Early‐life telomeres are influenced by environments acting at multiple temporal and spatial scales

An individual's telomere length early in life may reflect or contribute to key life‐history processes sensitive to environmental variation. Yet, the relative importance of genetic and environmental factors in shaping early‐life telomere length is not well understood as it requires samples collected from multiple generations with known developmental histories. We used a confirmed pedigree and conducted an animal model analysis of telomere lengths obtained from nestling house sparrows (Passer domesticus) sampled over a span of 22 years. We found significant additive genetic variation for early‐life telomere length, but it comprised a small proportion (9%) of the total biological variation. Three sources of environmental variation were important: among cohorts, among‐breeding attempts within years, and among nestmates. The magnitude of variation among breeding attempts and among nestmates also differed by cohort, suggesting that interactive effects of environmental factors across time or spatial scales were important, yet we were unable to identify the specific causes of these interactions. The mean amount of precipitation during the breeding season positively predicted telomere length, but neither weather during a given breeding attempt nor date in the breeding season contributed to an offspring's telomere length. At the level of individual nestlings, offspring sex, size and mass at 10 days of age also did not predict telomere length. Environmental effects appear especially important in shaping early‐life telomere length in some species, and more focus on how environmental factors that interact across scales may help to explain some of the variation observed among studies.

model analysis of telomere lengths obtained from nestling house sparrows (Passer domesticus) sampled over a span of 22 years.We found significant additive genetic variation for early-life telomere length, but it comprised a small proportion (9%) of the total biological variation.Three sources of environmental variation were important: among cohorts, among-breeding attempts within years, and among nestmates.The magnitude of variation among breeding attempts and among nestmates also differed by cohort, suggesting that interactive effects of environmental factors across time or spatial scales were important, yet we were unable to identify the specific causes of these interactions.The mean amount of precipitation during the breeding season positively predicted telomere length, but neither weather during a given breeding attempt nor date in the breeding season contributed to an offspring's telomere length.
At the level of individual nestlings, offspring sex, size and mass at 10 days of age also did not predict telomere length.Environmental effects appear especially important in shaping early-life telomere length in some species, and more focus on how environmental factors that interact across scales may help to explain some of the variation observed among studies.

K E Y W O R D S
environmental effects, heritability, life history, parental age Consistent with these ideas, telomere length often positively predicts longevity (Wilbourn et al., 2018;Bichet et al., 2020), and this relationship is often apparent even during early life (Eastwood et al., 2019;Heidinger et al., 2012Heidinger et al., , 2021)).Some of the variation in early-life telomere length is due to genetic factors (Chik et al., 2022;Dugdale & Richardson, 2018).
However, exposure to a variety of environmental stressors can also accelerate the pace of telomere loss and reduce early-life telomere length (Boonekamp et al., 2014;Epel et al., 2004;Haussmann et al., 2012;Herborn et al., 2014;Marasco et al., 2021).These effects can even extend across generations as both parental age and stress exposure at the time of offspring production can influence early-life telomeres (Eisenberg et al., 2012;Entringer et al., 2011;Haussmann & Heidinger, 2015;Heidinger & Young, 2020;Young et al., 2022).Knowledge about the relative importance of genetic, parental and environmental factors in shaping variation in early-life telomere length will be essential for understanding the relationship between telomeres and life-history strategies, how variation in telomere length is maintained, and the potential for telomere length and loss to evolve.Yet, information about this topic remains limited because acquiring it involves long-term studies with known pedigrees and detailed developmental histories.Most prior studies have relied on approaches (i.e.parent-offspring regression) that (1) are unable to separate parental effects from other sources of variation, (2) have measured parent and offspring telomeres at different life stages (i.e.measured the parental generation as adults and the offspring generation as juveniles) despite the fact that telomere length often changes with age or (3) have used relatively small sample sizes subject to vagaries of sampling and measurement errors (Chik et al., 2022;Dugdale & Richardson, 2018).Larger sample sizes employing animal model statistical approaches (Henderson et al., 1959;Wilson et al., 2010) may account for additional sources of variance in natural systems because they can include additional environmental effects, make use of more information about genetic relatedness and can improve understanding of the genetic structure of focal traits (Chik et al., 2022).This approach also has the advantage of being able to partition hierarchical sources of environmental variation, allowing a more detailed examination of the specific environmental factors that may operate at some levels (e.g.among locations or years) but not others.
Despite these advantages, animal model analysis of telomere lengths has also uncovered considerable variation in estimates of heritability.Telomeres in jackdaws (Corvus monedula) exhibited a heritability of 0.74 (Bauch et al., 2022), common terns (Sterna hirundo) were estimated at 0.46 (Vedder et al., 2022), and Soay sheep (Ovis aries) lambs had a value of 0.28 (Froy et al., 2021) and Seychelles warblers (Acrocephalus sechellensis) and European badgers (Meles meles) values of 0.03 (Sparks et al., 2021) and<0.0001 (van Lieshout et al., 2021), respectively.Although methodological differences in measuring telomeres can account for some of this variation (Chik et al., 2022), variation in heritability can also arise due to variation in the mechanism of inheritance and/or sensitivity to environmental conditions, but the details of both possibilities remain largely unknown.Thus, additional studies that delve more deeply into these different sources of variance in telomeres are critical for understanding both the causes and consequences of telomere variability.
To address these uncertainties, we analysed samples from a long-term study of free-living house sparrows (Passer domesticus).
In this system, we have previously reported that early-life telomere length positively predicts lifetime reproductive success; longevity in females, but not in males, was associated with early-life telomere length (Heidinger et al., 2021).In addition, we have also found that parental and offspring stress exposures can interact to influence offspring telomere length (Young et al., 2022).Here, we analysed the sources of variation in telomere length in nestling house sparrows using marked individuals with known parents.We leveraged the fact that all individuals suitable for pedigree analysis were sampled at the same age and typically from nest sites for which we had considerable information about environmental conditions experienced during development, such as the age of parents, timing and weather conditions for each nesting attempt, and the number of offspring in the nest and the size of sampled individuals.We had three general goals for these analyses: 1. Assess the magnitude of genetic variation in telomere length and if it was related to other factors, such as the sex of the offspring or age of the parent at fertilization.

