Research ArticleAutism Spectrum Disorder

GABAA receptor availability is not altered in adults with autism spectrum disorder or in mouse models

See allHide authors and affiliations

Science Translational Medicine  03 Oct 2018:
Vol. 10, Issue 461, eaam8434
DOI: 10.1126/scitranslmed.aam8434

A window into the ASD brain

A leading hypothesis for autism spectrum disorder (ASD) is that this condition is associated with an imbalance between excitatory glutamate and inhibitory GABA neurotransmission in the brain. To investigate one aspect of GABA signaling, Horder et al. measured the availability of the GABAA receptor in adults with ASD using positron emission tomography. No differences were found compared to control subjects without ASD in two studies performed at different clinical centers. GABAA receptor availability was also normal in three different mouse models of ASD. However, adults with ASD did display altered performance on a GABA-sensitive perceptual task, suggesting that although GABAA receptor density seemed to be normal, GABA signaling pathways could be impaired.

Abstract

Preliminary studies have suggested that γ-aminobutyric acid type A (GABAA) receptors, and potentially the GABAA α5 subtype, are deficient in autism spectrum disorder (ASD). However, prior studies have been confounded by the effects of medications, and these studies did not compare findings across different species. We measured both total GABAA and GABAA α5 receptor availability in two positron emission tomography imaging studies. We used the tracer [11C]flumazenil in 15 adults with ASD and in 15 control individuals without ASD and the tracer [11C]Ro15-4513 in 12 adults with ASD and in 16 control individuals without ASD. All participants were free of medications. We also performed autoradiography, using the same tracers, in three mouse models of ASD: the Cntnap2 knockout mouse, the Shank3 knockout mouse, and mice carrying a 16p11.2 deletion. We found no differences in GABAA receptor or GABAA α5 subunit availability in any brain region of adults with ASD compared to those without ASD. There were no differences in GABAA receptor or GABAA α5 subunit availability in any of the three mouse models. However, adults with ASD did display altered performance on a GABA-sensitive perceptual task. Our data suggest that GABAA receptor availability may be normal in adults with ASD, although GABA signaling may be functionally impaired.

INTRODUCTION

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by impairments in social communication and interaction, together with restricted and repetitive behaviors and interests (13). The prevalence of ASD is estimated at 0.6 to 1.5%, with each case of the disorder being associated with projected lifetime care and loss of productivity costs of around $1.5 million (4, 5). Currently, there are no specific “disease-modifying” pharmacological treatments for ASD, and drug discovery has been hampered by poor understanding of the pathophysiology of the condition.

There is, however, increasing evidence for an excess of the brain excitatory neurotransmitter glutamate as well as a deficit in inhibitory γ-aminobutyric acid (GABA) in ASD (68). For instance, reduced expression of the enzyme that synthesizes GABA, glutamic acid decarboxylase, has been observed in postmortem ASD brain tissue (9). In vivo, alterations in glutamate and glutamine have been reported in the cortex and basal ganglia of children and in the basal ganglia of adults with ASD, as measured by [1H]MRS (proton magnetic resonance spectroscopy) (1013). Reductions in GABA concentration have also been reported by several [1H]MRS imaging studies of ASD children (14, 15). However, a limitation of [1H]MRS imaging is that it is unable to determine neurotransmitter receptor density.

Alterations in the expression of GABA receptors, of which GABA type A (GABAA) is the most prevalent type in the brain (1618), may be a key contributor to the excitatory/inhibitory imbalance in ASD. Reductions in the density of GABAA receptors have been reported in the hippocampus (19) and cingulate cortex (20, 21) of postmortem brain tissue from people with ASD. Although these studies were valuable first steps, their sample sizes were small, and they included individuals taking medications such as anticonvulsants that might alter GABAA expression, leaving it unclear whether their findings could be replicated by in vivo studies.

Two in vivo studies have reported reductions of GABAA in ASD. Using [123I]iomazenil single-photon emission computed tomography (SPECT), Mori et al. (22) reported decreased GABAA availability in children with ASD. However, this study included children with epilepsy and intellectual disability, and some participants were on anticonvulsant drugs. Hence, it remained unclear whether it was ASD, or a confounding factor, that was associated with reduced GABAA receptor availability. Therefore, we subsequently carried out a pilot study (23) using positron emission tomography (PET) with the tracer [11C]Ro15-4513. Our participants were medication-free adult males with ASD and no evidence of epilepsy or intellectual disability and control individuals without ASD. We found significantly reduced availability of the GABAA α5 subunit in ASD, with no alterations in the non-α5 subunits. However, this preliminary study was small, including three adults with ASD and three controls without ASD. Also, this study did not clarify whether it was overall GABAA, or GABAA α5 specifically, that was altered in ASD (23). Hence, we designed the present investigation to replicate and extend prior in vivo GABAA imaging studies in a new and larger participant sample; the data from the pilot study were not included in the present analysis.

