The Autistic Brain: Thinking Across the Spectrum (8 page)

I couldn’t wait to get my hands on my issue of
Nature.
After I got off the flight home, I drove straight to the post office, but the magazine hadn’t arrived. I can’t say I waited by the mailbox for the next few days, but as soon as it did arrive, I tore into it. The article
“The Long-Range Interaction Landscape of Gene Promoters” was of special interest, and I particularly enjoyed the concluding sentence of its abstract: “Our results start to place genes and regulatory elements in three-dimensional context, revealing their functional relationships.”

But after I’d finished devouring that issue of the magazine, I realized that the most important lesson wasn’t in any one of the six Encode articles. It was, instead, in the overall impression that the articles made on me. Taken together, they helped me realize how much we don’t know about genetics.

Like neuroimaging, the science of genetics is still in its infancy. In a hundred years, the state of our knowledge today will look primitive. Ask yourself what would happen if we sent a laptop and a flash drive back in time one hundred years. Would scientists be able to figure out how pictures are stored on a flash drive? Let’s be generous and give them one hundred laptops, so they can do some destructive testing. What these scientists would do is get inside the flash drive and take the chip out. They would map the anatomy of the chip. They would give all the parts highfalutin but silly Latin names. (
Amygdala,
the name of the brain’s emotion center? It’s from the Latin word for “almond,” because that’s what it looks like.
Hippocampus,
the name of the brain’s file finder? It’s from the Latin word for “seahorse,” for the same reason.) And these scientists would assume that all the parts put together are the Intel, because each PC has “Intel Inside” written on it. But these scientists would have absolutely no idea how the flash drive works.

That’s pretty much where we are today with the brain and the genome.

For a scientist, that lack of knowledge is thrilling. A new field to explore! A chance to do fundamental, big-picture research, before the field gets really narrow and specialized! Questions that lead to other questions! What could be more fun?

For a parent waiting for answers about an autistic child
today,
however, the lack of knowledge can be extraordinarily frustrating.

Fortunately, we do have the beginnings of a body of knowledge about the genetics of autism. Even knowing that genetics plays a role in autism is a vast improvement on where we were only a few decades ago. It might be difficult to believe now, but whether DNA had anything to do with autism was open to question as late as 1977, when the first study
of autism in twins was published. The sample size was small, but the results were nonetheless striking. The concordance rate—meaning that both twins share the trait—for infantile autism among pairs of identical twins was 36 percent (four sets of twins out of eleven total). But among ten pairs of fraternal twins, the concordance rate was zero. Both those numbers might seem low, but remember, this was three years before the
DSM-III
provided the first formal diagnostic criteria for autism. By today’s diagnostic standards—our current definition of autism—the concordance rates in that same sample would be 82 percent (nine sets of twins out of eleven) for identical twins and 10 percent (one set out of ten) for fraternal twins. A follow-up study
in 1995, using double the sample size, found a comparable result: 92 percent concordance rate for identical twins, and 10 percent for fraternal twins.

Because identical twins share the same DNA, these results strongly support the idea that the source of autism is genetic. But the influence of DNA is not absolute. If one identical twin has autism, the chance that the other one will have it too is very high. But it’s not 100 percent. Why not?

Well, we could ask the same question about other subtle differences in identical twins. Their parents can always tell them apart, and in some cases the differences are obvious enough that anyone can tell them apart. One reason is that even when the
genotype
—the DNA at conception—is identical in both twins, the genes might work differently inside the cell. The other reason is that the genotypes might not be identical at birth, due to spontaneous mutations in the DNA of one or both of the twins. Both sets of genetic differences contribute to an individual’s
phenotype
—the person’s physical appearance, intellect, and personality.

Knowing that genetics plays a role in autism, of course, is only a start. The next question is, Which gene or genes?

