• telomere
• life span
• species

Humans have relatively short telomere lengths from 5 to 15 kb (1⇓–3), and yet humans have much longer life spans than mice, which can start with telomere lengths around 50 kb (4, 5). Previous studies have suggested that the telomere shortening rate rather than the initial telomere length is the critical variable that determines species life span (4, 6⇓⇓⇓–10). In particular, we have previously shown that human telomeres shorten at a rate of ∼70 bp per y (1), which is in line with the rate published by other authors (3, 11⇓⇓–14), while mice telomeres shorten at a rate of 7,000 bp per y (4). These different rates of telomere shortening between human and mice could explain the different longevities of mice and humans. However, the telomere shortening rate has been investigated to date in few species (4, 6⇓⇓⇓–10, 15, 16), and using different techniques, which has prevented side-by-side comparisons of telomere shortening rates in phylogenetically distant species with different body sizes and life spans.

Here, to address whether telomere length and/or telomere shortening rates could explain species longevity, we measured telomere length in peripheral blood mononuclear cells from individuals of different species of birds and mammals at different ages in parallel, and calculated the telomere shortening rate per year in each species. A longitudinal study of telomere length was not considered here owing to the very different longevities of the species included in this study. Future studies warrant this type of analysis to understand telomere dynamics at the individual level. To measure telomere length, here we used a high-throughput quantitative fluorescence in situ hybridization (HT Q-FISH) technique, which allows to quantify individual telomere signals at a single-cell level (1) rendering data on both the average telomere length and individual telomere signals (SI Appendix, Table S1 and Fig. 1) (17, 18). In particular, we measured, in parallel, the telomeres of laboratory mice (Mus musculus) (Fig. 1A), bottlenose dolphins (Tursiops truncatus) (Fig. 1B), goats (Capra hircus) (Fig. 1C), reindeer (Rangifer tarandus) (Fig. 1D), American flamingos (Phoenicopterus ruber) (Fig. 1E), griffon vultures (Gyps fulvus) (Fig. 1F), Audouin’s gulls (Larus audouinii) (Fig. 1G), and Sumatran elephants (Elephas maximus sumatranus) (Fig. 1H). Laboratory mice were included as a control, as we had previously shown a rate of telomere shortening of around 7,000 bp per y, which is 100-fold faster than that reported in humans (4). The initial telomere length of the different species under study was estimated by linear regression (Fig. 1). Note that the initial telomere length value is only an estimate and telomere length dynamics may not be linear during the early stages of life (19). First, we confirmed a very high rate of telomere shortening in our current mouse cohort of 6,420 bp per y (Fig. 1A), similar to that previously described by us (4). Bottlenose dolphins showed a telomere shortening rate of 766 bp per y (SI Appendix, Table S1 and Fig. 1B) and an estimated initial telomere length around 90.7 kb (Fig. 1B). Goats showed a telomere shortening rate of 363 bp per y (Fig. 1C) and an estimated initial telomere length around 10.4 kb. Reindeer showed a telomere shortening rate of 531 bp per y (Fig. 1D) and an estimated initial telomere length of ∼ 19.8 kb. American flamingos showed a telomere shortening rate of 105 bp per y (Fig. 1E) and an estimated initial telomere length around 21.0 kb. Griffon vultures had a telomere shortening rate of 209 bp per y (Fig. 1F) and an estimated initial telomere length around 19.8 kb. Audouin’s gulls had a telomere shortening rate of 771 bp per y (Fig. 1G) and an estimated initial telomere length around 35 kb. Sumatran elephants have a telomere shortening rate of 109 bp per y (Fig. 1H) and an estimated initial telomere length around 36.3 kb. In the case of griffon vultures and Sumatran elephants, we were limited to the few available individuals at the Madrid Zoo; thus, in these cases the values obtained should be considered as a first approximation to the telomere shortening rates in these species.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Telomere measurements for various species. The telomeres were measured by HT Q-FISH in individuals of different ages for (A) mice (Mus musculus), (B) bottlenose dolphins (Tursiops truncatus), (C) goats (Capra hircus), (D) reindeer (Rangifer tarandus), (E) American flamingos (Phoenicopterus ruber), (F) griffon vultures (Gyps fulvus), (G) Audouin’s seagull (Larus audouinii), and (H) Sumatran elephants (Elephas maximus sumatranus). Each point represents the values for a different individual. The correlation coefficient (R²), slope (rate of telomere shortening in kilobases per year), and y intercept (initial telomere length) are presented on the graphs.

We next investigated relationships between telomere length, telomere shortening rate, and species life span. For the species maximum life span, we used the AnAge database (20). The average life spans were obtained from various sources (SI Appendix, Table S1). First, we did not find any correlation between the estimated initial telomere length and species longevity (Fig. 2 A–D). In particular, a graph of the species maximum life span versus the estimated initial telomere length resulted in an R² value of 0.0190 with a linear regression curve (Fig. 2A), and an R² value of 0.0407 with a power law regression curve (Fig. 2B). A graph of the species’ average life span versus the estimated initial telomere length resulted in an R² value of 0.125 with a linear regression curve (Fig. 2C), and an R² value of 0.145 with a power law regression curve (Fig. 2D). Note that there is even a trend for shorter life spans with longer initial telomere lengths with the low R² values just mentioned, and negative slopes in the regression line equations (Fig. 2 A–D). Also note that the inverse correlation between average life span and initial telomere length (R² = 0.125; Fig. 2C) was better than that between maximum life span and initial telomere length (R² = 0.019; Fig. 2A). These findings agree with a previous study that compared telomere length in more than 60 different species (21). Although the telomere shortening rate was not measured in that study, the authors concluded that the life span of a species could not be predicted from the initial telomere length and that there was a trend for short-lived species to have longer telomeres (21).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Species life span predictions with telomere parameters I. (A) Maximum life span vs. estimated initial telomere length fit with a linear regression line. (B) Maximum life span vs. estimated initial telomere length fit with a power law regression line. (C) Average life span vs. estimated initial telomere length fit with a linear regression line. (D) Average life span vs. estimated initial telomere length fit with a power law regression line. (E) Maximum life span vs. rate of telomere shortening. (F) The predicted life span vs. the maximum life span. The predicted life span is calculated by using the telomere shortening rate in the power law regression equation from E. (G) Average life span vs. rate of telomere shortening. (H) The predicted life span vs. the average life span. The predicted life span is calculated by using the telomere shortening rate in the power law regression equation from G.

Interestingly, when we plotted maximum life span versus the rate of telomere shortening for the different species, we obtained a power law curve with an R² value of 0.829 (Fig. 2E). The equation from this curve can be used to predict the life span of a species when given the telomere shortening rate without using any information about the initial telomere length with an R² value of 0.782 (Fig. 2F). The same graphs can be made using the average life span instead of the maximum life span (Fig. 2 G and H), and in this case the power law curve R² value is 0.934. The observation that life span versus telomere shortening rate fits a power law curve is in agreement with many natural phenomena fitting either a power law or exponential curve such as population growth, temperature cooling/heating, city sizes, species extinction, body mass, individual incomes, and the number of connections to nodes in a scale-free network, among others (22⇓⇓–25).

Alternatively, more linear life span predictions can be made using both the initial telomere length and the rate of telomere shortening. In this case, it seems unlikely that species die when their telomeres are completely eroded since the life spans predicted by complete telomere erosion are longer than the observed life spans for most species (SI Appendix, Table S1). Instead, we find here that the length of the telomeres when species die at the age of the maximum life span appears to be ∼50% of the original telomere length for that particular species, when considering the average of all of the species measured (SI Appendix, Table S2). Interestingly, when considering the timepoint of the average life span, the telomere length appears to be ∼75% of the original length (SI Appendix, Table S2). Therefore, we can calculate the life span of a species if we assume that telomeres shorten with a constant linear rate and that the time of death will occur once the telomeres have shortened to 50% or 75% of the original telomere length. The equation of the estimated life span if telomeres shortened to 50% of the original length is as follows: ((Initial telomere length) − (Initial telomere length) × 0.5)/Telomere shortening rate. A plot of the estimated life span at 50% original telomere length vs. the maximum life span yields an R² of 0.565 (Fig. 3A). The estimated life span at 50% original telomere length vs. the average life span yields an R² of 0.694 (Fig. 3B). Similar graphs are presented for 75% original telomere length (Fig. 3 C and D). With this dataset, the graph with 75% original telomere length vs. the average life span provides the most accurate results with an R² of 0.694. Although the R² value is the same as the value in Fig. 3B, this graph also has a slope that is closer to a value of 1, indicating a smaller shift between the actual and estimated life spans. Note that better correlation coefficients are obtained with the power law regression curves using the telomere shortening rate without taking the initial telomere length into consideration (Fig. 2 E–H).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Species life span predictions with telomere parameters II. (A) The estimated life span if telomeres shortened to 50% of the original length vs. the maximum life span. (B) The estimated life span if telomeres shortened to 50% of the original length vs. the average life span. (C) The estimated life span if telomeres shortened to 75% of the original length vs. the maximum life span. (D) The estimated life span if telomeres shortened to 75% of the original length vs. the average life span. The estimated life span is calculated using the following equation: (“Telomere length original” − “Telomere length original” × “percent of original length”)/“Telomere shortening rate.” (E) Graphical illustration which shows the main finding from this paper, which is that faster telomere shortening rates result in shorter species life spans.

