The Hunt for Autism Genes:

But so far it hasn't panned out.

For gene-hunters like Cook and his colleagues, brilliant hypotheses that don't work are more the rule than the exception. So it is impossible to exaggerate the excitement generated among autism researchers when, in just the course of the past couple of years, several leads finally began to pay off.

And thus far Ed Cook, of the University of Chicago, along with department chair Bennett Leventhal; Cathy Lord, co-creator of the Autism Diagnostic Interview;Valerie Lindgren, the team's cytogeneticist; and Courchesne and his group are central figures in the tale. Cook and his colleagues are the first to report a gene-not just a region on a chromosome, but a single gene-related to autism. This is the famous (or infamous, considering how slippery the thing has been so far) serotonin transporter gene, located on chromosome 17. Now, with his new findings concerning a connection between chromosome 15 and autism, Cook and his colleagues have published another paper that has placed the field of autism research squarely within one of the hottest areas of genetics research today.

Genetics 101: A brief review of genetic principles.

Autism is now known to be one of the most genetic of all genetic disorders. To many of us this news has come as a shock; in my own family's case my husband and I were told, in 1991, that although "statistically" our chance of having a second child with autism was approximately 3%, in actual practice, none of the genetic counselors on staff at the major urban hospital handling our case had ever seen a family with more than one child with autism. Needless to say, we weren't happy when, just a couple of years later, we started to hear that autism was not only "genetic" but that it was among the most strongly genetic disorders known to the scientific community. Today, depending upon whom you ask, parents are told that the chance of having a second child with autism varies anywhere from 5 to 10%-with a roughly 25% chance of having children with related problems. In the meantime we had gone on to have twins, one of whom-surprise-is indeed autistic. We parents need the science to be moving a great deal faster than it is.

Still, these percentages are not quite as bad as they sound. To begin with, Susan Folstein, the first researcher to study identi-cal twins with autism, believes that the 10% figure will turn out to be wrong; her work tells her that the correct figure will probably be 7 to 8% (Cook uses a range of 5 to 9%). Still too high, but not 10%. And of course this also means that a family's chance of the next child not being autistic is 90 to 95%; the odds are in parents' favor that it won't happen again.

Perhaps more importantly, the 25% figure for related problems does not mean that 25% of a family's non-autistic children will have severe and life-altering problems. Some of these kids will have problems with spelling, for instance, and that will be the end of it. Some of them will be poor readers who will improve with age, as almost all dyslexics do with proper intervention. Only about 15% will show social reticence, and this won't be social reticence at the level of HFA or Asperger's. And some of these "differences" in the siblings of children with autism will actually be beneficial.

In Folstein's words, "I think that some of these various traits are valuable traits-or at least they are not in themselves bad things. A case of autism results when you get all of them together. They don't just add up; instead, whatever negative effects they have multiply. "

In other words, the real problem for parents of a child with autism is the 7 to 8% chance of having a second child with autism- --not the 25% chance of having a child who can't spell, or who isn't the life of the party.

One last note. As parents, you will continue to see the 3% figure cited as the "prevalence rate" in siblings. What this means is that when researchers go out and simply count how many families have more than one autistic child, they find that approximately 3% (some say 3 to 5%) of families with one child with autism also have another child with autism. However, from the perspective of family planning, the 3 to 5% range underestimates the danger. The true risk of having a second child with autism should be the "recurrence rate," which is 7 to 8% (or 5 to 9%, depending upon the source). The reason there are fewer "multiplex" families than we would expect is that families who have a seriously handicapped young child-any handicap, not just autism-frequently stop having children. Researchers call this phenomenon "stoppage," and it obviously lowers the number of autistic siblings in families. If you stop having children after having a child with autism-or choose to have fewer than you would have otherwise- you limit the population. Sadly, the real risk of having a second child with autism is the higher figure.

We know that autism is highly genetic through studies of identical twins in which the index twin has been diagnosed with the disorder. Recent studies, such as that of Bailey, et. al., (1995), have found that the odds of a "co-twin" also having autism are as high as 73%; that is, if the index twin is autistic, there is a 73% chance that his or her co-twin will be, too. This is the "concordance" rate.

Schizophrenia, by contrast, which is understood by all to be a genetic disorder, has a 46% concordance rate. And diabetes, another complex disease passed down through generations, has only a 30% to 50% concordance rate, with a risk to siblings of 6%. So 73% clinches it, particularly when you compare this figure to the rate in the studies of fraternal, or non-identical, twins which is extremely low. (In both the original group of fraternal twins, studied by Folstein, et. al., and the group recruited later by Bailey, et. al., none of the fraternal twins were concordant for autism.)

