DISCUSSION
Much of the theoretical and empirical research on the mechanisms that shape variation in body size is based on life-history theory which suggests that individuals allocate their acquired resources between growth, reproduction, and self-maintenance (Stearns, 1989; Zera & Harshman, 2001; Roff & Fairbairn, 2007). In this experimental study we examined the consequences of artificial selection for larger and smaller parental body size and how this influences variation in offspring TL at early age, as well as the associations between TL and recruitment, longevity and reproductive success in two wild house sparrow populations. First, a negative correlation between nestling TL and tarsus length was evident under the artificial selection for both larger and smaller tarsi (Table 3). This link between TL and structural body size suggests that telomere dynamics might mediate a trade-off between investment in early-life growth and long-term somatic maintenance in the wild (Metcalfe & Monaghan 2003; Ringsby et al., 2015; Monaghan & Ozanne, 2018). Artificial selection for larger individuals in thehigh population caused TL to decrease significantly as tarsus length increased during the four years of selection (Fig. 3). Additionally, there was weak evidence that TL tended to increase as tarsus length decreased in the low population (Fig. 4). It is possible that the artificial selection for smaller body size in adults only caused a small change in offspring size because the proportion of additive genetic variance may be lower for small compared to large individuals (Charmantier et al., 2004). Thus, selecting for smaller body size for multiple years, as in our experiment, may accumulate individuals that are smaller than their predicted size due to for instance malnutrition or disease caused by e.g. environmental or parental effects (Angelier et al., 2015).
TL has been suggested as a biomarker monitoring health and stress exposure of organisms (Monaghan, 2014; Pepper et al., 2018; Chatelain et al., 2020), individual phenotypic quality (Bauch et al., 2013; Boonekamp et al., 2013; Le Vaillant et al., 2015), and as an integrated physiological marker of cumulative life‐history costs (Monaghan & Haussmann, 2006). The prevailing negative correlation between TL and body size documented in this study, indicates that TL is influenced by structural growth in free-living birds, which confirms the observation by Ringsby et al. (2015). The artificial selection pressure on body size was accompanied by a reduction in TL that was probably not counteracted within the nestling period by increased investment into telomere maintenance (i.e. canalization). Early-life changes in TL have been hypothesized to influence long-term somatic state (Eisenberg, 2011; Boonekamp et al., 2013; Vedder et al., 2017; Criscuolo et al., 2018a). The enzyme telomerase can elongate telomeres (Blackburn, 1991), but its activity is assumed to be a physiologically costly process (Hatakeyama et al., 2016; Criscuolo et al., 2018b) or with potential increased cancer risk effects (Seluanov et al., 2018). Accordingly, somatic telomerase activity is generally assumed to be repressed in birds (Gomes et al., 2010), though more investigation of this is needed since some somatic telomerase activity has been detected (Haussmann et al., 2007). In common with other non-mammalian vertebrates, birds have nucleated erythrocytes; thus, TLs derived from whole blood samples are mainly measured in erythrocytes, which are normally produced in the bone marrow. Compared to other tetrapods, avian erythrocytes have a relatively short lifespan of 1 month with ~3% being replaced each day (Glomski & Pica, 2016). Early-life erythrocyte TLs in house sparrows have been estimated to 15-20 kbp (Ringsby et al., 2015), which is thought to reflect TLs in hematopoietic stem cells (Vaziri et al., 1994). If 50-100 bp of telomeric DNA are lost with each cell division (Lansdorp, 1995), these early cells would have the potential of 150-400 divisions, many more than is needed for growth and maturation of the adult house sparrow (Sidorov et al., 2009). However, increased oxidative stress associated with acquiring and maintaining a larger body size (Alonso-Alvarez et al., 2007) could accelerate the shortening of telomeres significantly (Reichert & Stier, 2017) providing an explanation for the observed negative association between size and TL (see Fig. 1 and Fig. 5a).
The evolutionary significance of the observed changes in TL induced by the artificial size selection will depend on the heritability of TL, which has been shown to vary considerably among species and populations: Among bird species, TL heritability have been shown to range from 0 to 1 (reviewed in Dugdale & Richardson, 2018), but may be relatively low in house sparrows given the effects of growth and weather observed in this study. We have refrained from estimating heritabilities of TL in the present study, which would be biased by the non-random removal of individuals during the artificial selection events (Steinsland et al., 2014), but future studies may show whether the relationship between size and TL is underpinned by genetic correlations (Monaghan & Ozanne, 2018).
