Improving Precision in Investigating Aging: Why Telomeres Can Cause Problems

Abstract

AFTER many years of near unchallenged supremacy, the use of telomere length as a marker of bio-aging has hit the rocks of controversy. The root of this controversy centers on methodology and its appropriate use when telomere length analyses are translated from laboratory studies into the wider world of the epidemiologist and clinical researcher. In the latter environment, many confounders abound.

Results from studies using traditional Southern blotting methods are generally consistent. As individuals get chronologically older, their telomere length diminishes. In the presence of extant disease, this rate of attrition is generally accelerated. Twin studies have also indicated that telomere length predicts the likelihood of mortality (1,2), though this may not always hold among the “oldest of the old” (3,4). The reasons for this particular discrepancy may be more to do with changing selection pressure and survivor effects than with methodology.

The use of Southern blotting in large clinical or epidemiological studies is often precluded by the cost, scale, labor intensiveness and degree of expertise needed to undertake the analyses. A similar argument can be made about using fluorescent in situ hybridization variant methods or single telomere length analysis (reviewed in (5)), despite their elegant nature when applied on a laboratory scale.

Consequently, quantitative polymerase chain reaction has provided an effective high throughput alternative with which to measure telomere lengths in peripheral blood cells in larger clinical studies (6–8). Data derived from a flood of such studies employing this method are often equivocal. Furthermore, there is a worrying lack of consistency with data derived by Southern blotting. This has been discussed elsewhere in detail (9).

The exact cause of this equivocacy remains to be determined, but it is almost certainly multifactorial. One immediate observation that may impact heavily on this is the remarkable range in the coefficient of variation in many published quantitative polymerase chain reaction studies. This has been reported to be as large as 27% (10), which is so high as to render drawing any conclusions from such data set fraught with error. Some studies have attempted to reduce the degree of covariance through increasing replication and by adjusting for interbatch variation in sample analysis. Although this has been reported to reduce the coefficient of variation to less than 1%, the interindividual variation in relative telomere lengths, at any give age, varies by almost one order of magnitude (11), which raises questions about appropriate use of the methodology.

More recently, Mather and colleagues (12) have examined age-sensitive indicators of physical function (lung function, blood pressure, and grip strength) working on the hypothesis that decline in these functions with increasing chronological age should be reflected in shorter telomere lengths in peripheral blood leukocytes (PBLs).

Critically, this analysis is on two narrow age range and ostensibly healthy participant cohorts (aged 40–49 and 64–70 years, respectively), not confounded by possible molecular changes associated with analysis of PBLs from diseased participants. Their findings are consistent with previous reports, looking at the older participants (4,13), but worrying from the point of view that these cohorts comprise younger individuals. Methodologically, their use of the quantitative polymerase chain reaction method to measure telomere lengths in PBLs is sound and is reported to give data with an average coefficient of variation of 3.44% for their cohorts, yet the findings are not intuitive.

These authors also determined age-related kilobase lengths in their cohorts, which were in keeping with previous reports using similar methodology, but lower than those generated by mean terminal restriction fragments. The age-related decline in telomere length of circa 50 bp per year is also in keeping with many reports in the field.

Critically, given the general lack of agreement in the field, there is a need to replicate such studies and address the differences, but what is the best way to do so?

First, double-blind analyses of the same sample and control set, undertaken in multiple laboratories, should go some way to indicating just how comparable and reproducible the respective methodologies are.

Second, a standardized DNA preparation method that minimizes oxidative damage during preparation should also be used, but given the routine nature of this procedure in most laboratories, this may be not a major concern. Another consideration here is the failure of most studies employing Southern blotting to include a control hybridization with a small single-copy gene fragment to show that all samples are adequately digested. This has the propensity to show small-scale partial digestions, which can add multiples of 256 bp to terminal restriction fragments, as on average, type I restriction enzymes used in preparing DNA for terminal restriction fragment analysis cut once every 256 bp. These are not normally detectable just by looking for the appearance of microsatellite bands on the final blot. Consequently, mean terminal restriction fragment values may be overestimated. This may contribute to the discrepancies between the kilobase sizes generated by the two methodologies.

