Introduction
Overall, the genotypes and phenotypes of populations undergoing local adaptation are expected to be spatially congruent, i.e. genetic diversity is expected to follow similar differentiation patterns as adaptive phenotypic diversity (Coop et al. 2010, Tigano and Friesen 2016, Fris et al. 2018, Simmonds et al. 2020). This positive correlation is commonly known as ‘isolation by adaptation’, and it is usually studied by estimating ecologically adaptive among-populations divergence as a proxy of the divergent patterns of selection that cause local adaptation (Schluter 2000, Nosil et al. 2008, Funk et al. 2011). In theory, isolation by adaptation should be theoretically detectable in the form of divergence peaks among locally selected loci (or among loci genetically correlated with them), but if populations remain isolated long enough, selectively neutral loci can also become differentiated, which makes the detection of outlier loci empirically challenging (Krohn et al. 2018, Llanos-Garrido et al. 2019). However, discordances may also appear, whereby two populations may be genetically undifferentiated while showing evident phenotypic divergence (Moody et al. 2015, Palmer and Kronforst 2015, Shaner et al. 2015). This can happen, despite the existence of processes that blur overall genomic divergence (e.g. gene flow, recent divergence), only when natural (or sexual; Yang et al. 2018) selection is strong enough to overcome such processes, favoring the presence of locally divergent regions with variants under selection within an otherwise undifferentiated overall genomic background (Burri 2017, Wang et al. 2019).
While numerous studies have dealt with locally adapted populations occupying different environments or isolated by ecological barriers (Rosenblum 2006, Orsini et al 2013, Sexton et al. 2014, Zhao et al. 2020), only a few have focused on the lack of correlation between isolation and adaptation (Feder et al. 2013). Moreover, it has been suggested that there is a significant publication bias against such studies (Krohn et al. 2018). This possible under-representation of apparent negative results may lead to biased estimates of how frequently local adaptation occurs without isolation, and this is precisely the reason why replicated studies on species or populations with different degrees of isolation are needed (Feder et al. 2013, Talla et al. 2017, Sendell-Price et al. 2020). In fact, if studies that do not find any effect of environmental gradients on genetic differentiation are rarely published, examples of incipient ecological speciation and/or isolation by environment may artificially be deemed frequent or even widespread (Hendry 2009, Shafer and Wolf 2013, Sexton et al. 2014).
The aim of this study is to elucidate the patterns of genetic differentiation that underlie phenotypic divergence between two populations of the lacertid lizard Psammodromus algirus separated by a 600-700 m altitudinal gradient. This gradient is associated with a large number of habitat differences, including forest type (deciduous vs. perennial) or average annual rainfall (1170 vs. 438 mm; see the Methods section for a detailed explanation of these habitat differences). The analysis of mitochondrial DNA sequences has shown that these populations present very little genetic differentiation (Verdú-Ricoy et al. 2010, Díaz et al. 2017), even though they differ in a wide variety of adaptive phenotypic characteristics, including many life history traits (Iraeta et al. 2006, 2010, 2011, and 2013, Llanos-Garrido et al. 2017; Table 1). The evidence for such adaptations is based on previous studies that have shown, through reciprocal transplant and common garden experiments, that these adaptive phenotypic differences are not sustained solely by environmental effects (Iraeta et al. 2006, 2013). Therefore, there must be a genetic basis to determine such phenotypic differences between these apparently undifferentiated populations, even if such genetic basis does not lead to an environmentally based isolation. To define the degree of genetic differentiation between the two populations, we used a GBS genomic scan based on 73,291 SNPs that allowed us to analyze the genetic structure and distance between them. In addition, we used a Bayesian method of detection of SNPs with genetic distances between populations greater than expected given the degree of background genomic differentiation (i.e. FST based outlier test; Bayescan: Foll and Gaggiotti 2008). With this approach, we tried to define polymorphisms possibly associated with the patterns of divergent selection that promote the observed adaptive differences (Bonhomme et al 2010). Thus, if local adaptation has led to isolated populations, we expect to find a genome-wide differentiation standard that could hamper the detection of divergence peaks at adaptive loci. Alternatively, we might find an undifferentiated genomic background where such peaks should be easy to detect, at least in theory.
We used an approach in which all genetic variants are used to infer the basal level of genomic differentiation, defining outliers as SNPs with a greater degree of divergence and with allelic frequencies deviated from the expected under neutral selection (Lewontin and Krakauer 1975). The reason why comparing differentiation peaks with adjacent regions should facilitate this task is that the degree of differentiation is heterogeneously distributed throughout the genome (Campagna et al 2015). Therefore, detecting a divergence peak in a specially conserved region is challenging using delocalized genetic variation, which may find such degree of divergence even below the background genomic differentiation, but which may actually be very divergent within its genomic location (Lawson y Petren, 2017). This scenario is especially common in the coding regions where the genetic basis of phenotypic diversity is located, so that approaches with delocalized SNPs do not usually respond to what is the genetic basis of a given phenotype, but describe the general patterns of genetic differentiation that lay behind the process of local adaptation (e.g. Tigano et al. 2017, Llanos-Garrido et al. 2019). Thus, while uncovering the specific genetic basis behind already published adaptations between these populations would be specially challenging given the methodology we used, it is still possible to elucidate whether such local adaptation is accompanied by any degree of isolation (or genome-wide differentiation) or not (Krohn et al. 2018).