Mapping the Molecular Pathway of Autism

John P. Hussman, Ph.D.
All rights reserved and actively enforced.
Reprint Policy

It was well past midnight, and the only light in the house glowed from my study, down the little hall from the kitchen. Stacks of papers and notes were balanced in the narrow space between my computer and the edge of the desk, and my scribbled drawings, which could have been mistaken for the work of a first-grader, littered the other side of the keyboard. "Just a few more - this can't be a coincidence." Now and then, I would topple the papers over looking for a word or phrase I had never seen before in my life, and now had begun to see every few minutes. Another hour. "This can't be a coincidence." I felt my head jerk back, unconsciously catching myself from falling asleep and smashing into the keyboard. One by one, again and again. "This can't be a coincidence."

When the moment comes, you're supposed to be sitting under a tree. Smelling the fresh, slightly bitter scent of cut grass in the breeze. The apple falls, and you yell "Eureka!" Or maybe you jump out of the bath and go running through the streets. When the pieces fit together, you expect to dance. But my only thought as I sat in near-darkness was of this innocent little boy, trying to move his body, his mouth, like the other children. Crying in the corner of the gymnastics center at a little girl's birthday party, hands cupped around his ears because it was too loud, while movement and speech and friendship came so effortlessly to the children around him. Eventually, they would choose not to invite him at all.

My breath grew heavier as years of memories shot through my mind, all the tiny victories that other parents so easily take for granted. The giggle he lets out when his attempt at a whole sentence is understood. Rocking back and forth smiling when he types a new idea independently. I leaned my elbows onto the edge of the desk and stared at all the papers and scribbles scattered about. The pencil drawing in front of me seemed more and more real. I remembered the sight of him bundled up in his crib, before we knew, and caught myself whispering out loud - "his brain... his brain.. my baby." My back teeth locked together as wave of heat raced up my face and my steady breath broke into random heaving. I pressed my palms against my jaw, and as my fingers covered my eyes, sixteen years of tears poured into them.

But parents of children with autism don't dwell in tears for very long. Sooner or later, preferably sooner, we realize that our kids are far more capable and intelligent than they may be able to demonstrate, and that they needn't be anything other than who they are to be loved, accepted and presumed to be competent. Though they could use some assistance learning how to translate intent into action, and would appreciate if we turned down the volume and dropped the baby voice, they're more like us than their challenges trick us into believing. They don't lack empathy nearly as much as we do. Just because social interaction may be overwhelming, it doesn't mean that they don't want friends - they just need it a little bit more on their terms. Jamie Burke, a family friend with autism, says it simply - "The soul must be loved as it is." Eventually, if we're really fortunate, we realize that the things we like the most about ourselves are there because that person with autism is there.

After a few minutes, I splashed some cold water on my face. I looked up through my bloodshot eyes, and squinted at the mirror like I was staring down those molecules, proteins and autism genes. My boy's dad imitating Clint Eastwood, in front of the sink in the middle of the night. "We've got you little bastards now."

Last week, in collaboration with the Hussman Institute of Human Genomics at the University of Miami, we reported a new molecular pathway for autism. The study, entitled "A Noise-Reduction GWAS Analysis Implicates Altered Regulation of Neurite Outgrowth and Guidance in Autism ", was published in the open-access, peer-reviewed journal Molecular Autism (the paper is immediately available as a PDF, with diagrams and tables following the references. The formatted publication version should be online within a week or two).

The UM School of Medicine published a nice piece about it: Economist Revolutionizes Mapping of Genetic Pathway for Autism.

Why would an economist be doing genetics research?

At this point, some background is probably a good thing. Just over two years after my son JP was born, he was diagnosed with autism. That's shorthand, of course, because as any autism parent knows, the period between feeling that something is wrong and the time we accept a diagnosis involves such an endless parade of doctors, therapists and other professionals that it might as well include floats, Disney characters and a marching band. As the son of two physicians, having "jumped the track" from medicine in college, I decided that there was something inherently wrong with a world where parents were essentially told "Sorry, but your child has autism. We don't know what causes it, so we have very little advice to offer, except for early intervention - good luck with the cost of that - and a few medications. There are some charlatans and snake-oil salesmen down the street you might look into. And thanks for stopping by."

