INTRODUCTION
Dispersal is a complex life history trait that influences demographic and genetic processes, hence dispersal plays an important role in the eco-evolutionary dynamics of geographically structured populations (Legrand et al., 2017; Van Dyck & Baguette, 2005). Dispersal affects evolutionary processes by leading to gene flow that increases genetic variation within and affects the genetic structure of populations (Holsinger & Weir, 2009). Variation in dispersal also impacts local population sizes and densities, habitat use and (re)colonization in fragmented populations, and, ultimately, these effects can influence the viability and persistence of populations and species (Clobert, Baguette, Benton, & Bullock, 2012; Saastamoinen, 2008). Because of its fundamental importance, it is essential to understand whether and how fast dispersal rates can evolve (Ronce, 2007; Saastamoinen et al., 2018). Estimating the heritable genetic component and examining the genetic architecture of dispersal is needed to understand causes of variation in the dispersal phenotype and for predicting its adaptive evolutionary potential (Orr, 2005; Zera & Brisson, 2012).
Dispersal-related traits (such as wing shape, locomotion performance or speed) have previously been shown to be heritable in birds and insects with an average heritability (h2 ) of 0.35 (Saastamoinen et al., 2018). Another meta-analysis revealed that the average heritability of movement behavior over 15 different studies (including both dispersal and migration) was found to be 0.46 (Dochtermann, Schwab, Anderson Berdal, Dalos, & Royauté, 2019). However, estimating the heritability of dispersal or dispersal syndromes (i.e. traits associated with dispersal) is a challenging task due to the complexity of the dispersal event itself. Dispersal propensity may be affected not only in one or more of the dispersal stages (departure, transfer and settlement) but also by dispersal-related phenotypic traits and their interactions with the environment (Bowler & Benton, 2005; Ronce, 2007; Saastamoinen et al., 2018). Due to the need for accurate identification of dispersers and resident individuals, which relies on the quality and extent of mark-recapture data over sufficiently large geographic areas to cover normal dispersal distances, estimating heritability of dispersal and dispersal related traits is challenging, but such estimates have been obtained in birds and insects more often than any other taxa (Brown, Phillips, & Shine, 2014; McGaugh, Schwanz, Bowden, Gonzalez, & Janzen, 2010; Saastamoinen et al., 2018; Waser & Jones, 1989; Zera & Brisson, 2012).
Additive genetic variance (σA2 ) and the proportion of the phenotypic variance explained byσA2 (i.e. narrow sense heritability; h2 ), reflect the heritable genetic component of a trait and determine the potential rate of any evolutionary response to selection acting on the trait (Lande, 1979). A specific linear mixed effects model called the “animal model” uses information on the relatedness of individuals with phenotypic data and is widely used to estimate additive genetic variances of phenotypic traits of domestic animals as well as wild populations of many species (Kruuk, 2004; Lynch & Walsh, 1998; Wilson et al., 2010). However, most animal models assume that the populations under study are genetically homogeneous, which is often not the case in natural populations, and this assumption may therefore introduce biases in estimates (Muff, Niskanen, Saatoglu, Jensen, & Keller, 2019; Wolak & Reid, 2017). A recent extension called genetic groups animal model (GGAM) enables us to account for genetic admixture within and between populations and allows estimating heterogeneous and population-specific mean genetic values (basic GGAM; Wolak & Reid, 2017) and additive genetic variances (extended GGAM; Aase, Jensen, & Muff, 2022; Muff et al., 2019).
Genome wide association studies (GWAS) are commonly performed to investigate underlying genetics of phenotypic traits and to detect Quantitative Trait Loci (QTL; Korte & Farlow, 2013). In relation to dispersal, it has for instance been shown that a foraging gene inDrosophila melanogaster is linked with locomotion behavior, causing adults with the dominant ‘rover’ allele to have longer dispersal distances (Edelsparre, Vesterberg, Lim, Anwari, & Fitzpatrick, 2014). Similarly, the Pgi gene in the Glanville fritillary butterfly (Melitaea cinxia ) codes for a metabolic enzyme associated with cellular energetics (Mattila & Hanski, 2014), and has an allelic variant that causes a higher flight metabolic rate and dispersal propensity (Haag, Saastamoinen, Marden, & Hanski, 2005; Niitepõald et al., 2009; Niitepõld, Mattila, Harrison, & Hanski, 2011). However, research on genetic variation in dispersal in natural populations, as well as other complex life history traits, indicates that underlying genetic variation is often caused by many genes of small effect (i.e. are polygenic; Saastamoinen et al., 2018; Tiffin & Ross-Ibarra, 2014; Zera & Brisson, 2012). Polygenic traits may covary with several different fitness traits and are often influenced by multiple environmental factors and can hence show complex evolutionary trajectories (Remington, 2015).
Studies on the genetic architecture of dispersal pave the road to a better understanding of the ecological and evolutionary consequences of dispersal and movement in fragmented populations and species invasions, and hence the capacity to spread and ultimately survive in the face of environmental change (Saastamoinen et al., 2018). In the present study, we used successful natal dispersal between islands as the phenotypic trait in order to investigate the heritable genetic basis of dispersal in an insular metapopulation of a small passerine bird, the house sparrow (Passer domesticus ). Previous studies have shown spatial differences in dispersal rates related to island habitat type (Ranke et al., 2021; Saatoglu et al., 2021). Initially we therefore assumed that the heritable genetic variation in dispersal was similar across islands but allowed the mean genetic values of dispersal to differ between island habitat types, and used a basic genetic groups animal model (basic GGAM) to estimate the σA2 of dispersal probability. Subsequently, we used an extended GGAM to allow for different σA2 of dispersal for the two habitat types. Lastly, we used GWAS to identify genes that might explain variation between individuals in dispersal probability. To achieve these goals, we used high-quality information on dispersal and high-density genome-wide single nucleotide polymorphism (SNP) genotype data from over 2500 individuals in a long-term study of house sparrows on islands in a metapopulation off the coast of northern Norway, where relatedness is available through a genetically determined multi-generational pedigree (Lundregan et al., 2018; Niskanen et al., 2020; Saatoglu et al., 2021).