ABSTRACT
Leukocyte telomere length (LTL) is a heritable trait with two potential sources of heritability (h2): inherited variation in non-telomeric regions (e.g., SNPs that influence telomere maintenance) and variability in the lengths of telomeres in gametes that produce offspring zygotes (i.e., “direct” inheritance). Prior studies of LTL h2 have not attempted to disentangle these two sources. Here, we use a novel approach for detecting the direct inheritance of telomeres by studying the association between identity-by-descent (IBD) sharing at chromosome ends and phenotypic similarity in LTL. We measured genome-wide SNPs and LTL for a sample of 5,069 Bangladeshi adults with substantial relatedness. For each of the 7,254 relative pairs identified, we used SNPs near the telomeres to estimate the number of chromosome ends shared IBD, a proxy for the number of telomeres shared IBD (Tshared). We then estimated the association between Tshared and the squared pairwise difference in LTL ((ΔLTL)2) within various classes of relatives (siblings, avuncular, cousins, and distant), adjusting for overall genetic relatedness (ϕ). The association between Tshared and (ΔLTL)2 was inverse among all relative pair types. In a meta-analysis including all relative pairs (ϕ >0.05), the association between Tshared and (ΔLTL)2 (P=0.002) was stronger than the association between ϕ and (ΔLTL)2 (P=0.45). Our results provide strong evidence that telomere length (TL) in parental germ cells impacts TL in offspring cells and contributes to LTL h2 despite telomere “reprogramming” during embryonic development. Applying our method to larger studies will enable robust estimation of LTL h2 attributable to direction transmission.
INTRODUCTION
Telomeres are DNA-protein complexes at the end of mammalian chromosomes that maintain genome stability by preventing recombination, end-to-end fusion, and DNA damage at chromosome ends 1. In somatic cells, shortening of the DNA component of the telomere occurs with each round of cell division due to the “end-replication problem”. In stem cells, extension of telomeres by telomerase counters this shortening, but gradual shortening still occurs with age, with critically short telomeres contributing to cellular senescence. Consequently, leukocyte telomere length (LTL) is considered an indicator of biological aging and a potential biomarker of risk for age-related diseases 2; 3. Short LTL is associated with increased risk for several age-related diseases in observational studies, including cardiovascular disease, hypertension, liver disorders, diabetes, atherosclerosis, and overall mortality 2; 4-7. In contrast, Mendelian randomization studies suggest that longer telomeres increase risk for several types of cancer, including lung adenocarcinoma 8, melanoma 9, glioma 10, neuroblastoma 11, and chronic lymphocytic leukemia 12.
LTL is a highly heritable trait 13-16 and is known to be effected by inherited variation in non-telomeric regions, e.g., SNPs (single nucleotide polymorphisms) that influence telomere maintenance 17. However, LTL has a unique additional potential source of heritability (h2): variability in the lengths of telomeres themselves in the parental gametes that produce the offspring zygotes 18, under the assumption that gamete telomere length (TL) partially determines TL in adult tissues (including leukocytes and germ cells). This second source of h2 is described as “direct” transmission of telomeres (Figure 1) and is hypothesized to be the primary source of TL h2,18. Telomeres undergo “reprogramming” in the embryo, initially by recombination-based mechanisms and then by telomerase at the blastocyst stage and later 19, but is unclear to what extent reprogramming alters the impact of parental germ cell TL on TL in offspring cells.
Evidence supporting an impact of parental TL on offspring TL includes the observed associations of advanced paternal age at conception and increased offspring LTL 16; 20-26. This association is often attributed to age-related telomere lengthening in spermatogonia, which express high levels of telomerase 20, providing evidence that acquired variation in TL in the male germ line contributes to offspring TL. Also supporting this notion are studies of families with genetic disorders affecting telomere maintenance, in which affected parents who carry a mutation that renders low telomerase activity and short telomeres have non-carrier children whose telomeres appear shorter as compared to the general population, despite not inheriting the mutation 27-29. These findings are also supported by data from telomerase deficient mice 29.
