DISCUSSION
Our study is the first investigation of the relationship between telomere length and DNAm dynamics in wild animals exposed to naturally variable early life conditions. In our longitudinal analysis we found a negative relationship between individual changes in telomere length and changes in the level of methylated CpG loci (92 CpG loci in our study) across early life, such that individuals with accelerated rates of telomere shortening were associated with increases in DNAm levels. The observational nature of our field data means that we were unable to establish whether the negative relationship between telomere length and DNAm changes reflected coincidental or causal effects. That is, whether DNAm and telomere length respond comparably to the same intrinsic/extrinsic conditions experienced across early life and/or whether potential regulatory effects between DNAm and telomere length may exist. How changes in telomere length and DNAm relate to each other and respond to early life conditions warrants further investigation since it may help identify potentially distinct aspects of biological age that DNAm and telomere length reflect. It is important to note that while variation in DNAm levels may reflect variation in the early life environment (Angers, Castonguay, & Massicotte, 2010; Sheldon et al., 2018; 2020; Makinen, van Oers, Eeva, Laine & Ruuskanen, 2020), the relevance of early-life, ‘genome-wide’ hypo or hyper methylation in the context of biological age remains unclear. Thus, while our MS-AFLP analyses are useful to compare DNAm and telomere length associations in different environments, next-generation sequencing data (used to generate an ‘epigenetic clock’) are necessary to compare DNAm and telomere length predictors of biological age in different environments (Horvath & Raj 2018; Banszerus et al., 2019). We did not detect a relationship between cross-sectional measures of DNAm and telomere length at either day 3 or day 11 post-hatch. However, telomere length at day 3 was weakly, but significantly positively associated with clutch size. Given this relationship was positive rather than negative, it suggests that the within clutch decline in telomere length across the laying order that has been seen in other studies did not have much effect here. The positive relationship could perhaps be due to parental age if older parents had a reduced clutch size and their offspring inherited shorter telomere lengths or lost more up to this point (Monaghan and Metcalfe, 2019).
Evidence is accumulating to suggest that longitudinal measures of within-individual change in telomere length may represent a more useful biomarker of ageing/individual condition than static, cross-sectional telomere length measures (Wood and Young, 2019; Boonekamp et al., 2014; Tricola et al., 2018; Wilbourn et al., 2018. However, to our knowledge, no other study has conducted longitudinal analyses on the relationship between within-individual change in telomere length and DNAm across early development, in any species. Although it may be expected that factors associated with telomere length would also be associated with telomere length changes , telomere length at a given timepoint is the outcome of both initial telomere length and subsequent attrition/elongation, which may not be associated. Indeed, in our supplementary analysis (Supplementary Analysis 2) the initial length of telomeres (at day 3) did not affect the rate of telomere length change across early life after controlling for effects of regression to the mean (Verhulst et al., 2013) (however, other studies have detected this trend (Salomons et al., 2009; Aviv et al., 2009)). Hence, variation in initial telomere length, which may be more influenced by heritable genetic factors (Broer et al., 2013, but see Voillemot et al., 2012), may confound the relationship between environmental factors associated with telomere length versus telomere length change , and indeed, may confound the accuracy of telomere length as a biomarker of individual condition later in life (Boonekampp et al., 2014; Wood & Young 2019). In line with this, in our study we found that genetic and/or familial effects (represented by nest ID) explained the greatest amount of variance in our cross-sectional telomere length measures at both day 3 and day 11, while telomere length change was better described by DNAm changes and growth rate, factors that are extremely sensitive to the environment.
Contrary to other studies (Stier et al., 2020; Fitzpatrick et al., 2019), we did not find an effect of ambient temperature on telomere length or telomere length changes across development. Temperature has been shown to effect within individual changes in DNAm (Sheldon et al., 2020) and telomere length (Stier et al 2020; Fitzpatrick et al 2019) however, it is unclear if temperature effects these DNA modifications comparably. The temperatures (average daily maximum of 27.6oC) in our study may not have been ‘stressful’ enough to impact telomere dynamics, and indeed temperature has not affected telomere length in studies on other taxa (McLennan et al., 2018; Boonekamp et al, 2020).