Identify general sources of environmental variation and probe
these further to uncover the specific mechanisms of environmental effects on telomere length.

Test for potential relationships between nestling size and off-
spring telomere length, given potential physiological effects of faster growth on telomeres.

| Study population
This study made use of a library of blood samples collected from 1993 to 2014 from a population of house sparrows located at the University of Kentucky's North Farm Agricultural Research Station.Westneat et al. (2002Westneat et al. ( , 2009) ) and Heidinger et al. (2021) provide details about this population and the standard methods used in the field to mark individuals and monitor their reproduction and the collection of and storage of blood samples.Briefly, all nestlings were given uniquely numbered metal bands at 10 days of age, and both independent juveniles and adults were given unique combinations of coloured-plastic bands during trapping sessions throughout the year.The present data consist of the families of breeding adults who had hatched in focal nest boxes, were banded as nestlings and had a blood sample taken when they were 10 days old.While we had samples from any nesting attempt of these focal birds that reached 10 days of age in one of our boxes, we often targeted a subset of these.If the focal adult produced only one brood, we analysed the DNA of all offspring.If they had multiple broods but only within a single season, we analysed the brood with the most offspring.If they had broods across more than one season, we sampled the largest brood in each season.Some additional broods were analysed for other reasons, and these were also included in the analysis.We identified breeding partners of target individuals, and if a blood sample was available, we measured the telomere length from that sample.
Most such partners were captured as adults, so their exact age was unknown, but we knew relative age from the date of capture.This sampling regime resulted in 1591 individuals with telomere lengths of which 1500 had been sampled at 10 days of age.

| Telomere measures
Some of the telomere measures used in this analysis came from the dataset used in Heidinger et al. (2021).Subjects in that analysis had all been sampled at 10 days of age and resampled at least once, but many did not breed.The present analysis included breeders only and also added individuals sampled at 10 days of age that were not recaptured and resampled.The birds used in the Heidinger et al. (2021) dataset were randomly assigned to assays that were carried out by A. Kucera, whereas the remaining samples were randomly assigned to assays and processed by R. Young.All assays used the same reference sample (described below).
Telomeres were measured in whole blood samples, a highly replicative tissue that can be nondestructively sampled and is well suited for telomere analyses, especially in birds which have nucleated red blood cells.We extracted DNA using Qiagen DNeasy Blood and Tissue DNA Extraction Spin Column Kits and following the manufacturer's instructions.DNA concentration was quantified using a Nanodrop 8000 Spectrophotometer (Thermo Scientific).DNA quality was assured using electrophoresis on a 2% agarose gel.
We used real-time quantitative polymerase chain reaction (qPCR) on an Mx3000P (Stratagene) to measure relative telomere length.The methods followed those of Criscuolo et al. (2009) adapted for house sparrows (Young et al., 2022).To measure relative telomere lengths (T/S ratios) for each sample, we ran separate qPCR reactions for telomeres and the single copy control gene, Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), on different plates.The suitability of GAPDH as a control gene in this species has previously been verified using a melt curve analysis (Young et al., 2022).All the samples were run in duplicate and randomly distributed across plates.
Individuals with longer telomeres cross this threshold more quickly than individuals with shorter telomeres, relative to the same house sparrow reference sample used on all plates.This same reference sample was also used to create a 5-point standard curve (40, 20, 10, 5 and 2.5 ng) to ensure that all samples fell within the bounds of the standard curve and to calculate reaction efficiencies.In total, we ran 53 plates, and in all cases, the efficiencies were between the recommended 85%-115% (mean ± SEM: telomere: 96.5 ± 0.85% and GAPDH 100 ± 0.99%).This assay produces highly repeatable T/S ratios (ICC: 0.86-0.88)when samples are run in random well locations across plates (Heidinger et al., 2021;Young et al., 2022).