We reasoned that if the GABAA receptor availability was abnormal in ASD, it might provide a useful therapeutic target for drug development. To provide a rational screen for new treatments, however, drug development would require animal models that recapitulate any GABAA receptor alterations observed in humans. Therefore, we set out to examine whether GABAA receptor alterations were present in animal models of nonsyndromic ASD as well as in individuals with nonsyndromic ASD. Altered GABAA receptor expression has been reported in rodent models of ASD associated with syndromic disorders including Fragile X syndrome (24) and Rett syndrome (25, 26). However, about 90% of ASD cases (and all of the cases in the present study) were nonsyndromic (27), thus necessitating that we used nonsyndromic rodent models of ASD.

A final motivation for this study was to investigate GABAergic function in adults with ASD. This was based on our reasoning that the functional integrity of GABAergic circuits might be abnormal in ASD even if GABAA receptor availability was unchanged. Our aims in the present investigation were (i) to confirm whether GABAA receptor density was altered in individuals with nonsyndromic ASD compared to control individuals without ASD; (ii) to determine whether any putative changes were specific to the GABAA α5 subunit; (iii) to discover which, if any, of the three rodent models of nonsyndromic ASD displayed the same pattern of GABAA receptor alterations as potentially observed in individuals with ASD; and (iv) to determine whether ASD was associated with functional impairments in GABAergic circuits using a proxy measure of GABAergic function in individuals with ASD.

We therefore conducted an imaging study of GABAA receptor availability in both adults with ASD and three rodent models of ASD. In humans, we carried out two PET imaging studies in two independent clinical centers: in Stockholm, Sweden using [11C]flumazenil as a tracer (28) and in London, United Kingdom using [11C]Ro15-4513 as a tracer (29, 30). We chose [11C]flumazenil as a nonselective tracer because it binds to α1, α2, α3, and α5 subunits of the GABAA receptor (31) and thus has approximately equal affinity for most of GABAA receptors in the brain (32). The tracer [11C]Ro15-4513 was chosen for its 10-fold higher affinity for the α5 subunit over the other subunits α1 to α4 and α6, for which it has approximately equal affinity (31). The α5 subunit of the GABAA receptor has been linked to ASD (23, 33, 34). Both PET imaging studies formed part of the Innovative Medicines Initiative (IMI) European Autism Interventions Multicenter Study (EU-AIMS) and it’s successor AIMS-2-TRIALS (tables S1 and S2) (35).

We selected three commonly used, etiologically diverse, rodent models of ASD that represent a range of genetic risk factors for ASD in humans: mice lacking Cntnap2 (36), mice lacking Shank3 (37), and mice carrying a genetic deletion equivalent to deletion of human chromosome 16p11.2, one of the most common copy-number variants found in ASD (38). Using the same radiotracers that we used for our PET imaging studies, we performed autoradiography to quantify GABAA receptors in the mouse brain. Autoradiography was used rather than small-animal PET imaging because it provides better spatial resolution and because PET imaging would have required the animals to be anesthetized during imaging, which could affect GABA receptor availability. Finally, to investigate the functional balance between excitatory and inhibitory neurotransmission using a visual psychophysics task, we used the paradoxical motion perception (PMP) task known to be a proxy measure of GABA function in individuals with and without ASD (39, 40).

RESULTS

PET imaging with [11C]flumazenil shows no difference in GABAA receptor availability among adults with or without ASD

Using the [11C]flumazenil tracer for PET imaging, we found no evidence of group differences in estimated [11C]flumazenil binding potential (BPND) in the whole brain of adults with ASD versus control individuals without ASD (independent samples t test: t28 = 1.304, P = 0.203, d = 0.476; Fig. 1A). We also found no differences in BPND in brain tissue gray matter (t28 =1.283, P = 0.210, d = 0.468) or in any of the 13 specific brain regions we examined (all P > 0.16; Figs. 2 and 3A and table S3). This suggested that there was no difference in GABAA receptor availability between adults with ASD and those without ASD. There was no difference in [11C]flumazenil tracer dosing between the ASD and control non-ASD groups (table S1). Sensitivity analyses including correction for partial volume effects, inclusion of male participants only, and exploratory voxel-wise analysis did not affect the results (tables S5 and S6). There was no difference between the London and Stockholm cohorts in terms of age or intelligence quotient (IQ) (Table 1).

Fig. 1 PET imaging with the [11C]flumazenil and [11C]Ro15-4513 tracers in adults with ASD or control individuals without ASD.

Estimated group mean for (A) the [11C]flumazenil tracer BPND and (B) the [11C]Ro15-4513 tracer distribution volume (VT) in adults with ASD compared to control individuals without ASD. PET images were normalized into MNI-152 standard space. PET images are displayed overlaid on the 2-mm resolution template anatomical magnetic resonance (MR) image. (A) PET imaging with the [11C]flumazenil tracer revealed GABAA receptor availability in the human brain (n = 15 ASD adults and n = 15 control adults). (B) PET imaging with the [11C]Ro15-4513 tracer showed GABAA receptor availability in the human brain (n = 12 ASD adults and n = 16 control adults). BPND was calculated using stationary wavelet-aided parametric imaging. VT was calculated using the Logan method.

