What a Mouse Model Can and Cannot Tell Us About Autism

What the review found

The laboratory mouse is a social animal. It distinguishes one cagemate from another, prefers company to an empty chamber, and will cross a cage to investigate a stranger. Those preferences can be measured in seconds and centimetres, which is why most of what is known about the brain circuits underlying social behaviour has been learned from rodents rather than from people.

That literature was drawn together by Masaaki Sato, Toru Takumi and colleagues working across Hokkaido University, Kobe University, McMaster University in Ontario and the RIKEN Center for Biosystems Dynamics Research, in a 2023 review for Molecular Psychiatry. The authors describe their coverage as selective rather than exhaustive, and it is a narrative survey of 159 papers, with no pooled statistics and no single estimate of how large or how reliable any effect is. Nearly every experiment described in it was performed on mice.

Mice isolated from other mice through the juvenile period become adults with impaired sociability, and a pathway running from the medial prefrontal cortex to a small region of the thalamus fails to activate as it should. Activating that pathway in adulthood, long after the isolation had done its work, restored the sociability the animals had lost.

The same pattern recurs across unrelated models. Mice lacking a working copy of Shank3, a gene that builds the junctions where neurons meet, show impaired social behaviour; restoring the gene in the anterior cingulate cortex of adult animals improved that behaviour, as did stimulating those neurons directly. Mice carrying a single working copy of PTEN show excessive connectivity between prefrontal cortex and amygdala alongside social impairments, and those impairments were reversed both by a drug given during development and, separately, by reducing that circuit’s activity in adulthood.

Sato and Takumi propose a rule for when timing matters. Whether an intervention must arrive early or can arrive late depends on whether the molecule concerned is involved in building neural circuits or in the ongoing function of mature synapses. That is a claim about mechanism, and every result supporting it comes from an animal.

What it means for you and your child

A mouse model of autism is made by introducing into an animal a genetic change found in some autistic people and then measuring whether its social behaviour shifts. The assumption that such an animal models human autism rests on the shared genetics, and this review does not test that assumption or address the long history of rodent findings that failed to translate to people. Human work appears in it only as scattered supporting citations, with no sample size, age range or country given for any of them.

Those human citations are associations. Individuals diagnosed with autism show the same prefrontal-to-amygdala hyperconnectivity seen in the PTEN mice; children with autism have been reported to have lower plasma oxytocin. A single crossover trial of intranasal oxytocin is cited as restoring insular activity and improving the reading of others’ social emotions, and the review reports neither its size nor its dose nor how large the change was. Oxytocin, serotonin-targeting drugs and a gut bacterium have all corrected social behaviour in mice, which measures how far the animal work has run ahead of anything demonstrated in a person, and is not a reason to try any of them.

The bearing on a parent is conceptual, and it is an inference rather than a finding. When a difficulty is described as developmental in origin, the word is often heard as a verdict about timing, as though a window had opened early, closed quietly, and taken the possibility of change with it. In mice, that inference has now failed repeatedly. Circuits shaped by early deprivation or by a mutation present from conception have been altered in adult animals, and the behaviour changed with them.

Whether the same holds in children is unknown. No study in this review addresses it, and the difference between a mouse cortex and a child’s decade of learning, language and relationship is not a difference the animal work is equipped to bridge. The field has not established that any developmental window in humans is critical, nor that it is not; the question is open, and the review’s own authors describe several of the relevant circuits as understudied.

That is grounds for patience while the human work catches up. If a clinician or a school tells you that a developmental origin fixes a ceiling, they have gone past the evidence, and so has anyone who tells you the ceiling has been lifted. What has been shown is that in mice the timing is more forgiving than it was assumed to be, and you are entitled to hear that claim delivered with the animal still attached to it.