Autism is a complex neurodevelopmental disorder characterized by impaired reciprocal social interactions, impaired verbal and nonverbal communication, and stereotyped patterns of repetitive behavior, with clinical symptoms developing in early childhood [1]. The cause of autism is unknown, but a combination of genetic, epigenetic, and environmental factors has been proposed to play a major role in the etiology of autism. Certain genetic disorders are associated with a particularly high incidence of autism, e.g., it is as high as 69% among individuals who have a duplication of chromosome 15q11.2-q13 of maternal origin [dup(15)] [2, 3].

Altered metabolism of amyloid precursor protein (APP) in autism is indicated by higher plasma levels of secreted APP—two or more times—in children with severe autism and aggression than in children without autism, and by lower levels of amyloid-beta (Aβ) 40 than in controls [4, 5]. Hence, an increased processing of APP by alpha-secretases has been proposed to contribute to autism [5, 6]. Our recent postmortem studies demonstrated intracellular deposits of Aβ truncated on the amino-terminal side—the Aβ17-40/42—in neurons and glia in the brain cortex, cerebellum and in subcortical nuclei. The percentage of amyloid-positive neurons as well as intraneuronal Aβ load were significantly higher in individuals diagnosed with idiopathic autism and in subjects diagnosed with dup(15) and autism spectrum disorder (ASD) than in controls [7]. The pathological consequences of intracellular accumulation of N-truncated Aβ by neurons and glia are not known. However, accumulation of full-length Aβ 1-42 and Aβ 1-40 in Alzheimer’s disease and Down syndrome is associated with production of reactive oxygen species and contributes to oxidative stress [8–10].

Increased formation of reactive oxygen species has been implicated in the pathophysiology of autism, and markers of oxidative stress have been detected in autism, even at an early age. Peripheral blood in children with autism contains elevated levels of malondialdehyde (MDA), an indicator of lipid peroxidation [11], and increased levels of thiobarbituric acid–reactive substances (TBARS) [12]. Increased oxidative stress in children with autism was also suggested by increased nitric oxide (NO) levels in red blood cells [13] and higher urinary excretion of TBARS, lipid hydroperoxides, 4-hydroxy-2-nonenal (HNE), and protein carbonyls along with low levels of urine antioxidants [14]. Severity of autism appears to be correlated with urinary excretion of 8-isoprostane-F2 alpha [15]. A link of oxidative stress in autism to malfunction of anti-oxidative mechanisms is indicated by reduced serum levels of ceruloplasmin and transferrin—the major proteins with anti-oxidative properties [11], as well as significantly lower ratio of reduced/oxidized glutathione in the plasma, decreased methionine cycle turnover [16, 17], increased plasma activities of glutathione peroxidase [13], xanthine oxidase and superoxide dismutase (SOD) [12], and decreased catalase activity [12].

Oxidative stress in the brain in autism is indicated by oxidative DNA damage: increased levels of 3-nitrotyrosine (3-NT) [18–20] and 8-oxo-deoxyguanosine [18]. The distribution of 3-NT indicates brain region–specific enhancement of oxidative stress, particularly in cerebellum and cortical areas involved in speech processing, sensory and motor coordination, emotional and social behaviors, and memory [20]. Other biomarkers of oxidative stress—reduced glutathione (GSH) levels, and decreased ratio of GSH to oxidized glutathione in cerebellum and temporal cortex—indicate a role of deficient glutathione antioxidant defense in certain brain regions in the development of oxidative stress in autism [18, 21]. The pathomechanisms of oxidative stress in autism have not been determined; however, it can be attributed, in part, to activation of the immune system, which has been confirmed in the brain [22] and in the peripheral blood [23]. Significantly higher levels of 3-chlorotyrosine—a biomarker of a chronic inflammatory response—in cerebellum and temporal cortex may link oxidative stress in the brain in autism with a chronic inflammatory response, whereas a decreased aconitase activity in the cerebellum in autism indicates increased mitochondrial superoxide production [18].

The aim of this project was to test the hypothesis that intracellular deposits of N-truncated Aβ are not biologically neutral, but may be a source of reactive oxygen species in brain cortex neurons in subjects with idiopathic autism and in subjects with dup(15) with autism.

