Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

High-Density Genomewide Linkage Analysis of Exceptional Human Longevity Identifies Multiple Novel Loci

  • Steven E. Boyden,

    Affiliations Department of Genetics, Harvard Medical School, Boston, Massachusetts, United States of America, Program in Genomics, Division of Genetics, and The Manton Center for Orphan Disease Research, Children's Hospital Boston, Boston, Massachusetts, United States of America

  • Louis M. Kunkel

    Affiliations Department of Genetics, Harvard Medical School, Boston, Massachusetts, United States of America, Program in Genomics, Division of Genetics, and The Manton Center for Orphan Disease Research, Children's Hospital Boston, Boston, Massachusetts, United States of America, Howard Hughes Medical Institute, Children's Hospital Boston, Boston, Massachusetts, United States of America



Human lifespan is approximately 25% heritable, and genetic factors may be particularly important for achieving exceptional longevity. Accordingly, siblings of centenarians have a dramatically higher probability of reaching extreme old age than the general population.

Methodology/Principal Findings

To map the loci conferring a survival advantage, we performed the second genomewide linkage scan on human longevity and the first using a high-density marker panel of single nucleotide polymorphisms. By systematically testing a range of minimum age cutoffs in 279 families with multiple long-lived siblings, we identified a locus on chromosome 3p24-22 with a genomewide significant allele-sharing LOD score of 4.02 (empirical P = 0.037) and a locus on chromosome 9q31-34 with a highly suggestive LOD score of 3.89 (empirical P = 0.054). The empirical P value for the combined result was 0.002. A third novel locus with a LOD score of 4.05 on chromosome 12q24 was detected in a subset of the data, and we also obtained modest evidence for a previously reported interval on chromosome 4q22-25.


Our linkage data should facilitate the discovery of both common and rare variants that determine genetic variability in lifespan.


Many common diseases of adulthood increase in prevalence with age. These morbidities accompany an exponential increase in mortality rate that is maintained until approximately age 90, whereupon it starts to decelerate [1]. The reduction in observed versus expected mortality may be due to demographic selection, whereby individuals with alleles predisposing them to an early or average age of death, once deceased, leave behind a robust cohort depleted of detrimental alleles and/or enriched for alleles that promote longevity [2], [3]. Centenarians often reach old age with delayed onset or absence of geriatric diseases [4], possibly benefitting from a “compression of morbidity” that confines these diseases to a short duration at the end of their life [5]. This correlation between exceptional longevity and healthy aging suggests that common genetic factors may underlie both traits. The epidemiology and phenotypes characteristic of human aging and results of candidate gene association studies have been reviewed elsewhere [6][8]. While many variants have demonstrated preliminary evidence of association to exceptional longevity [7], the only confirmed associations are those of the APOE (MIM 107741) haplotypes [9][12]. Multiple recently reported associations between variants in FOXO3 (MIM 602681) and longevity are also quite promising [13][18]. Despite these important discoveries, additional alleles that may regulate aging in humans and allow a minority of the population to attain extreme old age have likely yet to be identified.

Estimates of the heritability of normal human lifespan range from 10% to 58%, averaging about 25% [19]. The genetic contribution to lifespan grows markedly after age 60, indicating the heritability of exceptional longevity may be substantially higher than these estimates [20]. The relative survival probability for siblings of centenarians increases steadily with age, until male and female siblings have a 17-fold and 8-fold increased chance, respectively, of reaching age 100 compared to others from their birth cohort [21]. Moreover, while natural lifespan is likely a complex trait controlled by many genes with small effect sizes, extreme longevity may be determined by fewer genes of stronger effect [22], [23], and may therefore be amenable to linkage analysis. The only previous genomewide scan for linkage to longevity was conducted in part by a member of our group (LMK) and identified a region on chromosome 4q22-25 as significantly linked in 137 sibships of centenarians and nonagenarians [24]. A subsequent genomewide scan for healthy aging in a smaller and younger cohort provided weak support for the chromosome 4q22-25 linkage [25], whereas a targeted study of 164 sibships of nonagenarians did not find linkage to the locus [26], nor did a genomewide scan on bone characteristics as a biomarker for biological aging [27]. All these studies used microsatellite markers with 5–10 cM spacing. To assess the linkage to chromosome 4 and identify new loci, we performed the most powerful linkage scan to date on exceptional longevity. Though the evidence for linkage to chromosome 4 remains equivocal, several novel loci were discovered in our scan, including a region on chromosome 3p24-22 with an empirically genomewide significant LOD score of 4.02 and a region on chromosome 9q31-34 with a LOD score of 3.89.


Ethics Statement

Subjects were recruited through Elixir Pharmaceuticals, the New England Centenarian Study (NECS) now of Boston University Medical Center, Beth Israel Deaconess Medical Center (BIDMC), and Children's Hospital Boston (CHB), as described previously [24], [28]. All participants provided written informed consent and the study was approved by the Institutional Review Boards of the above institutions. All samples were de-identified and were either available from BIDMC or CHB, or were purchased from Elixir Pharmaceuticals or NECS for a processing fee.


All subjects provided proof of age. There was a predominance of female subjects in our cohort, likely reflecting the original ascertainment criterion of having a proband of at least 98 years of age regardless of gender [24]. Only self-identified white or Caucasian subjects (the vast majority of our cohort) were analyzed, since population stratification can confound nonparametric linkage analysis when parental genotypes are unobserved [29]. We had available gender, age at last contact, and alive versus deceased status as of last contact. Age at last contact was not a suitable approximation for our phenotype of interest, age at death, because 70% of our cohort was living. To produce a more homogeneous phenotype that would better estimate age at death, we calculated an expected age at death, which for deceased subjects was equal to their actual age at death, and for living subjects was equal to their age at last contact plus their age-specific and gender-specific life expectancies from life tables for the 1900 birth cohort [30]. The median year of birth for our cohort was 1901.