But as it turns out, the 73% concordance rate for autism is not the whole story. When researchers went back to revisit the original twins from the very first twin study, who were now grown, they discovered that the typical twins were no longer quite so typical. Most of them, in one researcher's words, had "something genetic" going on. They were not autistic; they would not even qualify for the label "high-functioning autistic" or "Asperger's." But only 1 of 7 had married and was living independently, and just 3 had managed to achieve full-time employment.

Looking at these grown non-autistic twins of index twins with autism, the researchers concluded that the concordance rate for this group was 60% for a diagnosis of autism-but 92% for "a broader spectrum of related cognitive or social abnormalities."

Thus, autism may well be more strongly influenced by genetic makeup than schizophrenia, possibly even more strongly genetic than manic depression, which was previously thought to be the most powerfully genetic of all the mental illnesses.

This much makes sense, but for parents confronting such data the immediate question is: if our child inherited his autism from us, why aren't we autistic?

The simple answer is that autism is a "complex disorder"; in many or most cases it is "oligogenic," meaning that it takes more than one gene to develop the disorder. Say autism requires a precise combination of 5 genes: if a husband has three of these genes and his wife has two, their child can be autistic while they themselves are not. Moreover, some of the 5 genes may be dominant, others recessive-which would mean that for the recessive gene the child would have to inherit two copies, one from each parent.

Alternatively, either husband or wife-or both-could have some symptoms of autism, but not enough to actually be autistic. One spouse might have had a language delay as a child, one spouse might be obsessive, one might be anxious or depressive-any or all of these traits might signal the presence of a gene or genes for autism. A student of Cook's recently constructed a Venn diagram with five intersecting circles, each circle representing a gene for autism. She showed that at the point where all five circles intersect you get autism, but at the point where only 3 circles intersect you get something else, at the point where two circles intersect you see a different symptom, and so on.

But Cook says what is most interesting to him is that you can also show that it would be possible to have four of the intersecting circles and yet show no symptoms at all. This is a particularly intriguing possibility in terms of what it would take to treat or to cure some cases of autism: if it is possible to have four autism genes with no symptoms, theoretically all you would need to treat is the one gene that "tips" the person into the disorder. A child or adult with autism, after having just this one gene remediated, could-with proper education and behavioral support-climb out of his autism even though four of his autism genes are still fully functional.

Which brings us to another critical point concerning "bad" genes and the mischief they work in human beings: the very same gene can have variable "expression" from one person to another. A really bad gene, for a progressive, wasting disorder, say, might send one child to an early death, while leaving another child only minimally affected.

This is why you can have such incredible variation in identical twins, who share the same genes. Identical twins are essentially clones, and yet one twin can suffer from a terrible genetic disorder while the other does not (although this is rare). Cook cites the even more startling case of identical octuplet mice who have had a critical gene "knocked out" or removed from their genotype. When you have eight genetically identical baby mice, they are all phenotypically different-they come out looking different. Cook recalls a set of identical octuplets who had a genetic mutation that should have caused the mice to be born with malformed ears. Some of the mice had no ears, but others had just one ear missing, on one side; still others had one ear missing on the other side; and another one or two had some deformities to the ear, but only very mild. "What this tells us," says Cook, "is that there's something that's not genetic about even very simple developmental biology-it could be as simple as where the mouse was carried in the womb, but we often don't know. This is an example where genetics may provide focused approaches to studying environmental effects."

In fact, data on non-concordant identical twins may one day tell us how to use the environment to prevent autism in the first place. Most of us, when we hear the word "prevention," think "abortion"-but abortion is a drastic and tragic means of preventing a genetic disorder. Down the line we may be able to use the environment to protect children with autism genes in the same way the environment may have protected Folstein's non-concordant twins. Remember: these are children who carry all of the genes for autism-who are genetically identical to their co-twins who do have autism-and yet they themselves are not autistic. When we know why, we may be able to use this knowledge.

Getting back to parents: a parent could have all kinds of autism genes, and yet have been lucky enough that they were not expressed in him or her. Instead, the bad luck hits his child. Even more intriguing: a parent might carry autism genes that have actually benefited him in his own life-given him special skills or talents he would not have had otherwise. Cook himself pointed out, after reading this paragraph, that he wouldn't have been able to spend his weekend finishing the team's latest updated analysis without being willing to give up social contact and focus on the details of his work "in a somewhat repetitive manner. " This is easily the kind of useful and productive ability that could come from an autism gene that doesn't result in autism.

Last but not least, the parents could have few to none of the autism genes themselves, and yet still end up with an autistic child because of random mutations that take place during the complicated and sometimes perilous process of recombination that unites the mother's DNA with the father's. (And of course the reality of environmental pollutants and toxins adds another dimension to the story: a child can become autistic because of a spontaneous mutation in his parents' genes caused by a virus or a toxin.)

