Abstract

For over a century, theories of Angiosperm origin have been framed within the assumption of seed plant monophyly. In addition, woody growth from a tubular cambium has been assumed to be the shared primitive state among seed plants. It followed that primitive woody dicots were seen as the living descendants of early angiosperms. However, if we imagine polyphyly of both the seed and woody growth among seed plants, it becomes possible to hypothesize early angiosperms as having arisen directly from pteridophytes--with monocots the living descendants of early flowering plants. We argue that a Monocots-First scenario has greater explanatory power than does the conventional Dicots-First scenario. Because the two scenarios are so different, genomic data should be able to determine which is more likely to be correct.

Introduction

Experimental analysis rarely offers resolution to questions in historical evolutionary biology. Instead, we create historical scenarios that attempt to explain contemporary diversity. We test these scenarios with fossil, developmental and comparative data. Over the last one-hundred years, a single uncontested scenario has informed flowering plant origins. Unfortunately, in such an instance we find only two classes of data: data that are meaningful within the accepted paradigm, and data that are meaningless. Contradictory data require a contradicting framework in which such data find their meaning. Currently, acceptable scenarios for angiosperm origin are all situated within the single paradigm of seed plant (Spermatophyta) monophyly (Crane 1985, Doyle & Donoghue 1986, Loconte & Stevenson 1990, Nixon et al. 1994, Rothwell & Serbert 1994, Loconte 1996, Niklas 1997, P. Soltis & Soltis 2004, Hilton & Bateman 2006, Cantino et al. 2007, Frohlich & Chase 2007, APG III 2009, Stuessy 2010).

Nevertheless, and despite more than a century of research, we still have not identified a generally accepted sister group for angiosperms, either in the fossil record or in molecular analyses of the living flora. Frohlich and Chase (2007) have commented on this enigma in a short discussion titled: "After a dozen years of progress the origin of angiosperms is still a great mystery." These authors opine that "Some extinct 'gymnosperm' groups must be closely related to angiosperms." What they fail to make explicit is that this view is embedded within the assumption that all living seed plants are members of the same monophyletic assemblage. Seed plant monophyly accords with William of Occam's teaching: simple explanations are to be preferred over more intricate explanations. The seed-syndrome, with its small but complex female gametophyte, enclosed within an ovule, fertilized by a highly reduced male gametophyte--the pollen grain--seems unlikely to have evolved more than once. Reproduction by seeds, together with woody growth and elaborate branching, characterize gymnosperms and many dicotlyedonous flowering plants. Bound together by such complex character traits, woody seed plants have been accepted as a perfectly natural assemblage for more than a century (Eames 1911, Sinnott & Bailey 1914a, Jeffries 1917, Metcalf & Chalk 1950, Stebbins 1965, Crane 1985, Doyle & Donoghue 1986, Rothwell & Serbet 1994, Cantino et al., 2007).

Sharing seeds, pollen, and woody growth, flowering plants were easily incorporated within the conceptual framework of a monophyletic Spermatophyta. With their seeds neatly enclosed within a nurturing and protective ovary, using double fertilization to initiate the formation of endosperm, and stamens with two pairs of pollen sacs, angiosperms are easily distinguished from all other seed plants (Crane 1985, P. Soltis & Soltis 2004). Because of their many woody growth forms, the dicotyledonous angiosperms (dicots) were considered to be linked with other seed plants, while morphologically simpler monocotlyedons (monocots) were seen as a clade arising from within early dicots (Sinnott & Bailey 1914b, Cronquist 1968, p. 128, Takhtajan 1969 p. 108, Stebbins 1974, p. 323, Dahlgren & Clifford 1982), Dahlgren & Rasmussen 1983, Loconte 1996, Chase 2004, Kim et al. 2004a, P. Soltis & Soltis 2004).
 
Campbell (1911, pp. 146, 162) concluded that seed plants were polyphyletic, and he commented on the near-identity of embryo structure and early development between Isoetes, Ophioglossum and many monocotyledons. Later, Campbell (1930) discussed a variety of possible origins for angiosperms but reached no conclusion. In more recent time, and with seed plants generally accepted as monophyletic, the possible origin of angiosperms from non-seed plants has received only scant attention (Burger 1981, Jacques-Felix 1988, Kato 1990, Raju & Nambudiri 1995).

Materials and Methods

We downloaded DNA sequence data from GenBank for five gene regions: 18S, atp1, atbB, matK, and rbcL. The total data set comprised 6307 characters and 110 OTUs. The 110 species of vascular plants included 6 pteridophytes, 10 gymnosperms and 94 non-Eudicot angiosperms. Core alignments were obtained and additional sequences were added to the alignments using ClustalX (Thompson et al. 1994). Improvements for variable regions were performed using Sate (Liu et al. 2009). The concatenated alignment was visualized in BioEdit v. 7.0.1 (Hall 1999) where ambiguous regions were excluded. Topological incongruencies were checked in single gene data sets by comparing reciprocal 70% neighbor-joining bootstrap support. These conflicts were considered significant when a clade was supported as being monophyletic in one tree but supported as non-monophyletic in the other tree. Gene trees for the test of incongruencies were obtained by using neighbor-joining with maximum-likelihood distances in PAUP v4b10 (Swofford 1993) and 500 bootstrap replications following Miadlikowska et al. (2006).

SplitsTree V.4 (Huson & Bryant 2006) was used to calculate phylogenetic networks and to visualize the conflict of phylogenetic signal in the combined data set. We used the neighborhood algorithm (Bryant & Moulton 2004) to infer the phylogenetic network. Conflicts in the phylogenetic signal are shown as net-like relationships, while the graph has a tree-like structure when the phylogenetic signal is stronger than contradicting evidence.

