American Journal of Botany 96(1): 22–66. 2009. R ECONSTRUCTING THE ANCESTRAL ANGIOSPERM FLOWER AND 1 ITS INITIAL SPECIALIZATIONS Peter K. Endress 2 ,4 and James A. Doyle 3 2 Institute of Systematic Botany, University of Zurich, Zollikerstrasse 107, 8008 Zurich, Switzerland; and 3 Department of Evolution and Ecology, University of California, Davis, California 95616 USA Increasingly robust understanding of angiosperm phylogeny allows more secure reconstruction of the fl ower in the most recent common ancestor of extant angiosperms and its early evolution. The surprising emergence of several extant and fossil taxa with simple fl owers near the base of the angiosperms — Chloranthaceae, C eratophyllum , Hydatellaceae, and the Early Cretaceous fossil Archaefructus (the last three are water plants) — has brought a new twist to this problem. We evaluate early fl oral evolution in an- giosperms by parsimony optimization of morphological characters on phylogenetic trees derived from morphological and molecu- lar data. Our analyses imply that C eratophyllum may be related to Chloranthaceae, and A rchaefructus to either Hydatellaceae or Ceratophyllum . Inferred ancestral features include more than two whorls (or series) of tepals and stamens, stamens with protrud- ing adaxial or lateral pollen sacs, several free, ascidiate carpels closed by secretion, extended stigma, extragynoecial compitum, and one or several ventral pendent ovule(s). The ancestral state in other characters is equivocal: e.g., bisexual vs. unisexual fl owers, whorled vs. spiral fl oral phyllotaxis, presence vs. absence of tepal differentiation, anatropous vs. orthotropous ovules. Our results indicate that the simple fl owers of the newly recognized basal groups are reduced rather than primitively simple. Key words: ancestral fl owers; angiosperm phylogeny; ANITA grade; A rchaefructus ; basal angiosperms; C eratophyllum ; Chloranthaceae; fl ower evolution; Hydatellaceae; water plants. The question of the structure and biology of the ancestral lecular results, have linked glossopterids, P entoxylon , Bennet- angiosperms, and especially their fl owers, is an enduring riddle. titales, and Caytonia , with or without Gnetales, with angiosperms Although we are continually gaining new insights from new ( Bateman et al., 2006 ; Doyle, 2006 ; Friis et al., 2007 ; Frohlich fossils and new studies on phylogeny, morphology, and devel- and Chase, 2007 ), but there is no general agreement that any of opmental genetics in extant plants, we are still far from a fi nal these taxa are related to angiosperms. The question of still answer. There are gaps at different levels. First is the uncer- closer angiosperm stem relatives is still a void because there are tainty concerning which other seed plants are the closest rela- no fossils that undisputedly represent this part of the tree. tives of angiosperms, particularly extinct groups because most Fortunately, there has been much more progress in recon- molecular analyses indicate that no living group of gymno- struction of the fi rst crown group angiosperms. Recent work on sperms is any closer to angiosperms than any other. Second, early fossil angiosperms (reviewed by Doyle, 2001, and Friis even if known fossils can be recognized as angiosperm stem et al., 2006 ) and on extant “ A NITA grade ” angiosperms relatives, all such groups are morphologically well removed ( Endress, 2001 , 2008a) has provided new insights. Problems at from angiosperms, so there is still a major gap that can only be this level have become easier to tackle thanks to analyses of liv- fi lled by the discovery of closer stem relatives. Third is the ing angiosperms, particularly using molecular data, which have problem of the original morphology and early evolutionary dif- clarifi ed relationships within the crown group with a degree of ferentiation of crown group angiosperms. precision and statistical confi dence barely imaginable two de- Identifi cation of seed plant relatives of the angiosperms has cades ago. These analyses have consistently rooted the an- been one of the most contentious issues in plant systematics and giosperm phylogenetic tree among the ANITA lines, namely evolution, both before and after the introduction of phyloge- Amborella , Nymphaeales, and Austrobaileyales ( Mathews and netic methods ( Crane, 1985 ; Doyle and Donoghue, 1986 ; Nixon Donoghue, 1999 ; Parkinson et al., 1999 ; Q iu et al., 1999 ; Renner, et al., 1994 ; Doyle, 1994 , 1996 ). Molecular analyses contradict 1999 ; Soltis et al., 1999, 2000 ; Barkman et al., 2000 ; Graham one of the few points on which morphological analyses agreed, and Olmstead, 2000 ; Zanis et al., 2002 ), which has focused at- that Gnetales are the closest living relatives of angiosperms tention on these taxa as particularly likely to yield insights on ( Donoghue and Doyle, 2000 ; Burleigh and Mathews, 2004 ; the fi rst angiosperms ( Doyle and Endress, 2000 ; Endress and Soltis et al., 2005 ), but they say nothing about fossil relatives. Igersheim, 2000a , b ; Endress, 2001 , 2004 , 2006, 2008a; Fried- Several recent studies, some of which take into account mo- man and Williams, 2003 , 2 004; Williams and Friedman, 2004 ; Friedman, 2006 ; Endress and Doyle, 2007 ). The main uncer- tainty is whether A mborella and Nymphaeales form two succes- 1 Manuscript received 7 February 2008; revision accepted 12 September 2008. sive branches or a clade ( Barkman et al., 2000 ), with some The authors thank E. M. Friis and M. Frohlich for useful discussions and recent support for the latter hypothesis from mitochondrial genes suggestions that improved the manuscript. J.A.D. thanks P. Garnock-Jones ( Qiu et al., 2006 ), but the former supported by recent analyses of and the School of Biological Sciences, Victoria University of Wellington, entire plastid genomes ( Jansen et al., 2007 ; Moore et al., 2007 ). for facilities and a supportive environment during preparation of this paper. An alternative rooting based on plastid genomes of fewer taxa, This work was facilitated by travel support from the NSF Deep Time with grasses the sister group of all other angiosperms ( Goremykin Research Coordination Network (RCN0090283). 4 Author for correspondence (e-mail: [email protected]) et al., 2003 ), appears to be an artifact of low taxon sampling and long branch attraction ( Degtjareva et al., 2004 ; Soltis and Soltis, doi:10.3732/ajb.0800047 2004 ; Stefanovic et al., 2004 ; Leebens-Mack et al., 2005 ). 22 January 2009] Endress and Doyle — Ancestral flowers 23 All other angiosperms form a strongly supported clade, named could be said to be typical at a relatively “ basal ” level of angio- Mesangiospermae by Cantino et al. (2007) , but relationships sperms (groups other than monocots and eudicots, or “ Magno- among several lines in this clade remain poorly resolved, prob- liidae ” in the paraphyletic sense of Takhtajan, 1964 ), but ably as a result of very rapid radiation ( Moore et al., 2007 ). One because they were scattered in different taxa (e.g., anther open- important area of current uncertainty is the position of Chloran- ing by valves, spiral fl oral phyllotaxis, inner staminodes, trim- thaceae, which have been the subject of much discussion be- erous fl owers), it was possible to entertain several alternative cause of their extremely simple fl owers. Combined analyses of models for the ancestral fl ower (e.g., Endress, 1986a ). However, morphological and molecular data ( Doyle and Endress, 2000 ) especially since 1999, a more precise discussion is possible be- and some molecular studies ( Qiu et al., 2005 ; Duvall et al., cause phylogenetic reconstructions are generally more advanced, 2006 ; Mathews, 2006 ) have placed Chloranthaceae at the base and specifi cally the topology of the basal grade of extant angio- of mesangiosperms, but they are nested within mesangiosperms sperms is well supported and can be used as a basis for discus- in most molecular trees, including most of those found in analy- sions on evolution. As emphasized by Crisp and Cook (2005) , ses of complete plastid genomes ( Jansen et al., 2007 ; Moore it cannot be assumed that single low-diversity “ basal ” lines are et al., 2007 ). Suggestions that fl owers of Chloranthaceae were plesiomorphic in any given character, but when several lines primitive based on the abundance of apparently related fossils branch sequentially below the vast bulk of a clade, as is appar- in the Early Cretaceous (reviewed by Eklund et al., 2004 ; Friis ently the case for angiosperms, and these lines share the same et al., 2006 ) have faded with fi rm establishment of the basal character state, this state can be reconstructed by parsimony ANITA grade, but if Chloranthaceae are sister to the remaining