Since we could not predict what minimum age requirements would provide optimal power, ten sets of gender-specific minimum expected age at death requirements were applied to all subjects, with the hypothesis that as the cutoffs increased, the loss of power due to the decreasing number of families might be partially offset by an increase in genetic homogeneity and/or magnitude of effects in older subjects. As designated hereafter, Categories 1 to 10 cover the range from the upper 5% to the upper 0.2% tail of the birth cohort, corresponding to a minimum expected age at death of 90 to 100 for males and 95 to 104 for females (Table 1). Subjects that did not meet the age criteria for a category were removed from the analysis; for sibships with more than two siblings, individual siblings were excluded while retaining the sibship, whereas once either member of a sibling pair was eliminated, the entire sibship was removed.

Table 1. Subject characteristics by age category for Total group.

In addition to analyzing the complete Categories 1 to 10 (the “Total” group), we divided the sample set by two criteria into two subgroups for each. To explore differences between this study and the first scan [24], we analyzed (in Category 1) 129 of the 137 families used in the original cohort (the “Previous” subgroup) separately from 150 families recruited since that study was completed (the “New” subgroup). Because life expectancies show a gender bias, we also split the Total group into 140 sibships (in Category 1) with at least one male member (the male-containing, or “MC” subgroup) and 139 sibships comprised of only females (the female-only, or “FO” subgroup) (Table S1). Results that apply across all ten categories, in any group, are referred to as “Overall” for that group (for example, the Overall maximum LOD score).

Genotyping and Filtering

Genomic DNA samples were purified from blood as described previously [24], and were subjected to whole genome amplification using the GenomiPhi kit (Amersham). Amplified DNA was genotyped at 10,204 single nucleotide polymorphisms (SNPs) on the GeneChip Human Mapping 10K 2.0 Array (Affymetrix). Samples that did not achieve at least a 95% SNP Call Rate were re-genotyped or excluded. Genotype concordance checks were performed to verify sibling status and eliminate duplicate samples, monozygotic twins, and unrelated subjects. Seven sibships were found to be comprised of at least one pair of half-siblings, which were included. Among 642 successfully genotyped samples, of which 632 subjects in 279 sibships qualified for Category 1, the mean SNP Call Rate was 98.81%, ranging from 95.03% to 99.87%. There were 109 SNPs that were not assigned to a chromosome, 170 SNPs that had a Hardy-Weinberg equilibrium P value below 10−3 among 282 unrelated subjects, and 267 SNPs that had a call rate of below 90% across all 642 samples. A total of 453 SNPs meeting one or more of these criteria was eliminated, resulting in a final panel of 9751 SNPs.

Linkage Analysis

Linkage analysis was performed using MERLIN/MINX v1.1.2 [31]. Genotype data were converted into MERLIN-compatible input files using the Affymetrix tool GDASPort. All siblings were encoded as affected and their ungenotyped parents were encoded as phenotype unknown. Marker map positions based on the deCODE Genetics sex-averaged genetic map and marker allele frequencies for a Caucasian population were provided by Affymetrix. The Total group was also analyzed allowing MERLIN to calculate founder marker allele frequencies in each category separately, and the results were negligibly different than with the Affymetrix frequencies (data not shown).

We computed a multipoint nonparametric Kong and Cox allele-sharing LOD score with the Sall scoring function and the exponential allele-sharing model [32], since the commonly used nonparametric linkage (NPL) score is overly conservative with missing data [32][34] and our parental genotypes were universally absent. Hereafter, “LOD score” refers to these conditions. A parametric heterogeneity LOD (hLOD) score was also computed under both a dominant and recessive model, along with a conventional parametric LOD score. The penetrance for non-susceptible genotypes (phenocopy) under each model was arbitrarily set to one-tenth the birth cohort trait prevalence for that age category, and the penetrance for the susceptible genotypes was set to 1 since we analyzed only affected subjects. The “disease” allele frequencies were then calculated to fit the prevalence and penetrance model (Table S1). To limit the multiple testing burden, the results were screened only on the basis of the nonparametric LOD scores; parametric hLOD and LOD scores were noted only in genomic regions for which the nonparametric LOD scores were considered interesting (LOD≥2). The hLOD scores were used to verify that the nonparametric results were robust to different assumptions and provide information about the possible mode of inheritance of the putative longevity allele. Empirical P values were determined solely on the basis of nonparametric LOD scores.

We used the --rsq option in MERLIN [35] to accommodate linkage disequilibrium (LD) between markers. SNPs for which the pairwise coefficients of determination (r2) exceeded 0.16, above which LOD score inflation due to inter-marker LD becomes appreciable when parental genotypes are unobserved [36], were clustered with intervening SNPs and treated as a single multi-allelic marker. The error checking and Pedwipe functions of MERLIN were used to remove unlikely genotypes as implied by close double crossovers. Mendelian inheritance errors could not be identified due to the lack of parental genotypes.

Empirical P Values

To account for the multiple partially-dependent hypotheses represented by the ten age categories, we calculated an Overall empirical P value for linkage peaks from any category in the Total group that met the significance threshold of LOD  = 3.6 for a single genomewide scan with a fully informative marker panel [37]. Based on the error-wiped pedigrees from Category 1, MERLIN was used to generate 1000 replicates of simulated genotype data under the null hypothesis of no linkage, applying the same marker map, allele frequencies, LD-corrected cluster definitions and haplotype frequencies (generated from the original scan with the --cfreq option), and missing data pattern as the actual data. The replicate pedigree files were successively trimmed nine times each to produce the corresponding pedigree files for Categories 2 to 10. These 1000 sets of ten replicates were analyzed identically to the actual data, and the number of distinct replicates or genomic locations in like replicates across all age categories that met or exceeded the observed LOD score for a given peak was recorded. If a replicate exceeded the observed LOD score in multiple categories at the same location, it was counted only once, just as in the original analysis.