Five different analyses of the 110 taxa data set were also performed. ME-GTR-G is a minimum evolution tree employing a general-time-reversible substitution model. ME-LogDet is a minimum evolution tree employing LogDet transformation. RAX-ML is a maximum likelihood program with a gamma shape and estimation of invariable sites. Garli-ML is a maximum likelihood analysis using Garli which employs a general-time-reversible substitution mode with a gamma shape and estimation of invariable sites.  The nhPhyML analysis used the nhPhyML program with a starting tree obtained from the ME-LogDet analysis, allowing for biases in GC content and long branches.

Results and Discussion

The concatenated five-gene analysis (figure 1 [see PDF below]) shows long branches separating angiosperms from gymnosperms and ferns. Long branches are well-known to cause problems in elucidating phylogenetic relationships and this may be a cause for the uncertainty in indetifiying the basal lineage of angiosperms in our analyses and may explain conflicting results in previous studies. Further, the relationships among all major groups show strong evidence for conflicting phylogenetic signals with virtually all relationships showing box-like relationships. Among the five other analyses incorporating the same five gene regions, two (RaxML and Garli-ML) placed Amborella as basal among angiosperms, one (nhPhyML) placed Ceratophyllum as basal, and two (ME-GTR-G and ME-LogDet) designated monocotyledons as basal. The different topologies yielding from different methods used suggest that the molecular data analyzed are not sufficient to reject any of the obtained topologies. On this account, we conclude that a basal placement of monocotyledons cannot be rejected, and propose an evolutionary scenario based on this possibility. In doing so, we set aside the assumption that woody seed plants are monophyletic, and argue that the earliest angiosperms were, in fact, herbaceous monocot-like plants.

A Monocots-First scenario: an alternative perspective

A Dicots-First scenario for angiosperm origins has been widely accepted for more than a hundred years. It is concordant with the assumption that the seed, the pollen grain and wood (produced from a tubular bifacial cambium) are shared synapomorphies. Dicots and some gymnosperms also share similarities in embryo structure, forming two or more cotyledons in early embryogeny, and the embryonic root can grow into a taproot. Overall, living gymnosperms and angiosperms appeared to be a natural, monophyletic, assemblage. It followed logically that the largely herbaceous monocots had arisen from woody dicot-like ancestors (Coulter & Chamberlain 1903, Sargant 1903, Henslow 1911, Sinnott & Bailey 1914b, Cronquist 1968, pp. 315-320, Takhtajan 1969, pp. 108-121, Stebbins 1965, Stebbins 1974, pp. 317-324, Guignard 1983). This Dicots-First scenario continues to serve as the sole theoretical framework in discussions of angiosperm evolution. But is it possible to construct an alternate scenario, also capable of bringing diverse facts together within a coherent historical narrative? We believe, a Monocots-First scenario can achieve this goal in a convincing manner. However, a Monocots-First hypothesis abandons the assumption of seed-plant monophyly, and denies a close sister-group relationship between angiosperms and other living seed plants. By viewing angiosperms as having arisen independently from a pteridophytic source, the Monocots-First perspective can explain many aspects of angiosperm diversification, beginning with the early embryo itself.

Early Embryogeny . Both the form of the early embryo and early plant development are very similar in monocots and in living pteridophytes (Campbell 1930, Jacques-Felix 1988, Vallade, Bugnon & Ibannain 1993, Raju & Nambudiri 1995). Having a terminal first leaf at the apex of the embryo, the shoot apex lateral to the axis of the embryo, and a short-lived primary root that is replaced by secondary roots are an unusual array of characteristics shared by living pteridophytes and monocots. In monocots the base of the cotyledon "often remains in contact with the endosperm as a digestive and absorbing organ very suggestive of the 'foot' of Pteridophytes" (Coulter & Chamberlain 1903, p.6; Vallade, Bugnon & Ibannian 1993). In both Ophioglossum and monocots root inception and development take place well before the development of the shoot apex, and the early leaves are clasping at their base (Campbell 1911). These similarities involving both form and development suggest evolutionary homology; they seem unlikely to have arisen independently through convergence. It follows that monocots do not possess a true cotyledon or seed-leaf (Jacques-Felix 1988). Their cotyledon is simply the first leaf of the plantlet, just as it is in living pteridophytes.

The cotyledon of monocots is usually of the same form and in the same phyllotactic spiral as are the succeeding leaves. This is characteristic of a reiterating series of homologous structures, and very different from the early growth of dicots. Chemical treatments do not alter monocot development as is the case in some dicots (Haccius & Lakshmanan 1967). Aberrant embryos are more common in dicots than in monocots (Eames 1961, p. 360, Palser 1975). Reports of dicotyledonous monocots supported the interpretation that monocotyledony within angiosperms was derived (Solms-Laubach 1878, Coulter & Land 1914, Eames 1961, pp. 328, 343, 345, & 360). However, these citations of unusual embryos have not been reinforced by subsequent studies. Excepting for these few reports, the monocot embryo appears to be conservative in form, resistant to chemical alteration, and remarkably consistent over a wide range of genera (Tillich 1992).

The similarity of embryo form and early embryogeny between monocots and pteridophytes has not been a subject of serious discussion for almost a century. Viewed from the perspective of having flowering plants deeply imbedded within a monophyletic Spermatophyta, embryonic similarities between monocots and pteridophytes can be said to "have no phylogenetic meaning." Textbooks separate discussions of seed plants and non-seed plants by many pages, allowing for little detailed comparison. From the Monocots-First perspective, similarity in early growth provides salient evidence for the likelihood that flowering plants arose directly from a pteridophytic base, and are a lineage separate and distinct from other living seed plants.

The recent transfer of Trithuria (Hydatellaceae) to Nymphaeales and close to Amborella is noteworthy (Rudall et al. 2007, Saarela et al. 2007, APG III 2009, Rudall et al. 2009b). These small aquatic sedge-like plants had been placed within Centrolepidaceae (Hutchinson 1959, p. 700, Melchior 1964, vol. 2: 559), as a distinctive order near the Cyperales (Cronquist 1981, p. 1148) or Restionales (Dahlgren & Clifford 1982, p. 36), near Xyridaceae (Davis et al. 2004) and as a member of Poales (Chase et al. 2006). Seed-provisioning prior to fertilization in these plants, and seed-nutrition largely supplied by perisperm (Hamann 1976, Friedman 2008), appear to confirm their primitive nature. More importantly, Trithuria possesses a typical monocot embryo (Tillich, Tuckett & Facher 2007), thus placing monocotyledony near the base of current phylogenies.

Viewed from a Monocots-First perspective, the dicotyledonous embryo marked a major advance in angiosperm morphology. Cotyledons of dicots are usually of different form and in a different phyllotaxy than are the succeeding leaves of the same plantlet; most have petioles. Differences in placement, form and arrangement indicate that the cotyledons of monocots and dicots are not homologous structures (Burger 1998). This is a deep and profound morphological distinction, used for over two centuries as the primary dichotomy among flowering plants. Similarity of the dicotyledonous embryo with the embryos of certain other seed plants have been used to support seed-plant monophyly (Eames 1961), but such similarities are convergent from a Monocots-First perspective.

Early Plant Form . The Monocots-First scenario implies that all early angiosperms, including the earliest dicotyledons, were herbaceous or weak-stemmed plants (Burger 1981). This view is similar to that of Taylor and Hickey (1990, 1992, 1996) who contended that the earliest angiosperm was a "rhizomatous to scrambling herb." The Monocots-First scenario differs in viewing early angiosperms as lacking secondary growth entirely. This is consistent with Tomlinson's claim (1995) that the earliest monocots lacked a cambium. Rather than a "loss" of stelar organization in monocots, our scenario posits a major advance in dicots. And because soft-stemmed and herbaceous plants are unlikely to be fossilized, an herbaceous ancestry helps explain the paucity of early angiosperm fossils (Taylor & Hickey 1992, 1996).

In basal ferns, such as Ophioglossales, the persistent shoot of the young sporophyte arises as an adventitious bud on the ephemeral primary root and the vasculature of the stem is largely, if not entirely foliar in origin (Campbell 1911, 1930, Mesler, Thomas & Bruce 1975). Such a growth pattern might have characterized early angiosperms. Carlquist (1975, p. 75) noted that in ferns all mechanical supporting structures arise from the primary plant body and that monocots, characterized by a similar growth pattern, are limited to a similar range of growth forms. In addition, vessel occurrence in ferns, though quite rare, resembles the roots-first pattern found in monocots (Fahn 1954). Similarly, Sarcandra (Chloranthaceae) has vessels in the secondary xylem of its roots but not in the secondary xylem of its stems, as is characteristic of monocots (Carlquist 1974, 1996, p. 79). To find this monocot-like pattern in a basal dicot lineage accords well with a Monocots-First scenario.

Double fertilization is a unique angiosperm trait, allowing both rapid and efficient seed production. Gymnosperms often produce their seeds in advance of fertilization and, when pollination fails, bear seeds that cannot sprout. Unfertilized energy-laden seeds are a cost small short-lived plants cannot sustain (Doyle & Hickey 1976). Double fertilization eliminates the possibility of producing unfertilized seeds, as well as reducing the time between germination and reproduction (Taylor & Hickey 1992, 1996). Double fertilization allowed small herbaceous angiosperms to succeed as seed plants, and this trait is fundamental to a Monocots-First scenario. As Stebbins (1965) remarked: "... selective pressures which brought about the angiospermous reproductive condition would have acted more strongly on smaller, short lived plants ..." Contrary to Friedman (1992) and in agreement with Niklas (1997, p. 208) and Stuessy (2004), we view intra-archegonial double fertilization in Ephedra as an unrelated convergence.

Early Leaf Form. The Monocots-First scenario implies that simple undifferentiated leaves with clasping leaf-base were the primitive condition in early angiosperms. Clasping or sheathing leaf-bases are the widely shared condition within monocots, and common within herbaceous dicots as well. Early ontogeny supports the primitive nature of this character state: all angiosperm leaf-primordia begin with a flanking base (Hagemann 1970, Ihlenfeld 1971). Carpels are recognized as leaves because they share this same initial form (Eames 1961, p. 227, van Heel 1983). In addition, the petiole is usually the last part of the leaf to differentiate during leaf development in angiosperms (Majumdar 1956, Richards 1983). These ontogenetic patterns support the claim that clasping leaf bases, poorly differentiated petioles, and linear to simply expanded blades with parallel or subparallel venation are primitive states. Reviewing gene expression patterns, Floyd and Bowman (2010) conclude: "While the ancestral seed plant leaf was very likely to be complex, the ancestral angiosperm leaf is predicted to be simple, implying that complex angiosperm leaves are derived."

