Our previous studies showed neuronal accumulation of N-tr-Aβ in idiopathic autism and in dup-15 with autism [48] and demonstrated almost complete co-localization of neuronal N-tr-Aβ deposits with markers of oxidative stress [16]. The morphological analysis of N-tr-Aβ deposits and oxidatively modified lipids suggested that N-tr-Aβ accumulation initiates oxidative stress. Hence, according to our hypothesis, intraneuronal deposition of N-tr-Aβ in autism in childhood causes enhanced formation of oxygen free radicals and lipid peroxidation products, which leads to the further formation of Aβ in a self-enhancing vicious circle contributing to neuron dysfunction in autism [16]. Now we demonstrate that in idiopathic and dup-15 autism this pathological process targets a specific subpopulation of GABAergic neurons, those expressing PVA. We also suggest that accumulation of N-tr-Aβ in GABAergic cells and synapses is a significant contributor to dysfunction of the GABAergic system that has been reported in autism [7, 13].
Detection of Aβ by the immunohistochemical method was based here on the use of the monoclonal antibodies 4G8 and 6E10, which may also bind full-length APP and its fragments, as reported previously [51]. However, in formalin-fixed and PEG–embedded human brains, the reaction is limited to Aβ [14,15,16, 48], as demonstrated also here in the double immunostainings (Fig. 1), which combine mAb 4G8 or 6E10 with APP-specific antibody R57.
The antibodies specific for the Aβ40 and Aβ42 species used for Aβ detection by immunoblotting do not react with APP or APP fragments after cleavage by α- or β-secretases, as they recognize the epitopes only when exposed as C-terminus. The usefulness of these antibodies in immunohistochemical staining in fixed sections, however, is limited, because of a lower sensitivity than mAb 4G8 — probably due to organization of Aβ species in the brain into complexes, oligomers, and aggregates (Fig. 2). Aβ peptide oligomerization reduces accessibility of the epitopes for the C-terminal–specific antibodies while preserving the reactivity with mAb 4G8 [19]. Polymorphism of Aβ aggregation states, dependent on peptide species and oligomerization/aggregation conditions, is known to affect the reactivity with numerous antibodies, including mAbs 4G8 and 6E10 [19]. It remains to be established if any particular Aβ oligomerization pattern is associated with specific neuron and synapse subpopulations in autism. The presence of lipofuscin, which is abundant in some neurons in autism [31], may generate problems with non-specific antibody binding. Thus, for detection of specific reactions essential are: optimization of the staining protocol and digital image collecting, and verification of fluorescence co-localizations in all three channels. The findings that mAb 4G8 and the Aβ-42– specific antibody R226 label only a fraction of autofluorescent granules, none of which are immunostained with mAb 6E10 (Suppl Fig. 1), indicate that our immunostaining protocol successfully prevents non-specific antibody binding in brain sections. The method of measurements of the N-tr-Aβ load that we applied here — immunohistochemistry, followed by confocal microscopy digital imaging and Image J analysis — has been shown to allow a reliable protein quantification in the model of cytochrome C aliquots embedded in gelatin [5].
Accumulation of N-tr-Aβ in GABAergic neurons
The loads of N-tr-Aβ in GABAergic neurons and in GABAergic synapses in prefrontal cortex in idiopathic and dup-15 autism significantly exceed those found in controls. The mechanisms responsible for the observed accumulation of N-tr-Aβ may include altered processing of APP [2], as well as a decreased peptide clearance that involves transport through the perivascular drainage system and local enzymatic degradation, particularly by IDE, endothelin-converting enzymes (ECE)-1 and ECE-2, and neprilysin [37]. The two latter enzymes are mainly expressed in GABAergic neurons: ECE-2 primarily in SST-expressing neurons and synaptosomes, and neprilysin — mostly in synapses of the PVA-expressing interneurons. Hence, synapses of GABAergic neurons were suggested to be the sites of Aβ degradation [37]. Accumulation of N-tr-Aβ mainly in the PVA+ but not the SST+ subpopulation suggests that dysregulation of neprilysin expression in the former subpopulation might be a part of the pathomechanism of the observed accumulation of N-tr-Aβ in autism. Neprilysin is an important protective factor for neurons and neuronal progenitor cells against the damaging effects of Aβ [36].
Several pathophysiological consequences can emerge from the accumulation of Aβ in neurons, which have mainly been studied in the Alzheimer’s disease context. It should be noted that there is little knowledge about the effects of N-tr-Aβ and particularly Aβ-pE11, and that peptides’ truncation and N-terminal modification may significantly alter their biological effects. Soluble Aβ oligomers, even in low nanomolar concentrations, increase neuronal excitability by disrupting glutamatergic/GABAergic balance, thereby impairing synaptic plasticity [30]. Aβ injected into the hippocampus depresses the functional activity of the GABAergic neurons responsible for the propagation of the theta rhythm without causing any actual cell damage [45]. Intraneuronal accumulation of Aβ peptides leads to a deep learning deficit detected in animal models, the mechanism of which is associated with a reduced nuclear translocation of the CREB co-activator, CRTC1, and decreased expression of the CRTC1-dependent genes associated with synaptic plasticity: Arc, c-fos, Egr1, and Bdnf [50].
Our finding of lower GAD67 in neurons that contain a high load of N-tr-Aβ suggests a reduced production of the GABA neuromediator and may signal a dysfunction of this fraction of GABAergic cells. Deficiency of GAD67 levels in PVA interneurons results in increased excitability of pyramidal cells and cortical dysfunction [29]. Reduction of the levels of GAD67 protein leading to a selective dysfunction of GABAergic interneurons can be induced by excessive stress during early development, as detected in a rat model of chronic unpredictable stress [3], and prenatal exposure to maternal stress specifically depresses precursors of PVA+ GABAergic interneurons [44]. These changes may be substantial in the pathophysiology of various stress-related disorders, including autism — known to be associated with prenatal stress and maternal immune dysregulation (reviewed: [6, 28]). We hypothesize that oxidative stress initiated by accumulation of N-tr-Aβ in neurons [16] may be responsible for most, if not all, of the above pathomechanisms.
Nuclear N-tr-Aβ
In this study, the N-tr-Aβ–immunoreactive granules were also detected in the nucleus. In idiopathic and dup-15 autism, the nuclei of neurons contained between 14 and 20% of the total neuronal load of N-tr-Aβ. Full-length Aβ has been detected previously in the nucleus by biochemical methods, confocal microscopy, and electron microscopy in cultured neuroblastoma cells that internalized Aβ [4]. Aβ1–42 appears to have a role in nuclear signaling that is distinct from that of C-terminal APP, by specifically interacting as a repressor of gene transcription with LRP1 and KAI1 promoters. The nuclear translocation of Aβ1–42 impacts the regulation of genes, of which the most studied are the genes important in the context Alzheimer disease pathogenesis [4]. Aβ accumulation in neurons may repress the expression of multiple genes linked to synaptic plasticity, e.g., Arc, Nur77, and Zif268, in mouse models [11]. Zif268 in turn may regulate expression of GAD67, as the GAD67 promoter region contains a Zif268–binding site. Thus, accumulation of Aβ may affect genes’ expression also indirectly, e.g., through regulation of the gene Zif268, the equivalent of which in humans is the EGR1 gene. Deficient EGR1 mRNA expression was detected in schizophrenia and was correlated with significantly lower levels of GAD67 [27]. Aβ1–42 in the nucleus of cortical neurons may also affect gene expression through a newly discovered mechanism: by affecting expression of miRNAs, the regulatory short RNA molecules [12]. It should be stressed that little is known about the effects on nuclear functions of N-tr-Aβ and pyroglutamate modified at glutamate-11; yet their nuclear presence in autism suggests they may also act as regulators of transcription in some neurons and possibly also in glia.
Functional consequences of N-tr-Aβ in the PVA+ subpopulation of GABAergic neurons
The deposits of N-tr-Aβ and pyroglutamate–modified Aβ-pE11 were found primarily in the PVA+ subpopulation of GABAergic neurons. Inhibitory synapses of the PVA+ and SST+ GABAergic neurons are regulated by excitatory neurons through different postsynaptic proteins — either the L-type or R-type calcium channels, respectively [21], are regulated through distinct acetycholine receptor modulators [10] and have distinct effects on spatial working memory [24]. The fast-spiking parvalbumin interneurons in the medial prefrontal cortex appear to be involved in coordination of the activity in the local network during goal-driven attention processing [25]. Dysfunctions of the PVA+ GABAergic interneurons in the prefrontal cortex have been linked to cognitive deficits in schizophrenia [35] and other psychiatric disorders [10]. A significantly reduced density of the PVA+ neurons, but not the interneurons expressing calbindin or calretinin, was reported in the prefrontal cortex in autism as compared to control subjects [18].
We found a significantly higher accumulation of N-tr-Aβ in the PVA+ neurons; yet there was a substantial variability in the peptide load in individual cells in this subpopulation. The PVA+ neurons in the prefrontal and frontal cortex represent a diverse population that consists of basket and chandelier cells that in layer 3 form a circuitry with pyramidal cells. Thus, the variability we observed may represent either distinct functional subpopulations of PVA+ cells, or distinct stages of N-tr-Aβ accumulation.
Several differences in the accumulation of N-tr-Aβ and its pyroglutamate-modified form have been detected here between idiopathic autism and dup-15 with autism. These differences may result from the fact that human chromosome 15q11–13 contains a cluster of three GABAA receptor subunit (GABR) genes, GABRB3, GABRA5, and GABRG3. Deletion or duplication of 15q11–13 GABR genes occurs in multiple human neurodevelopmental disorders, including Prader-Willi syndrome, Angelman syndrome, and autism. In humans, all three GABR genes are biallelically expressed, i.e., are not imprinted in normal human cortex. However, in autism, expression of one or more GABR genes is frequently monoallelic or strongly skewed allelic, indicating that epigenetic dysregulation of these genes without cytogenetic modifications may be relatively common in autism [20].
N-tr-Aβ in GABAergic synapses
The presence of N-tr-Aβ deposits in the GABAergic synapses — according to our study, in as many as 7% or more in autism, both idiopathic and dup-15 — may be a marker of dysfunction of the GABAergic system in autism. Aβ in soluble and aggregated forms has already been postulated as being responsible for synapse dysfunction. In cultured neurons, endogenous Aβ42 binds to a subset of synapses — more to glutamatergic than to GABAergic ones [49], and aggregated Aβ may damage axon terminals, even though the GABAergic neurons appear to be less vulnerable to Aβ toxic effects than cholinergic and glutamatergic ones [8]. It should be noted, however, that the toxic effects of low, even picomolar, doses of Aβ oligomers on neurons can be greatly enhanced by inflammatory response to infections during critical stages of embryonic development and early postnatal life, when activated microglia cause synapse damage and cognitive impairment [17]. This modification of microglia function may be significant in the context of autism pathogenesis, in which prenatal and early postnatal infections have been postulated as triggering factors for development of autism [28, 42].
Processing of APP yields several products of distinct, yet only partially known functions. Alterations of APP processing in autism result in higher levels of not only N-tr-Aβ but also secreted APP-α [2, 38, 43]. The latter product in the brain may further affect the GABAergic regulations by suppressing presynaptic vesicle release through direct binding of sAPP extension domain to the GABA type B receptor subunit 1a [39]. This may be another APP-related mechanism of GABAergic dysregulation in autism.