SHARE

 Rather than the conventional approach to interviews, which is to start with the comfortable questions, let’s start with the obvious uncomfortable one: why has deCODE been in such dire financial straits, and what’s the current status?

When I founded the company in 1996, my idea was to use genetics to develop diagnostics, but at that time the technology to systematically make discoveries to be turned into diagnostics was unavailable, and it wasn’t available for several years. So we had to fiddle with the business model and stretch ourselves to survive financially. We collected debt. For six years we had to be very creative in the way we raised money to keep us afloat. So we were very stretched when the recession hit the western part of the world. And when Lehman Brothers collapsed, when that company went bankrupt, we lost substantial amounts of money.

 Lehman Brothers was handling your operating funds?

Yes, and they had actually taken a cut without our consent and turned it into securities that they underwrote themselves, really toxic papers. Once we realized that, I went and negotiated with them. We were going to buy these papers back. We shook hands on that deal on a Friday, but they went bankrupt on Sunday. So we went through Chapter 11. We are coming out of Chapter 11 much more focused. We sold off our drug-discovery and development outfit and are now focused solely on diagnostics and partnerships based upon our unique capabilities in genetics. We are backed by two of the original investors in deCODE, who have remained faithful to the concept and who believe in it. So we are now reasonably well financed. We have a much more focused business model, centered totally on DNA-based diagnostics, which is within an arm’s reach of the genetics that we do.

 How much of original business plan was based on drug development?

The original business plan that we began with in 1996 was not at all dependent on drug development. That was an endeavor that came later that we went into during the times when the technology to make the discoveries in genetics that could be transferred into diagnostics really didn’t exist. What we needed was the ability to genotype a lot of single nucleotide polymorphisms, or SNPs, on the SNP chips, and they came later. But to be able to take full advantage of that, we had to found the company earlier; we needed phenotypic data to make it work. And, yes, we were founded too early, no doubt about that, but the consequence was that, when the SNP chip technology arrived, we were so ready that if you look at what has been discovered with that technology, an awfully large part of it has come from us. We had gathered so much data by then that we were ready to take advantage of it almost in real time. And now we have information on about 60% of the adult population of Iceland—all volunteers who have given us informed consent to work with their health-care information and phenotypic information.

 Many of deCODE’s highest-impact papers are on type 2 diabetes . What have you discovered, and what have we learned from that about the disease?

We isolated the TCF7L2 variant, the first high-impact reproducible association to type 2 diabetes, and it turns out to be a variant of a gene that affects insulin secretion (see table, paper #3). We positionally cloned that—it didn’t come out of genome-wide association studies. Our next big contribution to type 2 diabetes came in a paper we published in Nature last December in which we showed that the same sequence variant, when it comes from the father, increases the risk of type 2 diabetes, but when it’s inherited from the mother it protects against type 2 diabetes (A. Kong, et al., 462[7275]: 868, 2009). It’s an example of parent-origin influence and the effect of the sequence variant on a common complex trait. This is very important. What’s interesting is that the TCF7L2 variant is the one that has the biggest impact on risk of type 2 diabetes, at least of any variant discovered so far. And the one we described that is under the parent-of-origin influence is the variant with the second-largest impact on type 2 diabetes.

 Has the discovery of these variants taught us anything new about the disease itself?

When I was in medical school—and that wasn’t yesterday—and up until fairly recently, the prevailing hypothesis was that type 2 diabetes was caused more by insulin resistance than by the failure of insulin secretion. But most of the variants that have been discovered so far through genome-wide associations affect the risk of type 2 diabetes by influencing secretion of insulin. What we see today is basically that the body outgrows the pancreas, and it only outgrows the pancreas in those who have relatively limited secretion reserve. Those who have more reserve in their pancreas don’t develop type 2 diabetes.

 Do you know how the parent of origin affects the way in which the variant influences diabetes risk?

We have a lot of data indicating that this is because of methylation, the imprinting. And the reason we could isolate this, and no one else has, is that we can determine the parent of origin of all of the variants in the genome. We can do it because we can trace the entire genome; we can determine what comes from the mother and what comes from the father. And currently no one else in the world can do that.

 How do you plan on using the kind of genetic information you’re discovering to do diagnostics?

I can give you an excellent example of a clinical problem that people are struggling with today, where genetics can be incredibly useful. You know that the test for PSA, prostate specific antigen, has been under a lot of attack recently. This is, of course, supposed to be a test to assess the probability of whether a man has prostate cancer.  It measures the concentration of this prostate specific antigen, and the idea is that if you have a concentration over around 4.5, that’s a sign you may have prostate cancer and you should be biopsied. But a very large proportion of those who are biopsied turn out not to have cancer, and a significant proportion of the men who have PSA results below 4.5 turn out to have prostate cancer. So the question is, why is that?

Well, we’ve done a very large study of the genetics of gene expression, which we published in Nature a few years back, and one of the things we learned is that a very large percentage of genes in our genome have a normal distribution of expression (V. Emilsson, et al., Nature, 452[7186]: 423, 2008). If you look at the population, you get a bell curve of expression for these genes. There are those who have naturally very low expression of a gene and those who have very high expression. We’ve also done a very large study looking for genes that have impact on the expression of the PSA gene. And what we found is a coding sequence variant in the genotype; those who have that variant have a mean concentration of PSA only 50% of the mean of the population in general. And there’s another four variants that if you have all of them you have naturally 50% greater expression of the PSA than the population in general (J. Gudmundsson, et al., Nature Genetics, 41[10]: 1122, 2009).