In light of the evidence supporting the contribution of “direct” transmission of telomeres to LTL h2, in this work, we attempt to detect the effect of direct transmission on LTL in a population-based setting. Our approach is to estimate the association between identity-by-descent (IBD) sharing at chromosome ends (a proxy for sharing telomeres IBD) and phenotypic similarity in LTL between relative pairs. Prior family-based and SNP-based h2 studies of TL have not attempted to disentangle the effects of “direct” inheritance of TL from other sources of h2, and it remains unclear whether the h2 of TL is primarily driven by inherited genetic variation in non-telomeric regions (e.g., SNPs that impact telomere maintenance) or whether TL is substantially influenced by the length of the telomeres present in the parental germ cells 18. The latter hypothesis implies that TL is not completely “reset” during embryogenesis and early development.
EXPERIMENTAL METHODS
Study participants
DNA samples were obtained from participants in the following Bangladeshi studies: The Health Effects of Arsenic Longitudinal Study (HEALS), an expansion of the HEALS cohort (ACE), and The Bangladesh Vitamin E and Selenium Trial (BEST). HEALS is a prospective study of the health outcomes associated with arsenic exposure through drinking local well water in 11,746 adults (18-75 years of age). An expansion of the HEALS cohort (ACE) occurred between 2006 and 2008, and an additional 8,287 participants were recruited. Details of HEALS have been described previously 30. BEST is a chemoprevention trial that assessed the effects of vitamin E and selenium supplementation on non-melanoma skin cancer risk. Details of the BEST study have been described previously 31. The HEALS and BEST study participants included in this work are primarily from the Araihazar area, a rural community east of Dhaka. Multiple members of extended families were often recruited, resulting in a substantial number of relative pairs in these cohorts. These studies have been approved by the Ethical Review Committee of International Center for Diarrheal Disease Research, Bangladesh, The Bangladesh medical Research Council, and the Institutional Review Boards of the University of Chicago and Columbia University.
DNA extraction and genotyping
Details of sample collection, DNA extraction and genotyping have been described by Pierce, et al. 32; 33. In brief, genomic DNA for HEALS (including ACE) and BEST samples were extracted from clotted blood using the Flexigene DNA Kit (Cat # 51204) and whole blood using the QIAamp 96 DNA Blood Kit (Cat # 51161), respectively. DNA samples were processed on HumanCytoSNP-12 v2.1 chips. Prior to quality control, our genotype data consisted of 5,499 individuals with 299,140 SNPs measured. We removed individuals with call rates < 97% (n = 13), gender mismatches (n =79), and technical replicate samples or duplicates (n = 53). We excluded SNPs that had low call rates, i.e. (< 95%) (n = 20), were monomorphic (n = 39,798), did not have rs identifiers (n = 941), and had HWE p-values < 10−10 (n = 634). The HWE threshold used is not stringent because our sample includes relatives, leading to inflated HWE test statistics. In an unrelated set of participants, HWE testing confirmed no SNPs used in the analysis had an HWE P value < 10−7.
Measurement of LTL using qPCR
For 2,203 genotyped HEALS samples, LTL was measured using quantitative(q)-PCR using two different plate designs (Design 1 and 2) on 96-well plates. The relative telomere length is represented as a ratio of the telomere repeat copy number to single-gene (RPLP0) copy number, i.e. T/S ratio. Details of this methodology have been described previously 34. Plate design 1was based on Ehrlenbach et al. 35 and consisted of triplicates of 14 participant DNA samples per plate and six replicates of reference DNA sample, with telomere sequence and RPLP0 measured on the same plate. Plate design 2 was based on Cawthon et al. 36, and we used paired plates for amplification of telomere and single-copy gene of each DNA sample, each consisting of triplicates of 31 participant DNA samples and triplicates of reference DNA sample.
The coefficients of variation (CV) were calculated as the standard deviation of the sample replicates divided by the mean of the replicates. To assess the reproducibility of the measured T/S ratios for Plate Design 1 and 2, we re-ran 37 and 31 samples on separate days and calculated the inter-plate CVs to be 11.7% and 9.8%, respectively.
