Material and methods
Our earlier study examined associations between telomere variables and
aspects of life history for bird species (Criscuolo et al. 2021). Our
comparative analysis followed the recommendations of the preferred
reporting items for systematics reviews and meta-analysis (PRISMA)
statement (Liberati et al., 2009).
In the present study, we focused on 30 species for which adult telomere
length, TROC, and life history variables were all recorded. For these
species, telomere lengths were estimated in kilobases (kb) using
electrophoretic separation of telomere restriction fragments or by
quantitative fluorescent in situ hybridization (respectively, TRF and
Q-FISH; Remot et al. 2021). All studies used samples of erythrocytes.
TROC was measured as the slope of telomere length regressed over age
(kb/year) for each of the species (Haussmann et al. 2003; Tricola et al.
2018), but excluding the telomere lengths of hatchling chicks or
yearlings (ages of 0-1 year). We followed Dantzer and Fletcher (2015)
and Tricola et al. (2018) in using maximum lifespan in nature to typify
longevity and mean adult female body mass to typify body size, and
log-transformed both variables. Among the 30 species of birds,
body-size-independent aspects of longevity (longevity associated with
the pace of life) were estimated as the residuals of maximum lifespan
regressed on mean female body mass, and these residuals were checked for
normality using a Q-Q plot and Shapiro-Wilk test. When multiple sources
were available, a mean value of adult telomere length or TROC was used.
All sources are reported in the online supplementary information (ESM).
We divided lifespan into two statistically independent parts, based on
the regression of lifespan onto body size (as estimated by mean adult
female body mass for each species). The first variable represented
mass-associated aspects of lifespan, as indicated by the predicted
values of lifespan from the regression, and termed “mass-predicted
lifespan.” The residuals from the analysis were termed
“mass-independent lifespan,” and these two measures of lifespan were
statistically uncorrelated. The latter variable estimated lifespan at a
given body size, and may be interpreted as a measure of lifespan on the
slow-fast continuum of the pace of life (Gaillard et al. 1989; Dobson
and Oli 2007). Associations of telomere and lifespan variables were
examined by comparison of both correlation and by phylogeny-adjusted
correlation (after Price 1997).
A phylogeny was obtained from BirdTree (Figure 1), with 100 phylogenetic
trees downloaded from http://www.bird.tree.org (De Magalhaes and Costa
2009; Jetz et al. 2012) using ape, apTreeshape and caper R packages.
Branch lengths were estimated using the coalescent method, thus
reflecting an estimate of relative divergence times for the phylogeny
(Rannala and Yang 2003). Associations of the bird phylogeny with adult
telomere length, TROC, adult female body mass, maximum lifespan, and the
residuals of lifespan on body mass were estimated using Hadfield and
Nakagawa’s (2010) Markov chain Monte Carlo generalized linear mixed
model (MCMCglmm; Hadfield 2010) package in R (R core team 2020). The
MCMCglmm package was also used to produce phylogeny-adjusted estimate of
associations of telomere, body mass, and longevity variables.
Adjustments for within-species sample sizes of telomere variables were
entered into the phylogenetic models as,\(mev=\ \frac{1}{\left(N-3\right)}\) , where N is the
number of samples in each study.
MCMCglmm was used to produce two types of results: 1) an estimate of the
phylogenetic pattern in each of the study variables (viz., \(\rho\) =
the proportion of variance that could be statistically accounted for by
a matrix of the phylogenetic pattern); and 2) degree of correlation of
pairs of variables with statistical adjustment to remove the
phylogenetic pattern. Pearson’s correlations, unadjusted for phylogeny,
were also calculated for comparison with phylogeny-adjusted results.
Where directions of associations were specified by prediction, we used
one-tailed tests. Cohen’s (1988) criteria for effect sizes of
associations were applied: small r = 0.1, medium r = 0.3, and large r ≥
0.5.