analysis as ancestral. We took advantage of the new evidence mesangiosperms they could still be relevant to reconstruction on rooting in an analysis of basal angiosperms (including basal of the original fl ower and its initial modifi cations. monocots and basal eudicots), in which we used parsimony op- Comparative studies of fl oral developmental genetics repre- timization on a tree based on morphological data and r bcL , sent another growing fi eld that promises to provide new insights atpB , and 18S rDNA sequences ( Soltis et al., 2000 ) to estimate on early fl oral evolution. Such studies have already been used for ancestral states and trace character evolution ( Doyle and interpolations between angiosperms and other living seed plants Endress, 2000 ). In later articles we concentrated on implica- and within angiosperms ( Frohlich and Parker, 2000 ; Frohlich, tions of this data set for evolution of pollen morphology ( Doyle, 2003 , 2006 ; Baum and Hileman, 2006 ; Irish, 2006 ; S oltis et al., 2005 ), leaf architecture ( Doyle, 2007 ), fl oral phyllotaxis ( Endress 2006 ; Frohlich and Chase, 2007 ; Theissen and Melzer, 2007 ). and Doyle, 2007) , and the position of Hydatellaceae ( Saarela Several spectacular new fi ndings have brought the aquatic et al., 2007 ). This “ angiosperm-centered ” or “ top-down ” approach habitat to the center of the attention and debate on early angio- ( Bateman et al., 2006 ) can be questioned on the grounds that in sperm evolution (e.g., Sun et al., 2002 ; Friis et al., 2003 ; Crepet theory outgroup and ingroup relationships cannot be addressed et al., 2004 ; Feild and Arens, 2007 ). These are (1) the recogni- separately. However, in practice this seems less problematical tion of new aquatic angiosperms in the Early Cretaceous fossil than anticipated, thanks to the increasingly robust rooting of record, including Archaefructus ( Sun et al., 1998 , 2002 ; Friis angiosperms based on molecular data. et al., 2003; Ji et al., 2004 ), which had fertile axes bearing paired In the present paper we discuss changes in our perception of stamens, single or paired carpels, and no perianth; Monsechia , the fl ower in the most recent common ancestor of living angio- tentatively interpreted as a bryophyte when it was fi rst described sperms (the crown group node) and its initial evolutionary mod- ( Gomez et al., 2006 ); and S cutifolium , assigned to Cabom- ifi cations, using an updated version of the Doyle and Endress baceae ( Taylor et al., 2008 ), in addition to previously recog- (2000) data set and taking into account new evidence from phy- nized water plants such as N elumbites ( Doyle and Hickey, logenetic and structural studies on extant plants, fossils, and 1976 ; Upchurch et al., 1994 ; Mohr and Friis, 2000 ; Wang and evo-devo studies. It is unlikely that this “ ancestral fl ower ” was Dilcher, 2006 ); (2) the discovery that the submerged water plant the “ fi rst fl ower ” in a morphological sense, which may have family Hydatellaceae belongs not in monocots but rather to originated much earlier on the angiosperm stem lineage. We Nymphaeales in the ANITA grade ( Saarela et al., 2007 ); and will not consider the origin of the angiosperm fl ower in relation (3) indications from some (though not most) molecular analy- to reproductive structures in outgroups, which requires consid- ses that the aquatic genus C eratophyllum (= Ceratophyllaceae), eration of fossil seed plants, a topic treated elsewhere ( Doyle, which had a brief period of fame as the inferred sister group of 2006 , 2008). all other angiosperms in analyses of rbcL ( Chase et al., 1993 ), Our previous study ( Doyle and Endress, 2000 ) presented a may belong just above the basal angiosperm grade, with Chlo- list of inferred ancestral states for all characters, but this needs ranthaceae ( Duvall et al., 2006 ; Mathews, 2006 ; Qiu et al., reassessment in light of new data. Since 2000, we have been 2006 ). The importance of fresh-water habitats and fl ood plains revising our data set by adding new characters, refi ning old ones, in early angiosperm history has long been recognized by paleo- adding new taxa, and splitting taxa into more homogeneous botanists ( Doyle and Hickey, 1976 ; Taylor and Hickey, 1992 ), units to analyze character evolution in more detail and fi ll in and continues to be a major topic of discussion by paleoecolo- less well sampled parts of the tree. We have not yet performed gists ( Mart í n-Closas, 2003 ; Feild et al., 2004 ; Coiffard et al., a new combined analysis of morphological and molecular data, 2007 ; Feild and Arens, 2007 ). The impression that Early Creta- but we have made changes in the tree used as a framework for ceous angiosperms included a large number of water plants discussion where recent data provide robust evidence for differ- may be partly due to a bias in favor of fossilization of aquatics ent relationships. For example, our previous combined analysis over other plants, but water plants were clearly more common linked Piperales with monocots, but accumulating molecular data than expected under the old view that the initial diversifi cation ( Zanis et al., 2002 ; Sauquet et al., 2003 ; Qiu et al., 2005 , 2006 ) of angiosperms involved woody plants (cf. Doyle and Hickey, consistently associate them with Canellales (Canellaceae, Win- 1976 ). teraceae), Magnoliales, and Laurales, in a clade named M agnolii- Until about ten years ago, only a vague recognition of more dae by Cantino et al. (2007) , not to be confused with Magnoliidae widespread features in basal angiosperms was possible. They in the paraphyletic sense of Takhtajan (1964) and others. 24 American Journal of Botany [Vol. 96 Our most important change in taxon sampling is the addition phylogenetic analyses, particularly those related to rooting, to estimate ances- of two aquatic groups: Hydatellaceae, now linked with Nympha- tral states. Our assumptions on rooting of taxa and the publications on which eales ( Saarela et al., 2007 ), and C eratophyllum . In Doyle and they are based are cited in the taxon list. Improved information on relationships within taxa has led us to make many minor changes in scoring of taxa since Endress (2000), we omitted C eratophyllum because many char- Doyle and Endress (2000) . acters in our matrix were lacking or uninterpretable due to re- duction, its position was unstable in preliminary analyses, and Taxa — Besides adding Ceratophyllum and Hydatellaceae, we have in- we assumed it would have a minor effect on inferences on char- creased our taxon sampling in other groups. In Chloranthaceae, we split C hlo- acter evolution because of its specialized, reduced nature. How- ranthus and S arcandra , treated as one taxon in Doyle and Endress (2000) , into ever, omitting C eratophyllum is no longer justifi able in light of the two genera and rescored several characters based on Eklund et al. (2004) . In the increasing number of other near-basal taxa with simple Ranunculales, we have added C ircaeaster and split H ydrastis and G laucidium from “ core ” Ranunculaceae (their sister group according to Hoot et al., 1999 ) fl owers and the suggestion that they represent a prefl oral state because they differ substantially in fl oral features and may thus affect recon- ( Friis and Crane, 2007 ). Whether such fl owers are reduced struction of fl oral evolution. For the same reason, we split Magnoliaceae into should be tested rather than assumed. The claim that C erato- Liriodendron and Magnolioideae ( Magnolia s. l. of recent authors) and Trocho- phyllum is the sister group of eudicots, based on analyses of dendraceae into Trochodendron and T etracentron . We have modifi ed the scor- complete plastid genomes (J ansen et al., 2007 ; Moore et al., ing of Platanaceae (now Platanus ) and Buxaceae, in which we previously 2007 ), also needs to be evaluated in light of morphological data included presumed fossil relatives, to apply strictly to the extant crown groups, in anticipation of testing the position of the fossils. We have increased our sam- and its implications explored. pling of Alismatales for future tests of comparisons with fossils. We have added Another major change is addition of the Early Cretaceous Nartheciaceae because they appear to be related to but less modifi ed than Di- fossil plant A rchaefructus , which a cladistic analysis by Sun oscoreaceae ( Caddick et al., 2002a ), and Melanthiaceae as a relatively ple- et al. (2002) identifi ed as the sister group of all extant angio- siomorphic exemplar of Liliales ( Chase et al., 2006 ). sperms (i.e., a stem relative). This interpretation was questioned In Ceratophyllum , the fertile structures have been variously interpreted by Friis et al. (2003) , who interpreted Archaefructus as a crown ( Endress, 1994b ; Iwamoto et al., 2003 ): at one extreme, as female fl owers with a single carpel surrounded by tepals and as male fl owers with tepals and numer- group angiosperm with reduced unisexual fl owers, but reaffi rmed ous stamens; at the other, as female fl owers with no perianth but bracts lower by Crepet et al. (2004) . In either case, its unusual combination on the axis and as spikes with basal bracts and numerous male fl owers consist- of characters make it potentially relevant to reconstruction of ing of one stamen, with no perianth or individual subtending bract ( Endress, the ancestral fl ower. The addition of Archaefructus is part of a 2004 ). For the purposes of this analysis, we have provisionally accepted the general effort to integrate fossils into the phylogeny of living second interpretation. First, in the female structures, single carpels occasionally basal angiosperms ( Doyle and Endress, 2007 ). Results of our occur in the axils of the sterile appendages ( Aboy, 1936 ; Iwamoto et al., 2003 ), suggesting that the latter are bracts rather than tepals and the system is a re- analyses concerning other fossil taxa are not presented here be- duced infl orescence. Second, the stamens have an extremely labile phyllotaxis cause they have less impact on reconstruction of the ancestral and marked acropetal delay in maturation ( Endress, 1994b ), which is a common fl ower. Some of these fossils are too deeply nested within mag- pattern in fl owers of spicate infl orescences but anomalous within the androe- noliids and eudicots to affect inferred ancestral states; others cium of a multistaminate fl ower. may be more basal (e.g., taxa apparently related to Chloran- Our scoring of A rchaefructus is based on whole plants of A . liaoningensis thaceae: Eklund et al. [2004]) but add few new elements be- and A . sinensis ( Sun et al., 1998 , 2001 , 2002 ; Friis et al., 2003 ); features are generally consistent in A . eofl ora ( Ji et al., 2004 ), which may represent either a cause they resemble their presumed extant relatives in fl oral smaller species or a younger stage of A . sinensis . We analyzed the position of features. Archaefructus using two alternative scorings, one (A rchaefructus inf) assuming Besides considering implications of current phylogenies for that fertile axis was a raceme of male and female fl owers consisting of usually evolution of individual fl oral characters and character com- two stamens and one or two carpels, the other (A rchaefructus fl o), following Sun plexes, we stress several specifi c broader issues. These include et al. (2002) , that it was a bisexual fl ower or prefl ower with paired stamens be- (1) what present data say about evolutionary interpretation of low and carpels above. Friis et al. (2003) questioned whether the bodies that Sun et al. (2001 , 2002 ) described as pollen were in fact pollen grains, because of their the fl owers of Magnoliales and Winteraceae, once widely as- irregular size and shape, but we provisionally assume that at least some of them sumed to be primitive; (2) whether the simple fl ower structure are pollen and have scored them based on the most convincing specimen, illus- in some basal angiosperms (Hydatellaceae, A rchaefructus , Cer- trated in Fig. 2F of Sun et al. (2002) . To evaluate the alternative interpretation, we atophyllum , Chloranthaceae) is due to reduction of more “ com- have used a third scoring ( Archaefructus NP) that corresponds to Archaefructus plete ” fl owers, retention of a “ prefl oral ” state, or to breakdown inf with pollen characters treated as unknown. Sun et al. (1998 , 2002 ) described of the distinction between fl owers and infl orescences due to the carpels as conduplicate (= plicate), but in extant carpels of similar appear- ance this cannot be determined without developmental or anatomical evidence loss of fl oral identity, issues raised by Friis and Crane (2007) ( Friis et al., 2003 ; Endress, 2005 ). They described the fruits as follicles, but they and Rudall et al. (2007) , and if and how this might be related to did not actually report dehiscence. The seeds appear to have a palisade exotesta an aquatic habit; and (3) what the evolutionary consequences of as defi ned here (character 101, including not only radially elongated but also a position of Chloranthaceae just above the basal grade, with or shorter sclerotic cells): Sun et al. (1998 , 2 002) described the surface as consist- without Ceratophyllum ( Doyle and Endress, 2000 ; Duvall ing of epidermal cells with cutinized anticlinal and periclinal walls. et al., 2006 ; Mathews, 2006 ; Qiu et al., 2006 ), would be for in- Ji et al. (2004) interpreted seeds of A . eofl ora as orthotropous, but their pub- lished illustrations are not clear enough to determine whether the seeds were terpretation of the fl owers in these groups, and how these would anatropous or orthotropous; the fi gure of the end of a seed in Fig. 2C of Sun differ in the context of the plastid genome trees ( Jansen et al., et al. (1998) is actually more suggestive of an anatropous ovule. Hence we have 2007 ; Moore et al., 2007 ), where the two taxa are nested sepa- scored ovule curvature (93) as unknown. Ji et al. (2004) described one lateral rately within mesangiosperms. unit in A. eofl ora as bisexual, with one stamen and two carpels, and they also interpreted one unit in the type specimen of A . sinensis ( Sun et al., 2002 ) as bisexual. Under our character defi nition, A. eofl ora might be scored as uncertain (0/1) for fl ower sexuality (26). However, we believe it would be premature to MATERIALS AND METHODS rescore A rchaefructus as a whole in this way. Lists of taxa and characters and the data matrix are presented in Appendix 1. Characters — In this study we have not included all the characters in our In dealing with characters that vary within taxa, we have not simply scored most recent version of the Doyle and Endress (2000) data set, many of which characters as uncertain ( “ polymorphic ” ) but have made use of results of are not relevant for our present purposes, where we have used fi xed backbone January 2009] Endress and Doyle — Ancestral flowers 25 constraint trees as a framework for placement of Ceratophyllum and Archae- Scoring the number of perianth and stamen whorls (34, 43) is straightfor- fructus and reconstruction of fl oral evolution, and would require excessively ward in whorled taxa (except for seemingly tetramerous fl owers that actually lengthy documentation and argumentation. We have included all fl oral charac- have dimerous whorls, as in Proteaceae, T etracentron , and Buxaceae, as in- ters, including those of stamen, carpel, and ovule morphology. In addition, we ferred from the fact that the stamens appear to be opposite the tepals: von Balt- include all those nonfl oral characters needed for analysis of the position of C er- hazar and Endress, 2002a; Chen et al., 2007 ), but again the treatment of spiral atophyllum and Archaefructus . Characters omitted because they do not exist or taxa poses problems. Many spiral taxa (and N elumbo , with chaotic stamen in- are inapplicable in C eratophyllum include aspects of secondary xylem and sertion on an androecial ring meristem: Hayes et al., 2000 ) have numbers of phloem, leaf anatomy, and the inner and outer integuments, which are reduced tepals and/or stamens that are comparable to those of taxa with more than two or fused into a single integument. whorls, so we have scored them accordingly. We also used the number of series Some of the most important and complex arguments for decisions in defi ni- (in which a certain number of parts fi lls the circumference of the fl ower; En- tion of characters and scoring of particular taxa are discussed in this section, dress and Doyle, 2007 ) as a rough substitute for number of whorls. others that are less problematic or signifi cant in Appendix 1. Because of space In most basal angiosperms, all perianth parts are best described as tepals limitations, we cite only general sources of information for particular character ( Hiepko, 1965 ; Walker and Walker, 1984 ; Endress, 2001 ; Ronse De Craene, sets and especially important references on particular taxa, and reserve more 2008 ) because they are less strongly differentiated than the typical sepals and detailed documentation and resolution of differences between our interpreta- petals of core eudicots ( Pentapetalae of Cantino et al., 2007 ). These tepals may tions and those of other workers for elsewhere. We concentrate on references be uniform (either sepaloid or petaloid) or differentiated into outer sepaloid and for new taxa and characters; for those used in Doyle and Endress (2000) , read- inner petaloid parts, distinctions recognized in character 35. We