Genome-wide association studies (GWAS) were hailed as a landmark change in the way genetics was carried out; finally there was a tool that could be used to probe the genetics of common complex diseases based upon the common disease-common variant hypothesis. For those not familiar with this idea it was hypothesised that susceptibility to common diseases was likely influenced by multiple genetic variants that were generally common amongst the populace. People who have a disease may have multiple variants, but at the same time some people may also have these variants and never suffer from a given disease; there were key environmental triggers. Either way scientists would be able to uncover these variants and hopefully use them to predict who might suffer from a disease, or else use them to predict the severity.
A major GWAS was published by the Wellcome Trust Case-Control Consortium back in 2007 that covered seven common diseases and utilised a cohort of common controls to test against each disease. Needless to say multiple variants were detect, all with rather modest effects. Since its publication the WTCCC associations have been replicated and validated, and new variants have also been detected that are associated with risk of a number of common diseases, including rheumatoid arthritis, prostate cancer, cardiovascular disorders and more. My own interests lie in the field of autoimmunity, developed when I was on my undergraduate work placement. It soon became obvious that if these common variants did truly increase a persons risk of developing disease then it would involve the complex interactions of multiple pathways and multiple tissue and cell types - that's a needle in an Atlantic ocean of needles to put it lightly.
So the question is do common variants actually increase a persons risk of developing a given disease that is associated with it? Based on the evidence so far I'd say that this may well be the case, but, and this a very important but, not all of them. A paper published last year simulated the effects of rare variants in linkage with associated common variants and showed how they could potentially inflate the frequency of common variants to the point where they reach genome-wide statistical significance. Now I wont pretend to fully understand this paper in question, but I do understand the potential implications; rare variants may have a role to play in disease risk, and GWAS are not the tool to uncovering those variants.
So why am I even talking about this. Most people in the genetics community will be more than aware of the current grumblings and misgivings about GWAS. The reason is this paper published online in Nature this week:
Surolia et al, (2010) Functionally defective germline variants of sialic acid acetylesterase in autoimmunity, Nature Advanced Online publication doi:10.1038/nature09115
Nature News has coverage of this article, and explains things very concisely, but I'd still like to have a go at my own coverage (after all its good practise for when I start my PhD in October).
The justification for investigating this particular gene, SIAE, is due to previous functional work on siae mutant mice which display a defect in B-cell tolerance. They hypothesise that potentially variants in the human SIAE gene may contribute to autoimmunity, despite no variants within this gene having been associated by GWAS. In essence this is a good old-fashioned hypothesis driven candidate-gene approach.
They re-sequenced the SIAE exons (10 in total) in a number of autoimmune patients of European ancestry and detected 2 rare variants, one in a patient with rheumatoid arthritis and another in a patient with Crohn's disease. They then extended their re-sequencing efforts to include a total of 188 cases and 190 controls and uncovered a number of point mutations, including previously characterised single nucleotide polymorphisms (SNPs). This cohort was then extended to 923 autoimmune cases (varying autoimmune conditions including RA, type 1 diabetes, systemic lupus erythematosus and inflammatory bowel disease, amongst others). 14 previously uncharacterised SNPs were detected in the initial phase, including several cases homozygous for a non-synonymous mutation M89V. In their control cohort of 648 volunteers they detect 17 people carrying one of 8 non-synonymous SNPs, none of which were homozygous for the M89V variant.
The next step was to determine whether variants were defective in either their catalytic activity or secretion. They generated FLAG-tagged cDNA constructs for each variant using site-directed mutagenesis and transfected them in (at least) triplicate into HEK-293T cells in order to quantify protein levels by quantitative western blotting and an esterase activity assay. To cut a long story short 24 of the 923 subjects carried catalytically defective alleles, described as an activity 50% below that of the wild-type. Others had a profound defect in secretion. Amongst the secretion defective alleles was the previously mention 89V allele which retained catalytic activity, but whilst detected in 9.7% of controls in the heterozygous state, only cases were found to be homozygous for this mutation suggesting a recessive mechanism for this particular allele. In addition only cases displayed statistically significant departure from Hardy-Weinberg expectations, thus the 89V variant is enriched in patients suffering from autoimmune conditions.
The other alleles were detected for a dominant-negative effect by inhibition of the wild-type in a Murine-based assay, seven of which were classed as such.
So we start to see a picture building of multiple rare variants being detected in autoimmune patients that have loss of function mechanisms, and thus may play a role in the aetiology of these diseases. But, wait, we aren't finished yet. This was a rather comprehensive paper. Only one of 11 catalytically defective variants wasn't conserved between primates and rodents, thus providing further evidence of the impact of these variants.
So how much do these variants contribute to autoimmune disease? This is where the odds ratios come in. Surolia et al calculated OR for a number of disorders, including RA and type 1 diabetes. The verdict? RA OR 8.31 (95%CI=1.69-40.87) and type 1 diabetes OR 7.89 (95%CI=1.58-39.30). So we see that they span quite a large range, but significantly none of the 95% confidence intervals fall below 1.0, so there is a genuine effect on disease. The large ranges are most likely due to interactions with other loci, stochastic variation, different effect sizes for each variant, in other words the complexity of the disease modulates the effect of a given variant. Overall they compute their OR for all autoimmune conditions as 8.62 (95%CI=2.03-36.62).
These are pretty significant findings, particularly with respect to my own disease of interest, rheumatoid arthritis. Previous associations from GWAS haven't detected risk ratios for any variants above 1.2-1.6. The only known variants to exceed such odds ratios are those of the shared epitope variants which account for about 20-30% of RA genetic susceptibility. In short, this is a big finding. It is, however, not the end of the story. This is just a snap-shot of the rare variants associated with autoimmunity in one particular gene, the question is, can this approach be extended to try and find the other variants that affect disease risk?
My opinion as a lowly student? It is highly likely that rare variants contribute much more to common diseases than was previously thought, but finding them is going to require some new tools. What about all those common variants associated with disease? How do we find the causal variants? Well the answer is at the beginning of this paper by Surolia et al...re-sequencing of associated regions in cases and controls. They show in their power calculations that to reach a power or 0.8 they require 550 cases and 550 controls, that is a hell of a lot easier to achieve than the predicted 10,000+ cases and controls predicted to detect the most modest odds ratios using the GWA approach.
But there are other implications to think of too, notably those of costs. GWAS require multi-centre collaborations (granted this study did too), but they also require the use of thousands of micro array chips to genotype the entire genome, and even then there are certain areas that are notably missing (I'm thinking of the FCGR locus here which is badly represented on whole-genome chips because of its copy number polymorphic state and the high homology between the genes at this locus). These are very expensive studies to carry out and utilise some very complex mathematics I'm not even going to pretend to understand. So it seems that studies such as this may turn out to be more cost-effective, but only if they produce results on this scale. Seems like a tricky situation, common variants and GWAS versus rare variants and candidate gene approaches.
I've still not answered the whole question though. Re-sequencing studies are being carried out as follow-up to associated regions, but that still leaves those areas that may be implicated, but are not candidate regions. In all honesty I can't say what the answer is, but I'm hoping that the genetics community will come up with an answer sometime, sooner rather than later. All I do know is that this paper may well set the stage for more investigations into the role of rare variants in autoimmunity and other complex disorders.
References:
Dickson et al (2010) Rare Variants Create Synthetic Genome-Wide Associations. PLoS Biol 8(1): e1000294. doi:10.1371/journal.pbio.1000294
Wellcome Trust Case Control Consortium (2007) Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls, Nature 447, 661-678
Surolia et al, (2010) Functionally defective germline variants of sialic acid acetylesterase in autoimmunity, Nature Advanced Online publication doi:10.1038/nature09115
No comments:
Post a Comment