very significantly, but most R57 immunoreactive material is 4G8negative, and most 4G8-positive granules are negative for APP. These results indicate that in the examined material, mAbs 6E10, Figure 3. Enhanced accumulation of amino-terminally truncated Ab in autistic subjects astrocytes. Clusters of 4G8-positive astrocytes, especially numerous in the molecular layer (a, b); very frequent mitotic divisions (c, d); and extracellular 4G8-positive Ab deposits, with morphology of astrocytes' cytoplasmic aggregates (e) may reflect the enhanced proliferation, degeneration and death of Ab-positive astrocytes in the brain of autistic subjects. Confocal microscopy confirmed the presence of Ab (green; arrows) in the cytoplasm of GFAP-positive astrocytes (red). Cell nuclei were stained with TO-PRO-3-iodide (blue). doi:10.1371/journal.pone.0035414.g003 4G8 and 7F3D detect Ab but do not bind to neuronal APP detected with pAb R57.

Diffuse Plaque Distribution and Immunoproperties in the Brain of Autistic Subjects
Ab-positive plaques were detected in one of the nine examined subjects diagnosed with dup15 (AN11931), and in two of the 11 subjects diagnosed with idiopathic autism (AN17254 and BB1376). All three subjects were the oldest in each group. In the dup(15) group, a 39-year-old female with autistic features and intractable epilepsy (onset at 9 years of age) and whose death was epilepsy-related had clusters of plaques in several neocortical regions, including the frontal, temporal and insular cortex (Fig. 7). Plaques were also found in the brains of two individuals diagnosed with idiopathic autism, including a 51-year-old subject who had had only one grand mal seizure (Fig. 8), and a 52-year-old individual whose records do not contain information about epilepsy or brain trauma. In both brains, the postmortem examination revealed numerous plaques within the entire cortical ribbon (Fig. S2) and in the amygdala, thalamus and subiculum (not shown).
In all three cases, thioflavin S staining did not reveal fluorescence in the plaques (not shown), suggesting that the amyloid plaques detected in the examined subjects with autism/ dup (15) and idiopathic autism were nonfibrillar. However, positive immunoreactivity with all six antibodies used, including 6E10, 6F3, 4G8, Rabm38, Rabm40 and Rabm42 ( Fig. 7 and 8) and 6F3D (not shown), revealed full-length Ab 1-40/42 peptides. In the plaque area, numerous glial cells, mainly with the morphology of astrocytes, and less numerous, glial cells with the morphology of microglial cells, contained Ab-immunoreactive granular material. In contrast to the presence of full-length Ab peptides in plaques, the Ab peptides in both astrocytes and microglial cells in the plaque perimeter and surrounding tissue were mAb 6E10-and 6F3D-negative, indicating that they were the product of asecretase. They were positive for the three other antibodies, Figure 4. Ab in endocytic vesicles, autophagic vacuoles, lysosomes and mitochondria. Co-localization of Ab (4G8) in neurons in the frontal cortex of a 10-year-old subject diagnosed with autism/dup(15) (AN06365) demonstrates that a small portion of cytoplasmic Ab is stored in rab5positive endocytic vesicles and LC3B-positive autophagic vacuoles, whereas the largest proportion of Ab is colocalized with lysosomal Lamp1. Colocalization of a relatively large portion of cytoplasmic Ab with lysosomal markers appears to reflect the accumulation of products of intracellular degradation of Ab that originated from endocytic and autophagic pathways. The presence of only a few Ab-positive mitochondria immunolabeled with COXIV may suggest that this Ab makes the smallest contribution to the detected neuronal Ab accumulation and degradation pathway. doi:10.1371/journal.pone.0035414.g004 Rabm38, Rabm40 and Rabm42, demonstrating that both astrocytes and microglia accumulate Ab 17-40/42 .
The extracts from the areas of the cerebral cortex in which diffuse plaques were detected by immunohistochemistry contained Ab, mainly Ab1-42, revealed by immunoblotting as a 4-kD band reacting with pAb R226 and mAb 6E10. The levels of Ab1-42 in the samples exceeded 1.5 fmol per 1 mg of extracted proteins, whereas the levels of extracted Ab 1-40 were low, below 0.2 fmol per 1 mg of extracted proteins (Fig. 9).
Immunoblotting of lysates from the cerebral cortex of autistic subjects without plaques and age-matched control subjects detected Ab42 (Fig. 10) and Ab40 (not shown) as a 3-to 4-kD band reacting with the pAb R226 and pAb R162, respectively. The levels of Ab42 in the samples were in the range below 0.5 fmol per 40 mg of total proteins.

Neurofibrillary Degeneration
A very few neurofibrillary tangles (NFTs) were found in the entorhinal cortex and amygdala in a 43-year-old control subject and in the entorhinal cortex and cornu Ammonis of a 47-year-old control subject. A few NFTs were found in the entorhinal cortex, CA1 and parasubiculum in a 51-year-old autistic subject and in the entorhinal and temporal cortex and the amygdala of a 52-year-old autistic subject. Neurofibrillary changes were not found in the dup(15) autism cohort with the oldest examined subject who died at the age of 39 years.

Discussion
The accumulation of intraneuronal Ab is considered a first step leading to amyloid plaque formation in AD [14,[22][23][24]. However, our examination of control brains during the life span showed that intraneuronal Ab also occurs in normal controls and that almost all cytoplasmic Ab peptides are the product of aand c-secretases (Ab 17-40/42 ) [21], whereas, the majority of amyloid in plaques is the product of band c-secretases. This finding suggests that brain region-and neuron type-specific patterns of intraneuronal Ab 17-40/42 peptide accumulation in control brains are a baseline for detection and evaluation of increases associated with autism, FXS, epilepsy, brain trauma or age-associated neurodegeneration, such as AD.

Detection of Ab in Human Postmortem Material
The epitopes of mAbs 6E10 (4-13 aa of the Ab sequence) and 4G8 (17-24 aa) are present in full-length APP and various APP fragments. Recently, Winton et al. [25] demonstrated that neuronal APP is immunolabelled with these two antibodies in  (15) autism was characterized using mAbs 4G8 and 6E10 and pAb R226. Autofluorescent lipofuscin granules were 4G8-negative (cell 1) or partially positive (cell 2), but Ab was also accumulated in lipofuscin-free neurons (cell 3). The neurons revealed only traces of reaction with mAb 6E10 and a moderate amount of pAb R226-positive Ab42, which was partially co-localized with autofluorescent lipofuscin. doi:10.1371/journal.pone.0035414.g005 mouse brain fixed for 24 hours in 10% neutral buffered formalin. However, the pattern of immunostaining in human brain fixed in formalin for at least several months, dehydrated almost 3 weeks in ascending concentrations of ETOH, and embedded in PEG indicates that mAbs 4G8, 6E10 and 6F3D do not detect APP in tissue subjected to this process. The role of technical factors in the loss of access of these antibodies to their epitopes in APP was previously documented in studies of tissue fixed in formalin for 10 days and in studies of cultured cells [26,27]. Several observations in this report indicate that these antibodies do not detect APP. Massive immunolabelling of neuronal APP with R57 is in striking contrast with the presence of only traces of 6E10 and 6F3D immunoreactivity in these cells and the only partial colocalization of Ab and APP labeling in the amyloid-rich neurons of autistic subjects. These data indicate that in the examined material, APP is detected with the APP-specific antibody R57, but not with mAbs 4G8, 6E10,and 6F3D, which, however, detect Ab. One may assume that the epitopes of these antibodies, but not the R57 epitopes, are blocked or modified in APP molecules during long exposure to chemicals used for fixation, dehydration and embedding. Consistent with immunocytochemistry, Western blotting identifies 3-4 kD Ab not only in subjects with diffuse plaques, but also in autistic subjects without plaques and in control subjects.

Excessive Accumulation of Ab 17-24 in Neurons in Idiopathic Autism and dup(15) Autism
This is the first report documenting excessive accumulation of Ab in the neurons of subjects with idiopathic autism and an even more pronounced accumulation in the dup(15) autism cohort. Two patterns of excessive accumulation distinguish these two cohorts from control subjects and indicate that excessive accumulation is neuron type/brain region-specific. Type 1 of altered Ab accumulation is reflected in an increase in the percentage of neurons with strong Ab accumulation by 7.6-fold in the amygdala and thalamus and by 4.5-fold in the LGB in individuals with dup(15) autism in comparison to the control group. A similar (by 5.36, 6.36 and 3.96, respectively) and statistically significant increase was found in the idiopathic autism group. Type 2 of altered Ab accumulation is reflected in a more uniform increase in the percentage of neurons with combined strong, moderate and weak immunoreactivity. Again, this pattern is observed in both autistic cohorts in the pyramidal neurons in all three examined cortical regions.
These findings suggest that metabolic alterations are similar in both types of autism and that the severity of these alterations is less pronounced in idiopathic autism than in autism caused by dup (15). The significant increase in the percentage of neurons with enhanced cytoplasmic Ab load in idiopathic autism and the fact that almost all of this Ab is the product of a-secretase show the striking similarity to increased levels of sAPP-a in blood plasma in 60% of autistic children (6,16). In studies by Sokol et al. [6] and Ray et al. [16], aggressive behavior was identified as associated with increased levels of sAPP-a. Bailey et al. [9] also detected a significant increase in sAPP-a levels in 60% of autistic children but with no association between the severity of aggression, social or communication sub-scores and increased levels of sAPP-a. Due to the neurotrophic properties of sAPP-a, the authors proposed that an increased level of the products of a-secretase may help identify a subset of children in which early regional brain overgrowth is necessary and sufficient for the development of autism and may even represent a mechanism regulating overgrowth in autism. However, the most pronounced accumulation of amino-terminally truncated Ab observed in the dup (15) autism cohort with microcephaly [28] indicates that intraneuronal Ab accumulation of the products of a-secretase is not associated with brain overgrowth. Our data identify a dup (15) autism subcohort with microcephaly, more severe clinical phenotype, very early onset of seizures, a high percentage of intractable seizures, and a high prevalence of sudden unexpected death in epilepsy (SUDEP) as associated with the highest percentage of neurons accumulating a-secretase product. Trafficking of Ab [17][18][19][20][21][22][23][24]
This study revealed that another 20-30% of neuron Ab 17-40/42 is present in lipofuscin, which is the final product of cytoplasmic proteolytic degradation of exogenous and endogenous substrates. During the entire lifespan, lipofuscin gradually accumulates in neurons [39]. The age of onset and dynamics of lipofuscin deposition are cell type-specific [40,41]. Our previous study revealed that neurons in the inferior olive, dentate nucleus and lateral geniculate body start accumulating lipofuscin and Ab 17-40/ 42 early in life and that this accumulation progresses with age at region-specific rates [21]. The confocal microscopy study indicates that in spite of the known nonspecific binding of some antibodies to lipofuscin, the selection of the immunostaining protocol and the setting of proper thresholds in confocal imaging applied in this study reveal the selectivity of mAbs 4G8 and 6E10, and pAb R226 binding to some lipofuscin deposits.
The pattern of both Ab and lipofuscin accumulation can be modified in early childhood in subjects with autism and even more significantly in individuals with dup(15) autism. The difference is detectable as an increase in the percentage of Ab 17-40/42 immunoreactive neurons, the amount of immunopositive material per neuron, and the number of brain regions and neuron types affected in both children and adults. Detected changes in Ab accumulation may reflect abnormal accumulation of lipofuscin, as reported by Lopez-Hurtado and Prieto [42]. An increase in the number of lipofuscin-containing neurons by 69% in Brodmann area (BA) 22, by 149% in BA 39, and by 45% in BA 44, in brain tissue samples from autistic individuals 7 to 14 years of age, was observed together with a loss of neurons and glial proliferation. However, enhanced lipofuscin accumulation is not unique for idiopathic autism or autism/dup (15). It has been reported in Rett syndrome [43], an ASD, as well as in several psychiatric disorders, including bipolar affective disorder [44] and schizophrenia [45,46].
Enhanced lipofuscin accumulation and enhanced Ab 17-40/42 immunoreactivity in the majority of the examined brain structures in most of the individuals with autism and the subjects with dup(15) may be a reflection of enhanced oxidative stress. Oxidative stress contributes to protein and lipid damage in cytoplasmic components, their degradation in lysosomal and autosomal pathways, and the deposition of products of degrada-tion in lipofuscin or their exocytosis [47,48]. The link between oxidative stress, cytoplasmic degradation and lipofuscin deposition is supported by the presence of oxidatively modified proteins and lipids in lipofuscin [39,49,50]. A significant increase in malondialdehyde levels (a marker of lipid peroxidation) in the plasma of autistic children [51] and in the cerebral cortex and cerebellum [52] may reflect oxidative damage leading to enhanced degradation, and the possible increased turnover of affected cell components.

