• Open Access

Paternal age is positively linked to telomere length of children


Departments of Biochemistry & Molecular Biology and Oncology, Faculty of Medicine, University of Calgary, 3330 Hospital Dr. NW, Calgary, Alberta, Canada T2N 4N1. Tel.: +1 403 220 8695; fax: +1 403 270 0834; e-mail: karl@ucalgary.ca


Telomere length is linked to age-associated diseases, with shorter telomeres in blood associated with an increased probability of mortality from infection or heart disease. Little is known about how human telomere length is regulated despite convincing data from twins that telomere length is largely heritable, uniform in various tissues during development until birth and variable between individuals. As sperm cells show increasing telomere length with age, we investigated whether age of fathers at conception correlated with telomere length of their offspring. Telomere length in blood from 125 random subjects was shown to be positively associated with paternal age (+22 bp yr−1, 95% confidence interval 5.2–38.3, P = 0.010), and paternal age was calculated to affect telomere length by up to 20% of average telomere length per generation. Males lose telomeric sequence faster than females (31 bp yr−1, 17.6–43.8, P < 0.0001 vs. 14 bp yr−1, 3.5–24.8, P < 0.01) and the rate of telomere loss slows throughout the human lifespan. These data indicate that paternal age plays a role in the vertical transmission of telomere length and may contribute significantly to the variability of telomere length seen in the human population, particularly if effects are cumulative through generations.


Telomeric T2AG3 DNA repeats are progressively lost in most human cell types that replicate (Harley et al., 1990; Hastie et al., 1990; Lindsey et al., 1991; Chang & Harley, 1995) owing to the end replication problem of linear chromosomes (Olovnikov, 1971). Telomeric sequence is maintained by the activity of telomerase, a reverse transcriptase found in immortal cancer cells (Kim et al., 1994), embryonic tissues (Wright et al., 1996; Achi et al., 2000), lymphocytes (Broccoli et al., 1995), skin and intestinal stem cells (Harle-Bachor & Boukamp, 1996) and germ cells (Wright et al., 1996), or through a recombination-based alternative pathway (ALT) of telomere maintenance found in some cancer cells (Bryan et al., 1997). However, in the rapidly dividing stem cells of the skin, intestine and blood, telomerase enzyme activity is insufficient to maintain telomere length (Bodnar et al., 1996; Wright et al., 1996). When telomeres in normal cells erode, division ceases (Baird et al., 2003), but this can be reversed experimentally by expression of the telomerase enzyme that extends telomeres (Bodnar et al., 1998; Vaziri & Benchimol, 1998). Because replicating cells represent a considerable proportion of organ systems, limited replicative potential may contribute to age-related diseases, including infections where aging immune cells divide less efficiently in response to chronic damage or viral infection (Bestilny et al., 2000; Effros et al., 2003). A recent retrospective analysis of banked blood DNA samples (Cawthon et al., 2003) uncovered a clear correlation between short telomere length and increased probabilities of mortality from age-associated heart disease (> 3-fold) and infectious diseases (> 8-fold). Because the average lifespan of most human populations is steadily increasing and telomere length is linked to age-associated health issues and to lifespan (von Zglinicki et al., 2000), the mechanism by which telomere length varies is of biological, medical and social importance. Although the frequency of infection influences blood cell telomere length later in life, it is unclear why telomere length differs among individuals in utero or in the newborn (Youngren et al., 1998). In this study we investigated how telomere length varied between males and females, how telomere attrition varies with age, and what mechanism(s) might account for telomere length variability in the human population.

Results and discussion

For our cross-sectional study, between 2001 and 2002 we recruited 125 subjects (51 male and 74 female) from 30 to 80 years of age by random digit telephone dialing and stratified them in 10-year age groups. Subjects completed detailed in-person interviews and provided blood samples that were processed and coded without personal identifying information. DNA was purified from coded samples and mean terminal restriction fragment (TRF) analyses were performed in triplicate on each sample in independent experiments using a blind experimental protocol as described (Bestilny et al., 2000). Coded control samples were included to verify reproducibility. The association between the subject's telomere length and their age and gender, as well as the parental ages at subject conception was assessed by linear and multiple regression analyses. A typical TRF gel is shown in Fig. 1 and illustrates that telomere length in samples from older subjects is generally shorter on average than samples from younger subjects. Analysis of 30 such gels by scanning densitometry revealed a normal distribution, with telomeres ranging from 5780 to 9640 bp (mean length of 7020 bp). Linear regression revealed a mean telomere length decrease of 22 bp per year [95% confidence interval (CI) 13–29 bp yr−1, r = −0.42, P < 0.0001) (Fig. 2A). We also noted that telomere erosion declined with age (30–49 years = 51.3 bp yr−1, 50–80 years = 19.8 bp yr−1). Previous cross-sectional studies have shown a rapid loss of telomeric DNA during development in utero and in the first few years of life followed by a more gradual decline later (Frenck et al., 1998; Rufer et al., 1999). When our data are compared with that from many other studies (Table 1), it becomes evident that a variable telomere loss might also occur later in life, with the loss being greatest early in life and continually slowing with age (Fig. 2B). This may reflect decreasing cell turnover in the circulating component of the immune system with age and provides a basis for the different rates of telomere attrition reported previously by other groups.

Figure 1.

Telomere length assays. Autoradiogram showing terminal restriction fragment (TRF) length distributions from subjects of the indicated ages. Subject age was randomly distributed on gels because analyses were performed blind in all cases. M shows the migration of the molecular weight standards (HindIII-cut λ DNA) of the indicated sizes.

Figure 2.

(A) Telomere length of subjects at different ages. Mean terminal restriction fragment (TRF) length vs. age for 125 randomly recruited healthy individuals. The linear regression line is represented by the solid blue line, and slope, confidence intervals, r and P values are indicated on the graph. All data in this study were generated and initially analysed in a fully blind experimental protocol. (B) Rate of telomere loss slows throughout the human lifespan. The line shown was generated from values in this study. The circles show the various rates of telomere loss from blood in a variety of studies listed in Table 1, all of which have P values < 0.01. (C) Relationship between TRF length vs. age in males and females; males lose white blood cell telomere sequences more rapidly than females. The blue (male, open circles) and pink (female, closed squares) linear regression lines show that the difference in the rate of telomere sequence loss between males (31 bp yr−1, P < 0.0001) and females (14 bp yr−1, P < 0.01) is 17 bp yr−1. (D) Children with older fathers have longer telomeres. Telomere length of offspring vs. age of father is plotted (not age-adjusted). After age-adjustment with multiple regression, an increase of 22 bp is seen in the length of children's telomeres per year of increasing paternal age (P = 0.01). For males only, telomere increase was 29 bp per year of increasing paternal age (P = 0.06) whereas for females the increase was 12 bp per year of increasing age (P = 0.21). For mothers, an increase of 17 bp per year of increasing maternal age at birth of their children was noted but the difference was not significant (P = 0.08). All raw data used to generate panels A–D are available upon request from the authors.

Table 1.  Rates of blood telomere loss. These published data, in which a total of 1094 individuals were analysed, were plotted on the graph in Fig. 2(B). Each study was performed independently, has different statistical power and used varied experimental methodology, but in all cases were statistically significant (P < 0.01)
StudyRate of loss (bp yr−1)Age range (years)Median age (∼ years)n
Hastie et al. (1990)  3320–8552.547
Vaziri et al. (1993)  41 0–10748.3119
Slagboom et al. (1994)  31 2–9529.6123
Mondello et al. (1999)  42.626–7244 26
  15.650–10479.5 26
Rufer et al. (1999)1088 0–1.5 0.75 38
  52 1.5–9037.7246
von Zglinicki et al. (2000)  2018–9863.9186
Bestilny et al. (2000)  38.3 0–8245 15
Cawthon et al. (2003)  1460–9778.5143
This study  51.330–494236

Studies of fetal and newborn subjects have indicated that telomere length is similar in different tissues and in males and females during development (Youngren et al., 1998). Although we observed no mean telomere length difference between males and females in the age range examined (7020 and 7090 bp, respectively), males lost telomeric DNA at approximately twice the rate of females (31 bp yr−1, 95% CI = 17–43 bp yr−1, r = −0.56, P < 0.0001 and 14 bp yr−1, 95% CI = 3–24 bp yr−1, r = −0.30, P < 0.01, respectively). This trend shown in Fig. 2(C) agrees with a recent report (Nawrot et al., 2004) and might contribute to the consistently longer average lifespan of females seen in most societies. This observation may be due to higher levels of circulating estrogens in females because the promoter of the catalytic subunit of telomerase, hTERT, is responsive to 17 β-estradiol, resulting in increased telomerase activity (Kyo et al., 1999).

Telomerase activity prevents telomere erosion in germ line cells. During spermatogenesis, diploid stem cells are differentiated into haploid spermatozoa. Telomere length increases following spermatogonial proliferation to spermatocytes, then remains stable until maturation into spermatozoa, at which time telomeres again lengthen (Achi et al., 2000). The increase in telomere length in sperm of ∼70 bp yr−1 (Allsopp et al., 1992) is much different than what is seen in cells where telomeres are stable (post-mitotic) or are lost with increasing replicative age (mitotic). To test if increased sperm telomere length was transmissible, we compared the age-adjusted mean telomere length of our subjects to the ages of their parents at conception. Paternal age was positively and significantly associated with telomere length of both male and female offspring (Fig. 2D– unadjusted for subject age). On average, offspring had 22 more base pairs for each year older their father was at conception (95% CI 5.2–38.3 bp, r = 0.46, P = 0.01, n = 113 – after adjustment for subject age). In contrast, maternal age did not significantly influence the telomere length of either male or female offspring (r = 0.57, P = 0.30 and r = 0.26, P = 0.54, respectively – with or without subject age adjustment), although a trend towards children with older mothers having longer telomeres was noted.

These data support a mechanism in which telomere length in ova and variable telomere length in sperm, based upon initial telomere length and age of the father, co-determine the telomere length of offspring. Determination of progeny telomere length by the lengths of telomeres in both parents is also supported by observations in mice in which the F1 generation have telomeres intermediate in length between those of their parents (Manning et al., 2002). Previous studies have also indicated that telomere length is largely heritable (Rufer et al., 1999; Slagboom et al., 1994), and linked to the X-chromosome (Nawrot et al., 2004). One such study (Nawrot et al., 2004) found a correlation between telomere length in mothers and offspring and between fathers and daughters. This association might be related to a group of genes residing on the X chromosome (such as DKC1 and AGTR2), which have been found to influence telomerase through stable hTR accumulation and nitric oxide production, respectively. Together, these observations suggest a mechanism whereby human telomere length is affected by paternal age through vertical transmission and that X-linked genetic factors may serve to stabilize changes to the degree that telomere length appears to be up to 78% heritable (Slagboom et al., 1994). If telomere length is 78% heritable as calculated in previous studies of twins (Slagboom et al., 1994), then other factors would contribute ∼22% to determining telomere length in any one generation. A conservative estimate of the contribution of paternal age on variation of telomere length can be obtained from the square of the correlation coefficient generated from the multiple regression of mean TRF (mTRF) vs. paternal age, which is approximately 21%, consistent with estimates of heritability (Slagboom et al., 1994). Examined from another perspective, over 30 years, an increase of 70 bp yr−1 in sperm could theoretically generate a difference of 2100 bp in sperm from young vs. older fathers. With equal telomere contributions from mother and father, the paternal contribution could account for a maximum of ∼35 bp yr−1 or 1080 bp difference in 30 years. Our data approximate this theoretical value with offspring showing a 660-bp difference over the same time (∼22 bp yr−1 of father's age).

Could the variation in telomere length associated with paternal age in this study account for the much wider variation seen in telomere length at birth? Previous studies have measured a range of 10 200–12 400 bp in fetal and umbilical cord blood (Vaziri et al., 1994; Chang & Harley, 1995; Youngren et al., 1998; Bestilny et al., 2000). If a maximum variation of 660 bp is generated by different paternal age per generation as our data indicate, the variability in telomere length seen at birth could be generated in 3–4 consecutive generations by fathers 30 years different in age, if cumulative effects of paternal age on telomere length across generations exist. This estimate also assumes that inherited genetic factors maintain telomere length and do not serve to further alter telomere length after early development as suggested by studies of fetal development (Youngren et al., 1998).

Based upon these observations, paternal age is clearly linked to telomere length and could contribute to the variability seen at birth and thus later throughout the lifespan. However, to eliminate the possibility of unforeseen birth cohort effects, telomere length variation through multiple generations will need to be examined. Despite evidence that telomerase activity is present in most female reproductive tissues, there has been no study examining telomere length in oocytes of females throughout their reproductive lifespan. Such a study would add greatly to our understanding of telomere biology and germ cell telomere length maintenance. Additionally, determination of telomere lengths on Y chromosomes from individuals with fathers of different ages could further examine the mechanism of telomere length transmission. The mounting evidence that longer telomeres confer resistance to mortality due to age-associated diseases suggests that the tendency towards having children later in life, which correlates with and could result in telomere lengthening, may have beneficial medical effects particularly in the elderly.

Experimental procedures

Subject enrollment

Subjects were selected through random-digit telephone dialing. We were able to screen 74% of households for eligibility and among the 281 eligible subjects identified, 150 (53.4%) agreed to participate. All subjects completed an in-person interview that ascertained information about factors that may be related to peripheral blood mononuclear cell (PBMC) telomere length, such as physical activity (occupational, household, leisure), a brief medical history and tobacco use. Usable blood samples (10 mL Vacutainer tube containing EDTA) were obtained from 125 subjects from July 2001 to April 2002. Research was performed following approval by the Conjoint Health Research Ethics Board of the University of Calgary, Calgary, Alberta, Canada, and conforms to the Tri-Council and ICH Guidelines and with the Helsinki Declaration.

TRF assays and statistical analyses

PBMCs were isolated from whole blood using Ficoll-hypaque gradients. Cells were lysed and DNA was extracted using a phenol/chloroform extraction method. Five micrograms of DNA was digested with restriction endonucleases (Hinf1 and RsaI) for which there are no recognition sites within the telomeric DNA repeats. Visual quantification of an aliquot of digested DNA on a 1% agarose gel ensured equal loading in the final agarose gel. The digested DNA samples were separated on 0.6% agarose gels for 750 V h. Gels were denatured and neutralized prior to in-gel hybridization with [32P]-labeled synthetic oligodeoxynucleotide telomere probes of the sequence (CCCTAA)3. Following hybridization and subsequent washes, gels were exposed to X-ray film. Autoradiograms were photographed and analysed using Image J freeware (http://rsb.info.nih.gov/ij). The weighted centers of mass of density plot profiles were generated for each sample. Numeric values generated from the histograms were compared against values of a known molecular weight standard, results of which reveal the mTRF length. To determine the degree of interassay variation, control blood samples were randomly inserted throughout the collection of subject blood samples over the 10-month time period. The protocol was performed 2–5 times per sample and the raw TRF data from the multiple runs were compiled to generate a final mTRF value for each individual. This established that variability ranged minimally, and in a random manner. Relationships between telomere length, age, gender and parental age were assessed by single and multiple linear regression analysis. We also regressed subject age and paternal age on telomere length using a random effects model to account for both variation in telomere measurement per subject and the variable number of telomere measurements made per subject. As the random effects model provided nearly identical results to the linear regression/multiple regression models using mean telomere lengths per subject (because the between-subject variation was far greater than the within-subject variation), we used the more familiar regression models in the final analyses. Linear regression analysis was performed for mTRF vs. subject age (Fig. 2A) and mTRF vs. subject age by sex (Fig. 2C), and multiple linear regression was performed for mTRF vs. paternal age (subject age adjusted with age as a continuous variable).


We thank P. Round and L. Kmet for help with data collection, coding and analyses, and R. N Johnston and C. Sensen for critically reading the manuscript. L.S.C. is an Alberta Heritage Foundation for Medical Research (AHFMR) Health Scholar and Canadian Institutes of Health Research (CIHR) New Investigator. K.T.R. is an AHFMR and CIHR Scientist. This work was supported by CIHR grants to L.S.C. and K.T.R. The authors have no conflicting financial interests.