Introduction

Nuclear microsatellites became popular during the 1990’s as a powerful tool to assess patterns of genetic variation in populations (Allendorf, 2017; Ellegren, 2004). While they are still widely used, the development of Genotyping-by-Sequencing techniques, like RADseq (Baird et al., 2008; Miller, Dunham, Amores, Cresko, & Johnson, 2007) and similar techniques of genome complexity reduction (e.g. ddRAD, bestRAD), coupled with the decreasing costs of massive parallel sequencing, have extended the reach of massive single nucleotide polymorphism (SNP) genotyping to the study of non-model organisms (Baird et al., 2008; Davey et al., 2011; Peterson, Weber, Kay, Fisher, & Hoekstra, 2012; Allendorf, 2017; Andrews, Good, Miller, Luikart, & Hohenlohe, 2016; Putman & Carbone, 2014). This has led to a discussion about the relative benefits of using each type of marker in conservation and evolutionary biology (Allendorf, 2017; Hodel et al., 2017; Morin, Luikart, & Wayne, 2004; Puckett, 2017).
Mutation rates in microsatellites are several orders of magnitude higher than those estimated for SNPs (Dallas, 1992; Ellegren, 2004; Lynch, 2010; Weber & Wong, 1993; Zhang & Hewitt, 2003). Combined with the larger number of possible alleles for a single locus, microsatellites provide immense levels of polymorphism, yielding high statistical power in population genetic inference (Allendorf, 2017; Avise, 2004). Microsatellites are very sensitive to sudden, or recent, demographical processes, and are well suited to detect subtle population structure or recent bottlenecks (Haasl & Payseur, 2011; Luikart & Cornuet, 1998; Pereira, Teixeira, & Velo-Antón, 2018; Putman & Carbone, 2014). However, high polymorphism is usually associated with homoplasy (Garza & Freimer, 1996; Hedrick, 1999; Queney, Ferrand, Weiss, Mougel, & Monnerot, 2001), and poses difficulties in fitting adequate evolutionary models to heterogeneous mutation processes (Di Rienzo et al., 1994; Ellegren, 2004; Valdes, Slatkin, & Freimer, 1993; Weber & Wong, 1993; Webster, Smith, & Ellegren, 2002). This can lead to unreliable estimates of divergence times (Kalinowski, 2002; Queney et al., 2001) and underestimation of genetic differentiation between populations caused by high intra-populational heterozygosity (Hedrick, 1999). Furthermore, microsatellites are not well suited to reconstruct the evolutionary history of lineages or species under certain demographic scenarios, for instance during range expansions, when consecutive founder events and allele surfing processes in newly formed populations inflate genetic differentiation (Pereira et al., 2018). A microsatellite locus contains from four to twelve times more information than a SNP (Liu, Chen, Wang, Oh, & Zhao, 2005). However, current genotyping costs for SNPs are relatively low, so the lower per-locus information of SNPs is largely compensated by the sequencing of thousands of them at a similar cost than the genotyping of a few microsatellites (Hodel et al., 2016; Puckett, 2017). A large number of SNPs and their genome-wide distribution secure a range of mutation rates that can, in principle, provide sufficient information at different evolutionary scales, from recent demographic processes within-species to interspecies phylogenies (DeFaveri, Viitaniemi, Leder, & Merilä, 2013; Petersen et al., 2013).
The different molecular nature of SNPs and microsatellites is expected to impact their resolution power at different evolutionary scales, with microsatellites better reflecting recent demographic processes but rapidly losing resolution above the species level, and SNPs providing less information per locus but securing resolution of demographic processes over a wider evolutionary window (DeFaveri et al., 2013; Estoup, Jarne, & Cornuet, 2002; Haasl & Payseur, 2011). A review of the recent literature shows that thousands of SNPs are generally more powerful in detecting genetic structure than typical microsatellite datasets (Elbers, Clostio, & Taylor, 2017; Hodel et al., 2017; Jeffries et al., 2016; Malenfant, Coltman, & Davis, 2015; McCartney-Melstad, Vu, & Shaffer, 2018; Puckett, 2017; Puckett & Eggert, 2016; Rašić, Filipović, Weeks, & Hoffmann, 2014). The choice of marker (SNPsversus microsatellites) also seems to affect estimates of the proportions of individual ancestries and the inferred optimal number of clusters (Bradbury et al., 2015; Malenfant et al., 2015; Elbers et al., 2017; Bohling et al., 2019). These studies have made important contributions to our understanding of differences in patterns of genetic diversity and structure using both types of markers. However, the lack of comparable datasets, differences in the clustering methods used, and the absence of metrics allowing direct comparisons across marker types limit generalization of these results.
We present an explicit comparison of patterns of genetic structure and diversity based on comparable datasets of microsatellites and SNPs in two amphibian species: the Iberian tree frog, Hyla molleriBedriaga, 1889, and the Western Spadefoot, Pelobates cultripes(Cuvier, 1829). Both are nearly endemic to the Iberian Peninsula (with some populations reaching southern France), and their range-wide phylogeography has been previously investigated based on mitochondrial and microsatellite datasets (Gutiérrez-Rodríguez, Barbosa, & Martínez-Solano, 2017; Sánchez‐Montes, Recuero, Barbosa, & Martínez‐Solano, 2019). These studies linked their contrasting phylogeographic patterns with different demographic histories during the Late Quaternary. Hyla molleri is present in Continental and Atlantic Iberia, and its higher tolerance to colder conditions was hypothesized to account for their inferred demographic stability since the Last-Glacial Maximum (~ 21,000 years ago) (Sánchez‐Montes et al., 2019). In contrast, P. cultripes is a more thermophilous species present in southern and central Iberia, in areas with a Mediterranean influence. This species seems to have experienced important range contractions to southern glacial refugia during colder times in the Pleistocene, resulting in a south-to-north gradient of decreasing genetic diversity (Gutiérrez-Rodríguez et al., 2017). The availability of comprehensive microsatellite datasets and the contrasting demographic histories in a shared geographical area make these two species good study systems for a robust comparative assessment of patterns of genetic diversity and structure obtained with microsatellites and SNPs.