Genetic variation in human telomerase is associated with telomere length in Ashkenazi centenarians

Telomere Length, Heritability, and Longevity.

Phenotypes of centenarians, their offspring (approximate age of 70 years), and offspring-matched controls without a family history of unusual longevity, all of Ashkenazi Jewish descent, are presented in Table 1. The control and offspring groups have similar characteristics of lipid profiles, except that controls have a slightly smaller percentage of large LDL particle size, but these differences are not significant (Table 1). The centenarian group displays a lower mean body mass index and higher HDL levels as compared with controls (Table 1).

When controls and centenarians are combined to represent cross-sectional age groups of unrelated individuals from middle age to oldest old (range: 43–105 years), telomere length declines with age until the age of 85 years (Fig. 1). However, individuals older than 86 years of age as well as the centenarians show longer telomere lengths compared with unrelated individuals younger than 85 years of age, with a mean difference of 0.17 (P = 0.04) in the linear regression model adjusting for age and gender. The telomere length of offspring exceeds that of controls in a hierarchical linear model, on average by 0.082 [95% confidence interval (CI): 0.017–0.148; P = 0.014]. Because centenarians and their offspring are appreciable singleton families, we used the regression approach for nonindependencies of the observed subjects. To account for parametric model bias as well as for unknown intrafamily correlation, we used a sandwich variance estimator equation and a generalized estimating equation (GEE), respectively. Both analyses repeated our first estimation supporting the observation of significant telomere length differences between offspring and controls, on average, by 0.081 (95% CI: 0.022–0.139; P = 0.007) for the sandwich estimation and by 0.082 (95% CI: 0.017–0.147; P = 0.013) for the GEE. Moreover, unlike controls, offspring of centenarians do not show an appreciable decline in telomere length with age as compared with controls (Fig. 1). However, the difference in age-related decline did not differ significantly between offspring and controls (adjusted mean difference = 0.002, 95% CI: −0.007 to +0.010; P = 0.682) by the same hierarchical linear model described earlier. High heritability (86%; P = 0.005) for telomere length between centenarians and their offspring was observed, suggesting that telomere length has a strong genetic component in families with exceptional longevity and that these families have better telomere length maintenance than controls.

Telomere Length, Age-Associated Diseases, and Lipid Profiles of Healthy Aging.

Association analysis between telomere length and major age-related diseases among centenarians, their offspring, and controls indicated that significantly shorter telomere lengths (adjusted for age, gender, and group) are present in subjects with hypertension (P = 0.006), the metabolic syndrome (P = 0.03), or diabetes (P = 0.03) compared with subjects without these disorders (Fig. 2). Moreover, when we tested the relation between telomere length and cognitive function among centenarians, centenarians with impaired cognitive function [Mini-Mental State Examination (MMSE) score ≤25] have significantly shorter telomeres as compared with centenarians with normal cognitive function after adjustment for age and gender (Fig. 2; P = 0.02). We then tested whether telomere length correlated with the lipid profiles, which are known predictors of aging-related diseases such as coronary artery disease and metabolic syndromes (29). These analyses revealed that telomere length is associated with the lipid profiles of healthy aging, showing a positive association with increased particle sizes of LDL (P = 0.0001) and HDL (P = 0.03), percentage of large particle sizes of LDL (P = 0.0002) and HDL (P = 0.038), and levels of Apo-A1 (P = 0.005) and HDL (P = 0.04) (Table 2), whereas there is a negative association with very LDL levels (P = 0.008). Total cholesterol, triglycerides, LDL, and Apo-B levels show no significant correlation (Table 2). These results suggest that longer telomeres are positively correlated with the lipid profiles of healthy aging.

Variation in the Telomerase Genes, Longevity, and Telomere Length.

We performed a comprehensive sequence analysis of hTERT and hTERC genes to detect all possible genetic variants, including the rare variants that may be enriched in centenarians, throughout the coding exons and exon–intron flanking regions in centenarians (n = 100) and controls (n = 80) using 2D gene scanning and DNA sequencing (Fig. 3). We found a total of 15 sequence variants in the hTERT gene, among which 5 were previously unknown unique variants that were not reported in various SNP databases (Table 3). The locations of these variants are shown in Fig. 4. Nine variants in the hTERT gene are in the coding region, including two nonsynonymous variants, 893 G > A (Ala-279 Thr) and 3242 G > A (Ala-1062 Thr), which are rare and found only in controls (Table 3). In silico analysis using SIFT (sorts intolerant from tolerant substitutions) (30) and PolyPhen (31), which predict the effects of amino acid changes on protein function, indicates that these changes are likely to be benign. Interestingly, rare synonymous or intronic variants in hTERT are enriched in centenarians (n = 19) compared with controls (n = 3) (Table 2; P = 0.041). In addition, three previously undescribed intronic variants in hTERC were also found, two of which were rare and found only in centenarians (Table 3).