We used the formula P = (r+1)/(n+1) to calculate empirical P values, where r is the number of replicates reaching the observed LOD score, and n is the number of replicates analyzed [38]. This formula produces a biased estimate of the true underlying P value, but provides a more accurate estimate of the type I error rate than the conventional, unbiased (yet anticonservative) formula P = r/n [38][41]. Upper 95% confidence limits for empirical P values were calculated using the conservative Clopper-Pearson exact method for a binomial distribution [42], as implemented in an online calculator [43].


In the Total group we identified a region on chromosome 3p24-22 with an Overall maximum LOD score of 4.02 in Category 4, and a region on chromosome 9q31-34 with an Overall maximum LOD score of 3.89 in Category 8. No other intervals achieved a LOD score above 3.0 (Figure 1), and given the high probability that peaks below that magnitude represent false positives, they were not considered further. By averaging the simulation results across all ten age categories, we calculated the per-category empirical threshold for genomewide significance with a type I error rate of α = 0.05 to be a LOD score of approximately 3.1, likely reflecting the imperfect informativity of the 10K panel [37]. Correcting for all categories, a LOD score of approximately 3.9 was required for Overall genomewide significance. For the chromosomes 3 and 9 linkage peaks, respectively, we observed 36 and 53 distinct genomic locations among 1000 sets of replicates for which a LOD score in any age category met or exceeded the observed scores, resulting in corresponding empirical P values of 0.037 [upper 95% confidence limit 0.048] and 0.054 [upper 95% confidence limit 0.067]. Only a single replicate achieved two LOD scores of at least 3.89 at distinct locations across all categories, yielding an empirical P value for the combined result of 0.002 [upper 95% confidence limit 0.006]. The stability of the chromosomes 3 and 9 linkage peaks was evaluated by dividing the marker panel into two independent subsets containing every other SNP, and both regions retained strong evidence for linkage in the halved panels (data not shown).

Figure 1. Genomewide nonparametric LOD scores by age category for Total group.

Chromosome number is indicated at the bottom of each plot. Red line denotes empirical genomewide significance threshold across all age categories of LOD = 3.9.

The LOD scores for all linkage peaks varied considerably by group (Figure 2). The chromosome 4q22-25 linkage peak reported previously [24] produced an Overall maximum LOD score of 1.01 in Total Category 2. In the same region in the Previous subgroup we obtained maximum LOD scores of 2.14, 1.99, and 2.20 in Categories 1 to 3, respectively, whereas in the New subgroup, there was no evidence for linkage. For chromosomes 3 and 9, the Previous subgroup provided strong evidence for linkage with LOD scores of 3.90 in Category 4 (Overall maximum 4.49 in Category 5) and 3.43 in Category 8, respectively, while in the New subgroup the corresponding maximum LOD scores were 1.19 and 1.12. Analysis of the New subgroup also revealed a third novel locus on chromosome 12q24 with an Overall maximum LOD score of 4.05 in Category 6. This peak was completely absent from the Previous subgroup, and accordingly, the original cohort [24]. The Overall maximum LOD score for the chromosome 12 peak in the Total group was 1.11 in Category 5.

Figure 2. Nonparametric LOD scores by group for chromosomes 3, 9, 12, and 4 linkage peaks.

LOD scores for the entire chromosome are plotted for the Category that produced the highest LOD score in the Total group for chromosomes 3 and 9, and the highest LOD score in the New and Previous subgroups for chromosomes 12 and 4, respectively. A) Chromosome 3, Category 4. B) Chromosome 9, Category 8. C) Chromosome 12, Category 6. D) Chromosome 4, Category 3.

We did not discover any additional linkage peaks in the gender-stratified analyses. The chromosome 3 linkage peak was somewhat stronger in the FO group (Overall maximum LOD score of 2.77 in Category 5) than in the MC group (Overall maximum LOD score of 1.85 in Category 4), whereas the chromosome 9 linkage peak showed a greater difference in the opposite direction (Overall maximum LOD scores of 3.62 in MC Category 8 and 1.30 in FO Category 9, though the disparity may be explained by the greater number of MC than FO sibships in Categories 8 and 9). The subgroup characteristics are given in Table S1, and the complete nonparametric linkage data for the Total group and all four subgroups are provided in Text S1, S2, S3, S4 and S5.

The parametric analysis generally supported the results of the nonparametric analysis, showing all four linkage peaks in their respective groups and categories to be robust to different assumptions about the mode of inheritance. The chromosome 3 peak produced a higher maximum hLOD score under a recessive model (4.652) than a dominant model (3.334), whereas the maximum hLOD score for the chromosome 9 peak was higher under the dominant model (3.883) than the recessive model (2.680). The dominant model on chromosome 9 also yielded positive conventional parametric LOD scores of 1.047 in Category 8 and 2.840 in Category 9, suggesting reduced locus heterogeneity in the oldest sibships. The chromosome 12 peak achieved similar maximum hLOD scores of 3.915 and 3.875 under the dominant and recessive models, respectively, which likewise produced similar results on chromosome 4. We did not assess whether differences in hLOD scores under the dominant and recessive models were significant. Overall maximum nonparametric LOD and parametric hLOD scores, P values, and linkage peak locations for chromosomes 3, 9, 12, and 4 are given in Table 2.

Table 2. Characteristics of chromosomes 3, 9, 12, and 4 linkage peaks.


