Autism is probably one of the best known neurological disorders, in part due to promotion in Hollywood movies such as Rain Man. It is described in the On-line Mendelian Inheritance in Man (OMIM) database as being
‘characterized by a triad of limited or absent verbal communication, a lack of reciprocal social interaction or responsiveness, and restricted, stereotypical, and ritualized patterns of interests and behavior (Bailey et al., 1996; Risch et al., 1999).’
Related to autism is autism spectrum disorder, often referred to as ASD, a range of conditions with less severe affects and includes Asperger’s Syndrome.
(The OMIM entry for autism, entry 209850, while intended for researchers, is well worth reading if you want a primer on the genetics of autism. In particular, it gives you an interesting look at the changing views over time.)
The past few years have seen several genome-wide surveys*1 for possible genetic variations related to autism. The results to date have been mixed.
One suggestion is that the mixed results are because imprinting – an epigenetic process – confuses the analysis.
Consistent with this, some of the features of autism are also found in genetic disorders that have been associated with imprinting, suggesting that maybe this is true for autism too.
Delphine Frabin and her colleagues investigated at if an epigenetic event – parent of origin effects – might have a role in autism.
Imprinting, SNPs and parent-of-origin effects
Molecular genetics is full of terminology. Crazy stuff. Let’s de-fog some of it.
We’re diploid: we have two copies of every gene (except those on sex chromosomes).
For most genes, it doesn’t really matter which parent the gene comes from, they’re a DNA sequence coding for the same thing whichever parent they came from.
Very loosely speaking our genes are controlled in two main ways: they are made available to be used or not; gene that are available to be used are regulated as to how often they are used.
Genes that are not being used are packed away; those that are being used are ’open’, able to gotten at by the proteins (transcription factors) that regulate how often they are used.
You’ve heard of other kinds of imprinting.
A duckling imprints on it’s parent, following it everywhere.
A songbird’s song is imprinted onto it at a critical stage in it’s development.
Here’s another sort of imprinting, the parent ‘marking’ a gene by DNA methylation – adding a carbon (methyl group) to the DNA bases – to indicate how their gene is to be used.
(I prefer to think of this as marking for how the gene is to be used, rather than marking that it is from a particular parent although conceptually you might think of it either way.)
This DNA methylation can affect way the gene is organised, if it packed away or open, or what DNA loops can be formed, affecting how it is controlled and used. (See 5th & 6th figures of Epigenetics and 3-D gene structure.)
Several other disorders that have some common traits to autism are known to have parent-of-origin effects, such as Prader-Willi and Angelman syndromes.*2
These epigenetic effects might also explain why the results of the genome surveys so far were mixed, as the methylation marks are not part of the DNA sequence itself, but added to it; most of the earlier methods don’t consider it.
To measure what locations might be associated with autism, Frabin and her colleagues measured the linkage of SNPs, bearing in mind what parent (mother or father) the allele came from.
Linkage analysis exploits that neighbouring positions in the genome tend to more often be passed together on to a child, than ones that are further apart. (I written a little on this in Boney lumps, linkage analysis and whole genome sequencing and elsewhere.)
Positions in the genome that are passed together onto children that are associated with the trait (autism in this case) are likely to be close to variations in the DNA that contribute to causing the trait.
The DNA sequences used to follow what is passed on to the children aren’t the things that cause the trait itself, they are nearby sequences that are unique in the genome – ‘markers’ that geneticists know are located in one place in the genome.
This study watches how SNPs, single nucleotide polymorphisms – DNA sequences that differ in only one DNA base, are inherited, using them as the markers to indicate nearby DNA variants that might affect autism.
By tracing which adjacent base changes tend to be passed on together at the same time as the trait tested (autism), researchers can try locate regions of the genome that might carry an mutation (difference in the DNA sequence) that affects the trait.
I wrote earlier that for most genes, which parent the gene comes from doesn’t matter. In the case of parent-of-origin effects, obviously it does! In this case researchers need to have the DNA for at least one of the parents, ideally both, so that they can compare what was passed from the parent to the child.
The mathematics used is different from analysis that doesn’t look at parent-of-origin effects, too, but I’m going to skip that. (This is for a general readership.)
What did the researchers get up to?
Here’s their summary:
‘We have performed a genome-wide linkage scan that accounts for potential parent-of-origin effects using 16,311 SNPs among families from the Autism Genetic Resource Exchange (AGRE) and the National Institute of Mental Health (NIMH) autism repository. We report parametric (GH, Genehunter) and allele-sharing linkage (Aspex) results using a broad spectrum disorder case definition. Paternal-origin genome-wide statistically significant linkage was observed on chromosomes 4 (LODGH = 3.79, empirical p<0.005 and LODAspex = 2.96, p = 0.008), 15 (LODGH = 3.09, empirical p<0.005 and LODAspex = 3.62, empirical p = 0.003) and 20 (LODGH = 3.36, empirical p<0.005 and LODAspex = 3.38, empirical p = 0.006).’
OK… They took the families previously studied by the Autism Genetic Resource Exchange,*3 a large collection of DNA and data on autistic people, and tested about 16,000 locations*4 in the genome that are known to commonly differ by a single base in different people (SNPs).
Two different methods were used, the idea being that if a region was located as being interesting in two different, it was more likely to be real than a quirk of the data or methods being used.
They found three regions in the genome that have strong indications of parent-of-origin effects associated with autism in both the methods tested. These were on chromosomes 4, 15 and 20.
All three are from the father (i.e. are paternal). (And exception on chr 7 that is maternal is discussed later in their report.)
Linkage analysis is a game of narrowing down on targets.
There’s a lot of DNA in a genome to examine – 3 billion bases of it – so researchers need to break their search for the tiny changes that can cause disease into a series of steps, first looking for regions, then inspecting the regions more closely.
This is much faster than it used to be. One the DNA is at hand modern high-throughput approaches can carry out the experimental work in much faster than a few years ago. (In time we expect to see DNA sequencing take over this, by directly sequencing many individuals’ DNA and directly observing the specific changes made. This still has a number of technical challenges, in cost and bioinformatics.)
Nonetheless, in this case the results are not a one-shot ticket to locations that contribute to autism. It identifies markers that indicate that the region the marker falls within is likely to have a role in at least some cases of autism.
Researchers will want to look closer at the regions that they have identified, to find the specific DNA changes that are associated with autism. From there they will want to try understand the precise effect of the change, what is doing?
This may seem long-drawn out, but that’s the nature of genetic studies of complex disorders like autism. It’s progress, just a step at a time.
(These are pitched a bit higher than the article, mainly as asides to those who want a little more detail. Those who want a lot more detail should read the paper themselves!)
*1: 2009 saw at least 6 genome-wide screens, a busy year for autism genetics! I’m not going to compare the results of this study with earlier studies and putative autism loci that have already been identified.
*2: For these syndromes, one parent’s alleles in the 15p11-13 region are lost (deleted) and the other silenced through imprinting. In the case of the Prada-Willi syndrome, it is the paternal alleles that are lost; in the case of Angelman syndrome, the maternal alleles are lost.
*3: From the paper:
‘The samples used here were previously described by Weiss et al. . Nine hundred ninety three (993) families (896 affected sibling pairs) from the AGRE (Autism Genetic Resource Exchange) sample and 223 families (174 affected sibling pairs) from the NIMH (National Institute of Mental Health) Autism Genetics Initiative were included. AGRE families with a child diagnosed with an Autism Spectrum Disorder (ASD) based on evaluation by the Autism Diagnostic Interview-Revised (ADI-R)  were recruited from across the US. Further information on participant recruitment and study procedures has been described elsewhere  and is available on the program website (www.agre.org).’
’The combined data set, consisting of 1,216 nuclear families, was used for genetic analyses. All families used in our analyses had at least one genotyped parent; 89.4% had genotypes for both parents.’
*4: A 16K set of SNPs are used rather than a more complete set in order to have the average distance between the SNPs >2 centimorgans to avoid confounding the linkage methods, from the paper:
’We selected an extremely high quality set of SNPs for linkage analysis, including only SNPs genotyped in both data sets with 99.5% concordance and â‰¤1 Mendelian error. Linkage analysis involving high densities of markers, where clusters of markers are in linkage disequilibrium (LD), can lead to biased results . To alleviate these concerns, we analyzed a pruned set of 16,311 highly polymorphic, high-quality autosomal SNPs that did not contain any two nearby markers correlated with r2 >0.1, providing a marker density of 0.25 cM.’
Fradin, D., Cheslack-Postava, K., Ladd-Acosta, C., Newschaffer, C., Chakravarti, A., Arking, D., Feinberg, A., & Fallin, M. (2010). Parent-Of-Origin Effects in Autism Identified through Genome-Wide Linkage Analysis of 16,000 SNPs PLoS ONE, 5 (9) DOI: 10.1371/journal.pone.0012513
Other articles in Code for life: