- The Paleontological Society
Four vascular plant lineages, the ferns, sphenopsids, progymnosperms, and seed plants, evolved laminated leaves in the Paleozoic. A principal coordinate analysis of 641 leaf species from North American and European floras ranging in age from Middle Devonian through the end of the Permian shows that the clades followed parallel trajectories of evolution: each clade exhibits rapid radiation of leaf morphologies from simple (and similar) forms in the Late Devonian/Early Carboniferous to diverse, differentiated leaf forms, with strong constraint on further diversification beginning in the mid Carboniferous. Similar morphospace trajectories have been documented in studies of morphological evolution in animals; however, plant fossils present unique opportunities for understanding the developmental processes that underlie such patterns. Detailed comparison of venation in Paleozoic leaves with that of modern leaves for which developmental mechanisms are known suggests developmental interpretations for the origination and early evolution of leaves. The parallel evolution of a marginal meristem by the modification of developmental mechanisms available in the common ancestor of all groups resulted in the pattern of leaf evolution repeated by each clade. Early steps of leaf evolution were followed by constraint on further diversification as the possible elaborations of marginal growth were exhausted. Hypotheses of development in Paleozoic leaves can be tested by the study of living plants with analogous leaf morphologies.
Paleontology enjoys a rich tradition of research on the evolution of morphological diversity. Beginning with Raup's (1966) quantification of molluscan morphospace based on the geometry of coiled shells, paleontologists have used mathematical descriptions of shape (MacLeod 1999; Smith and Bunje 1999); continuous, quantitative measurements of distances among morphological points deemed homologous (Bookstein 1991); and discrete qualitative characterizations of morphology (Foote 1995) to illuminate the history of morphological diversity in invertebrates and skeletonized protists (see McGhee 1999 for review).
Developmental biology offers the prospect of understanding the genetic and physiological bases of morphology and, hence, of morphological evolution. To date, however, few attempts to integrate data from paleontology and developmental biology (reviewed in Shubin et al. 1997; Valentine et al. 1999; Knoll and Carroll 1999) have taken advantage of the possibilities afforded by morphometric analyses of the fossil record.
Vascular plants are particularly well suited for the integration of developmental biology and paleontology. The presence in plants of cell walls vastly increases the probability of anatomical preservation. Cell walls also prohibit cell migration, constraining the types of development that are possible in plants and facilitating the recovery of developmental pattern from fossils. (In structures such as the vascular cambium of seed plants, the tips of developing cells can grow intrusively between other cells, and cell contacts can be established between cell files on either side of a cambial initial that is lost. However, even in this special case of secondary growth, there is no actual cell migration; cambial ontogeny can still be traced readily in the cambium-derived wood [Barghoorn 1940]). This stands in marked contrast to animals, where ontogeny involves complex patterns of cell movement, changing cell contacts, and cell death.
Additional advantages arise because land plants essentially all make their living in the same way (Knoll and Niklas 1987; Niklas 1994). There are various specializations to deal with limitations of water, light, nutrients, symbionts, and substrates, but, with the exception of a few parasites, all plants gather sunlight, water, and carbon dioxide in order to conduct photosynthesis. As a result, there is, in general, far less uncertainty about the interpretation of functional morphology in fossil plants than there is with fossil animals. This uniformity of life strategy, in conjunction with developmental constraints, also increases the likelihood of evolutionary convergence. Roots, secondary growth, and laminate leaves each evolved multiple times in different tracheophyte lineages. Such repeated instances of convergent evolution permit developmental comparison of multiple independent origins of morphologically and functionally similar structures.
This combination of developmental constraint, cellular preservation, and convergent evolution makes plants unusually attractive subjects for morphological analysis. Leaves are particularly advantageous for studies of morphological evolution. Leaf compressions are abundant in fluviatile and lacustrine depositional systems, the leaf fossil record is well documented, and leaves are the one organ for which both overall morphology and details of vascularization are commonly available in the same specimen. Furthermore, laminate leaves are known to have evolved independently in several Paleozoic tracheophyte clades, and the degree of morphological convergence among these early leaves is high. Leaves produced by early pteridophyte and seed plant lineages were in some cases so similar to modern fern leaves that only in the early twentieth century did paleontologists recognize that some were borne by seed plants (reviewed in Scott 1909).
