News The latest developments in autism research.

Web of genes may hold clues for autism treatments

by  /  20 October 2014
Julia Yellow

Five years ago, scientists saw an exciting pattern emerge from sequencing studies of people with autism: Many of the mutations they picked up affect the function of synapses, the junctions between neurons.

There are trillions of synapses in the human brain, linking billions of neurons. Still, calling autism a ‘disorder of the synapse’ gave it encouraging specificity. More data, they hoped, would narrow this spotlight to pinpoint a handful of pathways that could be targets for autism drugs.

Instead, the spotlight has become a floodlight.

In studies over the past few years, most of the genes that have emerged as the strongest autism candidates have turned out to be regulators — meaning that they regulate the expression of hundreds, if not thousands, of other genes. What’s more, some of the target genes are themselves regulators, and may even loop back to influence the candidate gene.

“If autism has taught us anything, it’s that however complex you think it might be, it’s actually more complex,” says Stephan Sanders, assistant professor of psychiatry at the University of California, San Francisco School of Medicine.

That autism involves so many regulators shouldn’t come as a surprise. The defective gene in fragile X syndrome, which often leads to autism, controls more than 800 other genes, including about 90 autism candidates1. And MeCP2, the gene mutated in Rett syndrome, another autism-related disorder, regulates thousands of genes.

Many researchers see this complexity as an opportunity: One way forward, they say, is to home in on a few specific pathways that repeatedly turn up among the candidates.

“Instead of saying we’re going to follow one gene that really matters and see what it does, we need to say that each gene is like a little star that points in many directions,” says Sanders. “We need to use these multiple different genes and look for the process, time and place where they all point the same way.”

Building webs:

The closest thing to an ‘autism gene’ so far is CHD8, which emerged as candidate gene just two years ago2. CHD8 regulates gene expression by binding to DNA and changing its structure. In stem cells that give rise to neurons, CHD8 binds nearly 7,000 genes, according to a report earlier this month in the Proceedings of the National Academy of Sciences3.

Many of CHD8’s targets are themselves autism candidate genes, including FOXP1, DYRK1A and ADNP. The lists of autism genes also frequently overlap with the targets of FMRP, the protein affected in fragile X syndrome4. Researchers say that the more candidate genes they uncover, the more they find that their targets may converge at key points.

“If autism has taught us anything, it’s that however complex you think it might be, it’s actually more complex.”

“There’s going to be far fewer pathways than there are genes that have been implicated,” says Michael Talkowski, assistant professor of neurology at Harvard Medical School and lead researcher on the new study. “I do think they will converge at a number of final points — not just one, but enough that we’ll have targets we can try to manipulate.”

Aiming to find these points of convergence, Talkowski and others are mapping the targets of other regulators implicated in autism, such as AUTS2, FOXP1 and TBR1. Each of these genes has been found to carry rare, harmful mutations in people with autism. But researchers will probably need to identify common variants, found throughout the population, to be able to find points of convergence. “We need to layer different types of data,” says Sanders. “Then you look for convergence, and that convergence leads you toward the true etiology of autism, which leads to a true therapy.”

Some statistical tools take the different types of mutations into account when ranking autism candidates. For example, an algorithm called TADA considers both common and rare variants in a gene when rating its significance for autism5. TADA’s next iteration, DAWN, will include information on whether a gene responds to an autism-linked regulator such as CHD8, says Bernie Devlin, professor of psychiatry at the University of Pittsburgh, who developed both statistical models.

DAWN also takes into account when and where a gene is expressed during development6. These last factors acknowledge the fact that autism is a developmental disorder that likely begins in utero.

“If you’re interested in a specific gene, there may be a way to begin to narrow in on where and when you might want to look, in order to understand how that mutation may be contributing to autism,” says Matthew State, professor and chair of psychiatry at the University of California, San Francisco.

Time and place:

Last year, State’s team used an atlas of the developing brain to find genes expressed at the same time and in the same place as nine autism candidate genes, including CHD87. The networks they uncovered include many known autism candidates, and point to the prefrontal cortex during mid-fetal development as one birthplace for autism.

