Glossopetalon inhabits arid regions in the American west and northern Mexico on limestone substrates. The genus comprises four species: G. clokeyi ; G. pungens ; G. texense ; and G. spinescens . Three of the species are narrow endemics. The fourth, G. spinescens , is a widespread species with six recognized varieties. All six varieties are intricately branched shrubs that have been difficult to identify due to a lack of clearly delineating morphological characters. Characters typically used to differentiate the varieties of G. spinescens, such as stem coloration, leaf blade size, and presence of stipules, are highly variable within and among populations. A custom protocol of double digest restriction-site associated DNA sequencing (ddRAD) was used to resolve the phylogeny of Glossopetalon and address if population genetic data analyses (such as STRUCTURE, SVDquartets, and phylogenetic networks) support the recognition of six varieties of G. spinescens . Glossopetalon was fully supported as monophyletic and G. pungens was resolved sister to the remaining taxa in the genus. The varieties of G. spinescens were resolved as two distinct lineages corresponding to their biogeography – one to the northwest (lineage 1) and one to southeast (lineage 2). Glossopetalon clokeyi was resolved at the base of lineage 1 and G. texense was embedded within lineage 2 sister to var. spinescens . Taxonomic changes include the recognition of G. texense and G. clokeyi as varieties of G. spinescens and description of a unique population from northern Arizona as a new variety – G. spinescens var. goodwinii .

Genomic DNA was isolated from silica dried leaf material and herbarium vouchers using an amended Sorbitol protocol (Štorchová et al. 2000) with the exception of G. texense, which was extracted using a CTAB protocol with the addition of pvp-40 (Doyle and Doyle 1987). Preliminary DNA quality was assessed with 1% agarose gel electrophoresis. DNA quantifications and purity determinations were conducted via a Nanodrop 1000 Spectrophotometer (Thermo Scientific, Carlsbad, California) and PicoGreen quantification was conducted with a Synergy HTX Multi-Mode Microplate Reader (BioTek Instruments, Winooski, Virginia). All samples were normalized to 10ng/uL using 10mM Tris-Cl pH 8.0 before library preparation.

The libraries were prepared using an amended protocol of Peterson et al. (2012). Template DNA was digested with restriction enzymes MspI and EcoRI (NEB, Inc.). Adapter ligation was simultaneously conducted during the same reaction. Preparation of the adapters (Eurofins Genomics, USA) were as follows: the P1.1 EcoRI Adapter 5’-CCTATGTGGAGAGCCAGTAAGCGATGCTATGGT-3’ was annealed to P1.2 EcoRI Adapter 5’-[PHO]AATTACCATAGCATCGCTTACTGGCTCTCCACATAGG-3’ using a PTC-100 Programmable Thermal Cycler heated to 95°C for 5 minutes followed by a cool down to 25°C. Afterwards the EcoRI adapter was diluted to a concentration of 0.05μM with sterile water. The P2.1-MspI Adapter 5’-GTCAACGCTCACTACTGCGATTACCCAAGTCAG-3’ was likewise annealed to P2.2 Adapter 5’-[PHO]GCCCTGACTTGGGTAAGATAGCAC-3,’ but subsequently diluted to a concentration of 0.5μM using sterile water. The differences in concentration of the adapters were to account for the higher frequency of EcoR1 restriction enzyme sites. T4 DNA Ligase (NEB, Inc) was employed to ligate the adapters to digested DNA fragments. Furthermore, the reagents utilized for this reaction were: BSA (100x), EcoR1 10x Buffer, T4 DNA Ligase 10x Buffer with 10μM ATP, and sterile water. The reaction underwent 6 cycles of 37°C for 20 mins followed by 25°C for 20 mins and remained in the thermal cycler overnight at 10°C.

A 1:1 bead cleanup was performed with 25% PEG before the PCR indexing reaction. This amplification reaction consisted of Phusion HS II (Thermal Scientific), MgCl₂, custom primers, template DNA, and sterile water. Each sample was double indexed using distinctive forward and reverse indices. Indexing was performed over 25 cycles of 95°C for 1 min, 35°C for 15 secs, 55°C for 15 secs, 72°C for 30 secs, and 72°C for 7 mins. Now that samples were indexed, all samples were pooled and underwent a 1:1 bead clean up with 18% PEG. Samples were subsequently quantified using a Nanodrop 1000 Spectrophotometer and analyzed on an Advanced Analytical Fragment Analyzer (Advanced Analytical Technologies GmbH, Heidelberg, Germany). Based on the high presence of the fragments from 200-550 base pairs, a size selection at that range was conducted using the Pippen Prep (Sage Science, Beverly, Massachusetts) for the ddRAD library. These libraries were sequenced on a single lane on a HiSeq 4000 instrument (Illumina, San Diego, California) at the University of Oregon’s Genomics and Cell Characterization Core Facility using custom primers to produce single-end 150 base pair reads.

Sequence Data Preparation— Demultiplexing of raw data was conducted in accordance with akutils RADseq utility protocol using the module fastq-multx from EA-UTILS (Aronesty 2011; Andrews 2018). The demultiplexed data has been uploaded to Dryad. 

The demultiplexed data files are named by sample number. For example, r1.C1.fq, are the reads for sample tube C1. I uploaded a separate file indicating the taxon identified, collection locations, and collection date, for each sample tube ID. Each sample tube corresponds to an individual plant.