We performed the second genomewide linkage scan on families of exceptionally long-lived siblings, with substantial improvements in power over the first scan [24]. A filtered panel of 9751 SNPs with an average minor allele frequency of 0.27, equivalent to approximately 4000 microsatellite markers [44], provided a ten-fold effective increase in marker density over the original scan. The resulting advantage in information content and power of similar SNP arrays over microsatellite panels has been demonstrated in theory [45] and in practice [46]. Our analysis of 279 families in Total Category 1 included 129 of the 137 sibships in the previous study plus an independent cohort of 150 sibships. For a set of 300 sibpairs the power to detect linkage is nearly 100% for an allele conferring a two-fold increased risk to siblings of affected individuals [47], [48]. If genes with such magnitudes of effect exist they may be the most relevant to the study of human aging, and our scan would be well-powered to detect them.

Across ten categories of minimum expected age at death requirements, we identified three novel loci of interest: an Overall genomewide significant peak on chromosome 3p24-22 (LOD = 4.02, P = 0.037), a highly suggestive peak on chromosome 9q31-34 (LOD = 3.89, P = 0.054), and a peak on chromosome 12q24 (LOD = 4.05) in the newly recruited subset of our subjects. These linkage peaks are preliminary results that warrant replication studies, as several factors may influence their significance (Text S6). None of our scans provided substantial evidence for linkage to FOXO3 on chromosome 6, probably because of the relatively small effect sizes conferred by SNPs at that locus [13][18]. Consistent with previous studies [24], [49], the APOE gene on chromosome 19 was likewise not linked in any analysis, again possibly due to the low relative mortality risks of the ε2 and ε4 haplotypes [50]. In addition, the ε4 haplotype predisposes young individuals to elevated mortality rates and is thus depleted in the exceptionally old [11], but such negatively selected alleles do not result in significant excess allele sharing among long-lived siblings [49]. By contrast, the ε2 haplotype is enriched in the exceptionally old [11], but the apparent protective effect may be mediated by a heterozygote advantage mechanism [50]-[52]. The ε2 haplotype is then analogous to a rare dominant variant, for which the power of allele-sharing methods is also low [49], [53].

The only previous genomewide linkage scan for exceptional longevity used a minimum actual age at death or last contact of 98 for probands and 91 or 95 for additional male or female siblings, respectively [24], most closely matching our Categories 1 and 2. The previous study produced distinct linkage peaks at both the 3p24-22 and 9q31-34 loci reported here, but with relatively small LOD scores of about 0.8 and 1.0, respectively. Our corresponding Overall maximum LOD scores for the two peaks in the Previous subgroup were 4.49 and 3.43 in Categories 5 and 8, but the scores in Category 1 were only 3.22 and 0.18. The chromosome 12 linkage peak reported here only in the New subgroup was completely absent from the previous study. Therefore, compared to the previous scan the discovery of the three novel loci was probably most attributable to different factors in each case: the increased informativity of the 10K marker panel for chromosome 3, the use of a range of age cutoffs for chromosome 9, and the expansion of our sample set for chromosome 12. Other previous scans using age-related traits have provided some evidence for linkage to chromosome 3p [25], [27], but given the disparate phenotypes and coarse microsatellite maps in these studies, it is difficult to assess the relevance of their data to our results.

Future linkage studies may benefit in power from denser marker panels [54] and larger sample sizes, but detection of loci with small effects and gene identification require other methods [55]. An association study of genes under the previously reported chromosome 4q22-25 linkage peak [24] implicated a SNP in the promoter of MTTP (MIM 157147) as associated with longevity [28], but multiple attempts to replicate this result were unsuccessful [12], [26], [28], [56]-[58]. It has been suggested that these studies also cast doubt on the prior linkage evidence [59], when in fact the two issues are largely independent, and this study was the first adequately powered attempt to replicate the previous scan. Although we were unable to reproduce the chromosome 4 linkage in the Total group or independent New subgroup, the evidence was stronger when only sibships used in the original study were analyzed (Overall maximum LOD score of 2.20 in Previous Category 3), indicating heterogeneity between subgroups. Importantly, only 129 of the 137 original sibships were available for this study, which could help account for the failure of our LOD score in Previous Category 1 (2.14) to reach the LOD score of 3.65 reported previously [24]. However, the Maximum LOD Score (MLS) statistic [60], [61] with the possible triangle constraint [62], as implemented by GeneHunter [33] in the original report [24], was later documented to result in a slight anticonservative bias that was most pronounced in small sample sizes [34]. The Kong and Cox LOD score [32] used here was not subject to that bias [34], possibly indicating the LOD score from the previous scan was somewhat inflated relative to our scan.

Several candidate genes in the linkage peaks discussed here warrant mention. The gene TOP2B (MIM 126431) in the chromosome 3 linkage peak encodes an isozyme of topoisomerase II, which is potently inhibited by resveratrol and related compounds from grape cell culture [63]. Resveratrol can mimic the effects of caloric restriction, mitigate the symptoms of age-related diseases, and/or extend lifespan in a variety of model organisms, including mammals [64], [65]. Various human topoisomerase homologs have also been shown to regulate cellular senescence [66], [67], promote telomere stability [68], and interact with the RecQ helicases encoded by WRN (MIM 604611) [69] and BLM (MIM 604610) [70], in which mutations cause the progeroid disorders Werner syndrome [71] and Bloom syndrome [72], respectively. DBC1 (MIM 602865) in the chromosome 9 linkage peak directly interacts with SIRT1 (MIM 604479) and inhibits its activity [73]; SIRT1 is activated by resveratrol [74] and has been implicated in several age-related phenotypes in mammals [75]. The sirtuin family members are key regulators of lifespan in yeast [76], worms [77], and flies [78] and mediate the effects of caloric restriction [79], [80], the only behavior known to increase lifespan in a wide variety of organisms, including mammals [81]. In response to nutrient withdrawal, SIRT1 is stimulated by FOXO3 [82], in which variants have been reproducibly associated with human longevity [13][18]. In addition, different SNPs in or near TLR4 (MIM 603030) in the chromosome 9 linkage peak have been associated at least once, though often not reproducibly, with exceptional longevity in men [83], [84], a bone-related proxy for biological age [85], and various age-related diseases [86]. Also of note, mutations in ANK2 (MIM 106410), in the previously reported chromosome 4 linkage peak, cause long-QT syndrome [87], and common variation in ANK2 has been reported to regulate the QT interval [88]. QT interval prolongation is a risk factor for sudden cardiac death in healthy individuals [89] as well as those with ischemic heart disease [90] and chronic congestive heart failure [91], which are more common causes of death among centenarians than younger individuals [92]. Heterozygous Ank2+/− knockout mice displayed multiple signs of premature senescence and their lifespan was significantly reduced compared to wild-type littermates [93]. Finally, a nominally significant association was reported between a SNP in ALPK1 (MIM 607347) and both age at death and morbidity-free status at age 65 [85]. Both ANK2 and ALPK1 are within the 99% confidence (-2 LOD) [94] interval for the original chromosome 4 linkage peak [24] but are outside the 85% confidence (-1 LOD) interval tested in the study that identified MTTP [28], raising the possibility that ANK2, ALPK1, or another gene besides MTTP could explain the previous linkage result. The positions of all the above genes within linkage peaks suggest they merit attention in future gene identification efforts.

