Research articleIncreasing phylogenetic resolution at low taxonomic levels usingmassively parallel sequencing of chloroplast genomesMatthew Parks1, Richard Cronn2 and Aaron Liston * 1

Abstract

Background: Molecular evolutionary studies share the common goal of elucidating historicalrelationships, and the common challenge of adequately sampling taxa and characters. Particularly atlow taxonomic levels, recent divergence, rapid radiations, and conservative genome evolution yieldlimited sequence variation, and dense taxon sampling is often desirable. Recent advances inmassively parallel sequencing make it possible to rapidly obtain large amounts of sequence data, andmultiplexing makes extensive sampling of megabase sequences feasible. Is it possible to efficientlyapply massively parallel sequencing to increase phylogenetic resolution at low taxonomic levels?

Results: We reconstruct the infrageneric phylogeny of Pinus from 37 nearly-complete chloroplastgenomes (average 109 kilobases each of an approximately 120 kilo base genome) generated usingmultiplexed massively parallel sequencing. 30/33 ingroup nodes resolved wlth > 95% bootstrapsupport; this is a substantial improvement relative to prior studies, and shows massively parallelsequencing-based strategies can produce sufficient high quality sequence to reach support levelsoriginally proposed for the phylogenetic bootstrap. Resampling simulations show that at least theentire plastome is necessary to fully resolve Pinus, particularly in rapidly radiating clades. Meta-analysisof 99 published infrageneric phylogenies shows that whole plastome analysisshould providesimilar gainsacross a range of plant genera. A disproportionate amount of phylogenetic informationresides in two loci (ycf1, ycf2, highlighting their unusual evolutionary properties.

Conclusion: Plastome sequencing is now an efficient option for increasing phylogenetic resolutionat lower taxonomic levels in plant phylogenetic and population genetic analyses.With continuingimprovements in sequencing capacity, the strategies herein should revolutionize efforts requiringdense taxon and character sampling, such as phylogeographic analyses and species-level DNAbarcoding.

BackgroundMolecular phylogenetic and phylogeographic analyses aretypically limited by DNA sequencing costs, and this forcesinvestigators to choose between dense taxon samplingwith a small number of maximally informative loci, or

genome-scale sampling across a sparse taxon sample [1-4]. Balancing these choices is particularly difficult in stud-ies focused on recently diverged taxa or ancient rapid radi-ations, as taxon sampling needs to be sufficiently large todefine the magnitude of intraspecific variation and the

phylogenetic depth of shared alleles [5,6]. Similarly,broad genome sampling is necessary to offset the low levelof genetic divergence among individuals of recent co-ancestry and to overcome low phylogenetic signal to noiseratios characteristic of rapid radiations [6]. Next genera-tion DNA sequencing is poised to bring the benefits ofaffordable genome-scale data collection to such studies atlow taxonomic levels (genera, species, and populations).Massively parallel sequencing (MPS) has increased perinstrument sequence output several orders of magnituderelative to Sanger sequencing, with a proportional reduc-tion in per-nucleotide sequencing costs [7,8]. In principlethis could allow the rapid sequencing of large numbers ofentire organellar genomes (chloroplast or mitochondria)or nuclear loci, and result in greatly increased phyloge-netic resolution [9]. To date, comparatively few plant oranimal evolutionary genetic analyses have utilized MPS[10-12], due to associated costs and the technical chal-lenge of assembling large contiguous sequences frommicro-reads. These barriers have been largely eliminatedthrough four innovations: development of strategies fortargeted isolation of large genomic regions [9,13-15]; har-nessing the capacity of these platforms to sequence tar-geted regions in multiplex [9,14,16]; streamlining samplepreparation and improving throughput [17]; and devel-oping accurate de novo assemblers that reduce relianceupon a predefined reference sequence [18,19].

In this paper we demonstrate the feasibility and effective-ness of MPS-based chloroplast phylogenomics for one-third of the world's pine species (Pinus), a lineage withnumerous unresolved relationships based on previouscpDNA-based studies [20-22]. We also highlight thebroad applicability of our approach to other plant taxa,and remark on the potential applications to similar mito-chondrial-based studies in animals and plant DNA bar-coding. Using multiplex MPS approaches, we sequencednearly-complete chloroplast genomes (120 kilobases (kb)each total length) from 32 species in Pinus and four rela-tives in Pinaceae. Our sampling of Pinus includes bothsubgenera (subg. Pinus, 14 accessions; subg. Strobus, 21accessions) and species exemplars chosen from all 11 tax-onomic subsections [21] to evenly cover the phylogeneticdiversity of the genus. Taxon density is highest for a cho-sen subsection (subsect. Strobus) as representative of a spe-cies-rich clade lacking phylogenetic resolution in previousstudies [5,21-23]. Three species are also represented bytwo chloroplast genomes each (P. lambertuma, P. thunber-gii, P. torreyana).