“All this heterogeneity is actually an advantage, because you can use different methods to understand the systems underlying it,” says Jeremy Willsey, a postdoctoral scholar in State’s laboratory. “But we should do it in such a way that tells us something about the particular point in development and the particular region of the brain that’s involved.”

In his new study, Talkowski’s team found that CHD8’s targets tend to fall into the same networks that State’s team found.

These efforts all center on finding one, or a few, converging pathways among autism genes. Another hypothesis holds that each autism symptom has a separate genetic origin. If that’s true, researchers may be able to pare down the number of pathways involved by focusing individual symptoms.

For example, Smith-Magenis syndrome is a monogenic disorder characterized by intellectual disability and sleep problems. Another disorder, called brachydactyly mental retardation syndrome (BDMR), is often confused with Smith-Magenis syndrome but stems from a different gene. Sarah Elsea’s team at Baylor College of Medicine in Houston, Texas, discovered in 2010 that the gene mutated in BDMR regulates the Smith Magenis syndrome gene8.

Elsea is using the same approach to look at people who have autism, fragile X syndrome, Smith-Magenis syndrome or 2q23.1 deletion syndrome, all of which share problems with sleep and behavior9.

“Our hope is that we can find common pathways that are dysregulated in multiple disorders, which could then lead us to a common therapeutic intervention that might be able to alleviate some of their shared symptoms,” says Elsea, associate professor of genetics at Baylor.

Far from being discouraged by autism’s complexity, Elsea and others are thinking of creative ways to harness its diversity and find answers.

“There are two levels of complexity we see in autism: First, there are large numbers of genes. Second, each gene does many, many different things,” says Sanders. “Each one of those on its own is a disaster moving forward, but actually the combination might make this easier [to solve] than other disorders.”

Correction: This article was modified from the original. Matthew State is chair of psychiatry at the University of California, San Francisco, not professor of genetics at Yale University as the original version stated.


1: Ascano M. Jr. et al. Nature 492, 382-386 (2012) PubMed

2: Bernier R. et al. Cell 158, 263-276 (2014) PubMed

3: Sugathan A. et al. Proc. Natl. Acad. Sci. USA Epub ahead of print (2014) PubMed

4: Iossifov I. et al. Neuron 74, 285-299 (2012) PubMed

5: He X. et al. PLoS Genet. 9, e1003671 (2013) PubMed

6: Lui L. et al. Mol. Autism 5, 22 (2014) PubMed

7: Willsey A.J. et al. Cell 155, 997-1007 (2013) PubMed

8: Williams S.R. et al. Am. J. Hum. Genet. 87, 219-228 (2010) PubMed

9: Mullegama S.V. et al. Eur. J. Hum. Genet. Epub ahead of print (2014) PubMed

6 responses to “Web of genes may hold clues for autism treatments”

  1. IDorASD? says:

    Some of these network studies might be missing the obvious. They seek to identify convergence among particular candidate genes in individual studies. They go in with the presupposition that these genes are “autism genes” and that there exist “specific pathways.”

    However the so-called convergence among the genes discussed herein can be largely attributed to a loss of phenotypic robustness in broad brain development related pathways, as discussed here:
    and shown nicely in this figure:
    This would lead to general “bad brain development” and the observed spectrum of neurological/neuropsychiatric disorders including autism, intellectual disability, epilepsy, and schizophrenia which share mutations in many of the genes discussed in this article.

    Many papers point out that rare de novo variant carrying “autism” genes are also low-IQ or intellectual disability genes:
    and we already know that Fragile X is related to low IQ, so it is unsurprising that mutations in genes regulated by it during early development cause intellectual disability. Even for the very “strong” autism candidate CHD8, 60% of individuals have low IQ:

    Finally, it has been clearly shown that among individuals with IQ > 100 (“average” IQ), the rate of the “highly pathogenic” and rare de novo mutations focused on by many network studies discussed here is not above chance:
    this means that the rare mutation risk in the population resides almost exclusively in those individuals with low IQ and autism, not just autism.

