DISCUSSION
The evolutionary response to selection on telomere length depends on the additive genetic variance of TL and the strength and sign of any genetic correlations with other traits under selection. Dugdale and Richardson (2018) criticized past quantitative genetic studies of TL on the main grounds that 1) they applied basic regression analyses that did not consider environmental effects impacting TL and as a consequence of that, additive genetic effects may have been overestimated in previous studies; 2) TL changes with age, complicating the fact that parents and offspring are often sampled at different ages; and 3) sample sizes were too small to provide enough power to separate genetic and environmental effects using animal models. Here, we have accommodated this critique by 1) using mixed-effect animal models to partition genetic and environmental effects; 2) measuring early-life TL in both offspring and parents at the same time point in life (as around 11 days old fledglings); and 3) collect TL data from more than 3300 individuals across 4 populations, which represent a considerably larger sample size than those of previous wild animal studies.
We found that around 4% of the variation in early-life TL in house sparrows at the end of the nestling growth period was determined by additive genetic variation. The relatively small additive genetic variance and large year variance in early-life TL appears to be in accordance with the effects of relative growth and weather conditions on TL in similar sparrow populations (Pepke et al., 2021,submitted ). Similarly small but significant heritabilities of TL have been reported using animal models for e.g. nestling collared flycatchers, Ficedula albicollis(h2 =0.09, Voillemot et al., 2012), Seychelles warblers (h2 =0.03-0.08, Sparks et al., 2021) and adult greater mouse-eared bats, Myotis myotis(h2 =0.01-0.06, Foley et al., 2020), in which TL correlates with several weather variables. These studies also documented considerable year effects on TL (Foley et al., 2020; Sparks et al., 2021) similar to studies finding no heritability of TL in white‐throated dippers (Becker et al., 2015) and European badgers (Meles meles , van Lieshout et al., 2021). In comparison, studies based on parent-offspring regression have often found higher TL heritabilities in e.g. king penguins (h2 =0.2, Reichert et al., 2015), jackdaws (Coloeus monedula ,h2 =0.72, Bauch et al., 2019), and sand lizards (Lacerta agilis , h2 =0.5-1.2, Olsson et al., 2011). The heritability of TL in house sparrows is comparable to that of many life-history traits and considerably lower than many morphological traits (e.g. Mousseau & Roff, 1987; Visscher, Hill, & Wray, 2008), which may suggest that TL is under strong selection in the wild (Voillemot et al., 2012) or that there are considerable non-additive genetic or environmental influences on early-life TL. Curiously, Pepke et al. (2021, submitted ) reported indications of weak non-linear or negative associations between TL and various measures of fitness (survival and reproductive success) in house sparrows, suggesting that the environmentally pliant TL dynamics of these relatively fast-lived birds may be very different from several other bird species (reviewed in Wilbourn et al., 2018). In other species, positive associations between early-life TL and survival have been documented (Wilbourn et al., 2018), which may translate into an increased lifetime reproductive success (Eastwood et al., 2019; Sudyka, 2019; Bichet et al., 2020).
A considerable proportion of the phenotypic variance in TL could be attributed to brood and parental effects (Fig. 3). However, we did not find evidence that parental effects were transmitted through a parental age at conception effect (Fig. S2.1). Paternal age effects, which has been observed in several other species (Eisenberg & Kuzawa, 2018), may not manifest in these house sparrows because the mean age at reproduction was low (around 2 years). Parent-offspring regressions (Fig. S2.2) and parental genetic effects models (Table S2.2) suggested a stronger component of maternal heritability of TL, which is similar to the inheritance pattern found in several bird species (Horn et al., 2011; Asghar et al., 2015; Becker et al., 2015; Reichert et al., 2015) and some studies on humans (Broer et al., 2013). Maternal effects on offspring TL are expected to be strongest in early-life (Wolf, Brodie Iii, Cheverud, Moore, & Wade, 1998) and could act through e.g. yolk-deposited components in the egg (Criscuolo, Torres, Zahn, & Williams, 2020; Stier et al., 2020a) or post-laying through maternal care behavior (e.g., incubation and feeding rate, Stier, Metcalfe, & Monaghan, 2020b; Viblanc et al., 2020). Since TL is a heritable trait, a positive maternal effect on offspring TL may be expected to increase the expected rate of adaptive evolution of TL (Wolf et al., 1998; Räsänen & Kruuk, 2007). Parental and environmental effects documented in other studies (Monaghan & Metcalfe, 2019) suggest that some of the variation in TL may be inherited through epigenetic carry-over effects (Bauch et al., 2019; Eisenberg, 2019) that are not resolved by comparing early-life TLs. Thus, such effects may be more important in shaping nestling TL loss, rather than early-life TL (Heidinger et al., 2016). However, TL maintenance are at present not well known within house sparrows (Vangorder-Braid et al., 2021).