ResultsGenomic Assemblies and AlignmentAssemblies in subgenus Strobus averaged 117 kb, with anestimated 8.8% missing data (compared to P. koraiensisreference); subg. Pinus assemblies averaged just less than

120 kb (6% estimated missing data, compared to P. thun-bergii reference). Outgroup assemblies averaged just over119 kb (10.4% average estimated missing data comparedto P. thunbergii reference). Median coverage depth fordetermined positions was variable but typically high(range 21 to 156x) (Table 1, [also see additional file 1]).Full alignment of all assemblies was 132,715 bp in length,including 62,298 bp from exons encoding 71 conservedprotein coding genes (20,638 amino acids), 36 tRNAsand4 rRNAs.A high degree of co-linearity is inferred for thesegenomes due to the absence of major rearrangementswithin de novo contigs, and by the overall success of thepolymerase chain reaction-based sequence isolation strat-egy (indicating conservation of the order of anchor genescontaining primer sites). However, minor structuralchanges (a tandem duplication in two species [24] and theapparent loss of duplicate copies of psaM and rps4 in P.koraiensis) could not be confirmed. No evidence of inter-specific recombination was detected, consistent with therarity of recombination in plant plastomes [25].

The aligned matrix contained 7,761 parsimony informa-tive ingroup substitutions (4,286 non-coding positionsand 3,475 coding positions) (Table 2). Over one-half ofparsimony informative sites (55.0%) in protein codingregions resided in ycf1 and ycf2, two large genes of uncer-tain function [26], that accounted for 22% of all exonsequence (Figure 1A,B). No other exons in the pine plas-tome exhibit such a disproportionate number of parsi-mony informative sites (Figure 1C). These loci have anelevated nonsynonymous substitution rate (Table 3) andappear to have a substantial number of indels in Pinus,although it was not possible in many cases to confidentlyscore indels in these loci due to the inherent limitations ofreference-guided assembly of short reads in length varia-ble regions. Start codon position, overall length and stopcodon positions were nonetheless largely preserved inthese loci across the genus. In addition to substitutions inexons, 48 ingroup exon indels and 23 ingroup stop codonshifts were identified in 26 loci.

Phylogenetic Resolution in Non-Random and RandomizedData PartitionsFull alignment partitions yielded a higher proportion ofhighly supported nodes, with 88 to 91% (29 to 30/33) ofingroup nodes resolved with bootstrap support > 95% inlikelihood analysis. The four largest data partitions tested(full alignment and concatenated exon nucleotides, bothwith and without ycf1 and ycf2) yielded results that weretopologically identical with the exception of four taxa (P.albicaulis, P. krempfii, P. iambettiana N, P. parviflora) (Fig-ures 2 and 3). In addition, support for the branching orderof P. cembra, P. hotaiensis and P. sibirica was low in fullalignment partitions. Topological differences were foundto be significant according to Shimodaira-Hasegawa com-

parisons of the full alignment topology to two of the othermajor partitions (full alignment and exon nucleotideswithout ycf1 and ycf2). Trends in significance were moststrongly influenced by the two alternative positions of P.krempfii (Figure 2 vs. Figure 3A, C; Table 4). With theexception of P. krempfii, areas of topological uncertaintyreside in a single clade that historically has lacked internalresolution (subsection Strobus) [20-22]. Coalescent esti-mations suggest that these poorly resolved subsectionStrobus haplotypes diverged in rapid succession relative tothe age of their shared nodes (0.009 to 0.44 coalescentunits, or ca. 90,000 to 450,000 years) (Table 5). A putativechloroplast capture event in P. iambettiana previously doc-umented [5] was also supported with whole-plastomeresults. Substantial resolution was achieved in analyses of

ycf1 and ycf2 data partitions, however we observed severaltopological differences from the full alignment with highsupport (primarily involving the species discussed above)(Figure 4).

Of the 71 exon coding indels and stop codon shifts iden-tified, 35 mapped unambiguously to monophyleticgroups (that is, no accessions in a group were missing datafor that event) (Figures 5 and 6). All of these groups hadstrong support in nucleotide-based phylogenetic analyses(100% likelihood and parsimony bootstrap support). Theremainder of these events were primarily either putativelymonophyletic (missing data in one or more members of aclade) or showed strong evidence of homoplasy (Figures5 and 6).

In parsimony analyses of variable-sized jackknife samplesof our full alignment, nodal support showed a strong pos-itive correlation with the length of the nucleotide matrix(proportion nodes > 95% = -1.0808 + 0.38497*log10[matrix size, bp]; r2= 0.915, P < 0.0001) (Figure 7A). Res-olution of full alignment and exon nucleotide partitionswas indistinguishable from random jackknife samples ofcomparable size, indicating similar phylogenetic contentof these partitions and corresponding similar-sized ran-dom genomic subsamples. Partitions consisting of ycf1and ycf2 - in particular ycfl, and ycfl and ycf2 combined -showed significantly higher resolution than the genome-wide average (Figure 7A). The concatenated partition ycfl+ ycf2 (13.1 kb; 77.4% nodes > 95% bootstrap support)yielded only slightly less phylogenetic resolution than allexons combined (62.3 kb; 80.6% nodes > 95% bootstrapsupport) in parsimony analysis.