include both ers are referred to that article. tepals and more differentiated petals in the count of whorls, and staminodes as Our general philosophy on defi nition of characters is explained in more de- well as fertile stamens. However, we also introduced a separate character (36) tail in Doyle and Endress (2000) . Few of our characters are quantitative in the for presence or absence of typical petals (mostly in Ranunculales), defi ned on sense of continuous (e.g., pollen size, nexine thickness), but there are often se- more pronounced differences in anatomy and delay in development. Taxa with ries of conditions that could be grouped into many states or a few. In general, petals may show differentiation within the outer perianth whorls, such as we have tried to break the variation into a smaller number of states in ways that Nuphar , which has outer sepaloid and inner petaloid “ tepals ” or “ sepals ” and make morphological (especially developmental) sense and reduce the number much smaller petals. of uncertain ( “ polymorphic ” ) scorings of taxa (assuming that this reduction is We have made fewer changes from Doyle and Endress (2000) in characters evidence that the variations included in each of the states are related). Several of individual fl oral parts. Following Eklund et al. (2004) , to reduce uncertain important changes concern replacement of multistate with binary characters, scorings, we modifi ed the stamen base character (48) to combine short and wide which can sometimes improve resolution of relationships in cases where the and short and narrow in the same state, and the orientation character (53) to optimization of a multistate character would be ambiguous. In several cases we combine slightly introrse with latrorse. We previously treated modes of carpel previously used unordered multistate characters to combat the Maddison “ long sealing as a multistate character (corresponding to the four types of Endress and distance ” effect ( Maddison, 1993 ): where the ancestral state in one clade in Igersheim, 2000a ). However, carpel sealing has two potentially independent which a structure (more generally a character) occurs in two (or more) versions aspects, degree of postgenital fusion and secretion, which we have split into infl uences the polarity of the character in another clade that has the structure, two characters (76, 77). We separated types of papillae (82) from larger protu- even though the structure does not exist in the intervening lines and presumably berances (81), because pluricellular papillae and protuberances co-occur in A m- arose independently. This artifact can be avoided by treating lack of the struc- borella and Trimenia but not in other taxa and therefore appear to represent ture as one state of a multistate character and different versions of the structure independent characters. as other states. However, this procedure weakens the contrast between presence Many aspects of fl oral evolution were also treated by Ronse De Craene et al. and absence of the structure as an independent source of information on (2003) . They made less effort to ensure independence of characters: for exam- relationships. ple, lack of perianth was a state in three of their characters. This was not neces- An example that underlines the importance of the Maddison effect concerns sarily a problem in their study, in which they plotted characters on a molecular presence or absence of a perianth. In Doyle and Endress (2000), we treated the tree, and it may be useful in assessing the implications of different character number of perianth whorls as an unordered multistate character, with no peri- defi nitions. However, such redundancy poses problems if characters are used anth one of four states. A group where this may cause problems is Chloran- for tree reconstruction, since it may overweight what was presumably a single thaceae, where H edyosmum has one perianth whorl and the other genera have change — for example, loss of perianth. Because we intend to use our data set in no perianth. Since most outgroups have a perianth, its presence might appear to a future combined analysis and have used it to investigate the relationships of be evidence for the basal position of H edyosmum , or in other words its loss Ceratophyllum and Archaefructus in the current study, we have tried to mini- could be evidence for the monophyly of the remaining genera (which is sup- mize redundancy among characters. ported by molecular evidence). However, when presence and number of whorls Infl orescence characters deserve special attention as an area where we have are treated as a single unordered multistate character, scoring Hedyosmum as made major modifi cations. In our previous analysis ( Doyle and Endress, 2000 ), having one whorl does not favor a basal position in Chloranthaceae because the we recognized a relatively crude infl orescence character emphasizing degree of outgroups have two or more whorls, states that are not recognized as any more branching, with three states: solitary fl owers; racemes, spikes, and botryoids; similar to one whorl than to none. For this reason, Eklund et al. (2004) split the and more richly branched infl orescences such as panicles and compound infl o- Doyle and Endress (2000) character into two — one for presence or absence of a rescences of racemes, spikes, and botryoids. Thus in infl orescences of the sec- perianth, the other for number of whorls, with taxa lacking a perianth scored ond state, we did not recognize the standard contrast between indeterminate and as unknown — and we have followed this solution here (characters 31, 34). determinate infl orescences, or the related distinction of Troll (1964) and We- Maddison (1993) pointed out cases where this procedure is unlikely to cause berling (1989) between polytelic systems with no terminal fl ower (racemes, problems, notably where loss of a structure occurs in a terminal clade. With spikes, thyrses) and monotelic systems in which all axes terminate with a fl ower relationships largely inferred from molecular data, cases where this may cause (botryoids, thyrsoids, panicles — all sometimes imprecisely described as artifacts can usually be recognized and treated in discussion. “ cymes ” ). This was because of a perception that the two types intergrade within Additional important changes involve other characters of fl oral organization. taxa, such that many taxa would have to be scored as uncertain. However, In Doyle and Endress (2000), we recognized phyllotaxis and merism (merosity) closer examination has led us to conclude that a different grouping of traditional as separate characters in both the perianth and the androecium, but this poses types into basically monotelic and polytelic states (in character 22, Fig. 1 ) leads problems for scoring of merism in spiral taxa. Our solution was to treat merism to fewer problems than we had thought and is more informative: one state in- as a multistate character, with spiral taxa scored as (0) irregular and whorled cludes units lacking a terminal fl ower (racemes, spikes, thyrses), the other those taxa as (1) trimerous or (2) dimerous, tetramerous, or pentamerous. However, with a terminal fl ower (botryoids, thyrsoids, panicles). Although taxa often this may introduce bias due to redundancy of spiral phyllotaxis and irregular vary between types within each of these states, there is less variation between merism. One solution would be to combine phyllotaxis and merism into a single types belonging to the different states. character, but as discussed in Endress and Doyle (2007) , the distinction between Racemes and thyrses differ on whether the lateral units on the indeterminate spiral and whorled appears to be consistent and independent enough to be axis are single fl owers or cymes. Cymes are branching systems that can have treated separately. Our solution is to retain both characters (32, 33 for the peri- one to several branching orders (i.e., a main axis and lateral branches of one to anth; 41, 42 for the androecium) but score spiral taxa as unknown for merism. several orders formed by repeated branching of these laterals), but each axis has Optimization of this character across the tree produces artifactual reconstruc- not more than two lateral branches of the next higher order, and all axes are tions of merism in spiral taxa, but this can be considered in discussion. usually determinate. Botryoids are the other way around: they have only one 26 American Journal of Botany [Vol. 96 state (in character 22, Fig. 1 ): there is variation between these two extremes in many taxa, such as Austrobaileya , Eupomatia , Magnoliaceae, and Calycan- thaceae. If the axillary branch (pedicel) bearing the fl ower has no appendages or at most one or