I spent four years at Stanford for my doctorate in Economics. JP is nearly 17. If you're passionate about something, you can learn a lot in that amount of time. By 2001, I had written a piece for the Journal of Autism and Developmental Disorders (Suppressed GABAergic inhibition as a common factor in suspected etiologies of autism). GABA is the main inhibitory neurotransmitter in the brain, and reduces the tendency of neurons to "fire" too easily. The basic idea of the GABA paper was to connect multiple lines of research which, taken together, suggested that autism might involve an imbalance between excitatory and inhibitory activity in the brain. A couple of years later, a genetics team led by Dr. Margaret Pericak-Vance (now at the University of Miami) identified an association between autism and a number of GABA receptor subunits. Perhaps because Dr. Pericak-Vance (Peggy) is a parent with her own special mission - visit www.jjvance.org - we instantly became friends, and have collaborated ever since.

I spend almost all of my time in one of three activities - finance, family or charitable projects through the Hussman Foundation. In recent years, much of the foundation's research has been centered on autism. Meanwhile, the finance research has been centered on "ensemble methods" to integrate the information from multiple data sets, and to better measure both risk and uncertainty*. As it happens, statistical methods can be adapted to approach difficult problems in both genetics and finance. So as we developed various approaches to integrate multiple data sets in our finance research, it was natural to extend those methods to deal with genetics data.

* In case separating "risk" from "uncertainty" seems like a distinction without a difference, risk is the extent to which possible outcomes are spread around a known average outcome, given that you know which "state of the world" you're in. Uncertainty is the extent to which the "state of the world" is itself in doubt, so even the average outcome is in question. For example, risk measures the chance that you'll roll something other than 7 given that you know that two six-sided dice are being thrown. Uncertainty is the possibility that the dice themselves might instead have eight, twelve, or twenty sides without your knowledge. Similarly, risk is the probability that the market will decline, given that you know that the economy is in a typical post-war economic cycle. Uncertainty is the possibility that the true economic environment might correspond more closely to the Great Depression than to the post-war period. The difference isn't trivial.

Genome-Wide Association Studies

Our genome is made up of 23 chromosomes, of which each of us have 2 copies, one from each parent. Along each chromosome are segments that carry the genetic code for various proteins ("coding exons") as well as regions in-between that incorporate regulatory elements and may have other unknown functions. Each gene is like a long message in Braille or Morse Code that provides instructions on how to make a particular protein. Along the genome, there are millions of individual locations where normal variations occur between people, and these "single nucleotide polymorphisms" or SNPs can be used as signposts to mark specific points on each chromosome.

A genome-wide association study (GWAS) analyzes how these SNPs are transmitted to people having conditions such as autism, Alzheimers disease, multiple sclerosis, and so forth. Basically, a GWAS is like a huge coin-flipping experiment involving hundreds of thousands of SNPs, to find out whether any of the coins are flipped to the person with the disease more often than can be explained by random chance alone.

For instance, suppose at a given SNP, there are two "flavors": 1 and 2. Your dad might be a 1/2 and your mom might be a 1/2. But if you've got two siblings with autism, both who received 2/2 and you're typical and received 1/1, and the same pattern was observed in a significant fraction of the families under study, you would think of that SNP as a potentially unfair coin, and flavor 2 as the "risk allele" at that location. So you would look at which gene or genes are very close to that SNP, and what functions (if known) are reported for the associated protein. If the evidence was strong enough, you would also follow up by sequencing every single location in that region, to see if there are any variations in the DNA that have the effect of changing the way the protein is made (which would be a "functional variant").

There are at least two statistical problems with GWAS studies. One is that since you're looking at hundreds of thousands of locations, some are going to come up as "significant" by pure chance alone. The other is that if you look at GWAS results one marker at a time, you may be throwing away information. That's because if you think about the structure of the genome, it's clear that those coin flips aren't independent. They're tied together by a string of genetic material, and the closer they are in proximity, the greater the correlation you'll typically see. Based on the strength of these correlations, we can define short stretches of DNA that tend to be transmitted more or less as a single unit from parent to child, without much "recombination" from generation to generation. These segments are called "LD-blocks."

