Boise State University ScholarWorks

Biology Faculty Publications and Presentations Department of Biological Sciences

1-1-1993 Phylogenetics of Seed : An Analysis of Nucleotide Sequences from the Plastid Gene rbcL James F. Smith Boise State University

This document was originally published by Biodiversity Heritage Library in Annals of the Botanical Garden. Copyright restrictions may apply. Image courtesy of Biodiversity Heritage Library. http://biodiversitylibrary.org/page/553016 See article for full list of authors. PHYLOGENETICS OF SEED Mark TV Chase,' Douglas E. Soltis,' PLANTS: AN ANALYSIS OF R ichard C. Olmstead;' David Morgan,J Donald H Les,' Brent D. Mishler," NUCLEOTIDE SEQUENCES Melvin H, Duvall,' Hobert A. Price,' FROM THE PLASTID Harold G. Hills,' Yin· Long Qiu,' Kathleen A. Kron,' Jeffrey H. Hell.ig,' GENE rb ell Elena Cont.':, 10 Jeffrey D. Palmer, 1l James R. ft1an/wrt,9 Kenneth}. Sylsma, 1O Helen J . Michaels," W. John Kress, " Kenneth C. Karol, 1O W. Dennis Clark, 13 lvlikael J-I edren,u Brandon S. Gaut,? Robert K. J ansen, l ~ Ki.Joong Kim ,'5 Charles F. Jf/impee,5 James F. Smilh,I2 Glenn N. Furnier,lo Steven 1-1. Strauss,l? Qiu- Yun Xian g,3 Gregory 1.1. Plunkeu,J Pamela S. Soltis ,:" Susan M. Swensen, ]!J Stephen /, ', Williams," Paul A. Gadek,'" Christopher }. Quinn,20 Luis E. /:.guiarle,7 Edward Golenberg," Gerald 1-1. Learn, jr.,7 Sean W. Graharn,22 Spencer C. H. Barreu," Selvadurai Dayanandan,23 and Victor A. Albert'

AIJSTRACf

We present the results of two exploratory parsimony analyses of DNA sequences from 475 and 499 of seed plants. respectively. representing all major taxonomic groups. The data are exclusively from the chloroplast gene rbeL. which codes for the large subunit of ribulose-l .5-bisphosphate carboxylase/ oxygenase (RuBisCO or RuBPCase). We used two different state-transformation assumptions resulting in two sets of cladograms: (i) equal.weight ing for the 499-taxon analysis: and (ii) a procedure that differentially weights transversions over transitions within characters and codon positions among characters for the 475·taxon analysis. The degree of congruence between these results and other molecular. as well as morphological, cladistic studies indicates that rbc L sequence va riation contains historical evidence appropriate for ph yloge netic analysis at this taxonomic level of sampling . Because the topologies presented are necessarily approximate and cannot be evaluated adequately for in ternal support, these results should be assessed from the perspective of their predictive value and used to direct future studies. both molecular and morphological. In both analyses. the three genera of Gnetales are placed together as the sister group of the no ..... ering plants. and the anomalous aquatic Ce ratoph),lIllm (Ceratophyll aceae) is sister to all other flo ..... ering plants. Several major lineages identified correspond well with at least some recent taxonomic schemes fo r angiosperms. particularly those of Dahlgren and Thorne. The basalmost within the angiosperms are orders of the apparentl y polyphyletic subclass Magnoliidae sensu Cronquist. The most conspicuous feature of the topology is that the major di vision is not monoco! versus di cot. but rather one correlated with general pollen type: uniaperturate versus triaperturate. The Dilleniidae and Humamelidae are the only subclasses that are grossly polyphyletic; an examination of the la tter is presented as an example of the use of these broad analyses to focus more restricted slUdies. A broadly circumscribed Rosidae is para phyletic to Asteridae and Dilleniidae. Subclass Caryophyllidae is monophyletic and derived frolll within Rosidae in the 475-taxon analysis but is sister to a group composed of broadly delineated Asteridae aud Rosidae in the 499-taxon study.

, The authors acknowledge the following support: U.S. National Science Foundation (NSF) grant BSR-8906496 to MWC; 8SR·9007614 to DES; BSR·900232I to MRD; 8SR·9107827 to RGO; 8SR·881 7992 to DHL "nd CFW; BSR·9107484 to 8DM; BSR·9007293 and BSR·9020055 to KJS; 8SR·882 1264 to KAK; 8SR·8717600 "nd 8SR . 8996262 to lOP; 8SR·9020171 to RKJ ; BSR·8817953 to SHS "nd RAP and BSR·8957023 to SHS; BSR·880193 to W. Alverson; a Sloan Postdoctoral Fellowship to KJK: BSR-89l4635 to VAA; a grant fr om the American Philosophical Society to SEW; Australian Research Council grant AD903 185 1 to PAG and CJQ: an operating grant to SCHB and a graduate scholarship to SWG from the Natura l Sciences & Engineering Research Council of Canada; a grant to MWC and doctoral fellowship to VAA fr om the American Orchid Society; and a postdoctoral scholarship

ANN. MISSOURI BOT. CARD. 80: 528- 580. 1993. Volume 80, Number 3 Chase et al. 529 1993 Phylogenetics of Seed Plants

Curre nt assessments of hi ghe r-level relationships taxa fr om 135 families, slill less than a thi rd of in seed pla nts a re based largely on informed judg­ angiosperm families. I f insuffi cient sampling of taxa ments of the relati ve value of variolls reproductive or characters (i.e., sequence length, acknowledged a nd vegeta tive characters (including secondary as a problem with the rbcS data studied by Martin chemistry) and to some exte nt on historical pre­ & Dowd, 1991) are indeed fa ctors, then no valid cedent. Authors of recent taxonomic schemes (for investigation of the "informativeness" of a given example. Dahlgren. 1980; T ak hlajan. 1980, 1987, gene sequence exists for seed plants. Thus we are Cronquist, 198 1 ~ Thorne, 1983, 1992) ha ve syn­ le ft wi th only an unfounded assessment of sequence thesized an enormous a mou nt of information. Ne v. data as having the potential to aid in estimating ertheless, their taxonomic decisions have been guid­ higher. level relationships. ed by estimations of whic h characters a re reliable Suggestions that the chloroplast gene rbcL, whi ch indica tors of relationships. These diffe ring judg­ codes for the la rge subunit of ribulose· I ,5-bis­ ments are responsible for radical differences in phosphate carboxylase/ oxygenase (RuBisCO or delimitation and relative ra nks of taxa in each RuBPCase), was an appropriate locus to use in system of classification. Recently, a number of phylogenetic studies began with Ritland & Clegg explicit, cladistic hypotheses have been developed (1987) a nd Zurawsk; & Clegg (1987). In;l;al al· at inclusive hie rarchi cal levels (e.g., Crane, 1985, tempts to evaluate relationships used only a doze n 1988; Dahlgren & Bremer, 1985; Dahl gren ct al. , or so sequences representin g a ll land plants (Palm er 1985, Doyle & Donoghue, 1986. 1992, Bremer et a I. , 1988; Ciannasi et al. , 1992). Other rcccnt et aI., 1987; Donoghue & Doyle, 1989; Loconte studies have been restricted to single families (Doc­ & Stevenson, 1991; Martin & Dowd, 1991; Ham­ bley et al., 1990; Kim et aI., 1992) or putatively by & Zimmer, 1992; Hufford , 1992; Olmstead et closely related families (Soltis et aI., 1990; Les et al.. 1992, Taylor & H; ckey. 1992). Such clad;sl;c aI. , 1991; Donoghue et a I. , 1992; Olmstead et al. . studies have previously been limited in scope; some 1992, Rett;g el a I. , 1992). Mosl of Ih e la ttce data matrices contain significant taxonomic gaps, studies began the process of incorporating signifi ­ and in others characters for some taxa are missing. cantl y greater sampling to enhance their phylo­ Both of these fa ctors may have unpredictably mis­ genetic perspec ti ve. The use of rbcL was spurrcJ leading effects (Nixon & Davis, 199 1; Plat nick et by C. Zurawski, wh o generously made avai lable a aI., 1991). Despite a great deal of investigation set of internal sequencing primers. The advent of and analysis, seed. phylogenetics is, at best, temperature cyclers and hi gh.temperature- resis. in a prelimina ry stage of in vestigation and knowl­ tant DNA polymerases (sometimes terme<\ Poly. edge. merase Chain Reaction or PCR) has greatly en­ Molecular data, spec ifi cally DNA sequences, hanced rates at wh ich gene sequence data a rc ha ve received a great deal of attent ion as a potential accumulating, so that effects on phylogenetic es­ source of "phylogenelically informative" cha rac­ timates of more intensive sampling of sequence ters Iha t are putatively less ambiguous than non· variation can now be investigated. molecular characters. Such pronouncements suffer We wish to examine here the degree to whi ch from the limitation that, at higher taxonomic levels, a representative sa mpling of sequence variation for no extensive sampling and phylogenetic description rbc L contains evidence of the evolutionary history of DNA sequence variation has taken place. The of seed plants. In this stud y, we address the

from the OGAPA. Univcrsidad Nacional Autonoma de Mexico. Mexico. to LEE. We also thank David L. Swofford and John M. Mercer fo r assistance and advice in running the parsimony analyses; William S. Alverson. Michael T. Clegg. Juli e Dowd . E. Allcn Hcrre. Peter G. Martin. and James E. Rodman. for access to their unpublished sequences; Jeffrey 1. Doyle. Robert f . Thorne. Peter Westin. Michael J. Donogh ue. and several allon ymous reviewers for their helpful suggestions; and the Royal Bota nic Gardens. Kew. for support in the fin al stages of the prepara tion of the lila II uS(: n pl. footnotes 2- 23 follow the Literature Cited. 530 Annals of the Missouri Botanical Garden

pressed in cladograms a re diffi cult 10 relate direc tly Cornaceac, . Ma gnoli aceac, Zi ngibe ra nae). 10 diagrams and state me nt s of progenilor/ descclI + wh ereas ot hers are poo rl y sa mpled (e.g., di llc niid dan. rela tionships ust:d ill llIallY taxonomic sche mes. orders, especia l1 y a nd Theales). Despi te Dahlgre n ci al. ( 1985) a nd Da hl gren & Bre me r lac k of a coordina tcd effort , all subclasses and (1985) have published analyses most similar 10 the orders ha ve a t least some representati ves. Al1 spe­ ones presented here but they are no l of a similar cies that are used in this issue of the A"nals 0/ scope . Allhough some specific topological compo· IIIi' ,\'Iissollri /J ol.aniraf Carden are listed a lpha­ nents can be compa red to other cladistic sltHJic:; be ti call y by ramily in a fin al Appendix along wi th (Con ti e l a L, 1993: Rodman e l <1 1. , 1993: Qiu c t other information conce rning voucher sta tus, se· a I., 1993; and several of the ot her studies in this que nce gaps, lit erature c ita tions ror published se­ issue). morphological phylogene tic studies a t this que nces, and fi gures or this paper in whic h each level wi th simi la r, broad taxonomic sa rnpli ng do taxon occurs. Some laxa thai were included in the nol exisl. The computa tional diffi culties of evalu­ tJ.75 -taxon da ta se t we re excluded fr om the 499- ating inte rnal support (e.g .. the bootstrap. r eisen­ taxon ma trix . Thus. in the second a nalysis ce rtai n stein, 1985 ; decay analysis. Bre ine r, 1988) or fa milies ha vc fewer representatives. but th e overa ll de parture from matrix ralldomness (Archie, 1989: representa tion of lineages is greate r. Fai th & Cranston, 1991 ; Kiille rsjo e t aI. , 1992) Due to the large nu mber or laboratories tha t for Sti ch la rge numbe rs of taxa li ke wise prevent Ll i'l contributed unpublished seque nces, no sta ndard­ from addressin g extensively th ese issues he re . \Ve ized procedure wa s used to produce Ih e sequences do present an analysis of families traditionall y re­ analyzed he re. A generali za tion would be the fol. fe rred to I-Iamamelidae as an example of how ill ­ lowing: a fragmcnt containing rbd. was amplified te rnal support for specifi c clades Inight be exam­ from a total DNA extrac t using primers tha t fl a nk ined. (See also othe r papers in this issue th a t address or a re near the ends of the coding region: this internal a nd e xt ernal support for subse ts of the fr agme nt was the n directl y sequenced usin g one of general results.) Thus. the broad relationships de­ scveral differe nt procedures or was clolled usin g sc ribed he re can be lI sed to foc us IHo re restric ted standa rd recombina nt DNA techniques; dideoxy (with fe wer taxa). a lld the refore Itlore rigorous, sequencing gene rall y included both strands for a t in vest iga I iOlls. least . I~ of the minimally 1428-bp gene. SOllie work ­ W e expect that patterns prese nted here wi ll ers used more closel y spaced primers to seque nce c ha nge somewha t as seque nces of morc spec ies a rc onl y one strand of DNA : eit hc r strategy uppea rs added or if Illethodologit:al improvelli ellts pe rmit to provide rea80nably error-free se(luenccs. Mosl exact solut ions (for a disclIi'sion of progress, sec ex traction proto('ols relied on fr esh or freshl y dried Pe nny el a l. . 1992). These results none theless have lear samples, but some sarnples were ampli fi ed from great va lue. both frOIn he uristic a nd rne thodological DNA ex tracted fr om he rbarium specim ens as old pe rspectives, alt hough the prelirninary nature of as 20 years. these studies precludes a de ta il ed e xa mi na tioll of Ampli fi cation or rllC L from some taxa produced implications for seed-pl a nt ta xonomy a nd r-haractr-r two differe nt products. Somc of these produc ts we re e\'olution. as well as in vestigations of gene a nd different e nough in size to observc on a ll a garose protein evolution . gc l. whereas othcrs wc re dctcc ted initia ll y because We have generally foll owed the taxonomic cir­ lIIultiple bands occurred a t the sa me points in a u­ cumsc riptions of Cronquist ( J 981 ) for di cots bc­ toradiograms, indicating thnt more thnn one telll ­ cause this systcm COli forms closely to those used plate wa s presc nt. Nearl y all cycads produced two in IHOst textbooks alld flo ras. For mOllocots. WI:' loci (Hill s & Chase. unpublished), wh ic h we re sep­ have adopted the syste m of Duhlgrcn e t al. (1985) arately clolled a nd sequenced to charac te rize both but have changed the supe rorder ending "-inorac" copics. In cycads examincd, one copy contai ncd used by Dahlgren to the more a ppropriate " -anac" deletions tha t disruptcd the reading frame. a n in­ (Thorne. 1992). dication that this copy ma y represent a "psc udo­ gene." In Convolvulaceae. Olmstead (unpublished) de tected two copies of rbe L. one copy of wh ic h

contained dele ti ons. In (:(1(1('1/(1 (Cane lla ccae; Qiu M ATE:tHAt..5 AND M ETH O D~ e l a1.. 1993) a nd Ca/pliimi(l (: M. Selection of ta xa ror tltis study wa .'; 1I0t guided W . Chase, 1-1 . C. Hills & W . H. Ande rson, UIl ­ by 8. spec ifi c plan. Close e xa mination or gene ra published), two size -consc rved copies or rl)(" L wcre included in the a nalysis wi ll reveal a ll uneven tax ­ also e ncountered. In C(l/pitimi(/, one copy clearl y onomic distribution; some groups a re wel l reprc­ ('ontained substitutions a t numerous sit es othc rwise se nted (e.g .. Aste raceae a nd Aste ridae ill ge neral. conse rved a mong angiospe rms. sugges ting that this Volume 80, Number 3 Chase et al. 531 1993 Phylogenetics of Seec Plants

copy wa s a pseudogenc. In CandIa, no such un­ Although the assumptions of the Albert et aJ. model usual substitutions we re observed in eit her copy. are admittedly simplisti c, we nonetheless support Two, reading-frame-i ntact copies of rbe l are also the investigation of weighting approaches to nu­ reported in Ulmaceae (E. Conti & K. J. Sytsma. cleotide data and view the model used here as a unpublished). For Canella. we included bot h se­ justifiable first approximation. It must be recog­ quences in all analyses, hut because the Iwo se­ nized that giving a ll categories of molecular change que nces a re always eac h other's siste r we show equal transforma tional weight is also an assump­ only the position of "Cari ello" (in fact, there arc tion, but one that the investiga tion of Albert et al. two terminals here; the comple te ma trix used in (1993) found to be adequate if sampling effects Search II thus has 500 termi nal s but onl y 499 were not a fa ctor. Actual weights applied in this taxa). study faU wit hin stich a narrow and minimally Sequences of rbel we re easil y a li gned by sighl. asymmetrical range that deviation of results fr om Among la nd plants, the coding region contains litt le those using the " equal weighting" criterion of Fitch size variation through base position 1428 (num­ ( 197 1) ;5 I;kely to be sl;ght (see Albert et aI., 1993). be red from the first nucleotide of methi oni ne in the Hecently, concern has focused on the probabilit y start cod on, A UC). Positions 1426, 1427, and that "islands" of equally parsimonious exist, 1428 are the most common stop-codon among land partic ul arl y in large data sets (Maddison, 199 1). plants: longer reading frames, up to 1452 in some Because of their enormous size, ou r analyses do 1Il0nocots and 1458 in the (Kim el aI., not use methods of multip le random taxon-addition, 1992). appear to be due to insertions, oft en of a whi ch ha ve been suggested to uncove r sti ch dis­ short repeating sequence. Most laboratories ha ve junct, equally optimal islands. This topic is ad­ sequenced past this codon, but for phylogenetic dressed in most empirical pape rs in this number. analysis we have terminated a ll sequences at po­ The search for pa rsimonious trees consisted of sit ion 1428 to be confid ent tbat we have analyzed several separate but li nked heuristic searches using hornologous regions (the portion of the gene in­ either PA UP 3.0r (Sear ch I; 475 taxa) or 3.0s cluded in this analysis for each species is also (Search II ; 499 taxa). All searches included the included in the Appendix). All sequences were en­ full data matrix (a ll codon posit ions). Search I was tered illto a text fli e in NEXUS format (used by performed on a S PAHC II (Sun, Inc.) wo rk station PA UP 3.0; Swofford, 199 1) as complete sequences (PA UP 3.0 for non -Macintosh computers is avail ­ and then analyzed directl y in nucleotide form . Ma­ able only by special arrangement with D. Swofford). trices used in both searches are availa ble from the for Search II , a Macintosh Quadra 800 wi th 20 fir st author upon receipt of request and a diskette MB RAM was used. (A lthough slower than a S UIl for each matrix. computer, the more interacti ve na ture of a search on a PC is preferred by many workers.) In Search I , an init ial heuristic search wi th character-state changes given equal we ight (i.e .. THEE-SEA RCH STRATEG Y "unordered" status), S IMPLE data addition se­ Parsimony-based methods permit direct exam­ quence, STEEPEST DESCENT, and NN I (nearest­ ination of hypothesized cha racter-s tate changes on neighbor interchange) bra nch swapping algoritllll"l the reconstructed . and Ihis information can was used to find a single tree (M LPARS oplion be used in studies of molecular evolution. (These deacti vated). The second phase used this single are. however, likely to be underestimates of se­ tree obtained fr om phase one as a starting tree for quence change and could be misleading for this another heuristic search, this ti me using the S PH reason.) Numerous empirical studies have shown (subtree pruning- regrafting) branch swapping al­ that not a ll classes of substi tutions are equally li ke· gorithm wit h MULPARS deacti vated again. The Iy. and this kind of in formation may be incorpo­ third phase para lleled the fir st and second; the rated into weightng schemes for nucleotide data sin gle S PR tree from phase two was used as the (Swolford & Ol sen, 1990; Albert & M;shler, 1992: starting topology for a heuristic search using the among others). Various models of molecula r evo­ TBR (tree bisec tion-reconnect ion) branch swapping lution exist, and appropriately circumspect use of algorithm, again with MULPARS deac ti va ted . these may assist in the separa tion of historical signal The fourth phase used the sin gle TBR tree as from homoplasy. The characte r- state weight ing the starting point for a heuristic search employing model of Albert et al. (1993, this issue) uses prob­ a cha racter-slate weighting crit erion (wi th a dif­ abilit y formulae to calculate weights for different ferent step matrix for each codon positi on; cf. classes of molecul ar change. \Ve have used their Swofford , 199 J; Albert & M ishler, 1992: Albert method in the 475-taxon analys is presented here. et aI., 1993). This time MULPAHS wa s activa ted . 532 Annals of the Missouri Botanical Garden

Transversion substitutions were weighted over duced by using the CLOSEST addition sequence transitions differentially by codon position (Albert wi th the HOLD option set for fi ve trees (this in el al.. 1993). The specific weights used were the effect permitted initial swapping on se\'eral differ­ following: fOT transitions, 5520 (fi rst positions), ent starting topologies). Approximately 120 hours 6368 (second positi ons), 4039 (third positions); for were required merely to add a ll taxa in this manner. 1r ansversions, 6620 (fir st posit ions), 7470 (second The initial search 0) used NN I swapping and positions), 5 127 (third positions). These weights STEEPEST DESCENT with M ULPARS olf. The were calculated from empi rically deri ved param­ shortest single tree found was then swapped on eters (see Albert et al., 1993, for a complete de­ using (ii) TBH , whi ch generall y found a shorter scription and justifi cation). Because use of step tree, at which time (iii) NN I (with MULPAHS) wa s matrices is CP U-intensive, this search wa s exe­ used. \Vhen use of MULPAHS resulted in 3,900 cuted with the simplest branch- swapping algorithm, trees, which used up avail able RAM , a sin gle tree NN I. Additionally. because of dynamic RAM lim­ wa s randomly selected (iv) 10 swap on with TBH it a tions (a tree of 475 terminals occupies a great (M LPARS ofl). If this resuhed in a shorter tree deal of Inemory), we restricted our search to a being found, the search wa s then stopped and reo maximum of 500 trees. Although the initial TBR started (iii, again) using N N I and M U LPA HS and tree was by default a member of a single island, this shorter tree as start ing point (iii and iv were we hoped to provide a bridge to shorter trees through repeated until no shorter trees we re found). The use of th e STEEPEST DESCENT option (scc Mad· shortest tree length found with this method was dison, 199 1; Swofford, 1991). This fin al step yield · 16,305. Three randoml y selected trees from the ed 500 eq ua lly parsimonious, weighted trees. Be· 3,900 saved at this length were swapped in suc­ cause the maximum prespecifi ed number of trees cession to completion wit h TBR (no M ULPAHS), wa s found, many others probably exist at the same and no shorter trees were found . A strict conscnsus length . Search I required approximately 200 hours tree was computed, and branch lengths for one to complete . randomly selected tree we re calculated using the Search II was performed for three reasons. ACCTRAN optimiza tion. Search II thus used no First, the li kely existence of ot her islands of equal rela ti ve weighting; it required approximately four or greater parsimony prompted us to use a strategy weeks to compl ete. that would be more likely to find shorter trees and perhaps a different topology. Second , we we re con· CAVEATS cerned about the effects of the Albert el al. ( 1993) weighting scheme upon the resulting topology . Methodologically, these searches suffer fr om (i) Third, we wished to eX31lli ne positions of additional uncert.aint y about max imum parsimony, (ii ) un · taxa (a nd make use of updated sequences) that questionable absence of many trees at the same became available a ft er Search I was completed; level of optimalit y, (iii) id entifi cation of onl y a sin gle many of these be longed to previously unrepre· topology, and (iv), in Search I, in complete branch sent ed lineages. Differences between these two sets swapping on any of th e shortest trees found. \Ve of trees could thus be due to different taxon salll · would never recommend Ihesc search strategies for piing, corrections or comple tions of sequences after smaller data sels. but several options were seriously Search I was fini shed, or sea rch strategy. We did restricted by the number of taxa included (these not intend these two searches to be controlled, a re reputedly thc largest PA U P analyses attempted direct tests (i.e., wi th onl y one va riable differing to dale). SpecifiC sections of the gcncral topologies, between them); we show th em both, rather than when analyzed in a I'nore " localized" manner, pro· simply the one that we judge to be beller (Search vid e different sets of relat ionships (see Mi chaels et II ), because their similarities, despite variation in a I. , 1993; Morgan & Soltis, 1993; both this issue). taxon composition and search strategy, arc con­ The broader taxOJI distribution of the general anal­ siderable. They each represent results of searches ysis (thus with far greater outgroup information) that in their own context are worthy of publication, ma y be assessing character·state change on the and their differences should be viewed as reasons immediate branch leadin g to a specifle ingroup to be skeptical of both and as cause for future differentl y than in morc restricted analyses. C reat· stud y wi th more rb('L data as well as other char. er outgroup information could " improve" the in ­ acters. group analysis or add spuriollsness to it; it is gen· In Search II , we we re able to save more trees erally best to implement tree searches both wit h a t the shortest length found: 3,900 rather than and without outgroll ps to examine their effects. If only 500 trees. The init ia l sta rting tree was pro· more restric ted analyses differ from those pre· Volume 80, Number 3 Chase et al. 533 1993 Phylogenetics of Seed Plants

sented he re, we would certainly favor the forme r pe ndix). Many tissue sa mples were provided by because of inc reased confide nce ill findin g parsi. botanical gardens, a nd , if vouche rs arc not included monious solutions. W e nonetheless vicw these a nal­ wi th samples, in vestigators a re depe ndent upon yses as instruc tive about seed -plant rela tionships id entification of these pl a nts by the respective or­ a nd the utilit y and lim it s of rbe L information (see ganizations (see Cold blatt et a I. , 1992: it is c ritical Discussion). tha t a voucher sample taken from the same plant The major limiting factor in our studies is clearl y lIsed for DNA extraction be included a t the time ma trix size. When confronted with the largest mo­ of collection; accession numbers or vouc he rs taken lecula r data base relevant to seed plants, compu­ previously a re subject to la ter events, such as lost tational trade-offs inevit ably a rose. The amount of la bels or collector mista kes). Because a nu mber of lime spent on these approximations may nol be the sa mples used in these two a nalyses are un­ directl y proportional to the lime taken 10 genera te vouchered (see Appendi x), re producibilit y is com­ a ll the rbe L sequences, but no method exists 10 promised. In cases wi th multiple species from a predic t how ma ny trees a l how many steps shorter fa mil y, we should be able to recognize grossly mis­ might have been found after additional months or identified sa mples (but th is is true for only 37% of even years of continuous computa tion. 'Ve have the 265 families re presented). opt ed for the approxima ti ons present ed here rathe r than commit ourselves to an ope n-ended experi­ R ESULTS ment . Potential effects of errors in a utoradiogram read­ For di spl ay purposes, we show here the com­ ing and data entry should be conside red . '''(Ie de­ binable component consensus (Bremer, 1990) of tected a number of internal stop-codons in se­ the 500 equall y parsimonious trees found ill Searc h quences used in these analyses. a nd other workers I (Figs. I - I S, A series). Because these a re c har­ have re ported errors of va rious kinds. Most of these acte r-S late weighted trees, tree lengths a nd tree we re correc ted in the matrix L1 sed in Search II . statistics (e.g., consiste ncy index, e tc.) arc 1I0t com­ Certainly the potent ia l for e rrors is present , but parable to those of Fitc h Irees and arc not given the e ffects of such mistakes should not be extensive he re. For Fitch trees found in Search II . the le ngth because they a re li kely to be ra ndom. wa s 16,305 wit h a consistency index (C.!.. ex­ Id entifica tion of some taxa in this a nalysis lIa s cluding unique substituti ons a nd constallt c harac­ been questioned due to peculi a rity or placement. ters) of 0. 102 and a retention index (R .1. ) of 0.632. One example of this has been now ide ntified a mong The branch lengths, again wi th ACCTHA N opti­ the species used in Searc h I. W e ha \'e not expunged mization, a re shown on one of the 3.900 trees it from the illustra tions: the results represent the selec ted a t random (Fi gs. 1- 15, series Il). Ilmllches outcome of real tree searc hes a nd are instruc tive that colla pse in the stric t consensus tree are indi­ for that reason. The seque nce a nalyzed wa s all cH led by arrows on the 13 series (Figs. :~ - 1 5 B ). ac tual me mber of the dade ill to which it wa s pl aced . First we wi ll SUllHl'larize the topology round in but it wa s not the species to whic h it had been Searc h I ("A" seri es or fi gures). The results of a ttributed. The ma terial sent by a botanical garden Search II ("' 8 " series) have been inte rdigi tated wit h a nd labeled as Kirellw'shuma was evide ntl y mi s­ those of Searc h I to fa cilitate comparisons. After ide ntified . It was almost certainly a me mber of the describing the results of Search I, we briefly ex­ Parrwssia group ( se nsu lato; the amine major differences between the two results. posi tion it occupies in Fig. I I A) ra ther than a Note that a ll fi gures in the A series arc frolll the membe r of Hydrangeaceae. Seque ncing of another combinable component consensus tree, whereas in specime n of Kircngeshofrla and subseque nt dOJta the B series. Fi gures I Band 2B a re the stric t a nalysis revealed the e xpected placeme nt of this consensus tree (branch lengths in Fig. 2 8 , however. with othe r Hydrangeaceae (X iang & Soltis. were ta ke n from a single tree) and the remainder unpublished). Other " surprising" results also have (Fi gs. 3 13- 15 13) a re a sin gle tree randoml y selected been checked by obtaining a nother sample of the from 3,900 equally parsimonious Fitc h trees (this taxon in question a nd re-sequencing rbr L. The may be confusing; for example, whereas the po­ original sequence of purportedl y saxifragaceous lytomy among monosul cate clades in Figs. I Band MOlllillia (wh ic h nested in ) was checked 2 8 is due to varia tion among the 3,900 trees for and found to be accura te. Sti ll others (such as branches in this portion, the topology shown in 3 B Sargenfoc/oX(l. a mong ) a re still being is resolved because it is resolved in the single tree reassessed . Most worke rs have tried 10 make scleded). To indicate branc hes of the B series that vouche rs for each species in this study (see Ap- a re absent in the stric t consensus tree of the 3,900. 534 Annals of the Missouri Botanical Garden

we ha ve placed arrows to their right. T he bra nc h nolialcs. (ii) La urales, a nd (iii) "paleohe rbs" (here le ngths presented in the B seri es should unde r no de flned as composed of Aristoloc hiales, Pipe rales, cirCUlllsta nces be inte rpreted as meaningfu l mea­ a nd Nymphaea les; Fig. 4 A). Monocots (also wit h sures of support ; Ihus. in the exa mple of Ha ma · uni apertura te pollen a nd sometimes included in melidae provided. bra nches tha i deca y a l o ne step " paleoherbs"; Donoghue & Doyle, 1989) repre­ less parsimonious have le ngths tha t range from 2 se nt a fourth me mber of this . Anlong the 10 16 steps. whe reas those tha t deca y at four steps paleohe rbs are also nested seve ral proble ma ti c fa m­ less pa rsimonious ra nge from 7 10 14 sleps (Fig. ilies: Illi ciaceae, Sc hisand raceae (both lIIiciales), 16 ). The sole reaSon for providing branch lengths Ambore llaceae (La urales). a nd Austrobail eyaceae in the B se ries is to pe rmit readers 10 estima te (Magnolia les). Chlora nthaceae a re also a llied ci a· roughly relative degrees of di vergence a nd to iden­ disticall y wi th the paleoherbs, but Chlorflfllli u:; does tify cases in whic h long te rmi nal bra nches a re not form a monophyleti c group wi th other Pi pe r­ connected to short inte rna l branches (a sit uation ales. in whic h addi ng related taxa of le n radically a lt e rs Monocots a rc a we ll supported monophyle tic hypothesized rela tionships). group (see Duva ll e t a I. , 1993 , a nd Qiu e t a I. , Searc h I . Unless stated othe rwise. we ha ve 1993, both this issue) a nd a re de rived from wi thin used the ta xonomic circ umsc riptions of Cronquist Ill onosulca te Magnoliidae; the paleohe rbs a re their (1981 ) for dicots and Da hlgre n e t a l. (1 985) for immedia te siSle r group (Fig. 4 A). Within monocots monocots, a lt hough we ac know ledge that othe r re­ (Fig. SA, Il), Aranae plus Pleeo (of polyph yletic cent systems fi t these result s bette r. \Ve a rra nged M e la nt hia ceae ~ Lilia nae) a re basal-most, followed the unrooted trees of bot h searches with cycads by Alisma ttlllae plus l3urmoll fl ia (Bunnanniaceae) siste r to all othe r seed pla nt s (Figs. 1- ;-3) in accord a nd Alelris (Mela nthiaceae). form a pa ra­ wit h recent result s of seve ral lion-molecular cla­ phyle tic se ries of three li neages that correspond distic stud ies (C ra ne. 1985 . 1988; Doyle & Don­ we ll to , , a nd Asparagalcs of oghue, 1986 . 19 9 2). In Search I. conife rs a re Da hl gren e t a l. ( 1985), e xce pt in the placement pa ra phyle tic. but some trees (not shown) found in of certain fa milies (I ridaceae, Orc hidaceae. a nd Sea rch II ha ve a monophyle tic conife r lincage; the Smil acaccac) a nd genera (Chamaclirium of Mclan­ stric t conse nsus tree from Searc h II is unresolved thiaceae: Mcla nt hiales). Vcllozia (V e llo z i a c e ~lC: rega rding conife rs (Figs. I B. 2 13: we ha ve cit ed in Bromelia nae), F"re),cinctia (Panda naceae: Panda · this section the 13 se ries of fi gures along with the na nae). and Sp/wcraderl ia (Cycla nthaceae: Cy­ A se ries if they include the Sa me gene ral set of cla ntha nae) togethe r form a monophyle tic clade taxa). T he three gene ra of Cnetales. J~·ph e dr(l . that is collectively sister of the Lilialcs. The " com­ We/IVi l.schia. a nd e"e/llm . a rc highly di vergent melinoid" group of monocots (f ig. 6A, B) incor· from a ll othe r seed plants but we re none theless porates a ll of those tha t Harris & Ha rtlcy ( 1980) ident ifl ed as siste r of the angiospe rms (Fig. 3A. B). found to e xhibit fl uorescing ce ll -wa ll phenolics. In wi thin which Ce ra/ophy llilm (Cera tophyllaceae) both searc hes. this commelinoid clade includes alone is sister to a nd highl y di ve rgent from the rest monophyle tic Areca nae a nd Zingiberanae a nd (Fig. 4 A. Il). T he major feature of flowe ri ng plants polyphyletic Bromelia nae a nd Commelinall ae (Fig. (exclusive of Cera lophy llum ) is their separation 6A , B). Cyclanlhaceac have the sa me phenolic int o two major groups; thesc correspond well with biosyut he tic pa thwa y but do not accumulat e e nd distribut ions of the two major a ngiospermous pollen produc ts; they are not mcmbers of the cOllllnclinoid types, uni ape rt ura te (mollosul ca te and monosul ­ assemblage (Clark el a I. , in prep .). ca te-de rived) and triapertura te (tricolpa te and tri­ The two orders of Magnoliid ae wit h Iriape rtura te colpa te-de ri ved). Ce ratophy llllm has inapert ura te polle n, Hununculales a nd Pa pa ve rales, fon n a cl ade polle n (Cronquist. 1981 ). The major exception to tha t is sister to the rest of "eud;cots" (Fig. 7 A, this split is the presence of tricolpate poll en in B). The te rm "cudicot" has bec n va riously de fin ed Ill icia ceae a nd Schisa ndraceae. which fatl among in the lit era ture, but we use it he re to refer to a ll the rnonosulca te taxa (Fig . 4A. B; see Qiu el a I. , a ngiosperms with tria pe rtura te or tri apert ura te-de­ 1993, this issue). No morphological support for rived pollen (Donoghue & Doyle, 198 9 : Doyle & monophyly of the monosulcate clade has been rec­ Iiolt on, 199 1). T his is one of the best supported ognized in the lite rature (their polle n type exists clades a mong a ngiospe rms (Qi u et a I. , 1993 , this a mong nonfl owe ring seed pbliis a lld thus lIluSt be issue). T wo othe r basal clades within the cudi cOfS conside rCfI plcsiomorphic). (Fig. 7 A, U) consist of some Ha mamclidae (T roc h­ T hree monophyletic lineages wit hin uniape rtur­ odencl racc'lc a nd Tetrace ntraccae) a nd Pla ta na· a te we re ident ifi ed, a nd these corre· ceae. Sabi aceae, Nelumbonaceae, and P roteaceae. spond closely. a lt hough not exactly, to (i) Mag- \Vi thin . two la rge sister clades arc iel en- Volume 80, Number 3 Chase et al. 535 1993 Phylogenetics of Seed Plants

tifled, one that corresponds roughl y to Asteridae included in Gera niaceae (see also Price & Palmer, and the ot her to Rosidae (Figs. 1, 2). Membership 1993, and Morgan & Soltis, 1993, this issue). in both lineages is considerably expanded with re­ The tenuinuce ltate/ sympetalous clade that ter­ spect to their circumscription by Cronquist (1981), mina tes in Asteridae sensu Cronq uist (198 1) was although less so wit h respect to the circumscriptions also identified by Olmstead et a l. ( 19 92) in their of Dahlgren & Bremer (1985) and Thorne ( 19 92). effo rts to circumscribe subclass Asterid ae using These two major clades (Fig. 2A , B) renee! the rbc L sequences. This study greatly expands upon di vision of cudicots into two major groups (Young their sampling and id entifies as members of the & Watson, 19 70): sympc talousltenuinuccll atc and asterid clade two lineages of often polypetalous polypetalous/ not tenuinuccllate (aste rids a nd ros­ Rosidae (Figs. 2A, 12A): (i) plus Paeoni­ id s, respective ly, in our fi gures). The "crassi nu­ aceae and Gunneraceae and (ii) some famil ies of cella Ie" condition actua ll y consists of several dif­ Comales plus Hydrangeaceae (see Xiang et a I. , ferent states. Exceptions to this genera li zation e xist, 1993, this issue). The sister group of Asteridae is bu t these traits appear much less homoplastic here a clade (Figs. 2A , B, 13A, B; see Kron & Chase, than in most systems of classification. The basal­ 1993, this issue) that contains the dilleniid orders most lineage wit hin Rosidae includes Saxifragaceae Ebenales, Ericales, Primulales, Diapensiales, plus sensu stricto and Crassula ceae besides lower ha­ some members ofTheales (A ctinid iaceae and Thea­ mamelid s such as Cercidiphyllufft and IIamomelis ceae). Sarraceniaceae (Ncpenthales; DiJleniidae) and (Fig. 8A; see Morgan & Solti s, 1993, this issue, N oriduLa (but not BybLis of Byblidaceae of Ro­ for a di scussion of Saxifragaceae sensu lat o). The sales: Rosidae) a re also members of this li neage next-most-ba sal group contains Caryophyllid ae (in ­ (Albert et a I. , I 992b). Polemoni aceae (Solanales: cluding Plumbaginaceae a nd Polygonaceae; Rettig Asteridae) and Balsaminaceae (: Hosi­ et aL, 1992) plu s , Nepenthaceae (Ne­ dae) also belong to this e ricalean/ ebenalean group . penthales, Dilteniidae; see Albert et a I. , 1992b), (See also Olmstead et a l. , 19 93, this issue, for a Di lteniaceae (Dil leni idae), and (Fig. 9A ). treatment of the Asteridae sensu lato.) The remain ing Rosidae are split into two large Asteridae sensu Cronq uist split into Iwo major sister groups (Figs. 2A, lOA , I 1A). In one (rosid sister groups. In one of these (a sterid II ; Fig . 14 A, I) a re several families of higher Hama melid ae, ElI ­ B) are families of , Calycerales, Campan­ phorbiales, , Linales, Polygalales, and Ro­ ul ales, , and some Solanales (Menyan­ sales (Fi g. II A). Other members of this cl ade in ­ thaceae). Rosid taxa that are members of this cl ade clude a number of dilleniid families: Ochnaceae include , Aquifoliaceae (CelastraJes), some (Theales), Datiscaceae, Passifl oraceae, and Viola­ (), Pittosporaceae and Gros­ ceae (a ll Violales). Ordinal boundaries of this group sul ariaceae (both ). In the other major cl ade of Rosidae (sensu Cronq uist, 1981 ) are largely (Fig. 15A, B) fa ll ord ers Callitrichales, , unsupported; this assemblage is particularl y het­ , Rubi ales, Scrophula ri ales, and most So­ erogeneous. The largest pol ytomy in Ihe consensus lanales, alt hough these ordinal limi ts are nol always tree from Search I occurs at the base of this group, supported (see Olmstead et a!. , 19 9 3 , this issue). and sampling of the famil ies that potentially belong In this clade (Fig. 15 A, B) is a group that includes to this clade is the most spa rse in this analysis. Aucuba. (Cornaceae: Corn ales) and Carry a The other major li neage of Rosidae (rosid II ; Fi g. (Ga rryaceae: Cornales) and hamamelid Eli commia lOA) includes orders Myrlales (see Conti et aI. , (Eucommiaceae: Eucommiales), a ll of whi ch ac· 1993, this issue) and , for which ordinal cumulate aucubi n (Cronquist, 198 1) and sha re dis­ boundaries are reasonably intact. Malvaceae (Mal­ tinctive anatomical characteristics (E. Whee­ vales; Dilleniidae) and all but one of the mustard­ ler, pers. comm.). Thus the suite of Aoral oil families (those in Capparales plus others in Dil­ characteristi cs that have been in terpreted as sup­ leniidae; see Rodman et a I. , 19 9 3 , Ihis iss ue) are port for the monophyly of Asteridae sensu Cron­ associated wi th Sapindales. Geraniaceae are also quist (1981) appears either (i) to have twice a risen members of this clade, alt hough other members of independently from ancestors wit h rosalean and Geraniales appear elsewhere (Oxalidaceae with a cornalean fl oral traits or (ii ) 10 have undergone group of families in Rosales, Fig. 1 1 A; Balsami­ reversals in groups traditionall y included in Rosidae naceae with Ebenales, Fig. 13A, B; and mustard­ (sensu Cronquist; Donoghue et a I. , 1992; Ol mstead oil -producing a nd Tropaeolaceae et a I. , 1992, 19 93, this issue). wi th Capparales, Fig. lOA , B; Price & Palmer, Search II. A number of taxa from Search I 1993, this issue). Two members of Rosales, Creyia. were removed from Sea rch II (a ll marked with a (Greyiaceae) and Francoa (Saxifragaceae), appear "t" in the A series of fi gures) to accommodate the derived from wit hin Geraniaceae, if Viviania is representatives of additional lineages in Search II. 536 Annals of the Missouri Botanical Garden