two prophylls, we call the system a raceme; if it has more sterile appendages, we call the fl ower solitary. This defi nition allows most taxa to be scored unambiguously. Schisandraceae are still mixed (0/1), since the number of bracteoles varies between zero and three or more among species ( Weberling, 1988 ; Saunders, 2000 ). Special problems in interpretation of Nymphaeales are treated in the Discussion, since they make more sense in a phylogenetic context. When fl owers are unisexual and the infl orescences of male and female fl ow- ers differ in type, we have scored the taxon based on the more complex type. Thus, we have scored H edyosmum , with male spikes and female thyrses, as having thyrses; and C eratophyllum , with solitary female fl owers and male spikes, as having spikes. Except for three pollen characters (see Appendix 1), all multistate characters were treated as unordered. Analyses — Our analyses (all based on parsimony) used “ backbone con- straint ” trees, with Recent taxa fi xed into one of two topologies. Analyses were performed with the program PAUP* version 3.1.1 ( Swofford, 1990 ) and in- volved 10 or 100 heuristic replicates, stepwise random addition of taxa, and tree-bisection-reconnection (TBR) branch swapping. The relative parsimony of alternative relationships was determined by searching for trees less than or equal to a given number of steps and observing the trees obtained or by moving taxa manually with MacClade ( Maddison and Maddison, 2003 ). The fi rst backbone tree (henceforth labeled D & E) is a modifi cation of the tree found in our morphological and three-gene analysis ( Doyle and Endress, 2000 ), with changes where accumulating molecular data have most strongly and consistently contradicted relationships found in our previous study. Essen- tially this is a handmade supertree. Besides linking Piperales with Canellales, as already discussed, we have moved Euptelea from within Ranunculales to the base of the order, following Kim et al. (2004a) ; this position is actually more parsimonious in terms of morphology. Taxa added or split for the reasons dis- cussed earlier have been placed following Les et al. (1997) , Hoot et al. (1999) , Fig. 1. Sketches illustrating infl orescence types included in three Soltis et al. (2000) , Chen et al. (2004) , and Chase et al. (2006) . In a preliminary states of infl orescence character (22). (A, B) = state 0; (C – E) = state 1; analysis, we added C eratophyllum to the data set and constrained all other rela- (F – H) = state 2. (A) solitary, terminal; (B) solitary, axillary; (C) botryoid; tionships as described. The tree found in this constrained analysis is the modi- (D) panicle; (E) thyrsoid; (F, G) raceme; (H) thyrse. fi ed D & E backbone tree used in subsequent analyses. The second backbone tree (labeled J/M) incorporates relationships of major clades found in analyses of whole plastid genomes by Jansen et al. (2007) and branching order (i.e., a main branch and lateral branches of only the fi rst order), Moore et al. (2007), notably with Chloranthaceae linked with magnoliids, Cer- but the number of fi rst-order lateral branches is not limited until the terminal atophyllum with eudicots, and the latter two with monocots. The same relation- fl ower is formed; as in cymes the axes are determinate. When there is only one ships were found by Saarela et al. (2007) in analyses of a smaller plastid data branching order and axes have only one or two lateral fl owers, cymes and set. Relationships within clades (which were sparsely sampled in the plastid botryoids cannot be distinguished unless more highly branched units are found. studies) are the same as in the D & E backbone tree. Racemes and thyrses both occur within taxa such as Chloranthaceae, where To investigate the position of Archaefructus , we analyzed the data set with Hedyosmum has thyrses of female fl owers and spikes of male fl owers, and some Archaefructus added, using both backbone trees. To assess implications of the species of A scarina have spikes, others thyrses. We have recognized this dis- hypothesis that Amborella and Nymphaeales form a clade, we rerooted trees tinction with a separate character (23, lateral units single fl owers or cymes). We manually with the program MacClade version 4.03 ( Maddison and Maddison, have also distinguished racemes from spikes by introducing a character con- 2003 ). trasting pedicellate and sessile fl owers (24). We used MacClade to optimize character evolution on trees, reconstruct Another important distinction concerns the presence or absence of bracts or ancestral states, and identify characters supporting relationships. When we re- leaves (pherophylls) subtending the fl owers (25). In Archaefructus , Sun et al. fer to features as unequivocal synapomorphies of particular clades, this does not (2002) cited the absence of bracts below the paired stamens and carpels as evi- mean they are uniquely derived, but rather that the change in state unequivo- dence that the fertile axis was a fl ower (or prefl ower) rather than an infl ores- cally occurs at this point on the tree, as opposed to cases where the position of cence. However, subtending bracts are absent in several groups in the present change is equivocal (e.g., where an earlier origin followed by a reversal and two data set, such as Hydatellaceae, A corus , and Araceae. later origins are equally parsimonious, or where the character state in neighbor- Other problems concern the distinction between solitary fl owers and ra- ing taxa is unknown). cemes, specifi cally when solitary fl owers are borne in the axils of more or less unmodifi ed vegetative leaves. Solitary axillary fl owers are sometimes distin- guished from lateral fl owers in a raceme based on whether they are subtended RESULTS by normal leaves or modifi ed bracts, but this is more a matter of degree than a fundamental difference in organization. This problem is illustrated by cases in When Ceratophyllum is added to the updated Doyle and En- which fl owers are borne in the axils of bracts on an axis that then reverts to producing vegetative leaves (e.g., S chisandra , Euptelea ; Endress, 1969 ; We- dress (2000) tree, its most parsimonious position is as the sister berling, 1988 ). In our previous analysis we scored these as solitary. Alterna- group of Chloranthaceae (776 steps; Fig. 2A ). It is nested within tively, even systems with fl owers in the axils of normal leaves are sometimes Chloranthaceae in all six trees that are one to three steps longer. described as racemes ( Weberling, 1989 ). Because mode of branching seems A position as the sister group of eudicots ( Jansen et al., 2007 ; more fundamental than variation between bracts and leaves, we have adopted Moore et al., 2007 ) is nine steps less parsimonious (785 steps); this approach, grouping systems where fl owers are borne in the axils of bracts a position as the sister group of monocots is eight steps less and regular leaves as racemes. We group fl owers that terminate either a long shoot (the classic terminal condition) or an axillary short shoot in the solitary parsimonious (784 steps). January 2009] Endress and Doyle — Ancestral flowers 27 Fig. 2. Representative most parsimonious trees obtained after addition of Archaefructus to backbone constraint trees of Recent basal angiosperms. OM and OE indicate presumed positions of other monocots and other eudicots, respectively. (A) Using D & E backbone tree, from combined morphological and molecular analysis of Doyle and Endress (2000) , with modifi cations based on more recent data. (B) Using J/M backbone tree, with relationships among major clades found in plastid genome analyses of Jansen et al. (2007) and Moore et al. (2007) , but with relationships within clades as in Fig. 2A . Nymph = Nymphaeales, Aust = Austrobaileyales, Chlor = Chloranthaceae, Piper = Piperales, Ca = Canellales, Magnol = Magnoliales. 