In this paper, we present a morphospace analysis of Paleozoic leaves and interpret the results in light of developmental processes inferred from preserved morphologies.
PATTERNS OF MORPHOLOGICAL EVOLUTION IN PALEOZOIC LEAVES
During the later Devonian and Early Carboniferous, laminate leaves containing multiple veins evolved independently in seed plants, progymnosperms, ferns, and sphenopsids. The leaves of ferns, seed plants, and progymnosperms have traditionally been termed megaphylls and considered to be homologous. By definition (Gifford and Foster 1989), megaphylls are associated with leaf gaps in the stele of the supporting stem; they can be frondose or entire, and typically are laminate and contain more than one vein (unless secondarily reduced as in most conifers). Although widely applied, this megaphyll typology is an artifact of the depauperate living flora. Once fossils are included, no component of the megaphyll concept emerges as a synapomorphy uniting these lineages. In particular, the central tenet of associated leaf gaps is not relevant to the earliest fossil representatives of these lineages, all of which are protostelic (Taylor and Taylor 1993).
It is possible that some or all of these lineages inherited a lateral branch system with a broadly frondlike architecture from their common ancestor (Kenrick and Crane 1997). The likelihood of this is dependent on the phylogenetic placement of a few key taxa of ambiguous affinities. The traditional placement of the ferns and seed plants as sister taxa, with Equisetum as the closest outgroup, suggested that a frondose megaphyllous leaf was a synapomorphy shared by the ferns and the seed plants. However, the most recent phylogenetic work based on living plants places Equisetum and the Psilotales along with eusporangiate ferns as basal lineages in a clade containing all extant pteridophytes, save for lycopods (Pryer et al. 2001). Statements about last common ancestors, then, depend critically on how key Devonian plants without laminated leaves are added to this phylogeny.
Even if certain frond characteristics turn out to be synapomorphies of the clade that includes sphenophytes, ferns, progymnosperms, and seed plants, however, the terminal units on any fronds inherited from a common ancestor would have had little or no lamination. The earliest known leaves in each of the four clades are highly dissected structures composed of segments that were small, narrow, and with a single vein. In light of these fossils, our assessment of leaf evolution does not depend on any particular phylogenetic hypothesis.
A survey of the Paleozoic compression flora of North America and Western and Central Europe was carried out to investigate patterns of morphological diversification in the early evolution of leaves within and among groups. Each species was described from a single primary source, although stratigraphic ranges and taxonomic affinities were modified using the full list of sources (see Appendix 1). Taxonomic affinity was assigned only to leaves with documented connection to either fertile structures or stems with diagnostic anatomical features. Association of leaves and fertile structures at the same localities was not considered sufficient for taxonomic assignment.
Morphological similarity of a species to other leaf species of known taxonomic affinity also was not considered. For example, many Neuropteris species are listed as having unknown affinities despite the fact that some Neuropteris species are known to have been borne by seed plants. An exception to this was made in the case of the gigantopterids. Seed plant identity has been documented only for Asian gigantopterid species, which are beyond the scope of this study, but the unique construction of gigantopterids warrants placement of the two gigantopterid species included from North American localities as seed plants. Fossils that could not be identified to the species level and taxa for which photos with identifiable venation were not available were excluded from the analysis. Of the resulting list of 641 taxa, 52 are seed plants, 144 are pteridophytes, and 445 are of unknown affinity (many of these are probably but not demonstrably seed plant remains). Among the pteridophytes, there are 15 progymnosperms, 33 sphenopsids, 19 leptosporangiate ferns, 27 marattialean ferns, 15 zygopterids, and 35 eusporangiate species of other or unknown affinities. See Appendix 3 for a list of species, their stratigraphic ranges, and their taxonomic affinities.
