IMFAR 2011 Keynote: The Developmental Neurobiology of Autism: The First Steps and the World Ahead
Eric Courchesne, UCSD Autism Center of Excellence
The following is a modified transcript of Dr. Courchesne’s talk. I have not included all references to slides he presented, and I have left out some of the researchers he thanked and technical terms he cited. Emphases in bold are mine. Any additional omissions or errors are my own.
IMFAR introduction: Eric Courchesne is a leader on brain structure and abnormalities in autism, and has been pursuing research into autism before most IMFAR attendees “knew what autism was.” Dr. Courchesne had polio as a child, which left him unable to walk or stand. Medical intervention gave him the ability to walk without braces, and allowed him to eventually become a nationally ranked college gymnast. So, he has a personal stake in overcoming disability, and sincerely believes medical science can make a difference for kids with autism. He dedicated his IMFAR 2011 presentation to his mother.
Neurodevelopmental Studies of Autism: from MRI and fMRI, to Neurons and Gene Pathways
My talk will basically attempt to describe our little part of what we’ve done in looking at the neurodevelopment of autism. Imaging research has taken us from structure to functional and to what I believe are some new findings, very interesting findings on neurons and gene pathways.
We all know that autism is a neurodevelopment disorder. It begins very early in life. The first symptoms may be subtle, especially in the first year. The second year is usually when we see a variety of red flags displayed in kids who go on to develop autism.
Surprisingly enough, until just a few years ago, very little was known about the developmental neuropathology of autism. This is a well-known disorder, and yet, through most of its history, there were almost no imaging studies of the very young autistic brain. There were no functional studies. Dr. Geri Dawson’s lab did some of the early developmental neurophysiological studies examining how the brain operated in three year old and four year olds with autism, but except for a few groups such as hers, there really wasn’t a lot of research being done in that area.
I think it’s safe to say that there is not a body of literature on the developmental neuropathology of autism. So we really don’t know what the underlying neural defects are that cause this disorder. What we do know is a great deal about the adolescent and adult autistic brain, based on imaging studies and neuropathology. We know from those neuropathology studies that the amgydala in ten to forty-four year olds [with autism] shows a decreased number of neurons. We also know that the fusiform neuron numbers are reduced in older children, adolescents, and adults with autism.
In general, in adolescent and adult brains of people with autism, we see decreased amygdala neuron numbers and volume, dendritic arbors, minicolumn size.There’s also often thinning of the cortex (though not all studies find this), as well as reduced thickness in the corpus callosum. We know that there is evidence of glial activation in the older autistic brain as well as molecular signals of inflammation.
This has basically led people to think that neuroinflammation has something to do with causing autism. Well, maybe it does, and maybe it doesn’t. But it’s hard to see how we’ll know until we do lifespan or developmental studies. John Morgan has done to my mind the first (and so best) statistical and quantitative analysis of postmortem tissue, looking very precisely at parameters involving glial-neuron interactions and neuron-neuron interactions. His studies demonstrate that there is an age-related change in the autistic brain across the years from ages two to forty, that there is an increase in the relatively haphazard arrangements of neurons within the autistic frontal cortex. So, rather than a nice, orderly spatial distribution, it becomes more and more patchy, or haphazard. To my knowledge, that’s the first evidence that there’s an ongoing process of change in the autistic brain.
It’s high time for the field to realize that we don’t have a high volume of autistic brain tissue. There aren’t thousands of cases, there aren’t even hundreds of cases. It really comes down to a dozen or so really high quality cases with really good tissue. Those studies, even though they’re small, are pinpointing fundamental underlying defects that can’t be seen any other way in the autistic brain.
Morgan has gone on to study the relationship between microglia [the “main form of active immune defense in the central nervous system“] and neurons. We’ve seen signatures that microglia are activated, wrapping around neurons. These are signatures, known from basic research, that microglia are involved in either in the process of synaptic remodeling, synaptic removal, process removal, or neuron removal. It’s an active process of remodeling that’s taking place, during the lifetime, in autism.
We became interested in the question: what is the neurodevelopmental basis of autism? We know how it ends, based on our studies of adults and adolescents, but how does it begin?
About a decade ago, around the time of the first IMFAR, I published a paper that showed unusual early brain trajectories in autism, namely an overgrowth with larger brain volume in autism, as compared to the path of growth in a normal brain. In autism there’s a premature acceleration of growth, by about two to four year of age.
Autism is a lifespan disorder, and that pathology changes across the lifespan — in the older age range, something different is happening in autism, that has to do with neuron loss and remodeling, and activation of microglia.