Measurement of LTL using QuantiGene Chemistry on a Luminex Platform
The Affymetric-Panomics QuantiGene Plex is a probe-based non-PCR assay developed in our lab to quantify LTL in 1,825 BEST and 1,047 ACE participants. The details of this methodology have been described previously 37. The Luminex method has been validated in a blinded comparison to Southern Blot 38, the gold standard for TL measurement in DNA samples. Like qPCR, the Luminex-method method produces a relative telomere length measure represented by a ratio of telomere probe to reference gene probe, known as the Telomere Quantity Index (TQI).
LTL Standardization
To remove experimental variation in our measures of LTL, we utilized linear mixed-effects models (lme package in R), where position and plate were treated as fixed and random effects, respectively. We performed three separate mixed effects models; one for the Luminex-based LTL measures and one for each qPCR plate design. To harmonize the LTL measures from the two experimental methods, we standardized the residuals from mixed-effects regressions and adjusted for age, sex, and cohort in a separate regression model. These standardized residuals were used in all downstream analyses. Additionally, we generated two variables that represent the difference in LTL between relative pairs: the squared difference (ΔLTL)2 was our primary outcome (due to its utility for h2 estimation (see below)), and the absolute difference |ΔLTL| was our secondary outcome (presented as supplementary results).
Estimating telomeres shared
We performed LD pruning on 257,747 genotyped SNPs to produce a subset of uncorrelated markers (i.e. linkage equilibrium). In PLINK, we used the “indep-pairwise” command, where we specified three parameters: a 100kb window size, a 25kb count to shift the window at the end of each step, and a pairwise squared correlation (r2) threshold of 0.1 to eliminate SNP pairs with an r2 > 0.1 in each window, resulting in 52,066 uncorrelated SNPs. Using this set of uncorrelated SNPs, we restricted to the 180 SNPs closest to the end of each chromosome to estimate identity-by-descent (IBD) probabilities for each chromosome end, for all possible pairs of individuals (~12.8 million pairs). We chose to use 180 SNPs because this number gave us very clear separation among pairs who shared exactly 2 chromosome ends, exactly 1 end, and <1 end for all chromosome arms (Figure S1). When using fewer than 180 SNPs (e.g., 150 SNPs), for some chromosomes, we did not see a clear gap in the IBD proportion between the pairs sharing exactly one end and the pairs sharing <1 end. The genetic and physical distances spanned by these SNPs at each chromosome end are described in Table S1 and the frequency of pairs sharing chromosome ends IBD are described in Table S2.
We used the IBD proportion for each chromosome end, obtained from PLINK, to define a discrete value for the number of chromosome ends shared IBD, which represents likely sharing of telomeres IBD. For each chromosome end, we defined 0 shared telomeres if the IBD proportion was <0.45, 1 shared telomere if the proportion was between 0.45 and 0.9, and 2 shared telomeres if the proportion was ≥0.9. We summed the number of shared telomeres across all autosomal chromosome p- and q-arms to obtain a total number of ends shared IBD for each participant pair, which we take as an estimate of the number of telomeres shared IBD (Tshared) (Figure 2). We excluded the p-arm of chromosomes 13-15, 21, and 22 from our Tshared count, because those arms had a >14 Mb gap between the most telomeric SNP and the chromosome end (as defined by the Genome Reference Consortium Human build 37 (GRCh37/hg19)). For all other chromosome ends, this gap was < 0.3 Mb, except 1p which had a gap of 0.75 Mb. Recombination events are extremely rare in intervals of this size, implying that sharing a segment IBD at a chromosome end (based on 180 SNPs) is a strong proxy for sharing a telomere IBD. Thus, using only 39 chromosome ends results in a theoretical maximum of Tshared = 78 (for twins). We divided Tshared by 78 to represent the estimated proportion of telomeres shared IBD.
We classified each participant pair according to their relationship type (i.e., parent-offspring, sibling, cousin, avuncular, etc.) using the estimated genetic relationship (ϕ) and the IBD probabilities (P(IBD)=0 and P(IBD=1)), all estimated using the PLINK --genome command applied to genome-wide SNPs (Figure 3). Paris with ϕ=0.25 could in theory be grandparent-grandchild pairs or half siblings, but based on what we know regarding the demography of the population groups we are studying and the age difference of these pairs, the vast majority of these pairs should be cousins, so we classified them as such. The ϕ threshold separating cousin pairs from “distant” pairs (ϕ=0.125) was somewhat arbitrary due to lack of clear separation between these groups (in terms ϕ) at lower ϕ values; thus, there is likely contamination of the “distant” group with cousin pairs. We conduct sensitivity analyses (see below) to ensure our results are not sensitive to arbitrary decisions regarding this threshold.
Statistical Analyses
For each relative pair type, we used linear regression to estimate the association between the Tshared proportion and the squared difference in LTL ((ΔLTL)2), adjusting for each pair’s ϕ value. This regression was performed for 583 sibling pairs, 605 avuncular pairs, 945 cousin pairs, and 5,121 distant relative pairs. These are quasi-independent pairs, meaning that a given individual can be included in multiple pairs, but pairs are treated as independent. While quasi-independent pairs are routinely used in studies using sib-based IBD methods for h2 estimation 39-41, we also conducted analyses restricting to pairs with no duplicate individuals. Parent-offspring pairs were not analyzed because we expect no variation in the Tshared proportion (or ϕ) among these pairs.
To estimate the association between Tshared proportion and (ΔLTL)2 using data on all types of relative pairs, we combined the association estimates for each relationship type in a fixed-effects meta-analysis using the ‘metafor’ package in R. Fixed-effects meta-analysis requires two key assumptions: 1) homogenous effect-size across all groups and 2) no between-group variance. Additionally, we conducted a linear regression to test whether mean (ΔLTL)2 is different across relationship types (including parent-offspring pairs).
Using the meta-analysis association estimate, we used the approach described in Visscher et al., 40 to estimate the narrow sense heritability (h2) attributed to the proportion of shared telomeres IBD (Tshared). The regression parameter (β) for the association between Tshared proportion and (ΔLTL)2 is equal to twice the additive genetic variance, and h2 is equal to β divided by the twice the total phenotypic variance, multiplied by -1: We used this equation to convert the meta-analysis regression parameter β and its upper and lower bound to the h2 scale.
In order to determine if LTL reflects TL in germ cells (as expected under the causal diagram in Figure 1), used tissue-specific TL data from the Genotype-Tissue Expression (GTEx) project 42 to examine the correlation between TL in whole blood and TL in testicular tissue. Tissue collection from post-mortem tissue donors and DNA extraction has been described previously 43. TL was measured for GTEx DNA samples using the Luminex technology described above.
RESULTS
Characteristics of the 5,069 Bangladeshi participants included in our regression analyses are described in Table 1. Age was inversely associated with LTL across each cohort (HEALS: r = −0.20, P < 10−16, BEST: r = −0.23, P < 10−16, and ACE: r = −0.28, P < 10−16), as expected. The types of relative pairs observed in our sample are described in Figure 3. For each relative pair type, LTL showed significant correlation between relative pairs, with stronger correlation generally observed for closer relative pair types (e.g., siblings r = 0.22, P = 3×10−8) compared to more distant relatives (i.e., cousins r=0.07, P = 0.04) (Table 2). Similarly, the mean of our primary outcome of interest, (ΔLTL)2, increased with decreasing level of relatedness (Figure S2). Univariate regression analyses of (ΔLTL)2 on relationship type showed a similar trend, with a larger effect of relationship type for increasing degree of relatedness (Table S3).
The total number of chromosome ends shared IBD by relative pairs at each chromosome end (described in Table S2) showed evidence of a negative correlation with the cM distance at each chromosome end encompassed by the 180 independent SNPs used to measure IBD sharing (Figure S3). This is expected, as chromosome ends with SNP intervals that span a large genetic distance will have more recombination events occurring in that interval than ends with SNPs spanning a shorter genetic distance, resulting in less frequent observation of IBD sharing for the entire segment. The proportion of chromosome ends shared IBD for a relative pair, a proxy for shared telomeres (Tshared), showed a strong positive association with the estimated pair-wise genetic relatedness (ϕ, estimated using genome-wide SNPs) for each type of relative pair (Figure 4). This is also expected, as pairs sharing more of the total genome would be expected to share more telomeres.
Among each relative pair type, our estimate of the proportion of telomeres shared IBD (the Tshared proportion) showed a non-significant inverse association with (ΔLTL)2 (Figure 5). In contrast, the associations between genome-wide IBD sharing (ϕ) and (ΔLTL)2 were generally weaker and did not show a consistent direction of association (Table 3). We meta-analyzed the four groups of relative pairs (7,254 total pairs) using a fixed-effects model (test of heterogeneity Q = 2.41, P = 0.49), and the Tshared proportion showed a significant inverse association with (ΔLTL)2 (β = −2.76, P = 0.002) (Figure 5), while ϕ did not (β = −1.16, P = 0.448) (Table 3). Using a different threshold to separate cousin pairs from distant pairs (ϕ ≤ 0.08) or combining them into a single group had minimal impact on this result (Supplementary Figures 4-5). Restricting to relative pairs for which there were no duplicate individuals resulted in directionally consistent results for Tshared (P=0.07) and ϕ (P=0.49) in the meta-analysis, although power was reduced due to smaller sample size (Supplementary Table 4).
Regression analyses using |ΔLTL| as the outcome rather than (ΔLTL)2 produced similar results (P=0.0002) (Table S5, Figure S6). Scatterplots of the Tshared proportion against (ΔLTL)2 for each relative pair type are shown in Supplementary Figure S7-S8. Together, these analyses indicate that sharing telomeres IBD contributes to pairwise similarity in LTL.
We conducted similar analyses for five continuous control phenotypes to show that the observed association with Tshared was specific to pairwise differences in LTL. These phenotypes were height, BMI, mean arterial pressure, systolic blood pressure, and diastolic blood pressure. Difference measures for each phenotype were taken after each phenotype was adjusted for age, sex, cohort, and batch. We did not observe a statistically significant inverse association between the Tshared proportion and the pairwise squared difference or the absolute difference for any of the control phenotypes (Table S6 and S7).
According to Visscher’s method for h2 estimation using genetic data on full siblings 40, the narrow-sense h2 for the fixed-effects meta-analysis β coefficient is 1.38 (95% CI [0.49, 2.26]). This estimate is above the theoretical maximum for h2, likely due to the very large confidence bounds on this estimate. The lower bound for our h2 estimate is 0.49, suggesting that at least 49% of the variation in LTL can be explained by sharing telomeres sequence IBD with close relatives. However, variation due to age and other covariates was removed from our TL measures prior to analysis, so the true lower bound is likely lower than 49%.
In order to provide evidence that TL in germ cells and in blood cells reflect TL in the embryonic cells from which both cell types descend (Figure 1), we used data on 150 tissue donors from the GTEx project to examine the within-person correlation between TL in whole blood and TL in testicular tissue, which primarily consists of spermatagonia, spermatocytes, or spermatids (~80% of cells). Whole blood TL was positively correlated with TL measured in testicular tissue (r = 0.19, P = 0.02) (Figure S9). This observation supports the hypothesized causal relationships in Figure 1 and the interpretation of our results as trans-generational, direct inheritance of TL.
DISCUSSION
In this study of LTL measured in a genotyped cohort with substantial relatedness, we show that the extent of sharing chromosome ends IBD, a proxy for sharing telomere sequence IBD, is associated with similarity in LTL (smaller value of (ΔLTL)2) measured in a DNA sample. Our results provide evidence that direct transmission of telomeres from parent to offspring accounts for a substantial proportion of LTL h2. This implies that variation in TL in parental germ cells impacts TL in offspring embryonic cells (and adult cells) despite telomere reprogramming occurring during embryonic development. The novel approach we describe and apply in this work represents the first attempt to estimate the h2 in LTL due to direct transmission of telomeres in a human population.
The association between telomere sharing (Tshared) and pairwise LTL difference was estimated among sibling, avuncular, cousin, and distant relative pairs. Each of these analyses showed a consistent inverse association, and meta-analysis of all relative pair types produced a statistically significant inverse association. This association was consistently larger than the association between overall genetic relatedness (genome-wide) and LTL difference, suggesting the contribution of telomere sharing to LTL similarity between close relatives is larger than the contribution of sharing inherited variation in non-telomeric regions (e.g., SNPs that influence telomere maintenance).
Genome-wide association (GWA) studies of TL have demonstrated that common genetic variants contribute impact LTL 34;44-52. However, the SNPs identified to date account for a <4% of the variation in LTL17. Heritability estimates of TL typically range between 34-50% in family studies and have been as high as 82% in twin studies 13-16. SNP-based heritability (i.e., variation in TL explained by common SNPs) has been estimated to be ~28% 53. These findings suggest additional sources of h2, such as the direct transmission of telomeres investigated in this work.
Several prior studies provide support for the hypothesis that direct transmission of telomeres impacts TL. First, there are multiple studies that report a link between paternal age and offspring LTL 23;24;54-56. In light of the evidence that telomeres in sperm lengthen as men age, these results suggest that acquired variation in paternal germ cell TL impacts offspring TL in somatic tissues. Additionally, Eisenberg and colleagues 23 found the paternal age effect is cumulative across at least two generations, with grandfather’s paternal age (at father’s birth) significantly predicting grandchildren’s TL independent of father’s age at reproduction 23. Stindl 57 offers a different explanation for the paternal age effect that does not involve age-related telomere lengthening in sperm. Stindl proposes that the association between paternal age and offspring TL is the result of confounding by birth cohort, claiming that TL in the female germline decreases with age and oocytes used first (at younger ages) have longer TL. Consequently, older men have longer sperm TL because they are part of a generation born to younger mothers 57. Both explanations for the paternal age effect support the hypothesis that parental transmission of telomeres contributes to TL h2.
The work of Aubert et al. 27, Collopy et al. 28 and Chiang et al. 29 also suggests an effect of direct transmission of parental telomeres in humans. These studies report that parents with mutations in TERT or TERC that render low telomerase activity and shorter telomeres pass their short telomeres to offspring who do not carry the telomerase mutation 27; 28. However, these studies were conducted in relatively small samples of parents with telomerase deficiency and their offspring. In addition, these findings were complicated by differences in TL among individuals and the admixing of affected or unaffected paternal and maternal telomeres. 29 Thus, these papers provide consistent evidence of the impact of direct transmission on TL, but the implications of these results beyond telomerase-deficient individuals and their families are unclear.
It is worth noting that Chiang et al. 29 also reached a similar conclusion based on a telomerase deficient (Tert+/-) mouse model. These mice had critically short and dysfunctional telomeres, whereas their genotypically normal Tert+/+ offspring had functional telomeres that were protected from progressive shortening but had similar lengths to Tert+/-parents.
To our knowledge, this is the first study to attempt to assess the relationship between IBD sharing of telomeres and similarity in LTL. We focused our analysis on specific types of close relative pairs (siblings, avuncular pairs, cousins, and distant relatives) for several reasons. First, the strength of the presumed relationship between telomere sharing and LTL similarity may weaken as relatives become more distant due to “noise” in TL introduced by telomere reprogramming and other factors affecting TL in germ cells. Second, when examining all types of relative pairs in a single analysis, the correlation between genome-wide relatedness and telomere sharing is very strong (e.g. r=0.97, Supplementary Figure S10), and regression models are unable to estimate both parameters; the correlation is weaker for specific relative pair types (Figure 4), enabling estimation of both parameters. Also, very distantly-related pairs have an increased number of recombination events occurring since their most recent common ancestor, which makes inferring IBD relationships at chromosome ends more difficult due to more recombination events occurring within the span of SNPs used to estimate IBD. Lastly, traditional SNP-based h2 analyses that leverage very distant relationships are not ideal for detecting the effects of telomere sharing on TL similarity because we have no evidence showing that the last SNP measured on each chromosome will be in strong LD with variation in the TL on that chromosome (given the gaps can be hundreds of kb).
Our study utilized two different methods for TL measurement, qPCR and a Luminex-based method. Although we have demonstrated that both methods produce measures that are strongly correlated with Southern blot measures (the gold standard for TL measurement using extracted DNA) 37; 38, using multiple methods can introduce heterogeneity into our TL measurement. Therefore, we removed experimental variation from our TL measures through mixed-effects modeling, and standardized TL distributions across all batches and methods to harmonize these measures. In addition, the measures used in this study are a relative measure of average TL (i.e., telomere abundance) rather than an absolute measure of average TL.
Our study has limited power. Assuming there is attenuation in the effect of Tshared on (ΔLTL)2 with decreasing degree of relatedness, we may be underpowered to detect this difference. Our ability to estimate heritability through the method presented by Visscher et al. 40 is limited by our small sample size. Visscher et al. show that at least 10,000 sibling pairs are required to accurately estimate heritability using the Haseman-Elston-like regression approach. As expected, our h2 estimate obtained from the meta-analysis β coefficient has a very wide confidence interval, and produced a point estimate greater than 1. Thus, our work essentially provides a lower bound of the narrow-sense h2. Regarding generalizability, it is important to note that these results were obtained using data from a rural Bangladeshi population and cross-population differences in LTL h2 could exist due to genetic, environmental, and demographic factors.
To better understand the direct transmission of TL from parents to offspring, future studies should analyze larger numbers of relative pairs, which would enable more accurate estimation of h2 due to direct transmission. Our method could be applied to UK Biobank data, which includes a substantial number of relative pairs, as well as LTL data in an upcoming release. Larger studies would also allow researchers to determine if the effect of telomere sharing on pairwise TL differences is weaker for more distant relatives. More accurate methods for measuring TL could increase power for these analyses (e.g., Southern blot), and chromosome-specific measures would enable one to study the transmission of specific telomeres and the impact on chromosome-specific TLs, although these approaches are not currently practical for very large cohorts.
In summary, we provide evidence supporting the direct transmission of variation in TL from parent to offspring. This is the first study to attempt to detect and/or estimate the h2 in LTL that is due to direct transmission. Together with prior evidence 18; 23; 25-27; 54-56, our work implies that any effect on TL in germ cells – be it due to genes, environment, or aging – will impact offspring TL. Such effects could potentially shift TL in a population in response to demographic or environmental changes or amplify the impact of natural selection on variation (i.e., SNPs) affecting telomere maintenance 58.These findings are also relevant for complex diseases for which TL has a causal impact on disease risk, as the h2 for these traits may be impacted by trans-generationally inherited TL, a component of h2 that cannot be explained by SNPs or other variation in non-telomere sequences. Applying our method to larger studies of relative pairs will enable robust estimation of LTL h2 attributable to direction transmission.
SUPPLEMENTAL DATA
Supplemental data (included as a separate pdf file) include seven tables and ten figures.
ACKNOWLEDGMENTS
This work was supported by the National Institutes of Health [R01 ES020506, U01 HG007601, P42 ES10349, R01 CA107431, R01 CA102484, P30 CA014599]. The authors thank all the men and women who participated in HEALS and BEST and all research staff who contributed to data collection. The Genotype-Tissue Expression (GTEx) Project was supported by the Common Fund of the Office of the Director of the National Institutes of Health, and by NCI, NHGRI, NHLBI, NIDA, NIMH, and NINDS.
Footnotes
↵** Affiliated with Department of Public Health Sciences, University of Chicago, at time of involvement. Currently affiliated with Division of Foodborne, Waterborne, and Environmental Diseases, CDC