The claim that the lamina is formed from the upper primordial area in dicots and from the lower area in monocots (Keating 2002, p. 11, Kidner et al. 2002) appears to be an oversimplification. Laterally expanded and petiolate leaves in Saururaceae (Piperales) and Alismatales are very similar in both development and mature form (Rudall & Buzgo 2002). It is very likely that leaves with broad laminae and reticulate venation evolved along parallel tracks but independently in several lineages of monocots and early dicots (Bharathan 1996, Chase 2004). More telling, linear leaves precede highly differentiated petiolate leaves with broad lamina in the early development of species of Potamogeton (Esenbeck 1914), Sagittaria (Bloedel & Hirsch 1979), Eichornia (Richards 1983), and Victoria (Valla & Martin 1976); the latter is a dicot. In some species of Piper (Piperaceae) the early basal leaves have broadly sheathing monocot-like petioles, while distal leaves have more narrowly terete dicot-like petioles on the same plant. Each of these heteroblastic series is concordant with the Monocots-First scenario; they are not what we would expect if early angiosperms resembled Amborella.

Ontogenetic evidence also supports the notion that simple leaves with multiple parallel vascular bundles preceeded petiolate laminae with a single midrib. Ontogeny of Populus leaves indicate that the midrib "is a composite structure consisting of several independent vascular bundles, each of which eventually diverges into the lamina to become a secondary vein" (Isebrands & Larson 1980). From a Monocots-First perspective, differentiated petioles and broad thin laminae were later evolutionary innovations. A slender and flexible petiole allows the blade to twist under the stresses of high winds, and permits the blade to be oriented to maximize surface illumination. An abscision layer at the base of the deciduous dicot leaf was another significant advance; possible only after modification of the clasping leaf-base. Small, thin, readily deciduous leaves became especially important after dicots evolved a woody and highly branched architecture. This helps explain "...the apparent  rapidity with which modern-looking angiosperm leaves entered the fossil record..." (Friis, Pederson & Crane 2010). Later, high vein-densities in angiosperm leaves have allowed for higher transpiration rates, increasing both photosynthesis and precipitation in tropical rain forests (Boyce et al. 2009, Boyce & Lee 2010).

Stem structure . A Monocots-First scenario implies that the earliest angiosperms either lacked stems, or possessed herbaceous stems formed by the union of leaf bases. In this scenario stems with scattered vascular bundles are the direct result of having been formed from the fusion of leaf bases. In many monocots leaf traces arise from near the periphery of the stem, bend inward and upward to occupy the center of the stem, then abruptly bend outward to enter the leaf-base (Zimmerman & Tomlinson 1965, Tomlinson 1995). This pattern supports the inference of stems having evolved from united leaf bases in monocots. The absence of a bifacial cambium, and the paucity of complex branching are further consequences of their simple origins. This is consistent with Charles Bessey's view (1897 p. 163) that in monocots "... leaf and stem are not yet as fully differentiated as they are in dicotyledons." Also, with usually only a single leaf at each node, leaves are alternate in the vast majority of monocots and growth is monopodial; further evidence for a very simple Bauplan (Arber 1925, p. 179, Tomlinson 1970, 1995).

Because monophyly of woody seed plants appeared to be so parsimonious, a consensus developed viewing archaic angiosperms as woody trees or shrubs, with all herbaceous forms later derivations from woody ancestors (Hallier 1905, 1912, Eames 1911, Sinnott & Bailey 1914b, Bessey 1915, Jeffrey 1917, Stebbins 1965, Cronquist 1968, p. 128, Takhtajan 1969 p. 108, Stebbins 1974, p. 323, Dahlgren & Rasmussen 1983, P. Soltis & Soltis 2004). This consensus required a subsidiary hypothesis to explain how herbaceous lineages 'lost' both the ability to form a wood-building cambium and the genetic programs needed to construct a more complex, many-branched, architecture. It was generally thought that monocots had gone through an evolutionary bottleneck, probably as aquatics (Hallier 1903, Henslow 1911, Sinnott & Bailey 1914b, Cronquist 1968, p. 317, Les & Schneider 1995, Chase 2004, Qui et al. 2006). Moreover, to explain the mechanism of loss, Takhtajan (1959, 1976) suggested a neotenic process. In neoteny early stages of development are carried forward into the reproductive stages of the plant, as later stages are lost. In this way the ability to build a cambium and woody structures were lost in some lineages, producing short-lived herbs. "The herb may be considered a fixed juvenile phase of the tree" Takhtajan concluded (1991, p. 23). In contrast, a Monocots-First scenario requires neither an aquatic bottleneck nor neotenic reduction to explain herbaceous growth in angiosperms.

Tomlinson (1995) has argued that "...the absence of bidirectional (bifacial) cambium is ancestral in monocotyledons, and there is no homology of organization between the primary vascular systems of monocotyledons and dicotyledons." A monocots-first perspective accommodates this view by postulating the later development of radically new vascular organization in early dicots; a vasculature so different that it is now seen as non-homologous. The instantiation of dicotyledony may have been the critical precursor allowing anatomical restructuring in early dicot stems. This restructuring allowed for the development of a bifacial cambium, wood, and more varied branching. Enhanced axillary budding allowed early dicots to recover more quickly from browsing and trampling (Feild et al. 2004, Avens 2010).

In dicots multilacunar nodes are always associated with sheathing leaf bases (Sinnott & Bailey 1914a, Majumdar and Pal 1961). Howard (1974) suggested that "the multilacunar node of modern [flowering] plants with a sheathing stipule may well be the residual expression of the primitive nodal plexus." A transition from herbaceous ancestry with scattered vascular bundles to multilacunar nodes and, finally, tubular woody growth, conforms well with the Monocots-First scenario, and is consistent with interpreting the vascular system of angiopsperms as having a foliar rather than axial origin (Nast 1944, Philipson & Balfour 1963, Esau 1965). Also, the question of why vascular bundles should have become scattered has no simple answer in a Dicots-First scenario beginning with a woody ancestor having a tubular siphonostele (Eames 1911, Slade 1971, Beck, Schmid & Rothwell 1982). Hypotheses of 'medullation' were developed to explain the origin of multilacuar nodes and scattered vascular bundles in which central medullary (parenchyma) tissues invade and expand, forming gaps or lacunae in the vascular cylinder (Sinnott & Bailey 1914b, Metcalf & Chalk 1950, p. xxxv). Medulation is an unnecessary hypothesis within a Monocots-First scenario.

