A second pathway in ASD that could be modulated by drugs concerns the synaptic excitation–inhibition balance. Several genes associated with ASD seem to be involved in the formation of excitatory and inhibitory synapses, such as those encoding neuroligins and neurexins. In addition, mutations in genes associated with epilepsy, such as SCN1A, which encodes a voltage-gated sodium channel, have been found in patients with ASD76. In mice haploinsufficient for Scn1a, treatment with low-dose clonazepam, a positive allosteric modulator of GABAA receptors, completely reversed the abnormal social behaviors and deficits in fear memory77. Activation of the GABAB receptor by arbaclofen decreased mRNA translation in the cortex of fragile X mice and corrected the increased spine density78. In humans, a recent randomized, double-blind, placebo-controlled crossover study using arbaclofen in 63 subjects with a full mutation in the FMR1 gene showed a slight improvement of social abilities measured by the aberrant behavior checklist social avoidance subscale but no significant benefit in the primary outcome measures quantified by the aberrant behavior checklist irritability subscale79. Another example is provided by the Nlgn1 knockout mouse, which shows a reduced NMDA/AMPA ratio at corticostriatal synapses52. In these mice, administration of the NMDA receptor (NMDAR) partial co-agonist D-cycloserine rescued the repetitive grooming phenotype. In Shank2 knockout mice, direct stimulation of NMDARs with D-cycloserine normalized NMDAR function and improved social interaction51. Recently, an alternative and interesting new strategy to modulate the excitation–inhibition balance emerged that used the diuretic chloride-transporter antagonist bumetanide, which reduces intracellular chloride concentration, reinforcing GABAergic inhibition. A recent clinical trial in patients with ASD showed improvements calling for larger-scale randomized trials to test this drug80.

One of the main challenges for clinical trials in ASD is the considerable degree of heterogeneity, which could mask the effects of the tested treatment. Furthermore, the clinical scores most frequently used to measure treatment efficacy in clinical trials might be too focused on the core ASD symptoms and do not necessarily reflect improvements in quality of life91. For example, most of the pharmacological trials for ASD in ClinicalTrials.gov (n = 153) are using the clinical global impression scale (33%), which measures clinical improvement, and the aberrant behavior checklist (31%), which measures the comorbid problem behaviors associated with ASD. Both instruments do not take into account the quality of life of patients, and an improvement measured by these instruments does not always correlate with better functioning in everyday life92. Therefore, one possibility would be to have quality-of-life scales as the primary outcome scales and autism checklists as the secondary ones. Or, when designing new studies, researchers could flag from the beginning that both types of scales should be preset as the (combined) primary outcome variable.

We propose that three major steps are needed to improve the efficacy of clinical trials for patients with ASD and synaptic defects (Fig. 3): (i) obtain more homogenous subgroups of individuals with ASD, as determined on the basis of the affected genes or pathways rather than on the clinical phenotype; (ii) identify relevant clinical outcomes and biomarkers related to these subgroups (for example, neuronal, biochemical and brain imaging markers); and (iii) increase knowledge of the natural history of this subgroup of patients to identify, for example, periods of symptoms worsening.

A better knowledge of the factors that regulate each of these regulatory steps might help to identify new targets for restoring the amount of specific synaptic proteins in patients. In addition, given that several genes associated with ASD are regulated by neuronal activity20, it is most likely that a combination of pharmacological treatments (that restore the amount of synaptic proteins) and behavioral treatments (that stimulate specific neuronal circuits) might show the most benefit for patients.