| Parentage confirmation and pedigree construction
We confirmed maternity and paternity using genotypes obtained from a multiplex PCR of five highly polymorphic microsatellite loci adapted from well-established protocols (Dawson et al., 2012;Stewart et al., 2006).Generally, parents and offspring were organized by family, amplified on the same plate, and products analysed in the same round on an ABI 3730.If adults had multiple broods across several years or switched mates, they may have been run on a different plate than some of their offspring.In some cases, subjects were analysed 2-3 times on different plates and in all cases, their allele scores were within two base pairs of each other.Adults were excluded as parents if 2 or more loci diverged by more than 1 base pair.In some cases, adults were excluded from being the parent of a whole brood.This sometimes led to the identification of alternative adults that did match, and if these were not among our target individuals, that brood was excluded from further analysis.Only offspring that were genetically related to a target adult were included in the dataset.If the focal adult was a male and the offspring was extra-pair, it was excluded from our analysis.If the focal adult was a female, then any of her offspring sired by an extra-pair male were included in the analysis, but we did not identify the true sire in these cases.The parentage information allowed us to create a pedigree of individuals with known telomere lengths and their relatedness to other individuals.
The resulting pedigree of 1629 individuals had 1591 telomere measures (both nestlings and adults), 1500 of which were from individuals sampled at 10 days of age.This included 211 sires and 228 mothers, no cases of inbreeding, and 7 tiers (Table S1, Figure S1) with one lineage having six generations.The dataset covered 428 nesting attempts with at least one nestling sampled.Most mothers had only one nesting attempt (N = 152), but two were represented by offspring from 10 attempts.Mothers were typically part of only one pair (N = 215), meaning that they only had one male partner, but 44 had two or more partners with one having 6.Since extra-pair offspring were not assigned to a sire, these numbers reflect social pairings that produced within-pair offspring.We determined the sex of all nestlings using PCR of extracted DNA, following procedures described in Westneat et al. (2002).