M OSI of thc removed taxa were from monophyle tic 1I0liid species added in Search II represent addi­ families in whi ch six or more species we re present tional members of fa milies already prescnt ill Search in Search I (for example. I\ stc raceae, Erieaceae sen­ I, a nd all of these form monophyletic lIlI ilS with su lato, Magnoliaceac. and Poaccae). T wo ot hers. other members of their respec ti ve famili es (Fig. I3 l1rmlll",;a (Burma nniaccac) and /ly drolco (Hy­ 48). drophyllaceae), were removed because they arc Sister to a ll ot her monocots is Acortl s (Araceac; highly seque nce di vergent fr om a ll other taxa; ill see Duvall et al.. 1993, this issue). Calm·hor/lls separate, smaller analyses, these I wo appea r to be (CaloehortH ceclC) is sister to Liliaceae. /fem era- o involved in " branc h attractions" and allach ill rad­ callis (Hemerocall idaceae) is siste r to {:Jtfuraplty. icall y diffe rent positi ons as ot her taxa a re added /IUft (Anthe ri caceae), and Lomaflcira (Dasypogona­ or removed (ot her simi la rl y di vergent genera, for ceac) is a rn embe r of the oft e n arborcseelll clade e xample, Po con i.ll and GIIIII/('ra. do not cause these cmnposed of Aga vaceae, Asphodelaecae. and Xan· problems and were kepI: we admit to being rcla ­ thorrhoeacae wi thin Lilianae (Fig. 5 B). Additional li\'c ly a rbitrary in re moving onl y these two taxa). members of families in Searc h I a rc all placed as UlJ rmarlflia prese nt s a n inte resting case. Melnbe rs siSler laxa to their respec ti ve family represe nta­ of Burma nniaceae are ofte n nchlorophyllous: in ti ves. Among Ihe commclinoid clade idcntified in spite of being grecn, }Jllrmmlf/ia may still de rive Search I (Fig. 6A, B) are thc foll owing additional a great deal of its nutrition through it s mycorrhizal familie:;: Spargflflillfll (Sparganiaeeac) sister to Ty ­ associate. In suc h cases in othe r families, a number pha (T yphaceae) a nd Lachflocau/ofl (Erioea lll a­ of protein loci e xhibit highe r rates of sequence ceae) a mong the gra minoids. di ve rgcncc (as measured by relati ve branch lengths Among e udicots (Fig. 7 A. B), Pachysandra whe n compared to comple tely a utotrophic [n elll ­ (Bu xaceae) is sister to Troc hodendrales (Trocho­ bers of their lineages: C. de Panlphilis. pers. comm.). d endron and Tf' /rrf(·('fltrofl). Composilion of the Thus the high levels of seque nce dive rgence, whic h va rious ros ie! clades is somewhat diffe rent ill Searc h [nake it diffi cult to place these taxa accurately, arc II (Fig. 2 B: see helow), but the followin g additional a produc t of or a t Icast associated with their partial taxa arc placer! roughly among the rosid (( (Fig. he te rotrophy. 10 13) clade id entified in Search I: Shurf'a (Dipler. For Search II , additional tax a we re available ocurpaceae). 'l7H>o hrofT/rt (S te rcliliaeeile), '{"ilia (Til ­ (these species a re ma rked wi th aste risks in the B iaceuc), a nd Numbax (Bombacaeae) a re members series of fi gures). The placeme nts of these arc of a mulva lean clade (represented onl y by Cos­ described fir st. a nd then diffe rent a rra ngenlClits of .~)'pi/lfn in Search I) . .,.lkania (A kaniaceae) a nd taxa included in both searc hes a rc identified (the Hre /sellfl eidera (Bre tschne id eraceae), both Sap­ clades involved with major shift s of posit ion are ind ::lles, are Illcnlbers of the mustard·oil clude. Three marked with all .. § .. ill Fig. 2 (3 ). This la st sectioll additional me mbers of Ca pparaceae. Cleomf'. {{oe· describes onl y the major shifts of position, but many bl'rlin ia. a nd Setch('lIarllhus, along wi lh Cappa. "minor" shift s also occur wi thin cl ades (for ex· rij, c reate a pol yphyletic (Fig. l OB; a mple. 12 " minor" shifts ta ke place among the sec also Bodma n et a I. , 1993, this isslIe). /fY I)­ taxa in Fig. 15. a clade in whic h only a fe w ne w .~('(Jc hori .~ (Oxalidaceae) is siste r 10 a clade con­ species we re added). \Vha t constitutes a " major" ta ining many Ce ra niaceae (Fig. 108; see Price & versus a "minor" a lt e ration is, of course, a matter Palmer, 1993, this issue). AnlOng members of rosid of personal pe rspec ti ve. \V c wou ld therefore advise I (Fig. I I B, C) are representatives of the following readers to e xarnine carefull y th e trees from both new families: U f';fI/(IOrdlio (Linaeeae) is si::; te r 10 searches for taxa of specirlc inte rest, whi ch is one I 'io{rt (Violl.lceae ): Sargen/oc/oxa (Sargent odoxa­ reaso ll we presented and interca la ted the rcsult s ceae) is imbedded in Fabaceae; fl UIIIllIII .~ (Ca ll ­ of both analyses. nabinaceae). '/'f'{' flW (Uhnaceae), a nd /Ju phmnia Positions of additional taxa. Ginkgo (Cink - (Urticaceae) fall in to a n urtiealean clade wi th Mo­ goaceae) inte rcalates between cyca ds and conifers rtiS a nd Ficlls (hoth Moraceae), but Ulmaceae a re (Figs. I ~ 3 B ) . Tcu:ll s (Taxaceae), Cepltalola:(fl .~ paraphyleli c to Ca nnabaceae: Coria ria (Coria ria· (Ccphalotaxaceae), and Sciadopitys (Taxodiaceae cacae). /legonia (Begonia ceac), a nd three genera or in it s own farnil y) a re me mbers of the nOll · of Cucurbitaceae are placed wilh Datiscaceae: Ne/­ Pinaceae clade of conife rs. wit h Sr-iadopitys sister lila (Betulaceae) and Caryo (Jugla ndaceae) arc to Ihe rest of that clade (Fig. 3 B). \Vilhin magnoliid me mbers of the clade conl nining a nd angiospe rms. Lac /oris (La c toridaceae) is sister to Casuarinaceae: MOllriri and Osbeckia (Melasto­ Aristoloeltia (Aristolodliaceae). Other ne w mag· mataceac). Pllu;m (Punicaceae), Trapo (T rapa. Volume 80, Number 3 Chase et al. 537 1993 Phylogenetics of Seec Plants

ceae), and Neleropyxis (Myrlaceae) a re members Fa milies of M yrtales (plus Qualea of Vochysia ceae, of a clade of (mostl y ) tha t includes On­ Polygalales) and the clade containing Viviallia. agraceae and Comhretaceac (Fi g. ll C). In asterid rr1endlia (both Ceralliaceae), Creyia (C reyiaceae ), lineages, two additiona l families have been added: and Francoa (Saxifragaceae) a re shifted rrom rosid B)'blis (Byblidaceae) and Vahlia (Saxifragaceae) II to rosid I (Fig. I I B, C). and rela tionshi ps of a re placed among Schrophu la riales (Fig. 158 ). intermediate-level clades (containing several ra rll ­ Changes ill placemen ts of major clades. AI ilies, i.e., those or Urticales, , e tc.) are hi gher levels among the seed plants, Searc h II sornewha t modified from their posit ion in Searc h J. produced relatively few "major" differences in tax­ The third shift in volves the two taxa that we re on placements from the topology of Sea rch I. The siste r to the expanded caryoph yllid dade frorn conifers in some, but not a ll , 3,900 trees are mono­ Searc h I (Fi g. 9A): Dil/.e1l ia (Dill enia ceae) a nd 1';lis phyleti c. Pi perales- Arislolochia les were siste r to (V itaccae), whi ch become siste r to the rest of the all monosulcate angiosperms (Figs. 28 , 48 ). The la rger asterid clade (Fig. 12B; Vilis a nd /Jill(,llia "paleohcrb" clade of Sea rch J (Fig. 4A) was thus in a se nse exchanged positions with Sant alales). split in two with the portion containing Chl oran. Again, these groups have short internal branc hes, thaceae. Illiciaies. Au strobaileyaceae, · a nd these shifts wou ld require little c ha nge of ovcr­ ceae, a nd situa ted as sister to Mag­ a ll parsimony. nolia les. Although these !'na y seem to represent In several insta nces, the topology of Searc h II major shift s, extremely short branc hes separate (which we ravor because it was a more complete these clades (Fig. 48); thus nei ther topology has sea rch) is more similar to tha t found in il1\'csli ­ much inte rnal support (Qiu et al., 1993, this issue). gations of restricted nature (for example, outgroup \Vit hin monocots, onl y shifts a mong the groups in relationships of Asteraceae are id entical wit h those Aranae, Alismatanae, a nd Lil ianae occurred (Fig. round by Michaels e l aI., 1993, this issue, a nd 51l): Aranae (minus ) are sister to Alisma­ Olmstead e t aI., J 992, whe reas those or Search I tanae; the clade containing Pa nda naceae, Cyclan. devia ted in several ways). thaceae, and Velloziaceae is isolated and no longer siste r to Liliales, a nd Dioscoreales a re siste r to the DISCUSS tON commclinoid taxa (the positi on of Aspa ragales in the trees rrom Searc h I). Among commelinoid taxa Although rbe L sequences for several groups of (Fig. 6 8 ). Slegolepis (Rapateaceae) is sister to spore· bearing plants are availa ble (true mosses, , T yphales a re siste r to Juncaceae­ hornworts. liverv.'orts, Equisetum. Isoeles. I.y eo. Cyperaceae, a nd Flage/faria is sister to the othe r podium, Psiloltun. a nd both e u· and leplosporangi­ graminoicl clade. ate fe rns), their use as out groups is complicated Arllong eucl icots, three major shirts or taxa pres­ by extensive seque nce divergence rela tive to tha t e nt in Searc h I occurred . The fir st series or re­ in the seed-plant ingroup. No othcr extant lineages arrangements in volves the he te rogeneous astcrid or land plant.s a re likely to have shared a COI'I'lIn OIi V clade (Fig. 12A). GUIII/Na (Cunneraceae) in· a ncestor wit h seed pla nt s for we ll over 350 mill ioll te rcala tes as an isola ted lineage (Fig. 7 8 ) between yea rs, a nd a great deal or sequence c ha nge. muc h Trochodc nd ralcs (plus Paclry.wffulra or Buxaceae) of it in the form of multiple, unrecoverable sub­ and higher e udicolS (aslcrids. caryophyllids, a nd stitutions, has occurred. Analysis or these other rosids). Santalales, Plwrar/(,lI drurl (Viscaceae), la nd plant sequences produces topologies (not ShOWII , Sclwepfia (Olacaceae). a nd Os)'ris (Sanlalaceae) but see Hamby & Zimmer, 1992, for silllil a r re· become ~ i s t e r to Caryophyllidae- Droseraccac­ sults) that are radica ll y different from a ll previolls Nepenthaceae (Fig. 9 8 ), a nd this larger caryophyl­ hypot heses or relationships (e.g., Bremer e t aI., lid clade is shifted from a posit ion wi thin the rosid 1987). In contrast. seed-plant rela ti onships pre­ clade (rosid III; Figs. 2A, 9A) to siste r to the largcr se nted he re arc a t least congrue nt in gross aspec t aste rid- rosid clade (Figs . 2 13 , 98). The remaining with compara ble morphological studies (Cralle, mem ber of asterid V, Paeoflia (Paeoniaceae), is 1985, 1988; Doyle & Donoghue, 1986, 1992: deeply imbedded within rosid III clade (Fig. 88) Loconte & Stevenson, 1991). Addition or highl y as sisle r to . W e a llac h little signifl ' sequence.divergent out groups could be cxpec teJ cance to these shifts of position; int erna l branches to in crease ingroup homoplasy wi th unpredicta bl e of these g r ou p ~ are among the shortest supporting topological result s (Fe lse nstein, 1978). III these positi ons of major clades. analyses, we have c hose n to use the more cOllse r· The second series of shirt s occ urs within rosids. vative a pproach or

of seed pla nts, whi ch are alm ost certainl y mOllo­ and ribosomal studies (Hamby & Zimmer, 1992) phyle!;c (Do yle & Donoghue. 1986. 1992). are quite simila r to ours as wel l. I-low these la rger Likewise, effects of missing groups upon lopol­ groupings (cluste rs of families and in some cases ogics may be profound (Donoghue et a I. , 1989) orders) are inte r-re la ted is the point at whi ch the ir a nd unpredic table, and we expect the absence of si milarit y diverges. Although having quite diffe rent the numerous extinc t lineages of early land plants implications for angiospe rm origins and evolution. ill rnolecular da ta ma trices 10 pose pote ntia ll y se­ the prefe rred hypothesis in one of these studies is rious probl ems for elucidating rela tionships of e x­ not vastl y differe nt in relative pa rsimollY frolll those tant lineages. Perha ps more conservati ve gencs ravored in other in vestiga lions. For example, with sa mpled for more of the sequence varia tion presel1 t Ill ostl y morphological data, Do yle & Donoghue within e xtant groups rna y be able to "bridge" gaps ( 1986) discovered a " palcohe rb rooting" a t one caused by ex tinction . Features of genome orga ni ­ ste p less parsi moni ous. Constraining a paleohcrb zation (suc h as gene and inlron content and gene rooting for a ngiosperms. a t NY lllphaeales (as in orde r ; Downie & Palme r, 1992; Raubeson & J an­ lI amby & Zimme r. 1992). was also onl y slightly sen. 1992) may offe r more robust hypotheses than less parsimonious in the subset of angiospe rm rbeL gene sequences fo r such questions, but these a re scquences studied by Qiu e t al. ( 1993, this issue). lik e ly to be too few to provid e a full y resolved tree \V hen examined to address basal angiospe rm rc­ by themselves. A great deal more experimentation la tionships, a ll these da ta appear to la ck a strong with combined data se ts of morphological a nd 111 0- historical signal. lecula r cha rac te rs is obviously needed, a nd a num· A nu mber of phe nomena have been suggested ber of these studies are unde rway (both with rbe l to be capable of con rounding molecular phyloge. and rHNA/ rDNA data). Consideration of these ne tic studies. We consider below several of these proble ms he re is pre mature. ra c tors a nd exalliine their potential to affec t studies Siste r-group status of angiospe rms a nd Cnetales of rbc L seque nce va ri a tion a nd then address some (Fi gs. 2A. Band 3A. 13) is corrohorated by other additional concerns a bout future directions of mo­ cladisti c studies (C rane. 1985. 1988; Doyle &' lecular systema tic study. Donoghue, 1986, 1992; Locont e & Stevenson. 1991 ). The isolated position of C '",rophy llulII as HFECTS OF PAII ALLEL NUCL£OTJI)E siste r to a ll othe r a ngiosperms has been argued :-iU tlSTtTUTIONS IN tNDEPENDENT U NEAGE.o; previously (Les, 1988: Les e t al .. 199 1). I n studies An effect of unrecoverable (due to extinc tion) by Qiu et a l. (1993, this issue) in whi ch non­ or unobserved (due to insufficient sampling) char­ flowering seed plants we re re moved , this a rra nge­ ac te r- sta te cha nges is the introduc tion of spurious me nt was made equi vocal by the existe nce of a n· simila rities, whi ch may result in treatment of in· othe r equally parsim oni ous island in whic h (;Cf([· de pe nd ently de rivcd Ilucleotid es a t a givc n base rop"y llllln occurred in a radica ll y different position. position as hOlnologous (see Albe rt & Mishle r, Hamby & Zimme r ( 1992; rH NA) a lso found a yet 1992). Such mista ke n int erpretations of inde pen­ differe nt placement for Cer(/fop"yJ/llfT1 (but we de nt events can lead to -- branch at tractions" if an suspec t tha t the sparse r samplin g of the ir study analysis includes a great numbe r of such assess­ may be responsible for 1Il0st of the d ifferences from me nts (Felsenstein. 1978). Ad equa te taxon sa m­ those found with rbe L) . In in stances in whic h a pling, sometimes refe rred to as a ppropriate -- taxon taxon's morphology a nd anatomy are as di ve rgent de nsit y," is one In cans of reducing potentiall y in · a nd potentiall y modified as thosc of Cerar oph)'1- uccurate assessments of simila rit y, but de te Tlnin ing 111m , it s position becomes difllcult to address adc· a t what point sallipling is suffi cie nt has so rar onl y quately in cladistic studies. Cera rop"yllul1/ has been heen addressed in a ll a poste riori manne r. absent from many morphological cladisti c studies. The improveme nt a fforded to assessments of sllch as those of Doyle & Donoghuc ( 1986, 1992), c ha rac ter·state change by inc reased taxon sam­ so corroboration is currently precluded . pling is counte rbalanced by a decrease in CO l11 - The general groupings of angiosperms (exclusivc putational speed and abilit y to ascertain how ncar of Cera/opll)'llum ) ide ntified in these two a nalyses results are to maximum parsim ony. Intrafamilial are highly similar to each othe r and to those of studies a re not as like ly to be affected severely by most recent taxonomi c sche mes. pa rt ic ularl y thosc these problems because, in gene ra l, numbe rs of or Dahlgren (1980). Dahlgren el al. (1985). a nd ta xa a re lI ot as great a lld evcnness of sampling is Thorne (1992; this lasl has adlnitt edly in corpora ted 1Jt: lle r. At higher ta xonomic levels within sced pla nts. results of seve ral molecul ar investi gations). Fur· a paradoxical impediment to progress arises: if the rmore, results of rbcS (Martin & Dowd, 199 1) taxon number is great e nough to assess cha racter· Volume 80, Number 3 Chase et al. 539 1993 Phylogenetics of Seed Plants