We genotyped the four common hTERT variants, with a minor allele frequency (MAF) of >5% [IVS1–187 T > C, 973 G > A (Ala-305 Ala), 3097 C > T (His-1013 His), IVS16+99 C > T], in 73 centenarians and 49 controls. Information on telomere length was available for all these individuals. We did not find an association of the IVS1–187 T > C, 973 G > A (Ala-305 Ala), and IVS16+99 C > T with longevity. However, significant associations were detected with the 3097 C > T (His-1013 His) variant: the T allele is significantly enriched in centenarians (21.8%) compared with controls (8.2%; P = 0.026) (Table 3), and both CT and TT genotypes are significantly enriched in centenarians compared with controls (26.8% and 8.5% vs. 12.2% and 2%, respectively; P = 0.035) (Table 4). In contrast, no significant association between the hTERC common variant (IVS+63 T > C) and longevity was found (Table 4). Haplotype analysis [see SI Fig. S1 for a linkage disequilibrium (LD) plot derived from the four common variants] revealed that of the four most common haplotypes, Hap 1 is significantly depleted in centenarians compared with controls (34.9% vs. 53.1%; P = 0.0056), whereas Hap 3 is significantly enriched in centenarians compared with controls (13.7% vs. 8.2%; P = 0.007) when adjusted for age and gender (Table 5). Interestingly, a rare haplotype in controls, Hap 6, is found significantly more frequently in centenarians (1% vs. 6.8%; P = 0.024) (Table 5). Association analysis of common hTERT or hTERC variants with telomere length indicates no significant association (Table S1). However, of the two haplotypes that are enriched in centenarians compared with controls, Hap 3 shows a positive association with telomere length after adjusting for age and gender (P = 0.007), whereas Hap 6 does not reach a significant threshold (P = 0.59) because of its low frequencies in the tested population (Table 6). In contrast, a centenarian-depleted haplotype, Hap 1, shows no association with telomere length (Table 5).

Study Design and Subjects.

A cohort of Ashkenazi Jews with exceptional longevity was recruited as previously described (34). Probands were required to be living independently at 95 years of age as a reflection of good health, although at the time of recruitment, they could be at any level of dependency. In addition, probands had to have a first-degree offspring who was willing to participate in the study. Birth certificates or dates of birth as stated on passports were used to verify the participants’ ages. The Ashkenazi population is descended from a founder population (estimated to be several thousands) originating in the 15th century. This “founder effect” resulted in a population both culturally and genetically homogeneous, from which several disease-related genes have been successfully identified (45). A majority of these individuals were born in the United States or moved there before World War II. Telomere measurements (see below) were made on 86 Ashkenazi probands (aged 97.4 ± 0.3 years) defined as having exceptional longevity, 175 offspring of parents with exceptional longevity (aged 67.8 ± 0.6 years), and 93 controls who were offspring of parents with usual survival (aged 71.8 ± 1 years). By definition, parents of controls survived to the age of 85 years or less. The control group consisted largely of spouses of the offspring group. Informed written consent was obtained in accordance with the policy of the Committee on Clinical Investigations of the Albert Einstein College of Medicine.

For each subject, a detailed medical history, physical examination, and blood sample collection were performed as previously described (23, 32, 34). The presence of hypertension, the metabolic syndrome, and diabetes was determined using National Cholesterol Education Program (NCEP), Adult Treatment Panel III (ATP III) criteria (46). Scores on the MMSE (47) ≤25 were considered to indicate impaired cognitive function. Triglyceride, LDL-cholesterol (LDL), and HDL-cholesterol (HDL) levels and average particle sizes were determined by NMR spectroscopy at LipoScience, Inc., as previously described (34). Large lipoprotein particle sizes were defined as >8.9 nm for HDL and >21.3 nm for LDL.

Measuring/Estimating Length by Quantitative Real-Time PCR.

