|Home | About | Journals | Submit | Contact Us | Français|
Autism is a pervasive developmental disorder characterized by repetitive stereotyped behavior, social-emotional deficits and delayed or absent language abilities. There are known neuropathologies in the autism brain affecting limbic, cerebellar and cortical structures but the neurochemical profile of affected individuals, revealed in postmortem tissue studies, is only recently emerging. One major component that appears highly impacted in autism is the GABAergic system. It is now apparent that there are widespread significant effects in many distributed regions in the autism brain revealed by histochemical, autoradiographic and biochemical studies. The key synthesizing enzymes for GABA, glutamic acid decarboxylase type 65 and 67 (GAD65, GAD67), are decreased in the cerebellum and closer examination of mRNA levels revealed that it is largely due to decreases in Purkinje cells and a subpopulation of larger dentate neurons as measured by in situ hybridization studies. Other cell types had either normal GAD levels (Golgi cells, smaller dentate interneurons and stellate cells) or increased levels (basket cells). GABA receptor density, number and protein expression are all decreased in the cerebellum and in select cortical areas. GABAA and GABAB subunit protein expression was significantly reduced in cerebellum, BA 9 and BA 40. Benzodiazepine binding sites were significantly reduced in the hippocampus and anterior cingulate cortex (BA 24). Taken together, data from these studies suggest that there is a marked dysregulation of the inhibitory GABA system in the autism brain affecting particular biomarkers localized to specific cell types and lamina likely influencing circuitry and behavior.
There is increasing neuropathological and neurochemical evidence that the GABAergic system has major involvement in autism. From post-mortem tissue analysis, it has been long known that there is a decrease in the number of GABAergic cerebellar Purkinje cells in many individuals with autism (Williams et al., 1980; Bauman and Kemper, 1985; Ritvo et al., 1986; Bailey et al., 1998; Kemper and Bauman, 1998; Whitney et al., 2008). In the last decade, biochemical studies have revealed that there are highly significantly decreased levels of key synthesizing enzymes for GABA, glutamic acid decarboxylase type 65 and 67 (GAD65 and GAD67 isoforms) in the cerebellum and parietal cortex in autism (Fatemi et al., 2002) further suggesting that there are effects on the GABA system in the cerebellum and elsewhere in the autistic brain. To histologically localize these changes in the cerebellum, Yip and colleagues performed a series of in situ hybridization histochemical studies that radiolabels the GAD65 or GAD67 mRNA inside specific neuronal types.
In the first study, Yip et al. (2007) demonstrated a 40% reduction in GAD67 mRNA in Purkinje cells (PCs) in 8 adult autistic cases relative to 8 age-, gender, and post-mortem interval matched controls in the posterior-lateral cerebellar hemisphere in the Crus II region (blocks taken inferior to the horizontal fissue in the posterior lobe). This difference in GAD67 mRNA levels was present in remaining PCs from autism cases with decreased numbers of PCs as well as in autism cases that had the normal complement of PCs. Thus despite the neuropathological studies identifying subsets of autism cases with decreased PCs, the GAD67 mRNA reduction was found in every autism case examined in the study. Results from this study also suggested that PCs are especially vulnerable in the autistic brain and, that there is the potential for a major disturbance in cerebellar circuitry within the posterior lobe.
Recently, Whitney et al. (2009) performed a stereological count of basket cell (BC) and stellate cell (SC) GABAergic interneurons in the molecular layer of the posterior lateral cerebellar cortex in the same lobular location as the in situ studies were performed. Results from the study demonstrated that there was no significant difference in the relative density of either basket or stellate cells between control and autism cases despite there being a moderate or severe PC loss in 3/6 autism cases in the study. Thus, it didn’t matter whether there was a decrease in PCs or not, the density of these interneurons appeared unaffected.
In marked contrast to lowered GAD67 mRNA levels in Purkinje cells in adult autism cases, in the same cases there was significantly elevated GAD67 mRNA levels in basket cells and there was a trend for an increased levels in GAD67 mRNA in stellate cells compared to age- and postmortem interval-matched controls. These results were surprising since Fatemi et al. (2002) demonstrated an overall 50% decrease in GAD levels in the cerebellum in postmortem autistic tissue samples.
In the second in situ study, Yip et al. (2008) measured GAD67 mRNA levels in two types of cerebellar interneurons, BCs and SCs in the molecular layer of Crus II from the same 16 autism and control cases that were examined in the PC mRNA study. Since both types of interneurons innervate PCs (i.e., BCs send horizontal axons in the lower third of the molecular layer forming axonal “nests” around the soma of each PC whereas SCs innervate PC dendrites in the upper two thirds of the molecular layer), it was essential to determine whether the potential inhibitory modulation from interneurons was also disturbed. In this study, there was an overall highly significant increase of 28% in mean GAD67 mRNA expression in basket cells in the 8 autism compared to 8 control brains (autism 1.03 ± 0.05, controls 0.69 ± 0.05; t=−4.615, df=14, p<0.0001, two tailed, Levene’s test for equality of variance for homogeneity did not show a difference between the variance of the control and autism groups F= 0.061, p=0.808). It is noteworthy that one basket cell innervates up to ten PCs (Palay and Chan-Palay, 1974) and each BC has connectivity to other BCs that appear to be electrotonically coupled (Mann-Metzer and Yarom, 1999), thus they exert powerful modulatory inhibitory influence on PCs. It is not known whether the increased BC GAD67 mRNA translates to increased release of GABA to their PC targets but it is tempting to speculate that it may indeed be one of the influences that causes the decreased GAD67 mRNA levels in PCs resulting in reduced GABA release by PCs to cerebellar nuclear neurons, in this case, the dentate nucleus that is the target of PCs from the Crus II region.
There was a trend towards a small increase in GAD67 mRNA expression levels between autism and control cases in stellate cells which failed to reach statistical significance (mean ± SEM pixels of silver grains/neuron, autistic 0.88 ± 0.06, controls 0.77 ± 0.04; p=0.06, independent t test; Yip et al., 2008). Power analysis revealed about 12 cases per group would be needed to reach significance. It is likely that a repeat study involving a larger “n” might result in similar findings as the BC data. If there is also an increase in GAD67 mRNA levels in SCs, this could further enhance the increased modulatory effect exerted on the remaining PCs and potentially exacerbate the effects on the dentate neurons.
Intrinsic GABAergic Golgi type II neurons provide inhibitory input to granule cells and parallel fibers thus influencing Purkinje cell function. In contrast to the altered GAD67 levels found in Purkinje cells and interneurons, the GAD67 mRNA from Golgi cells in the granular layer from 8 adult autism cases compared to 7 cases from individuals with autism were found to be similar to levels in control cases (Blatt et al., 2009). This is an important finding because it demonstrates that in autism, specific subpopulations of cerebellar neurons are more vulnerable to GABA changes than others, an observation that was further confirmed when analyzing subnuclei within the dentate described in the next section.
A number of distinct dentate subpopulations were identified in the human cases based on cell profile estimates including those at approximately 10μm, 15μm, 20μm and 30μm in diameter (Yip et al., 2009). Of these, two distinct subpopulations contained GAD65 mRNA: 1) the larger-sized (about 20 μm) GABAergic dentate cells previously shown in animal studies to project to the inferior olivary complex (Graybiel et al., 1974; Tolbert, 1978; Saint-Cyr and Courville, 1981; Oertel et al., 1981, Mugnaini and Oertel, 1985; De Zeeuw et al., 1988, 1996, 1998; Nelson and Mugnaini, 1989; Ruigrok and Voogd, 1990; Fredette and Mugnaini, 1991), and 2) the smaller (approximately 10μm) GABAergic dentate neurons that are either exclusively GABAergic or perhaps also contain glycine (Chen and Hillman, 1993; Baurle and Grûsser-Cornehls, 1997) and project to other dentate neurons previously described in the monkey (Chan-Palay, 1977).
In this study of GAD levels in cerebellar neurons, quantitation of GAD65 mRNA labeling demonstrated a 51% reduction in signal in the larger sized dentate subpopulation in the autism group when compared to the control group (175.77 ± 19.60 control, 86.56 ± 17.40 autism; p=0.009; independent t-test; Levene’s test for equality of variance for homogeneity between variance F = 0.464). Surprisingly, however, the mean levels of GAD65 mRNA in the small cell subpopulation were similar between the two groups (155.69 ± 22.69 control; 138.25 ± 21.06 autism; independent t-test; Levene’s test for equality of variance for homogeneity between variance F = 0.031, p = 0.594). A correspondence between GAD65 mRNA and GAD65 protein levels including the dentate nuclei has been previously reported (Kaufman et al., 1991; Escapalez et al., 1993; Martin, 1993). With regard to the larger dentate subpopulation in the present study, the marked decrease in GAD65 levels suggests that there is reduced GABAergic input to the inferior olivary complex (IOC) that could result in excessive stimulation of IOC neurons, especially from the principal olive (PO). Since inferior olivary neurons are electrotonically coupled (e.g., Welsh, 2002) this might represent an altered asynchronous population response with potential effects on the firing of target PCs in the hemisphere. Misalignment of PO cells along the edge of the olivary ribbon in some autism cases (Bauman and Kemper, 1985, Thevarkunnel et al., 2006) may also contribute to abnormal physiological responses. In contrast, the normal GAD65 mRNA levels in the smaller dentate subpopulation infers that the local GABAergic circuitry within the dentate is preserved.
It is unknown with all the changes in the cerebellar GABA system in the autism cases as to whether there are additional consequences in the cerebellar nuclear projections to the red nucleus and the thalamus and consequently to the cerebral cortex. It would represent one or more potential pathway(s) that could influence motor and/or cognitive behaviors in autistic individuals. To complicate matters, it appears that the neurons within the four cerebellar nuclei may be disturbed. In a seminal study, Bauman and Kemper (1985) described a marked palor of Nissl staining in each of the four nuclei. This suggests that the number of projecting axons leaving the cerebellum may also be reduced further exacerbating the disturbed cerebellar circuitry.
The initial single concentration ligand binding studies were carried out in the hippocampus as a pilot study effectively to screen a number of different transmitter receptor types from four major systems: 5-HT, glutamate, GABA and acetylcholine (Blatt et al., 2001). Surprisingly, the results from this study found that most measures in adult autism cases were at normal receptor density levels including 5-HT1aR, 5-HT2R, muscarinic cholinergic type 1 and 2 (M1, M2), High Affinity Choline Uptake Sites (HACU), kainate and NMDAR in all major subfields of the hippocampus. The only two that were not at normal levels were the higher binding hippocampal subfields for 3[H]muscimol-labeled GABAA receptors and 3[H]flunitrazepam labeled benzodiazepine (BZD) binding sites (on the GABAA receptor) that showed a significant decrease in receptor density. A follow up study from this laboratory, further demonstrated that the 3[H]flunitrazepam labeled BZD binding reduction in the autism cases was due to a decrease in the Bmax, or number of receptors and not due to the Kd or binding affinity which was at normal levels (Guptill et al., 2007). A more recent study from our laboratory further demonstrated that there is an increased density of GABAergic hippocampal interneurons labeled with specific types of antibodies to calcium binding proteins. Lawrence et al., (2010) found a selective increase in calbindin-immunopositive interneurons in the dentate gyrus, parvalbumin-immunopositive interneurons in the CA1 and CA3 subfields and calretinin-immunopositive neurons in CA1. This agrees with the qualitative observations of Bauman and Kemper (1985) that the hippocampus gave a flattened appearance with smaller neurons and increased packing density. Taken together, these data infer increased density of neurons and interneurons, reduced neuropil, and decreased GABA receptors in select but key subfields that could collectively contribute to altered hippocampal circuitry, synaptogenesis, and ultimately function.
In the cerebellum, recently completed (article in preparation) studies by Thevarkunnel et al. have found in adult autism cases compared to controls, significant decreases in 3[H]flunitrazepam labeled BZD binding sites in the molecular layer and reduced 3[H]muscimol-labeled GABAA receptors in the PC layer in the posterior lateral cerebellar hemisphere. This cerebellar region receives fronto-pontine projections (via the granular layer) as well as olivocerebellar climbing fibers from the principal olive. These studies show that the GABAA receptor changes appear not to be limited to the hippocampus and, that there is another change in the GABAergic synaptic regulation in the cerebellum of the autism brain.
The next series of ligand binding studies was in the cerebral cortex and found a reduction of 46.8% in the mean density of 3[H]muscimol-labeled GABAA receptors in the superficial (supragranular layers I-III) and a reduction of 20.2% in the mean density in the deep (infragranular layers V-VI) of anterior cingulate cortex (ACC) in adult autism subjects compared to matched controls (Oblak et al., 2009). A similar significant decrease in the supragranular layers (28.9%) and in the infragranular layers (16.4%) for 3[H]flunitrazepam labeled BZD binding sites in the same cases (Oblak et al., 2009; Figure 1) was also demonstrated. These results demonstrate that the GABA changes are widespread throughout many areas and regions in the autism brain suggesting that selective and significant neurochemical changes may underlie the observed neuropathology as well as to likely contribute to social-emotional deficits.
Current studies are underway investigating the role of the 5-HT and glutamate systems in these areas as well as extending the studies to additional cortical areas. The changes in GABAA receptors using on-the-slide ligand binding autoradiography itself does not provide information regarding protein expression of this receptor type. In the sections that follow, studies by Fatemi and colleagues address this issue and demonstrate that protein expression does indeed follow the receptor changes for both GABAA and GABAB receptors in a variety of structures in the autism brain.
In the previous section we discussed autoradiography experiments dealing with mRNA localization and receptor densities/affinities for various GABA receptors in a number of brain sites in subjects with autism and matched controls. In this section of the article we will discuss localization and, specifically, protein expression studies for GABAA and GABAB receptors using conventional biochemical techniques of SDS- PAGE and western blotting in various brain areas of subjects with autism and matched controls (Fatemi et al., 2009a,b).
It is well known that cerebellar abnormalities in autism are more extensive than other areas in brain and include loss of granular and Purkinje cells (Ritvo et al., 1986; Bauman and Kemper, 2005) and atrophy of Purkinje cells (Fatemi et al., 2002), and may be responsible for motor associated dysfunction in autism (Nayate et al., 2005). Here, we measured levels of four GABAA receptors (GABRα1, GABRα2, GABRα3, and GABRβ3) and two GABAB receptors (GABBR1 and GABBR2) (Figure 2). All values were normalized against a housekeeping gene, β-actin and expressed as ratios to β-actin in order to provide a more accurate measure of specific changes in levels of GABAA and GABAB receptors. There were reductions of 27–63% in ratios of GABAA receptors/β-actin in cerebella of subjects with autism (Fatemi et al., 2009a). However, only GABRα1 and GABRβ3 reached statistical significance (p<0.007, and p<0.008, respectively; Table 1; Fatemi et al., 2009a). In the same tissues ratios of GABBR1 and GABBR2 receptors/β-actin were significantly reduced by 67% (p<0.0049) and 46% (p<0.026), respectively (Table 1; Fatemi et al., 2009b). Thus, it appears that four out of six examined GABA receptors were downregulated significantly throughout cerebella of subjects with autism.
Interestingly, GABRβ3 knockout mice show significant reduction in size of cerebellar vermal lobules II–VII when compared with control mice (DeLorey et al., 2008). This phenomenon may be related to the observed GABRβ3 downregulation. There is additional support in the literature for 15q11-q13 locus abnormalities in 1–4% of subjects with autism which includes a gene cluster for GABRβ3, GABRα5, and GABRγ3 receptors (Schroer et al., 1998; McCauley et al., 2004). The GABRβ3 abnormality in mice affects learning and memory and results in increased epileptogenesis (DeLorey et al., 1998, 2008). Moreover, GABRβ3 deficits are observed in other areas of the brains of subjects with autism as we will discuss further below.
