Nick Martin - Symposium 2003 - BMA Home -
Genetics of anxiety and depression
Nick Martin
Queensland Institute of Medical Research and Joint Genetics Program
Brisbane, Australia
nickM@qimr.edu.auI too would like to thank Max for the invitation and for what has been a very stimulating day, I think it is marvellous to get the sort of cross disciplinary interaction going we have here.
I hardly need to show this slide to this audience which points out that depression is set to become the second biggest cause of health burden by 2020; everybody knows this and, as has been indicated by Max and other speakers, almost everything you look at has got a strong genetic component. Not least anxiety and depression which are enormously large contributors to the total disease burden in the community and on which we and many others have done studies. In our case studies of thousands of pairs of twins on the Australian Twin Registry showing that the variance of susceptibility to anxiety and depression is about 40% heritable. But one of the interesting questions Ian Hickie raised with me and which stimulated us to do this work, is whether the same genetic influences are active throughout life. In particular, Ian's interest was whether there are genes that specifically influence late onset anxiety and depression. Because we have been studying these traits in twins in the Australian Twin Registry, since about 1980 have got multiple accessions of data on these people over a fifteen lifespan, so although we don't have longitudinal data from 20 to 70, in some twins we've got measurements at 20, 30, 35 and in others we've got them at 40, 50, 55 and so on; so what we have done is like a tiling experiment molecular biologists do when they sequence with BACs.
The numbers of twins for whom we have these measures of anxiety and depression at different ages are quite large as you can see, and this enables us to fit structural equation models like this simplex model for longitudinal development. I was very pleased this morning when Gary Egan introduced structural equation modelling in a different context. He is looking at signalling between parts of the brain while we are looking at development of psychiatric symptoms through different stages of life. At each age we allow for the possibility that new genes are being switched on to influence the trait, as opposed to the influence of the old genes that are being transmitted through all ages. So the scientific question is, do we just simply see simple set of genes that are here at the age of 20 (and presumably being switched on much earlier) that are transmitted all the way through life? Or do we see that there are new genetic influences coming at different ages? When we fitted this model to our female data which are the most numerous the answers were really very striking. In fact, yes indeed, there IS this set of genes for depression which are there at age 20 and are transmitted reasonably faithfully through life. But interestingly, there is a small but significant chunk of variance coming in here at old age which surely does support the idea that there might be genes specifically for late onset depression. Now I am not really going to say much more about that until towards the end of my talk and you will see why I started with that.
The problem with anxiety and depression is of course that psychiatrists dismiss these questionnaire measures we have used. I would much rather that we were using full diagnoses after a proper clinical interview, but being an irreverent geneticist who's done one first year course in psychology I have always been attracted to personality measures, and in particular to the neuroticism scale, which you can measure in about thirty seconds for about fifty cents instead of employing a very expensive psychiatrist. Using only a few items like this, you find these very high correlations of neuroticism with anxiety and depression. We can show, with very large samples of twins that the same genes that influence neuroticism are also influencing anxiety and depression - it is really just one genetic factor. So if you want to find a gene influencing anxiety and depression you may be just as well off studying neuroticism which, as I say, can be done very cheaply.
We were stimulated to ask whether we could find genes for anxiety and depression by this work that came out in the mid 90s from Jonathan Flint and colleagues, looking at a trait called emotionality in mice. They had several ways of measuring this, fairly simply and very low tech. Just put a mouse in a cage like this and you measure two things; firstly, how much it runs around the cage as opposed to just freezing because it is terrified; and secondly, how many turds it leaves at the bottom of the bucket after two minutes, which are politely called faecal boli, or the defecation score, and these two measures correlate very highly. This is the other measure which is the elevated plus maze and that really is the ratio of how much the mouse runs back and forth out here on this exposed part as opposed to running back and forth across that sheltered part; the less anxious mice will spend more time running back and forth along there and what Flint and colleagues did was to measure this on the F2 from a cross between two lines of mice. They then typed genetic markers across all chromosomes. When they did linkage analysis they found this huge lod score of 14 - that's odds of 1014 to 1 in favour of linkage here at the end of mouse chromosome 1. The top line there is for the open field activity (that's running round the bottom of the bucket), and here's the defecation score, down here and this is the maze measure here. Down here they had a control variable which is just activity in a non-anxious situation and you can see there is no linkage with that. They also found some evidence of linkage on mouse chromosome 12 and 15.
So we thought this was very exciting and we said, well if they can do that in mice can we find quantitative trait loci (QTLs), or genes of major influence, on neuroticism and hence anxiety and depression in humans?
The only trouble is if you do the power calculations for conventional linkage, the numbers of sib pairs that you would need to screen and genotype to do this study would be over 20,000 sib pairs to find a QTL accounting for 10% of the variance - which is actually bigger than the biggest QTL found for mice, and in mice you have already got selected lines so this is kind of depressing. In fact this has been known since the early 70's. People have done these power calculations and have written off the possibility that you could ever do linkage to find these genes of influence on quantitative traits.