Fig. 2 GABAA receptor availability in different regions of the ASD and control brain using the [11C]flumazenil tracer.

Estimated [11C]flumazenil BPND (mean estimated BPND) in different regions of the brain of adults with ASD (n = 15) and control individuals without ASD (n = 15) in the Stockholm cohort. There was no difference in tracer binding (BPND) between the two groups in any comparison. Error bars show ±1 SEM.

Fig. 3 Mean binding of tracer in the gray matter of adults with ASD and control individuals without ASD.

Scatterplots illustrate (A) the estimated mean binding (BPND) of the [11C]flumazenil tracer in gray matter and (B) the estimated mean binding (VT) of the [11C]Ro15-4513 tracer in gray matter for individual participants in the ASD and control groups. There were no differences in tracer binding between the two groups. There were no evident outliers or subgroups with either tracer.

Table 1 Participant demographic data and ASD clinical scores for the Stockholm and London cohorts.

Values are shown as means ± SD. P values are the outcome of independent samples t tests testing for a difference in group means within each study. Clinical scores were not available for control participants. Autism Diagnostic Interview–Revised (ADI-R) scores were not available (NA) for the Stockholm cohort. ADOS, Autism Diagnostic Observation Schedule.

View this table:

Behavioral assessments in ASD and control individuals undergoing [11C]flumazenil PET imaging

A battery of behavioral tests was administered to all subjects in the Stockholm [11C]flumazenil PET imaging cohort to explore executive function, central coherence, and social cognition. Executive functioning was measured with Delis-Kaplan Executive Function System Verbal Fluency, Tower of Hanoi, and Conner’s Continuous Performance Test. Central coherence was evaluated with the Embedded Figure Test (41) and the Fragmented Picture Test (42). Social cognition was assessed with Reading the Mind in the Eyes Test (43), Movie for Assessment of Social Cognition (MASC) Test (44), and Faux Pas Test (45). Wechsler Adult Intelligence Scale, fourth version (WAIS-IV) subtests Vocabulary and Matrix Reasoning were used to assess the general intellectual ability of individuals with and without ASD.

There were no correlations between GABAA receptor availability in the gray matter of the brains of adults with ASD or control non-ASD subjects and performance on this battery of ASD-related cognitive and social cognitive tasks (table S7). The performance of the ASD group was significantly lower than the control group on letter fluency (t = −3.27, P = 0.006) and alterations between categories (t = −3.96, P = 0.001). In social cognitive tests, the ASD group performed significantly lower on all tasks: Reading the Mind in the Eyes (t = −3.96, P = 0.001), identifying the number of situations in the MASC test (t = −4.74, P < 0.001), and Faux Pas Test (t = −3.89, P = 0.002). No significant group differences were found in the central coherence tasks.

PET imaging with [11C]Ro15-4513 shows no difference in GABAA α5 versus non-α5 subunit availability among adults with or without ASD

Using the [11C]Ro15-4513 tracer for PET imaging and the two-tissue compartment model (2TCM), we found in a primary analysis that there were no group differences between adults with and without ASD in total [11C]Ro15-4513 binding (VT) in the whole brain (independent samples t test: t25 = −1.392, P = 0.176, d = 0.574; Fig. 1B). We also found no group differences between adults with and without ASD in total [11C]Ro15-4513 binding (VT) in gray matter (t25 = −1.516, P = 0.142, d = 0.614) or in 11 of the 13 specific brain regions we examined (all P > 0.1; Figs. 3B and 4). In the cerebellum and the thalamus, nominally significant increases in binding were observed in the ASD group (t25 = −2.132, P = 0.043, d = 0.820 and t25 = −2.513, P = 0.019, d = 0.974, respectively), but these changes were not significant after Bonferroni correction for multiple comparisons (table S4). There were no significant differences in [11C]Ro15-4513 tracer dosing between the ASD and non-ASD groups (table S1).

Fig. 4 GABAA receptor availability in different regions of the ASD and control brain using the [11C]Ro15-4513 tracer.

The mean estimated binding (VT) of the [11C]Ro15-4513 tracer across the whole brain and in different brain regions of adults with ASD (n = 12) and control adults without ASD (n = 16) is shown. Estimated specific binding of the [11C]Ro15-4513 tracer was derived from the 2TCM of PET imaging data. Error bars show ±1 SEM.

In a secondary analysis of the [11C]Ro15-4513 binding data, we used nonlinear spectral analysis (23) to distinguish between GABAA receptor α5 and non-α5 subunit availability. We confirmed that there were no differences in the total brain availability of either non-α5 (t24 = −0.163, P = 0.872, d = 0.064) or α5 (t23 = −0.805, P = 0.429, d = 0.322) receptor subunits in adults with ASD versus control individuals without ASD. Across different brain areas, there were no regional differences in either non-α5 or α5 receptor subunit availability in any brain area. We did observe an apparent increase in α5 receptor subunit availability in the insula of adults with ASD compared to control individuals without ASD (t18 = −2.345, P = 0.029, d = 1.063), but this increase disappeared after Bonferroni correction for multiple comparisons (Figs. 1, 3B, and 4). Notably, we did not observe any differences in α5 subunit availability in the hippocampus or amygdala (P > 0.3), the areas in which significant differences were detected in our previous pilot study (23). Finally, there was no difference between adults with ASD compared with control individuals without ASD in terms of age or IQ (Table 1).