In our longitudinal analyses, we also detected a significant, negative relationship between tarsus growth and change in telomere length, such that individuals that grew faster lost more telomere length across the measured period of their early life. This relationship has been detected in previous studies, and could be due to increases in cell division, energy expenditure and oxidative stress associated with increased growth rate (Monaghan & Ozanne, 2018). The relationship between telomere length and tarsus growth occurred independently of temperature in our study. However, previous studies have shown that embryonic growth rate, manipulated by small variations (+-1oC) in incubation temperature, negatively affect telomere length in the laboratory (Stier et al., 2020). We did not detect a relationship between cross-sectional measures of telomere length and tarsus length in our attempt to explore proximate effects within individuals. However, in a study examining the same relationship (between telomere length and tarsus length) at an ultimate level, across generations, in a selection experiment on body size in the house sparrow, a relationship was detected (Ringsby et al., 2015).
In our cross-sectional analysis we detected a positive correlation between telomere length at day 3 and clutch size. We have previously shown that DNAm was affected by brood size during early life of wild zebra finches (Sheldon et al., 2018), and experimental work has also shown that telomeres are generally shorter when brood size is increased during early life (Voillemot et al., 2012; Reichert et al., 2014), potentially as a response to the stress of lower per capita food delivery rates to offspring (Costanzo et al., 2016). The positive relationship detected in our analysis is thus interesting and may reflect higher quality parental care (Bichet et al., 2020), lower rates of nestling/egg mortality, or optimal environmental factors among chicks developing in larger (unmanipulated) clutches (Van Noordwijk & de Jong, 1986).
In our study, telomere length was found to increase as well as decrease from day 3 to day 11 of post-hatch development. Increasing evidence from a range of avian taxa suggests this could be a somewhat common occurrence (Wood & Young, 2019; Kotrschal et al., 2007; Young et al., 2013; Uivari & Madsen, 2009; Haussmann & Mauck, 2008) possibly due to the actions of telomerase (Haussmann, Winkler, Huntington, Nisbet, & Vleck, 2004; 2007; Jaskelioff et al., 2011; Bateson & Nettle, 2017; Hoelzl et al., 2016). The extent to which mechanisms underlying telomere length elongation may weaken the utility of within‐individual telomere length variation and, indeed, the utility of telomere length as a biomarker of biological ageing, is unclear, as telomere repair by telomerase may itself be inhibited by the same factors effecting telomere length (e.g. oxidative damage; Ahmed et al. 2008).
Our measurements of absolute telomere length in wild zebra finches gave telomere lengths similar to those in captive birds measured by the same in-gel TRF method; measurements in the same lab and using the same equipment and protocol for laboratory birds (varying in age from 7 to 120 days) ranged from 10.1 to 16.35kb (Millet & Salmon, pers comm). These data suggest that, in contrast to the laboratory mouse which typically has telomeres several times longer than its wild counterparts possibly due to inbreeding (Manning et al 2002), but this is not the case for the zebra finch.
In conclusion, our study of wild zebra finches has confirmed previously documented effects in captive zebra finches of both growth rate and clutch size on telomere length dynamics during early life. This consistency is important given the differences in resources and variability in the environment of wild compared to captive animals. Most significantly, our study detected a negative relationship between early life telomere and DNAm dynamics, such that individuals with accelerated rates of telomere shortening have associated increases in DNAm levels. The nature of this relationship warrants further investigation since it may shed light on potentially clinically/ecologically distinct aspects of biological age reflected by DNAm and telomere length. It would be valuable for future work to focus on next-generation sequencing techniques (e.g., reduced representative bisulfite sequencing (RRBS)) to compare DNAm and telomere length measures of biological age across different environments and/or to compare potentially functional relevant or regulatory associations between DNAm and telomere length across different environments. Our results provide a base for further investigations to link and/or disentangle the relationship between the early life environment and DNA biomarkers of biological age.