8/6/2019 Amborella Not a Basal Angiosperm- Not So Fast - Soltis

1/5

997

American Journal of Botany 91(6): 9971001. 2004.

A MBORELLA NOT A BASAL ANGIOSPERM ?NOT SO FAST 1

DOUGLAS E. SOLTIS 2 AND PAMELA S. SOLTIS 3

2

Department of Botany, University of Florida, Gainesville, Florida 32611 USA; and3Florida Museum of Natural History, University of Florida, Gainesville, Florida 32611 USA

The sequence of the plastid genome of Amborella trichopoda , the putative sister to all other extant angiosperms, was recentlyreported ( Molecular Biology and Evolution 20: 14991505). Goremykin et al. used sequence data for 61 plastid genes from Amborellaand 12 other embryophytes in phylogenetic analyses and concluded that Amborella is not the sister to the remaining owering plants;the monocots instead occupy this position. The authors attributed their results, which differ substantially from all recent phylogeneticanalyses of angiosperms, to the increased character sampling (30017 nucleotides in their aligned matrix) in their analysis relative topublished studies that included fewer genes but more taxa. We hypothesized that the difference in topology is not due to limitedcharacter sampling in previous studies but to limited taxon sampling in the analysis by Goremykin et al. To test this, we conducted aseries of phylogenetic analyses using a three-gene, 12 (or more)-taxon data set to evaluate the topological effects of (i) including threevs. 61 genes for (nearly) the same set of taxa, (ii) analyzing different codon positions, (iii) substituting representatives of other basallineages for Amborella , (iv) replacing the grasses used to represent the monocots with other monocots, selected either for theirphylogenetic position or randomly, and (v) adding other basal taxa Nymphaea , Austrobaileya , magnoliids, and monocotsto the 12-taxon data set. Our results demonstrate that the monocots basal topology obtained by Goremykin et al. is not due to increasedcharacter sampling of the plastid genome; their topology was obtained using only two plastid genes or two plastid genes and onenuclear gene. This topology was also retained when either Nymphaea or Austrobaileya was substituted for Amborella , demonstratingthat any of the three basal lineages will attach to Calycanthus for lack of any other close branch. Furthermore, the monocots basaltopology is not robust to changes in sampling of monocots. Simply adding Oncidium , for example, places Amborella sister to theother angiosperms. Thus, limited taxon sampling, focusing on organisms with complete genome sequences, can lead to artifactualresults.

Key words: Amborella ; angiosperm phylogeny; basal angiosperms; taxon sampling; plastid DNA.

Goremykin et al. (2003) provide the complete sequence of the plastid genome of Amborella trichopoda (Amborellaceae),the putative sister to all other extant angiosperms (e.g., Ma-thews and Donoghue, 1999; Parkinson et al., 1999; Qiu et al.,1999; P. Soltis et al., 1999; D. Soltis et al., 2000; Savolainen

et al., 2000; Barkman et al., 2000; Graham and Olmstead,2000; Magallon and Sanderson, 2002; Zanis et al., 2002, 2003;Borsch et al., 2003; Hilu et al., 2003; Nickerson and Drouin,2004). This plastid sequence, 162 686 bases in length, is oneof fewer than 20 complete plastid sequences reported for landplants (including only 13 angiosperms; GenBank, 9/03) and istherefore an important contribution. However, Goremykin etal. then state that phylogenetic analyses of their data and othersequences obtained from databases indicate that Amborella isnot a basal angiosperm, a conclusion that contrasts sharplywith the results of recent molecular phylogenetic analyses (seereferences above). Although Goremykin et al. indicate a strongpreference for the Amborella not basal topology, they doacknowledge that further studies are needed to understand thecauses inuencing the position of Amborella obtained. . .

Goremykin et al. (2003) used only 13 taxa in their analysis,including three non-angiosperm outgroups and Amborella , Ca-lycanthus , and three monocots (all from Poaceae, a derivedlineage of monocots) as the only non-eudicot angiosperms.Adequate taxon sampling is crucial in phylogeny reconstruc-tion (e.g., Chase et al., 1993; Graybeal, 1998; Hillis, 1998;Pollock et al., 2002; Zwickl and Hillis, 2002). Furthermore,

1 Manuscript received 7 October 2003; revision accepted 10 February 2004.We thank Yin Long Qiu and Chuck Bell for helpful comments on earlier

drafts of the manuscript. This work was supported in part by DEB-0090283and PGR-0115684.

the constraint imposed by available complete plastid sequencesprecluded inclusion of other basal lineages that occupy im-portant phylogenetic positions. That is, in addition to Ambor-ella , Nymphaeaceae sensu APG II (2003) and Austrobaileya-les (see APG II, 2003; Trimeniaceae, Schisandraceae, andAustrobaileyaceae) are successive sisters to all other extantowering plants, but neither of these lineages is representedin the analysis of Goremykin et al.

Supporting our argument that the topology of Goremykinet al. is the result of limited taxon sampling are similar oddtopologies recovered by ourselves and others in early analysesof plastid and nuclear genes when few taxa were included. Forexample, in our initial analyses of small data sets of 18S rDNAsequences, we obtained topologies in which monocots, eitheras a clade or as a single taxon, appeared as sister to all otherextant angiosperms (D. Soltis et al., 1997). We recognized thatthe monocots sampled had long branches (see later), and wetherefore continued to add taxa, ultimately obtaining stabletrees in which Amborella , Austrobaileyales, and Nymphae-

aceae were sisters to all other owering plants (D. Soltis etal., 1997). Chase et al. (1993) similarly pointed out the errorsin the topologies recovered for angiosperms when small sub-sets of taxa were employed rather than large data sets. Like-wise, Naylor and Brown (1998) obtained incorrect relation-ships in animals with a small sampling of complete mitochon-drial DNA sequences.

The topology of Goremykin et al. for angiosperms differssubstantially from the consensus we obtained using three (P.Soltis et al., 1999; D. Soltis et al., 2000), ve (Qiu et al.,1999), six (Zanis et al., 2003), or 11 (Zanis et al., 2002) genesand from those reported by Mathews and Donoghue (1999),

8/6/2019 Amborella Not a Basal Angiosperm- Not So Fast - Soltis

2/5

998 [Vol. 91A MERICAN JOURNAL OF B OTANY

Fig. 1. Single most parsimonious tree (length 2770 steps; CI 0.644;RI 0.469) obtained in a phylogenetic analysis of three genes ( rbcL, atpB ,18S rDNA; data from D. Soltis et al. 2000). Bootstrap values are given abovethe branches. Branches without values received 50% bootstrap support.

Parkinson et al. (1999), Barkman et al. (2000), Graham andOlmstead (2000), Magallon and Sanderson (2002), Borsch etal. (2003), Hilu et al. (2003), and Nickerson and Drouin(2004). Goremykin et al. contend that their results are due tothe extensive plastid DNA data set (61 genes) they analyzed.However, we hypothesize that their topology has less to dowith extensive character sampling than with limited taxon

sampling.Goremykin et al. found Amborella and Calycanthus to besisters, with 100% bootstrap support. However, we hypothe-size that, given the taxon sampling by Goremykin et al., Am-borella had nowhere to go other than with Calycanthus , theonly other basal angiosperm included in their study other thanthe three grasses. If this hypothesis is correct, then a singlespecies of either of the other two basalmost branches of an-giospermsNymphaeaceae and Austrobaileyalesshould be-have as Amborella did in the analysis by Goremykin et al.

Furthermore, the placement of the monocots as the sister toall other angiosperms in the tree by Goremykin et al. and inthose obtained when additional samples of basal lineages andmagnoliids were added (see later), contra to all large-scalemulti-gene analyses (e.g., P. Soltis et al., 1999; D. Soltis et al.,2000; Qiu et al., 1999; Zanis et al., 2002, 2003; Mathews andDonoghue, 1999; Parkinson et al., 1999; Barkman et al.,2000), suggested that this result was due to the selection of two grasses (three in Goremykin et al., 2003) as the represen-tative monocots. Grasses are derived monocots and have longbeen known to have accelerated rates of cpDNA evolution(e.g., Gaut et al., 1992, 1996). Thus, the branch leading to thegrasses is quite long, given both the phylogenetic position of the grasses and their rapid rate of cpDNA evolution. We there-fore hypothesize that the selection of monocots other thangrasses, or the inclusion of additional monocots to shortenthe branch to the grasses, may alter the monocots basaltopology.

