By David Kirby
A new study written up in Science Daily (Mar. 2, 2009) and elsewhere says that researchers at the University of Southern California (USC) and Vanderbilt University have “identified a specific gene variant that links increased genetic risk for autism with gastrointestinal (GI) conditions.”
So far, reporting on the study, to be published in Pediatrics, has suggested that a polymorphism in the Met gene has been identified which, by itself, may explain both brain abnormalities and GI disturbances in children with autism spectrum disorder.
But there may be more to this story than that relatively simple explanation. Is it possible that these genetic variations produce susceptibilities to other environmental triggers, such as toxins? According to one of the lead authors of the paper, Patrick Levitt, PhD, autism cannot be explained by genetic factors alone.
Dr. Levitt made this comment when presenting his findings at the Neuroscience and Nervous System Disorders 2007 workshop on “Autism and the Environment - Challenges and Opportunities for Research,” of the Institute of Medicine.
(Please see: (HERE))
I hope everyone will look at these excerpts and read the full remarks (both are below) before drawing conclusions about these very significant findings.
STATEMENTS BY DR. LEVITT
1) Environmental factors “must” play a role in ASDs
“Gene–environment interaction is one of the unique properties of the brain. So, of course, regarding ASD, it is not genetic versus environmental, irrespective of whether you think there is a principal cause that is genetic or environmental. Because ASDs have at their core disrupted brain development, in terms of etiology, both genetic and environmental influences must play roles because this is in the basis for brain development.”
2) The Met gene is important for normal nerve and myelin development in the brain:
“It turns out that this gene is expressed in the brain during development and is important for a number of different processes, including cell migration, development of excitatory and inhibitory neurons, synapse formation, and myelination.”
3) The Met gene is also important for proper GI and immune function:
“Met is also involved in gastrointestinal repair, in immune response regulation, and some other peripheral functions that are consistent with the co-occurring medical issues that are described clinically for individuals with ASD.”
4) But manipulation of Met gene expression in animal models can change brain architecture and cause ASD behaviors
“We were studying Met in an animal model. The brain architecture changes we found when we manipulated levels of Met expression, together with long-term changes behaviorally, paralleled changes in ASD.”
5) Gene expression is controlled by transcription factors (proteins)
“Transcription factors are proteins that control how much of a gene is turned on in specific locations and at any particular time during development.”
6) Transcription factors in ASD-associated MET variants are disrupted by several toxins
It turns out that the transcription factor impacted by the ASD-associated variant in Met is SP1, which happens to be a transcription factor whose binding to DNA is disrupted by a number of environmental toxins
7) We need to learn where toxins impact gene expression in order to find the cause of autism
“Finally, if we are going to understand functional etiology of ASD, if we are trying to identify the genes that underlie risk, and we are searching for environmental factors that cause changes in brain development, we need to know where these candidate genes are expressed in the developing human brain, and where these environmental factors have their impact.”
Finally, in 2006, Dr. Levitt was quoted speaking about the same research. Noting that.
‘This variant is in the part of the gene that controls how much of the gene gets expressed…kind of like ‘volume control’ on a stereo.” He noted that people with two copies of the variant were 2.27 times more at risk for autism than the general population. People with just one copy were also at higher risk (1.67 times). “This is a relatively common variant, seen in about 47 percent of the population,” Levitt said. “So why doesn’t everybody have autism?” Because of environmental and other genetic contributions, he said. “Genes create a vulnerability that then gets coupled with some environmental disturbance — but right now, we don’t have any idea what those factors might be.”
FORUM ON NEUROSCIENCE AND NERVOUS SYSTEM DISORDERS
AUTISM AND THE ENVIRONMENT
Challenges and Opportunities for Research
WORKSHOP PROCEEDINGS - Board on Health Sciences Policy
INSTITUTE OF MEDICINE OF THE NATIONAL ACADEMIES- April, 2007
GENES AND THE ENVIRONMENT: HOW MAY GENETICS BE USED TO INFORM RESEARCH SEARCHING FOR POTENTIAL ENVIRONMENTAL TRIGGERS?4
Dr. Patrick Levitt
Dr. Levitt: I am going to provide for you a neurobiologist perspective on where we are in terms of genetics and what some of the opportunities are in terms of genetics and designing the kinds of research we might be doing to understand gene–environment interactions. The first slide basically depicts the fact that we all understand—complex genetic disorders are complex.
Complex genetic disorders are complex and what we are trying to understand are the combination of risk alleles, variations in gene sequences or in copy number of specific genes which, in combination, end up underlying risk or, in fact, directly perturb brain development that ends up generating the three core symptoms that are diagnostic of autism spectrum disorders.
You can see in the diagram that for any disorder, a combination of risk alleles may be correct, but there may be an intermediate phenotype rather than the features of the full disorder. We know that of the three major core symptoms that are used for an autism diagnosis, dysfunction in any one of these domains can run in families. There have been large twin studies to look at heritability independent of the autism diagnosis itself.
