Proc. Natl. Acad. Sci. USAVol. 91, pp. 5163-5167, May 1994Evolution

Chloroplast and nuclear gene sequences indicate LatePennsylvanian time for the last common ancestorof extant seed plants

(noer/gym ne/molecular dock/rge subit of rlbulos,S php rboe/18S rRNA)

LouISE SAVARD*t, PENG LI*, STEVEN H. STRAUSSt, MARK W. CHASE§, MARTIN MICHAUD*,AND JEAN BOUSQUET*¶*Centre de Recherche en Biologie Forestitre, Facultd de Foresterie et de G6omatique, Universit6 Laval, Ste-Foy, Quebec, Canada GIK 7P4;tDepartment of Forest Science, Peavy Hall 154, Oregon State University, Corvallis, OR 97331-5705; and iDepartment of Biology,University of North Carolina, Chapel Hill, NC 27599

Communicated by M. T. Clegg, February 17, 1994

ABSTRACT We have estimated the time for the last com-moancor of extant seed plants by using molecular clocksconstructed from the ofthe chloroplastic gene codingfor the large subunit of ribulose-1,5-blphosphate carboxyl-ase/oxygenase (rbcL) and the nuclear gene coding for the snailsubunit of rRNA (Rrnl8). Phylogenetic analyses of nucleoddesequences indicated that the earliest divergence of extant seedplants is likely represented by a split between conufer-cycadandansperm linea . Relative-rate tests were used to assesshomogeneity of substution rates among lineages, and annualanglosperms were found to evolve at a faster rate than othertaxa for rbcL and, thus, these sequences were excluded fromconstruction ofmolecular clocks. Five distinct molecular cockswere calibrated using substitution rates for the two genes andfour divergence times based on fossil and published molecularcock estimates. The five mated times for the last commonancestor of extant seed plants were in agreement with oneanother, with an average of 285 million years and a range of275-290 million years. This implies a subsantily more recentancestor of all extant seed plants than suggsed by sometheories of plant evolution.

Although progymnosperms are accepted as the ancestralgroup to all seed plants, the time when the five groups ofextant seed plants (cycads, conifers, Ginkgo, Gnetales, andangiosperms) shared their last common ancestor is unclearbecause of uncertainty about ancestor-descendant relation-ships between progymnosperms and seed plants (1, 2). Beck(2-5), although not discussing angiosperms, proposed thatthe two main lineages of gymnosperms, coniferopsids (coni-fers, Ginkgo, and Gnetales) and cycadopsids (cycads), di-verged independently from two lineages of the progymno-sperms, with coniferopsids from Archaeopteridales and cy-cads from Aneurophytales. Because Archaeopteridales andAneurophytales appeared in the Middle Devonian, Beck'shypothesis implies that the two gymnosperm lineages di-verged about 375 million years (Myr) ago. Alternatively,Rothwell (6, 7) hypothesized that the two gymnospermlineages both evolved from Aneurophytales via the seed ferncomplex, suggesting a more recent diversification of extantseed plants.

Phylogenetic analysis of conserved gene sequences fromextant plant taxa has contributed tangibly to our understand-ing of the early diversification of seed plants (8-11). Chlo-roplast and nuclear gene sequences have also been used toconstruct molecular clocks and to estimate the times for keyevents in the evolutionary history of flowering plants (12).

However, before applying molecular clocks to date evolu-tionary events, approximate constancy of evolutionary rateover time needs to be established (13) by using procedures forassessing rate homogeneity over taxa (14, 15) or over lineages(16).Here we estimate the time when extant seed plants shared

their lastcommon ancestorby using molecular clocks derivedfrom nucleotide sequences ofthe chloroplast gene coding forthe large subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (rbcL) and the nuclear gene coding for the smallsubunit of rRNA (Rrnl8). 11 Using four different landmarksfor the calibration of molecular clocks, we found this time tobe between 275 and 290 Myr, thus 85-100 Myr later than thatimplied by Beck's hypothesis.