Character sampling: 61 vs. three genes To test the effectof 61 vs. three genes, we constructed a 12-taxon data set usingthe same (or nearly the same) taxa used by Goremykin et al.and data from three genes: the plastid genes rbcL and atpBand nuclear 18S rDNA (angiosperms from D. Soltis et al.,2000; non-angiosperms from Pryer et al., 2001). Our pruneddata set contained the same outgroups ( Marchantia , Psilotum ,and Pinus ) used by Goremykin et al. (2003) and nine angio-sperms, rather than 10 as in Goremykin et al. (2003); Triticumwas not included in our analysis because it was not present inthe D. Soltis et al. (2000) data set. Brassica (also of Brassi-caceae) was substituted for Arabidopsis ; Pisum (also of Fa-baceae) was substituted for Lotus ; and Epilobium ( Cham-erion ; also of Onagraceae) was substituted for Oenothera .

Parsimony analysis (100 random taxon addition replicates,tree bisection-reconnection [TBR] branch swapping, saving allmost parsimonious trees; using PAUP* 4.0 [Swofford, 2003])was conducted using this three-gene, 12-taxon data set. Boot-strap analyses (100 replicates) were conducted with 100 ran-dom taxon addition replicates, using TBR branch swappingand saving all most parsimonious trees.

Parsimony analysis found a single shortest tree (Fig. 1) withexactly the same topology as that reported by Goremykin etal.: Oryza Zea were sister to the rest of the angiosperms, Amborella Calycanthus were sisters, and all other relation-ships were identical to those in the tree of Goremykin et al.However, constraining Amborella to be sister to all other an-

giosperms, the position it occupies in the many studies citedearlier, lengthened the tree by only seven steps (length 2777steps), an increase in length of only 0.25% over the uncon-strained tree (2770 steps). The monocots basal result wasobtained in analyses that included (i) all nucleotide positionsof the three genes, (ii) all codon positions of the two plastidgenes, and (iii) the rst and second codon positions of the twoplastid genes. Because our analyses used only three instead of 61 genes and all analyses, regardless of codon positions in-cluded, yielded this same result, this topology is not due tothe increased character sampling used by Goremykin et al. orto the exclusion of third positions. Instead, we argue that itmust result from differences in taxon sampling.

Taxon sampling To test the hypothesis that representativesof either of the other two basalmost lineages of angiospermswould attach exactly as Amborella did, we used the 12-taxon,three-gene data set described earlier. We replaced Amborellarst with Nymphaea (of Nymphaeaceae) and then with Aus-trobaileya (of Austrobaileyales) and conducted parsimonyanalyses as described. Further analyses to evaluate the effectsof additional basal lineages were conducted, including one ormore of the following magnoliids: Magnolia , Myristica , Cin-namomum , Asarum , and Drimys . In analyses in which either

8/6/2019 Amborella Not a Basal Angiosperm- Not So Fast - Soltis

3/5

June 2004] 999SOLTIS AND SOLTIS A MBORELLA AND TAXON SAMPLING

T ABLE 1. Randomly selected pairs of monocots (of the 102 monocotsincluded in D. Soltis et al., 2000) used in replacement analyses totest the effects of monocot exemplars on the placement of Ambor-ella vs. the monocots in 12-taxon, three-gene analyses.

Monocot PairSister to Rest

of Angiosperms

Zostera and Glomeropitcairnia

Bomarea and ChamaedoreaConvallaria and Oncidium Blandfordia and Monocostus Ixiolirion and Nolina

Zostera

Amborella Amborella Amborella Amborella

Behnia and Zingiber Aristea and Pleea Restio and Trillium Juncus and TaccaChlorophytum and Dasypogon

Amborella Amborella Amborella Amborella Amborella

Nymphaea or Austrobaileya was substituted for Amborella , thesubstituted taxon was sister to Calycanthus , with no otherchanges to the topology: Nymphaea Calycanthus were sis-ters (but with 50% bootstrap support), and AustrobaileyaCalycanthus received 100% bootstrap support (tree notshown). Therefore, the Amborella Calycanthus sister groupreported by Goremykin et al. is not due to the close relation-ship of these taxa (per authors of the 19 th and 20 th centuries,as claimed by Goremykin et al.) but results from limited taxonsampling: a single representative of any of the three basalbranches of angiosperms forms a sister group with Calycan-thus . The addition of multiple representatives of these basalbranches (e.g., Amborella , Nymphaea , Austrobaileya ) plus ad-ditional magnoliids ( Magnolia , Myristica , Cinnamomum , Dri-mys, Asarum ) broke up the Amborella Calycanthus sistergroup, forming a clade in which Amborella , Nymphaea , and Austrobaileya are successive sisters to a clade of magnoliids,which includes Calycanthus . However, the monocots Oryza Zea remained sister to all other angiosperms, even when all

of these taxa were added (trees not shown).To test the effect of the selection of two grasses as exem-plars for monocots on the overall topology, we conducted aseries of analyses that included alternative monocot exemplars.We (1) replaced the grasses with two other monocots (replace-ment analyses) and (2) added other monocots to the grassesto increase the diversity of monocot lineages represented andto break up the branch leading from the base of the monocotsto the grasses (addition analyses).

Two types of replacement analyses were conducted. First,we substituted Acorus (the sister group to all other monocots;e.g., P. Soltis et al., 1999; D. Soltis et al., 2000; Qiu et al.,1999; Zanis et al., 2002, 2003; Chase et al., 2000) and Spa-thiphyllum (of Alismatales, a basal clade of monocots) for Zeaand Oryza . Although both Acorus and Spathiphyllum them-selves have long branches (144 steps from the common an-cestor of monocots to Acorus and 205 steps from the commonancestor of monocots to Spathiphyllum on one of the three-gene, 567-taxon trees of D. Soltis et al., 2000), they are notas long as that leading to the grasses (447 steps from the com-mon ancestor of the monocots to the common ancestor of thegrasses; D. Soltis et al., 2000). Second, we randomly selectedtwo monocots from the 102 monocots included in the three-gene analysis of D. Soltis et al. (2000) and substituted themfor Zea and Oryza ; this was repeated 10 times (Table 1).

If the position of the monocots as sister to all other angio-sperms in the Goremykin et al. analysis (and our 12-taxon,

three-gene analysis described earlier) is an artifact of the longbranch leading to grasses rather than the result of real phy-logenetic signal, then the position of the monocots should shiftto its more conventional location as the monocot branch isshortened by the addition of taxa. We included one to six ad-ditional monocots selected for their phylogenetic placementsto test the robustness of the monocots basal topology to

changes in taxon sampling. Further analyses included two tosix additional monocots, along with Nymphaea , Austrobaileyaand additional representatives of magnoliids ( Cinnamomum Magnolia , Myristica , Drimys , and Asarum ).

In the replacement analyses in which Acorus and Spathi- phyllum were substituted for the grasses Oryza and Zea , Amborella is the sister to all other angiosperms (tree not shown).However, Acorus and Spathiphyllum do not form a clade asexpected, presumably due to limited sampling of monocots(see later), but are the successive sister groups to Calycanthus

the eudicots (although bootstrap support for the Spathi- phyllum (magnoliids eudicot clade) is 50%). However,when Nymphaea and Austrobaileya are also added to the anal-ysis, Acorus and Spathiphyllum form a clade that is sister tothe eudicots (tree not shown). The same result was obtainedwhen additional magnoliids ( Cinnamomum , Magnolia , Myris-tica , Asarum , and Drimys ) were included.

When randomly selected monocots were chosen to replaceOryza and Zea , the topology shifted from monocots basalto Amborella basal. In all cases but one, Amborella was thesister to the rest of the angiosperms, and the monocot pair waseither sister to Calycanthus (one of the nine remaining anal-yses), sister to Calycanthus the eudicots (six), sister to theeudicots (one), or in a trichotomy with Calycanthus theeudicots (one). In the single case in which Amborella was notsister to all other angiosperms, the randomly selected mono-cots, Zostera (of Alismatales) and Glomeropitcairnia (of Bro-meliaceae, the sister group of grasses), were successive sistersto all other angiosperms; Amborella was not paired with Ca-

lycanthus , however, as it was when the grasses were the mono-cot exemplars, but was sister to Calycanthus the eudicots.In only two of 14 analyses in which other monocot taxa

were added to the grasses was the monocots basal topologyobtained; in the other 12, Amborella was sister to all otherangiosperms. The monocots basal result occurred when (1) Acorus and Spathiphyllum were added to Oryza and Zea asmonocot exemplars and (2) Acorus and Puya were added toOryza and Zea . Both analyses also included Nymphaea , Aus-trobaileya , Cinnamomum , Magnolia , Myristica , Drimys , and Asarum . However, in all other analyses, regardless of monocotsampling and whether other basal taxa were added, Amborellaremained sister to all other angiosperms. For example, the ad-dition of (1) the orchid Oncidium alone to Oryza and Zea (Fig.2), (2) Puya alone, (3) Acorus and Oncidium , (4) Acorus andPuya , (5) Acorus , Oncidium , and Puya , (6) Acorus and Spa-thiphyllum , and (7) Acorus , Spathiphyllum , and Puya to the12-taxon data set produced trees with the Amborella basaltopology. Likewise, the other ve analyses with combinationsof additional monocots and additional magnoliids found the Amborella basal topology (Appendix, see SupplementalData accompanying online version of this article).

In only one of 13 replacement analyses and two of 14 ad-dition analyses is the monocots basal topology retained; inall other cases, the Amborella basal tree is obtained. Thus,the contention of Goremykin et al. that monocots form thesister group to all other angiosperms cannot be supported.

8/6/2019 Amborella Not a Basal Angiosperm- Not So Fast - Soltis

4/5

1000 [Vol. 91A MERICAN JOURNAL OF B OTANY

Fig. 2. Single most parsimonious tree (length 2949 steps; CI 0.623;RI 0.459) obtained when the orchid Oncidium is included as an additionalmonocot exemplar to break up the long branch leading to the grasses. Boot-strap values are given above the branches. Branches without values received

50% bootstrap support.

More thorough phylogenetic sampling of monocots, repre-senting basal lineages and the sister group of the grasses, forexample, all yield the Amborella basal tree. The only ex-ceptions occurred when the basal monocots Acorus and Spa-thiphyllum or Acorus and Puya , respectively, were added todata sets containing Oryza , Zea , Nymphaea , Austrobaileya ,and additional magnoliids. In these two replacement cases, thelong branches of grasses, Acorus , Spathiphyllum , and Puyagrasses (e.g., D. Soltis et al., 2000) presumably distort rela-tionships at the base of the tree; the addition of Puya , of Bro-meliaceae, the sister group to all other Poales, does not break up the long branch to the grasses. Furthermore, even analysesthat included randomly selected pairs of monocots found the Amborella basal tree in nine of 10 replicates.

Conclusions The topology obtained by Goremykin et al.,with the monocots sister to all other angiosperms and Ambor-ella sister to Calycanthus , is not due to the increased charactersampling in their study (61 genes vs. a maximum in previousanalyses of 17; Graham and Olmstead, 2000), but rather tolimited taxon sampling. Evidence supporting this conclusionis (1) analysis of three genes for the same (or nearly the same)12 taxa (omitting Triticum ) yielded the Goremykin et al. tree;(2) replacement of Amborella with either Nymphaea or Aus-trobaileya yielded trees in which these alternatives were sisterto Calycanthus , and addition of Nymphaea , Austrobaileya , andmultiple representatives of magnoliids to the original matrix

broke up the Amborella Calycanthus sister group; and (3)sensitivity analyses of the monocots basal topology, basedon the inclusion of grasses as the only monocot exemplars inthe original matrix, demonstrated that this position of themonocots is not robust to alternative sampling of monocots.Stefanovic et al. (unpublished data, Indiana University) re-cently added a nearly complete plastid sequence for the mono-

cot Acorus to the data set of Goremykin et al. (2003). Paral-leling our results, Stefanovic et al. also recovered Amborellaas sister to other angiosperms. Thus, available data cannot re-fute the position of Amborella as sister to all other extant an-giosperms.

A series of studies involving all three genomes and reason-able taxon sampling has demonstrated that Amborella (or Am-borella Nymphaeaceae; Barkman et al., 2000) is sister toall other owering plants (e.g., D. Soltis et al., 1997, 2000; P.Soltis et al., 1999; Qiu et al., 1999; Mathews and Donoghue,1999; Parkinson et al., 1999; Graham and Olmstead, 2000;Zanis et al., 2002, 2003; Borsch et al., 2003; Hilu et al., 2003).These analyses have sampled many nucleotides and many taxaand have applied diverse approaches, including parsimony,neighbor-joining, maximum likelihood, Bayesian methods, andcompartmentalization. For example, the bootstrap support for Amborella and Nymphaeaceae as successive sisters to all otherangiosperms is 91 and 98%, respectively, and the posteriorprobabilities inferred from Bayesian analyses for these sametwo nodes are 0.99 and 1.00, respectively (Zanis et al., 2002).Two recent analyses of fast-evolving plastid regions (Borschet al., 2003; Hilu et al., 2003) only strengthen the argumentfor Amborella as sister to all other angiosperms. In fact, matK alone provides 86% bootstrap support for these same two basalnodes. Recent phylogenetic analyses of B-class oral genesnot only recover Amborella and Nymphaeaceae as sisters toall other angiosperms, but also reveal structural features thatindicate that Amborella alone is sister to all other angiosperms(Kim et al., 2004). All of these studies have in common a

reasonable sampling of taxa (mostly over 100 taxa), especiallyof basal lineages. This reminder of the importance of sufcientand appropriate taxon sampling is particularly timely as at-tempts to use whole organellar genomes for phylogeny recon-struction are underway in several labs. As exciting as the ge-nomic data are, extensive character sampling cannot compen-sate for inadequate taxon sampling.

