Sunday, January 30, 2011

has holy grail molecular pathway for autism been identified?

An old acquaintance of mine sent me this interesting piece by economist and mutual fund manager, John Hussman, who has an autistic son and apparently moonlights as a genetic researcher with some other persons. The research involved genome wide sequencing of two different datasets, using a sophisticated computer algorithm that I don't really understand that studied what were called single nucleotide polymorphisms where there are normal variations in the genes of various people some of these can have risk alleles which are genes or a cluster of genes which can cause or predispose someone to conditions like Alzheimer's disease, diabetes, or autism.

From what I understand of the article, they were saying that there are so many possible locations that could result in genes which cause alterations to proteins or instructions on how to construct neurons, etc., it is infeasible to examine them all by conventional methods. However, many of the SNP's that could cause autism might be in very close proximity to one other and the problem could be confounded by a large "noise to signal ratio". The researchers Hussman was associated with apparently used an algorithm which helped with the signal to noise ratio in finding candidate alleles that could cause or predispose someone to autism. Two different data sets, one from the University of Florida and another from the autism genetic resource exchange were used with the idea is that you might be able to find locations for the same candidate genes in two different data sets, then this might make a clearer picture for a possible genetic etiology of autism.


I have written in the past about the simplistic statements that Simon Baron-Cohen and Temple Grandin and perhaps others have taken towards the genetics of autism in that perhaps there is only one or a few "autism genes" that have stayed in the population due to a protective effect and are responsible for every invention from the spear to the cell phone.

As Hussman correctly states, the problem with this thinking is that instead of a few genes, there are probably hundreds of genes that operate in concert with each other which could have a small effect on many individuals or a large effect on a few individuals. But what was intriguing is that these hundreds of genes could operate on a common biological pathway, which might make an understanding of the etiology of autism clearer.

With this computer algorithm, the researchers claim to have found a biologic pathway which could explain at least some aspects of autism. The biologic pathway regulates how axons and dendrites in a neuronal brain cell are formed and guides them to the appropriate places. For those who don't know, brain cells communicate with each other chemically and electronically by sending electrical pulses down what is called the axon of the cell. Then a chemical is released into a space called the synapse which is received by a dendrite of a neighboring brain cell. If something is amiss in genes that regulate this, then cells won't be able to communicate with each other properly and this lack of adequate transmission could cause the debilitating features of autism.

Though admittedly I am far from a science expert, one problem with this research that I see is that autism is likely not 100% genetic, since though we have high concordance rates in identical twins, it is not 100%. Likewise, as I remember correctly some studies have shown fraternal twins have higher concordance rates than siblings who are not twins; it would be the same if autism were 100% genetic. So, there may be something else, besides abnormalities in single nucleotide polymorphisms such as some environment insult during the prenatal period that could cause a person to become autistic, though it is possible this pathway found by Hussman and the others might predispose a person to autism.

I suppose Lorene Amet, a prolific commenter on autism's gadfly who has a doctorate in molecular biology can critique this research and correct me where I am going wrong if anywhere. Or possibly someone else better trained in science than I am.

However, this research holds out hope, maybe not for me personally, but perhaps other persons who have yet to be born. If there is a common pathway influence by genetic mutations that results in abnormal cytoarchitecture can be found for many cases of autism, perhaps interventions could be done to prevent the person from becoming autistic in the first place and having to live with this disability. Hussman seems to imply this is the case.

As I said, I wish I had more expertise on science so I could understand this stuff better and the possible implications it might have for helping or possibly even curing or preventing autism in the first place.

8 comments:

Anonymous said...

Off-topic for this particular post, but it's on-topic with some of your other posts so I think you might still find it interesting (but feel free to delete this comment or whatever if you don't find it interesting, it's not censorship to choose what you do and don't host on your own blog!):

Take a look at http://www.aspieweb.net/dubin-should-serve-life-in-prison/ (in comments on Jenn's blog post)

Some creep said that "There is something to be said for the effects of social exclusion, which is a basic civil rights abuse." (as if not wanting to be friends with someone who treats one in an unfriendly manner (when one isn't sure that someone doesn't have ASD), and not wanting to date someone who treats one in a manner that turns one off (when one isn't sure that someone doesn't have ASD), is abusing that someone's civil rights?!).

Another creep said that "I stand by my suggestion strongly that the way NT’s behave towards people with AS very much shapes their environment and future behaviours, and often for the worse sadly. You cannot disagree – NT’s usually treat Aspies quite differently to their NT friends, which is in fact the only thing that means Aspergers could be described as a “disability”. This is not trying to excuse his behaviour, he no doubt knew the possible consequences of his actions, but still chose to take them despite this. What I am saying is basically how I feel his behaviour may have arisen, and I feel that had NT’s behaved the way we would expect them to towards Dubin he would not have done what he did." (as if feeling mistreated as a child is any excuse for mistreating other children?!).

SM69 said...

