Across the Spectrum

In scientific research, we often deal with aggregates and averages, trying to get a picture of what a trait, process, or condition typically looks like. But in Thomas Bourgeron’s Human Genetics and Cognitive Functions lab at the Institut Pasteur in Paris, the concept of “typical” just doesn’t apply.

The lab, populated by scientists who span a variety of disciplines, is currently focused on trying to understand one of the most perplexing developmental disorders: autism. Once thought of as a singular disorder, autism is now understood as a wide range of disorders connected by two shared features: difficulty with social interaction and restricted interests. Outside of these core features, individuals with autism spectrum disorder (ASD) diagnoses show incredible variability in many domains: Some may have severe cognitive impairment, while others seem to have extraordinarily high IQs; some may have no language, while others show quite advanced verbal ability. When we talk about autism, Bourgeron explained in his keynote address at the 2017 International Convention of Psychological Science in Vienna, we are really talking about many different autisms.

During his talk, Bourgeron detailed the dimensional approach his interdisciplinary team is taking to better understand this complex spectrum of disorders. Together, the researchers are using various methods — including genetic analyses, brain imaging, mouse models, and even stem-cell applications — to identify the biological pathways that contribute to the phenotypic diversity that characterizes ASD.

Missed Connections

So far, the team’s findings suggest that some important clues can be found in genes that underlie the structure and function of synapses. In one study, for example, Bourgeron’s team examined the genetic profiles of three siblings — one child with ASD, one child with Asperger’s syndrome, and one child with no diagnoses. The genetic profiles of the affected siblings revealed mutations to genes associated with the presynaptic protein neuroligin. While the children’s mother carried the same mutation, her second X chromosome seemed to shield her from any downstream effects.

These results didn’t reveal the gene for autism, Bourgeron emphasized, but they did illuminate a potential pathway. If neuroligin genes are involved in ASD, the team speculated, maybe genes for related proteins are also involved.

Indeed, subsequent work pointed to a link between ASD and mutations on genes that code for SHANK proteins, proteins that serves a scaffolding function at the synapse. They also found evidence of mutations on genes that code for postsynaptic neurexin proteins. These proteins are all essential to synaptic function: Neurexin binds to neuroligin, resulting in a “handshake” that connects two neurons and actually forms the synapse.

The neurexin–neuroligin–SHANK pathway was one of the first genetic pathways to be implicated in ASD, and it opened up a whole field of genetic possibilities. This kind of work — identifying candidate genes — is labor intensive; Bourgeron noted that each of his presentation slides on genetic mutations represented approximately 5 years’ worth of work. But as the field of genomics has burgeoned, so have candidate genes. Some of these genes are known to underlie synaptic function, but others are involved in DNA transcription and translation or in various other processes.

Previously, critics might have argued that genetics research was fruitless in the context of ASD because there “is no gene for autism” — now, Bourgeron said, the criticism is more likely to be that there are too many genes for autism. While two individuals with ASD may share some mutations, it’s equally likely that they don’t share any.

From One to Many

Bourgeron noted that in some individuals, ASD could be monogenic, linking specifically to one gene or even alterations to a single copy of a gene. In one study, the team examined genotypic and phenotypic variation in a pair of siblings — the girl had lost a copy of the SHANK3 gene, while her brother had an extra copy. The girl was severely affected, with virtually no ability for speech; her brother, on the other hand, began speaking at a very young age and had developed a huge vocabulary, though he showed the characteristic difficulties with social interactions and restricted interests that many people with ASD display. These results highlight not only the possible role of SHANK3, but also its apparent dose-dependent nature.

“At the synapse, it looks like very tight gene dosage can [result in] autism or Asperger’s,” Bourgeron explained.

Data from thousands of individuals suggest that SHANK3 mutations may be one of the most robust genetic links to ASD, occurring in about 2% of individuals with ASD and intellectual disability.

In many cases, however, ASD is most certainly not caused by a single gene mutation but emerges instead as the additive effect of mutations to many different genes.

