Introduction
Telomeres are repeating DNA sequences found on the ends of chromosomes which shorten with age and are implicated in senescence. Past studies found that telomeres tend to shorten at faster rates in species with shorter lifespans (Dantzer & Fletcher, Reference Dantzer and Fletcher2015; Haussmann et al., Reference Haussmann, Winkler, O'Reilly, Huntington, Nisbet and Vleck2003; Tricola et al., Reference Tricola, Simons, Atema, Boughton, Brown, Dearborn, Divoky, Eimes, Huntington, Kitaysky, Juola, Lank, Litwa, Mulder, Nisbet, Okanoya, Safran, Schoech, Schreiber and Haussmann2018). Whittemore et al. (Reference Whittemore, Vera, Martínez-Nevado, Sanpera and Blasco2019), hereafter WEA, recently generated a dataset of cross-sectional telomere shortening rates across nine different tetrapod species using the quantitative fluorescence in situ hybridization technique to measure telomere length. They found a strong negative relationship between telomere shortening rates and lifespan. However, their analyses did not consider the phylogenetic relationships between the species they compared. It is now well established that it is necessary to consider the phylogenetic relationships between species in cross-species analyses (e.g. Garland et al., Reference Garland, Bennett and Rezende2005). Not controlling for the phylogenetic relationship between species implicitly assumes a “star phylogeny” in which all species are equally related to each other or that the traits in question are not influenced by phylogeny (Felsenstein, Reference Felsenstein1985). Neither of these are usually the case. Here we re-analyze the data from WEA using phylogenetically informed analyses. Our analyses are similar to the recent re-analysis of WEA by Udroiu (Reference Udroiu2020). Our analysis differs from Udroiu’s in two key ways. First, instead of independent contrasts and analyses of residuals we use phylogenetic generalized least squares (PGLS) regression with multivariate analyses. Second, we exclude the laboratory mouse in some of our analyses out of concern that laboratory mice have been shown to have considerably longer telomeres than their wild counterparts possibly due to artificial selection and/or inbreeding (Eisenberg, Reference Eisenberg2011; Kotrschal et al., Reference Kotrschal, Ilmonen and Penn2007; Manning et al., Reference Manning, Crossland, Dewey and Van Zant2002; Weinstein & Ciszek, Reference Weinstein and Ciszek2002).
Objective
While we appreciate these new data generated by WEA, and the questions they address with these data, we have concerns about their statistical techniques and inclusion criteria. We elaborate these concerns and discuss re-analyses of the data using appropriate phylogenetic corrections and exclusion of the laboratory mouse below.
Methods
We compiled species-level phylogenetic trees from sources reported in the supporting information. This supporting information also contains further details on the methods briefly described below. We reconstructed a time-calibrated supertree, which was pruned to match the species in WEA. We used PGLS (Grafen, Reference Grafen1989) to test linear correlations between lifespan, telomere shortening rate (TSR), initial telomere length (TL), body mass and heart rate following WEA. Most data are log-transformed for normalization, but we also include some analyses of the non-transformed data for comparison with WEA. We estimated Pagel’s λ with default bounds (0-1) using maximum likelihood.
Results
As illustrated in Figure 1, the species in question lie in two distinct groups, which can create a “worst case” scenario for cross-species analysis (Felsenstein, Reference Felsenstein1985). Re-analyzing the primary findings from WEA (their Figures 2, S1, & S2) with appropriate phylogenetic regressions leaves the general findings of WEA intact (Table 1). However, we find a strong phylogenetic association between body mass and TSR (Pagel’s λ = 0.97), but not among the other associations in Table 1. Re-analyzing the WEA dataset with phylogenetic corrections and excluding the laboratory mouse (C57BL/6 strain) weakens all results (Table 1). While the key association between TSR and lifespan remains in the same direction and significant, the magnitude of the effect is attenuated. The multivariate analyses of lifespan in WEA (their Tables S5 and S6) includes both body mass and heart rate, which strongly correlate (R2 = 0.94, p < 0.0001). WEA found that TSR, initial TL, body mass and heart rate all significantly correlate with average lifespan. To avoid multicollinearity, we excluded heart rate from the phylogenetic multivariate models, which show that only TSR significantly predicts average and maximum lifespan (Table 2). Excluding the mouse samples strengthens the positive association of initial TL with maximum lifespan.