Why We Need to Find the Genes

Most of us were taught, in high school or college, that genes are the blueprint for the human organism. Every middle-aged college graduate remembers the drill for eye color: you inherit a gene for blue eyes from your father, a gene for brown eyes from your mother (or vice versa) and bang: you've got brown eyes, end of story. In fact, eye color is more complicated than that, but this is what our college professors described to us as the one-gene-one-trait law of "Mendelian" genetics. In the popular view of genetics, a baby is conceived, its genes make it who it will be, and that's it.

As a result, most of us don't immediately see what is to be gained by discovering the genes for autism other than a prenatal test like the one for Down syndrome. What good does it do, for that individual child, to find out that a gene located on chromosome 15 may have caused him or her to be born autistic?

The answer is that finding the genes for autism may well mean finding treatments or cures for autism.

This is why. Broadly speaking-very broadly-there are two different classes of genes: "developmental" genes, and "operating system" genes. Developmental genes are the genes that guide the baby's development in the womb (though some are active throughout life). The genes that determine eye color are developmental genes. Developmental genes turn on, do what they were designed to do, then turn off and are not heard from again (except, interestingly, in cases of cancer which some re-searchers believe result from old developmental genes accidentally becoming active again and causing out-of-control cell growth).

Operating-system genes are different: operating system genes are the genes that are operating all day long; they are the genes that underlie and make possible everything we do. My own "operating system" genes make it possible for me to write this page; your operating-system genes make it possible for you to read this page. Other "system" genes make it possible for our lungs to breathe air, our hearts to pump blood, and our muscles to maintain a seated position in the meantime. Everything we do in life is "run" by genes.

This is where the possibility for treatment comes in. But first: another basic principle. Each gene "codes for"-or creates-one or more proteins. These proteins then go out to do their job: they might foster a chemical reaction in the brain or gut or heart or anywhere in the body; they might serve as receptors to allow one cell to receive a message from another cell; they might turn on another gene, which will produce another protein. But whatever they do in the body, proteins have to be shaped correctly in order to work. Just one tiny flaw in the gene can result in a fatal flaw in the structure of the protein. (It doesn't have to; a gene can undergo mutations that are entirely harmless. But some mutations are deadly. ) Babies born with PKU, for instance, are missing one enzyme, the enzyme that metabolizes phenylalanine, an ordinary amino acid found in food. Enzymes are made of proteins, and for PKU babies, that one missing enzyme, due to one mutated gene, causes profound mental retardation unless identified early so that dietary changes can prevent this fate. (For the sake of accuracy I should mention that PKU can be caused by either one of two different genes. But both genes cause the disorder alone, neither requiring the presence of the other.)

Or-and this has been covered extensively in the press-they might take the route of creating a gene therapy that would replace the bad gene with a new, good one. However, what is not often reported is the fact that gene therapy is currently the least attractive of these options, because it is the most complex-needlessly complex, in the view of many. Because genes not only produce proteins, but can also be turned off or on or slowed down or speeded up by proteins, most biotechnology companies are instead trying to create medications that will, like proteins, modify the gene's action. As Ed Cook says, "Gene therapy is something you turn to when you don't think you'll find an orally administered, more typical medication. "

In all likelihood, autism will involve both mutated operating system genes and mutated developmental genes. Patty Rodier of the University of Rochester is looking into a connection between the Hox genes and autism (her work is covered in the Summer 1997 issue of NAARRATIVE. ) A mutation in a developmental gene is more worrisome, because the developmental genes guide the creation of the brain in the first place, determining its structure. When a developmental gene is damaged, the brain ends up misformed -essentially, the baby is born with a birth defect of the brain. Many of the "thalidomide" babies of the 1960s were born not only with birth defects of the limbs and ears, but also with birth defects in the structure of their brains that caused them to have autism. This is not reason to lose hope, however, because structural differences in the brain can be, and have been, treated with medication in other brain disorders like schizophrenia and Parkinson's disease. More on this later.

The Serotonin Gene

The serotonin transporter gene has been a puzzler. Cook and his team looked at genes controlling serotonin in the first place because one of the most robust findings in the biochemistry of autism has been that approximately one quarter to one third of people with autism show abnormally high levels of serotonin in the blood. And sure enough, Cook and his team found, in three separate studies, a statistically significant association between autism and a shortened version of the promoter of the serotonin transporter gene, HTT.

However, while it was no surprise to find a serotonin gene involved in autism, it did surprise everyone involved that the short form of HTT turned up in all three studies. In simple terms, the "transporter" portion of the gene transports serotonin inside blood cells-and the long form is better at doing this than the short. Thus if people with autism have more serotonin inside their blood cells than average, which they do, you would expect that people with autism would also have higher levels of the long transporter than typical people. But this is not what Cook's three studies found.

The precise relationship between serotonin in the blood and serotonin in the brain is complicated, of course, but basically blood cells are analogous to brain cells-which means that the long form of the transporter would lead to more serotonin inside the brain cells, and less serotonin outside the brain cells. Generally speaking (and again this is a simplification) we want good levels of serotonin outside our brain cells where it is free to work its magic. All of the "SSRIs" (selective serotonin reuptake inhibitors), -Prozac, Paxil, Zoloft and Luvox-are thought to work by increasing the level of serotonin in the spaces, or synapses, between brain cells.

HTT was the first susceptibility gene for autism found using appropriate family-based controls, and it was big news. Without consulting Cook, the University of Chicago dispatched a press release to EurekaNet asking them to post it on May 1, the day of the paper's release. Just days before the release was to be posted, word reached Cook that Fritz Poutska, Annemarie Poutska, and K. Peter Lesch's group in Germany had found no evidence either way, for short or long form being associated with autism. Cook contacted Eureka at once, but it was too late; for some reason the service did not have a provision for altering an announcement just before it was scheduled to be posted, and thus all the world was given the impression that Cook and his team considered their finding to be absolute. Cook has been trying to explain the provisional nature of published research to parents and journalists ever since, causing his wife, in their Christmas letter, to call him "the boy who cried gene." (Though Cook fondly points out to his wife that she said the same thing about a gene for ADHD he and his colleagues identified a few years back which has now been replicated twice, albeit after a two-year delay . . . )

Naturally, these conflicting reports have led to confusion among parents trying to follow the science: is the HTT gene involved in autism or not? The answer, for the time being, is "maybe." Eric Courchesne speculates about where these findings may lead down the road:

"We're uncertain whether there is an association between autism and the short variant, or whether the short variant is a signpost that there is something somewhere else on that gene that is the real problem. Both groups are wondering whether these two findings may be telling us that it's not long vs. short that matters; maybe it's not the promoter region, but somewhere else in the gene that we should be looking. "

In other words, for parents, clinicians, and researchers alike the message is: stay tuned.

In the meantime, it is possible to draw some useful, although tentative, conclusions from work on HTT. To begin with, the HTT gene is probably normal; it is not a mutation. This means that it does not cause autism in and of itself, but may instead amplify the effects of mutations that do cause the disorder. The good news is that since the HTT gene is a normal variant (also called an "allele") we can use data collected from non-autistic people to think about people with autism.

In the normal population the short form is extremely common: over 60% of the general population carries at least one shortened form; 16% carry two. Furthermore, generally speaking, the short form is dominant over the long form: if you have one short and one long, behaviorally you'll act more "short" than "long." (Though researchers do not yet know whether the short is always dominant in all tissues of the body, or whether having two shorts is different behaviorally and emotionally from having one long and one short. )

In any case, the extremely common short form is associated with higher levels of normal anxiety. That is, on average, people who have the shortened form are more anxious than people who have two longs, but they are not pathologically anxious; they do not qualify for a diagnosis of generalized anxiety disorder (GAD), unless of course they do for other reasons. Ask a room filled with 500 people, half with the short form and half with the long, how they feel about speaking in public, and the group with the short form is going to be more anxious on average-though of course there will be plenty of "short-form" people with low anxiety, as well as "long-form" people with high anxiety. Nothing is absolute.

At this point, of course, any parent reading this account may be feeling confusion setting in for real, since many of us do see a great deal of anxiety in our autistic children-why shouldn't the short form, associated with higher anxiety, be exactly what researchers expected to see?

The answer is that, for the time being, there is no answer. That's the difference between designing a hypothesis according to behavioral data (autistic people have high anxiety) versus designing a hypothesis according to physiological data (autistic people have high blood serotonin). As we've said: the research is extremely complex.

"...serotonin may be involved in autistic learning and social deficits as well as in mood and aggression. This is an exciting possibility given that the large pharmaceutical companies...are spending billions trying to develop new and better serotonin medications."