Comparisons to Previous Pinus PhylogeniesPrevious cpDNA based estimates of infrageneric relation-ships in Pinus [20-22] sampled the same species and/orlineages as our study, and inferred relationships using2.82 to 3.57 kb of chloroplast DNA. Results of these stud-ies are largely consistent with our results, although highlysupported nodes (> :95%) accounted for only 13 to 23%of the total ingroup nodes (23% to 42% if [20,21]adjusted to match our species composition). The empiri-cal results of these studies fell within or close to the 95%prediction intervals established from our jackknife resam-piing response from our full genome alignment (Figure7A), indicating that the loci used in prior studies (prima-rily rbcLand matK) are similarly informative as a compa-rable sample of random nucleotides from the chloroplastgenome.

Meta-Analysis of Published Infrageneric StudiesFrom our sampling, infrageneric analyses in plants pub-lished from 2006 to 2008 were typically based on 2574

aligned bp (95% bootstrap confidence interval: 2,292,2,864) of sequence data, evaluated 3l.7 ingroup species(95% bootstrap confidence interval: 20.2, 43.2), andresolved 22.6% of nodes at > 95% bootstrap support(95% bootstrap confidence interval: l8.6, 26.5). Regres-sion analysis shows that the proportion of highly resolvednodes in these studies is significantly and positively corre-lated with matrix length (F1,96 = 18.032; r2 = 0.149; P <0.0001) but not the number of included taxa (F1,97 =0.546; r2= 0.006; P = 0.461), although there was a nega-tive trend in the latter (Figure 7B, C). Our current samplesize is typical in the number of taxa sampled, but bothmatrix length (132.7 kb) and the proportion of highlybootstrap-supported nodes (84.8% parsimony, 90.3%maximum likelihood) were substantially higher.

DiscussionOur results highlight that whole plastome sequencing isnow a feasible and effective option for inferring phyloge-nies at low taxonomic levels. Compared to previous chlo-roplast-based phylogenetic analyses in Pinus, our datamatrix contained approximately 60 times more phyloge-netically informative characters resulting in an approxi-mately two- to four-fold increase in the proportion ofhighly resolved nodes (after adjusting results of previousstudies to match our species composition) (Figure 8,Table 2). An important question arising from these com-parisons is whether the difference in resolution is entirelyattributable to the increase in nucleotides, or whether thegenomic partitions sequenced in prior studies were lessinformative on average than the rest of the genome. Infact, the resolution provided by loci used in previous Pinusstudies is indistinguishable from or slightly greater thanthat of comparably sized random genomic subs ampiesfrom our full alignment. Combined with the strong corre-lation between resolution and the size of randomgenomic subsarnple, this suggests that the increase in res-olution in this study is primarily due to the increase in

matrix length. This is further supported by a significantrelationship between resolution and matrix length in abroad sampling of chloroplast-based infrageneric phylog-enies. Based on these results, we predict that whole-plas-tome analysis will yield similar gains in phylogeneticresolution not only in the genus Pinus but for most landplant genera. On the other hand, it is apparent that eventhe entire chloroplast genome may be insufficient to fullyresolve the most rapidly radiating lineages. In this regard,our results are reflective of previous analyses of ancientrapid radiations wherein nodal resolution does not scaleproportionately to the length of sequence analyzed[27,28]. Notably, the position of P. krempfii was signifi-cantly different between the four largest data partitions(Table 4), even though this species does not appear to beassociated with a rapid radiation (Table 5). This result is

not completely unexpected, as this species has previouslybeen difficult to place phylogenetically [29,30]. An une-quivocal resolution of this species will likely require theinclusion of multiple nuclear loci [30].

When considering recent divergence, the disproportion-ately high mutation rate in ycfl (and ycf2, to a lesserextent) demonstrated here is of importance, and mirrorsfindings in other plant taxa [31,32] and recently in Pinussubsection Ponderosae [33]. These loci should be informa-tive for phylogenetic studies in recently-diverged clades orin population-level studies in a range of plant species. Dis-cretion is advised, however, as ycfl (and possibly ycf2)appears to be a target of positive selection at least in Pinusand may reflect adaptive episodes rather than neutralgenealogies. In likelihood analyses of ycfl and ycf2, we

observed several topological differences from the fullalignment at the subsectional level, further demonstratingthat caution must be taken in drawing phylogenetic con-clusions from these two loci. Although we were able toconfidently score small structural changes (indels andstop codon shifts) for all other exons, it was not possibleto score indels for ycfl and ycf2 due to the apparent highrate of indel formation in these loci. In all other loci exam-ined, small structural changes only delineated clades withconcurrent high support from nucleotide-based analyses(both in present study and [20-22]), and thus are likely tobe of limited use in species or population level discrimi-nation. It is not clear whether this will also be the case inycfl and ycf2.