But just a few months after the Flint paper there was a very nice paper published by Neil Risch pointing out that in fact you don't have to genotype every sib pair in the population. If you form your whole population of sib pairs whom you have measured for this trait to select out the extremes of the distribution you can get most of the linkage information. If you take your quantitative trait and assign everyone to a decile from 1 to 10, then it turns out that most of your linkage information is coming from the sib pairs who are most extreme, that is of 1-10s or 10-1s; you also gain a lot of linkage information from 10-10s at the extreme high end of the scale, or 1-1s at the extreme at the low end of the scale. When you think about it, it is intuitively obvious that that is where the information is going to come from, but it impressive to see the calculations and also depressing to see how you get virtually no information at all from the vast mass of sib pairs who are in the middle of that distribution who are just normal-normal, which is what most of us are.
So this paper of Risch's was a clarion call to people like myself who had access to large population data bases like the Australian Twin Registry already screened for neuroticism. At that stage we already had the neuroticism score for 23,000 people, about 10,500 sib pairs from 6,500 families. From those we were able to then select the people in these four corners of the sib-pair distribution and interview them. We used Gavin Andrew's telephone CIDI interview, the short CIDI interview that gives you a diagnosis for most of the axis-I disorders. We completed this interview for most people in our selected sample, and we also approached them for blood samples, and also their parents, and we got blood from most people. Interestingly, we actually did slightly worse at the top end of the distribution where they were more anxious, not surprisingly, but we did manage to get buccal swabs from those people as a backup, so there was some sort of operational validity to our measures of anxiety! Just to show that our selection worked, this is the prevalence of these diagnoses we got from the CIDI interview in the bottom quintile, and here it is from the top quintile of that distribution. You can see there is a greatly increased risk of each of these disorders in the top quintile, and this is just selected on a twelve item neuroticism scale measured ten years before we actually did the interview.
We actually got to this stage about three years ago and at that stage we made the tactical mistake of hooking up with a biotech company to try and get our genome scan done and they did about a third of the sample and then they got taken over by another biotech company who lost interest in it.
So what I have to show you are linkage data on about a third of the sample and we are now desperately trying to get funding to get the rest of the sample genotyped. So we don't actually have anything screamingly significant but there are a few interesting results. This is just stringing all the chromosomes together, it's about 3.3 mega bases long and we have a good peak on chromosome 19, but you see various smaller peaks here, none of which are quite reaching that magical level of 2 and so the interesting question is how many of these peaks are significant. What we have done here is to calculate the probability, just from emperical simulation, of the likelihood of getting the distribution of the number of T values we have observed. Simulating our study 2,000 times, dropping genotypes in there with the same allele frequencies that we have observed but unlinked to any disease gene, we wouldn't expect to see the number of peaks that we have got more than 17 in 2,000 times, or an empirical P value of around 0.01, so that suggests that we have got more suggestive linkages than expected by chance. The problem is deciding which ones of those are real and it is very hard to know how to do that. The ideal is that we would get the money to genotype the other two thirds of our sample and some of the linkages would then become screamingly significant, but in the meantime all we can really do is check the replication of our peaks with the locations from some other studies, and that is what I am going to show you.
This is a study that was published just a few months ago in the American Journal of Human Genetics from Iceland, where you all know they have this wonderful population resource. They collected a series of families with what they call anxious depression and by far their most interesting peak was on chromosome 9q, and it just so happens that one of our most interesting peaks is also there. Their peak was actually centred on here, ours is slightly further that way, but you will have to believe me when I tell you that in fact peaks shift around. Anyone who is in this game knows that if you redefine the phenotype or collect a few more families, or age correct in a slightly different way then unfortunately your linkage location will shift around quite a bit, so we are not at all concerned about this. The fact that we are finding a peak at the area so close to the best peak in Iceland we find very encouraging.
Another article, it's not actually quite out yet, I think it is still online, is from George Zubenko in Pittsburgh and they have been looking at recurrent early onset depression in a series of families. They have specifically been interested in recurrent early onset depression in females where their best linkage is on chromosome 2, and these were our results for chromosome 2. Just looking at depression in females, this is where Zubenko has his best linkage right here, which is where the CREB1 gene is. So we are quite excited about that.It is interesting that Zubenko makes the point that frequently he sees in these early onset depression families an increased frequency of stress stroke and neurodegenerative disease in all the members of those families. That is quite interesting in relation to our other finding I want to draw attention to which is on chromosome 19 which, as I already pointed out, was our best peak. Very interestingly, our very best peak is bang over the location of apoE, and there is a whole heap of literature out there which draws attention to the possible etiologic links between dementia in the elderly and late onset depression. So this brings me back to my earlier simplex analysis, just a biometrical analysis of our twin data where we are seeing evidence for genes specific to late onset depression in women. Could apoE be the cause? What we are trying to do now is firstly to refine our linkage analysis to see whether it is stronger if we restrict it to late onset cases, although we feel that our power will go down when we do that. And secondly to type some SNPs right into this apoE region in those samples and see whether that strengthens the linkage peak.
© 2004, Brain and Mind Australia Inc. - Copyright Notice -