DISCUSSION
Our re-analysis of the data from Gomes et al 2011 demonstrates that telomere length has evolved to be shorter with increasing body mass and lifespan, across 57 species of mammals spanning 15 orders (Fig. 2). This supports the hypothesis that short telomeres evolved in large, long-lived mammals to enable replicative senescence as a mechanism to ameliorate cancer risk conferred by a larger number of cells and longer time to accumulate oncogenic mutations (Gorbunova et al., 2014; Risques & Promislow, 2018). Alternatively, since TL was measured in adults, larger species may erode their telomeres faster than small species, giving rise to the negative correlation between TL and body mass (Monaghan & Ozanne, 2018). However, larger species have been shown to have lower telomere shortening rates across 9 tetrapod species, but this association was driven by inclusion of a domesticated mouse strain (Pepke & Eisenberg, 2020). Since telomere loss rates tend to be slower in longer lived species (Haussmann et al., 2003; Dantzer & Fletcher, 2015), TLs at conception are likely to be even longer in shorter-lived species than estimated in adulthood.
In contrast to our findings, past studies have shown no significant associations between TL with mass or lifespan across 15 rodent species (Seluanov et al., 2007) or between TL and lifespan of 19 bird species (Tricola et al., 2018). These differences might be explained by these rodents and birds being relatively small and short-lived compared to many of the species included here. Among 8 species of relatively large mammals and birds (all >0.5 kg), longer early-life TL predicted longer lifespan (the opposite of our findings), but only when accounting for body mass and telomere shortening rate (Pepke & Eisenberg, 2020). Gomes et al. (2011) failed to show the association between TL and mass that we found analyzing the same dataset. While the reasons for our differing results with Gomes et al. are unclear, they may be driven by a transformation of telomerase activity in the multivariate model (discussed below). Regardless, broader taxonomic studies that are attentive to model construction are needed to unravel the generality of these patterns.
Our re-analysis of Gomes et al. (2011) showed that telomerase activity had only inconsistent associations with lifespan and body mass (Tables 1 and S1-4), which call into question past results suggesting that high telomerase activity has been selected against in larger and longer-lived species. We found some hints that telomerase activity was inversely associated with mass when telomerase was either log-transformed or binary transformed (Tables S3-4 and Fig. S1). Thus, the associations between telomere biology, mass and lifespan are sensitive to modeling strategies. While we found an inverse association of binary telomerase activity with mass, the biological meaning of binary coded telomerase activity is unclear. Perhaps having any telomerase activity increases the level of cancer risk, but the particular level of telomerase activity greater than zero does not change cancer risk. This might suggest that telomerase activity variation greater than zero has no meaningful effects on the phenotype under selection with changing body mass. In contrast, Seluanov et al. (2007) found a strong negative linear correlation between (untransformed) telomerase activity and body mass across rodent species. Seluanov et al.’s analysis differed from this study in that they measured telomerase activity in spleen, liver, and kidney tissues, while this study depends on telomerase from cultured cells. Additionally, the rodents in Seluanov et al.’s analysis all expressed telomerase and had long telomeres (most species >20 kb), while this analysis includes many species without detectable telomerase activity and with shorter TLs. The fact that Seluanov et al.’s analysis showed non-transformed telomerase activity had an association with mass among animals with only non-zero telomerase activity argues against using telomerase activity as a binary variable. Furthermore, small levels of telomerase activity in normal stem cells is thought to be insufficient to fully maintain telomere lengths (Shay & Wright, 2010).
