E/I imbalance: an influential hypothesis of autism

Balance of excitatory and inhibitory activity is key to brain functioning, and an imbalance between them is believed to underlie autism (and other brain conditions, e.g. epilepsy and schizophrenia). Although this theory has been influential, findings are heterogeneous, and how this imbalance underlies autism is not yet well known. It could be due to excessive or insufficient excitation or inhibition, and it could also be that imbalances in different brain regions are responsible for different characteristics of autism.

We wanted to investigate links between the most common excitatory (glutamate) and inhibitory (GABA) neurotransmitters, and how they relate to differences in brain structure and behavior in autism. We did this looking at gene-sets of glutamate and GABA pathways (genes encoding proteins involved in the glutamate and GABA pathways in the brain), to cortical thickness differences (measured using MRI), and behavioral measures of autism (diagnostic interviews and questionnaires), in the AIMS-2-TRIALS LEAP cohort.

We investigated this by combining two analysis methods: competitive gene-set analysis linking genetic variation of the glutamate and GABA gene-sets to behavior (using MAGMA), and correlating gene-expression to cortical thickness differences (between autistic vs. non-autistic people). The competitive gene-set analysis tests whether the genetic variation in the glutamate and/or GABA gene-set is more strongly associated with the behavioral measure (scores on the interview/questionnaire), compared to the other genes in the genome. The latter, the gene-expression analysis, was done adapting analysis methods that have been used previously using gene-expression data from the Allen Human Brain atlas. We looked at whether the expression of our glutamate and/or GABA genes was correlated with cortical thickness differences between our autistic and non-autistic participants. The difference between these analysis methods is that the gene-set analysis looks at genetic variation, whereas gene-expression is about the activity and timing of coding proteins that are involved in the glutamatergic and GABAergic pathways in the brain. 

The findings

Glutamate and GABA genes were differently linked to autism behaviors. In the gene-set analysis we found that the genetic variation of the glutamate gene-set was linked to all the diagnostic interview scores; ADI-R (Autism Diagnostic Interview - revised) and ADOS-2 (these are interviews typically used to diagnose autism). The GABA gene-set was associated with sensory processing differences, although this was a weaker association that did not survive correcting for multiple tests. However, GABA concentrations have been linked to sensory processing previously as well. So, the genetic variation of the glutamate genes were linked to broader autism characteristics, whereas the GABA genetic variation was associated with sensory processing. In the gene-expression analyses we saw that more gene-expression of both glutamate and GABA genes was correlated with larger differences in cortical thickness. However, this effect was different across ages. In adolescence, this effect was positive, suggesting larger cortical thickness in the non-autistic participants, whereas in adults this effect was negative indicating larger cortical thickness in the autistic participants. You can see this in Figure 1, and in more detail across brain regions in Figure 2.

Taken together, our findings show that glutamate and GABA genes are differently linked to structural and behavioral characteristics of autism. More specifically, glutamate and GABA genetic variation links differently to behavioral characteristics, and their gene-expression links to cortical thickness differences in autistics but this effect differs across development.

Figure 1: Distributions of the correlations of cortical thickness differences (autistics vs non-autistics) and gene-expression in adults (A) and adolescents (B). The x-axes show the correlation coefficient between cortical thickness differences for all the genes in the gene-sets, the gray bars indicate the created null-distribution of the correlations. The dashed line indicates the correlation coefficient for the gene-set that was tested, if this dashed line is outside of the gray box that indicates a significant correlation between gene-expression and cortical thickness difference. In adults (A) this line is on the left side of the gray box, indicating a negative gene-expression correlation with cortical thickness difference. In adolescents (B) the dashed lines lay on the right side, indicating a positive correlation with cortical thickness difference. 

Figure 2: Gene-expression and cortical thickness difference of the highest correlating genes across brain regions. Here we plot the most significantly correlated glutamate and GABA gene in both adults (A) and adolescents (B) across brain regions, and the cortical thickness difference between autistics vs non-autistics on the left side panels. In adults (A) the gene expressions of the glutamate and GABA genes (solid black lines) are in an opposite direction compared to the cortical thickness differences (dashed lines) across brain regions, indicative of the negative correlations. In adolescents (B) the gene expression is following a similar direction to the cortical thickness differences, indicative of the positive correlations.

Why is this useful?

Getting a better understanding of the role of excitation/inhibition (im)balance in autism, and linking different aspects of the brain to behavior is particularly useful for pharmacological studies. While there are already medications that affect glutamate and GABA levels, studies investigating their effects on alleviating characteristics that autistics may want support with has so far had limited success. This could be due to the large diversity between autistic people, but also variability in outcomes. Here we also saw that glutamate and GABA related differently to behavioral characteristics of autism. We also saw that glutamate and GABA might have different effects at different ages, meaning that the effects of medication could be different depending on the age of the person taking them. By continuing to understand these relationships better, we can develop better targeted and personalized support options for those who may want it.

The next steps

It is important to keep in mind that genetic variation and expression of our glutamate and GABA genes do not directly translate to more or less concentrations of these metabolites in the brain. Genes can encode for both loss- and gain- of function, and we did not look into those functions here. A logical next step is to look at in vivo measures of these metabolites, and the way to do that in humans is by using magnetic resonance spectroscopy (MRS). In this cohort (LEAP, as mentioned above) and in an additional and related project called the Preschool Imaging Project (PIP), we are currently collecting a new wave of data that includes MRS measures of both glutamate and GABA. With this we will be able to investigate how concentrations of these metabolites link to all the brain and behavior measures investigated here, which will help us get an even more detailed understanding of the role of excitatory/inhibitory (im)balance in autism.