LITERATURE CITED

APG II. 2003. An update of the angiosperm phylogeny group classicationfor the orders and families of owering plants: APG II. Botanical Journalof the Linnean Society 141: 399436.

BARKMAN , T. J., G. C HENERY , J. R. M CNEAL , J. LYONS -W EILER , AND C. W.DEPAMPHILIS . 2000. Independent and combined analysis of sequencesfrom all three genomic compartments converge to the root of oweringplant phylogeny. Proceedings of the National Academy of Sciences, USA97: 1316613171.

BORSCH , T., K. W. H ILU , D . Q UANDT , V. W ILDE , C . N EINHUIS , AND W.BARTHLOTT . 2003. Non-coding plastid trnT-trnF sequences reveal a wellresolved phylogeny of basal angiosperms. Journal of Evolutionary Bi-ology 16: 558576.

CHASE , M. W., D. E. S OLTIS , P. S. SOLTIS , P. J. RUDALL , M. F. F AY , W. H.HAHN , S. SULLIVAN , J. JOSEPH , T. G IVNISH , K. J. S YTSMA , AND C. P IRES .2000. Higher-level systematics of the monocotyledons: an assessment of current knowledge and a new classication. In K. K. L. Wilson and D.Morrison [eds.], Proceedings of monocots II: the 2nd International Sym-posium on the Comparative Biology of the Monocotyledons, 316. CSI-RO Press, Sydney, Australia.

CHASE , M. W. ET AL . 1993. Phylogenetics of seed plants: an analysis of

8/6/2019 Amborella Not a Basal Angiosperm- Not So Fast - Soltis

5/5

June 2004] 1001SOLTIS AND SOLTIS A MBORELLA AND TAXON SAMPLING

nucleotide sequences from the plastid gene rbc L. Annals of the Missouri Botanical Garden 80: 528580.

GAUT , B. S., B. R. M ORTON , B. M. M CCAIG , AND M. T. C LEGG . 1996. Sub-stitution rate comparisons between grasses and palms: synonymous ratedifferences at the nuclear gene Adh parallel rate differences at the plastidgene rbc L. Proceedings of the National Academy of Sciences, USA 93:1027410279.

GAUT , B. S., S. V. M USE , W. D. C LARK , AND M. T. C LEGG . 1992. Relative

rates of nucleotide substitution at the rbc L locus of monocotyledonousplants. Journal of Molecular Evolution 35: 292303.GOREMYKIN , V. V., K. I. H IRSCH -ERNST , S. W O LFL , AND F. H. H ELLWIG .

2003. Analysis of the Amborella trichopoda chloroplast genome se-quence suggests that Amborella is not a basal angiosperm. Molecular Biology and Evolution 20: 14991505.

GRAHAM , S. W., AND R. G. O LMSTEAD . 2000. Utility of 17 chloroplast genesfor inferring the phylogeny of the basal angiosperms. American Journalof Botany 87: 17121730.

GRAYBEAL , A. 1998. Is it better to add taxa or characters to a difcult phy-logenetic problem? Systematic Biology 47: 917.

H ILLIS , D. M. 1998. Taxonomic sampling, phylogenetic accuracy, and inves-tigator bias. Systematic Biology 47: 38.

H ILU , K. W., T. B ORSCH , K. M ULLER , D. E. S OLTIS , P. S. S OLTIS , V. S AVO -LAINEN , M. W. C HASE , M. P OWELL , L. ALICE , R. E VANS , H. SAUQUET ,C. N EINHUIS , T. S LOTTA , J. ROHWER , AND L. C HATROU . 2003. Angio-sperm phylogeny based on matK sequence information. American Jour-nal of Botany 90: 17581776.