Taxa were coded for 63 unordered binary and multistate characters (see Appendix 2; a complete data matrix of character codings for each species is available from C. K. B.) describing individual pinnules rather than entire fronds and concerned primarily with venation and laminar structure rather than overall pinnule shape. Individual leaf species commonly display considerable laminar variability, even between the pinnules within a single fossil frond. This variability was included in the character codings by the use of “variable” as a state for many characters (see Appendix 2 for examples). Coding for variability introduces two important, but potentially negative effects. First, it requires the use of characters that are inapplicable to large subsets of the taxa. Second, taxa based on few or incomplete fossil specimens can be miscoded because there is less opportunity for actual species variability to be demonstrated in available examples. Despite these complications, inclusion of variability is preferable to its exclusion, because variation is such a common aspect of plant morphology (e.g., Knauss and Gillespie 2001) and because range of variation is potentially informative about development.
A principal coordinate analysis was used to provide a more comprehensible summary of the information recorded in the study. The original matrix of character codings for the 641 taxa was used to create a 641 by 641 matrix of the pairwise dissimilarities of all species calculated as the number of character mismatches divided by the number of characters that are not missing or nonapplicable. (Dissimilarity matrix and related statistics were calculated using software provided by R. Lupia; further details of methodology described in Lupia 1999. Mathematica was used for the principal coordinates analysis and all other calculations.) This matrix was then transformed to move the centroid of the dissimilarity distribution to zero (Gower 1966). Eigenvalues and eigenvectors of the transformed dissimilarity matrix were determined, and the component values of each eigenvector were used to position each taxon with respect to a particular principal coordinate axis (Sneath and Sokal 1973). The magnitude of the eigenvalue corresponding to each eigenvector gives an indication of the relative contribution of that axis to the summary of information from the original data matrix. The first two principal coordinate axes were plotted as a representation of morphospace occupied through time (Figs. 1, 2). These axes contain about 51% of the information in the original distance matrix, as extimated from the sum of the two corresponding eigenvalues divided by the sum of all eigenvalues of the transformed distance matrix (Foote 1995).
The taxa span the time interval from Middle Devonian until the end of the Permian. The Devonian through Early Permian (Autunian) is divided into intervals averaging approximately 15 million years and ranging from 8 to 21 million years in duration. The later Permian is not well represented for two reasons: (1) there are fewer productive localities within the geographic area covered, and (2) poor preservation and coriaceous habit commonly obscure venation pattern in specimens that are available (Kerp 2000).
Principal coordinate analyses provide a convenient method for visualizing large quantities of morphological information by geometrically summarizing as much of the variability between taxa as possible on a few axes in the form of a morphospace. However, because such an analysis is entirely dependent upon the overall composition of the data set, it can be highly influenced by taxonomic and morphological decisions. In particular, the gigantopterids bore the most complex and morphologically distinctive leaves of any plants in the Paleozoic, but there was no possibility of their leaves having coordinates distinct from other taxa on axes 1 and 2 because gigantopterids represented only about 0.3% of the included species and the characters that distinguish them were invariant among all other taxa. Placement of the diverse morphologies of the Devonian and Carboniferous leaves is, however, more amenable to morphological interpretation (Figs. 1, 2), and the overall pattern of diversification seen in the 50% of the information summarized in the two-dimensional morphospace plot represents well the data set as a whole (Fig. 3).
Three interesting patterns emerge from the resulting Paleozoic leaf morphospace (Fig. 1). First, the areas of initial morphospace occupation in the Devonian and Early Carboniferous remain occupied in the later record, but the taxonomic affinities of the plants exhibiting these leaf morphologies change over time. Second, ferns, seed plants, progymnosperms, and sphenopsids all share the same initial location in morphospace, diverging only with subsequent evolution. This suggests that the independent origins of leaves were based on modifications of a common underlying developmental system. Third, diversity and disparity (total occupied morphospace) initially increase in tandem, but after the Namurian, further increases in taxonomic diversity do not change the range of morphologies occupied, suggesting that the limits of a biologically constrained system had been reached. Early leaf evolution thus appears to be constrained at both ends of the radiation, first by initial architectures and later by the limits of leaf morphological variation.