Autoradiography in brain tissue from three ASD mouse models and WT mice shows no difference in GABAA receptor availability

We performed autoradiography in vitro to detect GABAA receptor availability in brain tissue from three mouse models of ASD: mice lacking Cntnap2, mice lacking Shank3, and mice with a 16p11.2 deletion. Brain tissue from six brain regions—the frontal cortex, cingulate cortex, caudate/putamen, dorsal hippocampus, cerebellum, and amygdala—was examined bilaterally. We observed no significant differences in the specific binding of either the [3H]flumazenil tracer or the [3H]Ro15-4513 tracer in brain tissue from any of the three mouse models compared to WT animals (all P > 0.05) (Fig. 5, fig. S1, and source data in table S8).

Fig. 5 GABAA receptor availability in mouse brain tissue using autoradiography.

(A to D) Illustrations of mouse brain regions of interest (ROIs) overlaid on representative autoradiography images from wild-type (WT) mouse brain are shown. ROIs were defined on the basis of Paxinos and Franklin’s mouse brain atlas. The numbers represent the specific regions that were analyzed. Averages were calculated from both brain hemispheres. (A) 1, frontal cortex; (B) 1, caudate/putamen; 2, cingulate cortex; (C) 1, dorsal hippocampus; 2, amygdala; (D) 1, cerebellum. (E, G, and I) Graphs show specific binding of the [3H]flumazenil tracer to brain tissue from three different mutant mouse models of ASD (16p11.2 deletion, Cntnap2−/−, and Shank3−/−) compared to WT control mice. (F, H, and J) Graphs show specific binding of the [3H]Ro15-4513 tracer to brain tissue from three different mutant mouse models of ASD (16p11.2 deletion, Cntnap2−/−, and Shank3−/−) compared to WT control mice. Bars show the group mean binding signal normalized such that the mean signal in the WT group was 100%. Error bars show SEM.

We subsequently examined the left and right amygdala separately, following Bertelsen et al. (46), who reported an increase in binding of the Ro15-4513 tracer in the left amygdala of rats with ASD induced by valproate. We found no significant differences in either [3H]flumazenil or [3H]Ro15-4513 tracer binding in any of the three ASD mouse models (fig. S1). However, in the Shank3-deficient mice, a trend of increased binding of both [3H]flumazenil and [3H]Ro15-4513 was observed (adjusted P = 0.09) in the left amygdala. In Cntnap2-deficient mice, there was a trend toward an increase in binding of [3H]Ro15-4513 in both the left and right amygdala (adjusted P = 0.06 and P = 0.1, respectively). No significant differences or trends were observed in the case of the 16p11.2 deletion mouse model (fig. S1).

A difference between adults with and without ASD in the PMP task

We used the PMP task as a proxy measure of GABA signaling in the visual cortex of adults with ASD compared to control individuals without ASD. We first verified that the PMP effect (39) was present in the control individuals without ASD. The effect was significant (one-sample t test: t17 = 3.99, P < 0.001; mean PMP ratio, 1.844; 95% confidence interval, 1.398 to 2.290, where a ratio of 1 is no effect), demonstrating successful implementation of the task (Fig. 6).

Fig. 6 PMP task undertaken by adults with ASD and control individuals without ASD.

The contrast-induced motion perceptual impairment ratio in each participant is shown, indicating the degree to which motion perception is impaired by large, high-contrast stimuli. Higher values indicate greater impairment. Impairment is considered a proxy for GABA-mediated inhibition, driven by high contrast, in the visual cortex. The line indicates a ratio of 1.0 representing no impairment. The control group shows greater impairment than does the ASD group; t28 = 2.487, P = 0.022, d = 0.921 (two-tailed t test).

The magnitude of the contrast-induced motion perceptual impairment ratio was reduced in the ASD group compared to the control group [independent samples t test with Welch-Satterthwaite correction for unequal variances: t28 = 2.487, P = 0.022, d = 0.921 (two-tailed); Bootstrapped t test: P = 0.028 with 10,000 samples; Fig. 6]. Adults with ASD were less impaired by the high-contrast stimuli than were control individuals without ASD. This group difference was not accounted for by low-level differences in the ability to detect motion because there were no group differences in perceptual threshold under the two control conditions (both P > 0.4). Furthermore, the paradoxical effect of contrast on motion perception did not correlate with age (n = 30, r = 0.026, P = 0.892) or IQ (n = 29, r = 0.243, P = 0.203). Thus, adults with ASD showed a selective enhancement in motion perception under conditions of high contrast, results considered consistent with a defect in GABA signaling.