Based on the literature as well as our own research, we proposed the theory that autism involved three phases of growth pathology:
- Arrested growth
- Possible decline and maybe even degeneration
The most consistently affected areas include the frontal and temporal cortices, as well as the amygdala. Those structures are very very important because without a doubt, these three domains of brain systems underly the higher order social, emotional, communication, and language processing deficits that exist in this disorder, in the very first years.
In order to study autism at a very early age as it occurs in the general pediatric population, and in an effort to transform the general pediatric screening for autism, Karen Pierce developed a one year well baby checkup screening that been featured in the media lately. It can be used in the office by any pediatrician, anywhere, any time. It’s a fast screen approach, it’s a very easy screen approach that investigates autism as it occurs in the general population. It not only catches individuals with autism but it catches individuals with other developmental disabilities, early on.
Using her screening, Karen went out and recruited 170 pediatricians in San Diego county, in order to get them to screen early, and identify a large cohort of autistic individuals at one year of age and up.
At our Autism Center of Excellence, because we have an outstanding neuroimaging team, we have now collected a large number of MRI scans of infants and toddlers with autism, almost 1000 scans total in my lab’s work — including controls: kids with developmental delays, language delayed kids, and siblings.
In the field, our clinical team has screened over 18,000 babies over the last couple of years, and we have over 420 infants that have been diagnosed [with autism] and tracked, and we have biomarkers on all of them. We have longitudinal blood draws on them from infancy on up through three or four years of age.
So, early on we discovered the phenomena of early abnormal brain overgrowth — we found that at birth, head circumference [in kids with autism] is slightly smaller, but by about twelve months the head circumference of kids who go on to have autism is larger than normal. We showed that these head circumference changes as compared to the CDC norms, this reflects an accelerated growth early on. That head circumference growth has been replicated by a number of independent research laboratories from around the world.
In our new studies, we replicated our original finding: growth in the twelve month to forty-eight month group compared to controls: a trajectory of early overgrowth of the brain.
In trying to figure out which part of the brain is exaggeratively overgrown, [we looked at] frontal and temporal lobes, because they’re the regions that underly social, emotional, language, and communication functions. We set about subdividing the cerebral cortex and we found that the frontal cortex and temporal cortex are abnormally enlarged [in autism], as compared to controls. And others have also found overgrowth in these areas and sometimes even in the parietal cortex.
If these structures are overgrown early in life, you would predict that there would be abnormalities of either neural or circuit connections, and we should be able to find those. So we undertook the largest study that’s ever been done with DTI [Diffusion Tensor Imaging] — there are no DTI studies of adults or even of children that are as large as the DTI acquisition studies we’ve done on infants and toddlers with autism, as well as controls.
We asked the questions: are there DTI differences in individual voxels*? Which fiber tracts are involved? Higher order social language and communication functions rest strongly upon the normalcy and the integrity and the development of long-distance fiber connections. And so we reasoned that DTI was a perfect tool for investigating the developmental growth of these fiber tracts — tracts that interconnect temporal and frontal regions, interconnect right and left hemispheres, fiber tracts that connect frontal regions with posterior regions.
And this is what we found: abnormal development of frontal and temporal circuits in autistic individuals as young as infancy, and in the toddler years. The fibers that interconnect within the frontal cortex and allow higher order integration of information within the frontal cortex are abnormally developing in autism. We found statistically significant abnormalities of striatal inferior frontal fiber tracts:
- The superior longitudinal fasciculus is a major fiber tract that is crucial for the development of higher order language functions and interconnects temporal with frontal regions.
- The uncinate fascicles is a fiber tract that’s crucial for interconnecting temporal cortices, including the amygdala and emotion systems with the frontal cortex.
- The forceps minor is a fiber tract in the frontal lobe that interconnects right and left hemispheres, so the two hemispheres can share information and cooperate in developing more sophisticated, higher order understanding of meaning, context, and social interactions.
Especially interesting is this is what the abnormality is: In the superior longitudinal fasciculus, in the early years of life, between twelve months of age and twenty-four months of age, we have an accelerated maturation of the superior longitudinal fasciculus in autism as compared to controls. We interpret this to suggest an overabundance of axons, and that these axons may be highly compacted, so that there’s a decreased fiber density, and an increased number of fibers.
But that increase is short lived in terms of its functional significance, because with age, there’s an underdevelopment of that same fiber bundle. This underconnectivity, this trajectory fits perfectly with a large body of literature on older autistic individuals — older children, adolescents, and adults with autism — as well as existing theories that autism in those age groups is marked by disconnection, underconnectivity, and functional disconnectivity. But [our] data show that autism doesn’t begin that way, it begins as an overabundance, probably of abnormal connections.
