Background
Autism is a lifelong neurodevelopmental disorder that is diagnosed in early childhood and is characterized by a core deficit in social interaction with impairments in communication, stereotypical movements and restricted behaviors [
1]. Converging evidence over the past 40 years indicates that immune dysfunction may be an important factor contributing to the development of a subset of cases of autism [
2‐
4]. Several studies have shown peripheral immune abnormalities in patients with autism [
5‐
7]. There is provocative evidence for an ongoing inflammatory response in some individuals with autism [
8,
9].
The possibility has been raised that some forms of autism may be due to an autoimmune process [
10]. There are a number of precedents for autoimmune diseases of the central nervous system. The best known among these are multiple sclerosis [
11,
12] and Sydenham's chorea [
13,
14]. The potential for an autoimmune etiology with respect to psychiatric disorders, including the pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections, remains intriguing but controversial [
14].
A number of reports have identified antibodies in individuals with autism that are directed against several central nervous system proteins. These include glial and neuron-axon filament proteins [
15] (but see [
16]), myelin basic protein [
17] (but see [
18]), serotonin receptor [
19], nerve growth factor [
20], cerebellar peptides [
21], brain-derived neurotrophic factor [
22], brain endothelial cells [
23] and the caudate nucleus [
24]. We have described an increased incidence of autoantibodies to brain proteins in autism compared with controls [
25].
In a previous study [
26], we used Western blot analysis and tissue immunohistochemistry to investigate the presence of autoantibodies to cerebellar tissue. Using Western blot analysis, we found that the plasma of 21% of individuals with autism demonstrated immunoreactivity to a protein of approximately 52 kDa from the human cerebellum that was present in only 2% of typically developing controls. When these plasma samples were then used as primary antibody sources for tissue immunohistochemistry using sections from the macaque monkey brain, 21% of the samples from children with autism spectrum disorder (ASD) compared to 0% of typically developing controls demonstrated intense immunoreactivity to what were morphologically determined to be Golgi cells of the cerebellum. This study utilized sections from the cerebellum based on early data that the cerebellum was preferentially involved in autism. However, it is now clear that autism affects many brain regions [
27], and the question arises whether the autoantibodies present in autistic individuals identify a broader class of neurons that are distributed throughout the brain. Thus, to build upon our previous findings, in the current study we used plasma from the same individuals who demonstrated positive cerebellar Golgi cell immunoreactivity to stain tissue sections from the full rostrocaudal extent of the macaque monkey brain. To evaluate conservation of the target antigens, we also examined brain sections from adult male mice. We were particularly interested in determining whether the autoantibody immunoreactivity would identify neurons from disparate brain regions and whether the identified neurons would have any unifying morphological features.
Methods
Participants and sample collection
The study protocol followed the ethical guidelines of the most recent Declaration of Helsinki [
28] and was approved by the Institutional Review Board of the University of California-Davis School of Medicine. All participants enrolled in the study had written informed consent provided by their parents and provided consent to participate if developmentally able. Subjects for this study were enrolled through the Medical Investigations of Neurodevelopmental Disorders (M.I.N.D.) Institute clinic. The current study was designed to be a detailed follow-up of our previous study that described the presence of Golgi cell immunoreactivity in the plasma of a subpopulation of children with autism [
26]. Our current study population consisted of a subgroup of children previously diagnosed on the autism spectrum (
n = 7) (age range, 2.5 to 7 years) who showed reactivity to an approximately 52-kDa protein in the cerebellum as determined by Western blot analysis as well as on the basis of intense immunoreactivity to Golgi cells of the cerebellum in the cynomolgus monkey (
Macaca fascicularis). We also examined seven participants from our previous study who were diagnosed on the autism spectrum but did not show immunoreactivity to monkey brain sections. Our control population consisted of typically developing children with an absence of reactivity to both the 52-kDa protein and to neurons of the cynomolgus monkey brain (
n = 6) (age range, 2.5 to 8 years). Any participants in the control group who were above criteria in either the Autism Diagnostic Interview-Revised (ADI-R) or the Autism Diagnostic Observation Schedule (ADOS) [
1,
29‐
31] were excluded. Samples from children with ASD with an absence of reactivity detected by both Western blot analysis and immunohistochemical staining were also examined. A diagnosis of ASD was confirmed in all participants on the basis of the ADI-R and ADOS score. The final ASD status is defined as meeting criteria on the communication, social interaction and repetitive behavior domains of the ADI-R with onset prior to 36 months of age and scoring at or above the social plus communication cutoff for autism on the ADOS module 1 or 2. The Social Communication Questionnaire [
32] was used to screen for characteristics of ASD among the typically developing controls. Children who scored above the screening cutoff were fully assessed using the ADI-R and ADOS.