MATERIALS AND METHODSFor the chloroplast gene rbcL, the following nucleotidesequences were used: monocots (Triticum aestivum, Oryzasativa, and Sorghum bicolor) (17, 18), annual dicots (Pisumsativum, Spinacia oleracea, and Nicotiana tabacum) (19-21), perennial dicots (Magnolia macrophylla, Carpentariacalifornica, and Itea virginica) (22, 23), conifers (Pinusedulis, Pinus longaeva, Pinus griffithii, Pinus pinea, Pinusradiata, Pinus thunbergii, Pseudotsuga mensiezii, andPodocarpus gracitior) (24-26), cycads (Zamia inermis, Bow-enia serrulata, Cycas circinalis, and Encephantos arenarius;GenBank accession nos. L12683, L12671, L12674, andL12676), a fern (Angiopteris lygodiifolia) (27), a liverwort(Marchantia polymorpha) (28), and a green alga (Chla-mydomonas reinhardtii) (29) as the outgroup. Complete rbcLsequences for cycads were determined from both DNAstrands following previously established procedures (30). Todetermine the node of earliest divergence among extant seedplants sampled, phylogenetic analyses were conducted by theneighbor-joining method (31) with numbers of nonsynony-mous nucleotide substitutions corrected for multiple hits (32)and by parsimony analysis (33) of amino acid sequences inorder to consider the same level of sequence information.The numbers of synonymous substitutions could not be usedfor this study because they were saturated between seed

Abbreviations: Myr, million years; rbcL, gene coding for the largesubunit of ribulose-1,5-bisphosphate carboxylase/oxygenase;Rrnl8, gene coding for the small subunit of cytoplasmic rRNA.tPresent address: Institut de Recherche en Biologie Vdgdtale, Uni-versitd de Montrhal, 4101 Sherbrooke East, Montr6al, Qudbec,Canada H1X 2B2.1To whom reprint requests should be addressed.Il7he sequences reported in this paper have been deposited in theGenBank data base (accession nos. L12683, L12671, L12674,L12676, L00970, L07059, and L12667).

5163

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Nadl. Acad. Sci. USA 91 (1994)

plants and nonseed plants (greater than one) and sometimeswere inestimable. A bootstrap procedure (1000 replicates)based on the neighbor-joining analysis of numbers of aminoacid substitutions was performed using the program NJBOOT2(a gift from T. S. Whittam and M. Nei, Pennsylvania StateUniversity).For the nuclear gene Rrnl8, the following nucleotide

sequences were used: monocots (Oryza sativa andZea mays)(34), dicots (Alnus glutinosa, Lycopersicon esculentum, Gly-cine max, and Arabidopsis thaliana) (34, 35), conifers(Metasequoia glyptostrobordes and Picea mariana; Gen-Bank accession nos. L00970 and L07059), a cycad (Zamiapumila) (34), and Chlamydomonas reinhardtii (34) as theoutgroup. For Metasequoia glyptostroboldes and Picea mar-iana, DNA was isolated from lyophilized needles following aCTAB method (36) and the gene Rrnl8 was amplified in twooverlapping fiagments by PCR using two sets ofprimers: NS1and NS4; and NS5 and NS8 (37). Sequencing was conductedon single-stranded DNA products obtained from asymmet-rical amplification (38), and complete nucleotide sequenceswere determined from both DNA strands using primers NS1to NS8 (37). To determine the node of earliest divergenceamong extant seed plants sampled, phylogenies were esti-mated by the neighbor-joining method (31) with numbers ofnucleotide substitutions corrected for multiple hits with theone- and two-parameter methods (39, 40) and by parsimonyanalysis (33) of nucleotide sequences. A bootstrap procedure(1000 replicates) based on the neighbor-joining analysis ofnumbers of nucleotide substitutions estimated with the one-parameter method was performed with NJBOOT2.To estimate the time at the node of earliest divergence