So there’s widespread abnormality of early accelerated connectivity, and circuits in frontal and temporal regions are failing to mature in a normal way.
In typically developing kids, early in life, before language is fully mature, there’s near-bilateral activation [in response to language stimuli]. As the brain develops it lateral specialization for processing the meaning of language, the left side of the typical brain becomes stronger, more active, and more responsive to language than does the right.
In autism, this doesn’t happen. From the get-go, there’s a failure of the left hemisphere to “come on,” and a failure of the left hemisphere to develop its specialization. And across time, there’s a failure of the left hemisphere to show strong leadership in the role of decoding meaning. Instead, it’s the right hemisphere.
If you were a little child with autism, you would say, “Well, that’s all good and well, but what caused this? Why do I have autism? What am I faced with when I first came into this world?” So we dedicated our UCSD Autism Center of Excellence to determining the identity of the neural and genetic defects that cause early brain overgrowth.
Among the things we’ve done are a host of studies of post-mortem studies of the brain at very young ages. Two year olds, three year olds, four year olds, eight year olds, and fourteen year olds. And one of our first studies examined something pretty obvious: If the brain’s too large, and there are too many axons early in life, there’s probably going to be too many neurons. But how do we test this?
We got a unique [imaging] sample of very young autistic brains, whole serial sections, on which we can do modern, state of the art, quantitative stereology and estimate the entirety of the neuron counts in the frontal cortex. We counted both neurons, and microglia.
Because I want to avoid any possible bias on our part, I had a completely independent, world-famous stereologist perform this. And we engaged a frontal cortex cytoarchitecture expert. These people at the time were totally unaware of the autism literature, unaware of the purpose of the study, and we coded and randomly mixed all slides. Everything sent was coded, blinded, and even blinded to the literature of the purpose of the study — and this is what we found:
In the dorsolateral prefrontal cortex [which plays an important role in executive functions], there were 65 percent more neurons than in controls — which is a gigantic increase. In the medial prefrontal cortex, there was a 25 percent greater number of neurons than in controls.
This huge increase is very important to think about, because there are no known mechanisms for producing such a gigantic difference in neuron numbers, post-natal in the human brain. The more neurons that were present, the greater the abnormal deviation from normal average brain weight for age. So, increased neurons; increased brain size.
In control children, and adolescents, and adults, the number of neurons in the dorsal lateral prefrontal cortex is pretty much constant — almost the same number in two year olds as it is in teenagers and adults. This fits with the literature that suggests that, by perinatal [immediately before and after birth] life, near-adult numbers of neurons have already been achieved.
By contrast, autistic individuals, these really young individuals, have a gigantic number of neurons — almost twice as many brain cells as there are in controls. This near-doubling by two to four years of age is a striking signature of what must be a very early, if not certainly prenatal dysregulation of systems that govern the total number of neurons that that brain is going to have.
Now, think back to the DTI data that showed premature overgrowth or development of axon circuits on long distance pathways [above]. By the calculation of computational neuroscientists, a doubling of the number of neurons means a four-fold increase in the total number of axons that are likely to be there. That means the brain in autism is starting with not only too many neurons, but a vast number of axons whose connectivity is likely to interfere with normal function.
But after that, autism becomes a story of decline. [Our study] suggests that autism is not steady — it suggests that if you measure a variety of parameters in the living as well as the postmortem autistic brain, that there will be this decline.
That is exactly what was predicted by John Morgan’s data — Morgan provides exactly this time frame, he shows this path, but he shows the inverse, which is the “ascendance of loss,” where there’s a change in the orderly organization of neurons, and there’s an increase in haphazardness of neuron organization. And he shows that microglial actions around neurons that are indicative of the molecular and cellular biology that must be taking place during this whole period of time in autism.
If there’s an excess of neurons, what might be causing that excess? From frozen brain tissue, also again from the dorsolateral prefrontal cortex [DLPC] — remember, the entire first part of my talk was about frontal overgrowth — we went into the front lobe once again, took samples from the DLPC, and these blocks of tissue were processed in two different ways.
First, we wanted to know what microstructure might underlie this excess number of neurons. Secondly, we wanted to know what genetic abnormalities might be responsible for generating an excess of number of neurons, and if there are microstructural defects, might there be a signature from the genetics that could tell us why, microstructurally, the defects are there, too.