We’re now in the process of turning this into a test, and we’re writing up a paper on it that we’ll submit shortly. We can identify 5% of the men in our society whose PSA is naturally so low that even if it doubles because of prostate cancer, it would still be within normal limits. And we can identify another 5% of men whose PSA is naturally so high that they are constantly being biopsied, even though they don’t have cancer. Men with those four variants are vastly over-represented in negative biopsies. So we have a test that allows us to genetically correct PSA—to apply a conversion factor based on the constellation of alleles that the person has – and we’ve already shown that it greatly increases sensitivity and specificity of the PSA test. This is one example of how you can use genetics to improve the sensitivity and accuracy of a cancer test.

 Is that the only test you have in the works at the moment, or are there others?

Another one is for breast cancer. Remember, the U.S. Preventive Services Task Force recently recommended against mammography screening for women in their forties because of the large number of women that need to be screened to save a single life—1,900 women—and the high rate of false positives that require biopsies even though no cancer is present. So the question is, how do you reduce the number of false positives and therefore the number of negative biopsies that result? Well, this all begins with an ambiguous lesion that requires a biopsy. So you take women who have one of these ambiguous lesions, and you send them out for a genetic test that determines whether they’re among the 1% of women who have quadruple the usual risk of breast cancer. If they are, if they have a high risk, then they get the biopsy. If they’re not, if they have low risk, they get sent home and told to come back in a year for another. You use this information in the decision-making process.

 Have you found that the medical community is interested in using these tests?

Yes. There is interest on behalf of the parties that are making mammography machines. There is interest on behalf of urologists doing PSA tests. There is interest on behalf of patient organizations. And we also have tests in the works that would determine, for instance, the likelihood of sudden cardiac death, and we’re aware of interest in that, as well.

 What unanticipated finding has emerged from your research that you feel hasn’t received the attention it deserves?

"Stefánsson and his deCODE colleagues have discovered more disease-related genes than any other biotech company or research organization, and Stefánsson’s citation record... with more than 10,000 collective citations over the last decade alone in the fields of Molecular Biology & Genetics and Clinical Medicine..."

There is hardly anything in my life that I believe has received the attention it deserves! Okay, more seriously, what we’ve done that I’m most pleased with is our work shedding light on human diversity. We published a paper a few years back in Nature Genetics on the connection between reproductive success and recombination rate (A. Kong, et al., 36[11]: 1203, 2004). We found that women who recombine a lot have more children than women who recombine little, which means there is a selection for the generation of diversity. This is the first time that’s been shown, and it’s terribly important. Humanity has been spread all over the globe, which means that it is exposed to maximum environmental diversity, and through this mechanism our species prepares itself as well as possible for natural disasters, be it volcanic eruption in Iceland, or whatever.

We also found an inversion on chromosome 17—a 900-kilobase inversion—that is under selection. About 17% of women have this inversion, and they have more children than women who have the opposite orientation of this inversion. That’s the first demonstration of selection in current human populations. We subsequently did a genome-wide association study for recombination rate, and we found a haplotype that associates with increased recombination in men but a decreased recombination rate in women. So the same variant , when it is passed through men, increases the recombination rate, and when it is passed through women, it decreases it. This is a brilliant way of containing the recombination rate, because, as I just said, selection for increased recombination rate is a way to increase human diversity, but if that continues forever, it could lead to genomic instability. So here’s a mechanism to contain that—a sequence variant, where every other generation increases recombination rate and every other generation decreases it. I think that’s an absolutely fascinating observation and we are actually, as we speak, writing a paper telling the next chapter in that story.

 Now that deCode has emerged from Chapter 11, what is your prognosis for the next five years?

As I said, we’re the only ones who can basically determine in the entire genome which gene comes from the mother and which ones come from the father. One of the benefits of that is we can impute genotype and sequence data using genealogies to effectively expand the amount of data we get for our studies. If I take a group of Icelanders, by sequencing some of them I can predict the frequency of virtually any genotype across the group with great statistical accuracy, and we can understand which markers are inherited from the mother and which from the father. That allows us to do the following project, which we’re just beginning: We’re going to sequence the complete genome of a couple of thousand Icelanders, and through imputation we’ll be able to expand the data we generate by roughly 100-fold. So, over the next year and a half, we’re going to be able to determine the sequence of the genomes of the entire Icelandic nation. I expect that this is going to lead to an explosion of discoveries, as big as anything ever done in biology. At that point in time, we will know everything about ourselves, so we can retire to some other planet.

KEYWORDS: Kari Stefansson, deCODE genetics, type 2 diabetes, prostate-specific antigen, PSA test, genetic diversity, TCF7L2 variant.

EMAIL   |   BACK TO TOP

• 2010
• September 2010 - Author Commentaries
• For Kári Stefánsson and deCODE, the Diagnostics Look Promising