Genetic factors involved in human aging become increasingly important1 and specific2 at oldest ages. Therefore, the analysis of subjects strongly selected for longevity provides a powerful tool to disentangle the complex gene network, which influences survival in the elderly.3 Owing to the multifactorial nature of the longevity trait, sampling designs targeted at non-parametric association analyses (both case–control and sib-pair designs) are currently used to discover the genes contributing to the trait. In case–control studies, the problem is that cases (for example, centenarians) and controls (younger subjects) belong to different cohorts, and this may introduce potential biases due to demographic and, to a larger extent, environmental changes occurring along the time. As compared to the case–control design, the sib-pair approach restricts the comparison within sibships, thus ensuring that subjects of each pair belong to quite close cohorts and are exposed in a comparable manner to environmental factors (at least those acting intrafamiliarly). Nevertheless, the possibility that both the sibs attained longevity more because of shared favorable environments than because of shared favorable genes should not be undervalued. A tool to check this possibility is to investigate, in addition to sib-pairs in which both sibs attained longevity (concordant pairs), sib-pairs in which a sole sib shows the longevity trait (discordant sib-pairs). In fact, both in concordant and discordant pairs, the sibs are expected to share intrafamiliar environmental factors. Therefore, the phenotypic concordance in long-living sibs should be due to both genetic and environmental factors; but if longevity is attained by a sole sib (discordant sib-pair), longevity genetic factors are expected to be restricted to the long-living sib only. Thus, different patterns of Identical By Descent (IBD) haplotype sharing are expected in concordant and discordant sib-pairs as chromosomal regions encompassing longevity genetic factors are regarded. In particular, for such regions, an over-sharing of IBD haplotypes (with respect to Mendelian expectation) should be observed in concordant sib-pairs, whereas the contrary (under-sharing) is expected in discordant sib-pairs. The use of discordant sib-pairs in linkage analysis is known in quantitative trait-locus mapping4, 5 but is never applied to explore the genetics of human longevity. Obviously, in such a case, the discordant sib is dead; however, it is possible to go down one generation and to apply IBD analysis to cousin-pairs born from sibs who are concordant or discordant for the longevity trait.

The aim of the present work was to check the feasibility and effectiveness of this novel design by using two chromosomal regions (6p21.3 and 11p15.5) encompassing genes having functional variants that are reported by literature to be associated with longevity.6, 7, 8, 9, 10, 11

The novel design, by implying the collection of offspring of centenarians, also provided us with the opportunity to check Mendelian transmission ratio distortion (TRD)12, 13 in the studied regions. In fact, TRD, which has been defined as ‘a statistically significant departure from Mendelian inheritance ratio expected regardless of the cause,’14 has been observed in these chromosomal regions,15, 16 and this may have an effect on IBD analyses.

The study was carried out in the frame of a European project aimed at investigating genetic and environmental factors affecting aging across Europe (ECHA project; We report here the results of an exploratory genotyping effort performed on the entire DNA collection that we assembled and focused on three and five genes located in chromosomal regions 6p21.3 and 11p15.5, respectively (Table 1).

Table 1 Genes and polymorphisms analyzed in the present study

Materials and methods


Two categories of sib-pairs were considered: concordant versus discordant for the longevity trait. A further requirement was that at least one child of both the proband and his/her sibling was available for sampling, in order to genotype cousin-pairs born from concordant/discordant sib-pairs. Hereafter, the proband, his/her child and the cousin are referred to as a trio. Table 2a summarizes the recruitment criteria for concordant and discordant ECHA families. Thresholds for determining eligibility for sampling were determined as follows:

Table 2 Criteria of recruitment (a) and number of samples collected (b) in Denmark, France and Italy


We set the lower age limit for longevity to the age corresponding to the surviving upper 0.05% of the population, approximately. In addition, as male/female (M/F) ratio among centenarians shows regional disparities,29 we adopted slightly different age windows for recruiting probands in Denmark (Southern Denmark), France (Languedoc-Roussillon) and Italy (Calabria) to get an equal representation of both genders in the study. This implied a 2-year M/F difference in the Italian subsample and a 3-year M/F difference in the Danish and French subsamples.

Concordant/discordant sib-pairs

The choice of age limits for the two groups was dictated by the need for obtaining a balance between a sufficient number of families and an efficient discrimination between the two groups. We then fixed thresholds based upon the survival distributions in each of the surveyed populations: if the proband's sib died at an age corresponding approximately to the range encompassing 40% of the survival in the general population, the sib-pair was considered discordant; sib's survival to an age corresponding to the upper 10% of the survival distribution determined the concordant status. This produced different cutoff ages both for men and women (due to sex-specific survival distributions) and for the three subsamples under study (Table 2a).

Identification of suitable trios

As an initial step to identify sib-pairs fulfilling the above criteria, population pools of 1.6, 2.3 and 2.0 million inhabitants were screened, respectively, to completion in southern Denmark, Languedoc-Roussillon (France) and Southern Italy. As to France, as no centralized register of residents was available, general practitioners and nursing homes of the entire Languedoc-Roussillon region (2.3 millions of inhabitants) were contacted by phone about possible long-lived probands. We started from the identification of the probands according to the requirements listed in Table 2a. Then, for each proband, the sampling of a complete trio was pursued; in the concordant pairs, we also sampled the proband's sib, when available.

Hindrances to sampling to completion

The most common cause of exclusion from sampling turned out to be the lack of the proband's children or their unavailability for sampling. In these cases, reconstruction of the proband's pedigree was not pursued further. It is then not possible to calculate among all sibships, including a centenarian, the proportion of those fulfilling concordance or discordance.

For each trio, information on familial, social and clinical data was collected, respectively, by two questionnaires specifically assembled for probands and offspring (reported at the ECHA web site). All information was properly coded to guarantee anonymity and stored in a dedicated database.

The ECHA design and procedures were approved by the local ethical committees of the participating institutions. Written informed consent was obtained for each recruited subject. Details on aims and tools of the ECHA project, strategies for identification of the sampling targets in the three countries and procedures, and fulfillment of ethical requirements can be found at the ECHA web site (see above).


From each subject, 6 ml of venous blood were drawn and used for DNA preparation according to standard procedures. Figure 1 shows markers and chromosomal regions under study. Protocols used in genotyping are available on request.

Figure 1
figure 1

Schematic representation of the markers localized at 6p21.3 (a) and 11p15.5 (b) chromosomes. Chromosome structure, contig numbers, genes and markers localization are reported. For each gene, exons are represented as filled boxes. The symbols > and < indicate the forward and reverse sequences, respectively. The figures are assembled according to

Data analysis

The reconstruction of haplotype phase and the likelihood of haplotype sharing were obtained by a Maximum Likelihood algorithm included in SimWalk2 software.30 Recombination frequencies between markers were estimated on the basis of the physical distances showed in MapView ( by assuming 1 Mb=1 cM. First, we estimated allele frequencies in each national sample by using one subject for each family (the niece/nephew of the proband). Then, we checked for possible allele frequency differences among the national groups (heterogeneity test by permutations). As the three samples had population-specific patterns of allele frequencies, both haplotype reconstruction and the following IBD analysis were carried out separately in each national sample. We calculated the likelihood of IBD haplotype sharing in each cousin-pair and counted the number of cousin-pairs having the same IBD likelihood. Possible differences between the likelihood distributions obtained in concordant and discordant cousin-pairs were checked by Mann–Whitney U test.

To verify whether haplotypic transmission deviated from the expected Mendelian inheritance ratio (1:1), we inferred the haplotypes transmitted (or not transmitted) from centenarians to offspring and applied a likelihood ratio test by SPSS software (Version 13.0, from SPSS Inc.).



On the whole, 851 and 1567 centenarians’ families have been screened, respectively, in Denmark and Italy. As for France, 750 general practitioners and 300 nursing homes have been contacted. Our effort resulted in a total of 737 subjects from 234 families collected in the three countries (Table 2b). The discrepancy between the total number of subjects and the number of families (234 × 3=702) is accounted for by a few concordant families with two or more sampled subjects in generation I and a few families with more than two sampled subjects in the following generation. As to generation II, the data show that the ages of the sampled subjects are very similar (Table 2b). This similarity was expected for concordant sibships. However, the fact that it is replicated in discordant sibships denotes that our samples represent subjects born at similar parental ages in the two categories of families.

Genetic analyses

All subjects were genotyped at the loci SIRT3, HRAS1, IGF2, INS, TH, TNFα, TNFβand HSP70.1. (Table 1 and Figure 1). Out of more than 5000 genotypings, parent–offspring incompatibility was observed in less than 0.2% of the cases. Re-typing of these samples dismissed incompatibility in all cases; an analysis of Y-chromosome (father/son) and mitochondrial DNA markers (mother/offspring) confirmed parent–offspring compatibility in these cases. Furthermore, a random subset of about 20% of the samples was typed twice, producing consistent results in all cases. Therefore, the impact of genotyping errors in our study can be considered minimal because neither incompatibility nor discrepancy was observed after re-typing.