Examining this situation, it seemed clear that we should be able to "amplify" the statistical signal provided by any given SNP by taking account of whether there were any other signals in the immediate neighborhood of that SNP. Furthermore, if you looked at more than one dataset, you could also amplify signals where you observed an association spike in the same LD block of two different sets of data. So instead of looking at the data marker-by-marker, you would give greater weight to signals in regions where you observed strong local correlation among "unfair coins" at nearby markers and across data sets. The objective of that statistical filtering would be to produce a reduced set of higher confidence candidates. If you're looking for a needle in a haystack, it helps to have a smaller haystack.

After writing the computer code for the new method, I first tested it simply by simulating millions of correlated coin flips. With strong initial results, we applied it to simulated genetic data, and then to actual family data. Dr. Ren-Hua Chung, one of the nation's brightest researchers in statistical genetics, tested the method using a wide range of assumptions about disease transmission (recessive, dominant, multiplicative, etc), and helped to develop additional procedures to rank the LD-blocks that we identified. Finally, we applied the method to two data sets in autism, one involving hundreds of families participating in research coordinated by Dr. Pericak-Vance at the University of Miami, and the other using data from the Autism Genetics Resource Exchange (AGRE). Dr. Anthony Griswold, a skilled post-doctoral researcher at Miami, did an enormous amount of bioinformatics work, including mapping the resulting LD-blocks to the corresponding candidate genes.

Other members of the UM research team included James M. Jaworski, M.S.; Daria Salyakina, Ph.D.; Deqiong Ma, Ph.D.; Ioanna Konidari, M.S.; Patrice L Whitehead, B.S.; Jeffery M. Vance, Ph.D., M.D.; Eden R Martin, Ph.D.; Michael L Cuccaro, Ph.D.; and John R Gilbert, Ph.D.. Dr. Jonathan Haines, director of the Center for Human Genetics at Vanderbilt University, also collaborated in the study. Their combined expertise in genetic research, clinical evaluation, genotyping, molecular analysis, statistical methods, laboratory operations, study design and other areas continues to advance urgent research in autism and other conditions. I shouldn't leave out our optimization and data analysis expert at Hussman Funds, Dongmin Kim, who never fails to find unimaginably clever ways to speed up my computer code by two orders of magnitude.

Mapping the Molecular Pathway of Autism

The question then became: "What do these candidate genes do", and "Do they operate in a coordinated way?" We didn't set out to look for "the" autism gene - the evidence suggests that there is no such thing. There may be hundreds of autism susceptibility genes. Any particular gene may have weak effect in a large number of individuals, or a strong effect in a small number of individuals, but in either case, the overall signal you get at the population level is almost always weak. So instead of trying to find one elusive gene, we were interested in whether there might be some biological "pathway" that was enriched with numerous candidate genes. When a specific biological pathway is involved in a disease, genetic signals that are relatively subtle on their own can combine to form a composite signal with very high statistical significance.

Taking a "pathway" approach is also important from a clinical standpoint. If you try to approach autism gene by gene, you may only be able to help a tiny fraction of people at a time. On the other hand, if you can identify the biological pathway that is affected in autism, you can look for more general ways to rescue that pathway and potentially improve many more lives.

The first task was to go through the published literature to identify the known functions of the candidate genes, as well as noting structural features (such as "binding domains") that provide clues about how they might interact with other proteins. Manually annotating gene function is like a huge game of Concentration. You flip a card over, read the clue, and then turn it back over. Then you flip another card over and try to remember where you saw that clue before.

In examining the candidate set, there were certainly some genes relating to synapses and neurotransmission, but that wasn't where the evidence clustered. Instead, numerous candidate genes were involved in functions such as "cell adhesion," "phosphatidylinositol signaling," "Rho-GTPase regulation," and had binding domains with names like "pleckstrin homology," "cadherin," and "fibronectin type-III." Working through the candidate list, the names, functions and binding domains quickly became old friends, though recognizing some of them was still easier than pronouncing them.