state changes accurately, then one reasonably can pled with managea ble numbe rs of taxa (from the expec t to reach onl y suboptimal phylogenetic so­ standpoint of computation of minimal trees) a nd if lutions, but if taxon number is restricted enough sa mpling with sufficient numbers of taxa precludes to gain confidence of the maximum parsi mony of assessing the parsi mony of results, the n we have the trees found , spurious assessments of c ha rac ter­ reached an impasse until improved methods of a nal ­ stale identity could obscure a ll but the closest re­ ysis a re developed. An apprecia tion of this proble m lationshi ps. has led us to be skeptical of the overa ll topologies To evalua te relationships of Plumbaginaceae and presented (Figs. I, 2), and competing ideas of re­ Polygonaceae to famil ies of Caryophyll alcs, Gian­ lationships should not be overlooked when per­ nasi e t al. ( 1992) used a number of phenetic and forming more restricted anal yses of these a nd other "phylogenetic" methods, including a Fitch-Ma r­ gene sequences. goliash de ndrogram based upon genetic distances (Kimura, 198 1; Felsenstein, 1990), maximum par­ INTERNAL SUPI'ORT FOR HIE BRO AD TOPOLOGY: simony (PH Y LI P, Felscnstein, 1990; PA UP. Swof­ AN AD HOC ANALYSIS OF SUBCLA SS HAMAMEUDAE ford, 1991), and maximum lik elihood (ML; Fel­ senstein, 1981). All three methods produced the For purposes of suggesting an appropriate mall­ same lack of resolut ion concerning rela tionships of ner to examine internal support of fa mi lies tradi­ these fa milies; to sta te tha t Plumbaginaceae and tiona ll y recognized as a na tural group, we selected Polygonaceae a re " not cl osely rela ted" (meaning one of the most controversial subclasses, Hama­ " closely sim ilar") to Caryophyll ales does not pre­ melidae sensu Cronquist (198 1 ), for which we ha ve clude them nonetheless fr om being closest relatives. da ta from 18 of 24 families. Numerous phyloge­ From the pe rspective of result s presented here (Fig. netic and systematic studies of the Hamamelidae 9A, B), spurious similariti es in the data analyzed have been completed (see va rious a uthors in Crane by Ciannasi e t a l. ( 1992) appear to affect equa ll y & Blackmore. 1989). The morphological features results of a ll three tree-building me thods: Gos.syp­ suggesting a close rela tionship a mong these families ill m simult a neously a ttracts highe r aste rids a nd a re largely those associated wi th the temperate magnoliids (in our trees Gossypium is well imbed­ amcntife rous syndrome, and these clearly could be ded a mong ros icls; rosid II, Fi g. l OA, 13). (See the result of parallel modificatio n in unrela ted lin­ Olmstead el a I. , 1992, 1993, this issue, for a n eages. In their developme ntal characteristi cs and example and discussion of the effec ts of taxon wood a na tomy, families of Hamamelid ae are pa r­ sampling.) ti cularl y heterogeneous (Cronquist, 1981; Crane Prospects for improveme nts of tree-building & Bl ackmore, 1989). In pe rformi ng th e broad methods with greater numbe rs of taxa exist (Pe nny a nal ysis, we sought to avoid a priori ideas about et aI. , 1992). In the e xa mple of Ciannasi e t a l. wha t consti tu ted monophyle tic subgroupings of a n­ (1992), none of the methods employed succeeded giosperms, but for this heuristic example we have in eliminating what we interpre t as branch allrac­ accepted ad hoc the outgroup rela tionships fou lld tions due to the small number of taxa sa mpled a nd in the gene ral stud y. use of distantly rela ted outgroups. Phylogenetic W e sel ccted 72 species that included all mem­ studies using Illorphological c ha racters for a closely bers of Hamamelidae and their immedi a te sister related group of rosid or asterid famil ies would taxa as identified in trees from both searches . No likely be a ffecled adversely by outgroups of mag­ a ttempt wa s made to sele ct spec ies that would noliids or monocots. This phenomenon is perhaps reproduce the particulars of the general topologies . e ven more probable with nucleotide da ta in which A tree searc h under the Fitch (equal we ights) cri. homology is initiall y assessed only by nucleotide te rion using 2,000 ra ndom sequence additions, positi on a nd character states are restric ted to the M ULPARS, STEEPEST DESCEN T, and NN I same four a lte rna tives. Chara <.:l e r " homology" for branch-swapping (but permitting onl y 10 trees 10 distantl y related taxa can be easily de te rmined wit h be held at each step) found only one isla nd at rbe L data (i.e., a given nucleotide positi on in this ma ximum pa rsimony (i. e ., a ll trces could be found size-conserved gene; "prim a ry homology," de by single branch swaps usin g anyone of them as Pin na, 1991 ), but this does not mea n that assess· a starting tree; d . Maddison, 199 1). Add it ional as ments of characte r·state homology (synapomor­ well as shortc r islands could stilt exist but are un­ phy) a re less subjec t to homopla sy than with othe r likely after 2,000 repetitions (th is type of search da ta. required about 24 hours to complete on a J\'1a c­ If ph ylogenetic a nalyses of nucleotide data can­ intosh Quadra 950 with 20 MB of RAl\'O . After not be expected to reveal rela tionships when sam- random addition searches we re comple ted , the trees 540 Annals of the Missouri Botanical Garden

found were used as starting points in a single anal­ arc shown to be grossly pol yphyletic. Besides the ysis with MULPARS on a nd TBR swapping to amentiferous syndrome, the major trait of Ha m­ completion; this process should find all equall y par­ amelidae is the prese nce of tannins, whi ch is like· simonious trees in the single isla nd identified. OUf wise compatible with a relationship to the other search found 36 equa ll y parsimonious Filch trees, tannin.cont aining families, Fabaceae, , three of whic h were optimal unde r the weighting Saxifragaceae sensu stric to, and Crassulaceae. criterion of Albert c l al. ( J 993, this issue; see Many a uthors (e.g., in Crane & Blac kmore, 1989) Materials and Methods). One of these was randoml y have discussed these families in terms of which are selected and is shown wit h Fitch branch lengths " lower" a nd which a re " higher" families. This (ACCTHAN optimiza tion; Fig. 16). The tree length distinction fi nds some support from these results; wa s 2,234 steps, the C. I. was 0.288 (excluding the " lower" groups either stand in an isolated po­ unique characters), and the H. I. wa s 0.532. The sition near the base of the eudicots (Eupteleaceae, maximum parsimony trees also were used to sear<.: h Pla ta naceae, a nd T etracentraceae- T rochode ndra­ for trees up to five steps less parsimollious unde r ceae) or basal within rosids (rosid IV, Fig. 2A; the Fitch c riterion; the FilTER TREES option wa s Daphniphyllaceae, Hamamclidaceae, a nd Cercid i. used to identify trees a t each length, and a strict ph ylla ceae), whereas most " hi ghe r" hamamclids consensus tree at each step was cotnput ed. The (Casuarinaceae, Fagaceae, Moracae, Ulnm ceae, and number of steps less pa rsimonious at wh ich each Urticaceae) demonstrate a we ll support ed relation­ topological component decayed was recorded (Fig. ship to Fabaceae or Rosaceae (d > 5 on two 16; "decay valu es" a re shown below the branches: branches a t the base of the largest clade in Fig. "dO" indica tes that the bra nc h is a polytom y in 16). The position of Lei/lleri.o in Sapindales ncar the strict consensus of the maximum parsimony Burseraceae is corroborated by a shared suite of trees, "d I " indicates tha t the branch is a polytomy secondary compounds and presence of intercellular in the consensus tree at one step less pa rsimonious, resin canals. etc. ). The example presented above is not in tended to This analysis identified eight lin eages illto which be more tha n a superficial phylogenetic treatment members of Hamamclidac fell (Fig. 16): 1','ucol1l ­ of families tradi tionall y refe rred to Ha mamclidae. m.ill, clade A, sister 10 Aumba (Cornaceae) and It is meant to serve as a n example of how the Ca.rryo (Carryaceae) and nested within an asterid general topology, whi ch itself is suspec ted of being clade; Hamamelid accae, Cercidiphyllaceae, elc., suboptimal and presently cannot be examined by clade 13 , in a se ries para phyle tic to Saxifragaceae­ decay analysis because of it s size, may ide ntify a Crossulariaceae; Fagales, etc., clade C, sister to relevant analys is within wh ic h questions of opti­ Fabaceae- : Urticales, cl ade D, sister malit y and rela tive support can be addressed. We to Rosaceae; Leitrlf'ria , clade E, nested within fam ­ are pleased that gene ral relationships found in the ilies of Sapindales; isolat ed Trochodendrales, clade broad a nalyses hold up we ll when addressed in this F, siste r to Buxaceae; I->Iatrlflll$, clade C, situated and other lnore restricted investigations, none of in a heterogeneous group; and the last, Ic'lIp l dea, whi ch have found vast ly diffe rent topologies. clade I-I, sit uated amoll g Ranutl culales. Many of these clades a re well supported internally, decaying ON T HE I NFOltMAT1VENES."; OF ALL SU BSTtTUTIONS at three or more ste ps less parsimonious (sonte, such as Trochodendrales- Ru xaceae and the PIa/­ The majority of c ha rac ter-state cha nges in pro­ anus assemblage a rc weak ly supported as mono­ tein-coding genes have been demonstrated to occur phyle ti c lineages but clearl y a re not members of a t third positions wit hin codons, and numerous other well supported groups, leaving thelll in iso­ empirical studies have shown third posit ion substi . la ted positions a pa rt from other I-Ia mamelidae). tutions to be more abundant in this and other data Results of this restricted a nalysis a re congruent se ts. Some worke rs ha ve experimented with dis­ with the topology found in the broad sea rches (Figs. carding third position substitutions from their anal­ I, 2) and indicate the level of int e rnal support yses or analyzi ng nucleotide sequences inferred demonstra ted by (h el data. Future studies could from amino acid data (w hic h sta nda rdi zes all syn­ combine morphological data with this Inolecul a r onymous substitut ions; Martin & Dowd , 1991). In matrix and perform constraint experiments in which several studies (Conti e t a I. , 1993, this issue; Don­ va rious rearra ngements of these taxa are examined oghue et a I. , 1992; Kim e t a I. , 1992; Smith el aI. , for their relati ve degrees of parsimony. 'rhe general 1993, this issll e). all three codon positi ons ha ve (;onclusion from this example is tha t the Ilama­ been found to ex hibit sim ilar levels of homoplasy melidae do not form a monophyletic lin eage; they (a nd pe rhaps similar ra tes of cha nge pe r sit e as Volume 80, Number 3 Chase et al. 541 1993 Phylogenetics of Seed Plants

well) . 11 has gene ra ll y been a ssumed that, for most for hypotheses tha t produc tion of di ve rse fl ower genes al some unspeci fi ed highe r t.axo nomic level, types took place ra pidly a nd earl y in a ngiospe rm third positio ns become saturated a nd tll creforc a rc evolut ion . more li kel y to be uninformative or even misleading Nonetheless, othe r explana tions could explain (Swoffo rd & Olsen , 1990), hence the "Iogic" for patterns of uneven distribution of c ha rac ter-s tate disca rding third positions and using only nonsyno n­ c hanges identifi ed by this and many other a nalyses ymous substitutions . In the empirical studies c it ed of molecula r da ta (e.g., Jansen & Pa lme r, 1988). above, multiple substitutions apparently have oc­ These include episodic c ha nges of rates of sequence c urred a l certain first a nd seco nd positions a l low divergence and effects resulting from a n uneven taxonomi c levels (i.e., with in families). Variable di stribution of sa mpled taxa tha t could cause cer· third positions a ppear as consistent on average as ta in c ha racte r·sta te c ha nges to be assigned to sev· va ria ble first o r second positions, thus making the m eral more terminal bra nc hes ra ther than to a single. as reliable al reconstructing relationships (as a class more inte rnal one. The laller effect could be ex­ pe rhaps they a re better because they a re more plained if subsequent substitutions over the long nume rous; Donoghue e t a I. , 1992). history of groups obscured or even e liminated ev­ Whethe r factors such as codon usage (tRNA idence of synapomorphic cha racte r states for these biases known to favor onl y a subset of the possible groups. Greate r sampling of varia tion within te r· third position nucleotides fo r a give n amino ac id ) minal groups ma y result in character sta tes beill g might contribute to the relatively highe r consiste n· optimized to a single, more basa l branch, wi th reo c ies of some synonymous substitutions is a topi c versals a nd further cha nges in terminal lineages. that should be investiga ted . Codon usage is usuall y as opposed to being optimized to appear as inde­ studied in a pairwise fa shion, but it would be bette r pende ntly arisin g in severa l te rminal groups. Sev. exami ned from a phyloge ne ti c pe rspec tive. Most e ra l other factors. including the specific me thod of molecular models, including the one (a modifi ed optimization and inte rpre tation of the robustness versio n of that of Kimura , 1980) that is the basis of groups with "weak" characte r·sta te suppo rt, a rc for the we ighting scheme used he re (A lbe rt e t a I. , conside rations relevant to the distribut ion of ho· 1993). a rc a rguably too simplisti c in the ir as­ moplasy. This is anothe r area in whic h fut ure em · sumptions concerning pa tt e rns of nucleotide sub· phasis should be placed beca use advances (s up. stitutions. Ultimately. we could imagi ne that knowl­ ported by theoret ica l c onsiderations) in edge of processes, specific to rbe L. responsible for understanding the e ffects of c ha rac te r·state opti . the highe r consistenc ies of some synonymous sub­ mization could improve eva luatio n of critica l inner­ stitutions mi ght permit construction of a lternative most bra nches. weigh ling mode ls tha t would extrac t more historical Lineage.specific ra te asymme try, a potential signal from rbc L (A lbe rt e t a l.. 1992a). Until we contributor to spuriolls branc h atlractions (He ndy bet ter unde rsta nd evolutiona ry constraints on all & Pe nny, 1989; Albert et al.. 1993 , this issue ). positi ons within a coding seque nce. it seems most is signifi cant for rbe L (Bousquet e t a l. . 1992), yet prude nt to use me thods tha t do not eliminate any may not be extensive enough be tween lineages to evid ence of the underlying process. be problematic . Initial estimates of total seque nce dive rgence ra te pe r year produced a ra nge of 5- 7 x 10- 10 for (i) panicoid versus pooid grasses [lR,\ NCH I.ENGTH INEQUA LIT I ES AND UNEQUAL RATES (Doeble y et a I. , 1990). (i i) Petunia versus To bacco, The foss il record suggests a n earl y. a nd pe rhaps and (iii) Colchicllm versus LiliulII (We ndel & Al ­ rapid, di versification of a ngiospe rms (e.g., Doyle bert , 1992, from data presented in Albert et a I. , & Hi ckey, 1976). Several diffe rent categories of 1992a). Subseque nt studies of woody magnoliid specialized flow e rs appeared nearl y simulta neously taxa with austra l disjunction pa llerns (and for which in geological te rms. These included strobiloid types pla te tectonics suggest appropriate divergence times) c harac teristic of , simple r fo rms simila r have indicated dive rgence rates ranging between to those of La urales, highly re

using rela ti ve rate tests (Gaul e l aI., 1992) delll ­ related taxa, and . like the effects of ancient hy. onstrated five-fold variation for rb el , whi ch still bridiza tion, those of ancient polymorphisms would fall s within the range cited above. Although 5- \ 0- li kel y be minor relative to subsequent gene tic di ­ fold differences in ra les wou ld appear potentially vergence. If lineages that contai n polymorphisms to contribute e normous bra nc h. length differe nces, di verged in a closely spaced manner and uneven the small overall rales in volved mean that great sorting did take placc, it is unlikely that any evi­ divergence times would be a more im porta nt fa ctor. dence of such variation within a progenit or could In addition, differences in rates a ppear 10 be highly be identified as such over the great amount s of lineage.correlated (for example, in graminoids; Caul time involved in this study. Furthermore. poly. cl 1.11., 1992) rathe r tha n ra ndom, and if more morphisms arc short ·li ved and are undocurnented extensive sampling at lower taxonomic levels is for conserva ti ve. single-copy loci, such as rbe L. possible in these " fa s' " clades, then effects of rate At most these effects wou ld be highl y loca li zed diffe re nces can be ofTsc l. among groups of closely rela ted termi nal taxa and. Most if no t a ll rbr L rales will probably exist with adequate taxon sampling. would not be ex ­ within a relati vely narrow "window," pcrhaps ap­ pected to pert urb greatl y the results.

proximating the ra nge ill ustrated above ( 10 10' Lateral transfers not in volving exchange or ga­

10 I I, Y_ A. Albert, M. \V. Chase & J. F. \Vclldcl. metes (by unknown mechanisms) may ha ve oc· unpublished). If tha t is correct. lineage-specifi c ra tc curred between major li neages in the past (i.e., of inequalities arc unlikely to be a primary factor in rb c; L fr om a purplc bacterium to a red alga l an· branch attraction : rather. asymmetrical di vergence cestor: Mord en et al., 1992). Such transfers pres· times wou ld be implicated because the product or ently appear rare among land plants and secd plants rate and time is the central pa rameter in consid ­ in pa rticula r; furthermore, artifi cial transforma­ era tions of potential systematic errors (see Albe rt tions arc relatively difficult and oft en havc a de· et a l. . 1993. this issue). Thus uneven sampling or stabilizing or transient effect on transformed plants. extinc tion of lineages l'I'l1.ly present greater problems While we must admi t that this is an unknown a rea than do differences in rates. that could have played a role in certain anomalous placements in the rb d . trees (e.g., MOll r;lI;a and Vah/ia of Saxifragaceae sensu lato among ; EFFECTS OF' LATERA L GENOME T ltA NS FEIl , ANCESTltt\l.. Fig. 15), at the same time we are not prepared to PO LYM O RPHI S ~I S , AND DIFFERENT advoca te it as a scena rio ull tit trees based on other MODES OF INII ERITANCE data demonstra te a pallern consistent wit h such The trees present ed here represent onl y infor. hypotheses (a s in the example of Mord en et a I. , mation from a single gcne, and factors pecul ia r to 1992). Usin g evid ence from studies of second ary its evolution could lead to erroneous results. Several chemistry and development , Morgan & Solti s of these phenomena are dis('u ssed below, but we ( 1993, this issue) build strong cases for the highl y feel that their impact is likely minim al. Genome dispersed groupings found in their stud y of Saxi­ tra nsfers would result in a ll or parts of a genome fr agaceae sensu lato. Furthermore, many of their being phylogeneli call y coherent (transferred as a fmd ings were also congruent with recent in vesti· unit ) at the time transit occurs (Doyle. 1992: Riese· gations of non-molecular characters. Rat her thall berg & Soltis. 199 1). bu t before and after move­ rcsort ing to explanations in volving lateral transfers ment most characters wi thi n genomes evolve in· of chloroplast.s by mechanisnls about which we call dependently (alt hough still lin ked if on the salllc onl y speculate. we would prefer first to exa mine chromosome) and ought to be ex pected to con tail! specifi c cases from a cladistic perspec ti ve rather historical evidence. Hybridiza tion is unlikely to in · than from thai of current laxonomic schemes. fl uenee ph ylogenetic analyses except at lower tax ­ An rbeL tree is not solely a ma ternal trec. Al­ onolHic levels. Even matings be tween divergent though chloroplast transmission is principally mao parents still occur wi thin port ions of families (and ternal and uni parenlal, several groups ex hibit a usually among closely related species in a genus) pa ternal (e.g., conifers) or biparental pattern (Neale ru ther than between fam ilies. Parental taxa at the et al.. 1986: Szmidt et a I. , 1987; Whatley, 1982; time of an a ncient hybridiza tion also were likely Wagner el a I. , 1987; Corri veau & Colcman, 1988; closely related, and in these gcnomcs hi ghly COII­ White, 1990; Owens & Morris, 199 1). Others that served loci, sllch as rl!l:L. wo uld have been similar have been thought most proba bl y ma ternal, such or even identical. as Uriod elldroll and Magnolia (Corriveau & Cole· Ancestral pol ymorphisms (Patnilo & Nei, 1988: man. 1988). cOll sistelltl y ex hibit 5- 15% paternal \'(Iu. 199 I; Doyle. 1992) also a ffect onl y closely inheritance (Sewell el al.. 1993). The potential Volume 80, Number 3 Chase et al. 543 1993 Phylogenetics of Seed Plants