Biological Activity of N-terminally Truncated Ab
The results of confocal microscopy suggest that on average, 30% of neuronal Ab is present in lysosomes and another 30% in lipofuscin. However, the biological consequences of accumulation of Ab, in the lysosomes or in lipofuscin are not known. Nterminally truncated Ab peptides exhibit enhanced peptide aggregation relative to the full-length species [53] and retain their neurotoxicity and b-sheet structure. Soluble intracellular oligomeric Ab (oAb) species inhibit fast axonal transport (FAT) in both anterograde and retrograde directions [54]. Inhibition of FAT results from activation of endogenous casein kinase 2. Altered regulation of FAT markedly reduces transport of synaptic proteins and mitochondria in the AD brain and in AD mouse models that accumulate oAb [55]. Dysregulation of FAT results in distal axonopathies with a reduced delivery of critical synaptic elements required for the integrity, maintenance and function of synapses [54].
The in vitro studies suggest that Ab 17-24 is toxic to neurons. Treatment of SH-SY5Y and IMR-32 human neuroblastoma cells with Ab 17-24 causes apoptotic death similar to in cells incubated with Ab1-42, whereas treatment with Ab17-40 results in a lower level of apoptosis, comparable to experimental exposure to Ab1-40. This apoptosis is mediated predominantly by the caspase-8 and caspase-3 pathways [56]. However, in vitro studies of the neuronal response to exogenous Ab peptides do not replicate the neuronal exposure to endogenous Ab 17-40/42 trafficking inside vesicles and vacuoles of lysosomal pathway.

Ab 1-40/42 in Diffuse Plaques of Autistic Subjects
The presence of diffuse nonfibrillar plaques in two autistic subjects who were more than 50 years old and in one 39-yearold subject with autism/dup (15) suggests that in the fourth/fifth decade of life, there is an increased risk of the second type of changes: activation of the amyloidogenic pathway of APP processing with band c-secretases, resulting in focal deposition of Ab 1-40/42 in plaques . It was hypothesized that Ab  peptides may initiate and/or accelerate plaque formation, perhaps by acting as nucleation centers that seed the subsequent deposition of relatively less amyloidogenic but apparently more abundant full-length Ab [53,57,58]. Gouras et al. [59] considered intracellular Ab 42 accumulation an early event leading to neuronal dysfunction. The Ab 1-40/42 -positive diffuse plaques in the brains of autistic subjects are different from the Ab 17-40/42 -positive cerebellar diffuse plaques detected in DS [57,60]. Diffuse amorphous nonfibrillar Ab deposits, called amorphous plaques [61], pre-plaques [62] or pre-amyloid deposits [63], are considered to be of neuronal origin [64][65][66][67] (15), autistic features, and intractable seizures (age of onset 9 years) and whose death was epilepsy-related, are 6E10-, 4G8-, Rabm38-, Rabm40-and Rabm42-positive. Reaction with Rabm38 and Rabm42 was weaker than with other antibodies. Almost all glial cells with the morphology of astrocytes detected in the plaque perimeter had a large cluster of granular material located usually at one cell pole and positive with all antibodies detecting Ab, except 6E10. doi:10.1371/journal.pone.0035414.g007 and are formed selectively in projection areas of distant affected neuronal populations [68]. Diffuse plaque formation in autistic subjects suggests the activation of the secretory pathway and the synaptic release of Ab 1-40/42 .
The presence of Ab 17-40/42 in astrocytes in Ab 1-40/42 -positive diffuse plaques suggests that the full-length Ab released by neurons is phagocytosed and processed by local astrocytes. One may hypothesize that the proliferation of Ab-positive astrocytes, the increase of cytoplasmic Ab immunorectivity in astrocytes, the presence of Ab in all astrocytes in the affected region, astrocyte death and the deposition of large aggregates of extracellular Ab in the cerebral cortex or hippocampus of autistic children and young adults is a response to the elevated levels of extracellular Ab 17-40/ 42 and/or Ab 1-40/42 . Therefore, the number of Ab-positive astrocytes may be an indicator of the local concentration of extracellular Ab not only in plaque-positive but also in plaquenegative brain regions, occurring decades before plaque formation. Cytoplasmic granular immunoreactivity (Ab17-23 and Ab8-17) was reported in astrocytes in AD [69]. In astrocytes, intracellular Ab appears in lysosomes and lipofuscin [70,71]. It defines the role of astrocytes in the uptake of different species of Ab in diffuse and neuritic plaques and their subsequent degradation in lysosomes and storage of products of degradation in lipofuscin [69].
In the examined autistic cohort, the early onset of intractable epilepsy and the epilepsy-related chronic and acute brain trauma appear to be additional risk factors for APP pathway activation and diffuse plaques formation. Repetitive brain trauma, including that related to epilepsy and head banging, produces a chronic traumatic encephalopathy with the associated deposition of Ab, most commonly as diffuse plaques [72][73][74]. In acute traumatic brain injury, diffuse cortical Ab deposits were detected in 30% to 38% of cases 2 hours after injury [75][76][77].
The presence of a few NFTs in the entorhinal cortex, cornu Ammonis and amygdala in 43-and 47-year-old control subjects and in these structures and in the parasubiculum and temporal cortex of 51-and 52-year-old autistic subjects is consistent with the topography and amount of age-associated neurofibrillary degeneration and NFT distribution observed in the general population [78].
In conclusion, this postmortem study of Ab distribution in the brain of subjects with idiopathic autism and dup (15)

Material, Clinical and Genetic Evaluation
The brains studied were from nine individuals diagnosed with dup(15) ages 9 to 39 years (five males and four females), 11 subjects with idiopathic autism ages 2 to 52 years (10 males and one female), and eight control subjects ages 8 to 47 years (four males and four males) ( Table 1). Medical records were obtained following consent for release of information from the subjects' legal Figure 8. Full-length Ab in diffuse plaques, and truncated Ab in astrocytes in idiopathic autism. Diffuse plaques in the frontal cortex of a 51-year-old subject (AN17254) diagnosed with idiopathic autism, who had had only one grand mal seizure and died because of cardiac arrest, are immunopositive when stained with all five antibodies (6E10, 4G8, Rabm38, Rabm 40 and Rabm 42), but granular material in the cytoplasm of glial cells is immunopositive for all antibodies used except 6E10. doi:10.1371/journal.pone.0035414.g008 guardians. The study was approved by the Institutional Review Boards for the New York State Institute for Basic Research in Developmental Disabilities; the University of California, Los Angeles; and Nemours Biomedical Research, duPont Hospital for Children, Wilmington. Clinical and genetic studies were performed as described previously [28]. Clinical characteristics were based on psychological, behavioral, neurological and psychiatric evaluation reports. To confirm a clinical diagnosis of autism, the Autism Diagnostic Interview-Revised (ADI-R) was administered to the donor family [79].
Molecular genetic evaluations, using antemortem peripheral blood samples and lymphoblast cell lines for eight of the dup (15) cases, included genotyping with 19-33 short tandem repeat polymorphisms from chromosome 15, Southern blot analysis of dosage with 5-12 probes, measurement of the methylation state at SNRPN exon a, as described [80], and array comparative genomic hybridization [81]. Duplication morphology was confirmed by fluorescent in situ hybridization [80].
In eight cases, tetrasomy, and in one case, hexasomy of the Prader-Willi/Angelman syndrome critical regions was detected. In eight cases, the origin of abnormality was maternal; in one case, the origin was not determined. In the examined dup(15) group, seven of nine subjects (78%) were diagnosed with autism or ASD, and seven had seizures. In six cases (67%), SUDEP was reported. In the idiopathic autism cohort, two subjects (8-year-old male, HSB4640, and 52-year-old male, BB1376), were diagnosed with the ASD (pervasive developmental disorder -not otherwise specified and high-functioning atypical autism, respectively). In all other cases, the clinical diagnosis of autism was confirmed with ADI-R. One brain hemisphere was preserved for neuropathological and immunocytochemical studies. Methods and results of neuropathological evaluations of developmental abnormalities have been summarized in our previous reports [28,82]. The mean postmortem interval varied from 23.9 h in the dup(15) cohort to 19.6 h in the idiopathic autism cohort and 15.0 h in the control group. One brain hemisphere from each subject was fixed in 10% buffered formalin for a period ranging from six weeks to several months, dehydrated in a graded series of ethanol, infiltrated and embedded with PEG (Sigma) [83] and stored at 4uC. Tissue blocks were then cut into 50-mm-thick serial sections and stored in 70% ethyl alcohol. Two brains (AN17254 and BB1376) were embedded in celloidin, as described [82] and were cut alternatively into 200and 50-mm-thick serial sections.