None the less, I think there is genetic susceptibility and yes I think at least AS more clearly runs in families. I would love to get the funding to capture that family tree transmission of AS, but there are a lot of ethical issues at stake. Think of the odd uncle or grand-father no one really fully understood but respected, would that person want to be diagnosed and would the rest of the family wish this too? Somehow I doubt it, why would someone wish to have a label when most of their life is behind them? With little benefit but stigma attached to it? Perhaps none of this is really important to demonstrate?

At Princeton University, I created several knock out mice which lacked 2 highly conserved secreted molecules called Midkines (Mdk and HBNF, also called PTN). These were thought to have a range of functions and based on their embryonic expression patterns to be involved in limb and brain development, as well as possible physiological cascade involving Tissue Plasminogen activator as well as being involved in cancer and being neurotrophic factors. One of them was also called pleiotophin because of this wide range of attributed functions. Very conserved from fly, Zebra fish etc to humans, 98% conserved across mammalian species. The knock outs appeared at first remarkably healthy, even the double KO mice were just fine, until I started to look in some details at the brain anatomy as well as synaptic transmission in collaboration with a Norwegian group. There were abnormalities in cerebellum development and in synaptic transmission. These mice are still being studied today and more functions have yet been identified, peripheral nerve injury model. A long story to say that these genes and mouse models have since been proposed to be involved in autism- I am terribly sorry I cannot find the damn paper that says this just now, but I will find it!

The point is that possibly genes with roles both brain development, synaptic transmission as well as being involved in various physiological function could be good gene models for autism, and possibly yes many genes can be characterised within these categories, which would cause a great variability in genotypes. Potentially, any gene interacting in these genes pathways could also be candidates and so on.

Then the question goes back to why would there be more de novo mutation to account for the increased in number of cases (which as you know I don’t solely attribute to better recognition etc)? Or epigenetic changes, and this reverts back to the environment. I think the increased de novo mutation must be looked at in details.

An additional notes- identical twins share the same genes AND environment and cannot help deciphering between environment and genes in terms of causalities, unless the twins are separated at birth. This is not 100% true, as we know very often twin pregnancies are uneven, with regard to sharing the placenta, blood supply and amniotic space/fluids. I have 2 pairs of identical twins on the spectrum who present quite differently in terms of their autism.

SM69 said...

Posting again the first part of my comment without the links as together with the other comment, it represents more fully my view on GWAS

I now have a working link for the Hussman study and will comment more specifically on this later on.

As you know, I don’t favour a genetic only model of autism- too many facets of the conditions and facts that cannot be explained with genes only. The Pinto study that is one of many large-scale genome wide screen for CNV (Copy number variant) failed to identify any change in 97% of the population tested (as the others did). What they have identified is the remaining 3% cannot conclusively be attributed to functional impairment either- for one the abnormalities can be at times detected in unaffected family members or conversely, cannot be found in the affected sibling. Second, a change is not necessarily causing a phenotype difference.
There is a growing concern within those using Genome Wide Association Studies (GWAS) to research various human diseases, that there is an insufficient characterization and understanding of the genetic changes that have been identified (reported for example cancer, various psychiatric conditions etc. links to Wikigene Post-WGWAS article removed))

The term of Post-GWAS has been coined to include further evaluations of the changes identified, whether they relate for example, to Single Nucleotide Polymorphism (SNP) or Copy Number Variant (CNV). It has been proposed that additional functional studies should be conducted with for example, the in vitro characterization of human tissues or the establishment of in vivo models of disease development. However, models are generally limited to study one variant at a time, and given that no consensus variant has been found in autism, this approach is not currently technically very feasible.

The Pinto work reported increased de novo mutation seen in the ASD group are potentially interesting, but an odd study design could account for this- the average age of the control group is 39 years old and is composed of 69% of females. You could hardly be more mismatched than this. I wrote an analysis of the paper that can be found on the Internet at various places (link removed).

SM69 said...
This comment has been removed by the author.
SM69 said...

Part I: I have now read the Hussman paper in this noise-reduction GWAS (Genome-wide association studies). I cannot claim to understand the paper in all its details, but I think I understand the gist of it. I’ll explain what I think the paper says before commenting on it.

Noise reduction is needed in GWAS because there is a great variation in single nucleotide polymorphism amongst tested individuals that will add noise and clutter the identification of disease-associated variants. For conditions such as autism, which have such a wide spectrum of presentations, and for which we can expect many potential candidates, the issue of background noise will affect even more the results. The system employed to reduce noise is to run a meta-analysis of individual analysis that provide p-values for association signals. The idea is that replicated associations across single analysis will add weight to the signals. A comparison to image enhancement techniques was made- a signal is detected on several sensors is accounted for leaving the other signals as part of the background noise. What varies also is the filter applied- and in the case of the GWAS, the filter will be equivalent to the setting of cut-off p values, the sensor is the equivalent of the maker sets.

I did not understand how the GWAS-NR (Noise Reduction) algorithm exactly computes a weight at each locus, based on the strength and correlation of association signals at surrounding makers, but I trust they know what they are doing. Of note, GWAS-NR cannot achieve better than a single joint analysis when an association signal is restricted to a single marker.

