Thursday, September 30, 2010

Genetics the right way....er, at least part-way!

A story about a gene for migraine headaches is making the rounds of health-related websites.  A paper published in Nature Medicine on Sept 26 reports that mutations in a gene that encodes the potassium ion channel TRESK K2P are responsible for common migraine with aura (one such aura is depicted in the picture to the left, though the artist says it doesn't do the temporal/spatial experience justice at all).  This ion channel was already known to have a role in pain pathways and as such has been a target for anesthesia, and other ion channel genes have been found to be associated with rare forms of migraine, which is why the authors decided to investigate KCNK18, the gene that codes for this particular ion channel.  So far so good.

The authors found that a mutation in this gene 'segregates perfectly' in a large family -- that is, everyone with migraine in that family had the mutation (which causes a frame-shift that results in a truncated channel), and no one without migraine had it. They then looked for expression of this gene in 'migraine-salient areas', and found it where it was expected. So, this was a good use of a priori knowledge about the biology of a trait to zero in on a gene that may cause migraines, and of course the possible sequel to the story is that it may thus be a therapeutic target.

'Segregates perfectly', however, raises the question about whether all those who did not have migraine in the family did not carry the mutant allele.  As it turned out, no one without migraine in the family had the mutation, so it did segregate perfectly (we have to assume that no unaffecteds in the family later became affected, which may not be a trivial point)  But then the researchers went on to sequence this gene in unrelated people with and without migraine0--a case-control comparison. This, it turns out, was problematic.

Unfortunately you have to go to the Supplementary information to find this out -- and we have to say that burying important data in the supplementary tables, without clearly spelling out its implications in the regular text that everybody (including reporters) reads, can be a cynical way to take the focus off serious issues.  We hope that we haven't somehow misunderstood the supplementary data, but of the hundreds of KCNK18 genes these authors sequenced, they found 14 different mutations in the gene, most of them very rare but more importantly, none of them statistically more common in those with migraine than in those without.  The frame-shift mutation that may be responsible for migraine in the one family is very rare as well, and certainly comes no where close to being 'the' gene, or even a primary gene, for migraine, even if it does explain it in this one pedigree.

Now, the fact that migraine generally takes a trigger -- or people with genetic susceptibility would have excruciating headaches all the time -- means that people with a genetic predisposition but who have never met their trigger may still be susceptible.  But it also means that this is one of those disorders that's 100% environmental, and 100% genetic.  And it makes the choice of controls difficult or even uninformative, when anyone with a truly causative mutation might be unaffected because they haven't been exposed to the provocative environment.

But this is a trivial problem with respect to this study, since so few cases or controls have a mutation in this gene. In this family, too, it could be that the effect is manifest only in some genetic background that was not varying in the family which would generate a 100% concordance pattern but would implicate the particular gene only in that background.  In other words, suppose a variant at some other gene, X, was shared by all members of this family, and the detected effect at KCNK18 only occurs in people also carrying the gene-X variant. That is by no means implausible.

We don't want to complain all the time, although resistance to poor practices, vested interests and the like is important, in at least a small way, to help keep science on as close to an optimal track as can be done in the realities of human society.  We don't think people, especially scientists, should wander in dreamland, intentionally or avoidably oblivious to known truths.  There are enough unknown truths to deal with!

But you don't get published in Nature Medicine by being too forthright with caveats.  The current reward system encourages burying caveats in Supplemental material, or minimizing them.

We still say that, in this case, the discovery process seems to have been done the right way -- by putting biology first.  The investigators used the kinds of understanding of basic biology, as being about signaling interactions among cells, that we try to describe in our Mermaid's Tale book.  Neurological traits are about interactions among neurons, and they signal to each other by a limited variety of means, ion channels being a major component of that.  But it would still be wrong to yield to temptation and call this 'the' migraine gene, since it explains so few cases in the population, and is found in controls as well.  Neither can it be called a 'migraine gene', since one would assume that its function and evolution were not to cause migraines.

Searching candidate genes, in this case known ion channel genes for mutations, rather than blind genomewide association studies (GWAS), is a sensible way to find genetic factors.  Indeed, GWAS for common migraine with aura had not worked -- and that's no surprise.  In a family with clearly inherited migraine problems, to search the limited genome space occupied by ion channel genes makes biological sense, and that's what seems to have worked in this one family.  But if there are many causes of this trait, this one might not have been frequent enough in cases vs controls sampled from the general population, to be statistically detectable.  Or, if the variant is rare, to even have been included in the set of cases that were analyzed in the GWAS.

While we don't believe these authors have done much for migraine sufferers in general, in spite of the widespread attention this paper got (and this column in The Guardian explains why the attention was so sloppy), at least the approach was a right one.

7 comments:

Holly Dunsworth said...

Thanks for this!

Anne Buchanan said...

You're welcome. We do wish we'd been able to draw different conclusions!

James Goetz said...

I'll take a novice conjecture at this. The primary point that I see is that the respective rare frame-shift causes a "complete loss of TRESK function." Likewise, if other frame-shifts in the same gene don't cause a "complete loss of TRESK function," then the importance of the discovery stands. However, my problem is that I don't have enough background in genetics to interpret the Supplemental Information.

Anne Buchanan said...

Right, it's not the biology we're questioning. In fact, it's progress if they've found further evidence that mutations in ion channel genes contribute to migraine. It's just that the particular mutation, or even the particular gene they have studied doesn't seem to contribute to migraine very generally. It seems, if this finding holds up, that they've explained the disease in just one family.

But this is a typical finding for complex diseases -- a contributing gene or mutation in one family isn't found in others, or is found to explain just a small fraction of the disease.

James Goetz said...

Okay, I suppose that one angle of your main point is that the complete loss of TRESK function contributes to only a small or trivial percent of migraine sufferers.

Anne Buchanan said...

Yes, in fact the surprising result would have been if this particular mutation _did_ explain a lot of cases.

Ken Weiss said...

'Trivial' may be the wrong word. But if it's a true finding, then it could be very important to those affected via this mechanism. It's unlikely this opens up really new 'pathway' information, since ion channel genes have been and are being heavily studied.

More sobering is that if we've interpreted the result correctly, it simply doesn't make a major explanatory breakthrough, and instead confirms the widespread finding about the complexity of complex traits.

To us, fully accepting that is needed if we're to make some kind of conceptual progress, but most people resist such acceptance, because complexity is harder to deal with, clinically, etc..