among extant seed plants, molecular clocks were constructedby using numbers of nonsynonymous substitutions (32) forrbcL and by using one-parameter numbers of substitutions(39) for Rrnl8. For the gene rbcL, four distinct rates ofnonsynonymous substitution per site per year (molecularclocks) could be estimated from four landmark events: 1, thediversification of the Pinaceae (140 Myr ago) (41, 42) (eventEl, Fig. 1); 2, the split between monocots and dicots (200 Myrago) (event E2, Fig. 1)-this date was based on molecularclock estimates from chloroplast DNA (12), nuclear rRNAs(12), and mtDNA (P.L. and J.B., unpublished results), whichis about 75 Myr earlier than the oldest angiosperm fossils (1),but nearly 150 Myr later than other more controversialmolecular clock estimates (43-46); 3, the split between fernsand seed plants (395 Myr ago) (1) (event E3, Fig. 1); and 4,the split between liverwort and vascular plants (440 Myr ago)(47, 48) (event E4, Fig. 1). From the set of complete Rrnl8gene sequences available, we could estimate only one rate ofsubstitution per site per year from event E2 (Fig. 1).For a particular node-landmark used to derive a molecular

clock, an average substitution rate per site was obtained fromall possible pairwise sequence comparisons between the twogroups above the node-landmark. This rate was then dividedby 2T, to obtain a rate per site per year, where T is thelandmark time for the divergence between the two lineages.The standard error was calculated as the square root of theaverage variance of all pairwise sequence comparisons used.To obtain an estimated time at the node of earliest divergenceamong extant seed plants, the various rates per site per yearwere applied to half of the average (±SE) pairwise substitu-tion rate per site for that node, which was obtained similarlyas above.

Before constructing molecular clocks, lineage relative-ratetests (16) were used to assess homogeneity of substitutionrates. In these tests, the sequence from a green alga (Chia-mydomonas reinhardtii), a close relative to land plants (1),was used as the reference taxon. Ifa group of taxa was foundto evolve at a significantly different rate, it was excluded from

the construction of molecular clocks and in the estimation ofdivergence times.

RESULTSDeternination of the Node of Earlest Divergence Among

Extant Seed Plants. For rbcL, both neighbor-joining analysisof nonsynonymous rates of nucleotide substitution and par-simony analysis of amino acid sequences resulted in twoclades of extant seed plants: (i) cycads and conifers and (ii)angiosperms (Fig. 1). The bootstrap procedure based on theneighbor-joining analysis of numbers of amino acid substitu-tions showed moderate to strong support for the seed plantclade (bootstrap value of 80%o), the angiosperm clade (78%),and the conifer-cycad clade (67%). Therefore, node A (Fig.1) likely represents the node of earliest divergence amongextant seed plants analyzed. We also included the rbcLsequences from Ephedra tweediana (a Gnetales) (GenBankaccession no. L12667) and Ginkgo biloba (sequence of 93%of the gene; a gift from R. A. Price and J. D. Palmer, IndianaUniversity) in phylogenetic analyses to further ascertain thatnode A (Fig. 1) represents the earliest divergence amongextant seed plants. Neighbor-joining and parsimony analysesplaced Ginkgo in the conifer-cycad clade (bootstrap value of68%). Ephedra emerged as a sister group to angiosperms inneighbor-joining analysis of nonsynonymous rates of substi-tutions (bootstrap value of 39%o), while it was placed withinthe conifer-Ginkgo-cycad dade in parsimony analysis ofamino acid sequences. In no case was Ephedra found tobranch out before node A. Because of uncertainties on thephylogenetic position of Ephedra whether in the gymno-sperm or the angiosperm clade, and because the availableGinkgo sequence was <95% ofthe open reading frame, rbcLsequences from these two taxa were excluded from con-structing molecular clocks and estimating the time for theearliest divergence among extant seed plants.For Rrnl8, both neighbor-joining analysis of one- and

two-parameter numbers of nucleotide substitutions and par-simony analysis of nucleotide sequences resulted in twoclades of extant seed plants: (i) conifer and cycads and (ii)angiosperms. The bootstrap procedure based on the neigh-bor-joining analysis of numbers of nucleotide substitutionsestimated with the one-parameter method indicated a strongsupport for the conifer-cycad lade (bootstrap value of 95%)and moderate support for the angiosperm clade (68%). There-

E2 Angiosperms

E3 ConifersCycads

Ferns

Liverworts

FiG. 1. Consensus phylogenetic relationships among groups ofland plants deduced from analyses of rbcL and Rrnl8 gene se-quences. In the circles, A is the last common ancestor among extantseed plants, to be dated, and El to E4 represent the four landmarkevents used for calibrating molecular clocks.

5164 Evolution: Savard et A

Proc. Natl. Acad. Sci. USA 91 (1994) 5165

Table 1. Results from lineage relative-rate tests

Lineages compared Rate difference per site* ProbabilityrbcLPseudotsuga/Pinaceae -0.00247 ± 0.00325 0.447Pinaceae/Podocarpus -0.00011 ± 0.00435 0.976Pinaceae/cycads 0.00164 ± 0.00358 0.646Podocarpus/cycads 0.00175 ± 0.00413 0.674Perennial angiosperms/gymnosperms 0.00243 ± 0.00396 0.542Fern/perennial angiosperms + gymnosperms 0.00117 ± 0.00527 0.826Liverwort/perennial angiosperms + gymnosperms -0.00197 ± 0.00612 0.749Annual angiosperms/gymnosperms 0.01092 ± 0.00374 0.004Annual angiosperms/fern 0.01161 ± 0.00552 0.036Annual angiosperms/liverwort 0.01240 ± 0.00565 0.029

Rrnl8tAngiosperms/gymnosperms -0.00754 ± 0.00578 0.194Monocots/dicots -0.00323 ± 0.00532 0.542

Chlamydomonas reinhardtii was used as reference taxon in all tests.*For rbcL, nonsynonymous numbers of substitutions per site (14); for Rrnl8, one-parameter numbersof substitutions per site (39). A negative value indicates that taxon or group on the left side of thepairwise comparison has a slower rate of substitution, while a positive value indicates a faster rate ofsubstitution.

tFor Rrnl8, because the estimated numbers of nucleotide substitutions per site obtained from the one-and two-parameter methods (39, 40) were essentially the same, only results obtained with theone-parameter method are presented.

fore, node A (Fig. 1) was also confirmed by RrnJ8 genesequences as the likely earliest divergence among extant seedplants analyzed.

Assessment of Rate Homogeneity. Before constructing mo-lecular clocks, rate constancy over lineages was assessed fornonsynonymous rates of substitution for rbcL and one-parameter numbers of substitutions for Rrnl8. For rbcL,rates were found homogeneous within gymnosperms as wellas between gymnosperms, perennial angiosperms, fern, andliverwort (Table 1). As previously reported for nonsynony-mous sites of rbcL (24), annual angiosperms were found toevolve more rapidly than other taxa (Table 1) and, therefore,they were excluded from the construction of all rbcL molec-ular clocks. ForRrnl8, substitution rates were homogeneousbetween monocots and dicots as well as between an-giosperms and gymnosperms (Table 1).

Determination ofthe Time at the Node ofEarliest DivergenceAmong Extant Seed Plants. The four nonsynonymous rates ofsubstitution per site per year obtained for the chloroplastgene rbcL differed by <6% (Table 2). For this gene, the timewhen extant seed plants shared their last common ancestor(node A, Fig. 1) was estimated by dividing halfofthe averagenumber of nonsynonymous substitutions per site betweenperennial angiosperms and conifers and between perennialangiosperms and cycads (0.0264 ± 0.051) by the four esti-mated substitution rates per year (rbcL, El to rbcL, E4)(Table 2). The four time estimates ranged from 275 Myr to 290Myr (Fig. 2).For the nuclear gene Rrnl8, the time when extant seed

plants shared their last common ancestor (node A, Fig. 1) was

estimated by dividing half of the average number of substi-tutions per site between angiosperms and conifers and be-tween angiosperms and cycads (0.0803 ± 0.0075) by theestimated substitution rate per year (Rrnl8, E2) (Table 2).The estimated time obtained (287 Myr) was within the rangeof time estimates derived from rbcL molecular clocks (Fig.2).

Alternative Hypotheses. The time when extant seed plantsshared their last common ancestor was also estimated underthree alternative topologies, as suggested by cladistic anal-yses of morphological characters (49, 50) (Fig. 3). Theestimated times for topology I correspond to those in Fig. 2.Time for the earliest divergence among extant seed plantsunder topologies II and III (node A) was estimated in a similarfashion as for topology I. For topology IV with the chloro-plast gene rbcL, the time for the earliest divergence amongextant seed plants was obtained by dividing half of theaverage number of nonsynonymous substitutions per sitebetween perennial angiosperms and conifers, between pe-rennial angiosperms and cycads, and between conifers andcycads (0.0245 ± 0.0048) by the four estimated substitutionrates per year (rbcL, El to rbcL, E4, Table 2). For topologyIV with the nuclear gene Rrnl8, the time for the earliestdivergence among extant seed plants (node A) was obtainedby dividing half of the average number of substitutions persite (one-parameter method) between angiosperms and co-nifers, between angiosperms and cycads, and between coni-fers and cycads (0.0756 ± 0.0074) by the estimated substitu-tion rate per year (Rrnl8, E2, Table 2). For the topologysupported by our phylogenetic analyses of rbcL and RrnJ8

Table 2. Molecular clocks for rbcL and Rrnl8Divergence No. of nucleotide Substitution rate per

Gene Landmark event time, Myr substitutions per site* site per yearrbcL El, Pinus-Pseudotsuga split 140 0.0130 ± 0.0034 4.64 ± 1.21 x 10-11rbcL E2, Monocot-dicot split 200 0.0096 ± 0.0018t 4.80 ± 0.90 x 1O-1'trbcL E3, Fern-seed plant split 395 0.0361 ± 0.0058 4.57 ± 0.73 x 10-11rbcL E4, Liverwort-vascular plant split 440 0.0399 ± 0.0061 4.54 ± 0.69 x 10-11Rrnl8 E2, Monocot-dicot split 200 0.0558 ± 0.0059 1.40 ± 0.15 x 10-10*For rbcL, nonsynonymous numbers of substitutions; for Rrnl8, one-parameter numbers of substitutions.tFor molecular clock rbcL, E2, fast-evolving annual angiosperms were excluded (see text), and the rate per year wasestimated by dividing the average path length (0.0096) from the node at the origin of the dicots to extant perennial species(Magnolia, Carpenteria, and Itea), as estimated by the neighbor-joining method (results not shown; method described inref. 24), by the divergence time between monocots and dicots.

Evolution: Savard et aL

Proc. Natl. Acad. Sci. USA 91 (1994)

Time(myr)

Present -

100 -

200 -

300 -

400 -

5166 Evolution: Savard et al.

Geologicalepoch

Tertiary

Cretaceous

Jurassic

Triassic

Permian

PennsylvanianMississippian

Devonian

Silurian

Ordovician500 -t -

Cambrian

(nEc(U-J

Estimated time for event A:last common ancestor of

extant seed plants

Molecular clocks

wNw* w L .

1=-1 I

IA

U)

c ) U) U)40 4-e

~ 4", c 0

G)mV) U) 's~~~~~~~U

9, I. E E

FIG. 2. Estimated time when extant seed plants shared their lastcommon ancestor.

sequences (topology I, Fig. 3), the estimated times weregreater than those under the three alternative topologies (Fig.3).

DISCUSSIONPhylogenetic analyses of gene sequences from rbcL andRrnl8 resulted in two clades of extant seed plants: (i) cycadsand conifers and (ii) angiosperms (Fig. 1). This confirmedprevious results based on nuclear 18S rRNA and 26S rRNAsequences (8) as well as rbcL sequences (9). Given thatGinkgo and Ephedra (a Gnetales) did not branch out beforethe angiosperm-conifer-cycad lade (Fig. 1), a result alsoreported from morphological and molecular studies (8, 9, 49,50), the earliest divergence among extant seed plants is likelyto be represented by the split between the conifer-cycadlineage and the angiosperm lineage (node A in Fig. 1).

Ephedra rbcL sequence was not included in divergencetime estimates because of uncertainties on its position,whether within the gymnosperm or as a sister group to theangiosperm lade. Although phylogenetic analysis of 18SrRNA and 26S rRNA sequences and cladistic analysis ofmorphological characters placed Ephedra and other Gnetalesas a sister group to the angiosperms (8, 49, 50), results fromrbcL sequences are more variable, with the Gnetales as eithera sister group to the angiosperms (10) or nested within thegymnosperms (9).An important assumption for the calculation of our esti-

mated dates was the observed monophyly of conifers andcycads in the phylogenetic analyses, which was also sup-ported by previous analyses of nuclear 18S rRNA and 26SrRNA sequences (8) as well as sequences from the chloro-plastic genes rbcL (9) and chiB (R. Boivin, Laval University,personal communication). However, cladistic analyses ofmorphological traits did not support the monophyly of coni-fers and cycads and revealed several alternative topologieswith regard to conifers, cycads, and angiosperms (49, 50),which differed by only one or few steps. We have consideredthese alternative topologies in our study. When these alter-native topologies were considered with regard to the para-phyly or monophyly of conifers and cycads, the estimateddivergence times were all more recent, with some estimatespostdating the first reliable appearances of the five groups ofextant seed plants in the fossil record (cycads from the LowerPermian, conifers from the Lower Triassic, Ginkgo from thePermian, Gnetales as early as Middle Permian, and an-giosperms from the Lower Cretaceous) (1, 41, 49-51). Thus,estimates under topology I most likely represent the time forthe earliest divergence among extant seed plants.By using the various molecular clocks, the estimated times

when extant seed plants shared their last common ancestorwere in good agreement. For the chloroplast gene rbcL, thefour estimates come from four different landmark events andare highly congruent, while for the nuclear gene Rrnl8, theestimate for the initial divergence of extant seed plants waswithin the range of rbcL estimates. Because they are basedon different genomes and four calibration events, the odds ofobtaining such high congruence among the five estimates bychance alone are small.

I

Angiosperms

II III

,Angiosperms Angiospem

Conifers

Molecular clocks

rbcL-El

rbcL-E2

Rrn18-E2

rbcL-E3

rbcL-E4

284±55

275±53

287±27

289±55

290±56

262±52

253+50

265±27

266±53

268+53

Cycads

245+51

236±49

259±25

248+51

250±51

IV

ns Angiosperms

.SIF22*~. Cycads

Conifers

264±52

255±50

270±26

268±53

269±53

Estimated time for node A

FIG. 3. Estimated time when extant seed plants shared their last common ancestor under four topologies of conifers, cycads, andangiosperms. The estimated times for topology I correspond to those in Fig. 2.

I

Proc. NatL Acad. Sci. USA 91 (1994) 5167

We consider the more recent estimate derived from onemolecular clock (rbcL, E2) less reliable than the estimatedtimes from the other four molecular clocks (rbcL, El; rbcL,E3; rbcL, E4; and Rrnl8, E2). This is so because of slightacceleration of nonsynonymous rate of substitution of rbcLin perennial angiosperms (24), although not shown significantin our relative rate tests. Furthermore, it is unlikely thatsaturation effects contributed to the observed differencesamong rbcL molecular clocks, because all pairwise numbersof nonsynonymous substitutions per site used to constructthese clocks were below 0.05.Our results indicate that the last common ancestor to all

extant seed plants occurred most likely during the UpperPennsylvanian. This postdates by about 85-100 Myr thatimplied under Beck's hypothesis (2-5) and is slightly earlierthan the first appearances of the five groups of extant seedplants in the fossil record. Our estimated time when extantseed plants shared their last common ancestor supports theconclusion from cladistic and stratigraphic analyses of ex-tinct and extant taxa (49, 50) that all major groups of extantseed plants, with the possible exception of angiosperms, wereproducts of diversification that occurred between the LowerPennsylvanian and the Upper Triassic (51). The recentness ofthis event also implies that all extant seed plants, includingconifers and cycads, are likely to have evolved from seedferns, which derived from one progymnosperm lineage, mostprobably the Aneurophytales (6, 7). The identification of theseed fern group from which extant seed plants derived awaitsfurther studies on the morphology and relationships of Palae-ozoic and Mesozoic seed ferns (51).

We thank R. A. Price and J. D. Palmer (Department of Biology,Indiana University) for use of their unpublished Ginkgo rbcL se-quence, M. Nei and T. S. Whittam (Institute of Molecular Evolu-tionary Genetics, Pennsylvania State University) for providing com-puter programs, and anonymous reviewers for their suggestions toimprove the manuscript. This work was supported by Qu6becResearch Fund (FCAR) and Canadian Natural Sciences and Engi-neering Research Council grants to J.B., U.S. National ScienceFoundation grants to S.H.S. and M.W.C., and a Canadian NaturalSciences and Engineering Research Council fellowship to P.L.

1. Stewart, W. N. & Rothwell, G. W. (1993) Paleobotany and theEvolution of Plants (Cambridge Univ. Press, Cambridge,U.K.), 2nd Ed.

2. Beck, C. B. & Wight, D. C. (1988) in Origin and Evolution ofGymnosperms, ed. Beck, C. B. (Columbia Univ. Press, NewYork), pp. 1-84.

3. Beck, C. B. (1966) Taxon 15, 337-339.4. Beck, C. B. (1970) Biol. Rev. 45, 379-400.5. Beck, C. B. (1971) Am. J. Bot. 58, 758-784.6. Rothwell, G. W. (1981) Rev. Palaeobot. Patynol. 32, 103-121.7. Rothwell, G. W. (1982) Rev. Palaeobot. Palynol. 37, 7-28.8. Hamby, R. K. & Zimmer, E. A. (1992) in Molecular System-

atics of Plants, eds. Soltis, P. S., Soltis, D. E. & Doyle, J. J.(Chapman & Hall, New York), pp. 50-91.

9. Hasebe, M., Kofuji, R., Ito, M., Kato, M., Iwatsuki, K. &Ueda, K. (1992) Bot. Mag. Tokyo 105, 673-679.

10. Chase, M. W., Soltis, D. E., Olmstead, R. G., Morgan, D.,Les, D. H. et al. (1993) Ann. Mo. Bot. Gard. 80, 528-580.

11. Clegg, M. T. (1993) Proc. Natl. Acad. Sci. USA 90, 363-367.12. Wolfe, K. H., Gouy, M., Yang, Y.-W., Sharp, P. & Li, W.-H.

(1989) Proc. Natl. Acad. Sci. USA 86, 6201-6205.13. Kimura, M. (1968) Nature (London) 217, 624-626.14. Wu, C.-I. & Li, W.-H. (1985) Proc. Natl. Acad. Sci. USA 82,

1741-1745.15. Muse, S. V. & Weir, B. S. (1992) Genetics 132, 269-273.

16. Li, P. & Bousquet, J. (1992) Mol. Biol. Evol. 9, 1185-1189.17. Terachi, T., Ogihara, Y. & Tsunewaki, K. (1987) Jpn. J. Genet.

62, 375-387.18. Doebley, J., Durbin, M., Golenberg, E. M., Clegg, M. T. &

Ma, D.-P. (1990) Evolution 44, 1097-1108.19. Zurawski, G., Whitfeld, P. R. & Bottomley, W. (1986) Nucleic

Acids Res. 14, 3975.20. Zurawski, G., Perrot, B., Bottomey, W. & Whitfeld, P. R.