It is reasonable to ask whether increased resolution isworth the effort of assembling whole plastomes. Consid-ering the conservative nature of bootstrap measures [34-37], systematists often accept bootstrap values of > 70% as

reliable indicators of accurate topology [36]. Simulationstudies [34], however, have demonstrated greatlyincreased accuracy (approximately 42 x) with bootstrapvalues > 95% versus > 70%, and the initial formulation ofthe phylogenetic bootstrap used > 95% as the thresholdfor topological significance [38]. Our results similarlysupport using a 95% bootstrap support cutoff for conclu-sive evidence as in both areas of topological differences,more than one clade received bootstrap support > 70% byanalysis of alternate data partitions. It is probable thatconflicting topologies with > 70% but < 95% bootstrapsupport accurately reflect data partitions yet may not rep-resent the plastome phylogeny, and here the use of entireorganelle genomes makes it possible to adopt more con-servative criteria of nodal support. There are further bio-logical reasons why an organellar phylogeny (essentially asingle-gene estimate) may not accurately represent theorganismal phylogeny; these include interspecific hybrid-ization, incomplete lineage sorting, and stochastic proper-

ConclusionPlastome sequencing is now a reasonable option forincreasing resolution in phylogenetic studies at low taxo-nomic levels and will continue to become an increasinglysimple process. As sequencers evolve to even higher capac-ity and multiplexing becomes routine in the near future,this will allow more extensive taxon and genomic sam-pling in phylogenetic studies at all taxonomic levels. It isestimated that sequencing capacity on next generationplatforms will approach 100 gigabase pairs per sequenc-ing run by the end of 2009. For perspective, this is suffi-cient sequence capacity to produce all 100 genus-leveldata sets used in our meta-analysis (including ours) atgreater than 100x coverage depth in a single sequencingrun. Based on the estimates of Cronn et al. [9], thissequencing capacity would also allow the simultaneoussequencing of several thousands of animal mitochondria,which could greatly benefit low-level taxonomic or popu-lation-based studies in animals that currently tend to relyon relatively short sequences from many individuals [39].It is also clear that these improvements could enable otherpursuits that are currently hindered by limited sequencingcapacity, such as identification of plants by diagnosticDNA sequences (DNA barcoding). The recently agreed

ties of the coalescent process. Nonetheless, phylogeneticreconstruction based on complete organellar sequencesmay facilitate the detection of such phenomena, by reduc-ing errors and uncertainty due to insufficient sampling ofDNA sequence.

upon two locus chloroplast barcode for plants claims only72% unique identification to species level[40]. Based onresults herein, whole plastome sequences have the poten-tial to be more highly discriminating and efficient plantDNA barcodes; in fact, the possibility of plastome- andmitome-scale barcodes has been raised previously [41].Results in this area (as well as in phylogenetic and phylo-geographic analyses) will be impacted particularly ifadvances in target isolation and enrichment [13-15] andstreamlining sample preparation [17] prove globallyeffective.

MethodsDNA Extraction, Amplification and SequencingDNA extraction, amplification and sequencing aredescribed in and followed Cronn et al. [9], with 4 bp mul-tiplex tags, replacing the original 3 bp tags (Table 1). Forone sample, P. ponderosa, additional reads from three non-multiplexed lanes of genomic DNA were also included.

Sequence Assembly and Genome AlignmentsSequence assembly and alignment are described in andfollowed Whittall et al. [42]. An analysis of interspecificrecombination was conducted using RDP(RecombinationDetection Program) v. 3.27 [43]. Rather than using thefull genomic alignment, which was too memory-inten-sive, concatenated. nucleotide sequences for 71 exonscommon to all accessions were used (reflective of order onthe plastome). Subgenera were investigated separately asmembers of opposing subgenera appear incapable ofhybridization [44]. Each subgenus was checked forrecombination events using standard settings for severalrecombination-detection strategies, including: RDP [45],GeneConv [46], Chimaera [47], MaxChi [48], BootScan[49], and SiScan [50]. A total of 24 putative recombina-tion events were identified. On close investigation, allevents involved one or more of the following: misalign-ment, autapomorphic noise coupled with missing data,and amplification of pseudogenes. In cases of misalign-ment, alignments were corrected prior to subsequent phy-logenetic analyses. In cases of amplification ofpseudogenes, the entire amplicon for the accessioninvolved was turned to Ns. Inspection of the alignmentalso revealed that some amplicons in some accessions hadfailed to amplify, or amplified apparently paralogous loci(evidenced by substantially higher divergence). Theseregions were masked in affected accessions. The locusmatK was determined to be a putative paralog in severalaccessions, and in four (P. armandii, P. lambertiana S, P.albieaulis, and P. ayaeahuite) it was replaced with Sangersequence [5]. We also replaced 2180 bp of poor qualitysequence of the locus ycf1 in P. ponderosa with Sangersequence. In all accessions amplified by PCR, the regionsadjacent to primer sites typically had low coverage, whileprimers had very high coverage, thus primer-flanking

regions (where problematic) and the primers were alsoexcluded. It was also determined through Sanger sequenc-ing that a 600 bp region of the previously published P.hotaiensis plastome (positions 48808 to 49634 in Gen-Bank AY228468) is apparently erroneous. This region wasremoved and reference guided analysis was rerun for thisamplicon.

Aligned sequences were annotated using DOGMA (DualOrganellar Genome Annotator) [51] with manual adjust-ments to match gene predictions from GenBank and theChloroplast Genome Database http://chloroplast.cbio.psu.eduj. Exons were evaluated for readingframe and translations, and validity of exon mutationswas judged based on presence in de novo sequence, effecton the resulting polypeptide sequence, and sequence cov-erage depth.

Phylogenetic AnalysesSequence data was analyzed using all genome positionsand concatenated nucleotide sequence from 71 exonscommon to all pine accessions; both partitions were ana-lyzed with and without the loci ycf1 and ycf2. A relativelyshort (approximately 630 bp) repetitive stretch of thelocus ycf1 of subgenus Strobus accessions was masked inall analyses due to alignment ambiguity. The loci ycf1 andycf2 (ca. 14 kb combined) were also analyzed individuallyand together.

Maximum Likelihood (ML) phylogenetic analyses wereperformed through the Cipres Web Portal http://www.phylo.org/portallHome.do using RAxMLbootstrap-ping with the general model of nucleotide evolution(GTR+G) [52] and automatically determined numbers ofbootstrap replicates. Bayesian inference analyses (BI) wereperformed using MrBayes v. 3.1.2 [53] using the GTR+G+Imodel, which was selected using MrModelTest v. 2.3 [54]under both Aikake Information Criterion and Hierarchi-cal Likelihood Ratio Test frameworks. Each analysis con-sisted of two runs with four chains each (three hot andone cold chain), run for 1000000 generations with treessampled every 100 generations. The first 25% percent oftrees from all runs were discarded as burn-in. Unweightedmaximum parsimony analyses (MP) of data partitionswere conducted in PAUP* (Phylogenetic Analysis UsingParsimony (*and other methods)) v. 4.0b10 [55] by heu-ristic search with 10 replicates of random sequence addi-tion, tree bisection and reconnection branch swappingand a maxtrees limit of 1,000. Non-parametric bootstrapanalysis was conducted under the same conditions for1,000 replicates to determine branch support.

Topological differences between the full alignment topol-ogy and each of the three other largest data partitions (fullalignment without ycf1 and ycf2, and exon nucleotides

both with and without ycf1 and ycf2) were tested for sig-nificance using the Shimodaira-Hasegawa test [56] withresampling estimated log-likelihood (RELL) bootstrap-ping (1,000 replicates) under the GTR+G model of evolu-tion. To further determine which topological differenceswere most influential, tests were repeated with the posi-tions of topology-variable accessions alternately modifiedto match the full alignment topology. In total, the fullalignment data set was compared to nine different topol-ogies.

Exon indels and stop codon shifts were mapped onto thetopology determined by ML analysis of the full alignmentby parsimony mapping using Mesquite v. 2.6 (Maddisonand Maddison, http·//mesquiteproject.org). Tests of selec-tion for exons were performed in MEGA v. 4.0 [57] usingthe codon-based Z-test for selection, with pairwise dele-tion and the Nei-Gojobori (P-distance) model; variance ofthe differences were computed using the bootstrapmethod with 500 replicates.

Estimation of Divergence Times for Poorly Resolved NodesDivergence times for four nodes with topological uncer-tainty (P. albicaulis - P. lambertiana N - P. parvifiora, P. sibir-ica - P. cembra - P. koraiensis, P. krempfii-sectionQuinquefoliae of subgenus Strobus) were estimated accord-ing to Pollard et al. [58]. Chloroplast mutation rate wasestimated by averaging maximum and minimum muta-tion rates for Pinaceae chloroplast genomes from two pre-vious studies [59,60] and assuming a generation time of50 years [61]. Two estimates were calculated for each nodeusing eitherlow (10,000) or high (100,000) effective pop-ulation size [23].

Effect of Character Number on Phylogenetic ResolutionEmpirical data from Pinus genomesVariable-size random subsamples of the full alignmentwere tested under the parsimony criteria using PAUP* v.4.0b10 (the faststep option was used for all but the twosmallest partitions due to time considerations). Elevenpartition sizes were tested (2.5, 5, 10, 20, 30, 40, 50, 60,80, 100 and 120 kb) in five replicates each, with resolu-tion measured as the percentage of ingroup nodes pro-duced with > 95% jackknife support. Relationshipsbetween partition size and ingroup resolution were esti-mated using least squares regressions, and 95% confi-dence limits for individual points were estimated basedon linear regression using SAS JMP 7.0.1 (S.A.S. Institute,Inc., http://www.jmp.com/).Our full alignment exonnucleotides and ycf1/ycf2 partitions were analyzed underthe same parsimony criteria for comparison, as were thealignments of [20-22]. Accessions from Gernandt et al.and Eckert et al. [20,21] were pruned to include only taxacommon to our sampling; the original analysis of Wang et

al. [22] was used since this data matrix was not availablefor alternative phylogenetic analyses.

Meta-Analysis of Published StudiesWe evaluated 99 phylogenetic analyses from 86 studiespublished between 2006 and 2008 in Systematic Botany,Systematic Biology, American Journal of Botany, Taxon,Molecular Phylogenetics and Evolution, and Annals of theMissouri Botanical Garden [see additional file 2]. Analy-ses were selected based on: 1) the presented phylogenywas based solely on chloroplast DNA sequence; 2) theanalysis included > 10 species. from a monophyleticgenus; 3) there were more inter- than intra-specific taxaanalyzed within the genus; 4) parsimony-based bootstrapor jackknife values were presented. Ingroup branches withbootstrap support > 95%, the number of ingroup taxa andthe aligned base pairs used in the analysis were recordedfor each case. The authors' taxonomic interpretations wereaccepted in instances of taxonomic uncertainty. Conspe-cific clades were treated as one taxon unless clearly differ-entiated from one another, and internal bootstrap valueswere disregarded. The number of branches with bootstrapsupport > 95% was regressed both on the number ofaligned base pairs and the number of taxa (both log-trans-formed to meet assumptions of normality and equal vari-ances).

Data DepositionIllumina sequencing reads and quality scores have beendeposited in the NCBI SRA database as accessionSRA009802. New sequences have been deposited in Gen-Bank as accessions FI899555-FI899583.

Accession numbers cited in manuscript[GenBank FI899555-FI899583, EU998739-EU998746,SRA009802]

AbbreviationsBI: Bayesian Inference; bp: base pairs; cpDNA: chloroplastDNA; kb: kilobase; ML: maximum likelihood; MP: maxi-mum parsimony; MPS: massively parallel sequencing.

Authors' contributionsMP obtained and assembled the plastome data, con-ducted the phylogenetic and statistical analyses, con-ducted the meta-analysis and drafted the manuscript. RCand AL conceived the study and contributed to data col-lection, data analysis and manuscript writing. All authorsread and approved the final manuscript.

AcknowledgementsWe thank Mariah Parker-deFeniks and Sarah Sundholm for lab assistance,Uranbileg Daalkhaijav, Zachary Foster and Brian Knaus for computingassistance, Linda Raubeson for providing a chloroplast isolation of Larix occi-dentalis, Christopher Campbell and Justen Whittall for DNA samples, DavidGernandt, Chris Pires, jonathan Wendel and Mark Fishbein for editorialcomments, and Steffi Ickert-Bond for timely questions. We also thank MarkDasenko, Scott Givan, Chris Sullivan and Steve Drake of the OSU Centerfor Genome Research and Biocomputing. This work was supported byNational Science Foundation grants (ATOL-0629508 and DEB-03171 03 toAL. and R.C), the Oregon State University College of Science VentureFund and the US Forest Service Pacific Northwest Research Station.

ReferencesI. Moore Mj, Bell CD, Soltis PS, Soltis DE: Using plastid genome-

scale data to resolve enigmatic relationships among basalangiosperms. Proc Natl Acad SciUSA 2007, 104: 19363.

2. Delsuc F, Brinkmann H, Philippe H: Phylogenomics and thereconstruction of the tree of life. Nature Rev Genet 2005,6:3'61-375.

3. Philippe H, Frederic D, Henner B, Lartillot N: Phylogenomics. AnnuRev Ecol Evol Syst 2005, 36:541-542.

4. Jansen RK, Cai Z, Raubeson LA, Daniell H, Depamphilis CW, Lee-bens-Mackj, Muller KF, Guisinger-Bellian M, Haberle RC, Hansen AK:Analysis of 81 genes from 64 plastid genomes resolves rela-tionships in angiosperms and identifies genome-scale evolu-tionary patterns. Proc Natl Acad Sci USA 2007, 104: 19369.

5. Liston A, Parker-Defeniks M, Syring jV, Willyard A, Cronn R: Inter-specific phylogenetic analysis enhances intraspecific phyloge-ographical inference: a case study in Pinus lambertiana. MolEco12007, 16:3926-3937.

6. Whitfield jB, Lockhart Pj: Deciphering ancient rapid radiations.Trends Ecol Evol 2007, 22:258-265.

7. Hudson ME: Sequencing breakthroughs for genomic ecologyand evolutionary biology. Molecular Ecology Resources 2008,8:3-17.

8. Mardis ER: The impact of next-generation sequencing tech-nology on genetics. Trends Genet 2008, 24: 133-141.

9. Cronn R, Liston A, Parks M, Gernandt DS, Shen R, Mockler T: Mul-tiplex sequencing of plant chloroplast genomes using Solexasequencing-by-synthesis technology. Nucleic Acids Res 2008,36:e 122.

10. Gilbert MTP, Drautz DI, Lesk AM, Ho SYW, Qi j, Ratan A, Hsu CH,Sher A, Dalen L,Gotherstrom A: Intraspecific phylogenetic anal-ysis of Siberian woolly mammoths using complete mito-chondrial genomes. Proc Nod Acad SciUSA 2008, 105:8327.

I I. Moore Mj, Dhingra A, Soltis PS, Shaw R, Farmerie WG, Folta KM,Soltis DE: Rapid and accurate pyrosequencing of angiospermplastid genomes. BMC Plant Biol 2006, 6: 17.

12. Ossowski S, Schneeberger K, Clark RM, Lanz C, Warthmann N,Weigel D: Sequencing of natural strains of Arabidopsis thalianawith short reads. Genome Res 2008, 18:2024.

13. Gnirke A, Melnikov A, Maguire j, Rogov P, LeProust EM, BrockmanW, Fennell T, Giannoukos G, Fisher S, Russ C: Solution hybridselection with ultra-long oligonucleotides for massively par-allel targeted sequencing. Nature Biotech 2009, 27: 182-189.