| Goal 1: Variance partitioning
Our initial goal was to partition the variance in telomere lengths among sources given the structure of the data.The initial animal model included all adults with a telomere measure, regardless of when they were sampled.We repeated the analysis restricting the set of adult telomere measures to those that were sampled at 10 days of age.These models were fitted with a set of random effects and no fixed effects.We included the relatedness matrix as a key term for assessing covariances among individuals by relatedness, the assay identity as a measure of lab artefacts, year (cohort) to assess effects of year-to-year environmental variation (note: prior analyses showed no effect of latency between sampling and analysis on telomere length, Heidinger et al., 2021), breeding attempt identity to assess all environmental effects common to a breeding attempt (e.g.date in the season, weather conditions during embryo and nestling development, size of the brood), nesting location (labelled as 'barn' since nest boxes were clustered on the sides of farm buildings) to capture local environmental effects common to offspring reared in the same location and pair identity to capture consistent joint influences of the male and female associated with breeding attempts.We did not include maternal identity alone, as males can provide extensive care during incubation and chick rearing, and most adult females in the dataset were members of only one pair and had only one breeding attempt, so the combination of pairD, motherID and attemptID (or attemptID, motherID and sireID) would have overspecified the model.In equation form, our initial model was: We reanalysed the initial model to include only data from individuals sampled at 10 days of age.Variance partitioning models were fitted twice in the R computing environment (R Core Team, 2019), first using brms (Bürkner, 2017) with two chains, default priors, 10,000 iterations with a burnin of 1000 and a sampling every five iterations.
We used this to assess how well the model performed and get initial insight into the magnitude of variance components.The models took a long time to run in brms, so once we confirmed that models were fitting well (no divergent chains and rhats less than 1.1), we reran models in the program MCMCglmm (Hadfield, 2010) using weakly informative priors (V = 1, nu = 0.002, Lemoine, 2019), 1.1 × 10 5 iterations, a burnin of 10,000 and sampling every 50 iterations.
We then analysed the data in more detail to assess three possible influences on genetic variance in telomeres.First, we asked if parent age at the time of fertilization influenced telomere length (sex-specific parent age entered as two fixed effects).Second, we asked if the genetic variance in telomere length was a function of parent age via separate analysis of sire-specific (1-9) or mother-specific (1-6) age as a linear random slope in the 'animal' random effect (Class et al., 2019).In these models, we dropped any random effects that could not be distinguished from 0 and replaced pair identity with either mother or sire identity depending on which sex was in the random slope term.Thus, for mother-specific effects, the model was where fixed included both mother's and sire's age.We allowed the model to estimate covariances between the animal intercept and slope.These models were fitted in brms with a linear slope, Cauchy priors (0,2), 10,000 iterations with a burnin of 1000, and a sample every five iterations.
Finally, we assessed the role of offspring sex, both as a fixed effect to repeat a test for any sex differences in telomere length (Bauch et al., 2020, Heidinger et al., 2021, Le Pepke, Kvalnes, Ranke, et al., 2022) and to assess sex-specific heritability.The former was tested by adding a fixed effect of offspring sex to the models described below that explored specific environmental factors.For sex-specific heritability, we fitted a bivariate model with the telomere length in male and female nestlings as two responses to investigate differences in heritability (Chik et al., 2022;Olsson et al., 2011) and estimate genetic correlations between the sexes.
We also included assay identity, year and attempt identity as random effects and modelled the covariance between male and female for each of these as well.Models were fitted in brms with each of three prior types for variance/covariances: default, V = diag(2), nu = 1.002, and V = diag(2) × (0.002/1.002), nu = 1.002.Results were affected only slightly.All models had 10,000 iterations, burnin of 1000 and a sampling of every 5 iterations.

| Goal 2: Influence of environmental effects
We examined types of environmental effects in more detail using a sequenced approach.
Step 1 was to assess the magnitude of variance explained by the 'environmental' variance components in the initial animal model described above.Our plan was to then explore specific environmental variables in more detail by adding fixed effects that varied most at the requisite level.For example, we expected from Le Pepke, Kvalnes, Lundregan, et al. (2022) that cohort (year) might explain some variation.If so, then yearly differences in weather might be relevant, so we gathered year-specific mean temperature and precipitation in either the 'spring' months (February and March) preceding each breeding season or 'summer' months through the period of breeding (April-August).Similarly, any among-attempt variation might be influenced by date in the season in which attempts were started, brood size at hatch, or the specific mean precipitation that occurred during the 25 days after the first egg was laid in that breeding attempt.Because temperature for a specific nesting attempt is highly correlated with date, and previous analyses of both variables revealed that date was a better variable to include for some traits (Westneat et al., 2009), we left attempt-specific temperature out of the model.Because these models focussed on environmental sources of variance, we omitted the pedigree and included year and attempt identity as random effects.Each of these investigations was modelled in brms using default priors, 10,000 iterations, a burnin of 1000 and a sample every five iterations.
We also explored the potential for interactive effects among environmental factors in two ways.To gather general evidence of interactions, we paired down the mixed model to just the informative and relevant random effects (omitting PairID, Barn and the pedigree).We then split either the breeding attempt random term or the residual via grouping by a higher order random effect.For example, we calculated among-breeding attempt variance and among-nestling-within-attempt (residual) variance in telomere for each year.We also assessed whether among-or within-attempt variance differed among parental ages by converting parental age to a random effect (lumping older ages into a 5+ category) and splitting among-attempt and residual variance by age, with each parent tested in separate models.These models were fitted in a frequentist framework in SAS 9.4 (SAS, 2015) Proc Mixed given the ease with which SAS coding allows this (see Appendix S1).
The magnitude of improved fit was assessed using a likelihood ratio test against the base model without group-specific variances.
We also tested a set of two-way interactions as fixed effects that we reasoned were likely to explain variance at the appropriate level.
For instance, if variation among attempts was important, then date by brood size seemed a likely influence given prior work showing that date affects clutch size (Westneat et al., 2009) and nestling growth (Mock et al., 2009) and brood size influences telomeres in a North Dakota population (Young et al., 2022).Similarly, at the residual level, brood size or parent age by the sex of the nestling might be important.We constructed these models after running the initial random effect models.Because these proposed fixed effects differed dramatically in measurement scale, we standardized ((x-mean)/ SD) all values of all variables and analysed them in a mixed model with a reduced set of random effects (omitting Barn, PairID and the pedigree) in brms with default priors, 10,000 iterations, a burnin of 1000 and sampling every five iterations.2022) found that telomere length was negatively correlated with offspring size with age controlled, suggesting that higher growth led to shorter telomeres.