28 American Journal of Botany [Vol. 96 When A rchaefructus is scored as having an infl orescence of supported by loss of bracts subtending the male fl owers (25) unisexual fl owers (A rchaefructus inf) and added to the D & E and dry fruit wall (97). backbone tree, its single most parsimonious position is as the A sister group relationship of C eratophyllum and eudicots, as sister group of Hydatellaceae (782 steps; Fig. 2A ). Its next best found in the plastid genome analyses of Jansen et al. (2007) and position (one step worse) is sister to the remaining Nymphae- Moore et al. (2007) and many other molecular analyses (e.g., ales (henceforth designated “c ore Nymphaeales ” ). Seven posi- Saarela et al., 2007 ), is nine steps less parsimonious and would tions are two steps worse: sister to all Nymphaeales, C abomba , be supported by only one unequivocal morphological synapo- Ceratophyllum , the Chloranthaceae- Ceratophyllum clade, all morphy, dry fruit wall (97), a highly homoplastic character. It is mesangiosperms except the Chloranthaceae-C eratophyllum eight steps less parsimonious to link Ceratophyllum with mono- clade, and either E uptelea or Circaeaster in the eudicots. cots, which would be supported by loss of cambium (4). The J/M backbone tree based on plastid genome data ( Jansen Whether this parsimony differential is suffi cient to overrule the et al., 2007 ; Moore et al., 2007 ) is 10 steps longer than the D & E molecular support for a relationship with eudicots needs to be tree (786 steps). When Archaefructus (inf) is added to the J/M tested by future combined analyses. However, it should be backbone tree, its most parsimonious position is sister to Cer- noted that bootstrap support for the link between C eratophyl- atophyllum (791 steps; Fig. 2B ). Next best are positions linked lum and eudicots is only modest (71% in Moore et al., 2007 ; with Hydatellaceae (one step worse) and sister to core Nympha- 74 – 89% in Saarela et al., 2007 ); that analyses by Moore et al. eales (two steps worse). (2007) using various methods and subsets of data gave different If pollen characters of A rchaefructus are scored as unknown topologies, some with Chloranthaceae in a more basal position; ( Archaefructus NP), its most parsimonious position with the and that other molecular analyses have linked C eratophyllum D & E backbone (781 steps; not shown) is sister to the eudicot with Chloranthaceae ( Duvall et al., 2006 ; Mathews, 2006 ; Qiu genus E uptelea (Ranunculales). Its next most parsimonious posi- et al., 2006 ). The strength of the morphological synapomor- tions (782 steps) are sister to Hydatellaceae, Ceratophyllum , Cer- phies might be questioned on the grounds that they largely rep- atophyllum plus Chloranthaceae, Ranunculales other than resent reductions and simplifi cations from ancestral states in Euptelea , Circaeaster (also Ranunculales), and the clade consist- angiosperms, which might be expected to give similar results ing of eudicots, monocots, and magnoliids. With the J/M back- regardless of their starting point. However, this is not in itself bone, omitting pollen characters strengthens the association of evidence that they are systematically worthless: without C er- Archaefructus with Ceratophyllum (789 steps), which becomes atophyllum all these features are valid synapomorphies of Chlo- three steps rather than one step more parsimonious than its next- ranthaceae, which are independently supported as a clade by best positions (792 steps), which are sister to Hydatellaceae, molecular data. Euptelea , Ranunculales other than E uptelea , and Circaeaster . These results suggest the intriguing possibility that C erato- When Archaefructus is scored as having a bisexual fl ower phyllum is an aquatic derivative of a terrestrial stem relative of ( Archaefructus fl o) and added to the D & E backbone tree, it has Chloranthaceae that already had many features of the crown three most parsimonious positions (783 steps; not shown): sis- group. Many additional changes would have to occur on the ter to Hydatellaceae, C abomba , and core Nymphaeales. Seven line leading to Ceratophyllum : origin of a protoxylem lacuna positions are one step worse, including not only elsewhere in (2), loss of cambium (4), loss of pericyclic fi bers (6), dissection Nymphaeales but also sister to Magnoliales plus Laurales, of the leaves (20) and shift to dichotomous venation (18), loss Magnoliaceae, C ircaeaster plus Lardizabalaceae, and C ir- of pollen aperture (62) (and almost total reduction of the exine: caeaster . When it is added to the J/M backbone tree, it has Takahashi, 1995 ), loss of stigmatic papillae (82), reduction or four most parsimonious positions (793 steps), including those fusion of the integuments to one (94), and large embryo (109). found with the D& E backbone and as the sister group of Chloranthaceae and their extinct relatives are emerging as one Ceratophyllum . of the fi rst successful angiosperm lines ( Eklund et al., 2004 ; Characters supporting these relationships of C eratophyllum Feild et al., 2004 ), which included greater diversity than would and Archaefructus are presented in the Discussion section. A be inferred from the four living genera alone. Our results con- list of inferred ancestral states in angiosperms for fl oral charac- cerning Ceratophyllum therefore raise the possibility that some ters is presented in Table 1 , with differences among eight trees, Early Cretaceous carpels or pollen that resemble Chloran- involving all combinations of the D & E vs. J/M backbone trees, thaceae might actually be closer to C eratophyllum and might Amborella sister to all other angiosperms vs. A mborella and therefore provide evidence on steps in its origin. Nymphaeales forming a clade, and exclusion vs. inclusion of Our analysis provides provisional support for the speculative Archaefructus . Parsimony optimizations of selected characters suggestion of Saarela et al. (2007) that A rchaefructus is related to on various trees are presented in Figs. 3– 12 . Hydatellaceae. This is the most parsimonious position of A rchae- fructus when its fertile axis is interpreted as a raceme of unisexual D ISCUSSION fl owers and C eratophyllum is associated with Chloranthaceae, as Phylogenetic results — Our inference that C eratophyllum is with the D & E backbone. Unequivocal synapomorphies of the related to Chloranthaceae is supported by fi ve unequivocal sy- two groups are loss of fl oral subtending bracts (25) and loss of napomorphies: sessile fl ower (character 24), one stamen (40), perianth (31). The fact that the fl owers are unisexual is consistent embedded pollen sacs (51), one carpel (74), and orthotropous but not indicative because the polarity of this character is equivo- ovule (93). Synapomorphies of Chloranthaceae that are not cal. Other features of A rchaefructus that support a relationship to found in Ceratophyllum and thereby place Ceratophyllum out- Nymphaeales as a whole are palmate venation (17) (reduced to side the family are sheathing leaf bases (12), interpetiolar one vein in Hydatellaceae), boat-shaped pollen (61), and palisade stipules (13), and stigmatic protuberances (81). Remarkably, it exotesta (101). This result would suggest that Hydatellaceae may is only one step less parsimonious to nest C eratophyllum within be what became of one member of the Archaefructus group after Chloranthaceae, where its best position is sister to H edyosmum , 125 Myr of further reduction in an aquatic habitat. January 2009] Endress and Doyle — Ancestral flowers 29 Table 1. Most parsimonious ancestral states for all characters concerning infl orescence and fl oral structure (see Appendix 1 for complete defi nitions), given different backbone trees (D & E vs. J/M), rooting with A mborella alone sister to all other angiosperms (A, N) vs. Amborella and Nymphaeales forming basal clade (A+N), and Recent taxa only vs. Recent taxa and fossil A rchaefructus (R, F). When the reconstructed ancestral state is identical for all trees, it is given only once. D&E J/M A,N R A+N R A,N F A+N F A,N R A+N R A,N F A+N F 22. Infl orescence 0 solitary, 1 botryoid etc., 2 raceme etc. 1/2 2 1/2 2 1/2 2 1/2 2 23. Infl orescence units 0 single fl ower, 1 cymes 0 solitary fl ower 24. Pedicel 0 present, 1 absent (sessile) 0 present 25. Bracts 0 present, 1 absent in male, 2 absent in all 0 present 26. Sex of fl owers 0 bisexual, 1 unisexual 0/1 bi/unisexual 27. Hypanthium 0 absent, 1 present, 2 inferior ovary 0/1 0 0/1 0 0/1 0 0/1 0 28. Receptacle 0 short, 1 elongate 0 short 29. Cortical vasculature 0 none or P, 1 A, 2 A plus G 0 none 30. Floral apex 0 used up, 1 protruding 0 used up 31. Perianth 0 present, 1 absent 0 present 32. Perianth phyllotaxis 0 spiral, 1 whorled 0/1 spiral/whorled 33. Perianth merism 0 trimerous, 1 dimerous, 2 polymerous 0 trimerous 34. Perianth whorls 0 one, 1 two, 2 more than two 2 more than two 35. Tepal differentiation 0 sepaloid, 1 sep + pet, 2 petaloid 0/1 1 0/1 1 0/1 1 0/1 1 36. Petals 0 absent, 1 present 0 absent 37. Inner perianth nectaries 0 absent, 1 present 0 absent 38. Outer perianth fusion 0 free, 1 fused 0/1 0 0/1 0 0/1 0 0/1 0 39. Calyptra 0 absent, 1 present 0 absent 40. Stamen number 0 more than one, 1 one 0 more than one 41. Stamen phyllotaxis 0 spiral, 1 whorled 0/1 spiral/whorled 42. Stamen merism 0 trimerous, 1 dimerous, 2 polymerous 0 trimerous 43. Stamen whorls 0 one, 1 two, 2 more than two 2 more than two 44. Stamen position 0 single, 1 double 0 single 45. Stamen fusion 0 free, 1 connate 0 free 46. Inner staminodes 0 absent, 1 present 0 absent 47. Food bodies 0 absent, 1 on staminodes 