Telomere length was a strong predictor of neoplasia rates across 22 species spanning 11 orders (Fig. 3a). Neoplasia rates are a good proxy for malignancy rates (Fig. S2). In turn, malignancy rates likely reflect the incidence of development of cancers with fitness consequences (Ewald & Ewald, 2015). Thus, natural selection against longer telomeres may be driven by cancer-related mortality. Our findings are consistent with the hypothesis that shorter telomeres have evolved as a cancer resistance mechanism across mammals. Larger and longer-lived species did not have higher neoplasia incidence rates (Fig. 3c), which is a demonstration of the well-known Peto’s paradox. The negative correlation that we found between TL and body mass may partly underlie Peto’s paradox; i.e. larger species have evolved shorter telomeres as an anti-cancer mechanism to compensate for otherwise higher cancer risks (Caulin & Maley 2011). The optimal TL in mammals may therefore have been shaped by a trade-off between the increased cancer risk of long telomeres and the increased susceptibility to age-dependent degenerative diseases conferred by short telomeres (Tian et al., 2018; Aviv et al., 2017). We also found some evidence that binary telomerase activity was associated with higher cancer rates (Fig. 3d), although this effect was not significant when controlling for TL (Table 2). High telomerase activity is a prerequisite for most cancer cells (Wright & Shay, 2005) and may decrease the threshold for cell immortalization leading to increased cancer risk in species with somatic telomerase activity.
Although species with short telomeres have lower cancer incidence rates, some species with long telomeres appear to have lower cancer rates than expected given their TL. This may be explained by the evolution of various telomere-independent cancer defense mechanisms in these typically long-lived species (Seluanov et al., 2018; Erten & Kokko, 2020). The most striking outlier in our analysis is the tiger (Panthera tigris ), having a large body size (162 kg) despite having very long telomeres (50 kb), high telomerase activity (10%), but a moderate neoplasia incidence rate (18%), suggesting the potential for discovery of novel tumor suppressor mechanisms within this species.
We provided the first evidence that telomeres are longer across multiple domesticated species spanning 5 orders (mainly members of Artiodactyla, but also of Perissodactyla, Carnivora, Rodentia and Lagomorpha). Long telomeres may form part of a “domestication syndrome” in mammals (Darwin, 1868; Wilkins et al., 2014), although the mechanisms generating this trait change are not well-known (Eisenberg, 2011). Alternatively, long telomeres were ancestral in domesticated species, but somehow connected to the early stages of domestication (Lord et al., 2020). We found that domestication has resulted in an estimated 38.5% increase in TL (Table 3), which is equivalent to the effect an 98.4% decrease in body mass would have on TL (Fig. 1a). There does not appear to be a single directional change in body sizes of domesticated animals (Lord et al., 2020). However, domesticated species may have experienced changes related to fertility, reproductive lifespan, growth rate, degree of inbreeding, parasite load and energetic constraints (Diamond, 1997; Lord et al., 2020), which are all effects known to shape intraspecific variation in TL (Weinstein and Ciszek, 2002; Manning et al., 2002; Bebbington et al., 2016; Monaghan & Ozanne, 2018; Sudyka, 2019; Giraudeau et al., 2019). Comparisons of TL and other phenotypes in domesticated species with their wild or feral cousins might provide more insights into the particular causes of TL changes with domestication.
Our re-analyses of the valuable dataset generated by Gomes et al. (2011) showed considerably divergent results from those shown by Gomes et al. (2011) and that many results are not robust to varying modeling strategies. Unlike Gomes et al. (2011), we showed that telomere length decreases with increasing body mass. While Gomes et al.’s results suggest a reduction in telomerase activity with increasing body mass, we only found weak evidence for this. Our results suggest that the co-evolution of telomere biology with lifespan and body mass across species needs to be reconsidered. The negative association between TL and mass and lifespan could be explained by the increased cancer risk we found in species with longer telomeres. Finally, we found that domesticated species appear to have evolved longer telomeres – suggesting new avenues to understand telomere evolution through closer examination of the domestication process and artificial selection experiments.
Acknowledgements: We thank Amy Klegarth for collaboration on earlier exploration of these data and Jerry Shay and Chris Venditti for collegial discussion of our discrepant results. We thank Pat Monaghan and Dan Nussey for organizing the Diversity in Telomere Dynamics Workshop, which facilitated the authors meeting and collaborating on this project. M.L.P. thanks the Research Council of Norway for funding through its Centres of Excellence scheme (223257).