14. Porreca Gj, Zhang K, LijB, Xie B,Austin D, Vassallo SL,LeProust EM,Peck Bj, Emigq, Dahl F, Gao Y, Church GM, Shendure j: Multiplexamplification of large sets of human exons. Nature Meth 2007,4:931-936.

15. Herman DS, Hovingh GK, lartchouk 0, Rehm HL, Kucherlapati R,Seidman jG, Seidman CE: Filter-based hybridization capture ofsubgenomes enables resequencing and copy-number detec-tion. Nature Meth 2009, 6:507-510.

16. Craig DW, Pearson jV, Szelinger S, Sekar A, Redman M, Corneveauxjj, Pawlowski TL, Laub T, Nunn G, Stephan DA: Identification ofgenetic variants using bar-coded multiplexed sequencing.Nature Meth 2008, 5:887-893.

17. Quail MA, Kozarewa I, Smith F, Scally A, Stephens Pj, Durbin R,Swer-dlow H, Turner Dj: A large genome center's improvements tothe lIIumina sequencing system. Nature Meth 2008,5: 1005-1 0 1O.

18. Hernandez D, Francois P, Farinelli L, Osteras M, Schrenzel j: Denovo bacterial genome sequencing: millions of very shortreads assembled on a desktop computer. Genome Res 2008,18:802-809.

19. Zerbino DR, Birney E: Velvet: algorithms for de novo shortread assembly using de Bruijn graphs. Genome Res 2008,18:821-829.

20. Eckert Aj, Hall BD: Phylogeny, historical biogeography, andpatterns of diversification for Pinus (Pinaceae): Phylogenetictests of fossil-based hypotheses. Mol Phylogenet Evol 2006,40: 166-182.

21. Gernandt DS, Lopez G, Garcia SO, Liston A: Phylogeny and clas-sification of Pinus. Taxon 2005, 54:29-42.

22. Wang XR, Tsumura Y, Yoshimaru H, Nagasaka K, Szmidt AE: Phylo-genetic relationships of Eurasian pines (Pinus, Pinaceae)based on chloroplast rbcL, matK, rp120-rps 18 spacer, and trnVintron sequences. Am J.Bot 1999, 86: 1742-1753.

23. Syring j, Farrell K, Businsky R, Cronn R, Liston A: Widespreadgenealogical nonmonophyly in species of Pinus subgenusStrobus. Syst Biol 2007, 56: 163-181.

24. Lidholm j, Gustafsson P: The chloroplast genome of the gymno-sperm Pinus contorta : a physical map and a complete collec-tion of overlapping clones. Curr Genet 1991, 20: 161-166.

25. Palmer jD: Comparative organization of chloroplastgenomes. Annu Rev Genet 1985, 19:325-354.

26. Drescher A, Ruf S, Calsa T, Carrer H, Bock R: The two largestchloroplast genome-encoded open reading frames of higherplants are essential genes. Plant J 2000, 22:97-104.

27. Fishbein M, Hibsch-jetter C, Oltis DES, Hufford L: Phylogeny ofSaxifragales (angiosperms, eudicots): analysis of a rapid,ancient radiation. Syst Biol 2001, 50:817-847.

28. Wortley AH, Rudall Pj, Harris Dj, Scotland RW: How much dataare needed to resolve a difficult phylogeny? Case study inLamiales. Syst Biol 2005, 54:697-709.

29. Wang XR, Szmidt AE, Nguyen HN: The phylogenetic position ofthe endemic flat-needle pine Pinus krempfii (Pinaceae) fromVietnam, based on PCR-RFLP analysis of chloroplast DNA.Plant Syst Evol 2000, 220:21-36.

30. Syring J, Willyard A, Cronn R, Liston A: Evolutionary relationshipsamong Pinus (Pinaceae) subsections inferred from multiplelow-copy nuclear loci. Am J Bot 2005, 92:2086-2100.

31. Neubig KM, Whitten WM, Carlsward BS, Blanco MA, Endara L,Wil-liams NH, Moore M: Phylogenetic utility of ycf1 in orchids: aplastid gene more variable than matK. Plant Syst Evol 2009,277:75-84.

32. Chung SM, Gordon VS, Staub jE: Sequencing cucumber (Cucumissotivus L.) chloroplast genomes identifies differencesbetween chilling-tolerant and-susceptible cucumber lines.Genome 2007, 50:215-225.

33. Gernandt DS. Hernandez-Leon S. Salgado-Hernandez E. Perez de laRosa jA: Phylogenetic Relationships of Pinus Subsection Pon-derosae Inferred from Rapidly Evolving cpDNA Regions. SystBot 2009. 34:481-491.

34. Alfaro ME.Zoller S. Lutzoni F: Bayes or bootstrap? A simulationstudy comparing the performance of Bayesian Markov chainMonte Carlo sampling and bootstrapping in assessing phylo-genetic confidence. Mol Bil Evo/2003. 20:255-266.