Even into the early years of the twenty-first century, some researchers held out hope that autism might be the result of one or just a handful of gene deviations in an individual’s DNA. Maybe autism was like Down syndrome, which, as researchers discovered in 1959, is directly attributable to an extra copy of chromosome 21—the first time that a copy number variation was recognized as a cause of intellectual disability. In the case of Down syndrome, the relationship between cause and effect is clear: This particular chromosome causes that particular syndrome. Geneticists have had some success in locating specific cause-and-effect genes in autism-related disorders. In Rett syndrome—a disorder of the nervous system that leads to developmental reversals that are often diagnosed as symptoms of autism—the cause is a defect in the gene for a particular protein, MeCP2, located on the X chromosome. In tuberous sclerosis—a genetic disorder that causes tumors to grow and is accompanied by ASD in nearly half of all cases—changes in one of two genes, TSC1 and TSC2, are responsible. Fragile X syndrome—the most common cause of mental retardation in boys, and one that can lead to autism—is due to a change in the FMR1 gene on the X chromosome.

By and large, though, the genetics of autism isn’t that simple. Nowhere near.

After the Human Genome Project and Celera Genomics mapped the human genome in 2001, dozens of institutions in nineteen countries banded together to form the Autism Genome Project, or AGP.
Using a database of 1,400 families, these scientists deployed the gene chip, a new technology that worked at a much higher level of resolution than previous methods and that allowed them to look at thousands of DNA variants on a single chip all at once, rather than on a one-by-one basis. The researchers used this technology to look at each subject’s entire genome—all twenty-three pairs of chromosomes—as well as particular areas that earlier research had pinpointed as possibly being of interest.

When phase one of the Autism Genome Project came to an end, in 2007, the consortium published a paper
in
Nature Genetics
that did identify several specific areas of the genome as likely contributors to autism. Among the more promising avenues for further research is a mutation in the gene that codes for a protein called neurexin, which links directly with a protein called neuroligin to control how two brain cells connect across the synapse between them. During development, these interactions are crucial for directing neurons to their proper targets and for forming signaling pathways in the brain. This finding by the AGP reinforced earlier research indicating that mutations in the SHANK3 protein, which interacts with neuroligin protein at the synapse, are associated with an increased risk of ASD and mental retardation.

But in addition to serving as a direction for further research, the paper demonstrated the effectiveness of the strategy that AGP scientists had used to detect these mutations. They searched for copy number variations, or CNVs—submicroscopic duplications, deletions, or rearrangements of sections of DNA. These variations, which can vary in length and position on the chromosome, have the potential to disrupt gene function.

Where do these copy number variations come from? Most are inherited. At some point, an irregularity entered the gene pool, and it was passed down through the generations. But some CNVs aren’t hereditary. They arise spontaneously, either in the egg or sperm before fertilization or in the fertilized egg shortly afterward. These are called de novo mutations, from the Latin words for “from the beginning.”

Many CNVs are benign. And geneticists estimate that each genome—each person’s unique DNA—might contain as many as several dozen de novo mutations. They’re part of what makes each person unique. But might de novo CNVs be associated with autism?

This is the question that a 2007 study
of 264 families, published in
Science,
set out to answer. The authors concluded that such mutations do pose “a more significant risk factor for ASD than previously recognized.” The study found that 10 percent of autistic children with nonautistic siblings (12 out of 118) had de novo copy number variations, but only 1 percent of controls who had no history of autism (2 out of 196) showed CNVs. In the following five years, this paper, “Strong Association of De Novo Copy Number Mutations with Autism,” would be cited more than 1,200 times.

The hope that autism could be traced to one or even a few gene variations became less and less realistic. By the time phase two of the Autism Genome Project—drawing on the DNA of 996 elementary-school-age children in the United States and Canada diagnosed with ASD, their parents, and 1,287 controls—came to an end, in 2010,
the collaborators had identified dozens of copy number variants potentially associated with ASD. By 2012, geneticists had associated ASD with hundreds of copy number variations.