Animals and fixation
Nonhuman primate tissue
We choose nonhuman primate tissue as our antigen source because of the highly conserved nature of autoantigens as well as the similar but compressed neurodevelopment of rhesus monkeys compared with humans [
33]. All procedures were carried out under an approved University of California-Davis Institutional Animal Care and Use Protocol and strictly adhered to National Institutes of Health policies on primate animal subjects. The brain sections were obtained from two different male adult cynomolgus macaques (
Macaca fascicularis): one was 9 years old and the other was 15 years, 4 months old at the time they were killed. The brains from these animals had been used for neural tract tracing studies, and library sections were used for the present study. No animals were killed expressly for these studies. The sections were obtained from animals for which no health problems were reported.
Tissue fixation and histological processing were performed according to procedures described previously [
34]. Briefly, the animals were deeply anesthetized and perfused intracardially with 1% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) at 4°C at a rate of 250 mL/minute for 2 minutes, followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) at 4°C at a rate of 250 mL/minute for 10 minutes. The flow rate was then reduced to 100 mL/minute for 50 minutes. The animal was then perfused with 5% sucrose in 0.1 M phosphate buffer (pH 7.4) to remove as much fixative as possible.
The brain was blocked stereotaxically, extracted from the skull and cryoprotected in a solution containing 10% glycerol and 2% dimethyl sulfoxide (DMSO) for 1 day, followed by 20% glycerol and 2% DMSO for 3 days. The fixed brain was then frozen using the isopentane method [
35] and stored at -70°C until sectioning. Frozen sections were cut in the coronal plane at a thickness of 30 μm with a sliding microtome and placed into a cryoprotectant tissue-collecting solution (30% ethylene glycol, 25% glycerin in 0.05 M sodium phosphate buffer). The sections were stored at -80°C until they were used for immunohistochemistry.
Additionally, tissue from one adult rhesus macaque (Macaca mulatta) whose perfusion included glutaraldehyde was used for anti-γ-aminobutyric acid (anti-GABA) immunostaining. Briefly, the perfusion consisted of 0.9% sodium chloride at a rate of 250 mL/minute for 2 minutes, followed by 4% paraformaldehyde/0.1% glutaraldehyde at a rate of 250 mL/minute and finally 4% paraformaldehyde/0.1% glutaraldehyde at a rate of 100 mL/minute for 50 minutes. Tissue was blocked and sectioned in a manner similar to that described for the cynomolgus macaques.
Mouse tissue
Two adult male C57BL/6 mice were perfused according to the methods described above for the cynomolgus macaque tissue, with an initial flow rate of 6 mL/minute followed by a flow rate of 4.75 mL/minute to account for the animal's smaller size. Tissue was cryoprotected in 10% glycerol and 2% DMSO for 1 day, followed by 20% glycerol and 2% DMSO for 2 days, and was sectioned on a sliding microtome at 30 μm as previously described for the monkey tissue.