| Goal 3: Phenotypic and genetic covariance between telomeres and body size
We used two bivariate analyses to explore these relationships in our dataset.One of the bivariate equations had telomere length as the response, including assay identity as a random effect and the important random effects from the analyses described previously (year and attempt identity) along with the relatedness matrix.The other equation had either nestling tarsus length, a standard measure of size or nestling mass as the response.These equations both included nestling age as a fixed effect since there was some variation in the age at which nestlings were measured, and year, attempt identity and the relatedness matrix as random effects.We set the models to extract covariances for each random effect, including the relatedness matrix to assess environmental and genetic correlations and the residual covariance.These models were fitted using brms using default priors, 10,000 iterations, a burnin of 1000 and sampling every 5 iterations.As a check, we reran these models with Cholesky priors (lkj(2)) as recommended for correlations among random effects in the set_prior notes for brms.We found no substantive differences in results.

| Variance partitioning
We partitioned the variance in 1591 telomere length measures taken from 10-day-old nestling house sparrows (Figure 1) and their parents.Measurement error (the assay unit of plates) accounted for approximately 16% of the variance.This source was accounted for in all analyses but omitted from estimates of heritability.We detected a low but important level of additive genetic variance (Figure 1).The heritability estimated from the model had a posterior mean of 0.09 and a credible interval of 0.002 to 0.211, indicating that very low values were relatively common in the posterior distribution (see also Figure S2).Some potential environmental sources (barn and pair identity) contributed low variance with a peak of estimates very close to 0 (Figure 1).However, both year and nest identity contributed important components of variation, accounting together for 22% of the total natural variance.Residual variance-the variance in telomere length among nestlings within a nest, accounted for 64% of the total natural variance.We found no difference in the estimates if we restricted the analysis to just individuals that had a telomere measured at 10 days of age.

| Genetic variance
We assessed whether heritability was contingent on parent age.
Our data revealed no evidence of an influence of parent age on telomere length directly (Table 1).We also obtained no evidence that genetic variance in telomere length depended on mother's age, but we did find support for genetic variance to increase with sire's age (Table 1).
We also tested whether genetic variation in telomeres differed by sex of the offspring.Bivariate analysis revealed no evidence of a difference in additive genetic variance between male (0.043

| Environmental variance
We assessed several sources of environmental variance in the initial animal model.Two sources, the general location of nest sites (Barn) and the combination of specific male and female (Pair ID), appear to explain little of the variance in telomere length (Figure 1) with posterior distributions piling up near 0 (Figure S3).The analysis of parent age also revealed little evidence of paternal or maternal effects on telomere length (Table 1).We thus did not investigate these sources further.
Two sources of variance, year and attempt identity, received strong support as important (Figure 1) and explained 6% and 16% of total variance, respectively.Both exhibited posterior distributions centred around values away from 0 (Figure S3).We investigated these sources in more detail.First, we found evidence for some potential interactive effects of the environment.The among-attempt variance itself varied significantly among years (Table S2, LRT test, chi-square = 100.8,df = 21, p < .0001)as did the variance among nestmates within nests (Table S2, LRT test, chi-square = 137.3,df = 21, p < .0001).By contrast, our general assessment of whether telomere variation among attempts or among nestmates varied by either maternal or paternal age classes revealed no evidence of interactive effects with parental age (Table S3).These results suggested adding interactions between the 4-year-level variables of weather (precipitation and temperature in spring and summer) and the attempt level variable of brood size given its effects on telomeres in another population (Young et al., 2022).
Our model of specific environmental effects revealed few identified causes of telomere variation.Higher mean temperature and precipitation over the prebreeding season (February and March) or over the breeding period (April-August) all tended to increase telomere length (Table 2), but only summer precipitation received sufficiently strong support to be considered important.
None of the interaction terms and neither date nor brood size received support as being influential, and these results were not sensitive to the presence of which interaction terms were in the model (Table S4).

F I G U R E 1
Median and credible intervals of posterior estimates of variance components in house sparrow telomere lengths partitioned among seven sources.Shown are two analyses from MCMCglmm (with weakly informative priors): (1) using data that includes telomere lengths of parents collected at any age ('All', blue) or ( 2) data on telomere lengths collected only from birds at 10 days of age (orange).'Assay' refers to the plate units on which the sample was analysed, 'Animal' is the additive genetic variance estimated from the pedigree.Inset graph shows the per cent of total variance from each source using the 'All' data.

TA B L E 1
Posterior medians ± SE and 95% credible intervals for parameter estimates of animal model analysis of telomere lengths from nestling house sparrows.Note: Results are from two models, identical to the base animal model (Figure 1) except that parent age (mother or sire) was added as a fixed effect and a random slope in the animal part of the model, Barn and Pair ID were omitted, and mother or sire identity was added.Random effect terms are given in standard deviations and bolded items received support as important.

| Residual variance
The dominant source of variance in telomere length in the initial model was among individual nestlings (Figure 1); residual variance composed 64% of the total variance in the base model.Some of the residual variance could arise from either additive or nonadditive genetic variance, but some forms of environmental effects could also be acting at the level of individual nestlings within broods.One variable with the potential to capture either genetic or environmental processes is the sex of the focal offspring.We found no difference between male and female offspring in the length of their telomeres (Table 2).

| Genetic and environmental covariances between telomere length and tarsus or mass
Bivariate animal models of telomere length and nestling morphology (tarsus or mass) revealed no correlations at any level of variance (Table 2), providing no evidence that offspring growth rate or condition might explain variation of telomere length among nestlings, among nest attempts or among years.Both tarsus length and nestling mass exhibited substantial variation among years and among nests (Table 3), accounting for 47% and 53% of variance, respectively.Both also exhibited weak evidence of some genetic variation although heritability for both was low (tarsus: 0.08 ± 0.08; mass: 0.08 ± 0.07) and quite similar to that of telomere length.

| DISCUSS ION
The length of an individual's telomeres early in life may reflect and/or mediate key life-history processes (Haussmann & Heidinger, 2015;Monaghan & Haussmann, 2006;Young, 2018), but the factors that contribute to this variation remain poorly understood.To address this, we used samples collected as part of a 22-year study on freeliving house sparrows to examine the sources of variation that contribute to telomere length during postnatal development (10 days of age).A small proportion (9%) of the variation in early-life telomere length was due to additive genetic variation.We identified three major sources of environmental variation across time and spatial scales that made up a larger proportion of the variation: among years (cohorts), among-breeding attempts within adult pairs and among nestlings within the same attempt.One implication of the results

TA B L E 2
Posterior means ± SE of fixed effect predictors and the remaining unexplained levels of environmental variance (in standard deviations of standardized scores) in telomere length among house sparrow nestlings using a mixed model with a reduced set of random effects and standardized telomere length and predictor variables.

TA B L E 3
Results from bivariate mixed effects models producing estimated environmental and genetic covariances between telomere length and either tarsus length (a measure of size) or mass of nestling house sparrows.reported here is that telomere length has the potential to evolve.
Another implication is that large environmental variation in telomeres is consistent with the idea that telomeres may be a biomarker of conditions experienced during early life.We also found that the magnitude of variation among-breeding attempts and among nestlings within a breeding attempt varied among years.Nested patterns of variance such as these suggest interactive effects among environmental variables.Although we were unable to uncover the specific combinations of environmental factors that contribute to these patterns, these findings may help to explain some of the conflicting patterns of the relative importance of genetic, parental and environmental effects in early-life telomeres reported among studies (Chik et al., 2022;Dugdale & Richardson, 2018).
Our finding of relatively low heritability of telomeres is consistent with values reported in another long-term population of house sparrows that live in an island population in the species' native range (Le Pepke, Kvalnes, Lundregan, et al., 2022).House sparrows are thus similar to several other species of songbird (Sparks et al., 2022;Voillemot et al., 2012) and mammals (van Lieshout et al., 2021) in having low heritability of telomeres.We also found that genetic variation did not differ between the sexes either in how it was transmitted from adults or expressed in the nestlings, and we found that the genetic variation increases with parent age, particularly in older fathers.
Both methodological and biological causes could contribute to relatively low heritability of telomeres.Two methodological issues could occur as a result of sampling.First, most measures of telomeres are from somatic tissues (red blood cells in our case), yet the telomeres that are present in the DNA received by offspring at fertilization are from gametes.Telomeres may differ across cell types but differences in telomere length appear to correlate in some species and in some tissues at some points during development (Demanelis et al., 2020;Reichert et al., 2013).Data on correlations between blood telomeres and gamete telomeres are scarce, but in adult house sparrows from North Dakota, the correlation between sperm and blood telomere length was 0.53 (Kucera, 2018), suggesting that blood can be a proxy of sperm telomeres in this species.
To our knowledge, there has been no study of how blood reflects telomeres in oocytes in any species.
A second methodological issue concerns the mechanism of heritability and the timing of sampling within the lives of the subjects.
Our understanding of the mechanism(s) by which telomere length is inherited is poor (Dugdale & Richardson, 2018;Haussmann & Heidinger, 2015).Two extreme possibilities exist.First, telomeres could be determined by genes elsewhere in the genome, such that offspring inherit a genetic programme for setting telomere length at any given age.For example, the action of telomerase early in life is known to influence telomere length early in development (Gomes et al., 2011;Young, 2018).Heritability in telomere length might then be due to genetic variation in the target telomere size of this developmental mechanism.In our study, 94% of the individuals in our pedigree were sampled at 10 days of age, and this sampling would provide the most comparable phenotypes within the pedigree.Major The alternative mechanism is that since telomeres are part of the DNA present at fertilization, offspring could inherit the telomeres directly from those in the gametes that fused to form the zygote.
That is, offspring directly acquire the telomeres of their parents at the time of fertilization.If this is the main genetic influence on telomere length for the rest of the offspring's life, then offspring inherit a trait that may have an acquired value.The fact that environmental effects in parents, such as parental age and stress exposure, influence offspring telomeres (Haussmann & Heidinger, 2015;Young et al., 2022) is suggestive of this mechanism.If offspring do inherit the telomere a parent has at the time of fertilization, then our estimates of parental telomeres, taken most often when they were 10 days old, may be less relevant for estimates of heritability and could explain the low heritability we observed.However, we detected no effect of including parents who were sampled as adults in the analysis, and we found no effect of parental identity or of parent age on offspring telomeres.Furthermore, offspring telomeres resembled their father's 10-day telomeres better when their father was older, a pattern opposite from what would be expected if offspring inherited telomeres of their parents at the time of fertilization and not at 10 days of age.This issue of the exact timing of sampling to assess heritability and whether it appropriately captures the underlying biology of telomere inheritance perhaps adds to other factors known to contribute to the wide variation in heritability estimates across studies (Chik et al., 2022;Dugdale & Richardson, 2018).
Heritability could be low due to strong selection (Fisher, 1930;Hunt et al., 2007).In this population, we have previously reported that early-life telomere length is positively related to lifetime reproductive success in females, and thus, it likely experiences some level of selection (Heidinger et al., 2021).However, the same relationship is not seen in males, and the relationship between telomere length and survival to breeding age is not known for this population.Thus, it is unlikely that strong selection explains low heritability.Conversely, many recent studies of the heritability of traits have focussed instead on variation due to environmental causes (e.g.Houle, 1992, Merilä & Sheldon, 2000, Teplitsky et al., 2009, Wheelwright et al., 2014).
Variation in heritability of telomeres among studies is large; our study adds to an array of differing results.For example, house sparrows differ dramatically from some other birds (Atema et al., 2015;Bauch et al., 2022;Morosinotto et al., 2022;Vedder et al., 2022), mammals (humans;Dugdale & Richardson, 2018) and insects (Boonekamp et al., 2022) in which heritability exceeds 0.4.That telomeres in some species are sensitive to environmental variation and others appear less so has several general consequences.Telomeres, a component of the genome, thus behave more like a phenotypic trait with multiple expressions, subjected to phenotypic plasticity.
The exact form of this plasticity, how adaptive it may be and what functional consequence it may have, remain to be understood in any organism.Our study narrows some possibilities but raises additional questions about environmental variation.