DISCUSSION

We conducted a PET imaging study at two different sites (London and Stockholm) to measure GABAA receptor availability in the brains of adults with ASD compared to adults without ASD. We also conducted an autoradiography study in three mouse models of ASD. Our objective was to test the hypothesis that ASD is associated with reduced amounts of GABAA receptors in the brain. We found no direct evidence in support of this hypothesis. Our results show that the unmedicated adults with idiopathic ASD but without intellectual disability in our study exhibited normal availability of both total GABAA receptors and the key GABAA receptor subunit, α5. There were no correlations between GABAA receptor availability and performance in the following social cognitive tests: Reading the Mind in the Eyes, MASC, and Faux Pas. Furthermore, we found using autoradiography that GABAA receptor availability was not altered in the brains of three mouse models of ASD: Cntnap2-deficient mice, Shank3-deficient mice, and mice with a 16p11.2 deletion. A visual perceptual proxy measure of cortical GABAergic signaling, the PMP task, revealed evidence consistent with a functional GABA signaling deficit in adults with ASD in the [11C]Ro15-4513 PET imaging study.

Our results contrast with reports of reduced GABAA receptor availability in children with ASD using SPECT imaging (22). The difference between those results and our data may indicate an age-specific alteration in GABAA receptor availability in ASD, but other methodological differences should also be considered. For instance, in the SPECT imaging study, the control group consisted of patients with epilepsy, most of whom were medicated with anticonvulsant drugs. In contrast, we recruited medication-free adults with ASD and control individuals without ASD who showed typical development to avoid any medical or medication confounds. The null findings in our PET imaging study also stand in contrast to the results of our prior pilot study (23) that reported reduced GABAA α5 subunit availability in three adults with ASD compared to three matched controls. It is possible that our prior finding reflected a type I error.

Although our current results suggest that GABAA receptor availability was not altered in ASD, other interpretations of our data should be considered. One possibility is that our two studies lacked statistical power to detect any differences in GABAA receptor availability. Both of our PET imaging studies were powered at 80% to detect a typical effect size (44). A second possible explanation for our negative result is that the PET imaging tracers we used were not sensitive enough. However, we consider this to be unlikely because both the [11C]flumazenil tracer and [11C]Ro15-4513 tracer have been used successfully to detect group differences in GABAA receptor availability, both in patients with other neuropsychiatric disorders (47, 48) compared to healthy controls and in controls differing in behavioral phenotypes (49).

We did not find any detectable differences in GABAA receptor availability using the tracers [11C]flumazenil and [11C]Ro15-4513 and PET imaging in adults with and without ASD. We did, however, observe an alteration in performance between the two groups on a visual perception task that has been proposed as an index of cortical GABAergic function (39, 40). A number of other groups have reported similar findings in individuals with ASD (50). This raises the possibility that the GABAergic system may be dysregulated in ASD despite an intact GABAA receptor density in the brain. For instance, it is possible that the synthesis or release of GABA is down-regulated in ASD, leading to impaired GABAergic signaling (51). Moreover, although GABAA receptors may be present in normal numbers, they may be abnormally localized on postsynaptic neuronal membranes, thus affecting their functionality, as has been observed in epilepsy (52). Future research studies should explore these possibilities.

Our rodent autoradiography results support our PET imaging findings in adults with ASD. GABAA receptor availability was not altered in the brains of any of the three genetic mouse models of ASD that we examined. Alterations in GABAA receptor availability might have been expected on the basis of the functions of the genes deleted in the three mouse models. Cntnap2 encodes a cell adhesion molecule belonging to the neurexin family (53) that is involved in dendritic arborization and neuronal spine development. Cntnap2 knockout mice show impaired excitatory synapse function and reduced inhibitory (GABAergic) transmission (50). Shank3 is another synaptic protein that plays a role in signaling to presynaptic neuronal membranes through the neuroligin-neurexin trans-synaptic signaling complex (54), and Shank3 has been linked to inhibitory neuronal signaling (55, 56). Finally, the 16p11.2 deletion is a chromosomal microdeletion associated with ASD in humans (57). In contrast to our results, increased binding of Ro15-4513 (but not muscimol) was recently reported in the left amygdala of rats treated with valproate (46). Our current study differed from that study in a number of ways. For example, we used mice not rats, we examined genetic mouse models of ASD rather than a valproate-induced rat model of ASD, and our animals were older (12 weeks versus 7 weeks of age). Therefore, our finding of no alterations in binding of the GABAA receptor tracer in our three ASD mouse models does not rule out the possibility that other animal models might show differences in tracer binding. We suggest, however, that our data at least establish that GABAA receptor availability changes are not a shared phenotype across all animal models of ASD.

There are a number of limitations to our study. These include the fact that because of prohibitions on exposing children to research PET imaging, we were only able to recruit adults; thus, we cannot exclude the possibility that there would be abnormalities in GABAA receptor availability in children with ASD. Similarly, we were only able to recruit males in the [11C]Ro15-4513 PET imaging study because of UK regulations restricting the exposure of women of reproductive age to radiation for research purposes. This limits the generalizability of our findings, although most of the individuals with ASD are males (3). Our ASD population was also free of medical comorbidities, with above average mean IQ scores and no cases of epilepsy. Although this represents a strength of the study in terms of avoidance of confounding factors, it also limits the generalizability of our findings. It is possible that GABAA receptor abnormalities are present in ASD cases who also have epilepsy or intellectual disability. Testing this hypothesis using PET imaging, however, could raise ethical issues over exposure to radiation in a vulnerable group of people who may be less able to give informed consent. Finally, we were able to selectively quantify the GABAA α5 receptor subunit, but we could not perform the same procedure for other receptor subunits such as α2 and α3 owing to a lack of suitable tracers.

A further limitation of our study is that we included only a limited range (three) of mouse genetic models of ASD for autoradiography. GABAA receptor alterations in other animal models cannot be ruled out. One could question whether the mouse models we used were appropriate models of the phenotype shown by our human ASD participants. For instance, many humans with Cntnap2, Shank3, and 16p11.2 mutations have epilepsy (58), whereas we excluded people with epilepsy in our two PET imaging studies. However, we do not think that this is a key limitation, given that we reported no differences in GABAA receptor availability in any of the three mouse models. Had we observed alterations in rodent GABAA receptor availability in mouse brain tissue, it would have been difficult to exclude the possibility that these alterations were correlates of epilepsy rather than of ASD per se; this concern does not apply to our null results. A final limitation of our PET imaging and autoradiography studies is that they were only able to quantify the density of GABAA receptors in brain tissue and not their function; other methods such as electrophysiology would be required to achieve this goal.

We used PET imaging and autoradiography to investigate the inhibitory GABAergic system in patients with ASD and in three mouse models of ASD, respectively. Our objective was to test the excitatory:inhibitory imbalance hypothesis, namely that GABAA receptor availability is reduced in ASD. Contrary to the hypothesis, we found normal availability of total GABAA receptors and the GABAA non-α5 and α5 subunits in adults with ASD compared to control individuals without ASD. We recapitulated these findings in three different genetic mouse models of ASD. We did, however, find behavioral differences on the PMP task between adults with ASD and control individuals without ASD, consistent with a functional GABAergic deficit in ASD.

MATERIALS AND METHODS

Study design

We carried out two human PET imaging studies in Stockholm (Sweden) and London (United Kingdom). Each study was designed to compare GABAA receptor availability in adults with ASD to control individuals without ASD. The two studies were intended to investigate GABAA receptor availability and to provide convergent evidence from two different clinical centers. In the Stockholm study, we measured brain GABAA receptor binding using PET imaging with the nonselective tracer [11C]flumazenil (also known as [11C]Ro15-1788) (59). In the London study, we measured GABAA receptor non-α5 subunit and α5 subunit availability using PET imaging with the α5 subunit–selective tracer [11C]Ro15-4513. We included data from 15 (11 males) adult control individuals who had normal development and 15 (11 males) adults with ASD in the [11C]flumazenil Stockholm study. We included data from 16 (all male) adult controls and 12 (all male) adults with ASD in the [11C]Ro15-4513 London study.

We also performed a quantitative autoradiography experiment to estimate GABAA receptor density in brain tissue from three mutant mouse strains commonly used as models of ASD: Cntnap2-deficient mice, Shank3-deficient mice, and mice with a 16p11.2 deletion (stock no. 013128, The Jackson Laboratory). Control mice included WT littermates. In all three models, there was n = 7 to 8 for the WT control mouse group and n = 8 for the mutant mouse group. Randomization was not used because this was an observational study. Analysis and quantification of images for PET and autoradiography were performed by individuals blinded to group status.

Participants and recruitment for PET imaging studies

The [11C]flumazenil PET imaging study was approved by the Ethics and Radiation Safety Committees of the Karolinska Hospital and was performed in accordance with the Helsinki Declaration. The [11C]Ro15-4513 PET imaging study was approved by the North London Research Ethics Committee 3 (reference no. 10/H0709/90) and by the Administration of Radioactive Substances Advisory Committee (certificate no. 630/3764/28579). The PMP task study was approved by Essex 2 National Research Ethics Committee (reference no. 04-Q0102/26).

In the Stockholm study, the ASD participants were recruited through the Stockholm County Health Care Services. In the London study, adults with ASD were recruited through the Behavioural Genetics Clinic at the Maudsley Hospital, a national referral service for the diagnosis of neurodevelopmental disorders. In each center, control individuals with normal development were recruited using local advertisements.

