Previous phylogenetic studies in oaks (Quercus, Fagaceae) have failed to resolve the backbone topology of the genus with strong support. Here, we utilize next-generation sequencing of restriction-site associated DNA (RAD-Seq) to resolve a framework phylogeny of a predominantly American clade of oaks whose crown age is estimated at 23–33 million years old. Using a recently developed analytical pipeline for RAD-Seq phylogenetics, we created a concatenated matrix of 1.40 E06 aligned nucleotides, constituting 27,727 sequence clusters. RAD-Seq data were readily combined across runs, with no difference in phylogenetic placement between technical replicates, which overlapped by only 43–64% in locus coverage. 17% (4,715) of the loci we analyzed could be mapped with high confidence to one or more expressed sequence tags in NCBI Genbank. A concatenated matrix of the loci that BLAST to at least one EST sequence provides approximately half as many variable or parsimony-informative characters as equal-sized datasets from the non-EST loci. The EST-associated matrix is more complete (fewer missing loci) and has slightly lower homoplasy than non-EST subsampled matrices of the same size, but there is no difference in phylogenetic support or relative attribution of base substitutions to internal versus terminal branches of the phylogeny. We introduce a partitioned RAD visualization method (implemented in the R package RADami; http://cran.r-project.org/web/packages/RADami) to investigate the possibility that suboptimal topologies supported by large numbers of loci—due, for example, to reticulate evolution or lineage sorting—are masked by the globally optimal tree. We find no evidence for strongly-supported alternative topologies in our study, suggesting that the phylogeny we recover is a robust estimate of large-scale phylogenetic patterns in the American oak clade. Our study is one of the first to demonstrate the utility of RAD-Seq data for inferring phylogeny in a 23–33 million year-old clade.

pyRAD output: c88_d6m4p2_wRE.readlociLocus file from pyRAD, clustering at 0.88 similarity, a minimum of 4 individuals per locus. This file can be read in using the read.pyRAD function of the R package RADami (http://cran.r-project.org/web/packages/RADami/index.html); structure of the data is described in the documentation to pyRAD (code.google.com/p/pyrad/; www.dereneaton.com/software). Briefly, each locus is represented by a sequence alignment of those individuals for which the cluster was found, and loci are delimited by a break line that indicates variable sites.c88_d6m4p2_wRE.readloci.zippyRAD loci output: c88_d6m10p2_wRE.readlociLocus file from pyRAD, clustering at 0.88 similarity, a minimum of 10 individuals per locus. This file can be read in using the read.pyRAD function of the R package RADami (http://cran.r-project.org/web/packages/RADami/index.html); structure of the data is described in the documentation to pyRAD (code.google.com/p/pyrad/; www.dereneaton.com/software). Briefly, each locus is represented by a sequence alignment of those individuals for which the cluster was found, and loci are delimited by a break line that indicates variable sites.c88_d6m10p2_wRE.readloci.zipPhylip file: d6m10.phyPhylip-style (sequential) multiple alignment file, generated from pyRAD loci d6m10 file.d6m10.phy.zipblast2go results: 2013-04-09.datablast2go results of d6m10 loci against three EST databases.2013-04-09.data.zipPhylogenetic analyses.EST.2013-03-28Results of phylogenetic analyses conducted on EST-length subsets of the original RAD dataset.analyses.EST.2013-03-28.zipBLASTN results: blasts.2013-02-21BLASTN results of m10 RAD locus consensus sequences against EST databases in NCBI.blasts.2013-02-21.zipR source file: PLOS.2013.analysesR source file for performing analyses presented in the paper that were not ported to RADami.PLOS.2013.analyses.RR source: PLOS.2013.blast2goR code for checking and summarizing blast2go results presented in paper.PLOS.2013.blast2go.R