Once the pieces initially came together, the evidence for a coherent pathway in autism kept accumulating. In the office of the Hussman Foundation, there's a 10-foot, two-sided whiteboard. On one side, my daughter has drawn a little heart and the words "craziness over there" next to an arrow. On the other side is a diagram with scores of proteins along with numbers referencing published studies that link them together. It is a pathway that regulates the way that neurons sprout protrusions - axons and dendrites - and navigate them to their appropriate locations. For publication, Katherine Savage, a talented painter and graphic artist who did all of the logo work for the Hussman Funds several years ago, generously helped to make the final, more simplified diagram presentable (I realized later that Kate used the Hussman Funds color scheme, which was a nice touch).

When we think of the brain, we immediately think of synapses, because they capture our imagination and and literally pulse with activity. In contrast, the cytoskeleton - the little set of toothpicks and Lego blocks that maintains the shape of the cell - just doesn't seem very interesting. But in fact, the cytoskeleton is dynamic - it's constantly being reorganized in response to cues that are received by numerous receptors and proteins - particularly adhesion molecules - at the surface of the cell. When nerve cells need to extend, to protrude, and to move those protrusions in this direction or that in order to create functional networks, it's cytoskeletal remodeling that gets the job done. Little rods made of actin or microtubules push forward under the membrane like a pencil under a sheet of rubber, while the membrane recycles and adds new material to keep up with the protruding neurite as it navigates toward its appropriate synaptic partners. That is the pathway - neurite outgrowth and guidance - that we propose is altered in autism.

We analyzed the same set of proteins using an objective functional annotation tool - the National Institute of Health Database for Annotation, Visualization and Integrative Discovery (DAVID), which provided consistent results. While we observed some enrichment related to neurotransmission (particularly excitatory-inhibitory balance), the highest ranked pathways implicated functions like cell adhesion, neuron projection, cell motion, and neuron morphogenesis (shape). Not all of the proteins had literature about their function, but of 183 autism candidates that were responsible for the 20 highest-scoring functional categories, 41.5% had established roles in the regulation of neurite outgrowth and guidance, including 51.3% of the "cell adhesion" candidates. Notably, all of this was consistent with neuroanatomical evidence. When researchers have examined actual brain tissue in individuals with autism, one of the most consistent findings in the literature has been alteration of dendritic structure. It just wasn't clear whether that was a cause or an effect.

The argument is not that people with autism have profound differences in the way their brains are "wired." People with autism are not much different from everyone else - the alterations are most likely subtle.

In my view, using the word "altered" is vastly better than the word "abnormal." This isn't simply a polite euphemism. The word "abnormal" is too often extended from describing some aspect of a person to labeling the person as a whole. The same thing happens when people are classified as "high functioning" or "low functioning" on the basis of their ability to physically demonstrate knowledge, initiate speech, or accurately move their bodies. As a result, we miss the real person and stop looking for their competence, because we've already put them in a tidy box.

As our friend Larry Bissonette says "We are more like you than not." We know from people with cerebral palsy that even difficulties initiating movement used to result in an assumption of mental retardation, yet more recent evidence is that a large proportion of CP individuals have normal or above-average intelligence, including many non-verbal individuals having access to technologies such as eye-gaze tracking that allows them to manipulate keyboards and other devices. Imagine how easy it would be to take Helen Keller or Stephen Hawking at face value. Presuming competence - then going about the work of finding it - is always the least dangerous assumption. Paula Kluth, who is also one of the directors of the Hussman Foundation, has some great resources for parents and teachers for that.

What's exciting is that even in the few weeks since the paper was finalized, we've seen new autism studies that are consistent with the pathway we've proposed. For example, the Wellcome Trust just released a study reporting microdeletions in the cadherin 8 gene in two families with autism. CDH8 is one of our top ten hits, and is known to regulate axon targeting in the hippocampus. Similarly, a joint radiology study from major universities including Stanford, Yale and MIT reported a study measuring central white matter volume in individuals with autism, with results suggesting alterations in long-range connectivity.

As for GABAergic inhibition, one intriguing point of intersection - In 2003, a researcher named Manuel Casanova identified alterations in the wiring or "microarchitecture" of cell minicolumns in the cortex of autistic individuals - specifically among GABAergic fibers that arrange to form inhibitory local circuits. So it may be that alterations in connectivity also affect the balance between excitatory and inhibitory tone. From my perspective, that's obviously an interesting possibility.