All participants were required to meet the following criteria: (i) full-scale IQ higher than 70; (ii) not currently taking any psychoactive medications and no history of psychoactive medication use within the 6 weeks before the study; (iii) no history of epilepsy, head injury or trauma (with loss of consciousness of more than 5 min), brain disease or infection, or serious medical illness by self-report; (iv) no history of bipolar disorder, schizophrenia, drug or alcohol dependency, or other DSM (Diagnostic and Statistical Manual of Mental Disorders) Axis I disorders; (v) was physically healthy and was able to safely undergo PET imaging; (vi) not pregnant (for female participants); and (vii) aged 18 years or older.

In addition, participants in the ASD group were required to meet ICD-10 (International Statistical Classification of Diseases and Related Health Problems, 10th revision) criteria for autism and to have received a clinical diagnosis of ASD by an experienced psychiatrist involved in the study. Diagnoses were informed by the ADOS in all participants and, where possible, the ADI-R schedule (10 of 12 ASD participants in the London sample). Details on the previous use of psychoactive medications and a measure of functional impairment (employment status) are presented in table S2.

For the [11C]flumazenil PET imaging study in Stockholm, IQ was estimated from the Vocabulary and Matrix Reasoning subscales of the Wechsler Abbreviated Intelligence Scale IV (WAIS-IV), administered in Swedish (60). For the [11C]Ro15-4513 PET imaging study in London, IQ was assessed using the English version of the Wechsler Abbreviated Scale of Intelligence (WASI) with four subscales: Vocabulary, Similarities, Block Design, and Matrix Reasoning.

[11C]Flumazenil and [11C]Ro15-4513 PET imaging data acquisition and processing

The following is a summary of the methods used in the two human PET studies (for more details, see the Supplementary Materials). [11C]Flumazenil was prepared as previously described (61) and injected in the antecubital vein as a 10-s bolus. [11C]Flumazenil PET was performed on a Siemens ECAT Exact HR 47 (Control Technology Inc./Siemens) run in three-dimensional (3D) mode with dual-energy Windows scatter correction. A 10-min transmission scan was performed to correct for attenuation and scatter. PET data were collected in a dynamic scan lasting for 63 min and binned into 26 time frames. PET images were reconstructed using filtered back projection and corrected for head motion using a frame to first frames approach. T1-weighted MR images were co-registered to PET using SPM5. MR images were segmented, and ROIs were delineated using FreeSurfer (62). Regional BPND values were calculated using the simplified reference tissue model, with pons acting as reference region.

[11C]Ro15-4513 was synthesized as previously described and injected in the dominant antecubital vein as a 15-s bolus. PET emission data were collected in 3D mode for 90 min. During the PET acquisition, arterial blood data were sampled via the radial artery using a combined automatic-manual approach. Dynamic PET data were corrected for interframe motion and aligned with the individual’s structural T1-weighted MR image. An atlas was co-registered on each subject’s image space, and a subset of regions including amygdala, hippocampus, and nucleus accumbens was considered. Two kinetic methods with metabolite-corrected plasma input functions were applied for the data analysis: (i) a 2TCM and (ii) a spectral-based quantification method, nonlinear spectral analysis to separate GABA non-α5– and α5-related components (V1 and V5, respectively) from the total tracer uptake (34).

Autoradiography in three mouse models of ASD

We performed quantitative autoradiography to estimate GABAA receptor density in brain slices from Cntnap2 knockout (Cntnap2−/−), Shank3 knockout (Shank3−/−) (both provided by M. Saxe, Roche), and 16p11.2 (deletion) (stock no. 013128, The Jackson Laboratory). Control mice included WT littermates. Mice were killed by decapitation, and their brains were rapidly removed and frozen in cooled isopentane. To approximate the participant demographics in the human ASD PET study as closely as possible, we used mature adult (12 weeks of age), male mice in all cases. In all three groups, there was an n = 7 to 8 for the WT mice and an n = 8 for the knockouts.

PMP task

To complement our PET imaging measures of GABAA receptor availability, we chose to investigate the functional balance between excitatory and inhibitory neurotransmission using the PMP task as a proxy measure. In this paradigm, high visual contrast normatively impairs perception of the direction of motion of sine wave grating stimuli, that is, paradoxically, lower-contrast stimuli are easier to perceive. This effect may reflect GABAergic lateral inhibition in the visual cortex (39).

During the PMP task, participants view brief presentations of a drifting sine wave gratings of varying size (small versus large) and contrast (bright versus faint). The task is to indicate the direction of motion. PMP stimuli were grayscale vertical sine wave gratings masked with a Gaussian centered on the middle of the screen. Gratings had a spatial frequency of 1 cycle/degree of visual angle at the viewing distance of 50 cm. Each grating drifted at 2°/s either left or right (randomly determined on each trial). Stimulus size, defined as the spatial SD of the Gaussian mask, was 0.35° for small and 2.50° for large stimuli. The bright stimuli used the maximum contrast of 127 provided by the Presentation software, while the dim stimuli used a contrast of 5, which was chosen in piloting as the lowest contrast that was clearly perceptible to all participants.