For his part, Cook, who is a clinician as well as a geneticist, has given a great deal of thought to what these findings may mean directly, in day to day life, for the children he sees: "My latest hunch is that the short/long distinction may be related to aggression. Aggression is one aspect of autism we don't currently have ratings of in our samples, because as a group we've been appropriately focused on whether our kids did or did not have autism, and there is nothing about aggression that is diagnostic of autism. The most aggressive people in the world, the kids with childhood onset conduct disorder, don't have a touch of autism. "

Fortunately, we know that medications that affect the serotonin sys-tem-the SSRIs, the older antidepressant Desyrel, and atypical antipsychotics like Risperdal-treat anger, irritability, and aggression in many clinical populations, including people with autism, and this is where Cook sees the HTT gene findings as eventually being useful:

"What I'm most interested in with this gene is whether it will give us a way to predict what dose of an SSRI a child needs. Some of these drugs are metabolized by enzymes that vary a great deal in the population. So in 90% of children and adults with autism the usual administration dose may make sense, but the other 10% might be completely different. There isn't a real predictable relationship between blood serotonin levels and clinical response, and I think there's a good chance we'll get some practical clinical data from this research soon. "

Apart from this, the serotonin-autism connection may give us clues to other aspects of autism quite apart from anger and aggression. Cook again, "In broad strokes, if there's more mental retardation, there's higher serotonin-though we do see high-functioning kids with high serotonin as well. "

Which points to the very real possibility that serotonin may be involved in autistic learning and social deficits as well as in mood and aggression. This is an exciting possibility given that the large pharmaceutical companies (which smaller start-up biotechnology companies call the "big pharmas") are spending billions trying to develop new and better serotonin medications.

Cook explains:

"Right now there are limits to how high you can push the serotonin system. If you push the dose too high you get a worsening of symptoms-in depression, autism, or any problem you're treating-and that's usually because you're triggering the 'autoreceptors.' The autoreceptors are like a thermostat in the system: they say, 'Oh-oh, there's too much serotonin, I have to shut the system down. '"

In other words, push the dose too high and you end up with less serotonin in the synapses, not more. Up to a point, an SSRI like Prozac will increase the amount of serotonin in the synapse; after that point the autoreceptors turn on and start pulling serotonin back out of the synapse.

Fortunately, the big pharmas are working feverishly to find a way around this barrier-not on behalf of people with autism, but in order to help people with depression, schizophrenia, obsessive-compulsive disorder, and other anxiety disorders.

Cook says:

"A lot of people are trying to figure out how to get around the autoreceptors, either to get a faster antidepressant response or to treat resistant depression. One model of doing this that has been shown to reduce the time to antidepressant response and treat non-responders is to add pindolol to the SSRI. Pindolol is a beta blocker normally prescribed for hypertension, but it has the 'impure' effect of also blocking the serotonin autoreceptors."

Unfortunately, when Cook has tried this combination in a few of his patients with autism, he has not seen any improvement. But he's confident that sooner rather than later we'll have something that can block the serotonin autoreceptors in our kids:

"It's possible that our understanding of the serotonin system is insufficient, but I'm very excited because Prozac is coming off patent this year or next, so Eli Lilly has to come up with something else. And when they do, we could start to see medications that can treat social and learning issues, too. "

Trouble on Chromosome 15

"This is the hottest story in autism genetics," says Eric Courchesne, speaking of the recent confirmation of a link between autism and chromosome 15-a connection that has now been found by two separate teams in three separate studies. With chromosome 15, we move directly into genes affecting the cerebellum, one of the main brain structures that UCSD's Courchesne (as well as, in Boston, Margaret Bauman, and earlier, at UCLA, Ed Ritvo) has found to be affected in cases of autism.

First off, it's important to remember that with chromosome 15 we are talking about a chromosome, not a gene. Chromosomes are the squiggly lines expectant parents see on their amniocentesis reports; the baby's 100,000 separate genes lie on these 46 chromosomes (which are arranged in 23 pairs, one from the mother, one from the father).

The new finding on chromosome 15 is of an affected region on that chromosome- a region that does not "look right." This puts the chromosome 15 finding in a different category from HTT: as Courchesne puts it, "I would bet a dinner at the nicest restaurant in San Francisco that this is a mutation, not a normal variant."

He and Cook looked at Chromosome 15 because Christopher Gillberg, author of The Biology of the Autistic Syndromes, suggested that 15 would have problems. In 1991 he reported several cases of autistic people with duplications of genetic material on 15. Sure enough, Gillberg was right. And while often in the history of "behavioral" genetics initial findings have not been replicated, so far this one has.

The biggest news about chromosome 15-the finding that suddenly places autism research in one of the hottest areas of all genetics research-is that children who develop autism due to an anomaly on chromosome 15 do so only if they received the anomaly from their mothers. In other words, for the first time ever, researchers have established a mode of transmission of autism-in this case, through the mother, not through the father. (Cook says fathers will get equal time once all the genes are discovered; it just so happens that this first anomaly comes from the mom.)