The linkage scans reported here should contribute to the analysis of both linkage and association studies on exceptional human longevity, which are underway in a massive data set [95]. Since cross-sectional study designs for longevity are subject to unique methodological complications [11] and longitudinal cohort designs can be prohibitively expensive, case-control studies will benefit from direction or corroboration by linkage scans. For example, the extensive multiple testing problem encountered in genomewide association studies can be partially alleviated by a weighted Bonferroni correction or Bayesian analysis that employs linkage data [96]. Alternatively, our results could help inform deep resequencing efforts to identify rare variants that influence lifespan, a particularly suitable approach for genes in which multiple variants exert individual effects too weak to be detected by association methods. Discovery and confirmation of human longevity genes will provide insight into the biology of aging and the genetic basis for resistance to age-related disease.

Supporting Information

Table S1.

Subject characteristics by age category for four subgroups, plus parametric analysis settings. Age, gender, and sibship composition statistics for the Previous, New, MC, and FO subgroups, plus parametric linkage parameters for all groups, are given across ten categories defined by gender-specific minimum requirements for expected age at death. The designations 2-ship, 3-ship, 4-ship, and 5-ship refer to sibships with two, three, four, or five siblings, respectively.

(0.14 MB DOC)

Text S1.

Nonparametric linkage data for Total group.

(10.21 MB TXT)

Text S2.

Nonparametric linkage data for Previous subgroup.

(10.05 MB TXT)

Text S3.

Nonparametric linkage data for New subgroup.

(9.74 MB TXT)

Text S4.

Nonparametric linkage data for MC subgroup.

(10.03 MB TXT)

Text S5.

Nonparametric linkage data for FO subgroup.

(9.75 MB TXT)

Text S6.

Supplemental discussion.

(0.05 MB DOC)


We thank Stephanie Brewster, Stevan Obradovic, Stacy Andersen, and Thomas Perls for providing DNA samples and phenotype information, and Eric Boyden, Lynn Boyden, and Karen Mowrer for helpful discussions and critical reading of the manuscript.

Author Contributions

Conceived and designed the experiments: SEB LMK. Performed the experiments: SEB. Analyzed the data: SEB. Wrote the paper: SEB LMK.