The diagram also shows that the correct combination of risk alleles might require specific environmental factors in order for the full-blown disorder to be expressed. There also are issues of incomplete penetrance where you may have the correct combination of genetic risk, but for reasons unknown, an individual has modifier factors that reduce the impact of the risk alleles. This means that one does not express the disorder.
So, I want you to keep something in mind. I take this from Daniel Weinberger, who studies schizophrenia at the NIMH and he makes this point, I think, very well. Genes are involved in the assembly of specialized cells to perform specific functions. Thus, there are no “social behavior” or “communication” genes. If we are looking for those, you might as well stop now because they don’t exist. Genes don’t know about social behavior. They don’t know about communication. What they know about are assembling tissues and cells to perform specific functions, and when there are mutations or changes in the sequence of those genes that affect function or expression levels, or differences in the copy number of those genes, we see alterations in the assembly of cells and the specific functions that they underlie.
So, what do we know from a genetics perspective? Well, there have been three approaches used: (1) linkage studies that look for excess sharing of genomic regions on chromosomes that track with the disease; (2) allelic association studies, where we look for excess sharing of alleles; this is accomplished by studying single nucleotide polymerphisms (SNPs), in which a single nucleotide is changed, or differences in microsatellite sequences at a single locus; and (3) a defined copy number variation (CNV), where we look for submicroscopic changes (thousands of bases, rather than macroscopically identified millions). CNVs thus are not obvious changes, such as chromosomal rearrangements, but submicroscopic changes that alter chromosome structure, which could be either deletions or duplications. Keep in mind that most of the chromosome is not occupied by sequences of bases that encode the transcript that will be translated into protein, but rather encode regions whose functions we really don’t understand, but we think may be involved in regulation of gene expression.
I want to mention here some of the previous and current caveats to what we know in terms of ASD and genetics. You need to keep these in mind as you read genetic studies to determine the degree to which you can rely on the findings and conclusions. First, there may and are likely ascertainment biases. This means that the subject population that has
been studied genetically may not necessarily be broadly representative, or perhaps they are broadly represented, but they represent one small domain of the spectrum. Second, until recently the sample sizes used in studies typically were small and underpowered.
Why is this a problem? Well, as my friend, Ted Slotkin, tells me, if you do enough comparisons, you will find something. For genetics, this means that if you try to find an association between many different SNPs and a disorder, eventually you will identify some relationship statistically—but one needs to correct for what we call Type 1 error, that is, false-positive results. There are debates regarding the best ways to correct, and many earlier studies may not have corrected at all, leaving us with nonreplicable findings.
Third, the accuracy and completeness of the diagnosis and characterization of the phenotypes are essential to understand who you are studying in terms of a cohort to be used in a genetic study. If this is not done at a high standard, an already heterogeneous disorder like autism becomes even more difficult to study genetically because the study population may be diluted with poorly defined subjects. Fourth, in the past there have been issues with technical quality control; that is, the quality of the assays used to identify SNPs and other changes. This is becoming a nonissue as technology advances.
Finally, and perhaps most important, there is for the most part a lack of assigned gene/variant function, in which the polymorphism does something to gene expression or gene function. Keep in mind that this concern comes from a biologist. One may identify a variant associated with autism, but if it is not a coding variation that would clearly change the coding of an amino acid, what does it do? You are stuck with that finding in terms of translating that to a biological substrate for the brain changes that may underlie the disorder.
So, what do we know about linkage? Well, the most recent autism genome project consortium identified a modest signal on chromosome 11p, and this is being followed up. In addition, by doing some data filtering, a few other loci seen in previous studies were seen, including regions on 2q and 7q. From previous work, there are in the literature dozens of other reports of linkage, but the bottom line is that with disease heterogeneity, as we have seen in schizophrenia, for example, and other disorders, the linkage signals are generally relatively small and there may be difficulty in replication from study to study. This is telling us something about the disorder, that there is locus heterogeneity. Thus, there are likely to be many different genes or combinations of risk alleles
that may underlie ASD.
Regarding CNV, there has been a lot of discussion over the last month about two studies, the AGP (Autism Genome Project) study and a study out of the Cold Spring Harbor group, essentially identifying somewhere between 8 and 10 percent of the individuals in their study having CNVs. The findings are exciting, as CNVs have been implicated in other disorders (e.g., certain cancers), but the findings are not without issues. There is no overlap in terms of chromosomal sites, as far as I could tell, between what was found in the Cold Spring Harbor study and in the AGP study and one does not know the biological significance yet. We are going to talk about that later perhaps, I think, in terms of what CNV might contribute to this disorder.
