Phylogenetic studies based on two nuclear ribosomal markers created the first phylogenetic framework in which to understand evolutionary relationships in Mentzelia section Bartonia (Loasaceae), but low molecular variation resulted in several large polytomies and an incomplete understanding of species relationships. We applied a genome-skimming approach to determine whether additional genetic variation generated from high-throughput sequencing could resolve relationships in one of the largest polytomies in the section. Among 20 species sequenced, a group of six species that have pinnatisect leaf morphology were previously hypothesized to be monophyletic and we tested whether additional data would resolve the group as monophyletic. For the chloroplast genome, reads were assembled using both de novo and reference-guided approaches, whereas reference-guided approaches were taken for the nuclear ribosomal cistron region and a single anonymous nuclear locus. Significant discordance was identified among all three gene trees. Exhaustive measures were taken to ensure phylogenetic and assembly-based errors were not responsible for the observed discordance among gene trees. We attribute much of the incongruence to a low phylogenetic signal-to-noise ratio that is likely due to the group radiating recently and rapidly and perhaps unique evolutionary histories among genomes. Despite incongruence, several well-supported relationships emerged across data sets, and all hypothesis tests for a monophyletic pinnatisect group among gene trees failed to reject monophyly of the group, suggesting the pinnatisect leaf form is a synapomorphy for the clades.

Genomic DNA was extracted using the CTAB method (Schenk et al., in press, their appendix 1). The DNA was quantified using an Agilent 2100 Bioanalyzer (Agilent Technologies) and Qubit 4 fluorometric quantitation (Invitrogen; Thermo Fisher). Whole genome sequencing was carried out in two reactions. The first reaction constructed libraries from total sonicated genomic DNA for 14 species with Illumina Nextera XT (Illumina Inc., San Diego CA U.S.A.) and was sequenced for 150 bp pair-end reads with the Illumina NextSeq platform (see Appendix 1) at the Georgia Genomics and Bioinformatics Core. The second reaction sequenced libraries for six additional species using NEB NEBNext® Ultra™ II DNA Library Prep Kit (New England Biolabs, Ipswich MA, U.S.A.) with enzymatic fragmentation and 150 bp pair-end reads were sequenced with the Illumina MiSeq platform at the Ohio University Genomics Facility.

Assembly and Alignment—Raw reads were filtered using Geneious v.10.2.2 (Biomatters, Ltd., Auckland, New Zealand). Reads were removed if they were duplicates, contained greater than four ambiguities, or were shorter than 75 bp. Mentzelia conspicua was first assembled using the Geneious de novo feature because it generated the greats number of reads, and resulting contigs that corresponded to the chloroplast and five anonymous loci were used in the reference guided assemblies. Regions that had less than 5x coverage were removed.

Both de novo and reference-guided methods were used to assemble the chloroplast genomes (see Table 1). The 14 NextSeq Mentzelia chloroplast genomes were assembled using GetOrganelle (Jin et al. 2020) with default settings and the M. conspicua de novo assembly as a seed. Reference-guided assemblies were performed in Geneious using M. conspicua as a reference for the six species sequenced with MiSeq because read depth was not great enough for de novo assembly. All reference-guided assemblies were conducted under medium/fast sensitivity for 25 iterations. A consensus sequence was assembled for each gene region or genome separately, filtering out sequences that had more than four ambiguities. A coverage filter of 65% was applied when assembling the consensus species.

Geneious and GetOrganelle.