Implications

Does this research mean that we've found a cure? No - rather, it gives us a clear focus, what Peggy describes as a "pipeline" of research, that should speed new discoveries. Frankly, some friends with autism aren't even comfortable with the word "cure," because they feel it implies that people with autism aren't acceptable as they are. That said, our hope is that by targeting the molecular pathway of autism, it will be possible to improve the quality of their lives. Though neural circuitry is established early in life, there's a great deal of evidence that neural development and "plasticity" continues through adulthood. Further research can pursue ways to influence or rescue compromised pathways, even if we can't address the probable multitude of individual autism risk genes one-by-one.

Another question is this - how does one explain the apparent epidemic of autism, with growing numbers of affected children? Well, my personal view is that we've gradually broadened the diagnostic criteria for autism, and also that more cases are being identified that might have previously gone undiagnosed. You know how if you double the radius of a circle, you quadruple the area? Well, back in the 1970's, the diagnosis of autism was generally restricted to what would be called "Kanner-type" individuals (rocking back and forth, twirling objects, lacking speech, and tragically, usually institutionalized). The diagnosis of "autism" is now used largely interchangeably with "autism spectrum disorder" or ASD, which includes a broad range of developmental conditions such as Rett syndrome, Asperger syndrome, and PDD-NOS (pervasive developmental disability, not otherwise specified). These individuals have legitimate challenges and are equally deserving of educational services and other intervention. But coupling a broader set of diagnostic criteria with better identification and ascertainment of affected kids probably accounts for much of the increased rate of autism diagnosis, in my view.

As for environmental factors, the controversial medical study suggesting involvement of MMR vaccines in autism was retracted a few weeks ago because the data were falsified and patients were misclassified. Though I respect the beliefs of other autism parents too much to completely dismiss the concern, my strong view is that if there is any involvement of vaccines at all, it would involve only a tiny sliver of cases. It's far more profitable to focus research on broader autoimmune mechanisms than to pursue MMR vaccines as any significant factor in autism. Just as the immune system can attack the cells of the pancreas in type-1 diabetes, and the myelin of nerve cells in multiple sclerosis, we can't rule out the potential for autoimmune responses that target neurite-regulating membrane proteins in a way that would insult the developing circuitry of the brain in some number of individuals with autism. Still, the evidence for a genetic component is far too strong to view autism as primarily environmental in origin.

In any event, there is still a great deal of work to do in sequencing gene candidates to identify functional variants (we have a large-scale targeted sequencing project already underway at the University of Miami). Similarly, efforts are underway to investigate alterations in the "neuroconnectivity" of people with autism using a variety of methods. My hope is that the research in this direction will advance quickly.

A decade ago, there were obscure mathematicians working on arcane problems that few people cared about - like how to compute the inverse of a massive sparse matrix. Then Google came along, and these guys suddenly realized that they were experts in internet search. With any luck, researchers with expertise in very specific fields of developmental neuroscience - particularly neurite regulation and growth cone function - will suddenly recognize that they may be experts in autism.

There's a lot of work ahead, but we're making important progress.

Market Climate

Since there's no natural transition back to discussing the financial markets here, imagine a huge cartoon foot descending while the British narrator from Monty Python announces "...and now for something completely different."

As of last week, the Market Climate for stocks remained characterized by an overvalued, overbought, overbullish, rising-yields syndrome that has historically been quite hostile for stocks, but with what I've called "unpleasant skew" - a seemingly relentless series of slight, marginal new highs, typically followed by an abrupt vertical plunge that wipes out weeks or months of progress in a few sessions. Our long-term shareholders have seen this before, and have seen how it ends, but in the meantime - it is what it is. The present instance has been compounded by some "basis risk" in that the strongest stocks have been speculative, richly-valued, highly leveraged and cyclical companies that we tend to avoid in any event, but our stock selection record speaks strongly enough for itself on a long-term basis.

While it's not certain that the present overbought conditions will be cleared immediately, it's notable that the S&P 500 hit a new high on Friday morning, then promptly erased nearly all of its gains since January 3rd in one session. While the overall Climate is still negative here, we took the opportunity to cover about 15% of our short call option hedges, leaving slightly out-of-the-money index puts in place. In the event that the market remains in this syndrome for a longer period of time, and speculative, cyclical and low-quality stocks remain in favor, we expect to experience less short-term discomfort. Overall, we continue to have put option coverage on 15% of our holdings, and a full hedge on our remaining investment position, so there's a good safety net in place if the market breaks substantially.

