3.1 GB-RFLP marker analysis for lake of origin assignment
As a first step, we chose RFLP markers that are specific for lakes (i.e., allowing to distinguish one population [ingroup] from all other populations [outgroup]) (Fig. 1 and Fig. 2). Although not all of them are (yet) described as separate species, genetic differentiation is very high among most of the lake populations — mainly because the crater lakes are completely isolated (and therefore not permitting gene flow) and because their founder populations were small (30–850 individuals), resulting in strong genetic drift (and thereby more alternatively fixed alleles) (Kautt et al., 2020). Most variants have likely been recruited from the standing genetic variation present in the source populations that was introduced into the crater lakes with the colonizers, as there was too little time for a substantial amount ofde novo mutations to occur — all crater lakes have been colonized less than 5,000 years ago (Kautt et al., 2020; Kautt, Machado-Schiaffino, & Meyer, 2016; Kautt et al., 2018). The two great lakes Managua and Nicaragua are not isolated from each other, but intermittently connected by a river (Rio Tipitapa) resulting in limited ongoing gene flow (based on demographic models from (Kautt et al., 2020) for A. citrinellus approximately 1 in 12000 individuals per generation).
To assess the performance of markers, we calculated their predictive power (see detailed description in Materials and Methods section) based on bootstrapping using allele frequencies from a previous genomic study that was based on 453 re-sequenced genomes (Kautt et al., 2020) as well as GB-RFLP assays of additional samples in this study (that were not part of the resequenced genomes from the (Kautt et al., 2020) study, but morphologically assigned to lakes and species). In line with the discussed potential effect of the described demographic parameters (Kautt et al., 2020), RFLP markers were most powerful for crater lakes. For ten out of the twelve RFLP markers for Crater Lakes Apoyeque (Fig. S3E), Apoyo (Fig. S3G, H), As. Managua (Fig. S3I, J), As. León (Fig. S3K, L), Masaya (Fig. S3M, N), Tiscapa (Fig. S3O, P), and Xiloá (Fig. S3Q) more than 90% of the samples were correctly assigned to the lake (Table 1). The second Apoyeque marker (87%, Fig. S3F) and second Xiloá marker (69%, Fig. S3R) performed less well. For the great lakes Managua (both 62%) and Nicaragua (87% and 69%) the percentage of correctly assigned samples was — as expected — lower (Fig. S3A–D), as many variants are shared between the two great lakes or can be found at least in one of the crater lakes (as they have been colonized from the great lakes).