0 absent 48. Stamen base 0 short, 1 long wide, 2 long narrow 1/2 1/2 0/1/2 0/1/2 1/2 1/2 1/2 1/2 49. Paired basal stamen glands 0 absent, 1 present 0 absent 50. Connective apex 0 extended, 1 truncated, 2 peltate 0 0 0 0 0/1 0/1 0 0 51. Pollen sacs 0 protruding, 1 embedded 0 protruding 52. Microsporangia 0 four, 1 two 0 four 53. Orientation 0 introrse, 1 latrorse, 2 extrorse 0/1 0/1 0/1 0/1 0 0 0 0 54. Dehiscence 0 longitudinal slit, 1 H-valvate, 2 fl aps 0 longitudinal slit 74. Carpel number 0 more than one, 1 one 0 more than one 75. Carpel form 0 ascidiate, 1 intermediate, 2 plicate 0 ascidiate 76. Postgenital fusion 0 none, 1 partial, 2 complete 0 none 77. Secretion 0 present, 1 absent 0 present 78. PTTT a 0 not differentiated, 1 differentiated, 2 multilayered 0 not differentiated 79. Style 0 absent, 1 present 0 0/1 0/1 0/1 0 0/1 0 0/1 80. Stigma 0 extended, 1 restricted 0 extended 81. Stigmatic protuberances 0 absent, 1 present 0/1 0 0/1 0 0/1 0 0/1 0 82. Stigmatic papillae 0 none, 1 unicellular, 2 pluricellular 1/2 uni/pluricellular 83. Extragynoecial compitum 0 absent, 1 present 0 present 84. Fusion 0 apocarpous, 1 paracarpous, 2 eusyncarpous 0 apocarpous 85. Oil cells 0 not visible, 1 intrusive 0 not visible 86. Unicellular hairs on carpels 0 absent, 1 present 0 absent 87. Curved hairs on carpels 0 absent, 1 present 1 0/1 1 0/1 0 0/1 0 0/1 88. Abaxial nectaries 0 absent, 1 present 0 absent 89. Septal nectaries 0 absent, 1 present 0 absent 90. Ovule number 0 one, 1 mostly two, 2 more than two 0 0 0/2 0/2 0 0 0 0 91. Placentation 0 ventral, 1 laminar-dorsal 0 ventral 92. Ovule direction 0 pendent, 1 horizontal, 2 ascendent 0 pendent 93. Ovule curvature 0 anatropous, 1 orthotropous 0/1 0 0/1 0 0/1 0 0/1 0 ap ollen tube transmitting tissue. 30 American Journal of Botany [Vol. 96 In contrast, with the J/M backbone tree, in which C erato- Archaefructus has racemes and Hydatellaceae have modifi ed phyllum is divorced from Chloranthaceae and associated with thyrses, as Rudall et al. (2007) argued, the order of fl owers in the eudicots, it is more parsimonious to associate Archaefructus two groups cannot be so easily compared. If the main axis of the with C eratophyllum , based on dissected leaves (20), dichoto- infl orescence in Hydatellaceae (as reconstructed by Rudall et al., mous venation (18), loss of fl oral bracts (25), unisexual fl owers 2007 , in fi g. 5D) is compared with the main axis in Archaefruc- (26), and loss of perianth (31). This position is four steps less tus , there is no difference in the relative position of male and fe- parsimonious with the D & E backbone, where C eratophyllum is male fl owers in the two groups. In both, the male fl owers (plus associated with Chloranthaceae, which have more features that female fl owers in Hydatellaceae) are borne on more basal lateral confl ict with those of Archaefructus , such as opposite leaves units (cymes in Hydatellaceae), while the more distal lateral units (9), pinnate venation (17), round pollen (61), and reticulate tec- are entirely female. Furthermore, the argument that an opposite tum (66). Better evidence on the position of Ceratophyllum order of male and female fl owers precludes a relationship is not could therefore have an impact on the best interpretation of Ar- compelling because analogies with other groups suggest that the chaefructus . Our results also depend on uncertain assumptions order of fl owers in bisexual infl orescences can reverse. For ex- concerning the morphology of the fertile structures of A rchae- ample, in Buxaceae, male fl owers are basal and female fl owers fructus . When the fertile shoot is interpreted as a fl ower or pre- terminal in B uxus and Styloceras kunthianum , female basal and fl ower ( Sun et al., 2002 ), which we regard as unlikely, one of male apical in S arcococca and P achysandra , and infl orescences the most parsimonious positions of A rchaefructus is still with are unisexual in other S tyloceras species. Based on inferred phy- Hydatellaceae, but it is equally parsimonious to place it else- logenetic relationships ( von Balthazar and Endress, 2002b ), ei- where in Nymphaeales. Confi rmation of the view of Ji et al. ther one or the other bisexual condition could be ancestral, but (2004) that the seeds of A rchaefructus were orthotropous would the other bisexual type would be derived from it. increase the relative parsimony of a link with C eratophyllum . Some of the evidence for a relationship of A rchaefructus Ancestral fl oral states and initial specializations — In the with Hydatellaceae comes from the report by Sun et al. (2001 , following sections, we consider the ancestral state reconstruc- 2002 ) of boat-shaped, tectate monosulcate pollen grains in Ar- tions in Table 1 and their general implications. Contrary to chaefructus , which was questioned by Friis et al. (2003) . With some expectations (e.g., Qiu et al., 2006 ), trees in which Ambo- the D & E backbone, removal of pollen characters weakens the rella is sister to all other angiosperms and those in which it is connection of Archaefructus with Hydatellaceae and favors a linked with Nymphaeales have only modestly different impli- link with the eudicot genus Euptelea , supported in part by ab- cations for ancestral states: all seven differences involve cases sence of a perianth (31) and one stamen whorl (43), as well as in which the ancestral state is equivocal with one rooting and palmate venation (17), shared with eudicots as a whole, and one of the same two states with the other. Addition of A rchae- several ovules (90), a synapomorphy of mesangiosperms other fructus has even less impact, with a few important exceptions to than Chloranthaceae and C eratophyllum . The possibility that be discussed. Finally, except for the positions of Chloranthaceae Archaefructus was related to eudicots was raised by Friis et al. and Ceratophyllum , the differences between arrangements of (2003) , based especially on the ternate, dissected leaf architec- mesangiosperm lines in the D & E combined and J/M plastid ture. Such a relationship would imply that A rchaefructus had trees ( Jansen et al., 2007 ; Moore et al., 2007 ) have generally tricolpate rather than monosulcate pollen, which would be sur- minor effects. This result seems due to two factors. First, infer- prising in light of its Barremian-Aptian age, when tricolpate ences on ancestral states are most dependent on relationships in pollen was exceedingly rare outside northern Gondwana ( Doyle, the ANITA grade, which are the same with both arrangements. 1992 ; Hughes, 1994 ; Hochuli et al., 2006 ). However, even in Second, very few morphological changes occurred on the inter- the absence of pollen characters, relationships with Hydatel- nodes between the three main lineages of mesangiosperms laceae and C eratophyllum remain almost as parsimonious with (magnoliids, eudicots, and monocots), however they are ar- the D & E backbone, and the link with C eratophyllum is strength- ranged, presumably because these lineages radiated in a very ened with the J/M backbone. These results underline the need short time ( Moore et al., 2007 ), the same reason their relation- for more convincing evidence on pollen of A rchaefructus . ships have been so diffi cult to resolve. Our analysis does not address the hypothesis that Archae- fructus is a stem relative of all living angiosperms rather than a Infl orescence organization — Because of varying views on member of the crown group ( Sun et al., 2002 ): it only specifi es interpretation of fl owers and infl orescences in taxa such as Ar- the most parsimonious position(s) of A rchaefructus if it belongs chaefructus , Hydatellaceae, and Chloranthaceae and recent in the crown group. However, the crown group hypothesis was suggestions that the distinction between infl orescences and supported by an analysis of living and fossil seed plants (Doyle, fl owers may be labile or problematic in basal angiosperms ( Friis 2008), including all the ANITA lines, Chloranthaceae, and and Crane, 2007 ; Rudall et al., 2007 ), we have considered char- three magnoliids. When Archaefructus was interpreted as hav- acters of infl orescences as well as fl owers. ing an infl orescence of unisexual fl owers, its most parsimoni- Based on our results, with A mborella basal ( Fig. 3 ), the an- ous position was with Hydatellaceae, and a position sister to all cestral infl orescence type (character 2 2 ) in angiosperms is living angiosperms was fi ve steps worse. When the fertile axis equivocal: either botryoids, as in Amborella ; or racemes (which was interpreted as a bisexual fl ower, it was again more parsimo- some authors might describe as stems with solitary axillary nious to place A rchaefructus in Nymphaeales than below living