There was evidence for additive genetic variance in the tarsus length of sparrow nestlings, but the heritability estimate (h2 =0.076, Table 2) was considerably smaller than those of adult house sparrows in a larger sample of populations in the same area (Jensen et al., 2008; Araya-Ajoy et al., 2019) and other avian species (Merilä & Sheldon, 2001). However, there was a large brood effect on nestling tarsus length suggesting common environmental effects within broods (e.g. Potti & Merino, 1994). For instance, variation in clutch size, seasonal differences in food availability, weather conditions (Ringsby, Sæther, Tufto, Jensen, & Solberg, 2002), and provisioning rates by parents (Ringsby, Berge, Sæther, & Jensen, 2009) may induce intra-clutch competition and variation in the degree to which nestlings are able to achieve their adult tarsus lengths at fledging (Naef-Daenzer & Keller, 1999; Metcalfe & Monaghan, 2001). Furthermore, measurement error is probably higher for the incompletely ossified nestling tarsi, which are covered by a soft fleshy skin tissue that contributes to the measured length.
Individuals with shorter tarsi (a proxy for structural size, Araya-Ajoy et al., 2019) were found to have longer telomeres, although the effect of tarsus length on TL was small and there was considerable variation in TL for a given size (Fig. 1). This confirms previous observations of a prevailing negative correlation between body size and TL within house sparrows (Ringsby et al., 2015; Pepke et al., 2021, submitted ) and other species (Monaghan & Ozanne, 2018). We did not find evidence for a significant negative genetic correlation between TL and tarsus length (Table 2). Instead, the negative phenotypic association between TL and tarsus length may be induced by common environmental effects that affects both traits in opposite directions. The lack of a genetic correlation between TL, tarsus length or body condition could also be attributed to selection acting simultaneously on some correlated, unmeasured trait (Merilä, Sheldon, & Kruuk, 2001). Both with and without controlling for the effect of tarsus length on TL, our GWAS on TL identified several genes involved in skeletal development, cellular growth and differentiation that may regulate body growth or size (e.g. GHRHR, Tmem120b, LMOD3, GH, POU1F1, SHCBP1, and FGFR2, Table 4, S2.5, and S2.7), which could, however, suggest some genetic basis of the negative correlation between TL and size. For instance, several growth factors were downregulated in telomerase deficient mouse bone marrow stromal stem cells (Saeed & Iqtedar, 2015) suggesting that short telomeres or telomere loss could also be a constraint on proliferation potential. Thus, because several of the genes that may regulate TL during early development appear to also be involved in cell proliferation or morphogenesis, such genes may have co-evolved.
None of the genes highlighted in our analysis have previously been linked to TL in GWA studies (reviewed in the introduction). However, theDrosophila orthologue of ZBED1 (dDREF) has been linked to telomere maintenance in Drosophila flies (Tue et al., 2017), but telomere biology in this taxon is very different from most other eukaryotes and does not involve telomerase (Casacuberta, 2017). Yet, the dDREF/ZBED1 is important for cell proliferation in bothDrosophila (Matsukage, Hirose, Yoo, & Yamaguchi, 2008), bats (Rhinolophus ferrumequinum , Xiao et al., 2016) and human cancer cells (Jiang et al., 2018, but see Hansen et al., 2018). Several of the identified candidate genes (ZBED1, AQP1, SHCBP1, CDCA4, ARL6IP5, UBA3, RNF34, RHOF, ANAPC5, and FGFR2) are involved in cell proliferation and apoptosis during which TL and telomerase activity invariably play an important role (Greider, 1998; Masutomi et al., 2003). The RHOF gene product functions cooperatively with CDC42 and Rac to organize the actin cytoskeleton (Ellis & Mellor, 2000). While the latter complex participates in the control of telomerase activity in human cancer cells (Yeh, Pan, & Wang, 2005), any direct link between RHOF and TL remains unexplored. CDC42 is activated by FGD4 (Chen et al., 2004), which was found within a major locus affecting TL in humans (Vasa-Nicotera et al., 2005). SNPa108592 was found near several genes involved in cell proliferation, differentiation, immune response, and ubiquitination (Table 4). Ubiquitination regulates several shelterin components and telomerase activity (Peuscher & Jacobs, 2012; Yalçin, Selenz, & Jacobs, 2017). The closest gene, ORAI1 (43 kb), the keeper of the gates of calcium ions (Homer, 1924), is crucial for lymphocyte activation and immune response (Feske et al., 2006). Although not linked to ORAI1 mutations, calcium ion levels can modulate telomerase activity (reviewed in Farfariello, Iamshanova, Germain, Fliniaux, & Prevarskaya, 2015).