In the science of genetics this phenomenon is called "imprinting," and it is one of the most exciting-and most active- areas in the field today. Courchesne explains: "Imprinting refers to the concept that the gene will become differentially active-or "expressed"-based on whether it came from the father or the mother. Some genes remember where they came from; they care whether they came from the mother or whether they came from the father. And that "memory" determines whether or not they are expressed in the child. "

With a maternally-expressed gene (or mutation) the gene has to come from the mother in order to be expressed in the child. If the child gets the exact same mutation from his or her father, nothing happens; the mutation is not expressed. With a paternally-expressed gene it's the opposite. The child has to inherit the gene or mutation from his father in order to have the traits that gene causes. Otherwise the mutation remains silent.

This is exactly what Courchesne and Cook found in the first family with the chromosome 15 abnormality whom they studied closely. There were three children, a girl with classic autism, a boy with atypical autism, and a third child, a girl, who was developing typically. (The unaffected sister was actually a step-sibling; the mother had remarried before conceiving her.) Both of the affected children had a duplication of material on chromosome 15. When Cook looked at the mother and the father of these two children he found that the father's chromosome 15 was normal; it was the mother who had the duplication.

But the mother herself was completely average; she showed no signs of autism at all, not even subtle ones. Cook then went back to her parents, and found that she had inherited the abnormal chromosome from her father-whose own version of 15 was normal. In the transmission of 15 from father to daughter, the chromosome had undergone a spontaneous rearrangement, which the daughter then passed on to her own children. The mom was normal because she had received the mutation from her dad; if she had received it from her mother, in all likelihood she would have been autistic, too.

Cook and Courchesne have now looked at 140 children with autism in all, and have found one more, a boy, with a duplication on 15. He inherited the duplication from his mother- who did not have the mutation herself. In this case the duplication arose "de novo" when the particular egg that was to become this boy was originally formed many years ago. (Which means that this mother's chance of having a second child with autism is near zero, since the duplication on 15 cropped up simply as a random mutation in a random egg). The 2-out-of-140 rate may be low, of course, because at this point researchers are dealing with anomalies on chromosome 15 so large they can be seen under a microscope- or picked up by having 3 alleles at a locus instead of the normal 2. Any of the other 138 children could also have duplications on 15 that are too small to be picked up in this way. (One note: in all, Cook and his group have found 3 children with the chromosome 15 duplication: the brother and sister from the first family, and then the boy whose mother did not have the duplication. But the official figure is 2-out-of-140 instead of 3-out-of-140 because the one very high-functioning boy was too mildly affected to meet the diagnostic criteria for inclusion in the 140. The research team picked him up by accident, after they found the duplication in his sister and so decided to look at the whole family. )

Interestingly, further evidence for maternally-imprinted duplications on chromosome 15 has just come from Browne's team in England, which has been studying the genetics of language disorders. They, too, found that the genes they were looking at had to come from the mother in order to produce a language disorder in the child. When their paper recently appeared in the December 1997 issue of the American Journal of Human Genetics, the authors mentioned Cook and Courchesne's paper and said that while they hadn't been looking for autism, now they would.

And finally, Cook's work has been duplicated and extended in the laboratory of Margaret Pericak-Vance (an expert in gene-mapping and 1997 NAAR Research Award winner) at Duke Univeristy. Previously Pericak-Vance had also found anomalies on 15; she has now replicated the maternal inheritance, and has added two important pieces to the puzzle:

1. Pericak-Vance found an "increase in recombination" on chromosome 15 in families with autism. "Every time people have a baby," Pericak-Vance explains, "it's like a deck of cards being shuffled. " Say you have a deck of cards with all four suits separated out from each other, and the numbers put in order. Then you shuffle that deck of cards once. The families that end up with an autistic child will show a much more pronounced "reshuffling" than the families that end up with a non-autistic child. In physical terms, Pericak-Vance and her team found that the autistic person's markers on chromosome 15 appear further apart than they are in the typical person. And: this difference came from the mother.

2. Having confirmed Cook's findings, Pericak-Vance then looked at chromosome 15 in families with a different neurological disorder, unrelated to autism. She found that in these families, this region of chromosome 15 was normal, further evidence that the duplication on chromosome 15 is specific to autism-not a general genetic anomaly you might find in many brain-based problems. More evidence for 15. (Cook, too, has a paper in press in which his team looked for duplications on 15 in over 250 non-autistic children with moderate to profound mental retardation, and did not find any duplications on chromosome 15, although they did find 4 cases of Angelman syndrome in which there was a deletion of the same portion of chromosome 15 that is duplicated in autism. )

Pericak-Vance notes that there are a number of different possibilities as to what could cause this anomaly. You might see perfectly normal genes that for some reason have been duplicated, giving the child an extra copy. Having extra copies of otherwise normal genes can be very damaging to the organism. This is the problem in Down syndrome. Or you might see some kind of incorrect rearrangement of otherwise normal genes; you might see a mutated gene that is causing the chromosomes to break and reshuffle. There are other possibilities as well.

