Seeking the Secrets of the Super Long-Lived

With analytical contributions from 23andMe Scientist David Hinds

For all human history people have been obsessed with their own mortality. Early explorers searched the globe for the mythical “fountain of youth,” and now scientists are turning that exploration inward to look at the genomes of people who’ve lived especially long lives.

Longevity runs in families, which suggests that the reason some people live longer than others has something to do with genetics. Indeed, some studies have identified a few genetic variants that occurred more frequently in people who live to be at least 95 years old, for example this small study in Japanese men. These few variants, however, don’t seem to contribute significantly to whether someone will live to be 100.

Larger studies looking broadly across the genome for common variants contributing to longevity have mostly come up empty. A four-study combined analysis published two years ago found no significant associations between common SNPs and longevity, and an initially more promising study published by Paola Sebastiani and Thomas Perls from Boston University was later retracted due to problems with the analysis.

Explorations of lifestyle factors impacting longevity have also turned up somewhat conflicting results. Most studies indicate that those who live longer have healthier diets, exercise more and experience delayed onset of disease, perhaps due to their healthier lifestyles. Smoking seems to lower one’s chances of living longer. Maintaining a healthy weight may also be associated with better cognitive function and lower risk of dementia. But other studies have found no difference in BMI or diet between long-lived individuals and the average population, suggesting that many other factors are involved.

Despite its elusiveness, the quest to uncover the secrets of longevity continues. As more people live longer, larger studies of the super long-lived will become more feasible. The New England Supercentenarian Study led by Boston University, for example, now consists of more than 60 people over the age of 110. Many of these individuals maintain physical independence and mental acuity well past the age of 100. By analyzing the DNA, lifestyle habits, and environmental surroundings of these extremely long-lived people, the researchers hope to learn about genetic as well as non-genetic factors contributing to longevity.

Unfortunately, their most recent findings — a pair of papers authored by the same group (Sebastiani and Perls) whose paper was retracted last year — contain little that can be generalized and have methodological shortcomings that call into question the conclusions they do make.

In the first study, an analysis of the genomes of two supercentenarians, Sebastiani and Perls suggest that the male supercentenarian’s extreme longevity may be due to an excess of common variants linked to longevity. They base this on the man’s enrichment for a set of previously identified longevity-associated genetic variants but without taking into account the fact that many of the variants in that set are correlated with each other. They also point to his high score on the predictive model described in their second paper, a result that is less surprising considering that the man’s data was used to develop that model in the first place. Several other of their conclusions also appear to be over-interpreted.

A close reading of Sebastiani and Perls’ other paper (the republished version of their previous retraction) reveals that the number of genetic variants that they identified for their model is actually quite close to the number of expected false positives (the authors misreport the number of expected false positives by a factor of 10). This means that many of the genetic variants that they suggest are associated with longevity are likely not to be associated with longevity at all. The model’s predictive accuracy is also very high when applied to the training data, but much lower when applied to individuals that were not included in the training set, suggesting that the model does not generalize well.

Studies like these certainly hint at the potential discoveries we could make from analyzing the genomes of people who live to be 100 or older. Can we identify common genetic variation predictive of longevity? Do rarer variants play a role? What genes and pathways are important? Ultimately, however, the latest foray into the genetics of longevity raises more questions about methodology and interpretation than satisfying answers or hypotheses. It appears the secrets of the super long-lived will remain secret for now.






  • Peter Ezzell

    in the republished Sebastiani and Perls’ paper which SNPs correlated to longer lives?

  • Florian

    Actually, there is one allele that is consistently negatively associated with Longevity: APOE4, the same allele that increases the risk of AD. Virtually all longetivity GWAS studies pick it up and in fact, just as with AD, the influence of that allele is so strong that most other novel longevity associated SNPs become insignificant when corrected for APOE4.

    • http://23andme.com Shwu

      In the republished paper, only rs2075650 in TOMM40/APOE (a close proxy to the APOE e2/e3/e4 variant) was significantly associated with longevity at a genome-wide level. Check back soon for a more technical discussion of this paper’s findings.

  • Florian

    The TOMM40 SNP is in LD with Apoe4/3/2. If the data are corrected for ApoE4, rs2075650 is no longer significant.

    There are 9 GWAS studies that have looked at longevity, and only APOE4 consistently scores.

  • Carolyn

    The current problem with all this research is that we know some genes seem to confer longevity … but we don’t know how.

    Maybe researchers should attack the problem using the new theory for longevity, that the very old have bad genes but they also have genes that protect them from the effects of bad genes that kill the rest of us.

    Start with the genomes of a very large number of super-old. Find their “bad genes” and group them accordingly. Who among them should have gotten cancer and maybe died from it many years earlier? Who should have been afflicted with diabetes or heart disease or aneurysms that might have caused death at much younger age? Use these groups to look for other, related, protective genes they have in common.

    I realize that this requires large numbers of subjects and the number of very old people is small. But there are more of them every year that passes, and communication links make it possible to reach people around the world.

    Many elders, I think, would be willing to contribute a little spit to the effort to find out what genes are protective, and this, in turn, might help researchers discover how they work to protect their lucky carriers from the effects of the bad genes they also carry, possibly leading to useful therapies.

  • edu

    Florian, what evidence do you have of LD between rs1054379 A/A and Apoe4? I am homozygous for rs1054379 A/A (the more longevity probability combination) and heterozygous for Apoe4 (e4/e3), so I do not see your linkage disequilibrium anywhere…

Return to top