Regarding rare mutations, we know that there are loss of function mutations that have been identified in a single individual that was part of a genetic study, or even in individual families in which there is an autism diagnosis. I have listed some of those genes up there. The reason I list those is because it turns out that a number of those mutations are found in genes that at least biologically have some things in common; they are involved in synapse formation and function. Keep in mind, however, that rare mutations generally do not translate into genetic variations across large segments of the affected population. They are important in trying to understand the genetic contribution to the neurobiological disruptions.
Genetic syndromes with co-occurring ASD diagnosis have often been overlooked in the past. My friend Art Beaudet talks about these all the time. Disorders such as Fragile X, Rett, Angelman, and Timothy have a relatively high prevalence of co-occurring autism diagnosis. In addition, there are some common themes in terms of the neurobiological changes known to occur in each of these disorders, related to the changes in neural development. Keep in mind that genetically the causes are quite distinct from each other, but the high co-occurrence suggests that there may be many genetic routes to impact negatively the three core functions used to diagnose ASD.
The literature is also replete with reports of association of common risk alleles with ASD; that is, gene variants have been identified from standard association studies that give us some clues regarding the impact of common variants on genetic risk. I have listed some of those on this slide: (1) nonfunctional risk alleles, meaning that there has been a change in the sequence of the gene, but we don’t know what that sequence change means. I have listed some of those genes up there. The neurexins, the GABA beta 3 subunits and Gral-2 and some risk alleles that have been identified, but not necessarily replicated in every study; and (2) functional risk alleles, that is, variants that have been identified that either change the function of the gene product or change how much of the gene product is actually produced. The promoter region of the reelin gene is one example, and I have placed a red circle around Met, a finding from my laboratory that I will tell you about in a moment.
I have posed some questions related to the influence of genetics on autism expression that might also be retitled “Gaps in Knowledge” (1) How much of genetic risk is due to direct impact of mutations on brain development? (2) How much of what we are talking about in terms of genetic influence is actually the combination of genetic mutations changing the trajectory or course of neural development in wiring the brain up? (3) How much of the risk is due to direct impact of mutations on peripheral functions, that is, other organ systems that influence brain development? I raise this as a possibility because we know that peripheral organ development and brain development are linked physiologically. (4) How much of the risk is due to genetically established sensitivities to environmental perturbations? We know this exists experimentally, but we really don’t understand it in the clinical population. (5) How much of the phenotypic heterogeneity of individuals with ASD are influenced combinatorially through genetic and environmental factors? That can be viewed as my red herring question.
So, here is my concept of where we are with understanding autism brain pathophysiology. I am being facetious, but that is a thimble in case you didn’t recognize that blurry image. In essence, we know very little about the changes in brain development and brain organization that underlie ASD. That is a real problem in trying to understand the causes. Genes, environment, or both? How can you answer any of the questions I posed without knowing what exactly is disrupted in terms of brain architecture and development? Part of the problem, in my opinion, has been that the gene–environment debate has been held in isolated silos, that is, separated disciplines in which there is rare exchange of ideas. The silos, or disciplines, need to interconnect. This harkens back, and we talked about it on the conference call among presenters, to when developmental neurobiologists spent an enormous amount of time trashing each others’ work because one was either in the “nature school” or the “nurture school” regarding brain development. Of course, that was silly because we know that the brain is built through a genetic blueprint that takes information from the outside world and utilizes it to direct the developmental course to wire up circuits. This gene–environment interaction is one of the unique properties of the brain. So, of course, regarding ASD, it is not genetic versus environmental, irrespective of whether you think there is a principal cause that is genetic or environmental. Because ASDs have at their core disrupted brain development, in terms of etiology, both genetic and environmental influences must play roles because this is in the basis for brain development.
So, here is a concept regarding what we might do to address mechanisms: Translational approaches that incorporate multiple technical strategies. There are a number of different strategies in which we are trying to link these domains experimentally. One approach is to focus on neurodevelopmental genes that have been characterized for altering the assembly of circuits that are likely to be disrupted in individuals who develop autism; it does allow investigators to move freely between animal models in which the biological functions of the genes are studied, and going back and working with human geneticists to try to determine whether there are meaningful relationships that would make sense in terms of variations of that gene that might underlie partial risk for ASD.
One also can begin from human genetic research data and develop model systems that probe biological functions, trying to make sense in terms of what has been identified as a variant associated with the disorder that carries genetic risk. I would suggest that it doesn’t necessarily help to knock out a gene in a mouse if the variant that has been identified in the human genetic research is not a complete null, but rather a variant that alters protein function or levels of gene expression. Genetic knockout studies may generate some very interesting biological findings, but these may not necessarily be relevant to the pathophysiology of ASD. Of course, with model systems, such as genetically engineered mice, you can do experiments. You can manipulate the system both genetically and environmentally at different developmental ages. I have diagrammed an example, in which one can expose genetically manipulated mice to different environmental factors that we know change the course of trajectory and development. The impact of exposure may be influenced by genetic variation and you can design experiments to do this in developing model systems.