K IM , S., V. A. A LBERT , M.-J. Y OO , J. S. F ARRIS , P. S. S OLTIS , AND D. E.SOLTIS . In press. Pre-angiosperm duplication of oral genes and reg-ulatory tinkering at the base of angiosperms. American Journal of Botany.

MAGALLO N , S., AND M. J. S ANDERSON . 2002. Relationships among seedplants inferred from highly conserved genes: sorting conicting phylo-genetic signals among ancient lineages. American Journal of Botany 89:19912006.

MATHEWS , S., AND M. J. D ONOGHUE . 1999. The root of angiosperm phylog-eny inferred from duplicate phytochrome genes. Science 286: 947950.

N ICKERSON , J. AND G. D ROUIN . 2004. The sequence of the largest subunitof RNA polymerase II is a useful marker for inferring seed plant phy-logeny. Molecular Phylogenetics and Evolution 31: 403415.

PARKINSON , C. L., K. L. A DAMS , AND J. D. P ALMER . 1999. Multigene anal-

yses identify the three earliest lineages of extant owering plants. Cur-rent Biology 9: 14851488.

POLLOCK , D. D., D. J. Z WICKL , J. A. M CGUIRE , AND D. M. H ILLIS . 2002Increased taxon sampling is advantageous for phylogenetic inference.Systematic Biology 51: 664671.

PRYER , K. M., H. S CHNEIDER , A. R. S MITH , R. C RANFILL , P. G. W OLF , J. SHUNT , AND S. D. S IPES . 2001. Horsetails and ferns are a monophyleticgroup and the closest living relatives to seed plants. Nature 409: 618622.

Q IU , Y.-L., J. L EE , F. B ERNASCONI -QUADRONI , D. E. S OLTIS , P. S. S OLTIS , MZANIS , Z. CHEN , V. S AVOLAINEN , AND M. W. C HASE . 1999. The earliestangiosperms: evidence from mitochondrial, plastid and nuclear genomes. Nature 402: 404407.

SAVOLAINEN , V., M. W. C HASE , C. M. M ORTON , D. E. S OLTIS , C. BAYERM. F. F AY , A. D E BRUIJN , S. SULLIVAN , AND Y.-L. Q IU . 2000. Phylo-genetics of owering plants based upon a combined analysis of plastidatpB and rbcL gene sequences. Systematic Biology 49: 306362.

SOLTIS , D. E., P. S. S OLTIS , M. W. C HASE , M. E. M ORT , D. C. A LBACH , MZANIS , V. S AVOLAINEN , W. H. H AHN , S. B. H OOT , M. F. F AY , M. A XTELL , S. M. S WENSEN , L. M. P RINCE , W. J. K RESS , K. C. N IXON , ANDJ. S. F ARRIS . 2000. Angiosperm phylogeny inferred from a combineddata set of 18S rDNA, rbcL and atpB sequences. Botanical Journal of the Linnean Society 133: 381461.

SOLTIS , D. E., P. S. S OLTIS , D. L. N ICKRENT , L. A. J OHNSON , W. H. H AHNS. B. H OOT , J. A. S WEERE , R. K. K UZOFF , K. A. K RON , M. W. C HASES. M. S WENSEN , E. A. Z IMMER , S.-M. C HAW , L. J. G ILLESPIE , W. J

KRESS , AND K. J. S YTSMA . 1997. Angiosperm phylogeny inferred from18S ribosomal DNA sequences. Annals of the Missouri Botanical Garden84: 149.

SOLTIS , P. S., D. E. S OLTIS , AND M. W. C HASE . 1999. Angiosperm phylogenyinferred from multiple genes as a research tool for comparative biology. Nature 402: 402404.

SWOFFORD , D. L. 2003. PAUP*4.0b10: Phylogenetic analysis using parsi-mony. Sinauer, Sunderland, Massachusetts, USA.

ZANIS , M. J., D. E. S OLTIS , P. S. S OLTIS , S. M ATHEWS , AND M. J. D ONOGHUE2002. The root of the angiosperms revisited. Proceedings of the National Academy of Sciences, USA 99: 68486853.

ZANIS , M. J., P. S. S OLTIS , Y.-L. Q IU , E. Z IMMER , AND D. E. S OLTIS . 2003Phylogenetic analyses and perianth evolution in basal angiosperms. Annals of the Missouri Botanical Garden 90: 129150.

ZWICKL , D. J., AND D. M. H ILLIS . 2002. Increased taxon sampling greatlyreduces phylogenetic error. Systematic Biology 51: 588598.