Oxidative stress has been detected in the brain and in peripheral organs of autistic subjects [11–13, 18–21]. Our data link intraneuronal accumulation of N-terminally truncated Aβ in dup(15)/autism and in idiopathic autism to elevated intracellular levels of lipid peroxidation products, the accepted markers of oxidative stress. Analysis of the relationship between deposits of Aβ of various sizes and oxidatively modified lipids in three-color immunofluorescence revealed almost complete co-localization of intracellular Aβ with lipid peroxidation products in autism, dup(15)/autism and control brains. However, the absence of HNE and MDA in the majority of the smallest Aβ immunoreactive granules, apparently representing the earliest steps of Aβ accumulation, suggests that cellular accumulation of Aβ precedes formation of lipid peroxidation products. Hence, Aβ deposits are most likely the source of oxidative stress, rather than oxidative stress being the trigger for Aβ accumulation.

Our present and previous [7] studies revealed higher levels of accumulation of N-truncated Aβ in neurons in individuals with idiopathic autism and with dup(15)/autism, than in controls. Detection of Aβ and its N-terminal truncation in both projects was based on immunohistochemical detection with mAbs 4G8 (17–21 aa of the Aβ sequence [30, 32]) but not 6E10 (4–13 aa). Even though these antibodies can recognize the epitopes in full-length APP and in APP fragments [35], in human brains fixed in formalin for at least several months, dehydrated in ethanol and embedded in polyethylene glycol the mAbs 4G8 and 6E10 do not react with APP, but detect only Aβ [7, 27, 28]. Thus, the deposited Aβ species are most likely mainly Aβ17-40/42 [36], Aβ11-40/42 [36, 37], Aβ11(pE)-40/42 [38], and possibly also other.

To evaluate the amounts of Aβ and also lipid peroxidation products we applied in this report measuremets of fluorescence in digital images, instead of morphological evaluation of intensity of the immunoreaction. This new approach—quantification of relative protein amounts basing on immunofluorescence imaging—has been recently shown to be a reliable method [39]. Using known aliquots of cytochrome C embedded in gelatin these authors revealed a high accordance of the amounts of protein detected with primary and secondary antibodies and fluorescence microscopy with the actual protein levels present.

The presence of large deposits of N-truncated Aβ inside neurons suggests that the peptide is, at least in part, aggregated. N-truncation of Aβ peptides is known to enhance their aggregation and formation of the β-sheet structure [40]. The biological effects of accumulation of N-truncated Aβ are not well characterized. The peptides have neurotoxic properties—especially the Aβ 17-42 species [40, 41]—leading to apoptosis mediated mainly by the caspase-8 and caspase-3 pathways [41]. However, we did not observe apoptotic nuclei in neurons, possibly because of the young age of the subjects. Aggregated and oligomerized full-length Aβ 1-40/42 is involved in the formation of reactive oxygen species through binding transitional metals copper and iron. N-terminal truncation of Aβ lowers the ability to form reactive oxygen species, because copper is bound to His6, His13 and His14, in the N-terminal sequence of Aβ, whereas the carbonyl of alanine-2 is an oxygen ligand [10]. However, methionine-35 can also be oxidized to form a sulfuranyl radical, which subsequently can cause lipid peroxidation [8, 9]. These data and the results presented in this study suggest that enhanced accumulation of intracellular N-truncated Aβ may result in increased production of reactive oxygen species and increased formation of lipid peroxidation products.

Our finding of higher levels of lipid peroxidation products in neurons in autism and dup(15)/autism than in controls is in agreement with the reported significantly increased levels of MDA in lysates of the cerebral cortex and cerebellum of autistic subjects [42]. The localization of lipid peroxidation products in almost all mitochondria, in some autophagic vacuoles and lysosomes, and in all lipofuscin granules most likely reflects the sites of formation of lipid peroxidation products, their intracellular trafficking, and storage of non-degradable components. Mitochondria generate the superoxide anion radical (O₂*-), hydrogen peroxide (H₂O₂), and hydroxyl radical (HO*) in the electron transport chain reactions (review [43]). The increased levels of lipid peroxidation products in mitochondria in idiopathic autism and dup(15)/autism (Figure  2), as well as their co-localization with N-truncated Aβ, may have very significant biological consequences for the neuron. In mitochondria, MDA has an inhibitory effect on mitochondrial complex I- and complex II-linked respiration and significantly elevates production of reactive oxygen species and protein carbonyls [44]. Thus, increased formation of HNE and MDA in neurons with N-truncated Aβ deposits may enhance the formation of reactive oxygen species in mitochondria and may be the cause of, or enhance an existing mitochondrial dysfunction in autism and dup(15)/autism. Impaired mitochondrial function has been detected in children with autism, with increased rates of hydrogen peroxide production [45], and reduced expression of numerous genes of mitochondrial complex I, III, IV, and V [46]. In autism the levels of HNE protein adducts are also increased in erythrocyte membranes and in plasma; the imbalance between pro- and anti-oxidative mechanisms may be linked to higher levels of unbound iron [47].