In a study led by graduate student Varun Warrier (University of Cambridge), the research team looked at genome-wide associations with psychological traits related to ASD. Posting a questionnaire on the website for the popular genetics testing company 23andMe, the researchers gathered data from tens of thousands of participants. Drawing from Simon Baron-Cohen’s empathizing–systemizing theory, the team used this massive trove of both genetic and psychological information to examine genetic links to participants’ ability to empathize, thought to be lower in ASD, and their orientation toward systems, thought to be higher in ASD.

Their analyses indicated that about 11% of the variance in participants’ empathizing scores and about 12% of the variance in their systemizing scores could be explained by genetic variation. But there was no evidence that any single nucleotide polymorphisms — differences in a single base pair in a DNA sequence — were correlated with either trait in the genome-wide analyses.

The takeaway from this and other work, Bourgeron said, is that “you can capture part of the variance by looking at the genome, but each gene will really contribute a very small effect.”

Out of Sight, Out of Mind

Identifying the diverse genetic pathways that contribute to a complex spectrum disorder is difficult enough, but Bourgeron and his team are also contending with the reality of the file-drawer problem. Not knowing how many times researchers have tried and failed to find a particular genetic association that has been published in the literature, scientists often end up on a resource-intensive wild goose chase. The consequences are especially problematic given that genome-wide analyses are a major undertaking — researchers must be able to marshal huge sample sizes to be able to detect relatively small effects.

These issues became particularly salient for Bourgeron’s team when they decided to look at brain volume in individuals with ASD. Previous research had suggested that, relative to their peers, individuals with ASD tend to have lower volume in the corpus callosum, the bundle of fibers that connects the left and right hemispheres of the brain. But after collecting brain-volume data from several hundred participants, the research team couldn’t find any evidence of such a difference. When they went back to the literature, they discovered that many of the previous studies had samples that were probably too small to reveal subtle differences, which led Bourgeron and his team to question the robustness of the finding.

In talking to colleagues and other researchers, they found that many had conducted the same investigation and achieved the same null results, never publishing their findings.

“I think it’s really a problem in the field of genetics, and in psychology, that we don’t know what’s going on because people have difficulties [with sharing their] data and [achieving] enough statistical power,” Bourgeron said.

He worries that without data sharing, existing theories that are conceptually appealing may collapse when they undergo further scrutiny. To help combat this problem, members of his lab are developing tools intended to facilitate open science at multiple levels of investigation.

At the genetics level, the lab has developed a tool that identifies all the mutations in a person’s genome and maps those mutations onto a protein–protein interaction network. The tool should help users construct a bigger and clearer picture of the roles that affected genes play in contributing to phenotypic profiles, and can be found here.

Expanding on mouse-model work spearheaded by postdoctoral student Elodie Ey, the lab is also developing a tracking tool that can keep tabs on a “little society” of mice. The tool, which depends on machine learning, enables users to identify and track each mouse so they can not only examine what each individual mouse is doing, but also can monitor how the mice cooperate and act together.

The proliferation of open-source tools and promotion of data sharing ultimately will help researchers achieve a more detailed portrait of a complex constellation of features, Bourgeron said. More software and tools are available on the Human Genetics and Cognitive Functions lab website.

Embracing Chaos

One conceptual knot that Bourgeron hopes to untangle in future research is understanding genetic risk and resilience in ASD.

“Some people are highly sensitive and a small number of rare variants will make the person autistic,” he noted. For others, “the genome is very robust and very resistant and you need a very strong mutation, like SHANK3, to have autism.”

Bourgeron wants to understand how some people seem to be unaffected despite having such strong mutations. Ultimately, his aim is to collect more data on patients and their genomes — but instead of aggregating these data to make group-level comparisons, he wants to focus on a variety of phenotypic dimensions to be able to stratify their analyses even further.

“I think we’re doing a lot of barplots with a little star and we have to think [of the] more dimensional aspects,” Bourgeron said.

In other words, the days of looking for a straightforward causal story are long gone — to make progress, researchers must appreciate the individual variation inherent in ASD.

Quoting the renowned autism researcher Lorna Wing, Bourgeron noted that “if you have seen one child with autism, you have seen one child with autism.”

It’s time to “accept a little chaos and complexity,” he concluded.