Like most altricial passerines, the growth and survival of house sparrow nestlings depend on early-life conditions such as habitat quality and insect food being supplied by the parents (Anderson, 2006). Larger sparrows have higher juvenile and adult survival (Ringsby et al., 1998; Jensen et al., 2008), and harsh weather during the nestling period increases juvenile mortality (Ringsby et al., 2002). The associations between TL and both body size and the weather proxy (NAO_30) in nestlings (Table 3) suggest that TL is determined by complex and potentially counter-acting effects of growth, nutrition and external factors (Angelier et al., 2015; Nettle et al., 2016). For instance, malnutrition may lead to arrested growth, but also increased oxidative stress and telomere attrition (Nettle et al., 2017). Also, indirect effects of weather conditions may cause foraging stress or maternal stress effects during breeding that negatively affect TL (Haussmann et al., 2012; Mizutani et al. 2013), and direct effects of weather may cause shortening of telomeres, such as thermal stress observed in e.g. brown trout, Salmo trutta (Debes et al., 2016), dark‐eyed juncos,Junco hyemalis (Graham et al., 2019) and greater-eared bats,Myotis myotis (Foley et al., 2020). Thus, generally habitat quality is important, with shorter telomeres in low-quality habitats (Angelier et al., 2013; Watson et al., 2015; Wilbourn et al., 2017). Spurgin et al. (2018) found a positive effect of seasonal insect prey abundance on TL in Seychelles warblers (Acrocephalus sechellensis ) when accounting for a negative correlation with tarsus length. In the same population, the amount of reactive oxygen metabolites in the territorial adult warblers, was shown to be higher in low quality territories than in territories of higher quality (van de Crommenacker et al., 2011), indicating that oxidative stress exposure is involved in telomere shortening (von Zglinicki, 2002). The regional NAO_30 index must be interpreted with respect to local conditions along the northern Norwegian coast but might be a better single proxy for the overall weather conditions by reducing complexity and avoiding problems of model variable selection (Stenseth et al. 2003; Hallett et al. 2004). Thus, a low NAO_30 index, which in our study area corresponds to a combination of low temperatures, strong winds and rainfall during a 30-day interval before TL sampling, was found to significantly reduce TL in nestlings, when correcting for body size (Fig. 5b). This is consistent with studies reporting shorter telomeres because of poor nutrition, competition, or thermoregulation (reviewed in Chatelain et al., 2020).
Natural selection against shorter telomeres may be driven by their negative effect on immune function and longevity (Wilbourn et al. 2018) or reduced cell replicative potential (Blackburn, 1991), while selection against longer telomeres is thought to be due to the high energetic costs associated with increased somatic maintenance (Eisenberg, 2011; Vedder et al., 2017) or increased cancer susceptibility (Aviv et al., 2017; Pepke & Eisenberg, 2020). Several ecological and epidemiological studies have reported a negative association between TL and subsequent mortality risk; mainly in birds (reviewed in Wilbourn et al., 2018) and humans (reviewed in Boonekamp et al. 2013; Wang et al., 2018). This association can be attributed to either the biomarker characteristic of TL reflecting cumulative environmental stressors (Monaghan, 2014; Nettle et al., 2017; Pepper et al., 2018; Angelier et al., 2018) or the direct effect of having short telomeres leading to cellular senescence and certain diseases (Blackburn et al., 2015; Young, 2018). However, this correlation is not universal across tetrapods, with some studies finding no correlation in birds (Boonekamp et al., 2014), mammals (Fairlie et al., 2016), and reptiles (Olsson et al. 2011), or that shorter telomeres correlate with higher survival in birds (Wood & Young, 2019), snakes (Ujvari & Madsen, 2009), and fish (McLennan et al., 2017). Ringsby et al. (2015) suggested that the changes in TL induced by the artificial size selection could underpin a trade-off between body size and lifespan if TL is related to lifespan (Heidinger et al., 2012). In this study, we found little support for an effect of TL on short-term survival (i.e. survival of juveniles until recruitment) after accounting for the positive association between tarsus length and survival (Table 4). Body size is likely to be an important component of juvenile mortality if the mortality is mainly due to extrinsic factors (Wood & Young, 2019; Eastwood et al., 2020), as expected in juvenile house sparrows (Ringsby et al., 1998). The artificial selection increased the range of body sizes across the populations, which may more clearly reveal effects of TL on fitness. The evidence for individuals with either short or long telomeres to have reduced mortality risk through life, controlling for the negative effect of tarsus length, was weak (Fig. 6b). While some correlative studies may have overlooked such weak disruptive selection on TL, such patterns can be confounded by (unmeasured) telomerase expression in somatic cells with high proliferation rates (Klapper et al., 1998; Ujvari & Madsen, 2009; Cerchiara et al., 2017). However, if TL is inversely related to telomere loss later in life (Verhulst et al., 2013; Bauch et al., 2014), measuring TL at a later age may generate the expected positive correlation between survival probability and TL (Wood & Young, 2019). Alternatively, in individuals with short telomeres, TL may be traded off against some unmeasured component of individual quality (Wilson & Nussey, 2010). Yet, when controlling for lifespan, short telomeres were associated with higher recruit production in thehigh population (Table 5). Telomere shortening rates in house sparrows are unknown, but we found little evidence for directional changes in TL across 5-17 days old nestlings (Table 2). The observation in humans that short telomeres are associated with age-dependent degenerative diseases and long telomeres with higher cancer incidence rates (Aviv et al., 2017), suggests the opposite of our findings (i.e. that both short and long TL is associated with higher mortality). However, there is probably little or no constraints on TL imposed by cancer or age-dependent diseases in free-living, short-lived sparrows (Møller et al., 2017). Combined, the adaptive significance of telomere length dynamics may be complex in wild populations with high juvenile mortality and no individuals surviving into very old age (the oldest house sparrow in this study survived until its 6thyear).
There was a weak negative effect of TL on reproductive success within individuals that survived until breeding in the population in which selection for larger size was imposed (high population, Table 5). This might suggest that there are additional negative impacts on TL associated with acquiring an artificially increased body size that deviates from the optimal body size under the prevailing conditions. Such impacts may act through increased competition when siblings are larger (Nettle et al., 2016) and increased oxidative stress during growth (Geiger et al., 2012). This indicates that in the highpopulation, high fitness birds were bigger and therefore had shorter telomeres.
Telomeres were longer in male than in female house sparrows, also when correcting for size (Table 3). We also note that males tended to have higher LRS (Table 5), but only in the high population, where just 6 males managed to reproduce at least one recruit. There were no sex-differences in survival in our study (Table S2.4), which has been suggested to underlie sex-specific telomere dynamics in humans, mice, and sand lizards (reviewed in Barrett & Richardson 2011). In similar Norwegian house sparrow populations, Holand et al. (2016) did not find any general sex-biased mortality or senescence patterns among adults. However, Cleasby et al. (2010) found females to have lower juvenile mortality than males in an English house sparrow population. In birds and mammals, adult mortality appears to be biased towards the heterogametic sex (Liker & Székely, 2005), which may be caused by the potentially unmasked expression of deleterious sex-linked alleles (Trivers, 1985; Hrdličková et al., 2012). In birds, females are the heterogametic sex, but sexual differences in telomere dynamics have only rarely been observed among bird species (but see Foote et al. 2011; Bauch et al., 2020). However, unmeasured sex-specific differences in growth dynamics (in house sparrows, Cleasby et al., 2011) or differential telomere loss under parasite infection (in blue tits,Cyanistes caeruleus , Sudyka et al., 2019) could also generate the observed TL sex differences.
Our study demonstrates the differential impacts of artificial body size selection on early-life TL during the important growth phase. TL was influenced by growth and weather and varied between sexes and populations. Body size was an important determinant of survival, but both short and long telomeres tended to predict lower mortality across the populations after the range of body sizes had been artificially increased. In individuals larger than their optimal size in the wild, TL was reduced, which may have been associated with an increased reproductive output. When selecting for smaller adult body size, we observed a smaller response in fledgling size and TL, and no relationship between TL and reproductive success. Thus, this study shows that the relationship between body size and fitness is complex, with larger body size giving rise to shorter TL. The fitness consequences of this interaction are not simple, and our experimental results suggest that evolution will optimize TL alongside phenotypic parameters.