*p < 0.05, **p < 0.01, ***p < 0.001
*p < 0.05, **p < 0.01, ***p < 0.001
Discussions
Telomere shortening across species has been shown to be strongly influenced by phylogeny in birds (Tricola et al., Reference Tricola, Simons, Atema, Boughton, Brown, Dearborn, Divoky, Eimes, Huntington, Kitaysky, Juola, Lank, Litwa, Mulder, Nisbet, Okanoya, Safran, Schoech, Schreiber and Haussmann2018) and telomere length to be strongly influenced by phylogeny in mammals (Gomes et al., Reference Gomes, Ryder, Houck, Charter, Walker, Forsyth, Austad, Venditti, Pagel, Shay and Wright2011). Particularly concerning is the fact that a past study suggests that telomeres might lengthen with age in wild mice (Ilmonen et al., Reference Ilmonen, Kotrschal and Penn2008), which can live longer than laboratory mice (Miller et al., Reference Miller, Harper, Dysko, Durkee and Austad2002). However, we find that corrections for phylogeny and excluding the potentially problematic laboratory mouse does not substantively alter the general findings of WEA’s analysis.
Conclusions
Together with past analyses of lifespan and TSR (Dantzer & Fletcher, Reference Dantzer and Fletcher2015; Haussmann et al., Reference Haussmann, Winkler, O'Reilly, Huntington, Nisbet and Vleck2003; Tricola et al., Reference Tricola, Simons, Atema, Boughton, Brown, Dearborn, Divoky, Eimes, Huntington, Kitaysky, Juola, Lank, Litwa, Mulder, Nisbet, Okanoya, Safran, Schoech, Schreiber and Haussmann2018) and another re-analysis of WEA (Udroiu, Reference Udroiu2020), it is clear that species with more rapid TSR tend to have shorter lifespans even when correcting for body mass. Nonetheless, we think it is important to share our re-analyses to help assure others with the same concerns about the WEA analysis we raised, to serve as a reminder of the importance of best practices in cross-species analyses, and to provide more reliable estimates of the associations in question.
Acknowledgements
We thank editor Reinder Radersma and two reviewers for valuable comments on a previous version of this manuscript.
Conflicts of Interest
None.
Author Contributions
M.L.P. analyzed the data and both authors wrote the paper.
Funding Information
M.L.P. is supported by the Research Council of Norway through its Centres of Excellence funding scheme (223257). The authors declare no competing interests.
Data Availability Statement
All species trait data is available in the online supporting information in Whittemore et al. (Reference Whittemore, Vera, Martínez-Nevado, Sanpera and Blasco2019). Phylogenetic data can be accessed through the sources provided in the supporting information for this study.
Supplementary Materials
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/exp.2020.18.
Comments
Comments to the Author: The paper correctly starts from the fact that interspecies comparison must be analysed with phylogenetic corrections, and uses a phylogenetic generalized least squares method to re-analyse data from Whittemore et al. (2019). The article is simple and clear and I have only minor concerns. 1) Analyses made by the authors: data are log-transformed, which is correct (and this was not done by Whittemore et al.). Why do the authors have used also non-transformed data? Since lifespan and heart rate are directly and inversely correlated with body mass, all analyses could be redundant. In this case, analyses of residuals are usually performed, e.g.: mass against lifespan, I get residual lifespan and the I test correlation between residual lifespan and telomeres. 2) Original data: Average lifespan is a quite bizarre quantity in my view: in the best case it is redundant with maximum lifespan. Data are inhomogeneous: some species comprise only adult, other ones adult+infant (this is important because telomeres shorten more rapidly in infants). Data for lifespans (and perhaps body mass) should be checked: e.g., value for Phoenicopterus ruber is 60 years, while in Anage (the source cited by Whittemore et al.) it is reported as “Not yet established” (and it reports 44 years for Phoenicopterus roseus and 36.7 years for Phoenicopterus chilensis...). 3) Considerations of the authors: In my view, the authors correctly excluded laboratory mouse. Indeed, data for telomeres are from laboratory mice, but the record of max. lifespan is from a population of feral mice (see Anage), which usually have much shorter telomeres (12 kb like humans).