After simulating the approach, they applied it to autism. They have used a number of inclusions and exclusion criteria based on age, diagnosis and presentation. This all makes sense. In total they have used 597 Caucasian families (99 multiplex and 498 singleton) from their own database and 696 multiplex families from AGRE. In total 1947 individuals with autism. The genotyping involved over 550,000 SNP markers.

The authors used haplotype blocks to define regions of interest- I am not sure how this is defined really, exact length unknown, but 3 markers across. In total they have used 2680 blocks based on the top 5000 markers. 725 LD (Linkage disequilibrium) were selected because they achieved a threshold of p<= 0.01, and additional 810 blocks achieved a threshold of p<0.05. They have used an additional 141 markers outside these blocks and all achieved the p value of less than 0.05 and were selected.

Then they look for genes within the 1535 significant LD blocks (based on data set of known start and end positions), and the search results in 431 unique genes that mapped back to 646 significant LD blocks and 50 single markers. The remaining LD blocks were analysed for the distance to the nearest gene, averaging 417,377bp with a range of 5296 to 5,547,466bp (bp stands for base pair, in other words nucleotide number).

After this, they look at LD blocks of ranking p values, with the top on being proximal to PUM2- What is known of PUM2 is that sort of info “Sequence-specific RNA-binding protein that regulates translation and mRNA stability by binding the 3'-UTR of mRNA targets. Its interactions and tissue specificity suggest that it may be required to support proliferation and self-renewal of stem cells by regulating the translation of key transcripts”. That does not seem to be specific enough to explain autism to me.

The authors then continue listing all range of potential candidates genes, focusing pretty much like in the Pinto study, on genes involved in neuronal cell function: Cadherin-catenin function, cell adhesion, ion channel, axon guidance, vesicle transport, post-synaptic scaffold, signal transduction, phosphatidylinositol signaling, cell polarity, Rho-GTPase signaling, cytoskeletal regulation and transcription.

SM69 said...

Part II: So where do it bring us to?

We are not further advanced in my opinion because none of these potential associations are anywhere near being defined a causality candidates, see pervious comment on Post GWAS I made earlier. The same problem applies.

There are many unanswered questions and the candidate genes selected cannot on their own explain the current features of autism.

Immune system? Digestive function?
Regressive autism? etc.

Why from an evolutionary point of view would we have so many potential mutations?

Yes potentially these genes if mutated could lead to an ASD phenotypes, but why would there be so many mutated in the current ASD population? That seems like extraordinarily improbable to me as a whole. An explosion in autism prevalence in the last 2 decades cannot be attributed to so many mutations occurring independently in affected individuals at the same time.

Why now? Is that de novo mutations? Should these be linked to other conditions also on the rise, Alzheimer, cancer etc?

Do we have on the whole a population that is becoming sicker and sicker?

None of these questions were even touched on in the discussion section.

I think these authors mean well, but they do not see autism in other ways than through their genetic and brain-only glasses. I am very unconvinced of the significance of this work. I am on the other hand very convinced that it does not help at all people with autism. This work represents huge funding money, let's not forget this.

But what is clear again and again- there is not one autism gene, there is not even 10 autism genes that can be found in this many individuals, because it that was the case, we would have found them by now.

I am afraid, I am in favour of looking at the biology and physiology of affected individuals as a way forward to understand what causes autism and that will be the most direct approach to provide treatment and remedial /accommodating interventions.

OK J- I have done the work you have asked me to do! Sorry if I remain skeptical. I know you want to understand your autism better.

jonathan said...

Lorene: Thanks for reading the study and commenting here. Unfortunately I am still kinda out of the loop and I guess I won't be finding out answers what caused this defect in my brain and how my brain does not function properly.

SM69 said...

I know you are taking part to various studies, I am amazed that the researchers involved don't give you any feedback. I find this quite upsetting actually. In the UK we tell people what we have found and how it compares to the group sampled. It is a shame there is nothing you can learn about yourself from there.

I'd opt for detailed MRI and SPECT scans. One assessment we have started to refer families to, is to look for the maintenance of primitive reflexes and presence of normal reflexes. Even very high functioning AS kids have retained abnormally some primitive reflexes. The work proposed to correct this is called'Developmental Reflexive Rehabilitation.' I am copying from the BIRD organization web site (b-i-r.d org uk) that does this in the UK-It is based on a series of movement patterns which facilitate neurological feedback to the brain. These movements inhibit the 'primitive' reflexes that have either not developed properly from the infantile reflexes present in everyone in the first year or two of life - or which have been re-released at a later stage through brain injury.

These primitive reflexes affect a person's development and functioning on a physical, psychological, social and cognitive basis.

I think this is a very interesting complementary approach and for once, one intervention free to parents here. Eye tracking, sensory motor integration etc.

Other than this, I think neurofeedback and for some FastForWords are also very valuable interventions that tell indirectly how the brain functions.

Finally go back with your parent's help to your development history, very often there are clues as to why you are who you are today.