In this gene expression study led by Maggie Chow for the last couple of years, she found 102 differentially expressed genes — that is, genes whose amount of expression or activity is either abnormally high or low compared to age/gender matched controls (all male) — that distinguished autism at a young age from controls at young ages. Which is very important, because remember — gene expression changes with age! So you need to have the young controls in order to get the kind of data that we’re getting.
What we found is dysregulation of pathways that govern cell numbers, and the functional integrity of cells. To be exact, what we found were pathways that showed dysregulation of the genesis of neurons, dysregulation of the way the cycling of cells that produces more and more neurons operates. Most importantly, we found downed regulation of DNA damage responses, and downed regulation of apoptosis [the process of programmed cell death].
Downed regulation of DNA cell damage is extremely interesting, because what it suggests is that in the process of cell cycling, not only are too many neurons being generated, but there’s the possibility that neurons that have defective DNA are not detected, and the DNA defect is not corrected — or if it can’t be corrected, it’s sent down a pathway of naturally occurring cell death. In the absence of that process, it’s possible that neurons are being generated that gather DNA defects, and that those defects and those cells are not gotten rid of but are retained, with their DNA defects, and that those DNA defects in turn are generating still more pathology.
So it could be that autism is a disorder of somatic mosaicism, involving DNA damage to cells that are not going through apoptosis, and that that occurs in situ [in place], in the brain in autism, in the developing cortex, producing DNA defects that will never be detected from blood, skin, or any other method except looking in the brain.
In addition we found downed regulation of neural patterning genes. We have a host of genes that are down regulated. Your brain has a normal left-right asymmetry, and you have normal left-right specialization for function, and you have normal anterior-posterior specializations, and in the cuneus, you have normal superior-inferior specializations — because you have genes during development that enabled your brain to distinguish left from right, up from down, and front from back. Those genes are responsible for local neuropatterning.
So, we found downed regulation of a number of genes that regulate the normal neural patterning of the brain. And this may be a signal in helping us understand why we have abnormal asymmetries in autism.
In an end of first trimester, beginning of second trimester [human] fetus, if you took a cross-section of the forebrain, you’d see cells lining the ventricles and proliferating — they’re generating more cells. If you went to a close-up, you’d see an active process of genesis of neurons. These neurons leave their location of genesis and migrate to the outer surface, and establish an orderly lamination pattern during this early period of life.
The majority of the brain cells that you have in your head right now, almost everything that you operate with, are brain cells that existed at the time you were born — and you have them your whole life. Those were generated in the second trimester, especially between week 13, and week 17 or 18. A gigantic number of neurons is generated by a novel zone that is specialized in humans — the outer ventricular zone. This zone is responsible for making the vast majority of neurons, so it’s very likely that it’s during this time that things are going wrong with this whole process of cell cycle regulation and cell number regulation.
In an effort to define the fine microstructure of what might be going wrong in autism, we got from exactly the same tissue that Maggie Chow worked with for the gene expression, and asked “how can we see the fine structure?” We got a panel of 25 markers, each marker that is specific for a layer or a specific cell type, or has autism candidate significance. And we asked the question, “are there normal layers or not in autism?” And this is what we discovered:
In autism, we find layer four running long, and then expression markers disappearing, and then coming back again, and markers for layer five running long, disappearing, and then coming back again. In these randomly selected, small blocks of tissue — one block per subject — roughly nine out eleven blocks — nine out of eleven autistic individuals showed this defect. A tenth individual showed not a loss, but in these zones a splaying out and a dispersion of markers, indicating a failure of normal migration, in patches. We also found this in temporal cortex in a small sample of two autistic and two controls, we saw an occipital cortex with no such abnormalities.
So it seems that probability of finding these patchy defects that are undoubtedly pathological is very high in the frontal cortex, and probably the area of the temporal as well. In novel three-dimensional reconstructions of each of these layers, we can actually visualize what’s going on.
Where there are patches of abnormally developed lamination, is there a reduction of neurons? The answer is “no.” So it’s not as though those patches are missing neurons, because there are a full complement of neurons there. It’s that those neurons either fail to fully differentiate and turn into their normal cell type, to establish normal connections and function, or they mis-migrated — and I suspect it’s probably going to turn out to be both. That it’s going to be a failure of cell differentiation and a migration defect.
For the first time, it’s possible to develop methods of three-dimensional reconstruction of the fine architecture of the human cerebral cortex to validate — no matter what the angle or section, that you have what you think you have.