DEVELOPMENTAL INTERPRETATIONS OF MORPHOLOGICAL PATTERNS
Insights from living organisms have long been applied to paleontological studies of plant function. Examples include biomechanical modeling of extinct species based on characteristics of living tissues (Roth and Mosbrugger 1996; Niklas 1997a,b; examples in Bateman et al. 1998) and estimates of past carbon dioxide levels based on stomatal indices derived from present-day plants (McElwain et al. 1999). In similar ways, paleobotanical studies of morphological evolution can benefit from advances in plant developmental biology. For example, following earlier suggestions (Scheckler 1976, 1978; Wight 1987), Stein (1993) modeled stem vasculature as a function of auxin diffusion from the stem apex and lateral primordia, successfully reproducing the observed stelar morphologies of some Devonian plant axes.
An understanding of leaf development in living plants may similarly inform our understanding of early leaf evolution. Comparative biology suggests that mechanisms of meristematic growth will likely impose constraints on leaf development and, hence, potential leaf morphologies. For this reason, fossil leaves may provide an indirect record of leaf meristematic capability through time. The study of leaf evolution has traditionally relied upon interpretation of frond architecture and the positional identity of laminar subunits of the frond. These traits are important for whole-plant reconstruction and for systematics, but they cannot illuminate the developmental capacity of the foliar meristems of these plants. The emphasis here is on the meristematic potential present in early leaf-bearing plants, using comparisons with living plants to constrain models of lamina development and evolution. Rather than considering positional homology and evolution, the focus is upon meristematic homology and evolution (Stein 1998).
Development can be considered in terms of two related processes: (1) growth, including cell division and the differentiation of individual cells, and (2) the patterning of differentiated cells to form functional tissues (Wolpert 1971; Sachs 1991). The leaves of extant plants display a number of different growth mechanisms. Anatomical studies of morphologically diverse ferns indicate leaf growth by meristems located at the margin of the developing organ (Pray 1960, 1962; Zurakowski and Gifford 1988; White and Turner 1995; Korn 1998). Marginal meristems consist of a peripheral row of dividing initials, which are the ultimate source of all cells in the leaf. Additional submarginal cell divisions are necessary both for cell differentiation and for building up the thickness of the developing leaf. The marginal meristem remains active until general pinnule morphology and the location of all procambium has been determined (Pray 1962). In contrast, angiosperm leaves grow diffusely, with meristematic activity throughout the developing leaf. Clonal analysis experiments corroborate anatomical work (Pray 1955) and SEM studies (Hagemann and Gleissberg 1996) in demonstrating that the marginal cells of an angiosperm leaf play almost no role in leaf growth (Nicotiana tabacum [Poethig and Sussex 1985a,b]) or, at least, play no greater a role than other parts of the leaf (Gossypium barbadense [Dolan and Poethig 1998]).
Marginal and diffuse growth are not necessarily mutually exclusive mechanisms; more likely, they represent end-members of a large and complex continuum. Although the few fern leaves for which development has been studied in detail possessed marginal meristems, others are likely to possess more varied and complex mechanisms of growth (such as those described as having “dilatory” leaf growth by Wagner ). Furthermore, there is much variety in what is here summarized as internal, as opposed to marginal, growth (Foster 1952; Hagemann and Gleissberg 1996). Although continuing research is needed to explore the complete developmental diversity of leaves in living plants, what matters for an initial assessment of Paleozoic plants is that their leaves could grow exclusively by means of a marginal meristem or could include extensive internal growth.
Although differing mechanisms of leaf growth exist, vascular patterning is broadly similar across all tracheophytes. Vascular plants all have a flux of auxins from distal meristems in the shoot system toward more proximal tissues. Auxin transport is accomplished by the pumping of auxin into cells from all sides and the preferential pumping of auxin out of cells proximally down the stem (Gälweiler et al. 1998; Steinmann et al. 1999; Berleth et al. 2000). Physiological studies have demonstrated the importance of auxins both for overall stelar patterning (Ma and Steeves 1992) and for finer scales of differentiation, including the establishment of cell polarity and the continuity of vascular strands (Sachs 1981). Sachs (1981, 1991) has hypothesized that the differentiation of individual vascular strands is based upon the distal to proximal flux of auxin, which determines cell polarity and induces the development of procambial characteristics in the files of cells through which it flows. These procambial characteristics increase the cells' capacity to transport auxin, further increasing flux through the cell file and, in consequence, decreasing fluxes through adjacent cell files. In this way, the widths of procambial strands can be limited without the necessary action of a second, inhibitory hormone.
The studies of vascular patterning and the auxin canalization hypothesis reviewed in the preceding paragraph are based principally on stems—molecular and biochemical studies of leaves are largely limited to investigations of vascular cell differentiation in Zinnia tissue cultures. Nonetheless, several considerations justify the assumption that leaves and stems follow similar mechanisms of vascular patterning and differentiation: The canalization hypothesis is the only hypothesis that has been documented in any part of the plant and it is consistent with all available information from leaves. Leaf primordia are known to be important sources of auxin (Ma and Steeves 1992; Stein 1993), and the disruption of leaf vascular patterning by auxin transport inhibitors (Sieburth 1999; Mattsson et al. 1999) and by mutations that disrupt auxin transport (Carland and McHale 1996) has been documented. (Leaf development is, of course, also influenced by auxin-independent factors [Carland et al. 1999]). The documentation of vascular patterning mechanisms in leaves remains an important research goal, and emerging techniques (Caruso et al. 1995; Moctezuma and Feldman 1999) suggest that new insights are on the horizon.
The growth and patterning mechanisms exhibited by living plants can be used to constrain hypotheses about developmental mechanisms present during the early evolution of leaves. The role of auxin fluxes from active meristems in the patterning of vascular tissue appears to be conserved throughout tracheophytes. We assume, therefore, that it applies to Paleozoic leaves. This assumption, in turn, provides a means of generating hypotheses about leaf growth in extinct plants. If the leaves of Paleozoic plants grew exclusively by marginal meristems, venation should be oriented toward leaf margins with all veins ending marginally, as observed in the living ferns for which marginal growth has been documented. If the leaf development included extensive internal, nonmarginal growth, venation would not be expected to have uniform orientation and vein endings should be dispersed throughout the lamina.
All Devonian and Carboniferous leaves, regardless of phylogenetic affinity, have exclusively marginal vein endings (Figs. 4, 5). This suggests that each origin of laminated leaves relied on marginal meristems. The only Paleozoic leaves with extensive internal vein endings were those of the late Early to Late Permian gigantopterid seed plants. In their case, internal tertiary veins are oriented toward one another and meet in discrete areas between the marginally ending secondary veins. This suggests a two-stage process in which marginal growth was followed by internal growth at discrete intercalary meristems. There is no evidence of true diffuse growth until the Mesozoic.
The early leaves of each lineage likely grew by means of marginal meristems, but our understanding of those meristems can be further refined. In some living ferns, a direct correspondence has been found between specific marginal initials and the cell types to which they give rise (parenchyma, or parenchyma and vasculature [Pray 1960, 1962; Zurakowski and Gifford 1988]). Other work with different ferns has found marginal meristems of uniform composition; in these plants, vascular patterning responds to marginal signals, but without reference to specific marginal initial cells (Korn 1998). Tissue patterning based on discrete ground meristem and procambial marginal initials would be expected to result in leaves with divergent venation, because the marginal initials corresponding to any two adjacent veins can only maintain their distance or grow farther apart as growth continues. Extra submarginal growth can distort the vein paths into convergence (Pray 1960), but in any event, a meristem of distinct initial types could never account for reticulately veined leaves.
Because marginal-meristem types place constraints on resulting leaf morphologies, we can draw inferences about meristem types from fossils with preserved venation. If the marginal meristems of Paleozoic leaves had distinct vasculature and parenchyma initials, divergent venation should dominate the early record. If the marginal meristems were homogenous, vein convergence and reticulation would be expected to be relatively common.
As shown in Figure 4, divergent venation was the first pattern to appear in each lamina-evolving clade. Progymnosperms show only divergent venation. Among seed plants and ferns, convergent venation first occurred in the Visean, and it did not occur among Sphenopsids until the Stephanian. Reticulate venation appeared in the Westphalian, but only in seed plants and leaves commonly presumed to have been borne by seed plants. (Ferns evolved reticulate venation during the Mesozoic Era.) The predominance of divergent venation in early leaves across all phylogenetic groups suggests that each group convergently evolved a marginal meristem with discrete initial types and only later diverged into other development types, including modified forms of marginal meristems with discrete initial types and homogenous marginal meristems.