Genetic similarity in cousin-pairs

As haplotype reconstruction is highly sensitive to the background allele frequencies, we preliminarily checked if the country-specific allele frequencies allowed us to pool the three national samples. As significant heterogeneity was found among the three sample groups (Supplementary Table 1), we reconstructed the segregating haplotypes in each national sample. We then calculated the likelihood that cousin-pairs shared IBD haplotypes and checked for possible differences between the likelihood distributions of concordant and discordant cousin-pairs. Neither at the 6p21.3 region nor at the 11p15.5 region did the cousin-pairs from concordant and discordant siblings show significantly different IBD patterns in any national sample (results not shown).

Transmission ratio distortion

By using country-specific allele frequencies, we first reconstructed the 6p21.3 and 11p15.5 haplotypes transmitted (or not transmitted) from each proband to his/her offspring in each national sample. Then, we analyzed Mendelian transmission in concordant and discordant families. In this case, as we were checking an extrinsic hypothesis (1:1 Mendelian transmission), we pooled the national data to increase the sample size. The results are reported in Supplementary Tables 2 and 3 (concordant and discordant families, respectively), where the rare class encompasses haplotypes present only once. In discordant families, the distribution of transmitted and not transmitted haplotypes significantly differed from the 1:1 Mendelian expectation both at 6p21.3 and 11p15.5 (P=0.007 and P=0.015, respectively), whereas in concordant families, the haplotypic transmission tended to be distorted at 6p21.3 only (P=0.06).


The choice of the optimal study design to discover gene-longevity associations in humans is a real challenge.31, 32 First, because of the effects played on lifespan by social and economic factors, longevity is a phenotype whose definition may change according to the place and year of birth of the cohort; second, true controls in candidate-gene association studies are missing; third, non-parametric linkage analyses require large sample sizes whose statistical power varies according to the dominant/recessive effect of the gene;33 and, finally, longitudinal studies in which cohorts of individuals are followed over time are very difficult to realize. By comparing genetic similarity in cousin-pairs born from sibships of long-lived subjects (concordant sib-pairs) with cousin-pairs born from sibships in which a single subject attained longevity (discordant sib-pairs), the ECHA design aimed at controlling for potential confounding effects of stratification and genetic demographic shifts in the communities from which the subjects were sampled.

As the aim of the work was not to discover new gene-longevity associations but to explore the potentiality of a novel study design, we analyzed functional polymorphisms of genes known to be associated with longevity by case/control studies6, 7, 8, 9, 10, 11 or in strong linkage disequilibrium with functional variants affecting biological pathways involved in aging (Table 1).

The ECHA sampling design (Table 2) required a huge screening effort that produced 737 subjects from 234 families starting from a basin of about 6 million people. However, despite the strong phenotypic selection, the results of IBD analyses did not reveal significant differences between concordant and discordant families. These negative findings, which contrast with the significant gene-longevity associations found by case–control studies, are most likely explained by sample sizes inadequate to reveal small genetic effects,33 which are further smoothed when searched in the children of the probands and their cousins (one generation away from the one selected). A further confusing effect may have been played by TRD, which has been observed in our samples for both the studied chromosomal regions. In this frame, it may be important to notice that the chromosomal regions that we studied were both known to harbor genes that play a role in longevity and turned out to have a significant TRD. Considering that a trade-off between longevity and fertility has been described34, 35 and that TRD is probably due to a low fertility of some allelic combination on specific chromosomal regions,36 it is possible to hypothesize that the chromosomal regions involved in longevity may be more prone to TRD than others. However, further studies are needed to clarify this point. On the whole, the novel design resulted in more complexity and less efficiency in revealing gene-longevity associations in comparison with classic non-parametric approaches.32, 33, 37

In conclusion, the ECHA study failed to evidence differences of IBD haplotype sharing in cousin-pairs born from concordant and discordant sib-pairs. It is likely that a larger sample size was required. On the other hand, the strict criteria of recruitment, although applied to a potential basin of about 6 million people, have not permitted the collection of a higher number of trios matching the ECHA criteria. If our findings will be confirmed by further replication studies, new sampling strategies that focus on the recruitment of large samples of centenarians and their offspring might be planned. In any case, the fulfillment of the ECHA sampling scheme can be used to dimension future campaigns to reach a desired power and/or level of selectivity.