MASON ASHER WAS AN easygoing baby. When he didn’t speak at 12 months, his mother, Sara Asher, didn’t give it a second thought. But when he reached age two and he’d said little more than “Mamma” and “Dadda,” Asher enrolled him in a special-education preschool for children with certain delayed behaviors. It was there that a teacher first recognized that Mason, who even as a baby would stare past people rather than look them in the eye, had more than delayed speech. At three, Mason was diagnosed with autism.

The Ashers’ second son, Jackson, was born four years later. Recognizing certain traits, they took him for screening at 18 months and found that Jackson was also autistic. The Ashers had a third son, Asa, in 2011, who started receiving help for his speech by 12 months. Six months later, Asa was diagnosed as autistic.

“It is interesting because no one on either side of the family has had a diagnosis of autism,” Sara Asher says.

According to the Centers for Disease Control and Prevention, one in 68 children in the United States has a disorder on the autism spectrum. Yet despite the condition’s prevalence, the causes of autism in a significant portion of children are still unknown. In the 1940s, a scientist first hypothesized that cold and distant mothering was the culprit. In the 1970s, however, studies of twins and family histories started to point to a strong genetic component. And while it is now thought that both genetic and environmental influences play a part, the scientific community has looked primarily to genetics for clues.

In the early days of that quest, in the late twentieth century, scientists tried to find the single gene, or perhaps a small group of inherited genes, that led to autism. That met with little success, but autism research gained more traction when the “genome era” of the early 2000s provided new tools to explore the body’s library of DNA. More recently still, the scientific community has been amassing genetic material from thousands of autistic children and their families that can be mined for clues.

With a vast amount of data to explore and better technology for sorting through it, researchers in the past decade have been able to ferret out many genes associated with autism. But as they dig, the genetic picture seems to become even more complex. One 2016 study led by UCLA scientists found that siblings with autism seldom share the same genetic variations, and other studies suggest that, in many cases, troubling mutations found in the child are not found in either parent. What has come into focus is a deeply complex picture in which some genetic mutations that lead to autism may be a product of genetic inheritance, and others come from neither parent. Environmental factors such as maternal infection or exposure to toxic chemicals during pregnancy may heighten the risk in ways that aren’t yet understood, and the environment may combine in other ways with genetic factors after birth to cause the condition.

Yet even if scientists remain far from achieving a detailed understanding of autism and its causes, significant progress is being made. “We are starting to provide some answers about what had been a complete enigma,” says Stephen Scherer, director of the Centre for Applied Genomics at The Hospital for Sick Children in Toronto.

fa17_autism_thumb4_protocoverfinal_630x420

IN 1943, LEO KANNER of Johns Hopkins University published a paper describing the unusual behavior of 11 children he had observed. From birth, these kids seemed to show no interest in other people, but they focused obsessively on particular aspects of their environments. Kanner called their condition autism, from the Greek word autos, meaning “self.” The children he observed were creating an isolated self.

Today, autism spectrum disorder (ASD) encompasses a range of neurodevelopmental disorders in which children (or adults) have trouble with social interactions, communication or language, and show rigid, repetitive behavior and interests. But within this broad definition, children with autism can act quite differently. One child may be silent and aloof, obsessively lining up blocks, while others talk endlessly about castles or football scores, failing to realize that they’ve lost their audiences’ interest.

Between 2005 and 2009, three large research projects found that when one identical twin had autism, there was as much as a 95% likelihood that the other twin would have the condition too. Other studies seemed to confirm this view. A child is almost seven times more likely to have an ASD if an older sibling has it.

That evidence seemed to point toward a genetic culprit, with suspect genes inherited from an autistic child’s parents. But efforts to trace that cause to a specific gene or genes found little success.

