I have every intention of living forever, but I’m deeply aware of a number of factors that stand in my way. I’m not female, for a start; I wasn’t born to a young mother; I enjoy my food far too much to ever consider caloric restriction; and I hate exercise with a passion. So right now my game plan is basically to rely on advances in medical science, and hope like hell that I have the right genes – bearing in mind that at least 25% of variation in life expectancy is genetically determined.
Finding the actual genes that influence longevity, however, has long proved problematic. In fact, the field of longevity genetics is scattered with the stinking corpses of reported genetic associations that have gone totally unreplicated, with the major exception being variation in the APOE gene (which is also associated with both coronary heart disease and Alzheimer disease, and thus makes good biological sense).
The overall poor track record of longevity genetics stems partly from sexy topic syndrome (topics that are intrinsically interesting are easier to publish crap studies in), and partly from a fundamental methodological problem. Basically, demonstrating an association between a genetic variant and longevity means showing that the variant is at a greater frequency in older people than in young ‘uns – but the frequency of a genetic variant may differ between young and old cohorts due to factors unrelated to longevity, especially through demographic changes (such as migration and population admixture). If a population has seen a large influx of young migrants over the last few decades, then naively comparing genetic variants between “young” and “old” segments of this population will simply give you a whole bunch of variants that are associated with ethnicity.
The best way to avoid this trap is to settle in for a long experiment: harvest DNA from a large, fairly genetically homogeneous group of people, and then sit back and wait to see which ones get voted off the island first (so to speak). At the end of the experiment, compare the DNA from the long-term survivors with those who clocked out early, and look for genetic variants with differing frequency between the two groups. Variants that tend to decrease in frequency as your population ages are good candidates for early death genes.
This is the basic approach adopted by a recent study in the journal PNAS, which studied variation in a set of genes associated with longer life in both worms and flies in a set of elderly Japanese men – and appears to have stumbled across one of the many genes that influence the probability of long-term survival.
The researchers collected DNA back in 1991-1993 from a cohort of 3,741 Japanese men living in Honolulu aged between 71 and 93. Now, fifteen years later, they selected two sub-groups for further analysis: a “survivor” set of 213 men who had survived past the age of 95 by August 2007 (putting them in the top 1% of the life expectancy curve), and a “control” set of 402 men who had died near the mean death age of 78.5 years.
They then made an educated guess regarding the types of genes that might be involved with longevity, selecting a set of five genes associated with the insulin/IGF-1 signalling pathway. This pathway has been implicated in the control of life-span in a wide range of organisms and experimental systems; in addition, mutations in the IGF-1 and IGF-1 receptor genes were recently shown to be over-represented in female Ashkenazi Jewish centenarians.
The researchers then analysed three common sites of variation in each of those genes to look for variants with differing frequency between survivors and controls. Four of the genes yielded nothing of interest, but the fifth – FOXO3A – had significantly altered variant frequencies for all three assayed sites.
Below, I’ve plotted the frequency of the “early-dying” version of the most strongly associated variant in FOXO3A, rs2802292, for individuals in five groups (three “young” and two “old”) scattered across a range of ages at death between 73 and 103. You can see that the youngest-dying individuals have a frequency of around 80%, with a fairly steady decrease to a value of just over 60% in the centenarians. That suggests a steady “purging” of carriers of this version of the gene as the population ages, resulting from a higher death rate in those carriers than in the population as a whole.
The effect is subtle – we’re not talking the death gene here – but it’s statistically significant, and certainly worth pursuing further.
How might FOXO3A influence life expectancy?
The long-term survivors differed from their less fortunate brethren in a number of ways: the survivors had a lower incidence of age-related diseases such as coronary heart disease, stroke and cancer, despite being an average of 11 years older; they were also leaner and had lower blood glucose and insulin levels. Simply comparing the survivors and the early bucket-kickers doesn’t tell us which of these differences are the result of FOXO3A genotype, however.
