Now we're finding changes in the brain that appear to have happened after birth. I found white-matter enlargement that was consistent with what other people were finding, which was that in autism the brain on average gets bigger after the child is born. This is one of the things I found that was outside of the model I'd started with. Subsequent work has found that there is a massive brain growth spurt in the first 2 years after birth, where the rate of growth of the autistic brain-and this may be a subgroup, but it's a substantial subgroup-shoots up and the brain gets way bigger by the time the child is 2 than in an average 2-year-old. And then the growth rate slows down relative to children without autism, so that there is no brain volume difference by adolescence, and by adulthood brains in autistic individuals are even a little smaller than those in typically developing individuals. So something is going on after birth, and though you can make up a story that it is triggered prenatally, no one has proved that or excluded the role of post- natal factors.

In my own imaging-based anatomy research I was looking at MRI scans from autistic children aged 5 or 6 to 11 years, and they had big heads. When I analyzed the data further I found that it was white matter that was big-and on yet further analysis I found that the enlargement didn't involve all the white matter, but it specifically affected the white matter right under the cortex that develops its white myelin coating the latest, well after birth, and not the deep white matter that myelinates earlier. There were more consistencies with other findings: the areas of the white matter that were bigger were the areas that were myelinating during the time that retrospective studies were identifying the brain growth spurt-and also during the time when "autistic regression"-the loss of skills like language and social interactivity and the onset of autistic behaviors like rituals and hand-flapping-tends to occur. More recently my group's finding regarding the distribution of white matter enlargement has been pursued by my colleague Carlos Pardo, a neurologist and neuropathologist at Johns Hopkins, who had already demonstrated activated microglia and acti-vated astroglia in brain tissue from autistic individuals-these are signs of innate immune activation. After reading my paper local-izing white matter enlargement, he went back and stained tissue in the same distribution as the areas I'd measured, and he detected cellular changes consistent with immune activation in the same parts of white matter where I had detected volumetric enlargement. This suggests that this white matter enlargement may be related to immune activation, which may be driving brain enlargement and impairing brain function.

Dr Pardo's findings of brain immune activation completely change the playing field of what is relevant to how autism works. In other words, if that's going on, then you have an ongoing chronic-disease process. There may or may not be early wiring changes, but you have an ongoing chronic-disease process. And that's a totally different ball game from what we've been thinking about autism-it adds a whole extra axis to the dimensions in which we need to characterize the condition.

To flesh out the implications of these chronic changes in autism, I wrote a paper called "Autism: A Brain Disorder or a Disor-der That Affects the Brain?" More recently I coauthored an article with my neurobiologist and neuropathologist colleague, Matt Anderson, about this called, "An Expanding Spectrum of Autism Models: From Fixed Developmental Defects to Reversible Func-tional Impairments," that will come out next spring in a volume edited by Andrew Zimmerman, a close colleague of Carlos Pardo's and an important pioneer in immune system research in autism. It's premature to say that the earlier model of fixed wiring deficits is wrong. But it is not premature to say that there are things going on later that could actively influence the level and type of functioning of the brain-all kinds of cellular changes that would affect the synapses and the blood flow and other things that can manifest as problem behaviors, either in addition to or even instead of early wiring diagram alterations.

AT: How did the TRANSCEND project come about?

Dr Herbert: I developed the TRANSCEND model because of the way a whole series of observations and study methods hung to-gether and yet were all being pursued separately. I was doing MRI research, and I noticed that the brains I was looking at were big. Then I added diffusion tensor and spectroscopy measures in MRI because they show you something about the tissue architecture and the tissue chemistry of the brains, but they still don't show you anything about function. So we added EEG and MEG (magne-toencephalography), which we chose rather than functional MRI because EEG and MEG are sensitive to time intervals as small as 1/1000, whereas MRI can only discern processes as short as 1 second. Synapses function at the level of 1/1000 of a second, and you wouldn't pick up problems related to neuronal malfunction with a functional MRI, but you could pick those up with EEG or MEG. We are planning collaborative projects with several neuropathologists, electrophysiologists, and technology designers.

What TRANSCEND collaborators are trying to find out is, are the parts of the brain where we will find abnormal tissue archi-tecture or abnormal tissue chemistry related to the parts of the brain where the timing and coordination of the connections is not typical? And if so, if we treat or modulate some parts of this, will the others change in concert? How do we best design experi-ments and measurement techniques to detect these changes, both experimentally and clinically? In order to do that, we have to have an interdisciplinary collaboration because no one researcher can do all of that. The next question, now going beyond the brain, is, are these individuals also, at the same time, showing signs of systemic metabolic abnormalities, immune disturbances, biochemical disturbances, and infectious disturbances? Could it be that dysregulation from immune or biochemical or other meta-bolic or infectious problems can have an impact across the blood-brain barrier and affect brain function? The technical word for this is encephalopathy. We are asking if autism is a dynamic encephalopathy, even a metabolic encephalopathy-that is, a meta-bolic disturbance of the way the cells can function in the brain.