  1. 1. Vaupel JW, Carey JR, Christensen K, Johnson TE, Yashin AI, et al. (1998) Biodemographic trajectories of longevity. Science 280: 855–860.
  2. 2. Vaupel JW, Manton KG, Stallard E (1979) The impact of heterogeneity in individual frailty on the dynamics of mortality. Demography 16: 439–454.
  3. 3. Vaupel JW, Yashin AI (1985) Heterogeneity's ruses: some surprising effects of selection on population dynamics. Am Stat 39: 176–185.
  4. 4. Hitt R, Young-Xu Y, Silver M, Perls T (1999) Centenarians: the older you get, the healthier you have been. Lancet 354: 652.
  5. 5. Fries JF (1980) Aging, natural death, and the compression of morbidity. N Engl J Med 303: 130–135.
  6. 6. Browner WS, Kahn AJ, Ziv E, Reiner AP, Oshima J, et al. (2004) The genetics of human longevity. Am J Med 117: 851–860.
  7. 7. Christensen K, Johnson TE, Vaupel JW (2006) The quest for genetic determinants of human longevity: challenges and insights. Nat Rev Genet 7: 436–448.
  8. 8. Martin GM, Bergman A, Barzilai N (2007) Genetic determinants of human health span and life span: progress and new opportunities. PLoS Genet 3: e125.
  9. 9. Davignon J, Bouthillier D, Nestruck AC, Sing CF (1988) Apolipoprotein E polymorphism and atherosclerosis: insight from a study in octogenarians. Trans Am Clin Climatol Assoc 99: 100–110.
  10. 10. Schachter F, Faure-Delanef L, Guenot F, Rouger H, Froguel P, et al. (1994) Genetic associations with human longevity at the APOE and ACE loci. Nat Genet 6: 29–32.
  11. 11. Lewis SJ, Brunner EJ (2004) Methodological problems in genetic association studies of longevity—the apolipoprotein E gene as an example. Int J Epidemiol 33: 962–970.
  12. 12. Novelli V, Viviani Anselmi C, Roncarati R, Guffanti G, Malovini A, et al. (2008) Lack of replication of genetic associations with human longevity. Biogerontology 9: 85–92.
  13. 13. Kuningas M, Magi R, Westendorp RG, Slagboom PE, Remm M, et al. (2007) Haplotypes in the human Foxo1a and Foxo3a genes; impact on disease and mortality at old age. Eur J Hum Genet 15: 294–301.
  14. 14. Willcox BJ, Donlon TA, He Q, Chen R, Grove JS, et al. (2008) FOXO3A genotype is strongly associated with human longevity. Proc Natl Acad Sci U S A 105: 13987–13992.
  15. 15. Flachsbart F, Caliebe A, Kleindorp R, Blanche H, von Eller-Eberstein H, et al. (2009) Association of FOXO3A variation with human longevity confirmed in German centenarians. Proc Natl Acad Sci U S A 106: 2700–2705.
  16. 16. Anselmi CV, Malovini A, Roncarati R, Novelli V, Villa F, et al. (2009) Association of the FOXO3A locus with extreme longevity in a southern Italian centenarian study. Rejuvenation Res 12: 95–104.
  17. 17. Pawlikowska L, Hu D, Huntsman S, Sung A, Chu C, et al. (2009) Association of common genetic variation in the insulin/IGF1 signaling pathway with human longevity. Aging Cell 8: 460–472.
  18. 18. Li Y, Wang WJ, Cao H, Lu J, Wu C, et al. (2009) Genetic association of FOXO1A and FOXO3A with longevity trait in Han Chinese populations. Hum Mol Genet.
  19. 19. Cournil A, Kirkwood TB (2001) If you would live long, choose your parents well. Trends Genet 17: 233–235.
  20. 20. Hjelmborg Jv, Iachine I, Skytthe A, Vaupel JW, McGue M, et al. (2006) Genetic influence on human lifespan and longevity. Hum Genet 119: 312–321.
  21. 21. Perls TT, Wilmoth J, Levenson R, Drinkwater M, Cohen M, et al. (2002) Life-long sustained mortality advantage of siblings of centenarians. Proc Natl Acad Sci U S A 99: 8442–8447.
  22. 22. Gudmundsson H, Gudbjartsson DF, Frigge M, Gulcher JR, Stefansson K (2000) Inheritance of human longevity in Iceland. Eur J Hum Genet 8: 743–749.
  23. 23. Kerber RA, O'Brien E, Smith KR, Cawthon RM (2001) Familial excess longevity in Utah genealogies. J Gerontol A Biol Sci Med Sci 56: B130–139.
  24. 24. Puca AA, Daly MJ, Brewster SJ, Matise TC, Barrett J, et al. (2001) A genome-wide scan for linkage to human exceptional longevity identifies a locus on chromosome 4. Proc Natl Acad Sci U S A 98: 10505–10508.
  25. 25. Reed T, Dick DM, Uniacke SK, Foroud T, Nichols WC (2004) Genome-wide scan for a healthy aging phenotype provides support for a locus near D4S1564 promoting healthy aging. J Gerontol A Biol Sci Med Sci 59: 227–232.
  26. 26. Beekman M, Blauw GJ, Houwing-Duistermaat JJ, Brandt BW, Westendorp RG, et al. (2006) Chromosome 4q25, microsomal transfer protein gene, and human longevity: novel data and a meta-analysis of association studies. J Gerontol A Biol Sci Med Sci 61: 355–362.
  27. 27. Karasik D, Hannan MT, Cupples LA, Felson DT, Kiel DP (2004) Genetic contribution to biological aging: the Framingham Study. J Gerontol A Biol Sci Med Sci 59: 218–226.
  28. 28. Geesaman BJ, Benson E, Brewster SJ, Kunkel LM, Blanche H, et al. (2003) Haplotype-based identification of a microsomal transfer protein marker associated with the human lifespan. Proc Natl Acad Sci U S A 100: 14115–14120.
  29. 29. Wang T, Elston RC (2005) The bias introduced by population stratification in IBD based linkage analysis. Hum Hered 60: 134–142.
  30. 30. Bell FC, Miller ML (2005) Life Tables for the United States Social Security Area 1900–2100. U.S. Social Security Administration. Available: Accessed 2009.
  31. 31. Abecasis GR, Cherny SS, Cookson WO, Cardon LR (2002) Merlin—rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 30: 97–101.
  32. 32. Kong A, Cox NJ (1997) Allele-sharing models: LOD scores and accurate linkage tests. Am J Hum Genet 61: 1179–1188.
  33. 33. Kruglyak L, Daly MJ, Reeve-Daly MP, Lander ES (1996) Parametric and nonparametric linkage analysis: a unified multipoint approach. Am J Hum Genet 58: 1347–1363.