In general, researchers attempted to work backward, starting with a particular behavior or trait and then seeking to identify genes with mutations that might be involved. This kind of detective work has been effective when applied to some simple traits—for example, the color of an animal’s coat—and in the early years of the 21st century, scientists scanned and compared the genomes of many people with and without autism. The holy grail would have been a single change in DNA—a substitution, deletion or addition of one chemical base in the genetic code, shared among autism patients but not seen in the control. Such a clear, isolated cause might have led to effective therapies or cures.

But this method of gene hunting proved to be of little use for identifying genes implicated in autism. “The approach failed completely on autism and doesn’t show much promise for very complex disorders in general,” says Michael Wigler, a geneticist at Cold Spring Harbor Laboratory in Cold Spring Harbor, N.Y.

WIGLER HAD STUDIED THE genetics of cancer since the 1970s, and he surmised that autism might arise from a bigger genetic event: the deletion or duplication of large sections of DNA. Such events, which can sometimes involve hundreds of thousands or even millions of base pairs of DNA—from among the three billion or so in the human genome—are known as copy number variants, or CNVs. These CNVs can have profound effects on a newborn, such as an imperfectly developed heart; they also have a role in Down syndrome.

Yet while Wigler believed CNVs might cause autism, he didn’t think they were passed down from the parents. “From what I had seen, these kids were so different from ordinary people that the differences couldn’t be the result of transmitted traits,” Wigler says. “They were so radically changed.”

Rather, he suspected that these CNVs arose spontaneously in the sperm, the egg or the combination of the two after fertilization—in other words, in genetic material that was unique to the child. Such spontaneous changes were known as de novo mutations. The CNVs that were involved in Down syndrome arose in this manner. A parent’s age may play a role—children have a greater risk of Down syndrome if they’re the offspring of older parents, and a child born to a 45-year-old father is 3.5 times more likely to have autism than a child born to a 24-year-old father.

Wigler’s theory gave autism researchers a new place to look for answers, and other researchers found that the genomes of autistic children did turn out to have de novo CNVs—large stretches of their genetic code had been deleted or duplicated. Wigler’s own investigation, published in 2007, first showed these results—that people with autism had more spontaneous CNV mutations than were found in the general population. The question that Wigler is investigating now is whether de novo mutations act alone, or in combination with other inherited factors. But there is little argument that the de novo discovery was a major breakthrough.

The early technology for locating de novo mutations was able to detect CNVs because these large structural changes in DNA could be seen more easily than individual genes. By 2010, much more powerful tools had become more affordable and efficient, and could analyze genomic sections as minute as a single base pair of DNA. That development allowed researchers to home in on the 1% of the genome that coded proteins—the exome—which had proved crucial in discerning the genetic roots of other diseases.

Researchers could now compare the exomes of autistic children with those of their parents and pinpoint many mutations that had occurred spontaneously—de novo variations—in the children. In addition to the large CNV mutations they had spotted earlier, they also found de novo changes in single base pairs of DNA—known as single-nucleotide variants (SNVs)—and small de novo insertions or deletions in chromosomes, all of which made the picture even more complex.

De novo SNVs aren’t particularly unusual—typically, a newborn will have about 70 such mutations across the whole genome that weren’t inherited from either parent—but by comparing the exomes of many autistic children, scientists were able to identify the de novo mutations that were consistently associated with the disorder. By 2015, this process had enabled them to link these mutations to 65 genes related to autism.

fa17_autism_thumb2_protofinal2a_630x420

There were other ways de novo mutations played a role. Michael Talkowski, a geneticist at Massachusetts General Hospital, has looked into so-called balanced chromosomal abnormalities (BCAs), which are small inversions, translocations and insertions of genes. In about a third of the children who had these de novo BCA mutations and a developmental disorder, the disorder showed a direct link to one of the genes that had been disrupted by the BCA. Talkowski’s team used this insight to identify a number of new genes related to the disease.

BUT GENETIC INHERITANCE STILL posed a problem. Autism links across generations and within families remained compelling—and many researchers have continued to explore the importance of the genetic diversity that autistic children inherit from their parents.