Secondary Growth and Wood . The Monocots-First perspective implies that secondary growth is a later development in angiosperm evolution. Likewise, more versatile axillary growth, and new modes of branching allowed dicots to produce a grand variety of forms. For angiosperms, the bifacial cambium was a new innovation, but the difficulty of having a tubular vasculature traverse the node was a problem for early dicots. This is seen today in woody Chloranthaceae and Piperaceae where the nodes are conspicuously swollen. More advanced dicot lineages solved the problem of bringing the tubular vasculature through the nodal plexus, with the result that their nodes are not noticeably distended. From the Monocots-First perspective, dicot paleoherbs, such as Nymphaeales and most Piperales, never "lost" their ability to make wood; they are simply the living descendants of early pre-woody dicots. Evidence for polyphylly of vessel-formation in early angiosperms (Bailey 1944, Cheadle 1953, Carlquist 1987, 1996, Schneider & Carlquist 1996) is concordant with a more recent elaboration of woody tissues. In contrast to the uniformity of conifer woods, innovations in tracheary elements and xylem parenchyma interlinkages were key drivers in dicot diversification (Carlquist 2009).

More fundamentally, Wilson and Knoll (2010) claim that a vascular cambium able to initialize both strong structural fibers and efficient water-conducting vessels was the key innovation allowing angiosperms to produce so great a variety of woody forms. A Monocots-First scenario places this innovation within early dicots. Fossil angiosperm wood has not been found in strata earlier than about 112 million-years-ago (Philippe et al. 2008). Even if angiosperms had a highland origin (Sinnott 1916, Stebbins 1974), their wood should have been fossilized in riverine deposits. From the Monocots-First perspective, the lack of angiosperm wood in the early Cretaceous is the simple consequence of their herbaceous origin. Perhaps the radiation of herbivorous low-browsing dinosaurs during the late Jurassic (Bakker 1978) was a consequence of monocot expansion at that earlier time. Later, by the upper Cretaceous, the expansion of angiosperm-dominated forests coincides with rosid diversification (Wang et al. 2009, D. Soltis et al. 2010).

The Monocots-First scenario also helps explain the unusual diversity of wood anatomy in early dicotyledonous lineages. The variety of stem structures and wood anatomy within basal dicots ranges far beyond that found in living gymnosperms (Benzing 1967). Six distinctive structural types of secondary xylem have been identified in early-diverging dicots (Gottwald 1977). Carlquist (1987) observed that "... vessel specialization has happened in a highly polyphyletic manner in dicotyledons." Such anatomical variety is not something one would expect in a lineage having had a woody ancestry. This grand diversity in dicot wood anatomy and growth habit is more likely the consequence of new anatomical innovations within related lineages that were radiating on the basis of these same new capabilities. From a Monocots-First perspective, similarities in the wood of Amborella, Winteraceae, and conifers (Evert 2006, p. 302) are superficial and convergent.

More significantly, the Monocots-First scenario addresses the issue of "Darwin's abominable mystery." Though this mystery is generally characterized as the inability to identify angiosperm ancestors, Friedman (2009) argues that, for Darwin himself, the abominable mystery was the "astonishingly sudden appearance" of angiosperms in the mid-Cretaceous. Because woody plants fossilize much more readily than weak-stemmed herbs (Herendeen & Crane 1995), this sudden appearance is documented largely by woody lineages. In our scenario several dicot lineages probably developed the woody habit independently not long after dicot origins, giving rise to the sudden appearance of new and distinct woody lineages. This is in agreement with an interpretation of the chloroplast genome that woody Laurales "emerged after the split of angiosperms into monocots and dicots" (Goremykin et al. 2003a). The Dicots-First scenario provides no insight into why so many different woody lineages should have radiated at this point in time.

Flowers . Our Monocots-First scenario does not address floral evolution directly. Angiosperms are distinguished by two very unusual synapomorphies. First: leaf-like carpels, enclosing the ovule or ovules, form the ovary in which seeds develop. Second: double-fertilization initiates endosperm formation and seed development. These distinctive traits are sufficient to distinguish angiosperms from all other lineages. The most complete early Cretaceous fossil, Archaefructus, bears simple terminal axes with paired stamens proximally and solitary or clustered carpels distally, without subtending bracts or perianth (Sun et al. 2002). Despite its early date of 124 million-years-ago, this simple arrangement has been interpreted as an example of "evolutionary reduction" from a more conventional flower (Friis et al. 2003, Friis, Pedersen & Crane, 2006, Friis & Crane 2007, Doyle 2008, Endress & Doyle 2009). A more likely scenario is that carpels, stamens and their subtending bracts or enclosing perianth have come together to form flower-like configurations independently in different lineages, and that the earliest angiosperms did not posses well-organized flowers, exactly as is the case in Archaefructus (Stuessy 2004). In fact, the stamen-pairs of Archaefructus may be homologous to the paired stamens seen in the early ontogeny of many monocots and Piperales (Burger 1977, Kaul 1993).