Immunocytochemistry and Confocal Microscopy
Brain Bank identification of the tissue samples is listed in Table 1, to maintain non-overlapping records of results of brains examined in different projects. Immunocytochemistry and confocal microscopy were applied to characterize (a) Ab distribution in cells in the cerebral cortex, subcortical structures, cerebellum and brainstem and in diffuse plaques; (b) the Ab peptide properties; and (c) Ab distribution in endosomes, lysosomes, autophagic vacuoles, mitochondria and lipofuscin ( Table 2). mAbs 6E10 (Covance, Inc., Princeton, Inc.) and 6F3D (Novocastra Lab. Ltd., Newcastle, UK) were used to characterize the N-terminal portion of Ab. mAb 6E10 recognizes an epitope in residues 4-13 of Ab [84,85]. mAb 6F/3D recognizes an epitope in residues 8-17 of Ab. The middle portion of Ab was detected with mAb 4G8, which recognizes an epitope in residues 17-24 of Ab [86]. The carboxyl terminus of Ab was characterized with rabbit monoclonal antibodies Rabm38, Rabm40 and Rabm42, which detect Ab 238 , Ab 240 , and Ab 242 , respectively [87]. The specificity of mAbs 4G8 and 6E10 for Ab was verified in the examined postmortem human brain tissue by double immunolabeling with pAb R57 detecting APP C-terminal aa 671-695.
To detect intracellular Ab peptides and amyloid in plaques, free-floating sections were treated with 70% formic acid for 20 minutes [88]. The endogenous peroxidases in the sections were blocked with 0.2% hydrogen peroxide in methanol. The sections were then treated with 10% fetal bovine serum in phosphate buffer solution (PBS) for 30 minutes to block nonspecific binding. The primary antibodies were diluted in 10% fetal bovine serum in PBS and sections were treated overnight at 4uC. The sections were washed and treated for 30 min with either biotinylated sheep antimouse IgG antibody or biotinylated donkey anti-rabbit IgG antibody diluted 1:200. The sections were treated with an extravidin peroxidase conjugate (1:200) for 1 h, and the product of reaction was visualized with diaminobenzidine (0.5 mg/mL with 1.5% hydrogen peroxide in PBS). After immunostaining, sections were lightly counterstained with cresyl violet. To detect fibrillar Ab in plaques, sections were stained with Thioflavin S and examined in fluorescence.
Double immunofluorescence for Ab (mAb4G8) and for astrocytes (GFAP; rabbit polyclonal antibody, pAb, Sigma) was carried out to confirm the presence of Ab in astrocytes. Confocal microscopy was applied to detect Ab localized in neuronal cytoplasmic organelles. Ab in lysosomes was detected by using lysosomal-associated membrane protein marker (LAMP1; Abgent, San Diego, CA). Early endosomes were immunodetected with rabbit pAb Rab5 (Ab13253; Abcam, Cambridge, MA), whereas autophagic vacuoles were immunolabelled with rabbit mAb LC3B (Cell Signaling Technology Inc., Danvers, MA). Mitochondria were detected with the rabbit mAb COXIV Alexa Fluor 488 conjugated (Cell Signaling Technology). To detect Ab, brain sections were treated with 70% formic acid for 20 min, washed in PBS 2x 10 min and double-immunostained using mAb 4G8 and antibodies detecting markers of cytoplasmic organelles. Affinitypurified donkey antisera against mouse IgG labeled with Alexa Fluor 488 and against rabbit IgG labeled with Alexa Fluor 555 (both from Molecular Probes/Invitrogen) were used as secondary antibodies. TO-PRO-3-iodide (Molecular Probes/Invitrogen) was used to counterstain cell nuclei. Absence of cross-reaction was confirmed as previously described [26]. Images were generated using a Nikon C1 confocal microscope system with EZC1 image analysis software.

Comparison of Intraneuronal Ab Accumulation in Examined Cohorts
Semiquantitative estimation of intraneuronal Ab was performed without knowledge of the subject's age, gender or clinical diagnosis or the neuropathological diagnosis of the tissue being analyzed. Evaluation was performed at a workstation consisting of Axiophot II light microscope, specimen stage with 3-axis computer-controlled stepping motor system (Ludl Electronics; Hawthorne, NY), CCD color video camera (CX9000 MicroBrightField Bioscience, Inc., Williston, VT) and stereology software (Stereo Investigator, MicroBrightField Bioscience Inc.). Grid size and the virtual test area were designated individually for each brain region to adjust to the region of interest size and shape. Intraneuronal Ab accumulation has been estimated by four neuropathologists in 12 brain structures including frontal, temporal and occipital cortex, amygdala, thalamus, lateral geniculate body, sectors CA1 and CA4, and dentate gyrus in the hippocampal complex, Purkinje cells and dentate nucleus in cerebellum, and inferior olive in the brainstem. The number of 4G8-negative neurons and neurons with weak (,10 immunopositive granules per cell), strong (condensed mass of indistinguishable small and large immunoreactive granules) and medium (.weak and ,strong) immunoreactivity was determined using a 640 objective lens. For each subject, from 100 to 180 neurons were examined per region of interest in sections immunostained with mAb 4G8. Inspection of the entire cell cytoplasm by using micrometer screw contributed to precise rating of amyloid load in each examined neuron.
Differences in the estimated cytoplasmic neuronal Ab load were examined using the Mann-Whitney U (Wilcoxon signed ranks) test or, for comparison of all three groups, the Kruskal-Wallis one-way ANOVA (an extension of the U test) [90]. Statistics were computed from pooled data from each group [dup (15) autism, idiopathic autism, control], where sampled neurons immunoreactivity was categorized as strong, medium, weak or none.

Western Blotting
Frozen temporal cortex samples from three control and three autistic subjects were homogenized in 106volume of 10 mM TRIS buffer containing 0.65% NP-40, 1 mM EDTA and Complete protease inhibitor coctail (Roche, Mannheim, Germany) in a Potter-Elvehjem homogenizer and sonicated for 2 minutes. Protein content in lysates was measured by BCA assay (Pierce). Forty mg of total lysate proteins were loaded per lane for PAGE in 8-15% gradient gels. Tissue samples from formalin-fixed PEG or celloidin-embedded brains of three subjects with diffuse plaques detected by immunocytochemistry (39-year-old female diagnosed with dup(15) autism, a 51-year-old autistic subject, and a 52-year-old subject with atypical autism) and two subjects without plaques (48-year-old autistic and a 47-year-old control subject) were used for protein extraction. From 50-mm-thick sections, approximately 120 mm 2 of affected cortex was dissected (approximately 6 mm 3 of tissue), rehydrated in PBS and homogenized in Potter-Elvehjem homogenizer in PBS containing 0.5% sodium deoxycholate, 0.1% SDS and 1% NP-40 (RIPA buffer). After sonication two times for three minutes, the material was centrifuged at 16,000g for 20 minutes, and supernatants were collected as RIPA extracts. Protein content in the extracts was measured by the BCA assay (Thermo Scientific, Rockford, IL). For Aß detection with R162, R226, and mAb 6E10, the amounts of extracted proteins loaded per lane were 3, 3 and 6 mg, respectively. The proteins were subjected to PAGE in 8-15% gradient gels, transferred onto nitrocellulose and probed with antibodies specific for C-terminus of Aß40 (R162) and Aß42 (R226), and N-terminus-specific mAb 6E10. Figure S1 Neurons with low and high amyloid load. In control brains, the percentage of Ab-positive neurons and their amyloid load is much lower in CA1 than in CA4 sector and is very low in the granule neurons in the dentate gyrus. The percentage of Ab-positive neurons and amyloid load is significantly higher in the dup(15) autism cohort than in the control and idiopathic autism groups (p,0.0001), but the difference between idiopathic autism and control is insignificant. The characteristic feature of the LGB, inferior olive and dentate nucleus of control subjects is the very high percentage of Ab-positive neurons and the highest amyloid load among the examined 12 structures. The increase of amyloid load is undetectable in the inferior olive and is minimal in the LGB and dentate nucleus of subjects with idiopathic autism and dup (15) autism. (TIF) Figure S2 Topography and morphology of neocortical diffuse plaques. Low magnification demonstrates diffuse plaques immunostained with mAb4G8 (17-24 aa) in frontal, temporal and occipital cortex (FC, TC and OC, respectively) in the brain of a 39-year-old female diagnosed with dup(15) autism, a 51-year-old autistic male, and a 52-year-old subject with atypical autism. (TIF)