The presence of lipid peroxidation products in lysosomes and in all lipofuscin granules in neurons reflects the sites of processing of oxidatively modified molecules, which are known to be partially degraded in lysosomes and deposited in lipofuscin [48, 49]. Detection of lipofuscin in brain neurons in children with autism, even younger than 10 years of age, is in agreement with the reported previously increased frequency of cells containing lipofuscin—particularly in Brodmann areas 22 and 39—in autism [50].

We have not found significantly increased levels of total HNE- and MDA-modified proteins in whole brain lysates, possibly because the Aβ-accumulating neurons constitute only a minor portion of the brain cortex, or because significant fractions of HNE and MDA detected by immunofluorescence were not protein-bound. However, lipid peroxidation products are known to modify proteins. The binding is aminoacid–specific: for HNE, it is histidine (particularly when flanked by basic amino acid residues), and less frequently Cys, and Lys. Proteins particularly susceptible to modifications by HNE include enzymes involved in glycolysis and ribosomal proteins [51]. Covalent binding of HNE to enzymes frequently causes their quick inactivation e.g., glyceraldehyde-3-phosphate dehydrogenase [52] and ion-transporting ATPases [53, 54].

Detection of HNE and MDA in the nuclei of numerous neurons points out to yet another possible biological effect of lipid peroxidation products—their influence on transcription. MDA reacts with DNA forming guanine derivatives, which in the transcribed DNA strand of expressed genes strongly inhibit RNA polymerase II [55]. MDA can also affect transcription through modification of aldehyde dehydrogenase 2—which in the nucleus plays an important role in transcription repression through its interaction with histone deacetylases—by inhibiting its nuclear translocation and its repressive activity in general transcription [56]. The great variability of nuclear reactions for HNE and MDA (Figures  4 , 6, 7, 8, 9) suggests that susceptibility for this effect of lipid peroxidation products may vary among cell subpopulations.

Cellular co-localization of lipid peroxidation products and Aβ may have significant biological consequences. HNE may covalently bind Aβ at multiple locations through 1,4 conjugate addition and/or Schiff base formation, which leads to covalent cross-linking of Aβ and formation of Aβ protofibrils [57]. Furthermore, oxidative stress can even trigger deposition Aβ, as shown for cultured brain vascular smooth muscle cells [28, 29]. Accumulation of HNE may also increase neuronal levels and secretion of Aβ by up-regulating expression of BACE-1—APP secretase-β—through activation of the c-Jun N-terminal kinases and p38 [58]. Increased Aβ levels and oxidative stress in neuronal progenitor cells may impair neurogenensis in postnatal brain. Full-length Aβ—even the less neurotoxic species 1-40—causes oxidative damage in human neuronal progenitor cells and impairs proliferation, migration, and formation of processes [59].

Analysis of the levels of Aβ and HNE in individual neurons indicates the existence of two neuronal populations in the frontal cortex with a distinct relationship between accumulation of Aβ and HNE: in one, the ratio is similar to that for Aβ and MDA; in the other, moderate contents of Aβ were accompanied by a massive accumulation of HNE. This significantly enhanced formation of HNE in a subpopulation of neurons could be related to some neuronal lineages and/or to a certain functional state of neuron. Further studies are necessary to characterize cells with these properties.

We propose that formation of deposits of N-truncated Aβ, which become a source of reactive oxygen species and lipid peroxidation products, causes neuronal dysfunction in autism. Accumulation of lipid peroxidation products causes/increases dysfunction of mitochondria and further increases Aβ accumulation leading to a self-enhancing process. Thus, autism appears to be an Aβ-associated disorder with enhanced non-amyloidogenic processing of APP and abnormal cytoplasmic accumulation and trafficking of N-terminally truncated Aβ, which lead to enhanced oxidative stress and mitochondrial injury, contributing to abnormal neuron development and function.