DISCUSSION
In the present study, we investigated the genetic architecture of dispersal in an insular metapopulation of house sparrows by estimating additive genetic and environmental variance components complemented by a genome-wide association analysis. Our house sparrow metapopulation is particularly interesting for such a study, as previous publications have shown that birds differ in dispersal probability depending on whether they originate from a farm or non-farm habitat type of island (Pärn, Ringsby, Jensen, & Sæther, 2012; Ranke et al., 2021; Saatoglu et al., 2021). We found that in this metapopulation, heritable genetic variation explained approximately 10% of the variation in individual dispersal probability. However, by using novel statistical methods that allow for mean and variance in heritable genetic variation to differ between genetic groups, we revealed that the farm and non-farm habitats differ in both mean breeding values and additive genetic variances for dispersal. Specifically, although phenotypic dispersal probabilities are higher in the non-farm habitat, the mean breeding value and the additive genetic variance (as well as the heritability) for dispersal was higher in the farm habitat than in the non-farm habitat.
It is challenging to obtain high quality data on dispersal because the study system needs to be sufficiently large to cover normal dispersal distances of the organisms, resident and dispersing individuals need to be individually recognizable, and to estimate either the heritable genetic component of dispersal or its fitness consequences, cross-generational data that include information also on the descendants of dispersers and residents are necessary (Cayuela et al., 2018; Holyoak, Casagrandi, Nathan, Revilla, & Spiegel, 2008; Millon, Lambin, Devillard, & Schaub, 2019). Despite these challenges, the genetic basis of dispersal phenotype and dispersal-related traits had been researched on occasion even in vertebrates in the wild, using either parent-offspring regressions or animal models. Parameter estimates derived from animal models, which account for all kinds of genetic relatives in the models and are regarded as potentially less biased by for example non-genetic environment effects than parent-offspring regressions (Kruuk, Slate, & Wilson, 2008), tend to report more modest magnitudes of σA2 andh2 than parent-offspring regressions (Saastamoinen et al., 2018). Furthermore, animal model-based heritable genetic parameters of dispersal propensity for Aves class members of natural populations were reported as 0.024 – 7.608 and 0.36 – 0.95 forσA2 andh2 , respectively (Supplementary Table S3). Moreover, the σA2 andh2 of dispersal were estimated to be 0.290 and 0.280, respectively, in an experimental cane toad study (Bufo marinus/Rhinella marina ; Phillips, Brown, & Shine, 2010), and in a semi-natural common lizards (Zootoca vivipara ) study as 1.148 and 0.170, respectively (San‐Jose et al., 2023). Hence, heritability estimates of dispersal for other species are usually higher than those we documented in house sparrows, although our estimate ofσA2 for the farm habitat was slightly higher than in the metapopulation as a whole, with approximately 12% of the phenotypic variance in dispersal explained by heritable genetic variation in this habitat type (Table 1). In combination, the relatively few studies on the heritable genetic basis for dispersal propensity that exist from natural vertebrate populations (Supplementary Table S3) suggest that this key life-history trait has the capacity for adaptive evolution on ecological time-scales if any selection is acting on it, but that its rate of micro-evolution may differ somewhat between species and even between populations within the same species. Indeed, in another study of the same house sparrow metapopulation we have shown that immigrants have higher fitness than resident individuals (as estimated by annual production of recruiting offspring and number of recruiting offspring produced over the life span; Saatoglu et al., in preparation). Consequently, dispersal rates are expected to increase across generations in our metapopulation.
Recently, gene mapping studies using a GWAS approach have been able to identify genes underlying phenotypic variation in various heritable life-history and fitness-related traits even in natural vertebrate populations (e.g. Barson et al., 2015; Husby et al., 2015; Johnston et al., 2011; Lawson & Petren, 2017; Lundregan et al., 2018; Tietgen et al., 2021). Here, we have revealed a single region on chromosome 15 that was linked with dispersal trait in the house sparrow metapopulation and it has been found that this marker was closest to the ADORA2A receptor gene. This receptor gene is located near the UPB1 gene both in the house sparrow genome (Elgvin et al., 2017) and the zebra finch genome (Warren et al., 2010). ADORA2A is involved in glycogenolysis (i.e. release of the glucose into the bloodstream; see González-Benítez et al., 2002), thus ADORA2A may influence energy dynamics. Glucose metabolism has been shown to affect dispersal rate in the Glanville fritillary butterfly (Melitaea cinxia) for which the Pgi gene explains variation in dispersal rate, and is involved in breakdown of glucose to produce ATP (Hanski et al., 2017; Niitepõld & Saastamoinen, 2017). Interestingly, another function of ADORA2A is to increase intracellular cAMP levels which are not only important in metabolism and wakefulness but are also an important aspect of the circadian regulatory mechanism that has direct influence on the clock phase (O’Neill & Reddy, 2012). A recent study on a semi-natural population of common lizards (Zootoca vivipara ) showed that expression of circadian clock genes differed between dispersers and residents. However, ADORA2A or UPB1 were not among the dispersal-related genes identified in these species (San‐Jose et al., 2023). Moreover, clock-linked genes may also influence migratory timing in the American kestrel (Falco sparverius ; Bossu, Heath, Kaltenecker, Helm, & Ruegg, 2022). Hence, although few studies exist and the functional relationship between putative genes and dispersal in most cases needs to be explored further, there appears to be some evidence that genes related to (flight) energy metabolism and circadian rhythms are related to the individual dispersal processes. However, it seems clear that dispersal propensity at least in our house sparrow metapopulation is a polygenic trait with a complex basis that involves both genes and environmental effects.