Further complicating the research was that many of the CNVs seemed to be, if not unique, at least extremely rare. The authors of the 2007
Science
paper seeking to link de novo mutations with autism had noted: “None of the genomic variants we detected were observed more than twice in our sample, and most were seen but once.” In 2010, upon the publication of the Autism Genome Project’s phase-two research, UCLA professor of human genetics and psychiatry Stanley Nelson said,
“We found many more disrupted genes in the autistic children than in the control group. But here’s where it gets tricky—every child showed a different disturbance in a different gene.” In September 2012, an article
in
Science,
“The Emerging Biology of Autism Spectrum Disorders,” recounted the stunning progress in the discovery of possible autism-related CNVs—but “with no single locus accounting for more than 1 percent of cases.”

Geneticists sometimes speak of a many-to-one relationship: many candidate mutations, one outcome. But what outcome, specifically? A diagnosis of autism? A symptom of autism? As is the case in neuroimaging, trying to understand autism through genetics is complicated by its heterogeneity. Autism manifests itself in numerous traits, and those traits are not identical from individual to individual. Why should we expect that the genetics of autism would provide a one-to-one correspondence between mutation and diagnosis?

In fact, researchers are finding that some mutations can contribute to a range of diagnoses, including intellectual disability, epilepsy, ADHD, schizophrenia—a one-to-many relationship. Again, heterogeneity is the problem, because the diagnosis of autism is based on behaviors, and autism shares those behaviors with other diagnoses. If researchers knew which traits—if any—were specific to autism, the search for a genetic cause might be a lot easier. As G. Bradley Schaefer,
a neurogeneticist at the Arkansas Children’s Hospital Research Institute, says, “The key is trying to figure out which differences are secondary versus which differences are salient to the condition.”

Until they figure that out, researchers have to adopt other methodologies to pinpoint autism-related genes. The Autism Genome Project, for instance, looked for a pattern among the mutations, or at least the beginning of a pattern. And the researchers found it: Many of the genes belonged to categories known to affect cell proliferation and cell signaling in the brain—a pattern that further reinforced the previous findings about the significance of the neurexin-neuroligin linkage and SHANK3.

In 2012, three groups of researchers that had independently devised an identical new approach to discovering de novo mutations published their complementary findings in an issue of
Nature.
Their strategy was to include only autistic subjects whose parents and siblings exhibited no autistic behaviors. They then used letter-by-letter sequencing of the exome—the protein-coding parts of the genome—to identify de novo single-letter mutations. If they found a de novo CNV in at least two of their autistic subjects, and if that CNV did not appear in any of the nonautistic subjects, then they considered that mutation a contributing agent to autism.

One of those studies,
led by Matthew W. State, a neurogeneticist at the Yale University School of Medicine’s Child Study Center, sampled two hundred autistic children and their nonautistic parents and siblings and found two children with the same de novo mutation, one that none of the nonautistic participants showed. At the same time, another study,
led by Evan E. Eichler at the University of Washington in Seattle, independently sampled 209 families and found a subject with the same de novo mutation as a subject in the Yale study. Again, it was one that neither study had found in their nonautistic subjects. The University of Washington study also identified another de novo CNV in two autistic participants in its own study. Then a third study,
led by Mark J. Daly at Harvard, looked for those three de novo variations—the one from State’s study, the one from Eichler’s study, and the one the two studies shared—in a separate sample of subjects and identified children with autism who had the same CNVs, indicating a possible correlation between that CNV and autism.

Another finding from that same trio of studies is worth noting—CNVs were four times more likely to originate on the father’s side than on the mother’s. This finding received reinforcement a few months later with the publication of a paper
in
Nature
that reported a correlation between a father’s age and the rate of de novo mutations. For me, that paper was one of those “Of course!” slap-yourself-on-the-forehead moments. Sperm cells divide every fifteen days, more or less, so the older a father is, the greater the number of mutations in his sperm. It’s like making a copy of a copy of a copy on a photocopier. And the greater the number of mutations, the higher the risk of a mutation that might contribute to autism.
³