(1981) Nucleic Acids Res. 9, 3251-3269.21. Shinazoki, K. & Sugiura, M. (1982) Gene 20, 91-102.22. Golenberg, E. M., Giannasi, D. E., Clegg, M. T., Smiley,

C. J., Durbin, M., Henderson, D. & Zurawski, G. (1990)Nature (London) 344, 656-658.

23. Soltis, D. E., Soltis, P. S., Clegg, M. T. & Durbin, M. (1990)Proc. Natl. Acad. Sci. USA 87, 4640 -4644.

24. Bousquet, J., Strauss, S. H., Doerksen, A. H. & Price, R. A.(1992) Proc. Natl. Acad. Sci. USA 89, 7844-7848.

25. Mukai, Y., Yamamoto, N., Odani, K. & Shinohara, K. (1991)Plant Cell Physiol. 32, 273-282.

26. Hipkins, V. D., Tsai, C.-H. & Strauss, S. H. (1990) Plant Mol.Biol. 15, 505-507.

27. Yoshinaga, K., Kubota, Y., Isshii, T. & Wada, K. (1992) PlantMol. Biol. 18, 79-82.

28. Ohyama, K., Fukazawa, H., Kohchi, T., Shirai, H., Sano, T.,Sano, S., Umesono, K., Shiki, Y., Takeuchi, M., Chang, Z.,Aota, S.-i., Inokuchi, H. & Ozeki, H. (1986) Nature (London)322, 572-574.

29. Dron, M., Rahire, M. & Rochaix, J.-D. (1982) J. Mol. Biol. 162,775-793.

30. Albert, V. A., Williams, S. E. & Chase, M. W. (1992) Science257, 1491-1495.

31. Saitou, N. & Nei, M. (1987) Mol. Biol. Evol. 4, 406-425.32. Li, W.-H., Wu, C.-I. & Luo, C.-C. (1985) Mol. Biol. Evol. 2,

150-174.33. Swofford, D. L. (1993) PAUP-Phylogenetic Analysis Using

Parsimony (Illinois Natural History Survey, Champai, IL),Version 3.1.

34. Neefs, J.-M., Van der Peer, Y., De Rjk, P., Goris, A. & DeWachter, R. (1991) Nucleic Acids Res. 19, 1987-2015.

35. Savard, L. & Lalonde, M. (1991) Plant Mol. Biol. 16, 725-728.36. Bousquet, J., Simon, L. & Lalonde, M. (1990) Can. J. For. Res.

20, 254-257.37. White, T. J., Bruns, T., Lee, S. & Taylor, J. (1990) in PCR

Protocols: A Guide to Methods and Applications, ed. Erlich,H. A. (Academic, San Diego), pp. 315-324.

38. Gyllensten, U. B. & Erlich, H. A. (1988) Proc. Natl. Acad. Sci.USA 85, 7652-7656.

39. Jukes, T. H. & Cantor, C. R. (1969) in Mammalian ProteinMetabolism, ed. Munro, H. N. (Academic, New York), pp.21-132.

40. Kimura, M. (1980) J. Mol. Evol. 16, 111-120.41. Miller, C. N. (1977) Bot. Rev. 43, 217-280.42. Miller, C. N. (1988) in Origin and Evolution ofGymnosperms,

ed. Beck, C. B. (Columbia Univ. Press, New York), pp.448-486.

43. Martin, W., Gierl, A. & Saedler, H. (1989) Nature (London)339, 46-48.

44. Crane, P. R., Donoghue, M. J., Doyle, J. A. & Friis, E. M.(1989) Nature (London) 342, 131.

45. Brandl, R., Mann, W. & Sprinzl, M. (1992) Proc. R. Soc. Lond.B 249, 13-17.

46. Martin, W., Lydiate, D., Brinkmann, H., Forkmann, G.,Saedler, H. & Cerff, R. (1993) Mol. Biol. Evol. 10, 140-162.

47. Taylor, T. N. (1988) Taxon 37, 805-833.48. Shear, W. (1991) Nature (London) 351, 283-289.49. Doyle, J. A. & Donoghue, M. (1986) J. Bot. Rev. 52, 321-421.50. Crane, P. (1985) Ann. Mo. Bot. Gard. 72, 716-793.51. Crane, P. R. (1988) in Origin and Evolution ofGymnosperms,

ed. Beck, C. B. (Columbia Univ. Press, New York), pp.218-272.

Evolution: Savard et al.