An alternative theory proposed that many inherited variations, each with a small effect, might, in the right combination, explain how and why children develop the disorder. Genetic variations are measured against the human reference genome. This genome was initially created by the Human Genome Project in 2003 and is now maintained by the Genome Reference Consortium. Since its release, the reference genome has been updated many times and represents the best knowledge about the locations of human genes and their normal variation.

On average, a person has about 3 million places in which the sequence of his or her genetic material differs from the human reference genome sequence. For instance, whereas the reference genome has type O blood, it is within normal variation to have type A, B or AB. Almost all mutations are relatively common, with more than 95% of them found in more than 5% of the population.

Still, some of those common variants may contribute in a small way to the risk of ASD, and if a child inherits a significant number of risky mutations, the combination might cross the threshold for developing the disorder.

In 2014, researchers from several institutes investigated that risk by analyzing genetic data mainly from a Swedish epidemiological sample, also pulling data from a separate Swedish family study, the Autism Genome Project, and another set of genetic information around autism, the Simons Simplex Collection. Their conclusion is that common inherited variations may account for about half of the risk of autism in the population. The researchers also determined that rare inherited gene variations—those found in less than 5% of people—contribute another 3% of autism population risk.

ARE THE GENES RESPONSIBLE for autism inherited or spontaneous? Do they occur in wide swaths of transposed genetic material, or exist as hundreds of tiny mutations across the genome? Current perspectives in autism research reside on their own spectrum. At one end are those who see the problem as multifaceted: that a range of common, inherited variations combine with spontaneous variations that arise from environmental factors, such as toxic exposures or infections during pregnancy. Together, these push a child over the threshold for developing autism. At the opposite extreme, some researchers believe that inheriting a single rare mutation could be sufficient to cause the condition.

fa17_autism_thumb3_protofinal1_630x420

Stephen Scherer says he understands the case for de novo mutations having a role in autism, but he believes that common genetic variations that also play into autism may be more common in the population than we know. “The behavioral attributes of ASD—such as a tendency toward socially isolated behavior—are part of a continuum that we see in every human being,” he says. A child’s unique mutations are what make the difference between what is considered relatively normal and what can be diagnosed as ASD.

Yet despite divergent points of view about what causes autism, most researchers are united on the need for more data, says Scherer, who also serves as research director of MSSNG, the world’s largest autism genome sequencing project. MSSNG, a collaboration between Google and the advocacy and research organization Autism Speaks, is an intentional misspelling of “missing,” with the omitted letters representing the crucial additional information about autism that the research program hopes to deliver. Its goal is to sequence the DNA of more than 10,000 families affected by autism—and families such as the Ashers, in which several children have the condition, may provide key clues for the next phase of research.

Using this approach, in 2017, Scherer and his colleagues have already sequenced the DNA of more than 5,000 people in the families of 2,600 children with autism. Their investigation found 61 genes associated with autism, including 18 that hadn’t previously been linked to the disorder. To add to the complexity of the puzzle, the team found that many of those genes themselves had multiple mutations.

Many other organizations are also seeking to amass genetic information. The Simons Foundation Powering Autism Research for Knowledge initiative is collecting data from 50,000 people with autism and their families, and by late 2017, more than 20 research centers in the United States will have access to the data.

The call for still more data is echoed by autism researchers everywhere. They’re searching for additional information about the extent and intensity of individual autistic traits that could be tied to the traits’ genetic underpinnings. And collecting longitudinal data about families—information from grandparents and even great-grandparents—may help scientists chart the development of particular risk factors and how they might interact with environmental triggers.

Stephan Sanders, a genetics researcher at the University of California, San Francisco, has shown that there could ultimately be as many as 1,000 genes that contribute to autism, and identifying those mutations and understanding their impact won’t happen quickly. Yet much has already been accomplished, especially during the past 10 years, using increasingly powerful technology and novel statistical techniques to untangle some of autism’s daunting complexity. “We have a long way to go,” says Wigler, “but it’s getting clearer how we need to proceed.”