Third, a further consideration, when moving from a small-scale laboratory study to larger epidemiological cohorts, is the inclusion of suitable controls. This is particularly relevant for clinical studies. Factoring in possible confounders such a socioeconomic status, lifestyle factors, psychological factors, and addressing their effects relative to chronological age should also be encouraged (14). At a fundamental level, a standard set of DNAs, or even synthetic telomeric DNA fragments of known sizes, might be employed as basic controls. Critically, for quantitative polymerase chain reaction analysis, a standardized set of primers, cycling conditions, and amplification reagents should also be included. Simple changes in amplification mixtures or the polymerase used may have dramatic effects on relative telomere size estimated in different laboratories. This is also confounded by different types of polymerase chain reaction systems (eg, Taqman & Light cycler) utilizing different chemistry. Uniformity is thus highly desirable, and a “standard operating procedure” would be good for the field.

Fourth, epigenetic status may act as a potential confounder. The proinflammatory state underlying many pathologies may also affect the epigenetic status of the individual. This is of direct relevance to the use of Southern blotting, which relies upon restriction-digested DNA to identify terminal restriction fragments. Methylase activity can be affected directly by the levels of inflammatory cytokines (15,16); hence, maintenance of, or changes in epigenetic status could result as a consequence of their activity. It is possible that such changes could, as a result of altered methylation status at restriction sites used to generate terminal restriction fragments, affect their size. Although this is a practical consideration, it remains to be determined if this is manifests as a real confounder.

Perhaps of more overriding importance is the question of just how good telomere lengths are as markers of bio-aging? At this juncture, they may be best viewed as effective but imprecise. They give a good indication of how many “miles are on the clock” and hence the likelihood that adverse biologic events may occur. Disentangling age-related telomere attrition from disease-related attrition, however, is difficult. The latter is not within the scope of this article and is covered in depth by Sprott (17).

In order for any such marker to be considered a validated marker of bio-age, it should at least fulfill the criteria specified in Baker and Sprott’s original postulate that

“A Biomarker of Aging (BoA) is a biological parameter of an organism that either alone or in some multivariate composite will, in the absence of disease, better predict functional capability at some late age, than will chronological age” (18).

Most markers fail this basic definition as they cannot predict adequately functional capacity. Telomere length may not be a particularly robust biomarker in this context. One weakness with its use, under such a definition, is that telomere attrition is a feature of cellular senescence. Extrapolation of cellular telomere biology to the level of the tissue or organ, or the whole organism, is not straightforward. To do so, one must take account of the number of senescent cells (generated by both replicative senescence and stress or aberrant signaling-induced senescence), their location, and similarly the number and location of cells lost through insult, in each respective organ or tissue, to gauge properly the effect on its functional capacity. Using PBL telomere length as a surrogate for the functional capacity of solid tissues and organs is thus imprecise and prone to error.

Such a postulate, however, may also go some way to explaining the lack of consistency in data from multiple clinical studies using different methodologies to determine telomere length. PBL telomere length may simply not always function adequately to explain what is happening in solid organs.

Perhaps, a better way to assess bio-age is the use of the cell cycle regulator CDKN2A, which functions to hold a cell in a state of growth arrest. An increasing number of recent studies have shown that this marker can better fulfill Baker and Sprott’s criteria (19–22). Indeed, the two markers have even been compared together. In this instance, CDKN2A has proven the more effective in predicting functional capacity (19). Its measurement can be either via reverse transcription polymerase chain reaction or by antibody detection and subsequent signal quantification. Generating a standardized assay is relatively straightforward and not technically complex. The data already generated by a number of labs appear to replicate well, showing increasing levels of CDKN2A transcriptional expression with increasing chronological age in PBLs and solid organs (19–21). Moreover, in solid organ studies, the levels again increase with decreasing organ function. It has not yet, however, been evaluated in the context of mortality. In support of its robustness, however, CDKN2A levels have been reported to positively associate with levels of proinflammatory cytokines, such as interleukin-6 (21). Telomere length also shows such an association (8), but this does not hold when evaluated for association with physical biomarkers of ageing, such as grip strength (12). Recent data from Tianen and colleagues (23), however, confirm that even among nonagenarians, levels of inflammatory markers correlate with grip strength and other markers of physical performance. It will be interesting to see CDKN2A levels evaluated in this context.

These analyses will be well supported by the development of participant cohorts, designed to address the impact of psychological, sociological, as well as biological determinants of aging and disease. In particular, those such as the psychological, social, and biologic determinants of ill health (pSoBid) cohort that compares individuals at social extremes and the Halcyon collaborative research programme, which combines multiple independent life course cohorts, with unprecedented power to detect statistical associations of small effect size should prove informative (24,25).

Until data from these and other longitudinal studies emerge, the utility of applying telomere length measurements to give a precise indication of bioage remains debatable. The use of alternative markers, such as CDKN2A, alone or in combination with telomere length, seems prudent.

References

Author notes