In BA9 (superior frontal cortex) only one GABAA receptor (GABRα1/β-actin) showed significant (p<0.012), 65% downregulation in brains of subjects with autism vs. matched controls (Figure 2; Table 1; Fatemi et al., 2009a). By the same token, of the two GABAB receptors only GABBR1/β-actin showed a significant, 70% reduction (p<0.021) in BA9 (Table 1; Figure 2; Fatemi et al., 2009b). GABBR2/β-actin, despite a 54% reduction, did not reach significance (Table 1; Fatemi et al., 2009b). In contrast, Samaco et al., (2005) observed GABRβ3 downregulation in BA9 of subjects with autism, which was later correlated with similar results by Hogart et al., (2007). These authors suggested that epigenetic dysregulation may be responsible for the downregulation of GABRβ3 levels in subjects with autism.
In BA40 (parietal cortex) of subjects with autism we observed significant downregulation of all four GABAA receptors: GABRα1/β-actin was reduced by 52% (p<0.018), GABRα2/β-actin was reduced by 39% (p<0.033), GABRα3/β-actin was reduced by 57% (p<0.005), and GABRβ3/β-actin was reduced by 38% (p<0.006) when compared with controls (Figure 2; Table 1; Fatemi et al., 2009a). Despite large reduction in GABBR1/β-actin (71%) and GABBR2/β-actin (79%), only the reduction GABBR1/β-actin reached statistical significance (Table 1; p<0.019) when compared to matched controls. Our data of global reductions in GABAA receptors as well as GABBR1 receptor reductions in BA40 of subjects with autism are unique since no previous published report discusses such reductions in parietal cortex. However, Princivalle et al., (2003) showed altered expression of GABBR1 (1A and 1B isoforms) and GABBR2 in the hippocampus of subjects with temporal lobe epilepsy. It must be stated that seizure disorder is found in 4–44% of subjects with autism (Tuchman and Rapin, 2002). The presence of seizure disorder in individuals with autism may interfere with cognition by causing disturbed vigilance, attention, and language production (Binnie, 1993). It should be pointed out, however, that the majority of patients with seizure disorders do not display cognitive disorders. Despite such high comorbidity for seizures and having had seven subjects in our collection comorbid with seizure disorder, none of the reductions in GABAA and GABAB receptors were correlated positively to be due to seizure production statistically (Fatemi et al., 2009b). There are also multiple animal models for seizure disorder which show a reduction in levels of GABBR1 and GABBR2 expression in various areas of hippocampus (Princivalle et al., 2003; Straessle et al., 2003; Han et al., 2006).
In conclusion, the protein expression data showed significant downregulation in multiple GABAA and GABAB receptors in three important brain sites in subjects with autism which may account for several clinical phenotypes of autism including presence of seizures, learning disorders, and perhaps mental retardation. It is hoped that future clinical trials can use these basic findings towards better use of drugs which can modulate GABA receptor activity in brains of subjects with autism.
Taken together, the autoradiographic and biochemical studies from the respective laboratories indicate profound alterations in the GABAergic system in autism. It is unclear whether these changes are a result of a mutation of the GAD gene(s), epigenetic mechanisms such as changes in DNA methylation or histone modification or other neurodevelopmental abnormalities such as a miswiring of afferents to GABAergic neurons and/or abnormal output of GABAergic neurons. Alterations in the firing rates of neurons may influence GAD mRNA levels, GAD gene expression, GABA release and GABA receptor densities. More studies are needed focusing on glutamatergic afferents in the cerebellum to better assess how the reported changes in the inhibitory system are influenced by or are a consequence of possible disturbances in the principal excitatory system in the autistic brain.
Human tissue was obtained from the NICHD Brain and Tissue Bank for Developmental Disorders; the Harvard Brain Tissue Resource Center, which is supported in part by PHS grant number R24 MH068855; the Brain Endowment Bank, which is funded in part by the National Parkinson Foundation, Inc., Miami, Florida; and the Autism Tissue Program and is gratefully acknowledged. Grant support by National Institute of Child Health and Human Development (1R01HD039459 and 2R01HD039459-05A1) to GJB and (5R01HD052074-01A2 and 3R01HD052074-03S1) to SHF is gratefully acknowledged. Additional grant support from The Nancy Lurie Marks Foundation and The Hussman Foundation to GJB is also greatly appreciated. Collaborative assistance from Drs. J-J. Soghomonian, T.T. Gibbs, J. Yip and A. Oblak to GJB and technical help from T. Folsom and T. Reutiman to SHF is also gratefully acknowledged.