    So in general, I am unsure what all of the talk of “autism” and “specific pathways” above is all about. It seems like people are just finding many ways to link many genes to each other and autism, which could just as well be substituted with intellectual disability in this discussion.

    • Stephan Sanders says:

      There is clearly overlap between autism spectrum disorder (ASD) and intellectual disability (ID). However, while a specific mutation may be associated with ASD, ID, and schizophrenia this association is not equally distributed (as would be expected under the ‘bad brain’ hypothesis. For example 15q11.2 maternal duplications are more common in ASD than in ID and schizophrenia, while 22q11.2 deletions are observed more frequently in ID/schizophrenia than in ASD. This review discusses this issue further:

      The key point of considering convergence, as discussed in the article above, is that by focusing on the many genes found in children with ASD, the etiology of ASD should become apparent. Similarly, by focusing on the many genes found in children with ID, the etiology of ID should become apparent. While the etiology of these two conditions may be similar, since there are children with only ID or only ASD, it follows that there must also be some etiological differences.

  2. ASDinID says:

    There is no doubt that a subset of mutations will cause specific diseases and others will not. The work discussed in this article focuses on genes related to rare loss of function events in proteins. Analysis, enrichment, and networks based on these do not seem to home in on biology specific to ASD, as the effect attributed to ASD over controls in the population resides in a subpopulation that is also afflicted by ID.

    This is very clear from looking at the data in the appropriate statistical framework:
    “We found that the rate of de novo loss-of-function mutation in ASD cases with a measured IQ above average was no different than the expectation (IQ ≥ 100; n = 229; 0.08 de novo loss-of-function mutations per exome in comparison to the expectation of 0.09; P = 0.59). By contrast, the rate in the rest of the sample was substantially higher than the expectation (n = 572; rate of 0.17 de novo loss-of-function mutations per exome; P = 1.17 × 10−10). Furthermore, when directly compared (rather than being compared to our expectation), these two groups were significantly different from each other (P < 0.001), confirming a difference in genetic architecture among ASDs as a function of IQ." Regarding the specificity of a particular rare event, the same mutations often cause variable phenotypes. This complexity is commonly attributed to epistatic effects, genetic background, and/or the environment. However, disruption of phenotypic robustness during brain development explains the data just as well if not more succinctly. It is noteworthy that some events do confer consistently larger or smaller head size, and some are associated with dysmorphias. This may be considered some level of specificity, but, like low IQ, it is also consistent with a loss of phenotypic robustness. In a sense individuals with a loss of phenotypic robustness may be a real subset of ASD, but that doesn't mean focusing on their biological changes will lead to specific pathways that cause ASD. It is just an alternative interpretation that seems to fit. Networks and pathway analysis are certainly valuable, but it seems like ASD specificity may reside with common genetic variation. Heritability studies already suggest that most of the liability is in common variation. So once GWAS are done with sufficient sample size, true ASD specificity may emerge.

  3. Autism Mother says:

    It feels like I have seen this movie before…..”New genetic discoveries will lead to exciting treatments!” Sorry, but as a mother of an autistic 15 yr year old these promises feel stale. The vast, vast majority of ASD people do not have these rare chromosomal abnormalities and the actual translational of these types of genetic research has fallen fall short of expectations. Parents would like to see greater emphasis on nuts and bolt here and now biology, such as treating immune based dysfunction.

  4. Autism Mom says:

    I have three daughters, 21, 23 and 24, the middle having autism. I would very much appreciate one of the professionals commenting on this website to tell me if the findings above indicate we are any closer to including “autism” genes in an amniocentesis finding. My major concern at this point is for my two typically-developed daughters who have a legitimate concern regarding having children with autism. They know first-hand how the condition affects a family. I would very much appreciate an informed opinion. Thank you.

    • wendychung says:

      The findings will improve the ability to diagnose some but not all cases of autism prenatally. Prospective parents will have to make individual decisions about whether or not to pursue that type of testing prenatally.

Leave a Reply

Your email address will not be published. Required fields are marked *