35. Douady CJ. Delsuc F. Boucher Y. Doolittle WF. Douzery Ej: Com-parison of Bayesian and maximum likelihood bootstrapmeasures of phylogenetic reliability. Mol Bil Evol 2003.20:248-254.

36. Hillis DM, Bull jj: An empirical test of bootstrapping as amethod for assessing confidence in phylogenetic analysis.Syst Bil 1993. 42: 182-192.

37. Suzuki Y, Glazko GV. Nei M: Overcredibility of molecular phyl-ogenies obtained by Bayesian phylogenetics. Proc Natl Acad SciUSA 2002. 99: 16138-16143.

38. Felsenstein J: Confidence limits on phylogenies: an approachusing the bootstrap. Evolution 1985,39:783-791.

39. Patenaude Nj, Portway VA, Schaeff CM, Bannister JL, Best PB, PayneRS, Rowntree Vj, Rivarola M, Baker CS: Mitochondrial DNAdiversity and population structure among southern rightwhales (Eubalaena australis). J Hered 2007,98:147-157.

40. Hollingsworth PM, Forrest LL,Spouge jL, Hajibabaei M, RatnasinghamS, Bank M van der, Chase MW, Cowan RS, Erickson DL, Fazekas Aj,et al.: A DNA barcode for land plants. Proc Natl Acad Sci USA2009,106:12794-12797.

41. Erickson DL, Spouge J. Resch A. Weigt LA, Kress jW: DNA barcod-ing in land plants: developing standards to quantify and max-imize success. Taxon 2008, 57: 1304-1316.

42. Whittall js, Syring j, Parks M, Buenrostro j, Dick C. Liston A, CronnR: Finding a (pine) needle in a haystack: chloroplast genomesequence divergence in rare and widespread pines. Mol Ecol2009 in press.

43. Martin DP, Williamson C, Posada 0: RDP2: recombination detec-tion and analysis from sequence alignments. 2005,21 :260-262.

44. Price RA, Liston A, Strauss SH: Phylogeny and Systematics ofPinus. In Ecology and Biogeography of Pinus Edited by: Richardson DM.Cambridge: Cambridge University Press; 1998:49-68.

45. Martin D, Rybicki E: RDP: detection of recombination amongstaligned sequences. 2000, 16:562-563.

46. Padidam M, Sawyer S, Fauquet CM: Possible emergence of newgeminiviruses by frequent recombination. Virology 1999,265:218-225.

47. Posada D, Crandall KA: Modeltest: testing the model of DNAsubstitution. Bioinformatics 1998, 14:817-818.

48. Smith jM: Analyzing the mosaic structure of genes. J Mol Evol1992, 34: 126-129.

49. Martin DP, Posada 0, Crandall KA, Williamson C: A modified boot-scan algorithm for automated identification of recombinantsequences and recombination breakpoints. AIDS Research &Human Retroviruses 2005, 21 :98-1 02.

50. Gibbs Mj,Armstrong jS, Gibbs Aj: Sister-scanning: a Monte Carloprocedure for assessing signals in recombinant sequences.2000, 16:573-582.

51. Wyman SK, jansen RK. Boore jL: Automatic annotation oforganeliar genomes with DOGMA. Bioinformatics 2004,20:3252-3255.

52. Stamatakis A: A rapid bootstrap algorithm for the RAxML webservers. Syst BioI 2008, 57:758-771.

53. Ronquist F. Huelsenbeck jP: MrBayes 3: Bayesian phylogeneticinference under mixed models. Bioinformatics 2003.19: 1572-1574.

54. Nylander jM: MrModeltest v2. Program distributed by the authorEvolutionary Biology Centre, Uppsala University 2004.

55. Swofford DL: PAUP*. Phylogenetic Analysis Using Parsimony(*and Other Methods). In Version 4 Sunderland, Massachusetts: Sin-auer Associates; 2000.

56. Shimodaira H, Hasegawa M: Multiple comparisons of log-likeli-hoods with applications to phylogenetic inference. Mol BiolEvo/1999,16:1114-1116.

57. Tamura K, Dudley j, Nei M, Kumar S: MEGA4: Molecular Evolu-tionary Genetics Analysis (MEGA) software version 4.0. MolBiol Evol 2007. 24: 1596-1599.

58. Pollard DA, Iyer VN. Moses AM, Eisen MB, McAllister BF: Wide-spread discordance of gene trees with species tree in Dro-sophila : evidence for incomplete lineage sorting. PLoS Genet2006, 2:e 173.

59. Gernandt OS, Magallon S, Geada Lopez G. Zeron Flores 0,WillyardA, Liston A: Use of simultaneous analyses to guide fossil-basedcalibrations of Pinaceae phylogeny. Int J Plant Sci 2008.169: 1086-1099.

60. Willyard A, Syring j, Gernandt DS, Liston A, Cronn R: Fossil calibra-tion of molecular divergence infers a moderate mutationrate and recent radiations for Pinus. Mol Biol Evol 2007,24:90-101.

61. Bouille M, Bousquet j: Trans-species shared polymorphisms atorthologous nuclear gene loci among distant species in theconifer Picea (Pinaceae): implications for the long-termmaintenance of genetic diversity in trees. Am J Bot 2005,92:63-73.