Some more useful clues came from analysing the effects of FOXO3A variation within the cohorts. In the individuals who cashed in their chips early, the “survivor” version of the gene was associated with significantly lower fasting insulin and a reduced risk of heart disease. Curiously, these associations weren’t seen within the cohort who clung to life into their 90s, who nearly all had low insulin levels regardless of FOXO3A genotype – this may simply reflect the fact that maintaining good insulin sensitivity is a prerequisite for membership of the hard-core grey hair brigade, so individuals who weren’t blessed with the right FOXO3A variant had to have reached that end-point through other avenues (such as better diet or exercise levels, or other genetic variants).
On this basis the researchers argue that FOXO3A probably influences life expectancy primarily through an effect on insulin regulation, which fits plausibly with its known role in the insulin signalling pathway.
Reasons to be cautious
There are always caveats with this type of study. First there are the mundane risks of any genetic association study, which are that the association occurred simply by chance or through some systematic bias in the study design (for instance, genotyping error that affects one cohort more than the other, or hidden population stratification). The researchers have done their best to minimise the risk of systematic bias, but the relatively small sample size by the standards of modern association studies (which often include thousands of participants) is enough to raise some alarm bells. No-one should consider this association to be definitive until it’s been replicated in independent cohorts.
Secondly, the effects of this variant are likely to differ between groups. According to the HapMap database the frequency of the “die early” (A) version of the variant varies substantially between populations: it’s around 76% in Japanese, slightly more common (81%) in Chinese, less common (58%) in Europeans, and quite rare (17%) in West Africans. Given that the Japanese have the highest life expectancy of any of these populations, this is a counter-intuitive result, which reinforces the fact that there are many genes and other factors influencing variation in longevity.
Asian populations also differ from other groups in traits related to body mass, fat deposition and diabetes risk (for instance, Asians tend to have greater predisposition to diabetes at a low body mass index), suggesting that there are important between-population differences in insulin pathway function that could modulate the effects of FOXO3A genotype. In other words, even if this association holds up in other studies, and even if I’m lucky enough to have the right FOXO3A genotype, my white male privilege may still foil my odds of taking full advantage of my genetic heritage.
Finally, there are important mechanistic details that need to be fleshed out. The association between the FOXO3A variant and age-related disease is fairly unconvincing (curiously, in the text the authors state that the variant was significantly associated with cancer and diabetes, but there’s absolutely no evidence of this in the data presented; there’s only a weak association with heart disease). In addition, it’s unclear what the molecular basis of the association might be. The variant identified in this study is almost certainly not causative (it sits in the middle of a long non-coding region of the FOXO3A gene), but is more likely to be a non-functional marker that “tags” a causal variant elsewhere in the gene. If this association holds up, there will need to be some serious investigation into exactly what that causal variant is, how it affects the FOXO3A gene, and how that results in effects on the insulin system and longevity.
This study and others like it have generated some intriguing preliminary results, but it will take much larger and better-controlled studies to pin down the complex, interacting network of genes and environmental factors that control longevity. On the bright side, such studies are now within sight: around the world, massive cohorts are currently being collected for a range of longitudinal analyses of the effects of genes on health outcomes. Combining information about health outcomes and environmental influences with large-scale genetic technologies – up to and including whole-genome sequencing – will provide powerful insights into the factors that predict illness and early death, hopefully allowing clinicians to intervene early in at-risk individuals.
You can bet that I’ll be taking full advantage of such studies once I have my own genome sequence (more on that later). Do I have the right genes to make it to a ripe old age, or will I actually have to take up exercise and a healthy diet to prevent myself from falling off the tree early? We’ll have to wait and see.
B. J. Willcox, T. A. Donlon, Q. He, R. Chen, J. S. Grove, K. Yano, K. H. Masaki, D. C. Willcox, B. Rodriguez, J. D. Curb (2008). FOXO3A genotype is strongly associated with human longevity Proceedings of the National Academy of Sciences, advance online publication DOI: 10.1073/pnas.0801030105