effects of a mixed pattern of inheritance on gene An additional implication of adding more taxa trees a rc a lso unlikely to affect studies at interfam­ is that the results presented he re may not be stable ilia I levels and above. or refl ective of those of an analysis of two, three, or four times as many sequences. We point to two regions of the general topology that a re morpho­ GENES VERSUS TAXA logicall y quite hete rogeneous and that we believe One of the more pe rsistent controversies sur­ a re possibly gene rated by undersa mpling: hama­ rounding molecul a r phylogenetic studies has cen­ melid II (Fi gs. 2A, B, 7 A, B) and cosid I (Figs. tered on whether results, al a given taxonomic 2A, B, I IA, 8 , C). In the former, the c ritical taxa level, will be "improved" more by adding additional needed to provide more appropria te relationships taxa seque nced for the same gene or by adding could be some of the still -unsa mpled families, but sequence analyses of additional genes for the same these groups (represented by Lambcrlia, Nclumbo, set of taxa (analyzed simult a neously or each per­ PLata nilS, and Sabia) could just as li kely represent formed inde pe nde ntly for assessments of congru­ the isola ted relicts of now la rgely extinct lineages. ence; Pamilo & Nei, 1988; Wu, 1991). From the In the case of rosid I, only a small percentage of standpoint of corroboration, phylogenetic studies the families has been sa mpled (for instance, only of othe r da ta sets a re a bsolutely cruc ia l. Never­ 7 of 24 families in the Violales: Dill elliidae), a nd theless, it seems quite clea r from our wo rk on these it is in such a case that we might predict " curious" data sets tha t ideas of relationships have changed sister-group relationships. Topological instabilit y considerably as more taxa have been added. We often occurs in cases where long te rminal branches suspect that no single gene sequence can provide are next to short internal branches, tha t is, hy­ reasonable hypotheses of relationships of seed plants pothesized rela tionships are drasticall y a lt ered by when sampled unc ritically and superficiall y, but the addition of related taxa that " break up" long perhaps e rroneous results from one loc us would be branches. Two surprisin g pairs of sister taxa from "corrected" by stronger signals present in the oth­ Search I, Erythroxy lum- Vio!a a nd Oellna- Dr),­ ers . Da ta to evalua te this most c ritical question do pct.cs, a re quite dive rgent and connected to eac h not exist: would separa tely a nalyz in g 20 gene se­ other by relatively short branches (Fig. II A; branch quences for the same sct of 25 to 30 taxa produce lengths show n onl y in II B) . In Searc h II , the we ll supported relationships? addit ion of R einwardlia displaced Erythroxylum W e are convinced tha t adding re presentatives from Viola to Drypctcs, leaving Oc/Uta to stand of the still numerous and diverse families absent isola ted fr om any other taxon (Fig. I I B). In suc h from this a nalysis has the potential to e nha nce situations, assessments of re la tionships a re difficult , assessments of rela tionships at all levels (despite but, as more closely related species arc added, obviolls computational compli ca tions). A great deal distinguishi ng synapomorphies for families and of the varia tion present wi thin the gene sti ll remains groups of families as distinct from autapomorphies to be sampled taxonomicall y. S tudies in several for individual species will become more re liable. groups of plants indicate that rbe L often can be W e would not argue that the rela tionships found valuable at rathe r low taxonomic levels (A lbert, in for rosid I a re " bette r-' in Search I than in Searc h press, in the slipper orc hid s; S. Gra ham, B. Morton II (o r vice versa), but ra ther point these out as & S. Barrett, unpublished, in Eichhornia; R. Price areas of the trees tha t require additional sampling & J. Palmer, unpublished, in Pclargonium; S. and in whi ch many of the rela tionships suggested Williams & M. Chase, submitted, in Droscra; Xi­ by these analyses of rbel have little or no mor­ ang et a l. , 1993, this issue, in Comus). W hat can phological support. be concluded fr om these studies is tha t a lt hough it is absolute ly critical tha t more genes be studied VALUE OF THE BHOAD ANALYSIS to provide corroboration, it is unlikely that any of these studies will make a contribution to under­ We view the relatively robust interna l support standing seed-plant rela tionships if the level of sam­ found by evaluations of portions of the broad a nal­ pling is too sparse. Only empirical studies will ul ­ ysis as an indication that rbe l sequences contain timately r esolve the "genes versus taxa" information relevant to the evolut ionary history of controversy. We estima te that by the time this a ngiosperms (Fig. 16, on Hamamelidae and most pape r is in print, more than 1,200 sequences of of the other papers in this issue; in particular see rbe L from seed pla nts will exist, and future studies Conti et al., 1993, and Hodman et a I. , 1993, for will undoubtedly benefit from this enormous da­ whic h comparable cladistic analyses of non-molec­ tabase. ular data are also available and compared). \" ell - 544 Annals of the Missouri Botanical Garden

characterized families and groups of fam il ies (as treated these groups in a manner somewhat simil a r evidenced in several morc restricted studies of mor­ to our topology); (iii ) relationships of Nepe nthaceae phology, anatomy. and secondary chemistry; e.g., and Droseraccae to Ca ryophyllid ae sensu la to (Fig. Rodman, 199 1a, b; Hufford , 1992) are largely 9A, B); and (iv) specific associations of numerous congruent with OUT results. Other evide nce sup­ problematic genera like Oillcnia (among rooids near ports the monoph yly of the a ngiospe rms (Doyle & Caryophyllidae. Fig. 9A, or near the base of the Donoghue, 1986, 1992), monocots (Da hlgren ct astc rids, Fig. 12 8 ; see Olm stead et a I. , 1993, this a!', 1985), and e udicots (Donoghue & Doyle, 1989). issue), Impatien s (nea r me rn be rs of Ebellales and Lik ewise, the posil ion of Ceratophy lfllfll as sister Ericales; Fig. 13A, B), a nd Ne/flmbo a nd Lam ­ to the rest of the angiosperms compa res favorably herti.a (among lower ha mamelids; Fi g. 7 A, B). wi th the fossil record (Les, 1988; fossil fruils from Pl acement of most of these genera a nd families 120 million years ago a rc id entical 10 extant fruils, has varied substa ntiall y among recentl y proposed D. Dilc her, pe TS. comm.). The status of the fa mil ies classifications (alt hough no one has suggested the of the Papavera les- Ha nuncula les collectively as rela tionships found he re), and their posi tions in this siste r to the rest of the eudicots (Figs. I, 2) also :; tudy will undoubtedly add to the controversies. find s al'nple exte rnal support (Donoghue & Do yle, Because addit ional lineages were present in 1989, among others). Search II a nd different me thods we re used to con· The results reported he re do find some ··sur­ structlhe trees, it is impossible to evalu

on distantly related groups (in recent ta xonomic these ana lyses represent a n a tte mpt to improve schemes) fell into or near their group of interest. both our unde rsta nding of seed -pla nt evolution and Prime examples of this are the close ph ylogenetic m e thods of phylogenetic inference. Corroboratio n relationship of Nepenlhaceae and Droscraceae to by other data sets a nalyzed in a similar fashion is families of Caryophyllidae, Pill osporaceae to Api­ by far the most significant measure of relationsh ips aceae- Araliaceae, and Capparales (both proposed he re, and we hope this process of eval­ "dilleniids") to Sapindales, and Coroki.a. (Coma­ ua tion will be inner vated by our e fforts. ceae) to Asle raceae. Broad analyses a re thus important in providing llTt:RAT URE CIT t]) evaluations of a priori assumptions about appro­ ALBERT, V. A. Phylogeny of the slipper orchids (Cypri. priate sets of study taxa for more foc used and pedioideae: Orchidaceae) f rOIIl congruent morpho­ logical and molecular data sets. Nordic J. Bot. (in rigorous studies; they should be formalized so th at press). someone takes the initia ti ve to perform them. Gov­ --- & B. D. MISHLt:H. 1992. On the rationale and ernmental funding agencies should facilita te stud ­ utilit y of weighting nucleotide sequence data. Cladis· ies, such as the one presented he re, tha t a rc well tics 8: 73- 83. ---, M. \V. CHASE & B. D. MI SIIL EH. 1993. Char. beyond the scope of individual laboratories. This acter·state weight ing for cladistic analysis of protein· s tud y demonstrates the pote ntial of this kind of coding DNA sequences. Anll. Missouri Bot. Gard. a na lysis, but no single individual or la boratory could 80, 752- 766. have received formal support for sampling this ---, B. D. MISHLER & M. W. CHASE. 1992a. Char. diverse set of taxa a nd performing the phylogenetic acter-state weighting for restriction site data in phy­ logenetic reconstruction, with an example from chlo. analysis; it would have been deemed by reviewers roplast DNA . Pp. 369- 403 in P. S. Soltis, D. E. too broad and too u nfocused. Allhough extra mura l Soltis & J. J. Doyle (editors), Molecular Systematics funds supported mos t of the other s tudies in this in Plants. Chapman & Hall, New York. issue, no suppo rt was received specificall y for the ---, S. E. W lI_UAMS & J\.t. W. CIIA::i £. 1992b. broad analysis. It is the investment in individual Carnivorous plants: Phylogeny and structural cvo­ lution. Science 257: 1491 - 1495. s tudies that justifies a further expenditure to sup­ ARCHI E, J. W. 1989. A randomization test for phy lo­ po rt syntheses that supply a n esse ntia l overall per­ genetic information in systematic data. Syst. Zoo!. spec tive, even though they may be necessarily 38,219- 252. approxima te. Bo USQU ET, J., S. H. STAAL'S."', A. O. DOER KSE;\' & R. A. PRIC£. 1992. Extensi\'e va riat ion in evolutionary The benefit s of performing this largest- yet phy­ rate of rbe l gene sequences among seed plants. Proe. logenetic s tudy of seed plants lie not onl y in support Nat l. Acad. U.S.A. 89: 7844- 7848. of specific relationships h ypo thesized by othe r s tud­ BA EM ER, K. 1988. The limits of amino acid sequence ies (e.g., the sis te r group sta tus of the les data in angiosperm phylogenetic reconstruction. Evo. a nd higher astc rids proposed by Hufford, 1992) lut ion 42: 795- 803. ---. 1990. Combinable component consensus. Cla­ and in identification of previously u nhypothesized dist ics 6: 369- 372. monoph yle tic groups. This stud y a lso presents a ---, C. J. H UMP HRIF-'i, B. D. MIStl LEA & S. P. CH UR­ comprehensive, explicit hypothesis fo r higher level CHI LL. 1987. On cladistic relat ionships in green relations hips, permitting a nd encouraging initia tion plants. Taxon 36: 339- 349. CONTt, E., A. FI SC IIBAGI & K. J. SYTS MA. 1993. Tribal of studies evalua ting other c ha rac te r systems that relationshi ps in Onagraccae: Implications frolll rbel may show cong rue nce wit h the major li neages of sequence data. Ann. Missouri Bot. Card. 80: 6 72- seed plants described here. \V e have performed 685. none of the essential expe riments tha t these results CORRIVEAU, J. L. & A. \1;' . COLEMA N. 1988. Rapid suggest (topolog ical constra ints, removal of c har­ screening method to detect potential biparental in­ heritance of plastid DNA and result s for over 200 ac te rs, combining morphological wit h molecular angiosperm species. Amer. J. Bot. 75: 1443- 1458. data, etc_) a nd have not developed implications CRANE, P. R. 1985. Phylogenetic anal ysis of seed plants these topologies m ay have for specific character and the origin of the angiosperms. Ann. Missouri Bot. transfo rmations in seed pla nts o r molecula r evo­ Card. 72: 7 16- 793. 1988. Major clades and rela tionships in the lution of rbc L or RuBisCO. Believin g that serious "higher" gymnosperms. PI' . 2 18- 272 in C. B. Beck consideration of the significance of these gene ra l (editor), Origin and Evolution of Gymnosperms. Co­ topologies is best ha ndled a t the more m a nifes t lumbia Univ. Press, New Yor k. level of othe r pape rs publis hed he re (and else­ --- & S. BLAC KMOIl E. 1989. Evolution, System. whe re), we have c hosen instead to emphasize rea­ atics, and Fossil History of the Hamamelidae. Sys­ tematics Assoc., Clarendon Press, Oxford. sons for caution. \Vc are content to present the CRONQ UIST, A. 1981. An Integrated System of Classi­ findings of these s tudies as examples of potential tication of Flowering Plants. Columbia Univ. Press, benefits and pitfalls of su ch exe rcises. At the least, New York. 546 Annals of the Missouri Botanical Garden

DAHLGREN, R. T. 1980. A revised system of classifI­ approach using the bootstrap. Evolution 39: 783- cation of the angiosperms. Bot. 1. Linn. Soc. 80: 791. 9\ - 124. 1990. PHYLiP manual, version 3.3. Univ. --- & K. BIIEMEIl. 1985. Major clades of the of California, University Herbarium, Berkeley, Cal· angiosperms. Cladistics I; 349- 368. ifornia. ---, H. T. CUFFOllD & P. F. YEO. 1985. The FITCH, W . M. 197 1. Toward defining the course of Families of the : Structure, Evolu­ evolution: Minimum change for a specific tree to· tion, and . Springer-Verlag, New York. pology. Sys\. Zoo!. 20: 406- 416. DoEIlLEY, J., M. D URBIN, E. M. GOL[NIlERC, M . T . C LEGG CA UT, B. S ., S. V. MUS E, W. D. CLA HK & M. T. CI.EGG. & D. P. MA . 1990. Evolutiona ry analysis of the 1992. Relative rates of nucleotide substitution at large subunit of carboxylase (rbcL) nucleotide se· the rbcl locus of monocotyledonous plants. J . Molec. quence among the grasses (Gramineae). Evolution Evol. 35, 292- 303. 44, 1097- 1 lOB . ClAN NA SI. D. E., C. ZUIlAWSKI. C. lEAH '" & M. T. CLEGG . DoNOCH UE, M. 1. & 1. A. DOYLE. 1989. Phylogenetic 1992. Evolutionary relationships of the Caryophyl. studies of seed plants and angiosperms based on lidae based on comparative rbe l sequences. Syst. morphological characters. Pr. 181 - 193 ill B. ­ Bot. 17: 1- 15. holm, K. Bremer & H. Jornvall (editors), The Hi· GOI.DB LATI, P., P. C. HoclI & L. M. MCCOOK. 1992. erarchy of life: Molecules and Morphology in Phy. Documenting scientific data: T he need for voucher logenetic Analysis. Elsevier Science Publishers, specimens. Ann. Missouri Bot. Card. 79: 969- 9 70. Amsterdam. HAMBY, R. K. & E. A. ZIMME,:H. 1992. Ribosomal RNA ---, R. C. OL~t STEA!), J. F. S~ II T H & J. O. PAUtEll. as a phylogenetic tool in plant systematics. Pp. 50- 1992. Phylogenetic relationships of Dipsacales based 91 in P. S. Soltis. D. E. Soltis & 1. 1. Doyle (editors), on rbel sequences, Ann. Missouri Bol. Card. 79: Molecular Systematics in Plants. Chapman & Ha ll , 333- 345. New York. ---, J. A. DOYlE, J. CAUTIIIElt, A. C. K LUGE & T. '·h.HH IS, P. 1. & R. D. H AHTLEY . 1980. Phenolic con· RowE. 1989. T he importance of fossils in phylog. stituents of cell walls of monocotyledons. Uiochem. eny reconstruction. Ann. Rev, Ecol. Sysl. 20: 431 - Sys!. Ecol. 8: 153- 160. 460. H ENDY, M. D. & D. hNNY. 1989. A framework for DoWNIE, S. R. & J. D. PALMEIl, 1992, Use of chlo· the quantitati ve study of evolutionary trees. Syst. roplast DNA rearrangements in reconstructing plant Zool. 3B, 297- 309. phylogeny. Pp, 14- 35 ill P. S. Soltis, D. E. Soltis HUfFOIW, L. 1992. Rosidae and their relationships to & J. J , Doyle (editors), Molecular Systematics in other nonmagnoliid : A phylogenetic Plants. Chapman & Hall , New York. analysis using morphological and chemical da ta. Ann. DonE, J. A, & M. J , DoNOGtHJE. 1986. Seed plant Missouri Bot. Card. 7 9: 218- 248. phylogeny and the origin of the angiosperms: An JA NSEN, R. K. & J. D. PALM EIl. 1988. Phylogenetic experimental cladistic approach. Bol. Rev . 52: 321 - implications of chloroplast DNA restriction sit e va ri· 431. ation in the Mutisieae (A steraceae). Amer. J. Bol. & , 1992. Fossil s and seed plant phy. 75, 751 - 764. logeny reanalyzed . Brillonia 44: 89- 106. KA LLEHSJO, M., J . S. FAil illS, A. G. KLlICE & C. Bull'. ----'- & L. J. HI CKEY. 1976. Pollen and from 1992. Skewness and permutati on. Cladistics 8: 2 75- the mid·Cretaceous Potomac Croup and their bearing 287. on early angiosperm evolution. Pp. 139- 206 in C. KIM . K.·J., R. K. JA NSEN, R. S. W AL LA CE, H. J. MtCHA ELS U. Beck (editor), Origin a nd Early Evolution of An· & 1. D. PALMER. 1992. Phylogenetic implications giosperms. Columbia Univ. Press , New York . of rbel sequence variation in the Asteraceae. Ann. Doyu:, 1. J. 1992, Gene trees and species trees: Mo· Missouri Bol. Card. 79: 428- 445. lecular systematics as one·character taxonom y. Sys\. KI MURA, M. 1980. A simple method for estimating Bot. 17, 144- 163. evolutionary rate of base substitutions through com· --- & HOtTO N, 1991. Di versification of early an· parative studies of nucleotide sequences. J. Molec. giosperm pollen in a cladistic context. Pp, 169- 195 Evol. 16, 111 - 120. in S. Blackmore & S, H. Barnes (editors). Poll en and 1981. Estimation of evolutionary distances Spores. Clarendon Press, Oxford. between homologous nucleotide sequences. PrO(; . Nat!. DUVALL, M. R .. M, T. CLEGG, M. W, CII ASE , W. D. Acad. U.s.A. 7B, 454- 458. CLA RK , W. 1. KH ESS, H. C. HI !.!....';, L. E. ECLIIAHTE, KIlON, K. A. & M. W. ClIASE. 1993. Systematics of J. F. SMITH, B. S. GAlIT, E. A. ZI~ tM EIl & C. H. the , Empetraceae, Epacridaceae and reo lEAHN. 1993. Phylogenetic hypotheses for the lated taxa based upon rbel sequence data. Ann. monocotyledons constructed from rbel sequence dala. Missouri Bot. Card. 80: 735- 7 41. Ann. Missouri Bol. Card. 80: 607- 6 19. l ES, D. H. 1988. The origin and affinities of the Cer· FAITH, D. & P. CHANSTO N. 1991. Could a cladogram atophyll aceae. Taxon 37: 326- 345. this short have a risen by chance alone? Cladistics 7: --- , D. K. CARVIN & C. F. WIMPEE. 1991. Mo· 1- 28. lecular evolutionary history of ancient aquatic an· FEI..SENSTEIN, J . 1978. Cases in which parsimony or giosperms. Proc. Natl. Acad. U.S.A. 88: 10119- compatibilit y methods will be positively misleading. 10123. Sy", Zool. n 401 - 410. lOCO NTE, H. & D. W. STEVENSON. 1991. Cladistics of ---. 1981. Evolutionary trees fr olll DNA sequen<.> the Magnoliidae. Cladistics 7: 267- 296. es: A maximum likelihood approach. 1. Molec. Evol. MADDISON, D. R. 1991. Discovery and importance of 17, 36B- 376. multiple islands of mosl'parsimoniolls trees. Syst. Zool. - - - 1985. Confidence limits on phylogenies: An 40, 315- 32B. Volume 80, Number 3 Chase et al. 547 1993 Phylogenetics of Seed Plants

MAIl TIN. P. G. & J. M. Do WD. 199 1. Studies of an · consequences of cytoplasmic gene flow in plants. giosperrn phylogenies using protein sequences. Anf). Evol. Trends PI. 5: 65- 84. Missouri Bot. Ca rd. 78: 296- 337. RI TLAN D, K. & M. T. CLEGG. 1987. Evolutionaryanal. MrCHAEI..5, H. 1.. K . M. Seon', R. C. OLMsn :AD, T. SZA IW, yses of plant DNA sequences. Amer. Naturalist 130: R. K . JANSEN & 1. D. P ALMER . 1993. Interfamilial 574-5100. relationships of the Asteraceae: Insights fr om rbe l RODM AN, 1. 1991 a. A taxonomic ana lysis of glucosi. sequence variation. Ann. Missouri Bot. Card. 8 0: nolate.producing plants, Pari I : Phenetics. Syst. Bot. 742- 765. 16, 598- 618. M OR DEN, C. W ., C. F . D EL WICIlE, M . K UHSEL & J. D. ---. 1991b. A taxonomic analysis of glucosinolate. PAL MER. 1992. Gene phylogenies and the endo­ producing plants, Part 2: Cladistics. Syst. Bot. 16: symbiot ic origin of plastids. Biosystems 28: 75- 90. 6 19- 629. 1\'10 IlG AN, O. R. & D. E. SoLTIS. 1993. Phylogenetic ---, R. PRICE, K. KAROL. E. CONTI. K. SYTSMA & relationships among members of Saxifragaceae sensu J. D. PALMER . 1993. Nucleotide sequences of the la lo based all rbe l sequence data. Ann. Missouri Bot. rbe l gene indicate monophyly of mustard oil plants. G" d. 80, 63 1- 660. Ann. Missouri Bot. Gard. 80: 686- 699. NEAU:, D. B., N. C. W"U:U::H & R. W . A UAIlD. 1986. SeIl WARZW ALD ER. R. & D. L. DIl.o~E R . 1991. System. Paternal inheritance of chloroplast DNA in Douglas· atic placement of the Platanaceae in the Harnamel· fi r. Canad. 1. Forest Res . 16: 1152- 1154. idae. Ann Missouri Bot. Ca rd . 78: 962- 969. NIXON, K. C. & C. I. DAV IS. 199 1. Polymorphic taxa. SEWE LL. M.. M. W. CII ASE. Y .. L. QIU & c. R. PAnKs. missing values, and cladistic analyses. Cladistics 7: 1993. Trace paternal inheritance of chloroplast DNA 233- 241. in LiriodendrOf! and Magnolia (Magnoliaceae). Amer. OL~I sT EAn, R. G., H. 1. MI CHAE15, K. M. ScoTr & 1. D. J. Bot. 80 (in press). PALM Ell. 1992. Monophyly of the Asteridae and S ~II T H, 1. F., W. 1. KnF..5S & E. A. ZEMME n. 1993. identification of their major lineages inferred from Phylogenetic analysis of the based on DNA sequences of rbe L Ann. Missouri Bot. Card. rbel sequences. Ann. Missouri Bot. Card. 80: 620- 79, 249- 265. 630. ---, B. BREME II , K. M. Scorr & J. D. PALMER. SoLTIS, D. E., P. S. SoLTIS, M. T. CLEGG & M. DUR. 1993. A parsimony analysis of the Asteridae sensu lllN . 1990. rbe l sequence divergence and phylo. lalo based on rbe l sequences. Ann. ~' I i ssouri Bot. genetic relationships in Saxifragaceae sensu lalo. G"d. 80, 700-722. Proc. NatL Acad. U.S. A. 87: 4640- 4644. OWENS. 1. N. & S. 1. MORRIS. 199 1. Cy tologica l basis SWOFFOR D, D. L. 1991. PA UP: Phylogenetic Analysis for cytoplasmic inheritance in Pseudotsuga men;;ie­ Using Parsimony, Version 3.0r. Computer program sii. II. Fert ilization and proembryo development. distributed by the Illinois Natural History Survey, Amer. 1. Bot. 78: 15 15- 1527. Champaign, Illinois. PAL MER, 1. D.. R. K. JA NSEN. H. 1. M[( :II ,-\EIS, M. W. --- & G. 1. OES EN . 1990. Phylogeny reconstruc· CHAS E & 1. M. MAN H.-\R T. 1988. Analysis of chlo­ tion . Pp. 4 11 - 501 in D. M. Hillis & C. Moritz roplast DNA va ria tion . Ann. Missouri Bot. Gard. 75: (editors), Molecular Systema tics. Sinauer Associates, 1180- 1206. Sunde rland, Massachusetts. PAMlLO, P. & M. NEt. 1988. Relationships between SZMEI)T , A. E., T. ALDEN & 1. ·E. HAtLGIlE N. 1987. gene trees and species trees. Malec. BioI. Evol. 5: Paternal inheritance of chloroplast DNA in Larix. 568- 583. PI. Molec. BioI. 9: 59- 64. PENNY, D. , M. D. HEN DY & M. A. STEEL 1992. Prog­ TAKHTAJAN, A. 1980. Outl ine of the classification of ress wi th met hods for constructing evolutionar y trees. flowering plants (Magnoliophyta). Bot. Rev. 46: 225- Trends Ecol. Evo l. BioI. 7: 73-79. 359. 1987. Systema MagnoliophYlorum. Nauka. DE PI NNA, M. C. C. 1991. Concepts and tests of ho· Leningrad. [In Russian. ] mology in the cladistic paradigm. Cladistics 7: 367- TAY tOR . D. W. & L. J. HI CKEY . 1992. Phylogenetic 394. evidence for the herbaceous origin of angiosperms. PLATN ICK, N. l.. C. E. GRISWOLD & 1. A. CODDINGTON. PI. Syst. Evol. 180: 137- 156. 1991. On missing entries in cladistic analyses. Cla­ THORN E, R. F. 1983. Proposed new realignments in distics 7: 337- 343. the angiosperms. Nordic J. Bot. 3: 85- 117. PRICE. R. A. & 1. D. PALM ER . 1993. Phylogenetic --- 1992. An updated classification of the flow­ relat ionshi ps of the Geraniaceae and Geraniales from ering plants. Aliso 13: 365- 389. rbel sequence comparisons. Ann. Missouri Bot. Gard. WAGN ER . D. 8.. G. R. FURNIEIl. M. A. SAG I1 At·MAIlOOF, 80, 661 - 67 1. S. M. WI LLI AMS, D. B. DANe1" & R. W. ALLAII D. QIU, Y.- L., M. W. CHASE. D. H. b :... & C. R. PAIlKS. 1987. Chloroplast DNA polymor phisllls in lodgepole 1993. Molecular phylogene tics of the Magnoliidac: and jack pines and their hybrids. !'roc. Nat l. Acad. Cladistic analyses of nucleotide sequences of the plas. U.s.A. 84, 2097- 2100. tid gene rbc L Ann. Missouri Bot. Card. 80: 587- WE NDEL, 1. F. & V. A. ALU ERT. 1992. Phylogenetics 606. of the cotton genus (Goss)'pillm): Character·state RAUIJESON, L A. & R. K. JA NSEN . 199 2. Chloroplast weighted parsimony analysis of chloroplast-DNA reo DNA evidence on the ancient e\'olu tionary split in st rict ion site data and its systematic and biogeograph. vascular land plants. Science 255: 1697 - 1699. ic implications. Syst. Bot. 17: 115- 143. RETflG, 1. H., H. D. W1150N & J. M. MANHAIlT. 1992. WH ATLEY, 1. M. 1982. Ultrastructure of plastid inher. Phylogeny of the - Gene sequence ita nce: Green algae to angiosperms. Bot. Rev. 57: data. Taxon 4 1: 20 1-209. 527- 569. R1ES EIJ ERG, L. H. & D. E. SoLTIS. 1991. Phy logenetic WtII TE. E. E. 1990. Chloroplast DNA in PillH S mOI/- 548 Annals of the Missouri Botanical Garden

fico/a. 2. Survey of within-species variabili ty and Current address for LEE: Centro de Ecologia. Universidad detection of heteroplasmic individuals. Theoret. Aprl. Naciona l Autonoma de Mexico, Apartado Postal 70·275. Gen. 79: 251 - 255. CU .. c.P. 0 451 0 . Mexico. D.F., Mex ico. Current address W IL.50N, M . A., B. GAUT & M . T. CLEGG. 1990. Chlo­ for BSG: Department of Statistics, North Ca rolina State roplast DNA evolves slowly in the palm family. Molec. University, Raleigh, North Carolina 27695. U.S. A. Cur­ BioI. Evol. 7: 303- 314. rent address fo r MR D: Office of Biotechnology. Iowa State YOUNG, D. 1. & L. W AT:"ON. 1970. T he classification IJlliversity. Ames. Iowa 500 11-3260. U.S.A. of dicotyledons: A study of the upper levels of hi­ " Department of Biology, Indiana University, 13100- erarchy. Austral. J. Bot. 18: 387- 433. mington. lndiana 47405. U.S. A. Current address for RAP: W II, C.-L. 1991. Inference of species phylogen y ill Department of . University of . Athens, relation to segregation of ancient polymorphisms. Georgia 30602. U.S.A.

Gene tic!' 127: 429- 435. Q Department of Biology, A & M University , XIANG. Q.. Y., D. E. 5o L TI ~ . O. R. Mow;"N & P. S. 5oLT r ~ . Coilege Station. Texas 77843. U.S.A . Current Address 1993. Phy logenetic relationships of COrtlll$ L. sensu for JH R: Department of Biology. Co ll ege of the Oza rks, lato and putative relatives inferred frolll r{w L se­ Point Lookout. Missouri 657 26. U.S.A. quence data. Ann. Missouri Bot. Card. 80: 723 - U> Department of Botany, University of Wisconsin, 734. Madison. Wisconsin 53706. U.S.A .

ZURAW SKt, G. & M. T. Cu :o :. 1987. Evolution of II Department of Biological Sciences. Bowling Green higher-plant chloroplast DNA·coded genes: Implica. State University. Bowling Green. Ohio 43403. U.S.A. tions for structure-function a nd phylogenetic studies. IZ Department of Botany, Smithsonian Inst itution. Ann. Rev. PI. Phys. 38: 391- 418. Washington. D.C. 20560, U.S.A. Current address fo r JFS: Department of Biology. Boise State University. Boise, Ida ho 83725. U.S. A. ,. Department of Botany. Arizona St ate University. Tempe. Arizona 85287, U.S.A.

~ Department of Biology, University of North Carol ina, I< Institutionen For Systematisk Botanik. Uppsala Chapel Hill, North Carolina 27599·3280 . U.S.A. Currcnt Universitel, Uppsala 5-751 2 \, Sweden. address for MWC: Laboratory of Molecular Systematics. ,5 Department of Botany, Universit y of Texas, Austin, Hoyal Botanic Gardens, Kew. Ri chmond. Surrey TW9 Texas 787 13. U.S.A.

3AB, Un ited Kingdom. Current add ress for KA K: De­ L<, Department of Forest Resources and Plant Biology, partment of Biology. Wake Forest University. Winston· University of Minnesota. SI. Paul. Minnesot a 55108- Salem, North Carolina 27 I 09, U.S.A. Currcnt address 81 12. U.s.A. fo r V t\t\: Institutionen For Systclllatisk Botanik. Uppsala I ' Department of Forest Science. FS L 020. Oregon Universitet, Uppsala S-751 21. Sweden. State University. Corvallis, Oregon 97331-7505. U.S.A. , Department of Botany. Washington State Universi ty, ' ~ Depar t men t of Botany. University of Tennessee, Pullman. Washington 99164-4238. U.S.A. Knoxville. Tennessee 37996-1 100, U.S.A. Current ad­ , Department of E. P. O. Biology, University of Col· dress: Rancho Santa Ana Botanic Garden, Claremont. orado. Boulder, Colorado 80309·0334. U.S.A. California 91711·3 101. U.S.A. , Department of Biological Sciences. University of \Vis· ,Q Depart ment of Bioiogy, Leba non Valley College, consin, Mi lwaukee. Wisconsi n 53201. U.S.A. Current Annville . Pennsylvania 17003·050 1. U.S.A. address fo r DH L Department of Ecology and Evolution­ ", School of Biological Sciences. University of New ary Biology. University of Connecticut. Storrs, Connect. South Wales. P.O. Box 1. Kensington. New South Wales, icu t 06269-3042. U.S.A. . " Department of Botany. Duke University. Durham. " Department of Biological Sciences, \V a yne State Un i­ North Carolina 27708·0338. U.S.A. Current address: versity. Detroit, Michigan 48202. U.S.A. Depa rtment of Integrative Biology. Uni\·crsity of Ca lifor. ~" Department of Botany. University of Toronto. To· nia. Berkeley. California 94 720. USA. ronto. Ontario M5S 3 B2. Canada. : Department of Botany and Plant Sciences. Universit y Z' Department of Biology. Boston University. Boston. of Cu lifornia, Riverside, California 92521-0124. U.s.A. Massachusetts 022 15. U.S.A. 550 Annals of the Missouri Botanical Garden

IA

Asteridae -- including Ericales, Primulales, Ebenales, Santalales, Apiales, Comales, and some Rosales

Rosidae -- including Violales, Malvales, mustard-oil families, higher Hamamelidae, and CaryophyUidae

hamame\ids ranuncuhds paleoherbs

monocots

Laurales Magnoliales Cerlltophyllaceae "' - gnelophyte§ "' Plnllceae other conlrers cycads FIGURE I . SUlIl lllo ries of the major clades identified in: (A) the cOlllbinaule component consensus tree of 500 equa ll y parsimonious trees found for 4 75 taxa using the character·state weight ing method of Albert e l al. (1993. this issue): and (8 ) the strict consensus tree of 3.900 equall y pa rsimonious trees for 499 taxa found using the Fit ch (even weights) criterion. These afC ingroup networks arranged a rbi trarily with the cycads sister to a ll other seed plants. Volume 80, Number 3 Chase et al. 551 1993 Phylogenetics of Seed Plants

IB

Asteridae -- including EricaIes, Primulales, EbenaIes, Santalales, Apiales, Cornales, and some Rosales

Rosidae -- including Violales, Malvales, mustard-oil families, and higher Hamamelidae

caryophyllids Gunncraccac hamamelids ranunculids paleoherbs [[ Magnoliales

monocots

paleoherbs I gnetophyles other conifers Pinaceae Ginkgo cycaas 552 Annals of the Missouri Botanical Garden

2A asterid I asterid II asterid III asterid IV asterid V rosid IV rosid III rosid II rosid I hamamelid II hamamelid I ranunculids paleoherbs monocots Laurales Magnoliales Ceratophyllum Gnetales Pinaceae other conifers cycads

FIGU HE 2. Summaries of the same topologies as ill Figure I . In B, Filch branch lengths are optimized from a single tree; optimiza tion on consensus trees is likely to overestimate branch lengths. Na mes of the specifiC clades identified do not conform to the composit ion of fam il ies used in most taxonomic schemes. but rather are designated with respect to the components of the major lineages (i. e .. by the subclass 118111e for the majorit y of taxa included, except for the heterogeneous ha mumclid I, which is so designated because of ils position and inclusion of PlatanDceae). Na mes of each clade correspond 10 groups shown in Figures 3- 15. Clades marked wi th a .o§" in n are those that differ significantly in position or composition from A. Volume 80, Number 3 Chase et at. 553 1993 Phylogenetics of Seed Plants

28 5 6 asterid I 4 3 asterid II 6 10 asterid III 4 5 asterid IV 6 3 asterid V§ 2 9 rosid I 3 5 5 rosid II 5 5 rosid III 9 caryophyllids§ 8 33 Gunnera§ 4 8 hamamelid II 18 5 hamamelid I 9 ranunculids 4 22 5 paleoherb II § 6 Magnoliales 1 24 monocots 59 11 Laurales 12 39 paleoherb I § 44 Ceratophyllum 18 61 Gnetales 16 other conifers 32 Pinaceae 28 23 Ginkgo§ cycads 554 Annals of the Missouri Botanical Garden

3A angIOsperms• Gnetum Gnetaceae Welwitschia Welwitschiaceae Ephedra Ephedraceae Abies --, KeteleerW ..r Pseudotsuga Larix Picea sitchensis Picea pungens Pinaceae Pinus griffithii Pinus radiata Cednts Pseudolarix Tsuga Callitris Cupressaceae Widdringtonia_ Taxodium Metasequoia Taxodiaceae Sequoiadendron Podocarpus Podocarpaceae Cycas Cycadacae --, Microcycast Zamia Chigua Bowenia

ull Macrozamiat a, I II Zamiaceae .u UI Encephalartos ut IV ... V Ceratozamia t rM IV f Oi III Stangeria ,f ..

3B 59 angIOspe• r ms 39 71 42 Gnetum G netaceae 52 61 Welwitschia Welwitschiaceae 70 Ephedra Ephedraceae 38 22 Y Sciadopitys* Taxodiaceae 27 6 Taxus* Taxaceae 28 Cephalotaxus* Cephalotaxaceae 16 36 29 24 Callitris 19 5 Widdringtonia--1 Cupressaceae 10 7 8 Metasequoia lO y 6 Sequoiadedron 12 9 Taxodiaceae Athrotaxis* 18 30 Taxodium 65 Podocarpus Podocarpaceae - 7 Abies 18 5 Keteleeria 1 4 Pseudolarix 25 7Y Tsuga 8 8 Pseudotsuga 13 4 Larix Pinaceae 20 32 Picea 7 15 12 Pinus griffithii 14 Pinus radiata 28 12 Cedrus 23 Ginkgo * Ginkgoacae 31 Cycas Cycadacae 18 Bowenia - 3 3 19 Zamia 2 5 Chigua ti

<~

F ICUIH: 4 . Gasu l portion of the overall analysis showing the positions of Ceratophyll accac (iuapcrturatc po ll en). monocots (uniaperturate pollen). eudicots (dicols with triapert urate pollen). and the three clades of "primitive" dicots (lIlonosulcate pollen). Note that. exclusive of Ceraloph)'lIuf1l, the angiosperms form two sister groups marked by the general pollen opert ure number (one versus three). (Numbers above the bronches in B are the numbers of substitu tions optim ized onto one tree randomly selec ted from the 3.900 saved in Search 11. ) Genera marked with a "t " in A were om itted frolll Search II ; genera marked with a n asterisk in B were not ava il able for Search I. The "Canella" in this figure represents the position of two sequences amplified fr om 3 total cell ular DN A template (sec Materials & Methods). Note that A is the consensus tree of Search I. whe reas 13 is a single tree with branches lIot present in the strict consensus of Search II marked by 3n arrow. Volume 80, Number 3 Chase et al. 557 1993 Phylogenetics of Seec Plants

4B ----.,... eudicots Victoria Nymphaea Nymphaeaceae • Nupluu Barclaya 8arclayaceae • -:- Brasenia ] Cabombaceae· Cabomba , Amborella Amborellaceae Schislmdra Schisandraceaer IUicium IUlciaceacf AustrobaUeya _-, Austroballeyaccaeb :,.,.. Hedyosmum· - d Chloranthus Chloranthaccac Sarcandra* ...".. ClUIellD Canellaceaeb ...... Drimys b BeUioium Winteraceae - Tasmannia· Magnolia saIki/oUa : Michelitz [j Tal.tuurao ovala ~=~=: Magnolia hypoleuca :. Manglietia Magnollaccaeb TaJauma dngaporensis Uriodendron tulipifera I...irWdendron chinensis· Degeneria Degcnerlaceae b GalbuUm;ma Hlmantandraceaeb '-:-:- Eupomalia Eupomatlaceaeb AnnOnD Asimina Annonaceacb Cananga· K",,,,,, Myrlstlcaccae b _ Cinnamomum ] Lauraceae(' Penta Hedycarya Monlmlaceae (' , Idiospermum Idlospermaceae Calyean/hus , ChimolUUllhus Calycanthaceac , Hernandio Hernandlaccac Gyrocarpus· -----=-----.,.,. monocots Asarum , SaruntIJ Aristolochlaccac -g ArislOlochia b - 12 Uzctoris· _-, Lactorldaccae ~ Saururus ::r Saururaccac d ... HouJtuynia ., Piper· d ::r Plpcraceae 44 Peperomia - L ______-= ______Cerarophyl/um Ceratophyllaccac • :: k IVI :==:a'... v ~: II 'Magnoliidae, Nyphaeales rOl 111 m bMagnoliidae, Magnoliales (' Magnoliidae, Laurales ham I pal'"" D d Magnoliidae, mOl • Magnoliidae, Aristolochiales ~.;; ~ .. -- 'Magnoliidae, Illiciales ,~ ."' ''"' ...'Y' 558 Annals of the Missouri Botanical Garden

SA commelinoids Xantho"hoea ] Xanthorrhoeaceae J KniphoFID Aloe J Haworthia Aspbodelaeeae SansevUria DracaenaceaeJ DafIfJe RuscaeaeJ NolrlUl recut'Wltll I N I J NoUIUI lindluimerill!!1J 0 Inacaceae 80wua HyaclntbacueJ Chlorophytum Anthericaceae J CliviIJ Amaryllldaceae j Iris --, Irklaceae I Anomatheca --..J CyalUls~rum TecopbllaeaceaeJ Curcullgo Hypoxis Hypoxklaceae J Neuwitdia Orchldaceae l Onci1.~m Smi SmHacaceae' Lilium ] Medeola LlIlaceael AlslroemuilJ Alstrocmeriaceael Colchicum ] BurchardiD Colchlcaceae I Chamaelirium Melanthiaceaeb Vellozia Velloziaceae. Freycindin Pandanaceael sp'haeradrnin Cyclanthaceaer Dwscorea Dloscoreaceae. Tocco Taccacea.' AkIns Melanthlaceaeb Burmanniat Burmannlaceaec Sagitlaria gram;nea Sagitlaria Ialifolia Allsmataceae" AlisnuJ POlamogelon PotamoRetonaceae" PI... Melanthlaccaeh Spa/hiphyllum ull Gymnoslochys ] Araceae· "'0utn. A5IIV ulV • fOIl IV Aranae fOIl 10 bLilianae, Melanthiales r(JI: II ., Alismatanae ~'MIh~U h~1 IIILilianae, Burmanniales ~ ,~ ~Lilianae, Dioscoreales moo~ _____ fCyclanthanae lau., ---...... -­ 'pandananae c",r "Bromelianae, Velloziales ,~ pin ILilianae, Liliales JLilianae, 'Y<•

FIGUIIE 5. The five basalmost lineages of the monocots, composed of the uroids, ali smatids. and Iilioid groups. (Numbers above the branches in B are the numbers of substitutions optimized onto one tree randomly selected from the 3,900 saved ill Search II.) Genera marked with a "t" in A were omitted fr om Search II : genera marked with an asterisk in B were not available for Search I. Note that A is the consensus tree of Search I. whereas B is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 559 1993 Phylogenetics of Seed Plants

5B Cypripedinm - .. 27 d Neuwiedia Orchldaceae

Oncidium ; " Lomandra - Dasypogonaceae i 17 .. Hemerocallis- Hcmcrocailidaceae 38 Chw rophytum Antherkaceae' 17 Xanlhorrhoea Xanthorrhoeaceaei • Kn iphofUl Aloe Asphodclaceae' 10 Haworthia 10 Scilla - .. 2. Hyaclnthaceae I 12 4 Bowiea Sansevieria Dracacnaceae i II Dantu Ruscaccac I • No Un a Nollnaccae i II Clivia AmaryllIdaceae i 21 12 Iris .. J Irldaccac d Anomalheca , .. Cyanastrum Tccophllacaccae 13 Curculigo I' 21 ] Hypoxldaceae i Hypoxis commelinoids 4 2' Dioscorea Dloscorcaceacd 27 , TaccD Taccaccilc d ~ Pandanus· Pundanaccac h • Freycinetia J " Sphaeradenia Cyclanthaccae 8 VeUozia V cllozlaccac r Sm;/ox Smllacaceac e 13 I' d Calochortus· CaJochortaceae 14 17 Liliu m d I. » 13 21 Llliaccac Medeola I' 24 Colchicum d 12 13 Burchardia Colchlcaccac J." d A Istroemeria Alstroemcriaccae J6 VeraJrum· • 13 , ChamlU!lirium Mclanthlaccac • '7 Aletris 22 I'leea Sagittaria graminea " b 3d • SagiJtaria latifoUa Alismataccac 44 24 Alisma Potamogeton Potamogctonaccacb IJ Spalhiphyllum 41 J Araccac· 7 Pistia· J' LemlUJ· Lcmnaccac • " 13 Gymnostachys JAraccac· 4' Acorus· I II UI IV V a I Aranae n [IJ bA lismatanae C Lilianae, Melanthiales dLilianae, Liliales cLilianae, Dioscoreales f Bromelianae, Velloziales gCycianthanae hpandananae I Lilianae, Asparagales 560 Annals of the Missouri Botanical Garden

6A A vena Puccinella t TriJicum Hordeumt Aegilops Poaceae l Cenchrus Pennise/urn t Neurachne OryUJ Elegia _ Restionaceae l luneus Oxychloe Juncaceae J Cyperus - Carex CyperaceaeJ Prionium - JuncaceaeJ FIaKellaria Fla~ellariaceael Typha Typhaceaeh Slegolepis Rapateaceae f Aechmeat Ananas Glomeropitcait'nia t CalopsisT BromeliaceaeA Tillandsia Puya Hechliat pallida Tradescanlia zehrina Commelinaceaer Tradescantia soconuscana Pontederia Pontederiaceaee Philydrum Philydraceaed Anigozanlhos Haemodoraceae~ Rovenaia StrelitZcia Strelitziaceaeb Phenakospermum MOTonla b Calalhea Marantaceae Hedychium Riedelin Zingiberaceaeb Zingiber Globba Costus Tapeinocheilos Costaceaeb Musa -' M usaceae" Orchidanlha Lowiaceaeb Heliconia -,Heliconiaceaeb Caryo/at Nypa Phoenixt Drymophloeus Arecaceae a astl Chamaedorea ast ][ Calamust ast III Serenoa ast IV ast V aArecanae ros ros IVII bZ 'mg. 'be ranae ros II ros I cBromelianae, Haemodorales dBromelianae, Philydrales ~==hamllham I ! raD eBromelianae, Pontederiales pal rCommelinanae, mOD lau gBromelianae, Bromeliales mag cer hBromelianae, Typhales gne, I Commelinanae, pm con iCommelinanae, Cyperales cyc FI CU RE 6. The terminal li neages of the monocots, composed of the palms. gingers, and commelinoids. (Numbers above the branches in B are the numbers of substitutions optimized onto one tree randoml y selected from the 3,900 saved in Search II .) Genera marked with a .. t" in A were omitted from Search II ; genera marked with an asterisk in B were not available fo r Search I. Note that A is the consensus tree of Search I, whereas B is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 561 1993 Phylogenetics of Seec Plants

6B ~A"'UJ Triticum AegiJops l Cenchrus Poaceae NeurachM 13 -'7- OryTJ) Bambusa* --:-;:;:--- LachllOcaulon* Eriocaulaceaef --::: "--- Ekgia Restionaceae I l -~- FlageiJoria Flagellariaccae ~-Juncus Juncaceae J Oxychlae Cyperus Cyperaceae J ,.... -- Carex 10 -- -...,.. Prianium Juncaceae J Sparganium * Sparganiaceaeh Typha Typhaceaeh 32 - Stegokpis Rapateaceae( - '-:- Tillandsia Anarws Bromcl iaceae g Puya -..:: -':"'" Commelino* -:- TradescantW pal/ida ' ...:.. TradescantW zebrina - Tradescantia soconuscan e Pontederia Pontederiaceae d Philydrum Philydraceae c Anigozanlhos Haemodoraceae ~ ROl'enala b II SlTe/it;ja Strelitziaceae Phenakospumum Moronta Marantaceae b Calathea .:.,... Hedychium ~Riedeka Zingiberaceaeb Zingiher Globba Costus Costaceaeb TapeinocheiWs 7:- Musa Musaceaeb --"'-- Orchidantha Lowiaccaeb __-,1;;;1 __ Heliconia HeI ICODIaceae" b Cluzmaedorea ",;,. ~~ Drymophloeus A recaceae a Nypa ::~h -"":'-- Serenoa ="lIIIuttV ~ ul V • Arecanae :::b b Zingiberanae ,~m 'Bromelianae, Haemodorales '" dBromelianae, Philydrales 'Bromelianae, Pontederiales :;-_ r Commelinanae, Commelinales "Bromelianae, Bromeliales hBromelianae, Typhales I Commelinanae, Poales JCommelinanae, Cyperales 562 Annals of the Missouri Botanical Garden

7A higher eudicots :r Trochodendron ' ~ 3 Tetracentron • Tetracentraceae'7-..-4 SabUz Sabiaceae c = :r LambertUz Proteaceae • ~ 3 Nelumbo Nelumbonaceaet • Platanus Platanaceaeb AkebUz Lardizabalaceaec MahonUz Berberidaceaec -- Caulophyllum ., - Ranunculus ~ == - Xanthorhiza Ranunculaceae C = '-- Caltha n== c =_. --- Cocculus Menispermaceae -Q. CIl '----- Euptelea Eupteleaceae b - Dicentra ---Fumariaceae" - Papaver Papaveraceae" __ __ '--.. , Sanguinar=Uz,---,- .,"11M III .. 'v ;::ri • Magnoliidae, Papaverales

~=~ :::' ~ C Magnoliidae, ~~ d Magnoliidae, Nymphaeales :::- • Rosidae, ~ , Hamamelidae, Trochodendrales

FIG1JHE 7. The basalmost lineages of the eudicots. corllposed of Ranunculales- Papavera les, Trochodendrllics. and a he terogeneous lineage (plus GtlflflCra in 0). (Numbers above Ihe branches ill B arc the numbers of substitut ions optimized onlo one tree randomly selected from the 3,900 saved in Search 11. ) Genera marked with an asterisk in B were not available for Search I. Note that A is Ihe consensus tree of Search I, whereas B is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 563 1993 Phylogenetics of Seed Plants

7B , , higher eudicots h JJ Gunnera Gunneraceae • 20 " Trochodendron Trochodendraceae& ;- II Tetracentron Tetracentraceae & •:3 • 3. Pachysandra'" Buxaceae' ---- .0 41 Sabia Sabiaceae C ::r , 37 Lambertia Proteaceae e ~ :3 ••, Nelumbo Nelumbonaceaed • 13 Platanus Platanaceal - c __3~2 __ Akebia Lardizabalaceae .8 27 Mahonia c 8 Berberidaceae 2' Caulophyllum ., ~ .0 l! Ranunculus ::s c c • 2. Xanthorhiza Ranunculaceae ::s 2. Caltha c/') _---=2!.,. --Cocculus Menispermaceaec -_. b Q. • -_....!•.::.... --Euptelea Eupteleaceae '" .2 l7 Dicentra Fumariaceat! II 39 Papaver a • Sanguinaria Papaveraceae __ _n-, _m aMagnoliidae, Papaverales _IV _v, bHamamelidae, Hamamelidales , CMagnoliidae, Ranunculales .. ~n}~ .... 1 ~ dMagnoliidae, Nymphaeales n. -.,olD eRosidae Proteales ., , Rosidae, Euphorbiales ...- gHamamelidae, Trochodendrales .-• hRosidae, Haloragales 564 Annals of the Missouri Botanical Garden

8A Boykinia Saxifragaceae C Astilbet Heuchera

Tetracarpaea Grossulariaceae C Myriophyllum Haloragaceae d

Saxifragaceae C Penthorum --, Sedum Dudleya Crassulaceae C Kalanchoe Crassula Ribes Grossulariaceae C Itea Rhodoleia Hamamelidaceae b Hamamelis Cercidiphyllum Cercidiphyllaceae b ... 1 a ... u Daphniphyllum Daphniphyllaceae ut 111 ... IV ulV ~IV ros III ...... ~II ~I rosid IV ".m II pml n" aHamamelidae, Daphniphyllales p_1 bHamamelidae, Hamamelidales -"- m_, .~ :Rosidae, Rosales pi""'. Rosidae, Haloragales ' M

88 9 Pterostemon* • Grossulariaceae' 21 Itea 10 7 Boykinia 9 , " Heuchera Saxifragaceae ' I5Saxl ifi raga 21 11 Tetracarpaea Grossulariaceae' 4 • 34 Myriophyllum Haloragaceae e

IS 21 Penthorum Saxifragaceae '

I Paeonia Paeoniaceae d I. tI 15 Sedum I' 10 Dudleya Crassulaceae ' 9 Kalanchoe 44 Crassula 22 5 Ribes Grossulariaceae' H Rhodoleia Hamamelidaceae • CercidiphyIIaceae • DaphniphyUaceae b 2 Hamamelidaceae •

... rosid III '""ham I[ llaml a Hamamelidae, Hamamelidales ""palll ."' "Hamamelidae, DaphniphyIIales ,,.""" 'Rosidae, Rosales '",M dDiIIeniidae, DiIIeniales '""pin eRosidae, Haloragales ,'"<,< 566 Annals of the Missouri Botanical Garden

9A Dillenia Dilleniaceaec Dianthus ----,Caryophyllaceae e Atriplex Chenopodiaceae e Spinacia ---I Amaranthus Amaranthaceae e Basella Basellaceae e Portulaca Portulacaceae e Alluaudia Didiereaceae e Mollugo Molluginaceae Phytolacca Phytolaccaceae e Mirabilis Nyctaginaceaee I Rivina Phytolaccaceaee Trianthema Aizoaceaec

Stegnosperma Phytolaccaceae C Nepenthes N epenthaceae b Plumbago Plumbaginaceaed Rheum PolygonaceaeC Droseraceaeb Vitis Vitaceae ."_I II _till _ , IV ., V rosid III roo. IV r o. lU~I-__ 1.-:: :' aRosidae, Rhamnales p.''" ,moo.. CCaryophyllidae, Polygonales m., dCaryophyllidae, Plumbaginales '" ."p" eCaryophyllidae, Caryophyllales '"'Y ' CDiIleniidae,

F1G UHt: 9. The li neage thai includes the Caryophyllidae. (Numbers above the brullches in B arc the numbers of substitutions optimized onto one tree randomly selecled from the 3,900 saved in Search 11. ) Genera marked with an asterisk in B were not a vailable for Search I. Note that the positions of this dade and thai of rosid IV (Fig. SA, B) differ significantly in the results of Searches I and II (see Figs. I , 2), Note that A is the conseIlSus tree of Search I. whereas B is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 567 1993 Phylogenetics of Seed Plants

98 r-- ~Dianthus _ Caryophyllaceae e -..!.!....- Atriplex

maranthus Amaranthaceae e - Basella Basellaceae e Portulaca Portulacaceae e Alluaudia Didiereaceaee - Mollugo Molluginaceae e '- Phytolacca Phytofaccaceaee Mlrabilis Nyctaginaceae e Rivina PhytoTaccaceaee - Trianthema Aizoaceae e --~-- Stegnosperma Phytolaccaceae e Plumbago Plumbaginaceae d Rheum Polygonaceae C b L- - Nepenthes _ epenthaceae Drosera filiformis Drosera spathulata* b - Drosera oetiolaris* - :""-Dionaea'* - Phoradendron Viscaceae" Schoepfia Olacaceae" ... :.... Osyris Santalaceae" ..._I ... IV ... v ~I caryophyllids ~II ::.u:~;...._ • C:'u b Rosidae, Santalales ...N, 1 Dilleniidae, Nepenthales ,.111 --. C Caryophyllidae, Polygonales ..••II «, d Caryophyllidae, Plmllbaginales ...roo e pi. Caryophyllidae, Caryophyllales p '1' 568 Annals of the Missouri Botanical Garden

lOA ------Gossypium MalvaceaeJ Leitneriaceae' - Ailanthus Simaroubaceaeh -- Poncirus Rutaceaeh Bursera Burseraceaeh _ - Acer Aceraceaeh -- Cupaniopsis Sapindaceaeh -- Schinus Anacardiaceaeh ------Tropaeolum Tropaeolaceaea Brassica Brassicaceae' Capparis Capparaceae' Reseda Resedaceae' Tovaria Tovariaceae' Batis _ Bataceaeg F/oerkea Limnanthaceae" Carica Caricaceal Moringa Moringaceae~ Qualea Vochysiaceae rlarkia Oenothera Epilobium Lopezia ' Hauya Fuchsiat Ludwigia ----' Lythrum Lythraceaec c Quisqualis Combretaceae b rrossosoma Crossosomataceae Pelargonium Geranium Geraniaceae" Monsonia_ "tI b ... u Greyia Greyiaceae aslln Francoa Saxifragaceaeb utlV a ... V ~-- Viviania Geraniaceae rOlIV rOlln '''' II ". I rosid II

..:~~==:h_n ~:' I "Rosidae, Geraniales f Dilleniidae, Violales

polmon "Rosidae, Rosales gDilleniidae, Batales la. m.. CRosidae, Myrtales h Rosidae, Sapindales on • an. dRosidae, Polygalales I Hamamelidae, Leitneriales pin ,on eDilleniidae, Capparales jDilleniidae, Malvales 'Y' FICU HE 10. One of the two "higher" rosie! lineages. (Numbers above the branches in B are the numbers of substi tutions optimized onto one tree randomly selected from the 3,900 saved in Search 11. ) Genera marked wi th a .. t .. in A were omitted from Search II ; genera marked with an asterisk in B were not available fo r Search I. Note that A is the consensus tree of Search I, whereas n is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 569 1993 Phylogenetics of Seed Plants

• lOB Leitneria Leitneriaceae ' Ailanthus C ~"- Poncirus Rutaceae C - .1-- Bursera Burseraceae C Acer Aceraceae C Cupaniopsis Sapindaceae C -Schinus Anacardiaceae C ----- Bombax* Bombacaceae h :"'-Tilia* Tiliaceae h ....:.~ Theobroma* Sterculiaceae h ~ Gossypium Malvaceae h Thespesia* Shorea zeylanica* Dipterocarpaceae g Shorea slipularis* Moringa Moringaceae d 14 Carica Caricaceae r Stanleya* - Brassicaceae d Brassica ~Cleome* -Capparis ~-Reseda ---Tovaria Koeberlinia* Balis Limnanthes Floerkea ----"'''----Setchellanthus* Bretschneidera* Akania* .:.... Tropaeolum Tropaeolaceae a -:-::- Crossosoma _ Crossosomataceae b - Pelargonium Geranium Geraniaceae a ..II Monsonia utll ut III - Hypseocharis* Oxalidaceae a ua IV ..tV rOil rosid II rOI II rOI [II ,.,,u. a Rosidae, Geraniales ham II bRosidae, Rosales ham I no C Rosidae, Sapindales pal n m.. dDilleniidae, Capparales m •• 'Dilleniidae, Batales 'au pall , Dilleniidae, Violales ,.,m ,.. gDilleniid"e, Theales pin bDilleniidae, Malvales ".'Y' 'Hamamelidae, Leitneriales 570 Annals of the Missouri Botanical Garden

llA Celtis Ulmaceae I Morus MoraceaeJ Rhamnus Rhamnaceae n Krameria Kramerlaeeae C Guaiacum ZygophyUaceae m Spiraea Photinia b Prunus Rosaceae Geum Myrica Myrlca~e o Fagus Fagaceae Casuarina Casuarin~ceae II. Chrysolepis Fagaceae Datisoo r Octomeles J DB tlscaceae Trigonia C licania --, b Chrysobalanu~ C hrysobalana""ae Erylhrorylum Erythroxyla..,.,.e h Viola Vlolaceae r Ochna Ochnaceae I Drypeus Euphorblaceae" Humiria Humlriaceaeb Euphorbia Euphorbla""ae" Passiflora Pas.slnoraceaer AcriJocarpus Thryallis Bunchosia Malplghlaceae' Mascagnio. Dicella Byrsonima Euon)mus --'Celastraceae It BrexlQ Grossulariaceae b Upuropetalon Saxlfragaceaeh Kirengeshomaf Hydrangeaceaeh Parnassia _-, Saxlrragaceae b Pisum Fabaceae d - =~=== AlbizjaMedicago ------Bauhinia = Securidaca Polygala Polygalaceae ' Cephawlus Cepbalota..,.,.e b Plntytluca Tremandraceae c: Bauera b CeratopetalumJ Eucryphia Eucryphlaceaeb OxallS OxaUdaceaea ...::h m alit IV rosid I ... V • :: fiI b Rosldae, Geranlales :: I ~ Rosldae, Rosales ; DilIenlidae, Theales ~==~ ~:: II C Rosidae, PolygalaJes II. HamameUdae, Fagales ;!i ~ Rosidae, Fabales I HamameUdae, Urticales ~n Rosldae, Hamamelidae, Casuarinales

n ~r: g Rosldae, Euphorbiales Rosidae, Rhamnales q~ hRosidae Llnales °HamameUdae, Myricales

f iGURE II. The second "higher" rosid lineage. (Numbers above the branches in Band C, see foldout. a rc the numbers of substitutions op timized onlO one tree randomly selected from the 3,900 saved in Search II.) Genera marked with a "t" in A were om itted from Search II : genera marked with an asterisk in Band C were not available for Search I. The taxon labeled " Kirengos!Jomu" was, subsequent to Search I, discovered to be misiden tified and was removed frolll Search II : its identity is unknown. Note that A is the consensus tree of Search L whereas Band C are a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 571 1993 Phylogenetics of Seed Plants

lIB lIC -----other rosid I Malpighiaceae I ::-- Cephalotus Cephalotaceae b -:':- Platy theca Tremandraceaed Bauera Ie ' b Euphorbiaceae D CeratopetalumJ unomaceae PassiOonliceae e Ochnaccae P ~ Eucryphia Eucryphiaceae b Humiriaceae 0 Dxalis TrigoniD Tl'igoniaceae Averrhoa' Oxalidaceae· Ucania Chry.obalanus ] Chrysobalanaceae I Denothera Reinwardtia* LinaceaeO Clarkia Viola Violaceaee - Epilobium Erytlu-oxylum Erythro~laceae 0 '- Circaea Onagraceae C Dryp"" Eur,horbaaceae D Euon}mus Ce astraceae m -- Lopezia BrexlII Grossulariaceae I -- Hauya LepuropdaJQn ] Saxifragaceae l '--""""':::'" Ludwigia Par1UlSsia Lythrum Lythraceaec SorgenlodoX4* -, Sargentodoxaceae r Pisum Trapa' Trapaceae C Medicago '- Punica' Punicaceae C Albizzio Fabaceae I ----.! '---- Heteropyxis' < Bauhinm Securidaca '- Qualea _ Vochysiaceaed Polygalaceae I Mouriri' Polygala c Humulus· - Cannabaceaed Dsbeckia' Melastomataceae Celtis d Quisqualis I... Trema' JUlmaceae Terminalis' _ Combretaceaec Boehnurw* Urticaceaed Ficus· Greyia Greyiaceae b JMoraccac d MOTUS Francoa Rhamnus R ham naccae J Viviania Krameria Kramcriaceac I Wendtia' Geraniaceae B GlUliacum -, Zygophyllac.. . h Photinio ------rosid II Prunus Rosaceae I Spiraea uUlll I lUI III Geum ..I IV Coriaria* -' Coriariaceae r a l V OclolfUks "'rIM. III rosid I (in part) Tet:rameks· , .. m Datiscaceae ~ VatiseD cannahina* '" DameD glomeraID fi::' D h~1 a Rosidae, Geraniales Lulla' Cucurhila* Curcurbitaceae ~ m.,~:fu bRosidae, Rosales Cucumis* moo --- Begonia- ~o n iacea \~ ,.. c Rosidae, Myrtales Myrica Myricaceac m Bdula· Betulaceae· ... . < <00 dRosidae, Polygalales Casuarina C asuannaceae pin Carya* -, J uglandaceae b ....y< Fagus Chry.oiepis Trigonobalanus* Fagaceae · halansae* Nothofagus dombeyi_'-,

rosid I (in part) U • Hamamelidae, Fagales ~ Rosidae, Polygaiaies I b Hamamelidae, Juglandales Rosidae, Rhamnalcs t Hamamelidae, Casuarinales ~ Ha~m elidae, Myricales d Hamamelidae, Urticales Rosidae, Fabales Ii OiIJeniidae, Vioiaies m Rosidae, Celastrales f Magnoliidae, Ranunculales " Rosidae, Euphorbiales I Rosidae, Rosales 0 Rosidae Linales h Rosidae, Sapindales POiJIeniidae, Theales 572 Annals of the Missouri Botanical Garden

12A .....------higher asterids Comus kousa Comus canadensis --Comus walteri Cornaceae d Comus mast Comus --- Alangiaceae d Dec umaria Carpenteria '--- Deutzia --Philadelphus '---- Hydrangea --Nyssa Nyssaceae d r- Camptotheca - '-----I Diplopanax Araliaceae e L-__ Davidia Nyssaceae d ,....---- Gunnera Gunneraceae c ,---- Paeonia Paeoniaceae b Phoradendron Viscaceae" Schoepjia Olacaceae a _'-" - Osyris Santalaceaea _,m_,n :=::: : ~ }..... -- r: r .. TV ~ ID a Rosidae, Santalales ~ U ~ I b Dilleniidae, Dilleniales == ....0_1 u ~ ..n •, ~ Rosidae, Haloragales ••• .~.. e Rosidae, Corn ales ,"... r Rosidae, Apiales .no" Rosidae, Rosales 'r<

FIGURE 12. The two hasalmost lineages of the general asterid clade. (Numbers above the branches in B are the numbers of substitutions optimized onto one tree randomly selected from Ihe 3,900 saved in Search II.) Note the different composition of aSlerid V in Searches I and II . Species marked with a ·'t '· in A were omitted from Search II . Note that A is the consensus tree of Search I, ..... hereas B is a single tree with branches not present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 573 1993 Phylogenetics of Seed Plants

12B ----.:-- higher asterids kousa Cornus canadensis Cornaceae d Cornus walteri Cornus florida -"-- Alangium ---, Alangiaceaed Nyssa --' Nyssaceae d ~Davidia 7 Diplopanax Araliaceae" 10 h d -:..:!-- Camptot eca ~ Nyssaceae 2 Decumaria Hydrangea Carpenteria Hydrangeaceae c Philadelphus ~ -Deutzia --17J:J Dillenia Dilleniaceae b , I ~ Vitis Vitaceae a ....""1 ....u" tll ~ =u"u}~.... 'IV ~ uOV ~I ~n -< roo III '" a Rosidae, Rhamnales

h~1 m b Dilleniidae, Dilleniales pa.lll mot moo,,, : Rosidae, Rosales '",M e Rosidae, Cornales '" Rosidae, Apiales ".~In '" 574 Annals of the Missouri Botanical Garden

13A - Clavija TheoPhrasta~eaej Ana$allis j ArdlSia Myrsinaceae -- Diospyros Ebenaceae b Diapensia Diapensiaceae I Polemonium Polemoniaceae b Symplocus Symplocaceae b ______Impatiens Balsaminaceae' Camellia Theaceae l Callunat Erica Be/aria Ericaceae' Elliottiat _ _ Leiophyllum t Ceratiala Empetraceae' Rhododendro n _-=== Cassiope Chamaedaphn e Zenobiat Ericaceae c: Daboecia Leucothoet Gaultheria Vaccinium Cyathodest Pentachondra Leucopogon t Epacridaceae' Dracol!hylllum t Epacns Arbutus Ericaceae' - Arctostaphylost L..... ______Pyrola Pyrolaceae' Enkianthus Ericaceae' Actinidia Actinidiaceae I Sarracenia Sarraceniaceae' ______= Roridula Byblidaceaed Cyrilla CyriUaceae' ------....L.... Styrax Styracaceae b ~ Clethra Clethraceae' r- Manilkara I b .... Chrysophl llum-.J Sapotaceae L....._ FouqUlena Fouquieriaceaea

utI ..to asterid III .RIU~ ... IV .R V • Dilleniidae, Viol ales ro-IV b rmlU Dilleniidae, Ebenales rmll ,wI 'Dilleniidae, Ericales d Rosidae, Rosales ~~h.muI ~ ".hom , Dilleniidae, Nepenthales p.1 ...- I Dilleniidae, Theales m •• 'Rosidae, Geraniales m h Asteridae, Solanaies 8!,~ p'. I Dilleniidae, Diapensiales ,~ 'Y' j Dilleniidae, Primulales FIGURE 13. The immediate sister lineage to the clade composed of traditional nSlcrids. (Numbers above the branches nre the numbers of substitutions optimized on the general semi·strict cOilsenslis tree in A and one tree randomly selected from the 3,900 saved in B.) Genera marked with a .. t " in A were omilled from Search II ; genera lIIarked with 311 asterisk in 13 were not avail able for Search I. Note that A is the consensus tree of Search I. whereas B is a !'i ingle tree wit h branches not present in the st rict consensus of Search II ma r k(.od by an arrow. Volume 80, Number 3 Chase et al. 575 1993 Phylogenetics of Seed Plants

I3B 31 Clavija Theophrastaceae j Anagallis Prim ulaceae j Ardisia Myrsinaceae j Styrax Styracaceae ' Clethra Clethraceae f Symplocus Symplocaceae' 18 Diospyros Ebenaceae' --"''':- -- Cassiope f Be/aria Ericaceae Ceratiola _ Empetraceae f 3 Erica Rhododendron Chamaedaphne Ericaceaef Gaultheria 12 Daboecia Vaccinium Pentachondra 8 Epacris Epacridaceae f Pyrola - Pyrolaceaef 3 Arbutus Ericaceae f --16""""'1""1 Enkianthus Sarracenia Sarraceniaceae I ~ Heliamphora*_ - Roridula BybJidaceaeh -----' ;",.---"""'" Actinidia Actinidiaceaeg Camellia Theaceae g Cyrilla CyriJIaceae f 16 Manilkara .:.. Chrysophyllum,,-- Sapotaceae e --" Impatiens Balsaminaceaed 4 Diapensia Diapensiaceae C Polemonium Polemoniaceae b -....::...;.- Fouquieria Fouquieriaceae a

1 a"ast II asterid III astIII ast IV 3Dilleniidae, Violales V a", .. I 1"05 II bAsteridae, Solanales 1"05111 ~, cDilleniidae, Diapensiales ~~1I ham I dRosidae, Geraniales ,an pal II eDilleniidae, Ebenales mag mOD lau fDilleniidae, Ericales pal I 'llilleniidae, Theales '"g"' COD pm• hRosidae, Rosales gm cyc iDilleniidae, Nepenthales JDilleniidae, Primulales 576 Annals of the Missouri Botanical Garden

14A Symphoricarpos Car.rifoliaceae I Valeriana Va erianaceae I Dipsacus Dipsacaceael Sambucus Caprifoliaceael Adoxa Adoxaceael - Viburnum Caprifoliaceael --;== Boopis Calyceraceaeh Lobelia Campanula ~CampanulaCeaeg -- Scaevola _.,Goodeniaceae g Stokesiat _ _ Piptocaryha Vernonia -;::= Gerbera _ _ , T~agoppgon Ochonum lac/ucat GaUlniat Echinopst Carthamnus Dimorphotheca -;:== Felicia Achillea Asteraceae r Chrysanthemumt Blenn9spermum~ SenecIO Tagetes Flaveriat Chromolaena Coreopsist Eupatorium Helianthust Barnadesia t Dasyphyllum ----i ------Villilrsia e Corokia Cornaceaec Hedera d Aralia --r----1Arahaceae . Pillosporum Pittosporaceae b -- Coriandrum J!pium d Loniumt Apiaceae ______~=~== GriseliniaSanicula Cornaceaec Berzelia Bruniaceae b ------Escallonia Grossulariaceaeb Helwingia Cornaceaec J1ex vomitoria t .•. • .... J1ex crenalll Aqullohaceae .seD ~ ___ .slID •• 1 IV •• 1 V asterid II 'MIV roslD • Rosidae, Celastrales 'MU 'M' b Rosidae, Rosales ==h~D ~ Rosidae, Cornales ~ ham'". p" Rosidae, Apiales mnn ,.. t Asteridae, Solanales m •• I Asteridae, Asterales '"8!'1~ gAsteridae, Campanulales ,no"'. hAsteridae, Calycerales 'Y' 'Asteridae, Dipsacales Volume 80, Number 3 Chase et al. 577 1993 Phylogenetics of Seed Plants

148 i

, Caprifoliaceae' , Valerianaceae' _---, Di psacaceae i Araliaceaeh

Apiaceaeh Pittosporaceae b Cornaceaec Berzelia Bruniaceaeb Boopis Calyceraceae g __ Goodeniaceaed Dimorphotheca Felicia Asteraceae f

3 _---,CornaceaeC Menyanthaceaee --4 Campanulaceaed --'Grossulariaceaeb Cornaceaec Grossulariaceaeb Aquifoliaceaea till uln ast III asterid II ast IV til V a ~I Rosidae, Celastrales ~n ~rn b Rosidae, Rosales ''''guo ham II C Rosidae, Cornales ham I ~ d Asteridae, Campanulales pain mag e Asteridae, Solan ales moo lau f Asteridae, Asterales pall ''''go. ! Asteridae, Calycerales ron pin , Rosidae, Apiales g;D 'Y' I Asteridae, Dipsacales FIGURE 14. One of the two clades of traditional Asteridae. (Numbers above the branche!> in B are the !lumbers of substi tutions optimized onto one tree randomly selected from the 3,900 saved in Search [I. ) Genera marked with a "t" in A were omi tted from Search II. Note that A is the consensus tree of Search I, whereas B is a single tree with branches not present in the strict consensus of Search II marked by an a rrow. 578 Annals of the Missouri Botanical Garden

lSA NicotUJna ycop~rsicon Solanaceaec PetunUl Convolvulus Convulvulaceae I; l.p..omf!e~ Monam.a = GrossuJariaceae I Borago _ 80raginaceae & Heliotropium H.)'drophyUum _ Hydrophyllaceae C EriodICtyon Callitric1ae Callitrichaceae h DiKitalis Atilirrhinum ~Schrophulariaceae r Nelsonia N eL~oniaceae r Jus/icwamericQIUJ Jus/icia odora _ __ Jlypoestu r Rutty. ApheliJndro Acanthus - Thunborgia Thunbergiaceaef _ __ Phaulopsist Ruellia Acanthaceae r narkria Lepid.gathis ---;:===: VerbenaSesamum Verbenaceae& Teucrium Lamiaceae& Clerodendrum Verbenaceae& Proslanlhera Sculellaria Salvia Lamiaceae& Glechomat PhysostegUl Pogpstemon 1 _~===~~== CalJiHarpagop'hytumcarpa VerbenaceaeJ'edaliaceae ( Proboscidea _oJ Budd/tja R!1ddlejaceae~ Catalpa Bignom8ceae ------Strepwcapus Gesneriaceaer ~~:%~= Lentibulariaceae r Jasminum Oleaceae f LiguslTum --I Genliana Gentianaceae~ Exacum ---' Apocynum Apocynaceae~ Asclepias Asclepiadaceae ~ !ljJigelia Gelsem;um ::::J Loganiaceae~ Pentos Rubiaceaed Hydrophyllaceaec ______GanyaAucuha Cornaceaeb Garryaceaeh - Eucommia Eucommiaceae· ::tb-- asterid I :: vlU a Hamamelidae, Eucommiales ul b :::: IX Rosidae, Cornales ~~ ,I C Asteridae, Solan ales ~===t:: P d Asteridae, Rubiales ..: ~ e Asteridae, Gentianales

~~& gAsteridae, Lamiales rE ~Asteridae, Callitrichales ~yc I Rosidae, Rosales

FIGU RE IS. The second clade of traditional Asteridae. (NumOers above the branches in B are the numbers of substitutions optimized onto one tree randomly selected from the 3,900 saved in Search II.) Genera marked wilh a ""t" in A were omiued from Search II ; genera marked with an aslerisk in B were nol available for Search I. Note Ihat A is the consensus tree of Search I. whereas B is a single tree with branches nol present in the strict consensus of Search II marked by an arrow. Volume 80, Number 3 Chase et al. 579 1993 Phylogenetics of Seed Plants

15B • • • Solanaceae I

• Convulvulaceae I ---.J _-, Grossulariaceae e Boraginaceae g ----I Hydrophyllaceae i ~ Callitrichaceae h Scrophulariaceae r -

7 Acanthaceae f

Thunbergiac~ae f Nelsoniaceae ~Lentibulariaceae f _ Gesneriaceae f

g

Verbenaceae g

f

6 lejaceae f Bi~noniaceae r Byblidaceae e ~Oleaceaef Saxifragaceae e

~ Gentianaceae C Rubiaceae d Apocynaceae C Asclepiadaceae C

~ Loganiaceae C Cornaceae b Garryaceae b Eucommiaceae a asterid I aHamamelidae, Eucommiales b Rosidae, Cornales c Asteridae, Gentianales dAsteridae, Rubiales e Rosidae, Rosales r Asteridae, Schrophulariales gAsteridae, Lamiales hAsteridae, Callitrichales i Asteridae, Solan ales 54 - Nicotiilna Eucommia II Aucuba Garrya arryaceae d5 .....: ;;- Rhododendron Ericaceae Hyt!rangea --, Hydrangeaceae Ph.ladelphus----.J Paeonia Paeoniaceae t Ribes Grossulariaceae :f Heuchera S 'f Saxifraga ax. ragaceae ...,.... CercidiJJhyllum Altingza LiljuiiJambar formosana - Liguidambar orientalis B ---'--- Rliodoleia , Daphniphyllum .:.:.. HamamelIS ....,..- Sqrgentodoxa I>~,a-rg-entodoxaceae PIS u m -----, F ..:.... Medicago --.-l abaceae .;;.... PolygalQ Mynca 4 Betuln Casuarina Carya Fagus Chrysolepis c , Trigono6alnnus Nothofagus • Nothofagus - Coriaria " Datisca 14 Luffa • Cucurbita =:J Cucurbitaceae 7 -"- Begonia Begoniaceae Te(rameles __...:D=al:..:i,scaceae Celtis Humulus ~ Trema Boehmeria D II Ficus 12 Morus --~ - Rhamnus .....: Geum • ---, Prunus Rosaceae spiraea Dicelln ." Byrsonima _----'._M_a_I'--pig~aCeae Leilneria -----ii-" _____ Bursera Burseraceae - ....:... "'ii- Acer Aceraceae 17 Schinus A nacardiaceae ------..:.:.------."... Gunnera Gunneraceae Trochodendr'7ion;;---'== Tetracentron:__ _ F dl Pachysandra uxaceae Sabia Sabiaceae Lambertiil Proteaceae ------:'------Nelumbo Nelumbonaceae , Plntanus dJ Akebia :==::;:~~IG Lardizabalaceae Mahonia Caulophyllum.,,-~ Berberidaceae Ranunculus - Xanthorhiza Ranunculaceae Caltha Cocculus Menispermaceae Euptelea I H Dicentra Fumariaceae Papaver -----, _ Sanguinaria --.-l Papaveraceae

flGUH£ 16. An example of lIsing the general analysis to focus a narrower study of internal support for a polyphyletic HUlIlamelidae. Numbers above the horizolltallincs indicate the number of substitutions optimized (ACCT· RA N) onto aile of the lIlost par!;imoniou!; tr~e!; found using character-state weighting (i.e., these are Fitch steps, equal weighting). The numbers below the horizontal lines (preceded with a "d") are the number of steps less parsimonious a t which a branch becomes a polytomy wi th the branch interior to it. Branches that are not present in the strict consensus of most -parsimonious Fit ch trees are indicated by "dO, " signifying that they " deca y" at Jna xi mum parsimony. Groups of taxa oft en considered to be members of Hamamclidae are bracketed and lettered to designate putatively independent lincages (for further explanation see text).