Time will tell-and most researchers feel we'll know sooner rather than later. The next step is to pinpoint a narrower region on the chromosome, or a single gene within this region that is key to the disorder. Ed Cook's prediction: "Within the next two years there's going to be some very hot and definitive information about specific genes involved in autism."

Where Does Your Child Fit In?

At present we can't tell the autistic children who have chromosome 15 duplications from the ones who don't simply by looking at them. However, there do seem to be characteristics specific to these kids. "The one I'm sure of," Cook says, "is increased epilepsy and epileptiform EEGs. One autistic woman we studied didn't have her first seizure until her late teens, but she had abnormal EEGs as a child in the way autistic kids often do. "

The chromosome 15 children studied so far also show regression. Between 12 and 24 months in their development, they lost skills. As well, these children have low muscle tone. "They walk on time," Cook says, "and they can eat OK; it's not severe. But they might have a little trouble holding their heads up as infants, and show a history of low tone in other ways. Most kids with autism aren't like that, so the floppy ones stand out a little bit. " He continues: "A lot of them visually look like Fragile X, with hyperextensibility of the joints, double-jointedness, and ears that may be a bit longer than normal, and incorrectly 'rotated' backward. "

As preliminary as these impressions are, they are extremely significant for any parent of an autistic child who is contemplating having another baby. Cook gives this advice to parents: "You can find this on an amnio, but most labs don't do it. You have to look very carefully. But people who are trying to get pregnant now, and already have one autistic child, should look for it. It's much more important than looking for Fragile X, though we still recommend checking for Fragile X, too."

Ed Cook's prediction: "Within the next two years there's going to be some very hot and definitive information about specific genes involved in autism."

Bear in mind, of course, that any lab that agrees to look for a duplication on chromosome 15 is going to come up with a large number of "false negatives," since at this point all anyone can look for is an anomaly large enough to be seen under a microscope. Bear in mind, too, that we don't know what a chromosome 15 duplication found on an amniocentesis is going to look like in the actual child. Of the two original children Cook and Courchesne studied, the sister was much more severely autistic than her brother, who was so mildly affected that the school system did not want to provide him with services. His IQ, language, and academic performance were normal, and the school system was not concerned with his narrow interests or poor social skills.

This is the mystery of gene expression, the mystery of why a gene mutation can be devastating in one person, only mildly troublesome in another, and silent in yet a third. Ed Cook comments:

"Even in this family the little girl probably has a second gene involved. So here is a major finding and you can't even use it to distinguish a child who is mildly retarded and has classic autism from a child who has normal intelligence and is only mildly autistic. "

Parents of children with autism who are contemplating having another child and would like to check for duplications on chromosome 15 should tell their physicians that a possible region may be 15q11-q13, so that the chromosomal analysis will be done with attention to this region as well as to the other chromosomes. Be sure to discuss this very early on in the pregnancy, since locating a lab that can do this test may take time.

Can Chromosome 15 Lead Us to a Treatment?

The prospects for chromosome 15 leading to a biomedical treatment for autism-not a "cure" (or not necessarily) but a genuine treatment-are high. This is so because the affected region on chromosome 15 contains three genes that code for the neurotransmitter GABA-and the pharmaceuticals are already pouring buckets of money into the GABA system, and have been for years. GABA, or gamma-aminobutyric acid, is the neurotransmitter involved in anxiety. Alcohol, anticonvulsants like Gabapentrin and Vigabatrin (note that the drug companies have been helpful enough to include "gaba" in the names of these two) and antianxiety medications like Xanax and Valium all work by attaching to the GABA receptor.

GABA is an "inhibitory" neurotransmitter; it prevents cells from firing. Some call it the brain's "braking system." This brings us to another line of converging evidence: in the cerebellum, the Purkinje cells-which Margaret Bauman has found to be diminished in number in the autistic brain-release GABA.

The problem with antianxiety drugs like Valium and Xanax, as anyone who has taken either for sleep knows, is that although they can work wonders at first, the effects do not last. Shortly before last Christmas, Cook used a GABA medication to treat a severely behaviorally disordered young man with autism, and it helped. But the effect was fleeting. As a result, the pharmaceuticals are engaged in an ongoing quest to develop a GABA drug that can work over the long term- the financial payoff would be enormous. And the chances that one or more of this new generation of improved GABA drugs could be helpful to our children are good. It is also possible that an existing compound-a medication that has already been developed and tested for safety but never marketed-could work for autism. Drug companies cannot legally test medications in humans without having a biological "target," and until now it was not known that GABA was involved in autism. As a result, none of the GABA medications has ever been formally tested in people with autism; the tests were all run on people with anxiety disorders. A medication that does not work for an anxiety disorder in fact might work for autism. It's possible.