Strategic International Equity also remains well hedged here. The Strategic Total Return Fund continues to carry a duration of about 1.5 years in Treasury securities, with about 8% of assets in precious metals shares, established on recent price weakness in that sector.

I remain convinced that QE2 and the speculation that comes with it is based far more on rhetoric than on any demonstrable or historical cause-effect relationship between the monetary base and the financial markets. In recent weeks, we've hopefully provided sufficient data and analysis to support that evaluation. If central banks had the power to prevent bear markets (we've seen two plunges of over 50% each in the U.S. over the past decade), we simply would not see them, because they nearly always involve economic turbulence. Like the story by Hans Christian Anderson, Ben Bernanke looks to be parading about in the Emperor's New Clothes. One would think that investors would learn something from Japan's quickly ineffective experience with QE.

Meanwhile however, there's certainly no denying the huge dose of "animal spirits" in recent market action. This is encouraging to trend-followers, but purely technical methods are not difficult to test, and they tend to be abruptly blindsided by the spike declines that usually end the "unpleasant skew" of overvalued overbought, overbullish, rising-yields conditions. Indeed, in postwar data, simply avoiding market risk during this syndrome of conditions boosts the cumulative total return of a simple 200-day moving average strategy by over 68%, while cutting the maximum drawdown in half. The required sentiment data isn't available for earlier periods, but popular moving-average systems typically produce a maximum drawdown loss exceeding 40% in Depression era data unless additional criteria such as valuations are added as a filter.

If we can clear some component of the present syndrome - most likely the overbought or the overbullish piece - and market internals don't break down considerably while doing so, we'll have the latitude to accept a moderate exposure to market fluctuations. Probably with a safety net using out-of-the-money put options, but greater exposure nonetheless.

As shareholders who regularly read these comments are aware, we've recently introduced some additional methods to expand the range of Market Climates we identify. This will allow us to accept moderate, transitory exposures more frequently as market conditions change over time, even in persistently overvalued markets. The methods are based on research over the past two years to integrate the information from multiple - possibly competing - data sets. I've regularly discussed the challenges that resulted from the conflict between post-war and post-credit crisis datasets, which had enormously different implications - particularly in 2009. Taking a weighted-average was inadequate. Suffice it to say that the expanded methods sensibly address that "two datasets" issue. Given that the fundamental issues in the credit markets have been obscured to a much greater degree than they have been resolved, I expect that this will prove useful in the coming years.

Our objective is to outperform our investment benchmarks, with smaller periodic losses over complete market cycles (bull peak to bull peak, bear trough to bear trough). Though we've achieved that, it has been challenging to achieve positive absolute returns in the Strategic Growth Fund during the most recent cycle. We do remain ahead of the S&P 500 from the 2007 bull peak to the most recent peak, and our worst interim loss was a fraction of the market's interim loss. In any event, the objective that matters most is to achieve strong absolute returns over the long-term. That continues to be our focus.

A few people have asked things like "Why talk about missed upside when the guys who were bullish right before the crisis act like it never happened?" or say "You don't need to apologize for being defensive here." But this sort of discussion is essential. It would be far easier - and might briefly feel much better - just to jump on in. So shareholders need to completely understand why we aren't doing so. If you're going to ask people to trust you, you have a responsibility to tell them why - with evidence to back it up; to be open about how you're responding to challenges; and to make sure that their concerns are also yours (which is why nearly everything I have is invested in the Hussman Funds).

We can't always make our strategies comfortable over the short-term, but that's why we regularly do "anti-marketing": Don't invest in the Hussman Funds if you feel a strong need to track the market over periods shorter than a complete market cycle. Please. We don't try to "fix" short-term performance by abandoning our discipline. Our objective focuses on long-term, absolute returns - specifically, to outperform the market, with smaller periodic losses, over the complete market cycle. I have no particular expectation that we'll match or outperform the market over the short-run, but as always, we continue to confidently pursue our long-term objectives.

A final note to our shareholders. From my family, and especially JP - thank you.