Understanding the Genetics of Autism

“Mutations in many different genes are individually associated with high risk of autism.”

First, mutations in many different genes are individually associated with high risk of autism. Some of these have been known for a long time, including mutations causing Fragile X syndrome or Rett syndrome, for example. Over a hundred additional risk genes have been identified in recent years through DNA sequencing technologies of many thousands of autistic individuals, as well as their parents and siblings. In addition, a large number of chromosomal deletions or duplications, which can affect many genes at once, have also been discovered to confer high risk of the condition.

Currently, around 25–35% of cases of severe autism, and 5–6% of less severe cases, can be assigned a genetic diagnosis — i.e., a presumably causal mutation can be identified. These numbers are increasing all the time, however, as more high-risk mutations of different types are being recognized. The condition is thus extremely genetically heterogeneous — different mutations are likely to blame across different individuals or families.

Currently, around 25–35% of cases of severe autism, and 5–6% of less severe cases, can be assigned a genetic diagnosis — i.e., a presumably causal mutation can be identified. These numbers are increasing all the time, however, as more high-risk mutations of different types are being recognized. The condition is thus extremely genetically heterogeneous — different mutations are likely to blame across different individuals or families.

Second, mutations conferring high risk are very often new ones that are not actually carried by either parent. Instead, they have arisen in the generation of the sperm or egg that gave rise to the affected individual. This kind of de novo mutation has been implicated in around 60% of severe cases and 35% of cases overall. This explains why so many severe cases are actually sporadic — that is, they occur with no family history and no other relatives affected.

Third, none of the high-risk mutations identified so far are really specific for autism. Instead, they increase risk across a range of psychiatric conditions and can be found in patients with intellectual disability, schizophrenia, epilepsy, or other diagnoses.

Fourth, the effects of high-risk mutations can be strongly influenced by the genetic background of individual carriers. Carrying such a mutation does not mean that a person will definitely develop autism — just that their risk is increased. Whether they actually develop the condition can be affected by the presence of other genetic variants that they carry, by their sex, and also by random variation in how neural development proceeds.

In some individuals, the condition may be caused by two rare mutations, each conferring moderate risk alone. In such cases, one or both of the mutations is likely to be inherited from an unaffected parent. In general, each of us carries hundreds of rare mutations, so there is high potential for this kind of interaction.

We all also carry millions of much more common genetic variants at polymorphic sites across the genome. These are positions in the genomic sequence where two different versions are both quite common in the population. Using genome-wide association studies across very large samples, researchers have found hundreds of such polymorphic sites where one version is at higher frequency in autism cases than controls. The statistical risk associated with carrying any one of these common variants is individually tiny — almost negligible, in fact. But because there are so many of them, their collective impact can be considerable (explaining about 10% of the population variance in risk). The overall burden due to these common variants differs among individuals and can be captured by a polygenic risk score, which tracks a small component of a person’s overall risk of the condition (notably, not enough to be useful for clinical prediction).

This kind of study has also been done for other psychiatric conditions, with one consistent message: The polygenic risk for all these conditions is highly overlapping. Polygenic risk for autism correlates especially strongly with that for ADHD but also overlaps with risk of schizophrenia and many other conditions. So, the familial epidemiology, the analyses of rare, high-risk mutations, and the characterisation of polygenic risk all support the picture of overlapping genetic risk of autism and a wide range of other psychiatric conditions.

An important finding is that risk from rare mutations and from the polygenic burden of common risk variants can combine to push a person over a threshold of total burden. Autistic individuals who carry a very high-risk mutation tend not to have a particularly high polygenic burden, because their risk is already very high. By contrast, those with mutations carrying lower statistical risk by themselves do tend to also have a high polygenic burden. The inference is that if they did not then they would not have developed autism. One way to think about this, biologically, is an increasing polygenic burden generally decreases the collective robustness of neural development processes, making it more difficult to buffer the effects of rare mutations.

“...The actual genes identified also strongly support a neurodevelopmental origin of the condition.”

Finally, the actual genes identified also strongly support a neurodevelopmental origin of the condition. The human genome contains about 20,000 genes, each of which encodes a different protein. Those proteins carry out different kinds of functions in the organism and are often differentially expressed (i.e., actually produced from the encoding DNA) in some tissues or developmental stages, but not others. If we look at the genes that are affected by both rare mutations and common genetic variants associated with autism we see, first, that they tend to be especially highly expressed during fetal development, particularly in neural tissues. And second, the encoded proteins are enriched for those that play roles in neural development, especially in the regulation of gene expression, and in the establishment and dynamic regulation of neuronal connectivity.

Collectively, these observations strongly reinforce the view that autism arises from genetic perturbations affecting the processes of neural development, especially at fetal stages. These perturbations may be effectively buffered in some individuals. But when the overall genetic burden is high enough, they may lead to trajectories of development that ultimately manifest as autism, or as one of a wide range of other possible clinical outcomes. Exactly which outcome results likely reflects some inherent randomness in the way that the processes of development actually play out in a given individual. This means that the residual variance in autism occurrence that is not explained by genetics is not necessarily due to something in the environment — it may instead reflect the inherent variability of the processes of brain development themselves.

The evidence that autism is a genetic condition affecting neural development is extremely strong and consistent. But the underlying genetics is highly complex. It is not just that autism can be caused by mutations in any one of hundreds of individual genes. It’s also that the effects of those mutations are dependent on the genetic background of the individual in hard-to-tease-apart ways. In many cases, especially at the less severely affected end of the spectrum, the genetics may be essentially multi-causal, to the point where no specific single cause can be identified. But at the more severe end, the number of individuals who can be given a genetic diagnosis is increasing all the time, providing families with a much sought after explanation, with growing implications for clinical management.