In addition to the immune findings, we have clinical reasons for thinking that this may be a metabolic disturbance-there are times when you see children get better very briefly. There's a paper coming out in Pediatrics from Andy Zimmerman's group documenting that a fair number of children with autism get a lot better when they have a fever. They start making eye contact; some of them talk. I had one mother of an 18-year-old autistic boy tell me that when he would get a fever when he was younger, he would become much more communicative. And she said she welcomed those times because she would get to visit with her son, as she called it. Another time you see these children do better is when they're going to have a colonoscopy or an endoscopy, and they have to go on clear fluids and have no solid foods. Some of them will perk up and make eye contact and talk; then, when you rein-troduce foods after the procedure, it all goes away.

It makes you think that the autism is not due to a broken system. It's a system that's capable of functioning, but something is suppressing the function. It's a really different model, but it fits a lot of the things that we see. So then the question is, what is sup-pressing the function? In some cases, it seems like the suppression is pretty easily reversed transiently, although it can't be kept re-versed very easily. In other cases, it seems to take a lot of metabolic tinkering to get a child to be able to pull out of it somewhat. But even so, in order to be motivated to do the hard work needed to make those kinds of improvements endure, you need to have a model of what you're doing, one that tells you to look for and address functional disturbances and not just hard-wired disturbances.

In TRANSCEND, we're trying to bring together different levels of research and different measures in an organized way so that we can see how they relate to each other. My colleagues who have partnered with TRANSCEND have been hungry for this kind of collaboration. The other thing we're trying to do is to develop a profile of measures that are sensitive to change, because we're look-ing at autism not as a fixed state, not as a trait, but as a state that can be changed. People have been treating it like it is a fixed entity, and consequently there aren't a lot of good, and particularly not a lot of validated, measures of improvement or change in autism.

When you start seeing change, you want to figure out, how do we measure that change? What domains is the change in? What parts of the brain? What sorts of functional tasks can you give a person so that their EEG will show the change? I don't know of anybody else who's doing that. I hope we get more company soon.

AT: This is fascinating on many different levels, and it brings up the question of autism-singular-vs autisms, plural.

Dr Herbert: This ties into the question of final common pathways. People have spent years meticulously defining the behavioral criteria for autism. You need to have impairment in communication, impairment in social interaction, and manifested repetitive behaviors or restricted behaviors or interests-all of that by the age of 3 in order to meet the full criteria for autistic disorder. Or you have to have pieces of it in order to be on the autism spectrum.

My feeling, and I'm not alone in this, is that these behaviors are a final common pathway, and you can get to the point that your brain will produce these behaviors by a variety of different biological pathways. So the problem is that if all you're looking for are the components named in the gene-brain-behavior model, then the links in between genes and brain and behavior become black boxes. In particular, people haven't been measuring enough of the biology that's in between the genes and the brain for us to be talk-ing enough about what biological subgroups we may have in autism.

Our TRANSCEND model is what we call a middle-out model. The gene-brain-behavior would be bottom-up, or if you start with the behavior, it's top-down. We're saying that the intermediary biology is the middle, and then you work out from looking at the biology, the physiology of the person, and you work back to what genetic and environmental things could have led to it, and you can work out to how this would lead to behaviors. At the core, you're grounded in the biology of the person.

As an example, some people think that there could be, for instance, a calcium channel abnormality. There are a lot of different genes that can affect calcium channels, and there are a variety of toxins that can affect calcium channels. Others talk about methy-lation abnormalities. For each of these, there are a variety of genes and a variety of toxins that can cause such problems-making these mechanisms potential final common pathways-but the behavioral outcomes seem to wind up similar, for reasons that it would be incredibly valuable to figure out.

The point is that you can end up in a similar place even if the specific environmental trigger or the specific genetic vulnerabil-ity is different from case to case. Autism-when thought of as a singular condition (which has really gone out of fashion, and for good reasons)-is meeting the criteria for behavior and even then, there are people who have more repetitive and restrictive behav-ior, people who have language problems, and people who don't. So even at the level of behavior, it's heterogeneous, it's interindi-vidually (as well as intra-individually) different. We started TRANSCEND because we have not had a research program that con-nects the people who are measuring the biological configurations with the people who are measuring the behavioral configurations to see whether there's a relationship between the two. It's been very fragmented. I think we need to rethink that. Our "middle-out" approach is about these different levels, and about rethinking the way we design research programs and collaborations.