,
16 tweets,
3 min read
Read on Twitter
A few general thoughts on mouse models of "autism". I think we need to be very careful in how we talk about these...
First, "autism" is not a thing. It's an umbrella term for lots of things that manifest with more or less similar cluster of symptoms. But it's very highly heterogeneous on a clinical level.
No person with autism is a good model of "autism". Why should we expect any particular mouse line to be a good one?
Also, autism is defined by traits, not states. We may arguably be able to model psychosis or mania or depression or anxiety or inattention or hyperactivity in animals, but these are states.
Many can be acutely induced by drugs or acutely treated with drugs - both in humans and animals. This is not the case for the symptoms that define autism
These include deficits in social communication and interaction... autismspeaks.org/dsm-5-criteria 
And restricted, repetitive patterns of behavior and interests
We don't really have much of a handle on the underlying neurobiology of either of those sets of behaviors, in humans, and it's not obvious what would constitute proxies in animals
There are certainly candidate behaviors we could look at in animals (measures of social interaction may be directly relatable, but having repetitive interests is hard to distinguish from compulsiveness or perseveration)
But how would you know you are looking at a valid behavioral construct or a relevant effect on that construct in any given mouse model? Without pharmacology as a reality check (flawed as that is), the arguments are sort of circular
Maybe genetics will come to the rescue! Well, kind of, but we need to be realistic about the specificity...
We now know of many rare mutations (copy number variants or single-gene) that strongly increase risk of autism. But none of these is specific.
There are no "genes for autism". There are mutations that cause developmental brain dysfunction that can manifest as autism in some, but not all, carriers.
And the range of severity can be enormous, from severe intellectual disability to much milder symptoms, with differences even between MZ twins.
Modeling the effects of these high-risk mutations in animals is interesting and important and will hopefully reveal some of the underlying neurobiology but they're not "models of autism".
That strategy (modeling validated high-risk mutations) is still the best option, IMO - we just need to be more circumspect in how we describe these and more realistic about the underlying heterogeneity and variability