Another trait that correlates with life span is body mass (26). In general, larger species such as elephants and whales have longer life spans than small species such as mice and rabbits. One investigation compared the mass and life span of 1,456 different species, and found a trend for longer life span with larger mass (R² = 0.397) (26). With the species in our dataset, we also observed a correlation between mass and life span (SI Appendix, Table S3). The species telomere shortening rate also correlated with body weight with an R² of 0.413 (SI Appendix, Fig. S1). Species with higher body weights tend to have lower telomere shortening rates and longer life spans.

Some authors have shown an inverse correlation between life span and heart rate, a variable related to organismal metabolism (27, 28), although more extensive studies do not seem to support this notion (29). Here, we set to address a potential correlation between heart rate and telomere length. First, we observed a correlation between life span and heart rate with our dataset (SI Appendix, Table S3). We also found a linear correlation between the telomere shortening rate and the heart rate with an R² of 0.974 (SI Appendix, Fig. S2 A and B).

Next, to investigate the effect of the multiple variables on life span when combined into the same model, we performed a multivariate linear regression. The input variables of telomere shortening rate, initial telomere length, body mass, and heart rate were fit to either the average life span or the maximum life span. The data used for the regression are presented in SI Appendix, Table S4. The log value of all of the data points was used for the regression instead of the original values. Each variable vs. the average life span or maximum life span had either a higher linear R² correlation coefficient when using log-transformed data, or there was no noticeable change in the correlation coefficient in the case of the initial telomere length variable. The model fit to the average life span resulted in an R² value of 0.997 and an adjusted R² value of 0.992 (SI Appendix, Table S5), demonstrating that these variables can predict the average life span. The P values (listed in the Pr(>|t|) column) were statistically significant for all variables. The telomere shortening rate was the most statistically significant variable (P = 0.000422). The model fit to the maximum life span resulted in an R² value of 0.950 and an adjusted R² value of 0.884 (SI Appendix, Table S6), demonstrating that the variables can also predict the maximum life span. In this case, only the telomere shortening rate variable was statistically significant (P = 0.0218). Again, we found an inverse relation between average life span and initial telomere length with a P value of P = 0.0302, with short-lived species having initial longer telomeres (SI Appendix, Table S5). Also, in the multivariate analysis, the relationship between the initial telomere length and the maximum life span was not significant in agreement with a weaker inverse correlation between initial telomere length and maximum life span compared with average life span (Fig. 2 A and C). Thus, these findings confirm that the telomere shortening rate (negative correlation), initial telomere length (negative correlation), body weight (positive correlation), and heart rate (negative correlation) can predict species life span and that, among these variables, the variable with the greatest power to predict the life span is the telomere shortening rate.

Finally, one caveat of studies with animals of different ages is that an effect can occur in which old animals with short telomeres selectively disappear due to death, and these telomeres are consequently not measured at older ages. Therefore, the telomere length could be artificially high at the older ages since only the animals with longer telomeres continue to survive at these ages. However, the fact that telomere shortening with age fits a linear regression in the majority of the species studied indicated that this phenomenon is not very distorting in our current study. Also, such disappearance of animals would only be expected to occur at very late ages, and the majority of the animals in this study were not extremely old (Methods).

• Copyright © 2019 the Author(s). Published by PNAS.