  34. 34. Cordell HJ (2004) Bias toward the null hypothesis in model-free linkage analysis is highly dependent on the test statistic used. Am J Hum Genet 74: 1294–1302.
  35. 35. Abecasis GR, Wigginton JE (2005) Handling marker-marker linkage disequilibrium: pedigree analysis with clustered markers. Am J Hum Genet 77: 754–767.
  36. 36. Boyles AL, Scott WK, Martin ER, Schmidt S, Li YJ, et al. (2005) Linkage disequilibrium inflates type I error rates in multipoint linkage analysis when parental genotypes are missing. Hum Hered 59: 220–227.
  37. 37. Lander E, Kruglyak L (1995) Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 11: 241–247.
  38. 38. North BV, Curtis D, Sham PC (2002) A note on the calculation of empirical P values from Monte Carlo procedures. Am J Hum Genet 71: 439–441.
  39. 39. Broman KW, Caffo BS (2003) Simulation-based P values: response to North et al. Am J Hum Genet 72: 496.
  40. 40. Ewens WJ (2003) On estimating P values by the Monte Carlo method. Am J Hum Genet 72: 496–498.
  41. 41. North BV, Curtis D, Sham PC (2003) A note on calculation of empirical P values from Monte Carlo procedure. Am J Hum Genet 72: 498–499.
  42. 42. Clopper CJ, Pearson ES (1934) The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika 26: 404–413.
  43. 43. Sauro J, Lewis JR (2005) Estimating completion rates from small samples using binomial confidence intervals: comparisons and recommendations. Proceedings of the Human Factors and Ergonomics Society 49th Annual Meeting 5: 2100–2104.
  44. 44. Kruglyak L (1997) The use of a genetic map of biallelic markers in linkage studies. Nat Genet 17: 21–24.
  45. 45. Sawcer SJ, Maranian M, Singlehurst S, Yeo T, Compston A, et al. (2004) Enhancing linkage analysis of complex disorders: an evaluation of high-density genotyping. Hum Mol Genet 13: 1943–1949.
  46. 46. Middleton FA, Pato MT, Gentile KL, Morley CP, Zhao X, et al. (2004) Genomewide linkage analysis of bipolar disorder by use of a high-density single-nucleotide-polymorphism (SNP) genotyping assay: a comparison with microsatellite marker assays and finding of significant linkage to chromosome 6q22. Am J Hum Genet 74: 886–897.
  47. 47. Fishman PM, Suarez B, Hodge SE, Reich T (1978) A robust method for the detection of linkage in familial disease. Am J Hum Genet 30: 308–321.
  48. 48. Risch N (1990) Linkage strategies for genetically complex traits. II. The power of affected relative pairs. Am J Hum Genet 46: 229–241.
  49. 49. Nemani M, Sahbatou M, Blanche H, Thomas G, Pascoe L (2000) The efficiency of genetic analysis of DNA from aged siblings to detect chromosomal regions implicated in longevity. Mech Ageing Dev 119: 25–39.
  50. 50. Gerdes LU, Jeune B, Ranberg KA, Nybo H, Vaupel JW (2000) Estimation of apolipoprotein E genotype-specific relative mortality risks from the distribution of genotypes in centenarians and middle-aged men: apolipoprotein E gene is a “frailty gene,” not a “longevity gene”. Genet Epidemiol 19: 202–210.
  51. 51. Utermann G, Hardewig A, Zimmer F (1984) Apolipoprotein E phenotypes in patients with myocardial infarction. Hum Genet 65: 237–241.
  52. 52. Hayden KM, Zandi PP, Lyketsos CG, Tschanz JT, Norton MC, et al. (2005) Apolipoprotein E genotype and mortality: findings from the Cache County Study. J Am Geriatr Soc 53: 935–942.
  53. 53. Tan Q, Zhao JH, Iachine I, Hjelmborg J, Vach W, et al. (2004) Power of non-parametric linkage analysis in mapping genes contributing to human longevity in long-lived sib-pairs. Genet Epidemiol 26: 245–253.
  54. 54. Shojaee S, Sina F, Banihosseini SS, Kazemi MH, Kalhor R, et al. (2008) Genome-wide linkage analysis of a Parkinsonian-pyramidal syndrome pedigree by 500 K SNP arrays. Am J Hum Genet 82: 1375–1384.
  55. 55. Risch N, Merikangas K (1996) The future of genetic studies of complex human diseases. Science 273: 1516–1517.
  56. 56. Nebel A, Croucher PJ, Stiegeler R, Nikolaus S, Krawczak M, et al. (2005) No association between microsomal triglyceride transfer protein (MTP) haplotype and longevity in humans. Proc Natl Acad Sci U S A 102: 7906–7909.
  57. 57. Bathum L, Christiansen L, Tan Q, Vaupel J, Jeune B, et al. (2005) No evidence for an association between extreme longevity and microsomal transfer protein polymorphisms in a longitudinal study of 1651 nonagenarians. Eur J Hum Genet 13: 1154–1158.
  58. 58. Neville MJ, Clarke R, Evans JG, Rubinsztein DC, Karpe F (2007) Absence of relationship between MTTP haplotypes and longevity. J Gerontol A Biol Sci Med Sci 62: 202–205.
  59. 59. Tan Q, Kruse TA, Christensen K (2006) Design and analysis in genetic studies of human ageing and longevity. Ageing Res Rev 5: 371–387.
  60. 60. Risch N (1990) Linkage strategies for genetically complex traits. III. The effect of marker polymorphism on analysis of affected relative pairs. Am J Hum Genet 46: 242–253.
  61. 61. Risch N (1992) Corrections to “Linkage strategies for genetically complex traits. III. The effect of marker polymorphism on analysis of affected relative pairs”. Am J Hum Genet 51: 673–675.
  62. 62. Holmans P (1993) Asymptotic properties of affected-sib-pair linkage analysis. Am J Hum Genet 52: 362–374.
  63. 63. Jo JY, Gonzalez de Mejia E, Lila MA (2005) Effects of grape cell culture extracts on human topoisomerase II catalytic activity and characterization of active fractions. J Agric Food Chem 53: 2489–2498.
  64. 64. Baur JA, Pearson KJ, Price NL, Jamieson HA, Lerin C, et al. (2006) Resveratrol improves health and survival of mice on a high-calorie diet. Nature 444: 337–342.