Likewise, simple reproductive units in Zannichelia, Lilaea, Peperomia and Chloranthus may not have been "reduced from earlier flowers," as is generally thought (Endress & Doyle 2009). The reproductive units of Trithuria can be interpreted as prefloral (Rudall et al. 2009a). Similarly, transitions between flowers and inflorescence in the distal spikes of Triglochin (Lieu 1979, Charlton 1981) and in Cyperaceae subfamily Mapanioideae (Richards, Bruhl & Wilson 2006.) may reflect the ambiguity of early floral construction. More fundamentally, flowers are formed by differing gene regulatory networks in monocots, basal dicots, and core eudicots (Johansen, Frederiksen & Skipper 2006, Liu et al. 2010). Pentamerous flowers also appear to have evolved independently several times in Eudicots (Gonzalez & Bello 2009). In all likelihood, what we call flowers are later evolutionary contrivances, evolved to protect and display the reproductive organs. This view echoes Stuessy's (2004) argument that the evolution of flowers came late in angiosperm history, and contradicts the claim of P. Soltis et al. (2009) that "The unifying feature of angiosperms is the flower..." More importantly, if angiosperms evolved apart from other living seed plants, similarities of reproductive structures in other seed plants are likely to be independently convergent.

Age and Diversification . A few molecular studies indicate that within-angiosperm splitting may date back well over 200 million years (Martin, Gieri & Saedler 1989, Troitsky et al. 1991, Martin et al. 1993, Savard et al. 1994, Kim et al. 2004b). Other studies give evidence for deep phylogenetic branching within monocots (Goremykin et al. 2004). However, Chaw et al. (2004) claim that constant-rate methods for estimating monocot-dicot divergence from chloroplast genomes overestimate the time when "fast-evolving monocots are included." Accordingly, these authors excluded monocot lineages to give "...estimates consistent with the known evolutionary sequence of seed plant lineages..."

Because of their limited morphological diversity, deep divisions within the monocots are little appreciated. Not only are there deep divisions, monocots lack any central phylogenetic core as is found in dicots. The hugely diverse Eudicots are clearly linked to the Ranunculales and a few related clades (Soltis et al. 2000, Hilu et al. 2003, Anderson, Bremer & Friis 2005, Qiu et al. 2006, Hansen et al., 2007, APG III 2009, D. Soltis et al., 2010). This is very different from the situation among monocots. Takhtajan (1969, p. 121) pointed out that "...there is no monocot order that could occupy the place in their family tree that is occupied by the Magnoliales among dicots." Similarly, Cronquist (1981, p. 1032) noted that within monocots "No one of the subclasses can be considered basal to the others." Along the same lines, Cronquist enumerated six subclasses within the dicots and five subclasses within monocots, despite dicots having four times as many species as monocots. More recently and using DNA sequences, Chase et al. (1995) arranged monocots into six major clades, while Davis et al. (2004) identified 15 mutually exclusive groups within monocots. Relationships among many of these clades remain weakly supported (Givnish et al. 2006). Petrosaviaceae, comprising two small genera, one achlorophyllous, are now seen as a very isolated order, not closely related to other monocots (Cameron, Chase & Rudall 2003, Graham et al. 2006, Tobe & Takahashi 2009). A long history may be a plausible explanation why monocots are so deeply divided, and why they lack the core-clustering seen in dicots.

There is additional support for a long monocot history. Evidence for sea grass communities is found in the Late Cretaceous (Brasier 1975). The only land plants to have reverted to a submerged marine environment, one can imagine that such an unusual adaptation required a long period of time. Phytoliths characteristic of differing lineages of grasses (Poacae) have been found in late Cretaceous dinosaur coprolites (Prasad et al, 2005), while bambusoid fragments have been reported in 100 million-year-old Burmese amber (Poinar 2004). Fossil flowers of the highly derived Triuridaceae have been found in 90 million-year-old strata (Gandolfo, Nixon & Crepet 2002). Paleoherb-like angiosperm pollen from the middle Triassic (Hochuli & Burkhardt 2004, Zavada 2007), and the poorly preserved but monocot-like Sanmiguelia (Cornet 1989) are also noteworthy. Biogeography also implies great age within monocots. The Strelitziacae (Zingiberales), with highly derived flowers and large distichous leaves, include three genera of restricted range: Phenakospermum in Amazonia, Strelitzia in South Africa, and Ravenala in Madagascar. Such Gondwanan disjunctions suggest an early origin for so highly derived a family (cf. Bremer & Janssen 2006).

Other data lend support to a basal position for monocots. Apocarpy--where more than one unicarpellate pistil is found within the same flower--appears to be an ancient trait within angiosperms. Apocarpy is restricted to magnolids, ranunculids and rosids among dicots, but apocarpy occurs in Arecaceae (palms), Alismatales (aquatics), Petrosaviaceae and Triuridaceae among monocots; a far more widely scattered distribution. (These apocarpous gynoecia are interpreted as independently derived by Remizowa, Sokoloff & Rudall, 2010.) Similarly, pteridophyte-type starch grains were found in 14 monocot families but only in Gunneraceae among dicots (Czaja 1978). This study revealed a progression in which the monocots generally had the widely shared states, the dicots generally had the most derived states, while Piperales and Nymphaeales were intermediate.

Gymnosperms and Angiosperms . Rooting angiosperms outside living gymnosperms implies a long and separate history. Some DNA studies have suggested a division as early as 300 million-years-ago (Martin et al. 1993, Goremykin, Hansmann & Martin 1997, Wang et al. 2007, Wu et al. 2007). So early a separation makes an independent acquisition of the seed habit in angiosperms more likely. In addition, many attributes of living gymnosperms contradict any close relationship with angiosperms. In gymnosperms pollination places pollen in contact with the ovule; in angiosperms pollination places pollen on the receptive surface of the carpel. The chemical nature of the pollen tube wall is very different in gymnosperms and angiosperms (Williams 2008). These differences are not what we would expect if pollination were a shared synapomorphy.

Gymnosperm lignins are derived primarily from guaiacyl monomers, whereas angiosperm lignins are guaiacyl/syringyl copolymers (Weng et al. 2008). The discrete shoot and root promoting WUS/WOX5 gene functions are unique to angiosperms and unlike development in gymnosperms (Nardmann, Reisewitz & Werr 2009). Angiosperms build their reproductive parts with only one copy of the Floricaula/LEAFY homeotic gene, while gymnosperms use two (Frohlich & Parker 2000). Similarly, angiosperms lack one of the three class III HD-Zip patterning genes found in gymnosperms (Floyd, Zalewski & Bowman 2006). Likewise, synonymous substitution ratios between chloroplast and mitochondrial genes are very different in gymnosperms and angiosperms (Drouin, Daoud & Xia 2008). In addition, gymnosperms do not display similar levels of polyploidy, or so grand a variety of chromosome rearrangements as do angiosperms (Ehrendorfer 1976, Leitch & Leitch 2008). In contrast, angiosperms and pteridophytes share a grand variety of ploidy levels (Hill & Crane 1982, Kejnovsky et al. 2009), while heterosporous ferns and angiosperms exhibit similar average chromosome numbers (Barker & Wolf 2010).  

Wood anatomy and stem development are only superficially similar in gymnosperms and angiosperms; xylem pitting is fundamentally different (Carlquist 1996). In angiosperm seedlings the early epicotylar protoxylem originates in the base of the leaf primordia and progresses basipetally (downward) to form the central protoxylem of the stem (Benzing 1967, Larson 1976). Early xylem differentiation is very different in conifer seedlings where the first epicotylar xylem elements are located just above the hypocotyl and mature acropetally toward the expanding primary leaves and basipetally into the hypocotyl (Tilton & Palser 1976). In addition, angiosperm leaf-traces usually originate from one to many nodes below the leaf they will innervate, whereas gymnosperm leaf-traces usually originate directly below the leaf (Devadas & Beck 1972). In conifers tension wood (reaction wood) forms under the area of stress, but in angiosperms it forms over the area of stress, resulting in differing tree architectures (cf. Heinrich & Gartner 2008). Resin chemistry is distinctly different in conifers and angiosperms (Bray & Anderson 2009). Taken together, these observations support the contention that wood has evolved independently in angiosperms.

A Monocots-First scenario helps us understand why so many attempts to relate angiosperms to extinct fossil forms such as the seed ferns have failed to be convincing (Taylor & Taylor 2009). These lineages are found as fossils because all were woody plants with tough foliage. Comparative studies seeking homologies between angiosperm reproductive structures and those of gnetopsids (Hickey & Taylor 1996) or other seed plants (Frohlich & Parker 2000, Doyle 2008, Specht & Bartlett 2009) are appropriate within a monophyletic Spermatophyta and a Dicots-First paradigm; but they are inappropriate from the Monocots-First perspective which roots the angiosperms outside other seed plants. An ancient and separate origin may explain why Hilton and Bateman (2006) found angiosperms as the basalmost lineage among living seed plants when using the fossil progymnosperm Cercropsis as the sole outgroup in a cladistic analysis, and why some nucleotide data place angiosperms at the base of seed plants (Burleigh & Mathews 2004). Phylogenetic analyses that place monocots before dicots--or angiosperms near the base of seed plants--need not be artifacts of "fast evolving monocots" or "long-branch attraction;" they may be correct.  

Discussion

In his introduction to a collection of papers on angiosperm origins, Beck (1976, p.4) remarked: "... it is only logical to expect that the pro-angiosperms should exhibit a range of characteristics intermediate between those of their precursors and typical angiosperms." A Monocots-First scenario satisfies this expectation: in both embryology and general growth habit the monocots do exhibit characteristics intermediate between simple pteridophytes, such as the Ophioglossales, and early dicots, such as the Piperales. This is consistent with the opinion that "In their reliance on vegetative growth, early angiosperms may have been more like ferns than other seed plants" (Feild et al. 2004). In addition, viewing monocots as the earliest of living angiosperms has an impressive precedent. In their encyclopedic series Die Naturlichen Pflanzenfamilien, Engler and Prantl (1897-1915) placed monocots before dicots. They undoubtedly saw monocots as having simpler overall structure, as well as an embryology reminiscent of ferns and lycopods. In contrast, currently accepted thinking interprets the simplicity of monocot growth as the result of evolutionary loss and neoteny. But such explanations can also be seen as subsidiary hypotheses, bringing monocots into conformity with a Dicots-First scenario, itself embedded within a monophyletic Spermatophyta.

The Dicots-First scenario has identified major angiosperm lineages in a satisfying way and explained dicotyledonous diversification concordant with fossil evidence and DNA similarities. Thus, with dicots comprising about 80% of living angiosperm species, the Dicots-First scenario has helped us understand the phylogeny of a majority of flowering plants very well. Despite a few ambiguities in basal lineages, DNA data has identified a primitive assemblage of dicots, from which the vastly more numerous eudicots have emerged. Tricolpate pollen, first appearing in the fossil record 125 million-years-ago, marked a major advance in angiosperm evolution and correlates well with early eudicot diversification. Overall, these results have led to a satisfying consensus (P. Soltis & Soltis 2004, APG III 2009, D. Soltis et al. 2010). Accordingly, monocots are seen as one of several early-diverging angiosperm lineages, and the importance of the monocot-dicot division within angiosperms has been discarded.

Nevertheless, DNA phylogenies have recognized monocots as a firmly supported monophyletic assemblage within flowering plants (Duvall & Ervin 2004, Chase 2004, P. Soltis & Soltis 2004, Chang et al. 2005,  Mardanov 2008, D. Soltis et al. 2010). While some studies have identified the genus Acorus as a basal monocot lineage rooted to early dicots (Duvall et al. 1993, Buzgo & Endress 2000, Chase 2004, APG III 2009, Givnish et al., 2010), specific monocot-dicot connections remain unresolved (Qui et al. 2006). Despite using an expanded multigene data set, Graham et al. (2006) concluded that "the issue of monocot placement in broader angiosperm phylogeny remains problematic."

Our DNA analyses were strongly affected by the inclusion of aquatic monocot lineages. We found that when these taxa are removed from some analyses, Amborella slips into the basal position among angiosperms. Likewise, D. Soltis & Soltis (2004) found monocots basal in an analysis that included the aquatic monocot Zostera. The importance of aquatic monocots in phylogeny reconstruction should not come as a surprise. Ancient forms are more likely to survive as specialists in increasingly competitive environments over time (Vermeij 1987). Among mammals, Echidna, like Platypus, is now thought to have had a semi-aquatic history (Phillips et al. 2009). Alternatively, the earliest angiosperms may have been semi-aquatic, not unlike early insects.

Overall, we believe DNA phylogenies that place monocots basally must be taken seriously. This view is concordant with a study by Donoghue and Mathews (1998) using phytochrome sequences which found that their monocot-rooted tree entailed fewer duplications and sorting events than did trees rooted among Magnoliales. A number of other genomic analyses have also placed monocots basally (Troitsky et al. 1991, Martin et al. 1993, Bezannilla et al. 2003, Goremykin et al. 2003a, 2003b, 2004, 2005, Chung 2006). These results have been dismissed because they do not conform to current thinking (D. Soltis & Soltis 2004, APG III 2009). In contrast, we argue for a basal position for monocots and claim that the structural simplicity of monocots is both ancient and primitive. This view is consistent with the rice (Oryza) genome lacking the paleo-hexaploidy found in grape, poplar and Arabidopsis genomes (FIPCGGC 2007). A much longer history for monocot lineages may also explain why the rice genome has 328 predicted host-sensors with non-RD Kinase domains while the Arabidopsis genome has only 35 (Ronald & Beutler 2010). More generally, the Monocots-First scenario outlines increasing morphological complexity within angiosperms over evolutionary time, with little need for hypotheses of morphological reduction and loss. We believe the diagram produced by our concatenated five-gene analysis (fig. 1) reflects a long history of phylogenetic isolation for angiosperms; sufficient time for angiosperms to have accrued their distinguishing synapomorphies (Stuessy 2004).

Moreover, a direct derivation of angiosperms from pteridophytic stock through monocot-like intermediates does not discount the many positive attributes of the traditional Dicots-First scenario. Thomas Kuhn (1970, p. 169) pointed out that "... the new paradigm must promise to preserve a relatively large part of the concrete problem-solving ability that has accrued to science through its predecessors." All that a Monocots-First scenario does is to remove monocots from within the 'basal angiosperms' and place them beneath the living descendants of the earliest dicots. The Dicots-First paradigm can continue to serve its explanatory role, with modification only at its base. Many studies, such as the recent review of Early Cretaceous fossils by Doyle & Endress (2010) remain unaffected. The Archaeoangiospermae in the sense of Stuessy (2010) are simply the living descendants of early dicot diversification. The essential distinction is that dicot-origin will no longer be conflated with angiosperm-origin, and sister group relationships will no longer be sought among living or fossil gymnosperms. From a Monocots-First viewpoint, the basal position of Amborella in so many DNA phylogenies results from chance convergences among inappropriate out-groups (cf. Graham & Iles 2009). In fact, the earliest dicots should have retained many ancestral monocot features, exactly as is the case in contemporary Nymphaeales and Piperales.

The Monocots-First scenario predicts similarity in genomic control of early embryogeny in monocots and pteridophytes. In contrast, the Dicots-First scenario implies closer similarity between dicots and other seed plants. Comparative genomics in early development should provide crucial data in deciding between the two scenarios. Unfortunately, by postulating a herbaceous origin, the Monocots-First scenario does little to clarify angiosperm origin. Because weak-stemmed herbs rarely fossilize, Paleozoic or Mesozoic fossils of Ophioglossales have not been found--nor any trace of the earliest angiosperms.

More importantly, the Monocots-First scenario reaffirms the monocot-dicot division as the primary dichotomy among living angiosperms, and recognizes dicot-origin as a major and later advance in flowering plant evolution. This is consistent with two studies indicating that increase in diversification rates occurred well after key angiosperm innovations were already in place (Sanderson & Donoghue 1994, Qiu et al. 2006). Newly equipped with woody stems, diverse branching, and deciduous leaves, dicots gave rise to an explosive radiation. Innovations in dicot anatomy fostered more complex architecture among angiosperms and more richly structured environments. Growing rapidly, spending less energy on defense and more on reproduction, and enticing sentient vectors to carry pollen between isolated plants, both monocots and dicots have fostered a terrestrial world far richer in species than any that has ever come before (Burger 2006, Novacek 2007, Benton 2010, Leigh 2010).

[Acknowledgements: All the molecular analyses mentioned at the beginning of the paper and the basis for Figure 1 were made by Thorsten Lumbsch, Chairman of the Botany Department here at Field Museum. Thorsten, a Lichenologist, declined to be a coauthor. His Figure makes clear the distinction of flowering plants. The National Science Foundation funded floristic studies on the Piperales of Costa Rica (1966-1972) which were basic to the ideas presented here.]