First, we were surprised to find limited effects of parental identity.Parents build nests, mothers stock the eggs with a variety of resources and signalling molecules (Giraudeau & Ducatez, 2016;Schwabl, 1993), and both parents care for eggs and nestlings.Parents thus have the potential to influence many aspects of nestling phenotypes.Evidence from a North Dakota population of house sparrows indicated cross-generational effects on telomeres of parents experiencing stressors indicating that parents can modify offspring telomere dynamics (Young et al., 2022).In the present study, we included pair identity or the identity of male or female alone in our analyses, and these should have captured consistent differences in parental effects among adults across breeding attempts.None were important sources of variance.That breeding attempt within families did explain significant variance leaves open the possibility for plastic parental traits to influence telomeres.Stressors that affect parents but vary among attempts would be one possible mechanism (Young et al., 2022).
The large amount of environmental variance we uncovered came from three undefined sources: among cohorts (years), among nesting attempts within years and families and among nestmates.We investigated several known environmental factors at each of these levels and surprisingly found few predictors.The most well-supported environmental factor we could identify was that telomere lengths were longer in years in which there was a higher level of daily precipitation over the breeding season.Le Pepke, Kvalnes, Ranke, et al. (2022) also found the effects of weather, although they used the North Atlantic Oscillation Index (suitable for their northern European and island populations) and found a curvilinear relationship.Weather has been implicated in several other studies (Axelsson et al., 2020;Foley et al., 2020;Seeker et al., 2021) A key result of our analysis was that telomere variation has  found no links between nestling size and telomere length at 10 days of age.We tested this at multiple levels; at the individual nestling level within breeding attempts, among attempts and among years, but we found no evidence for a relationship at any of those.It seems unlikely that our methods differed from the Norwegian groups' sufficiently to cause this difference in results, and there may be a long list of unknown but possible differences between the two populations of sparrows.This list includes differences in the development of telomeres posthatching (Bennett et al., 2022) and the timing of growth.The Kentucky population has a much longer breeding season with higher temperatures than Norway (Westneat et al., 2014).
More detailed studies of the development of telomeres may be needed to further understand these differences.
Finally, we note as well that the largest component of variation in telomere length was among nestlings within a breeding attempt.
Our assessments of within-assay measurement variance indicate this within-attempt variance is biological.We identified no specific factors that would explain this variation although this component of variation differed among years, suggesting interactive effects of variables differing by year with factors differing within broods, such as the extent of hatching asynchrony or the level of food competition between nestlings, both of which could vary by year.Young et al. (2022) showed that at least some of the within-nest variation was due to interactions between parent environments and individual nestling conditions, such as relative rank early on within the brood.Kinnard and Westneat (2009) also found that early inequities in the nest have long-lasting effects on both morphological and physiological attributes.Our data on nestling attributes for this analysis are limited to one measure of body size and mass at the time of sampling.Neither of those were correlated with telomere lengths at the residual level, so if relative status within a brood was important for telomeres in the Kentucky population, it did not produce a signature using these measures.We assessed the possibility of parental age influencing within-brood variation but found no evidence for the age of either parent affecting mean telomere length or the variation within broods.We therefore suspect again that interactions among levels of variance, such as environmental factors affecting parents, which then drives parental effects on individual offspring, to be a viable and intriguing hypothesis worthy of more focussed study.
In sum, we found that telomere length at 10 days of age varies due to both heritable genetic variation and a wide range of possible environmental influences that vary across years, among-breeding attempts within years and among nestlings within a breeding attempt.
The latter two levels of variance also vary in magnitude among years, which strongly suggests interactive effects of environmental factors, but we could identify few specific causes.Telomere lengths can be correlated with fitness; the circumstances in house sparrows and in many other species (Tobler et al., 2022) suggest telomeres are linked with important processes that translate environmental circumstances into biological performance.Both those processes and the environmental circumstances appear to be complex and require considerably more detailed investigation.
differences in environmental exposure due to age should therefore have been controlled.If so, our results indicate the heritability of that programme is low, perhaps due to considerable variation in the influence of the environment between fertilization and the age of 10 days.

a
nested structure.That is, the among-breeding attempt and among-nestmate variation both varied among years.This pattern could arise from two sources: nested structures of key environmental factors (e.g. more variation in food supply in some years than in others), or interactive effects of environmental factors that also vary at these different scales.One possibility, suggested by the results ofYoung et al. (2022), is that parents may do different things at different nests and that it is parental plasticity that affects offspring telomeres.Young et al. (2022) found that experimental stressors applied to parents interacted with natural stressors experienced by offspring to affect telomere length.Both kinds of stressors could be variable on different time scales (e.g.among years), producing effects only on individual nesting attempts or individual offspring that differ among years.Interactions among environmental variables that have different patterns of variation or that cross-generational boundaries might produce patterns of variation that are difficult to explain.For example, differences among years in the variation in predator abundance within years would produce among-year variation in the among-brood variance.While we found indirect evidence of environmental interactions, our explicit tests of interactions between date and brood size, brood size and temperature or precipitation revealed little evidence that these were important.Heterogeneous variances plus the large amount of unexplained variation among-breeding attempts indicate that not only are the processes whereby season or weather influence telomeres unknown but also additional factors exist that remain to be identified.