So, I just want to highlight for the last minute or two what we have done in our laboratory. There is only one data slide, and it summarizes work published in 2006 in the Proceedings of the National Academy of Sciences.
We took the approach of studying the role of a gene in brain development and then extending these to human genetic studies for several reasons. We were examining the role of a tyrosine receptor, Met, in cerebral cortical development. Met actually has been the focus of thousands of scientific studies because its dysregulation is implicated in certain kinds of cancers. It turns out that this gene is expressed in the brain during development and is important for a number of different processes, including cell migration, development of excitatory and inhibitory neurons, synapse formation, and myelination. We were studying Met in an animal model. The brain architecture changes we found when we manipulated levels of Met expression, together with long-term changes behaviorally, paralleled changes in ASD. We also realized the Met is located under a linkage peak on chromosome 7 in humans, a region implicated multiple times in studies of ASD.
The major finding is that we identified an SNP in the 5′ region of the gene that controls how much of the gene is expressed. We showed experimentally that it reduced how much of the Met gene is expressed, and we believe the mechanism for this is due to the 5′ SNP It turns out that the transcription factor impacted by the ASD-associated variant in Met is SP1, which happens to be a transcription factor whose binding to DNA is disrupted by a number of environmental toxins
Thus, the Met variant that is strongly associated with ASD actually had a functional outcome. It changed how much of the gene was actually produced. Met is involved in brain development, but we also thought more broadly about this when we were debating about doing the human genetics studies. Met is also involved in gastrointestinal repair, in immune response regulation, and some other peripheral functions that are consistent with the co-occurring medical issues that are described clinically for individuals with ASD.
We spent a lot of time with clinicians to talk to them about whether this made sense because it is not a small number of children who have co-occurring medical conditions. Though still unsettled, it may be a relatively large number. These detailed delineations of the population are telling us about disorder etiology and perhaps even the biology as well.
It turns out that the transcription factor impacted by the ASD-associated variant in Met is SP1, which happens to be a transcription factor whose binding to DNA is disrupted by a number of environmental toxins. So, here one can see that the possibility of combining environmental toxin work with this variant in a humanized mouse model, for example, or introducing the humanized mutations in cells, opens up the possibility of studies that examine combined genetic and environmental influences.
So, for one example, we have actually shown that if you expose cells that have either the G or C (ASD-associated) variants of this gene to BaP, which is a common environmental toxin, levels of gene expression are reduced quite dramatically for both the common and ASD-associated variants. Keep in mind that the common SNP (G) results in more than double the amount of gene transcription in the cells than the ASD-associated variant. I’ve added here a hypothetical threshold for when a disorder is expressed. If the toxin reduces levels of Met expression for both the G and C variants, but the C variant starts out lower, the environmental exposure will result in even lower levels of expression that reduce below the threshold. In this example, even with BaP exposure, expression of the gene with the G allele still does not drop below disorder threshold. Thus, BaP does not directly cause the disorder, but has differential effects due to genetic variation.
So, what do we need to do in the future? I’ve listed some suggestions here. We need to increase subject ascertainment, characterizing populations in great detail, which will allow geneticists, psychologists, and neuroscientists to stratify groups more accurately to determine if certain phenotypes are associated with specific genetic variants, including SNPs, CNVs, and other genetic changes. Given that we all agree that we need to be very careful about how we phenotype in doing the genetic studies, it simply doesn’t make sense to start out with a cohort of 1,000, because by the time you stratify based in different characteristics of ASD, or even life history, the study will be underpowered.
Deep sequencing to identify more functional variants will be important to pursue. If we are going to translate the genetics to more than just associations or statistical arguments, we have to translate the findings to biologically relevant changes. Thus, functional characterization of the variants is a very high priority.
There needs to be continued wise investments in model systems that will allow us to pursue gene–environment influences more rigorously than can be done in human populations.
Finally, if we are going to understand functional etiology of ASD, if we are trying to identify the genes that underlie risk, and we are searching for environmental factors that cause changes in brain development, we need to know where these candidate genes are expressed in the developing human brain, and where these environmental factors have their impact. There is a difference between mouse and human brains, and it is essential to keep in mind that one cannot always extrapolate findings between species because of fundamental differences, particularly related to brain areas that simply are not represented in the mouse, but which may be at the heart of ASD. For this type of information, there is an enormous gap in terms of understanding where key genes might be playing a role in neurodevelopment, and how their perturbation may impact the core features of ASD
David Kirby is author of Evidence of Harm, a founding contributor to Huffington Post and a contributor to Age of Autism. His next book, “ANIMAL FACTORY” – about the impact of industrial livestock production on our health and the environment – will be released within the year.