There were 260 trials, 65 each of four trial types (small dim, small bright, large dim, and large bright). Psychometric estimation was performed using a 2-up-1-down duration staircase. There were four staircases, one per stimulus type. Each staircase began at 250 ms, with the step up (harder) at −14.1 ms per two correct answers (that is, the stimuli were made briefer) and step down (easier) at 18.2 ms per one incorrect answer (that is, the stimuli were made longer). A failure to respond was counted as incorrect. Participants performed the task in a darkened room.

In line with previous work, contrast impairment on motion perception is defined as the ratio between the estimated threshold (in milliseconds) for large-bright compared to large-faint stimuli. Trial-by-trial results were analyzed using a custom Microsoft Excel macro to detect reversals. We calculated perceptual thresholds for each stimulus condition by taking the mean of all of the reversals because this is the best estimator of the psychometric threshold in a forced-choice staircase design.

Statistics

For all analyses, α was set at 0.05, and two-tailed tests were used. For the [11C]flumazenil and [11C]Ro15-4513 PET imaging studies, group differences in estimated receptor availability (BPND and VT, respectively) were compared using an independent samples t test for each of the brain regions. Results are reported as uncorrected P values, with remarks indicating whether significant results would survive Bonferroni correction over the number of brain regions within each study. For the PMP task, the magnitude of the contrast-induced motion perceptual impairment ratio was compared between groups using an independent samples t test with Welch-Satterthwaite correction for unequal variances. For the rodent autoradiography studies, nonspecific binding was subtracted from the total binding to give the specific binding values, which were compared between WT and mutant mice using unpaired t tests corrected for multiple comparisons using Šidák correction.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/461/eaam8434/DC1

Methods

Fig. S1. Autoradiography in the left and right amygdala of mice.

Table S1. PET tracer dosing in the two human studies.

Table S2. Further information on functional outcome (employment status) and medication history in the ASD participants in the two PET studies.

Table S3. [11C]Flumazenil BPND in the ASD and control groups.

Table S4. [11C]Ro15-4513 VT in the ASD and control groups.

Table S5. Sensitivity analysis: [11C]Flumazenil BPND in ASD and control groups using correction for partial volume effects.

Table S6. Sensitivity analysis: [11C]Flumazenil BPND in ASD and control groups, excluding female participants.

Table S7. Pearson correlations between cognitive performance and [11C]flumazenil BPND in gray matter.

Table S8. Source data for autoradiography experiments.

References (6371)

REFERENCES AND NOTES

Acknowledgments: We thank J. Myers for help with the PET imaging data analysis. We thank N. Gillan, S. Coghlan, D. de la Harpe Golden, L. Brennan, C. E. Wilson, C. Murphy, and D. Robertson for help with participant recruitment. Funding: This study was funded by EU AIMS and AIMS-2-TRIALS. EU-AIMS receives support from the IMI Joint Undertaking (JU) under grant agreement no.115300, resources of which are composed of financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013), from the European Federation of Pharmaceutical Industries and Associations (EFPIA) companies’ in kind contribution, and from Autism Speaks. AIMS-2-TRIALS received funding from the IMI 2 JU under grant agreement no. 777394. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA, Autism Speaks, Autistica, and the Simons Foundation Autism Research Initiative. J.H. was supported by the Wellcome Trust (grant no. 091300/Z/10/Z) and by the Biomedical Research Centre (BRC) at King’s College London. J.L. was funded by the Swedish Research Council (no. 523-2013-2982). D.G.M. was supported by the Mortimer D. Sackler Foundation and also by the National Institute for Health Research BRC at South London and Maudsley NHS (National Health Service) Foundation Trust and King’s College London. J.B. was supported by the Swedish Brain Foundation, the Stockholm Brain Institute, and the Thuring Foundation. O.D.H. was supported by Medical Research Council, UK grant no. MC-A656-5QD30. Author contributions: D.G.M. and J.B. jointly designed the PET imaging experiments. N.S. designed the autoradiography experiments. M.A.M., M.A., I.M., and S.S. performed participant recruitment and carried out PET imaging of individuals under the supervision of Ä.T., J.L., A.G., C.H., S.B., L.F., D.J.N., A.L.-H., and O.D.H. PET imaging data analysis was performed by M.V., F.T., and M.A. J.H. and M.A.M. implemented and carried out the PMP task and analyzed these data. N.S., T.S., D.C., and K.H. carried out autoradiography on mouse brain tissue under the supervision of A.G. N.S. analyzed these data. D.G.M., O.D.H., and J.B. secured funding. J.H. and M.A. performed statistical analysis and wrote the first draft of the paper. All authors contributed to the final version of the manuscript. Competing interests: S.B. consults for Shire, Medice, Roche, Eli Lilly, Prima Psychiatry, GL Group, System Analytic, Kompetento, Expo Medica, and Prophase. O.D.H. consults for AstraZeneca, Autifony Therapeutics, Bristol-Myers Squibb, Eli Lilly, Heptares, Jansen, Lundbeck, Lynden-Delta, Otsuka, Servier, Sunovion, Rand, and Roche. The other authors declare that they have no competing interests.
View Abstract

Navigate This Article