Dispersal in our house sparrow metapopulation occurs during the fledged juvenile phase, in the autumn before the juveniles’ first winter (Pärn, Jensen, Ringsby, & Sæther, 2009; Ranke et al., 2021; Saatoglu et al., 2021). Thus, it seems likely that environmental conditions related to the population density, weather or various habitat characteristics that offspring experience during development may also affect the propensity to disperse. Accordingly, we have previously shown in the same study system that dispersal rates were higher when springs were warmer, breeding started early, and when total population sizes at the end of the breeding season were higher (Pärn et al., 2012). Condition-dependent dispersal probabilities that are influenced by environmental conditions such as population density, prenatal/postnatal environmental conditions and/or physiological traits underlying the movement capacity have also been documented in many other studies of vertebrates (Boualit et al., 2019; Leon, Banks, Beck, & Heinsohn, 2022; Maag, Cozzi, Clutton-Brock, & Ozgul, 2018; Massot, Clobert, Lorenzon, & Rossi, 2002; Matthysen, 2005; McCaslin, Caughlin, & Heath, 2020; Messier, Garant, Bergeron, & Réale, 2012; Saastamoinen et al., 2018; Walls, Kenward, & Holloway, 2005; Wu & Seebacher, 2022). Interestingly, the relationships between dispersal and environmental conditions in our house sparrow metapopulation mentioned above actually differed between habitat types: dispersal rates were positively related to spring temperature, onset of breeding and total population density in non-farm habitat islands, while dispersal was independent of these environmental conditions in farm habitat islands (Pärn et al., 2012). Despite higher average dispersal rates in the non-farm habitat than in the farm habitat (Ranke et al., 2021; Saatoglu et al., 2021), the results in the current study that show lower estimated mean breeding values and lower additive genetic variances for dispersal in the non-farm habitat than the farm habitat (Table 1), suggest that when individuals make their dispersal decisions, environmental components are more influential than heritable genetic effects in the non-farm habitat.
Moreover, the contrasting results for the two habitat types such as lower mean breeding values but higher dispersal probabilities in the non-farm habitat type and differences between islands within each habitat type in dispersal probabilities (Supplementary Figure S3) suggest that there may be a genotype by environment interaction (GxE ) for dispersal in our house sparrow metapopulation. If such an interaction exists, one would expect that birds with genomes originating from non-farm islands will respond differently to the farm environment with respect to their dispersal probabilities and vice versa. Birds that disperse between habitat types and the descendants of such inter-habitat dispersers (see Supplementary Figure S1) can not only be used to separate additive genetic effects from environmental causes of observed differences in dispersal probabilities such as we have done here (Table 1; Figure 3), they also allow for examining GxE in dispersal. Testing whether there is a GxE for dispersal in our study metapopulation, and investigating any causes and consequences of such an interaction is however outside the scope of the current paper and should be examined in a future study.
Here we have shown that there is a habitat-dependent heritable basis for dispersal, which is an important life history trait because of its close connection with spatio-temporal ecological and evolutionary and dynamics across geographically structured populations (Clobert et al., 2012; Saastamoinen et al., 2018). The ability of evolutionary ecologists to partition a natural population’s phenotypic variance in key traits into a heritable genetic component and environmental components of variation advanced when animal models were introduced to the field approximately two decades ago (Kruuk, 2004; Wilson et al., 2010). Here we have exploited the recent development of genetic groups animal models that allow for exploring and quantifying spatial variation in heritable genetic variation (Aase et al., 2022; Muff et al., 2019) and show that the rate of any adaptive evolutionary change in dispersal may differ across space in a fragmented population. In a rapidly changing world, where many populations become increasingly fragmented and range shifts may be necessary to avoid extinction, quantifying such spatial variation and understanding its consequences for ecological and evolutionary processes is likely to be of increasing importance.