"I think the differences between the autistic brain and the normal brain are relatively subtle. Of course a structural difference can be small but critical, but even so I don't see anything in the neuroanatomical studies that says autism is untreatable."

As Ed Cook says, "Now we need to think about the GABA system as much as we think about serotonin. " Happily, more work on GABA is being done all the time. The Cook team's findings on GABRB3-a gene for one part (or subunit) of the GABA receptor-are in press, and will appear in May in the American Journal of Human Genetics.

Other Paths to a Treatment

As to the question of whether the missing Purkinje cells are the "real" problem, as opposed to a "chemical" anomaly in the GABA system-it is at least theoretically possible that an autism-specific GABA medication could compensate for missing cells by drastically increasing the GABA production of the Purkinje cells that are present. This has been done in other brain disorders like Parkinson's disease. Or, eventually, the structural differences in the autistic brain may be treated by "neurotrophic factors" or "nerve growth factors"-chemicals that cause new brain cells to grow. (See related story on p. 10 in the Summer 1997 issue of NAARRATIVE.)

But Cook believes-and here there is disagreement among researchers-that the structural flaws we see in the autistic brain are not drastic enough to be insurmountable: "There's nothing that's that abnormal in the brains of people with autism. If you compared a young autistic person's brain to the brain of his healthy 60-year old grandfather, the grand-child would have the better looking brain. " [Editor's note: Men's brains shrink with age-as do women's, though to a lesser degree.]

"I think the differences between the autistic brain and the normal brain are relatively subtle. Of course, a structural difference can be small but critical, but even so I don't see anything in the neuroanatomical studies that says autism is untreatable. With the right nerve growth factor, you might get maturation of those structurally different parts of the brain. " The fact is, it is possible to treat autism now: both Anafranil and the SSRIs have been shown to diminish core symptoms of the disorder, not just behavioral "add-ons. " (see article, p. 6.)

In Ed Cook's words: "The SSRIs are very exciting. With these medications we can treat something we couldn't touch just 10 years ago. I think there's a lot of excitement about where we can go with autism treatment medically, and in general I see SSRIs as giving us 5 percent of what we'd like to be able to do. Say we get 5% every five years-that doesn't sound like a lot. But there are going to be a number of kids out there who, with just a 5% bump up in functioning, will have their lives significantly changed. Then you keep adding onto that, and adding on, until you get as far as you can go.

"Will we eventually be able to cure autism? I don't know. Maybe there would always be something left over; maybe you could never give an autistic person the fluidity of thought and movement normal people have. But I don't see anything about the neuroanatomy that says we can't bring everybody up to Temple Grandin's level, except that the rest of us aren't as bright as Temple.

"But we're very far away from that today. "

This is where parents come in. What we can do-what we must do as parents-is to push the science forward. Raise the money, raise the awareness, make it happen. That is our job, and our hope.

Other Hot Spots: Chromosomes 7 and 16

The results of the first full genome-wide screen of autism were published this March in Human Molecular Genetics. This study, by the International Molecular Genetics of Autism group, reported linkage for chromosomes 7 and 16. Previously, Sue Smalley of UCLA had suggested a connection between autism and tuberous sclerosis (TS) that excites people; one of the genes for TS is on 16, though it looks as if this gene is not going to be the same one the International group is looking at on 16. Nevertheless, researchers feel there is fairly strong evidence for an autism gene on the long arm of chromosome 7, weaker evidence for an autism gene on the short end of chromosome 16. Geneticists are in the stage of working across groups to find out which gene hypotheses hold up and which do not; we'll report their discoveries as they emerge.

At this point we have no idea how many autism-susceptibility genes researchers will eventually identify. Assuming it takes a combination of 5 genes to produce the disorder, there is nothing to say that these 5 genes will be the same 5 genes in every person with autism. There could be 20 autism-susceptibility genes, with some people having one combination of 5, other people having other combinations of 5. And of course, it is likely that there also exist dominant genes for autism, genes that can cause autism acting entirely on their own. We just don't know yet. As Clarence Shutt, a structural biologist at Princeton and Executive Vice President of NAAR, says: "In science, everything's a mystery until it happens."

A Note From Ed Cook: We would like to thank the NIMH, NICHD, NINDS, the University of Chicago Brain Research Foundation Seed Grant Program, the Jean Young and Walden W. Shaw Foundation, the Irving Harris Foundation, the Daniel X. and Mary Freedman Academic Psychiatry Fund and the MRC in the UK. None of what has been done in our laboratories would be possible without this support.

Catherine Johnson, Ph.D., co-author with John Ratey, MD, of Shadow Syndromes, is a member of NAAR's Board of Trustees and the mother of two children with autism.