fl owers), as in Nymphaeales, Chloranthaceae (modifi ed to angiosperms, but by three steps rather than fi ve. spikes and thyrses), and basal eudicots and monocots. However, Rudall et al. (2007) cited the order of fertile parts in Archae- if Amborella is linked with Nymphaeales, the ancestral type can fructus (stamens basal, carpels apical) as an argument against a be reconstructed as a raceme. Both hypotheses imply that soli- relationship with Hydatellaceae, where the female fl owers in spe- tary fl owers, often considered ancestral in angiosperms, are in- cies with bisexual infl orescences are to the outside (assumed stead derived: from racemes in Austrobaileyales (with a shift to to be basal) and male fl owers are central (apical). However, if botryoids in T rimenia ), magnoliids, and Nelumbo , and from January 2009] Endress and Doyle — Ancestral flowers 31 botryoids in the Hydrastis - Glaucidium clade in Ranunculaceae. the base of the pedicel, but she argued that the N uphar condition In magnoliids, solitary fl owers may have evolved either once is derived, as a result of intercalary growth between the abaxial from racemes at the base of the Magnoliales-Laurales clade, tepal and the rest of the fl ower. This interpretation is less plausi- with a reversal in Myristicaceae and a shift to botryoids in Lau- ble in terms of outgroup comparison. We have therefore scored rales, or separately from racemes in Magnoliales and from ei- both Nuphar and Nymphaeoideae (N ymphaea , Euryale , and V ic- ther racemes or botryoids in Laurales (Calycanthaceae). toria ) as having racemes, with the fl oral subtending bract present Thyrses, distinguished from racemes and spikes by the in Nuphar but absent in Nymphaeoideae (which could be due lateral unit character ( 23 ; cymes rather than single fl owers), either to reduction or to incorporation into the perianth). The appear to be derived from racemes in Hydatellaceae, Chloran- condition in B arclaya is unknown, although it appears consistent thaceae, Aristolochioideae, and B utomus , and from either ra- with that in Nuphar and N ymphaea . cemes or botryoids in Siparunaceae and Hernandiaceae. S essile From this perspective, the whole shoot system of Nymphae- fl owers ( 24 ) were derived from pedicellate ones, resulting in aceae can be considered a giant raceme. If the pherophyll-bud spikes in Chloranthaceae and C eratophyllum (a synapomorphy primordium is viewed as a complex of two parts (cf. Chassat, with the D & E backbone, a convergence with the J/M back- 1962 ), in some cases, the pherophyll part develops into a foliage bone), the Piperaceae-Saururaceae clade, monocots (separately leaf, and the fl oral bud is suppressed; in others, the fl oral bud in Acorus , Araceae, and Aponogeton ), and Platanus (modifi ed grows rapidly after initiation, and the pherophyll is reduced to a into heads), and botryoids with sessile fl owers (i.e., stachyoids) thin bract or nothing at all. One could also suggest there is com- in Tetracentron . In all these cases, reduction of the pedicel is petition for space: either the fl ower or the leaf is reduced, and the correlated with general fl oral reduction. Loss of bracts ( 25 ; ar- other, more precocious part “ wins. ” This divergence in develop- rows in Fig. 3 ) occurred in several lines in which racemes were ment may be a function of the gigantism of the shoots, leaves, modifi ed to spikes (H edyosmum and C eratophyllum , either and fl owers of these plants, compared to their outgroups. once or twice, depending on backbone tree and optimization; Victoria and E uryale may provide indirect support for this Acorus , Araceae, A ponogeton , Platanus ) or thyrses of reduced interpretation. Borsch et al. (2007) identifi ed these taxa as the pedicellate fl owers (Hydatellaceae) and might also seem corre- sister group of N ymphaea , but subsequent analyses of more lated with reduced fl owers. However, this is not a universal rule markers ( L ö hne et al., 2007 ) indicate they are nested within because bracts were also lost within Nymphaeaceae, in which Nymphaea and are therefore unlikely to represent the ancestral fl owers are unusually large. condition in Nymphaeaceae. However, they too can be inter- Nymphaeales deserve special attention because interpreta- preted in terms of an underlying racemose pattern. In V ictoria tion of their infl orescence morphology is both particularly con- and Euryale the fl owers arise in a Fibonacci spiral. Each fl ower troversial and potentially relevant to ancestral conditions and is associated with a leaf, but it is located not in the middle of the early trends in angiosperms. Cutter (1957a , b , 1959 , 1961 ) de- leaf axil but rather toward the inner side, in terms of the direc- scribed Nymphaeaceae (N uphar , Nymphaea ) as having a unique tion of the spiral (anodic side; Cutter, 1961 ; Schneider et al., system of solitary fl owers borne in the same phyllotactic spiral 2003 ). Cutter (1961) described the leaves and fl owers as form- as leaves, with no subtending bracts (accepted by Schneider et ing two separate spirals, but an interpretation more consistent al., 2003 ), which she compared with conditions in ferns. This with normal angiosperm morphology may be that each fl ower is view was critiqued by Chassat (1962) , who interpreted Nympha- in the axil of a foliage leaf but slightly displaced ( Chassat, eaceae as having modifi ed racemes, with the apparent position 1962 ). This displacement might be due to the fact that both leaf of fl owers in the same spiral as leaves due to reduction of the (petiole) and fl ower (pedicel) are bulky, so an exact superposi- leaf (pherophyll) component of a leaf-bud primordium. tion would not allow enough space in the mature state. As in In a phylogenetic context, with Cabombaceae sister to Nymphaea , the abaxial tepal in Victoria develops fi rst; but the Nymphaeaceae, the interpretation of Chassat (1962) makes more fact that Victoria has a subtending leaf as well could be evi- sense because C abomba has racemes, with fl owers borne in the dence against identifi cation of the abaxial tepal in N ymphaea axils of peltate fl oating leaves ( Brasenia has not been studied in with the fl oral subtending bract. Because each pherophyll de- suffi cient detail for comparisons). It would also bring Nympha- velops into a leaf and each bud develops into a fl ower, the num- eaceae in line with the normal shoot organization in angiosperms ber of fl owers and leaves in a shoot is the same. In contrast, in and other seed plants. Closer examination of infl orescence and Nuphar and N ymphaea only the leaf or only the fl ower of the fl oral morphology in Nymphaeaceae supports this view. N uphar , pherophyll/fl ower “ complex ” develops to maturity, and the which is basal in Nymphaeaceae, has a bract near the base of the numbers of mature fl owers and leaves in a shoot are not neces- pedicel, on its abaxial side with respect to the main axis and thus sarily equal. If V ictoria and Euryale are nested in Nymphaea , in near the position of a subtending bract, and three outer tepals. which the subtending leaf is absent, their condition may repre- Nymphaea , however, has no bract on the pedicel and four outer sent a “ reactivation ” of the pherophyll portion of the leaf-bud tepals, with the fi rst-formed tepal abaxial relative to the main primordium, perhaps related to even more extreme gigantism. axis, like the bract in Nuphar . As discussed by Chassat (1962) , In Hydatellaceae, interpretation of the crowded infl ores- this structure might be derived from that in N uphar either by cences of extremely simple fl owers is made diffi cult by the lack complete reduction of the subtending bract or by its incorpora- of subtending bracts for the lateral branches. However, Rudall tion into the perianth as the abaxial tepal. The latter hypothesis et al. (2007) tentatively but plausibly interpreted the fl owers as would explain the change from trimerous to tetramerous organi- forming reduced thyrses. zation of the perianth. On the other hand, the earlier development Based on the inferred phylogenetic relationships, racemes of the abaxial tepal could be a function of the fact that the fl ower are ancestral in Nymphaeales, either as a synapomorphy or a is more developed on the abaxial side at the time the tepals are retention from the fi rst angiosperms. With bracts present in initiated and somewhat incurved. Cutter (1957b) also homolo- Cabomba and Nuphar , it is most parsimonious to assume that gized the bract in Nuphar with the abaxial tepal in Nymphaea , bracts were lost independently on the line to Hydatellaceae and noting cases in Nymphaea in which this tepal is displaced toward Archaefructus (if these two taxa form a clade) and within
Description: