To our knowledge, this is the first study that has compared, as impartially as possible, measurements of LTL by qPCR versus Southern blot analysis in two laboratories with considerable experience in performing these two methods. In comparing the methods one must balance precision and other features of a given telomere length measurement method with its practicality.
It is important to point out that the pre-analytical factors that impact the Southern Blot analysis and qPCR for telomere length measurement could be very different. Whereas the integrity of genomic DNA is crucial for Southern blots, it is less critical for qPCR. However, residual PCR inhibitors that may still be present even after DNA purification could contribute to variability in qPCR. The influences of different types of blood collection tubes, as well as individual donor differences have been observed in a separate study (J. Lin, J. Cheon, E. Epel and E. Blackburn, manuscript in preparation). The present study used DNA samples that have been examined for their DNA length (lack of detectable low-molecular weight DNA fragments, as assayed on agarose gels) as a quality control for sample integrity, but we did not address pre-analytical conditions, including blood collection and storage, DNA purification and storage, that might contribute to qPCR variability.
As shown in , we observed high correlations between the two sets of measurements of LTL performed by either the Southern blots of the TRFs or telomere DNA content generated by the qPCR. The measurement error, defined by the inter-assay CV, which on the two occasions tested here, were 6.45% for the qPCR, and 1.74% for the Southern blots. It is useful to regard this error in the context of the relation between LTL and age in the present study, and specifically with respect to epidemiological research and ultimately clinical practice.
The larger error of the qPCR relative to the Southern blots could explain the findings that age, across 30 years, accounts for 17.2% of the inter-individual variation in the mean T/S (qPCR method), while it explains 29% of this variation in the mean TRF length (Southern blot method) (). Thus, we suggest that measurement error should be a primary consideration, as well as any unique demographic or biological features, in large-scale cross-sectional studies that show little or no LTL shortening over a wide age range.
As an illustrative example, consider an adult whose LTL
kb. Measurement errors of 2 and 6% of 6
kb amount to 120 and 360
nt, respectively. Although the average rate of LTL shortening in adults is highly variable, in cross-sectional studies it on average amounts to ~30
nt/year, as also shown in . In equivalence of age-dependent LTL attrition, the error of the measurement by the Southern blot amounts to 4
years, while that of the qPCR amounts to 13
years. However, it is also important to note that the rate of LTL change over time has been shown in longitudinal studies to have a strong dependence upon baseline LTL, and thus base-line TL should also be co-varied in longitudinal analyses of telomere shortening (14–16
). Thus, in cross-sectional studies the shortening rate is only inferred indirectly based on group analysis, and masks the findings that the rate of LTL shortening in individual adults measured longitudinally is highly variable.
In addition, age-dependent LTL attrition is only one component of LTL dynamics; the other component is birth LTL. The range of inter-individual variation in LTL in adults after age adjustment, in cross-sectional studies, is
kb, and an important contributor to this variation might be not only the rate of age-dependent LTL shortening during adult life, but the variation in LTL across newborns, which is ~4–6
). Another significant contributor to LTL in adults is likely to be the rate of telomere attrition in the period from birth to early adulthood, which is much higher than that seen in adults in both humans (11
) and non-human primates (21
). Given this wide inter-individual variation in LTL, a 6% error in telomere length measurement might be acceptable in epidemiological research and clinical practice, if the goal of LTL measurement is not to determine the rate of LTL shortening but to rank individuals for susceptibility to a given disease based on having relatively short, average or long telomeres for their age.
displays deviation from linearity of the relation between the mean length of the TRFs and the T/S ratio, which was observed in other studies (2
). The addition of a quadratic term increased the fit of the data to the regression line. No deviation from linearity is observed between the mean TRFs and telomere DNA content generated by a non-PCR method of the strictly canonical telomeric repeats region of telomeres, suggesting that this phenomenon is PCR mediated (23
It is important to regard the ramifications of this work within the broader framework of epidemiology and clinical practice. Precision alone might not always justify the use of Southern blot method of the TRFs, since it is labor-intensive, costly, requires a large amount of DNA, cannot be performed in DNA that is even modestly degraded and requires expertise that might be available only in a few laboratories (1
). Furthermore, while the qPCR method measures strictly the canonical region of telomeres, the TRF length data include long stretches of the non-canonical region, which might be variable across the general population due to polymorphism in the restriction sites that are the targets of the enzymes generating the TRFs.
Epidemiologists have employed various methods to measure telomere length but there has been less discussion of the shortcomings of the methods used. Moreover, measurements errors of the same method often differ among laboratories. Therefore, the findings of the present study by two laboratories with considerable experience in qPCR and Southern blots might not represent those generated in less experienced laboratories. For these reasons, large-scale epidemiological studies comparing the Southern blot analysis versus qPCR and possibly other methods (measured in a number of laboratories for each method) will bolster understanding of aging-related maladies and gage the risk of developing them. This will more fully address matters related to the question: which is the optimal method to measure telomere length? Without such studies, identifying the optimal method for assessing telomere length unbiased by preconceived views will be hampered.