As a baby grows inside the womb, its brain does not simply expand like a dehydrated sponge dropped in water. Early brain development is an elaborate procession. Every minute some 250,000 neurons bloom, squirming past one another like so many schoolchildren rushing to their seats at the sound of the bell. Each neuron grows a long root at one end and a crown of branches at the other, linking itself to fellow cells near and far. By the end of the second trimester, neurons in the baby's brain have formed trillions of connections, many of which will not survive into adulthood—the least traveled paths will eventually wither.

Sometimes, the developing brain blunders, resulting in "neuro-developmental disorders," such as autism. But exactly why or how early cellular mistakes cause autism has eluded medical science. Now, Eric Courchesne of the University of California, San Diego, thinks he has linked atypical gene activity to excessive growth in the autistic brain. With the new data, he has started to trace a cascade of genetic and cellular changes that he thinks define autism. Although intrigued by Courchesne's work, other researchers caution that explosive neural growth is not necessarily a defining feature of all autistic brains.

Since 1998 Courchesne has been searching autistic brains for unusual structural features. His studies suggest that while in the womb, the autistic brain sprouts an excess of neurons and continues to balloon during the first five years of life, as all those extra neurons grow larger and form connections. Sometime after age four or five, Courchesne has also found, autistic brains actually start to lose neural connections, faster than typical brains.

In a study published November 2011 in JAMA, The Journal of the American Medical Association, Courchesne reported that children with autism have 67 percent more neurons in their prefrontal cortex (PFC) than typical children. Located in the area of the brain just behind the eyes, the PFC is responsible for what psychologists call "executive functions"—high-level thinking, such as planning ahead, inhibiting impulses and directing attention. In his 2011 study Courchesne sliced up brain tissue from six autistic children and seven typical children who had passed away and counted the number of cell bodies in the sections to estimate the total number of neurons in their PFCs.

Now, Courchesne and his colleagues have analyzed DNA and RNA in 33 cubes of brain tissue from people who passed away, 15 of whom were autistic (nine children and six adults) and 18 who had typical brains (seven children and 11 adults). Looking at the order of DNA's building blocks reveals whether individual genes have mutations; measuring levels of RNA indicates how often those genes were translated into proteins. Such gene expression, Courchesne and his colleagues found, varied between autistic and typical brains. In brain tissue from both autistic children and autistic adults, genes coding for proteins that identify and repair mistakes in DNA were expressed at unusually low levels. Additionally, all autistic brains demonstrated unusual activity levels for genes that determine when neurons grow and die and how newborn neurons migrate during early development. Some genes involved in immune responses, cell-to-cell communication and tissue repair, however, were expressed at unusual levels in adult autistic brains, but not in autistic children's brains. The results appear in the March 22 issue of PLoS Genetics.

By combining his new findings with his earlier discoveries, Courchesne has started to construct a kind of timeline of autism in the brain. Perhaps, as the brain of a future autistic child develops in the womb, something—an inherited mutation or an environmental factor like a virus, toxin or hormone—muffles the expression of genes coding for proteins that usually fix mistakes in sequences of DNA. Errors accumulate. The genetic systems controlling the growth of new neurons go haywire, and brain cells divide much more frequently than usual, accounting for the excess neurons found in the PFC of autistic children. Between birth and age five, the extra neurons in the autistic brain grow physically larger and form more connections than in a typical child's brain. Unused connections are not pruned away as they should be. Later, in adolescence and adulthood, the immune system reacts against the brain's overzealous growth, which might explain the unusual levels of immune genes Courchesne found in his new study and why, in earlier work, he had discovered that when autistic children become teenagers, some brain regions actually start shrinking compared with typical brains.

Not all researchers, however, accept that the patterns of brain growth Courchesne has discovered are relevant to everyone with autism. Nicholas Lange, a biostatistician in the psychiatry department at Harvard Medical School, says that Courchesne analyzed too few samples in his new study to generalize the results to the larger autistic community. Some researchers have surfaced evidence that around 15 percent of autistic children have smaller than usual heads, a condition known as microcephaly, which indicates an abnormally small brain. David Amaral of the University of California, Davis, has previously told reporters that in an unpublished neuroimaging study, he found that only about 11 of 114 autistic children had unusually large brains. Other researchers point out that, in his research with tissue samples from brain banks, Courchesne fails to compare the number of neurons in the cerebral cortex with other parts of the brain—it remains unclear why only the PFC would explode in growth.

But acquiring enough preserved tissue from brain banks to conduct meaningful studies is no easy task—they are incredibly coveted resources, and Courchesne's new study relies on a respectable sample. Looking at gene expression in postmortem brain tissue offers insights into the biology of autism that neuroimaging studies and analysis of DNA and RNA in blood cannot provide because different cell types express different sets of genes. Courchesne's newest findings at least partially echo earlier research by Daniel Geschwind of the University of California, Los Angeles, who also linked autism to unusual activity of genes that control immune responses and how neurons organize themselves in the developing brain. Although Courchesne's concept of autistic brain development is far from flawless or complete, it remains one of the most cohesive theories offered so far—one that suggests the possibility of treatment as well. If scientists definitively link autism to a characteristic sequence of changes in gene expression and unusual neural growth, then it becomes possible to target and reverse any one of the thousands of steps in that sequence.

"Each individual autistic person likely has their own specific profile of dysregulated [sic] genes," Courchesne says, "which means that autism is a very complicated problem. But it's now knowable. We are getting at core knowledge. If we confirm that the starting point is gene activity, we can do something about it, because gene activity can be modified."