DNA extraction, sequencing and data clustering
A total of 189 individuals from 33 populations (Fig. 1; Table S1) of the four species of C. gr. laevigata (plus one sample each ofC. mairei –sister group–, C. distans , C. boryanaand C. borbonica as outgroup) were sampled for Restriction-site-Associated DNA sequencing (RAD-seq). This technique has been proven as a useful High Throughput Sequencing (HTS) approach for investigating recent evolutionary processes in non-model organisms, including different scenarios of speciation among closely related species (Otero et al., 2022 and references therein), in which traditional Sanger-sequenced nuclear and plastid markers have been unable to solve phylogenetic relationships (Escudero et al., 2013b). A special effort was conducted for C. laevigata in order to obtain a wide intraspecific genomic sampling representative of chromosome number variation within and across populations. Thus, 130 individuals from 20 populations of C. laevigata , most of which with available chromosome counts, were included. DNA was extracted using a DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA). RAD-seq libraries preparation using restriction enzyme PstI were followed by sonication and barcoding performed by Floragenex (Eugene, Portland, OR, USA) following Bairdet al. (2008).
Raw data were demultiplexed based on the code for each sample and adapters were removed from the reads using ipyrad v0.7.24 (Eaton & Overcast, 2020) using a reference genome (Carex scoparia ; Planta et al., 2022). We used the matrix resulting from a clustering threshold of 90% similarity and a minimum sequencing depth of 6 sequences per locus for within-sample clustering (c90m6). We compared this matrix to others with a clustering of 95% and sequencing depth of 6 and 11, but no significant difference was found. Then we discarded those samples with higher missing data. Finally, we reanalyzed the data according to the c90m6 parameters and a minimum number of samples with data at a given locus of 90 out of 156 (ca. 60%). We discarded those samples with >95% of missing data for the phylogeny construction, so a final matrix of 156 accessions was retained (three outgroup taxa, one sister taxa and 152 Carex gr. laevigata samples). The phylogenetic relationships of the retained samples were reconstructed using IQTREE v1.6.11 software (Nguyen et al. , 2015; Hoanget al ., 2018).