Immunohistochemistry
Free-floating macaque brain sections were processed using methods employed and described previously [
34,
36,
37]. Serial sections were used for immunohistochemistry using samples from each subject individually. Briefly, 30-μm-thick coronal sections from the brain were taken from storage, placed into nets and then washed (3 × 10 minutes with 0.1 M phosphate-buffered saline (PBS)). The following steps were followed with agitation at room temperature, except for the primary incubation, which was performed at 4°C with agitation. Prior to primary incubation, the sections were pretreated with hydrogen peroxide (1.6% hydrogen peroxide in 0.1 M PBS (Fisher Scientific; Pittsburg, PA, USA) for 15 minutes to remove endogenous peroxidases and reduce background staining. During incubations, the tissue was maintained in solution at all times. The sections were blocked for 4 hours using 5% normal mouse serum (Pierce Biotechnology; Rockford, IL, USA) in 0.1 M PBS and 0.5% Triton X-100 (Fisher Scientific; Pittsburg, PA, USA). The nets were incubated with the primary antibody (subject plasma) at a dilution of 1:100 in 0.1 M PBS, 2% normal mouse serum and 0.3% Triton X-100 for 40 to 48 hours. The sections were then washed with 2% normal goat serum (Invitrogen; Carlsbad, CA, USA) in 0.1 M PBS (3 × 10 minutes) and incubated with biotinylated mouse anti-human monoclonal immunoglobulin G (IgG) (Zymed: Carlsbad, CA, USA) in 0.1 M PBS, 2% normal mouse serum and 0.3% Triton X-100 for 1 hour. Sections were then washed with 2% normal goat serum in 0.1 M PBS (3 × 10 minutes) followed by incubation for 45 minutes with avidin-biotin-complex (ABC) peroxidase (Biostain Super ABC/Peroxidase Basic Kit; Biomeda: Foster City, CA, USA) incubation in 0.1 M PBS. Following the first ABC peroxidase incubation, the sections were washed in 2% normal goat serum in 0.1 M PBS. The incubation with the secondary antibody (biotinylated mouse anti-human monoclonal antibody) was repeated for 45 minutes, followed by washes with 0.1 M PBS only. The second ABC peroxidase incubation was for 30 minutes, followed by washes with 0.05 M Tris buffer. Next, the sections were incubated with 3,3-diaminobenzidine (DAB) (Fisher Scientific) in 0.05 M Tris with 0.04% hydrogen peroxide for 30 minutes. The sections were then washed twice with 0.05 M Tris, once with 0.02 M potassium PBS (KPBS) and stored covered at 4°C in 0.02 M KPBS. Sections were then mounted from phosphate buffer onto gelatin-coated slides and dried with a fan, then placed in a 37°C incubator for 24 hours to dry. Mouse sections were processed in an identical manner, except for the substitution of 0.5% for 1.6% hydrogen peroxide and the additional use of an avidin and biotin blocking kit (Vector Laboratories; Burlingame, CA, USA) prior to incubation in the primary antibody solution.
Silver nitrate/gold chloride intensification
To increase the signal-to-noise ratio of the immunohistochemical reaction product, a silver nitrate/gold chloride intensification procedure was followed for immunohistochemical analysis for those sections stained with DAB. Intensification was not used for sections visualized by using fluorescence immunohistochemistry. Prior to intensification, the slides were defatted in a mixture of equal parts chloroform and 100% ethanol for a total of 4 hours. The sections were then hydrated through graded ethanols and placed in a 37°C incubator overnight. Following the overnight incubation, the slides were rinsed with running, deionized water for 10 minutes, followed by incubation in a 1% silver nitrate solution maintained at 56°C with the use of a water bath for 40 minutes. The slides were then rinsed with running, deionized water for 10 minutes and incubated for 10 minutes in a 0.2% gold chloride solution with agitation at room temperature. Following a second rinse, the slides were stabilized in a 5% sodium thiosulfate solution at room temperature for 15 minutes with agitation, followed by an additional rinse in running, deionized water for 10 minutes. Sections were then dehydrated through graded ethanols, followed by xylenes, and coverslipped using DPX mounting medium (Sigma Aldrich; St. Louis, MO, USA).
Slides were observed by bright-field microscopy on a Nikon Eclipse E600 (Nikon Americas; Melville, NY, USA) or a Leica Leitz DMRB microscope (Leica Microsystems; Buffalo Grove, IL, USA) and photographed using the Spot Diagnostic Instruments Digital Camera System and software (Spot Imaging Solutions; Sterling Heights, MO, USA). Preliminary identification of the various cell types was accomplished using morphological features such as their location, cell body size and dendritic configuration.
Fluorescence immunohistochemistry
Since it appeared that the majority of neurons that demonstrated specific immunoreactivity when incubated with plasma from individuals with autism had the morphological characteristics of GABAergic neurons, we carried out analyses designed to confirm this finding. First, we compared the distribution of the plasma-stained neurons with libraries of sections stained immunohistochemically for the distribution of glutamic acid decarboxylase 65 (GAD65) and GAD67, GABA or
in situ hybridization with probes to GABAergic neurons [
34,
37]. Second, we carried out studies in which sections were double-labeled for the presence of plasma-identified interneurons and GABA immunoreactivity. Finally, we double-labeled additional sections for plasma-labeled interneurons and the calcium-binding proteins calbindin, parvalbumin and calretinin.
For double-labeling with plasma and anti-GABA, sections from the glutaraldehyde-fixed macaque brain were rinsed in PBS, then blocked in 0.2% gelatin, 1% Triton X-100 and 5% normal donkey serum in 0.1 M PBS for 2 hours, and incubated with plasma from an individual with autism as well as monoclonal mouse anti-GABA diluted 1:400 (a gift from Ismo Virtanen) at 4°C for 72 hours. Tissue was then rinsed and incubated in Texas Red-AffiniPure donkey anti-human IgG (1:100; Jackson ImmunoResearch; West Grove, PA, USA) and fluorescein isothiocyanate-conjugated horse anti-mouse IgG (1:500; Vector Laboratories) for 2 hours at 4°C. For double-labeling with plasma and antibodies directed to calcium-binding proteins (calbindin, parvalbumin or calretinin), a Tyramide Signal Amplification (TSA) kit (PerkinElmer; Waltham, MA, USA) was used to enhance the fluorescent signal obtained from labeling with plasma. Sections were first rinsed in 0.1 M PBS, pretreated with 1% hydrogen peroxide in PBS for 30 minutes, rinsed in 0.1 M PBS and then blocked with 0.5% Triton X-100 and 5% normal mouse serum in 0.1 M PBS for 4 hours. They were then incubated overnight with a 1:100 dilution of human plasma in 0.1% Triton X-100 and 1% normal mouse serum in PBS. Sections were subsequently rinsed in PBS and incubated in Tris-NaCl-Tween buffer (TNT) buffer for 15 minutes, followed by a 30-minute incubation in TNB buffer (Blocking buffer supplied with TSA kit; 0.1 M TRIS-HCl, pH 7.5. 0.15 M NaCl. 0.5% Blocking Reagent), and incubated in a 1:1,000 dilution of biotinylated anti-human IgG secondary (Zymed; Carlsbad, CA, USA) for 1 hour. The secondary antibody was washed away with TNT buffer, and sections were then incubated in streptavidin-horseradish peroxidase (included in the TSA kit) diluted 1:100 in TNB buffer for 30 minutes. After being washed with TNT buffer, sections were incubated in tetramethylrhodamine tyramide (PerkinElmer) diluted 1:50 in amplification diluent for 8 minutes at room temperature, washed in 0.1 M PBS and incubated overnight in a solution containing an antibody against calbindin (Swant; CH-1723 Marly 1, Switzerland), parvalbumin (Swant) or calretinin (Swant) diluted 1:1,000 in 0.02% gelatin, 0.1% Triton X-100 and 1% donkey serum in 0.1 M PBS. The next day sections were rinsed in 0.1 M PBS, incubated in secondary antibody (donkey anti-rabbit) diluted 1:200 in 0.1% Triton X-100 and 1% donkey serum in PBS for 2 hours, rinsed in PBS again and mounted onto slides and coverslipped with VECTASHIELD HardSet Mounting Medium (Vector Laboratories). Fluorescently labeled sections were examined and photographed using an Olympus FluoView confocal laser-scanning microscope, and analysis was performed using FluoView version 3.3 software (Olympus: Hamburg, Germany). For identification of labeled cells, we used the same criteria described for sections processed with diaminobenzidine.
Discussion
We previously reported that plasma from children with autism contains autoantibodies that are reactive to cerebellar Golgi cells [
26]. The current study expands our earlier observations by exploring the pattern of immunohistochemistry in additional brain regions. We have demonstrated that the plasma from children with ASD contains antibodies that are reactive to neurons throughout the brain that are invariably GABAergic interneurons. We have also determined that not all GABAergic neurons are immunopositive, although we have not yet been able to determine in detail which categories of GABAergic neurons are positive for the plasma staining.
GABAergic cells form a heterogeneous population of neurons that are unevenly distributed in all layers of the monkey neocortex [
41]. At least 20% of the neurons in the primary visual cortex are GABAergic interneurons [
42], and the distribution in the monkey brain is quite similar to that in the human brain [
43]. The plasma from children with autism allowed us to identify populations of cells in all layers of the cerebral cortex, with a slight preponderance in superficial layers. Colocalization studies using a monoclonal antibody to GABA revealed that the plasma from children with ASD appeared to recognize GABAergic interneurons in V1. There were fewer plasma-labeled cells in the deep layers of cortex relative to GABA staining. Double-labeling with plasma and antibodies toward the calcium-binding proteins calbindin, parvalbumin and calretinin revealed that plasma-labeled cells commonly expressed calbindin. The autoantibody-positive cells expressed parvalbumin less commonly and calretinin only rarely. The distribution of GABAergic interneurons that colocalize one of these calcium-binding proteins is variable from one cortical area to the next, and certain interneuron types can colocalize more than one of the proteins [
44‐
47]. Defelipe [
48] provided a detailed description of the correspondence of specific interneuron types and immunoreactivity for a calcium-binding protein in the neocortex. In general, parvalbumin is associated with chandelier and large basket cells; calretinin is associated with bipolar, double-bouquet and Cajal-Retzius cells; and calbindin is associated with double-bouquet, neurogliaform, Martinotti and, in rare cases, Cajal-Retzius cells. This implies that the plasma autoantibodies affect neurons harboring calbindin more than those colocalizing the other calcium-binding proteins. However, it is not possible at this time to predict the specific functional significance of targeting these specific classes of interneurons. The question why some GABAergic interneurons are identified by the plasma autoantibodies and others is not very difficult to speculate on, given that the antigen is not known. However, recent neuronal transcriptome studies [
49,
50] have made it abundantly clear that tens to hundreds of proteins differentiate neuronal classes. Given the distinct morphological features of interneurons, it is highly likely that different classes of GABAergic neurons express different profiles of proteins and that one or more of these are being identified by the plasma autoantibodies.
Our finding that some children with autism produce autoantibodies directed against GABAergic neurons is all the more interesting, given the other indications that ASDs may be highly related to impaired GABAergic function. Previous studies have suggested that there is a depression of the GABAergic system, potentially due to altered GABA receptors and reduced GAD, in the etiology and maintenance of ASD [
51‐
53]. It has also been proposed that suppression of GABAergic function in the brain of individuals with ASD may result in compensatory mechanisms and increased activation of additional GABAergic receptor subtypes [
54]. The potential imbalance of excitatory and inhibitory neurotransmission in autism could influence cortical networks that control social behaviors and neuromodulatory systems [
55]. Interestingly, a reduction in GABAergic activity is also capable of causing epilepsy, which has been reported in up to one-third of individuals with ASD [
52,
56‐
60]. While the participants in the present study were not assessed for epileptic seizure activity, this would be of great interest for future longitudinal studies.
In a study by Fatemi and colleagues [
61], GAD, which converts glutamate to GABA and is a marker commonly used to identify GABAergic neurons, was found to be reduced by 48% to 61% in the parietal cortex and cerebellum in five individuals with autism relative to eight controls. More recent studies by Fatemi and colleagues [
62] have found downregulated expression of various combinations of GABA
A receptor subunits in the parietal cortex, cerebellum and superior frontal cortex in the brains of individuals with autism. These investigators also reported alterations of GABA
B receptor subunit expression in the same areas of the brain [
63]. Yip and colleagues [
64,
65] reported increased levels of GAD67 mRNA levels in cerebellar interneurons but decreased levels in Purkinje cells. Blatt
et al.[
66] examined the distribution and density of GABAergic, serotonergic, cholinergic and glutamatergic receptors in the hippocampus of individuals with ASD and typically developing controls. The only receptor system found to be significantly reduced in individuals with ASD was the GABAergic system [
66]. Decreased GABA levels have also been reported in the platelets of children with ASD [
67].
While it is intriguing that some children with autism demonstrate autoantibodies that are immunoreactive to central nervous system GABAergic neurons, it is not at all clear whether these antibodies play a pathophysiological role in the etiology of autism or are an epiphenomenon indicative of some other pathological process. In order for the systemic antibodies to interact with central nervous system neurons, they must be able to reach their antigenic targets. As antibodies do not readily cross the blood-brain barrier, a breach would have to occur to allow access. Cytokines in the peripheral circulation or cytokines produced by central nervous system cells are capable of altering the blood-brain barrier through endothelial cell activation [
68]. Disturbances to the blood brain-barrier can be caused by a number of other factors, including extreme stress, subclinical infection and even nicotine or epinephrine exposure [
69‐
71]. In studies by Diamond and colleagues [
72,
73] in which an animal model of systemic lupus erythematosus (SLE) was explored, antibodies directed against the
N-methyl-D-aspartate receptor were able to enter the brain following exposure to either lipopolysaccharide or norepinephrine. These antibodies not only produced neuropathological changes in the hippocampus or amygdala, respectively, but also led to impaired memory function or increased anxiety.
If we presume that the anti-GABAergic neuron antibodies produced by children with autism can gain entry to the brain, there are several possible mechanisms through which they may influence cell function. First, autoantibodies may act as an agonist or a receptor ligand, causing excessive receptor stimulation. This phenomenon has been demonstrated in approximately 27% of patients with epilepsy and 30% of individuals with SLE, as well as in some patients with encephalitis, in whom various autoantibodies to the glutamate receptor have been described [
74]. Further, both human and animal studies have demonstrated that the described glutamate receptor autoantibodies are pathologic, binding to neurons with a unique ability to activate their glutamate receptor antigen and leading to neuronal death, either through excitotoxicity or by complement fixation independent of receptor activation [
74‐
76]. While a decreased number of neurons has been reported in the amygdala [
77] and in the fusiform gyrus [
78] of postmortem cases of autism, it is not yet clear whether there is a preferential loss of any particular category of neurons, such as GABAergic interneurons.
The second possibility for autoantibody-induced pathology is through antibody-mediated inhibition of a critical receptor, ligand or pathway. This would be similar to the pathologic mechanism through which the autoantibodies to thyroglobulin, thyroid peroxidase and/or the thyroid-stimulating hormone receptor function in Hashimoto's thyroiditis [
79]. Whether through excitation or inhibition, autoantibody-mediated changes are capable of resulting in a number of permanent changes, including inhibition-excitation imbalance, altered neural circuitry and changes in receptor numbers and distribution. Finally, it is possible that the autoantibody directly induces cell and tissue destruction through complement activation such as that seen in patients with SLE [
80]. Each of the above-discussed mechanisms, either individually or in concert, is a possibility with respect to the autoantibodies described herein. However, additional studies, including an animal model, are needed to determine the pathologic relationship between these autoantibodies and ASD.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
JV, along with DGA, conceived of the study, participated in its design and coordination, and helped to draft the manuscript. SW carried out the investigation of plasma reactivity in the nonhuman primate tissue and helped to draft the manuscript. JB prepared tissues and helped with analysis of staining. CCR carried out the investigation of plasma reactivity in mouse tissue, processed and analyzed tissue double-labeled with plasma and anti-GABA, and participated in double-labeling with antibodies directed against calcium-binding proteins. VMC participated in colabeling experiments and manuscript preparation. PA helped with study design and preparation of the manuscript. All authors read and approved the final manuscript.