  65. 65. Harikumar KB, Aggarwal BB (2008) Resveratrol: a multitargeted agent for age-associated chronic diseases. Cell Cycle 7: 1020–1035.
  66. 66. Michishita E, Nakabayashi K, Ogino H, Suzuki T, Fujii M, et al. (1998) DNA topoisomerase inhibitors induce reversible senescence in normal human fibroblasts. Biochem Biophys Res Commun 253: 667–671.
  67. 67. Humbert N, Martien S, Augert A, Da Costa M, Mauen S, et al. (2009) A genetic screen identifies topoisomerase 1 as a regulator of senescence. Cancer Res 69: 4101–4106.
  68. 68. Temime-Smaali N, Guittat L, Wenner T, Bayart E, Douarre C, et al. (2008) Topoisomerase IIIalpha is required for normal proliferation and telomere stability in alternative lengthening of telomeres. EMBO J 27: 1513–1524.
  69. 69. Lebel M, Spillare EA, Harris CC, Leder P (1999) The Werner syndrome gene product co-purifies with the DNA replication complex and interacts with PCNA and topoisomerase I. J Biol Chem 274: 37795–37799.
  70. 70. Wu L, Davies SL, North PS, Goulaouic H, Riou JF, et al. (2000) The Bloom's syndrome gene product interacts with topoisomerase III. J Biol Chem 275: 9636–9644.
  71. 71. Yu CE, Oshima J, Fu YH, Wijsman EM, Hisama F, et al. (1996) Positional cloning of the Werner's syndrome gene. Science 272: 258–262.
  72. 72. Ellis NA, Groden J, Ye TZ, Straughen J, Lennon DJ, et al. (1995) The Bloom's syndrome gene product is homologous to RecQ helicases. Cell 83: 655–666.
  73. 73. Kim JE, Chen J, Lou Z (2008) DBC1 is a negative regulator of SIRT1. Nature 451: 583–586.
  74. 74. Howitz KT, Bitterman KJ, Cohen HY, Lamming DW, Lavu S, et al. (2003) Small molecule activators of sirtuins extend Saccharomyces cerevisiae lifespan. Nature 425: 191–196.
  75. 75. Haigis MC, Sinclair DA (2010) Mammalian sirtuins: biological insights and disease relevance. Annu Rev Pathol 5: 253–295.
  76. 76. Kaeberlein M, McVey M, Guarente L (1999) The SIR2/3/4 complex and SIR2 alone promote longevity in Saccharomyces cerevisiae by two different mechanisms. Genes Dev 13: 2570–2580.
  77. 77. Tissenbaum HA, Guarente L (2001) Increased dosage of a sir-2 gene extends lifespan in Caenorhabditis elegans. Nature 410: 227–230.
  78. 78. Rogina B, Helfand SL (2004) Sir2 mediates longevity in the fly through a pathway related to calorie restriction. Proc Natl Acad Sci U S A 101: 15998–16003.
  79. 79. Lin SJ, Defossez PA, Guarente L (2000) Requirement of NAD and SIR2 for life-span extension by calorie restriction in Saccharomyces cerevisiae. Science 289: 2126–2128.
  80. 80. Cohen HY, Miller C, Bitterman KJ, Wall NR, Hekking B, et al. (2004) Calorie restriction promotes mammalian cell survival by inducing the SIRT1 deacetylase. Science 305: 390–392.
  81. 81. Mair W, Dillin A (2008) Aging and survival: the genetics of life span extension by dietary restriction. Annu Rev Biochem 77: 727–754.
  82. 82. Nemoto S, Fergusson MM, Finkel T (2004) Nutrient availability regulates SIRT1 through a forkhead-dependent pathway. Science 306: 2105–2108.
  83. 83. Balistreri CR, Candore G, Colonna-Romano G, Lio D, Caruso M, et al. (2004) Role of Toll-like receptor 4 in acute myocardial infarction and longevity. JAMA 292: 2339–2340.
  84. 84. Nebel A, Flachsbart F, Schafer A, Nothnagel M, Nikolaus S, et al. (2007) Role of the toll-like receptor 4 polymorphism Asp299Gly in longevity and myocardial infarction in German men. Mech Ageing Dev 128: 409–411.
  85. 85. Lunetta KL, D'Agostino RB, Sr, Karasik D, Benjamin EJ, Guo CY, et al. (2007) Genetic correlates of longevity and selected age-related phenotypes: a genome-wide association study in the Framingham Study. BMC Med Genet 8: Suppl 1S13.
  86. 86. Balistreri CR, Colonna-Romano G, Lio D, Candore G, Caruso C (2009) TLR4 polymorphisms and ageing: implications for the pathophysiology of age-related diseases. J Clin Immunol 29: 406–415.
  87. 87. Mohler PJ, Schott JJ, Gramolini AO, Dilly KW, Guatimosim S, et al. (2003) Ankyrin-B mutation causes type 4 long-QT cardiac arrhythmia and sudden cardiac death. Nature 421: 634–639.
  88. 88. Sedlacek K, Stark K, Cunha SR, Pfeufer A, Weber S, et al. (2008) Common genetic variants in ANK2 modulate QT interval: results from the KORA study. Circ Cardiovasc Genet 1: 93–99.
  89. 89. Schouten EG, Dekker JM, Meppelink P, Kok FJ, Vandenbroucke JP, et al. (1991) QT interval prolongation predicts cardiovascular mortality in an apparently healthy population. Circulation 84: 1516–1523.
  90. 90. Puddu PE, Bourassa MG (1986) Prediction of sudden death from QTc interval prolongation in patients with chronic ischemic heart disease. J Electrocardiol 19: 203–211.
  91. 91. Barr CS, Naas A, Freeman M, Lang CC, Struthers AD (1994) QT dispersion and sudden unexpected death in chronic heart failure. Lancet 343: 327–329.
  92. 92. Gessert CE, Elliott BA, Haller IV (2002) Dying of old age: an examination of death certificates of Minnesota centenarians. J Am Geriatr Soc 50: 1561–1565.
  93. 93. Mohler PJ, Healy JA, Xue H, Puca AA, Kline CF, et al. (2007) Ankyrin-B syndrome: enhanced cardiac function balanced by risk of cardiac death and premature senescence. PLoS One 2: e1051.
  94. 94. Kruglyak L, Lander ES (1995) High-resolution genetic mapping of complex traits. Am J Hum Genet 56: 1212–1223.
  95. 95. Franceschi C, Bezrukov V, Blanche H, Bolund L, Christensen K, et al. (2007) Genetics of healthy aging in Europe: the EU-integrated project GEHA (GEnetics of Healthy Aging). Ann N Y Acad Sci 1100: 21–45.
  96. 96. Roeder K, Bacanu SA, Wasserman L, Devlin B (2006) Using linkage genome scans to improve power of association in genome scans. Am J Hum Genet 78: 243–252.