Genomic DNA was extracted from blood samples and stored at −80°C. The average (of triplicate) telomere length in leukocytes was estimated using a quantitative PCR (qPCR) method (48). This approach provides relative average telomere lengths in genomic DNA by determining the T/S ratio in experimental samples relative to a reference sample. Briefly, standard curves were generated for telomere lengths and for the single-gene copy amplification reactions from a reference DNA sample that was serially diluted with double-distilled water by 1.68-fold per dilution to produce 5 concentrations of DNA ranging from 7.7 to 61 ng/μL. For analysis of subjects’ DNA, triplicate PCR reactions using 2 μL of each DNA dilution were carried out in a 20-μL volume using the LightCycler FastStart DNA Master SYBR Green kit (Roche Diagnostics), with MgCl₂ added to a final concentration of 3 mM. Primers for telomeres and the single-copy gene HGB (β-globin gene) were added to final concentrations of 0.1 μM per 0.9 μM and 0.3 μM per 0.7 μM, respectively. The primer sequences are TEL 1b, 5′-CGGTTTGTTTGGGTTTGGGTTTGGGTT-TGGGTTTGGGTT-3′; TEL 2b, 5′-GGCTTGCCTTACCCTTACCCTTACCCTTACCCTT-ACCCT-3′; HGB1, 5′-GCTTCTGACACAACTGTGTTCACTAGC-3′; and HGB2, 5′-CACCAACTTCATCCACGTTC-ACC-3′ (48). The enzyme was activated at 95°C for 10 min, followed by 30 cycles at 95°C for 5 s, 58°C for 10 s, and 72°C for 40 s for the HGB reaction or for 20 cycles at 95°C for 5 s, 56°C for 10 s, and 72°C for 60 s for the telomere reaction. All transition rates were set to 20°C/s with the exception of the annealing transition rate in the telomere reaction, which was set to 4°C/s. Forty samples (mixed samples from probands, offspring, and controls) were analyzed by qPCR, both in our laboratory and by one of the authors (R.M.C.) for validation purposes, yielding a correlation of 0.89 (P < 0.001). In addition, to ensure reproducibility, when measuring previously unanalyzed samples, and to account for assay variability between batches, we used an anchor sample that was previously measured. The reproducibility between the two measurements was high (r = 0.94, P < 0.001). Results obtained using this qPCR were highly correlated with those obtained by the traditional terminal restriction fragment length approaches (48). One T/S ratio unit is equivalent to a mean telomere length of 4,270 bp in the leukocytes. The qPCR telomere length determination does not include subtelomeric lengths.

Mutation Screening and Genotyping of hTERT and hTERC.

We sequenced genomic DNA isolated from blood of centenarians and controls to discover all possible genetic variations in the exonic regions and intron–exon boundaries of the telomerase genes (hTERT and hTERC) by a 2D gene scanning (TDGS) method (49). To increase specificity, PCR amplification was designed in two steps. To begin with, triple-multiplex long-distance PCR coamplifies eight large fragments encompassing the coding and splicing junction regions of hTERT and hTERC. These products serve as a template for quadruple-multiplex short PCRs to amplify target regions in 18 fragments; 16 fragments for hTERT and two fragments for hTERC. Primers are listed in Table S2. The mixture of amplicons after heteroduplexing was separated on the basis of size and melting temperature in a 2D gel. Then, DNA fragments containing heterozygous sequence variation were reamplified from the genomic DNA and sequenced to identify the position and nature of sequence variation. For hTERT, exon 1 was analyzed by nucleotide sequencing and exons 2, 4, and 5 were separately analyzed by denaturing gradient gel electrophoresis (DGGE) because of ambiguous melting patterns on a TDGS gel. Genotypes of the common variants in the telomerase gene were determined by DGGE using the same primers designed for the TDGS assay.

Statistical Analyses.

Two comparisons have been established: (i) between probands and controls for whom crude data were adjusted for gender and (ii) between offspring and controls for whom crude data were adjusted for age and gender. Individual phenotypes that were not normally distributed were log natural-transformed for analysis and back-transformed for the current presentation. Associations of telomere length with other variables were estimated using regression on an indicator variable for group, age, and gender. In defining strata for age, cutoff points of 65, 75, 85, 95, and 105 years were used (Fig. 1). Telomere length distribution in probands appears to be nonnormally distributed; thus, to test for statistical significance, we applied nonparametric statistics analysis, and pairwise crude comparisons of telomeres between the study groups were carried out using the Wilcoxon signed-rank test. Parameters were estimated by maximum likelihood methods. The significance of the parameter (association) was evaluated using the likelihood ratio test by comparing the model in which the parameter is set to 0 with the model in which the parameter is estimated. In addition, we compared telomere length between offspring and controls using a hierarchical linear regression model to account for clustering within families. The regression equation adjusted for age and gender included an indicator variable distinguishing offspring from controls and an offspring vs. controls × age interaction term. We tested the hypotheses that offspring have higher mean telomere length than controls and that the age-related decline in telomere length is less in offspring than in controls. To determine the effect of lipid profile on telomere length, a general linear model adjusted for group, gender, and age at recruitment was used. Correlations of telomere length and lipid profile adjusted for group, gender, and age at recruitment were evaluated using Spearman rank correlation coefficients. Heritability of telomere length was assessed in two ways: First, it was computed in the ASSOC module of SAGE (v6.0.1) (http://darwin.cwru.edu/sage/).Second, narrow sense heritability was estimated from the slope of the linear regression of the traits of each parent on the mean value of the offspring group (16). Data are expressed as mean (SE) as appropriate. Statistical analyses were performed using SAS 9.1 (SAS Institute).

Comparison of the observed numbers of each genotype with those expected under Hardy-Weinberg equilibrium was tested using χ² tests for each of the three groups: centenarians, their offspring, and controls. Haplotypes were constructed using the PHASE algorithm (50). The haplotype trend regression analysis for logistical (case-control), linear (telomere length), and proportional hazard with telomere length was performed using JMP genomics 4 (SAS Inc., Cary, NC).