We identified a particularly interesting gene associated with TL, AQP1. The AQP1 channel not only conducts water across cell membranes, but also hydrogen peroxide, a major reactive oxygen species (ROS, Tamma et al., 2018), and nitric oxide (Herrera, Hong Nancy, & Garvin Jeffrey, 2006), which is an important regulator of oxidative stress (Pierini & Bryan, 2015) and a weak oxidant itself (Radi, 2018). Furthermore, increased availability of nitric oxide may activate telomerase and thereby prevent replicative senescence (in endothelial cells, Vasa, Breitschopf, Zeiher Andreas, & Dimmeler, 2000). Enhanced oxidative stress associated with endothelial cell senescence may also be mediated by AQP1-regulated nitric oxide flow (Tamma et al., 2018; Chen et al., 2020). In AQP1 knocked-out erythrocytes (where TL was measured) cell lifespan was shortened (Mathai et al., 1996) and angiogenesis is inhibited in AQP1 knocked-out chicken embryos (Camerino et al., 2006) and mice (Saadoun et al., 2005). Telomeres are particularly sensitive to ROS and shorten due to oxidative stress during growth (von Zglinicki, 2002; Reichert & Stier, 2017). For instance, Kim, Noguera, Morales, and Velando (2011) found a negative genetic correlation between growth and resistance to oxidative stress in yellow-legged gull (Larus michahellis ) chicks, which could be mediated by TL (see also Smith, Nager, & Costantini, 2016). Another candidate gene, OXR1, 76 kb from SNPa450086, has a well-described antioxidant function (Volkert et al., 2000; Oliver et al., 2011) and is upregulated in senescent human cells (Zhang et al., 2018). Knockdown of OXR1 increases ROS production and ultimately induces apoptosis (Oliver et al., 2011; Zhang et al., 2018), which could be due to telomere crisis.
Over-expression of AQP1 has been associated with several types of cancer (Verkman, Hara-Chikuma, & Papadopoulos, 2008), suppression of apoptosis (Yamazato et al., 2018) and may play an important role in tumor biology (Saadoun et al., 2005; Tomita et al., 2017). Other candidate genes including GHRHR, SAMD5, SHCBP1 (Tao et al., 2013), GH (Boguszewski & Boguszewski, 2019), and OXR1 (Yang et al., 2015) are also involved in tumorigenesis. Cancer prevalence is not well-studied in wildlife (Pesavento, Agnew, Keel, & Woolard, 2018), but tumors have been documented in house sparrows (Møller, Erritzøe, & Soler, 2017). Long telomeres or increased telomerase activity may increase the risk of acquiring an oncogenic mutation before cell proliferation ceases due to telomere crisis (Aviv, Anderson, & Shay, 2017; Pepke & Eisenberg, 2021). However, long telomeres also increase immune function required to combat cancers (Helby, Nordestgaard, Benfield, & Bojesen, 2017) and short telomeres can result in chromosomal instability leading to some types of cancer (Ma et al., 2011; Aviv et al., 2017). This TL paradox is not yet resolved (Eisenberg & Kuzawa, 2018). However, genes affecting both TL and cancer risk (Tacutu, Budovsky, Yanai, & Fraifeld, 2011; Jones et al., 2012) could underlie the antagonistic pleiotropy of trade-offs between long telomeres in early-life (with potential benefits to growth, reproduction, and other oxidative stress inducing processes) and later-life cancer mortality (Tian et al., 2018). For instance, Vedder et al. (2021) found a significant genetic correlation between TL and lifespan in wild common terns. Cancer is often viewed as a senescence-related pathology (Lemaître et al., 2020). However, the absence of cancer in early-life should not lead us to conclude that a somatic and potentially fitness-related cost is not paid to maintain that status (Thomas et al., 2018).
We have shown that TL is a heritable, polygenic trait with considerable environmental variation and a maternal inheritance component in a wild passerine. It is, however, important that future studies attempt to confirm the candidate genes identified here as associated with TL in other wild populations. Even though the additive genetic component was small, selection on variation in TL may produce evolutionary change in TL over time in wild populations. The large component of variation in early-life TL caused by annual environmental stochasticity suggests that this will generate heterogeneity in TL among cohorts. Although we did not find a negative genetic correlation underlying the negative phenotypic correlation between TL and body size, we may hypothesize that selection for larger nestling size, which may enhance survival until recruitment (Ringsby, Sæther, & Solberg, 1998), will be associated with selection for shorter early-life TL due to non-genetic mechanisms, which can ultimately influence lifespan or reproductive success.