We all know that brain development takes time. And what’s important to know is that at birth, near numbers are near normal in typical kids, but there’s not much connectivity — that connectivity takes place across the span of years. What I’ve shown you is evidence that has to do with roughly ages two years and out, autism neuron numbers, and control neuron numbers.
What we see in autism is either an excess proliferation, producing an overabundance of neuron numbers, or the excess might be due to a reduced ability to undergo naturally occurring cell death. Or it could be both. We don’t know which and our data don’t speak to that, although our data do suggest that it’s probably both.
Finally, our evidence shows that across time, there’s a prolonged period of apoptosis, removal and remodeling of circuits. In order to get back to where neuron numbers are supposed to be, it takes a very long time for the autistic brain. In the normal developing brain, this takes just a few months. In autism, it’s a couple of decades.
In the normal brain, life begins with a near-normal set of neurons and you’re good to go. The neurons are in there, they’re ready to make connections, and those connections will be driven by experience and learning. The [autism] system is different.
Prenatal life in autism begins, and it looks a little like this: there’s a period of excess numbers in autism, either due to excess proliferation or reduced apoptosis as compared to normal controls. Normal controls, typical babies, have to get rid of excess neurons as well — patches of pathological cortex exist. At perinatal life, neuron numbers have been reduced to near-adult normal levels, and the brain is ready to gain from experience and learning. The autistic brain is not.
We now know what the autistic child is faced with: a huge excess of neurons that preclude the possibility of rapidly taking advantage of the normal experience and learning environment that they come into. Furthermore, they’re faced with patches of defective cortex that may actively interfere with the capacity of even spared neurons to operate. Then, by the infant/toddler period, you have accelerated growth in the normal cortex, where connections are now being made based on learning experience, language is advancing, longscale connectivity is advancing.
In autism, they’re still at the starting point, they’re still at prenatal life in a certain sense, where they’re still trying to get rid of neurons and these patches — this is all my theory. I don’t want you to believe that my data *prove* this, but my data *suggests* this — and there are at least two other major alternative theories.
By childhood and adulthood, you have what John Morgan described — changes in clustering patterns, the presence of microglia that are wrapped around neurons and wrapped around processes. In the young adolescent and adult, there’s an active process of removal and remodeling. You have to remove and remodel before you can gain from learning and experience. And what are the chances of learning and experience and intensive intervention after all your circuits have been put into place? In disassembling hundreds of thousands of aberrant axons? Millions of aberrant neurons?
Early intervention works because it begins intensively to drive the effort for experience and learning to help make selections about the most adaptive connections out of this overabundance of possibilities. Sooner, rather than later, so circuits are being created; not waiting until later, when circuits have to be disassembled. It’s far more effective to undergo this early rather than later.
So why do some kids benefit and some kids don’t? I think what we should be thinking about is the genetics of recovery in autism, not just the genetics of cause. Removal, remodeling, and reorganizing circuitry is not about cause. The cause happened much earlier.
Treatment interventionists are operating in the domain of neural plasticity. Each and every autistic child comes into life not just with a defect that triggered their problem in the first place, but with varying possibilities for recovery based on the genetics that are not part of autism — genetics having to do with synapse remodeling and formation.
There’s a whole host of genes that have been affiliated with autism, that may well be genes that have something to do with the remodeling capacity in individuals who have [genetic] variants that are not suitable, or have loss of genes that are important. [Those individuals] are going to be at a greater disadvantage in recovery than those that do have genetic variants that improve the capacity to remodel the nervous system during this period of time.
The evidence we have suggests that prenatal life may not be a great place to detect biomarkers of autism, particularly if it turns out that you have in situ somatic mosaicism involving defective progenitor cells. There is no certainty from our data and our perspective that genetics or other biomarkers are going to be able to identify kids at certain risk for autism prenatally. At risk, yes, but at certain? It’s going to be a long time.
Postnatally is a different matter. You’ve got the [autism] system there, it’s a problem; it’s up to us to figure out how to fix it. We need to have a huge effort towards a new developmental neurobiology of autism that drives towards recovery. It’s not just genetics; it’s behavioral interventions early on! It’s early biomarkers detecting kids at high risk. It’s the suitable use of markers, hopefully by pharmaceutical companies that become interested in autism, to find molecular pathways that can be tweaked, that enhance neural plasticity. Then we will be making a difference for these kids.
Thank you very much.
Two quotes from Dr. Courchesne from the Q and A session following his talk:
- “We will learn more about ourselves through the study of autism.”
- “When we talk about autism as being part of a spectrum, it’s part of the spectrum of all of us. We are the same.”
*I hope this is the right link. -SR