Leaf evolution in each clade followed the same sequence of morphologies. Paleozoic seed plants first evolved dichotomizing, finely lobed leaves with a single vein per laminar segment, and only later evolved laminar multiveined leaves with divergent venation, followed by convergent venation, followed by reticulation. The other clades followed the same sequence to a varying extent. It is proposed that this repeated pattern reflects the limited number of ways that plants can form laminate photosynthetic surfaces. The range of morphological possibilities is further constrained by the common ancestry of the groups in question: evolution works by the modification of preexisting structures and developmental pathways, and the available underlying pathway common to the four clades and ultimately derived from a shared ancestor was stem development. Stems are indeterminate, cylindrical structures that grow from a discrete apical meristem, providing the source of an auxin gradient involved in vascular patterning. We hypothesize that stepwise modification of the growth and patterning employed in this ancestral, axial system produced marginal meristems and laminate leaves in each lineage (Table 1).
Just as the early steps of marginal meristem evolution were constrained by the scope of developmental possibility, the ultimate range of possible leaf morphologies must have been constrained by available developmental mechanisms. Early leaves consisting of linear, dichotomizing segments had limited potential for morphological variability: they could be three-dimensional or planar; dichotomies could be equally distributed, concentrated distally, or concentrated proximally; and the relative strengths of the two arms of a dichotomy could be altered (Fig. 6). Each of these modifications had been explored by the Late Devonian.
The possibilities of marginal growth are also limited. Simple marginal growth with point initiation will result in a fan-shaped leaf with vein endings along the distal margin, the only margin along which a marginal meristem would have been active. This system can be modified by symmetric or asymmetric alteration of the rate of growth parallel and/or perpendicular to the marginal meristem or by variation in the duration of growing time along the meristem (Fig. 7). Other possible modifications include the development of a midvein, and broad rather than point initiation of lamina growth. A rigorous documentation of possible leaf forms in terms of a theoretical morphospace (McGhee 1999) based on developmental mechanisms is needed, but it is fair to state that, even though the various possible combinations of these modifications continued to be shuffled within individual lineages, all specified variables had been explored by the Namurian. The lack of further increase in overall morphological disparity after the Namurian, despite further increase in taxonomic diversity, likely reflects limitations on possible elaborations of marginal growth.
The stratigraphic distributions of individual characters may provide crude tools for investigating the evolution of leaf development, but the results are compelling. Internal vein endings are not present for the first 100 million years of leaf evolution; convergent venation is absent for the first 50 million years. The morphological constraints apparent in early leaves strongly suggest specific mechanisms of development. Although multiveined leaves evolved independently at least four times, the close similarities of early leaves in all groups suggest parallel evolution by modification of common developmental mechanisms inherited from ancestors whose photosynthetic organs consisted of apically growing, bifurcating axial systems.
The evolution of laminated leaves may be related to the evolution of vascular systems competent to support high levels of evapotranspiration and other aspects of whole-plant function in increasingly stratified later Devonian and Early Carboniferous communities. It has also been proposed that appearance of laminate leaves was causally linked to a drop in atmospheric CO2 concentration through the Devonian (Beerling et al. 2001). Changing environmental conditions may well have removed a physiological barrier to the evolution of leaf lamination. Nonetheless, the staggered timing of lamina evolution among clades—in the Devonian, multiveined, laminate leaves occurred primarily in the Archaeopterid progymnosperms—suggests that both intrinsic and extrinsic factors are necessary to explain the observed patterns of leaf evolution.
After an initial period of parallel evolution, leaf morphologies within and among the groups diverged; however, by Namurian times the limitations of marginal meristematic development had been reached and further increases in taxonomic diversity did not increase morphological disparity. Innovations in frond architecture, leaf anatomy, and biochemistry continued to evolve, but the possibilities of marginal pinnule growth had been explored.
The use of vascular patterning as a proxy for developmental mechanism can illuminate other patterns of developmental evolution and convergence, including the evolution of diffuse leaf growth in post-Paleozoic time. Angiosperms radiated initially in the Cretaceous Period. One of the hallmarks of angiosperm morphology, at least among dicots, is highly reticulate leaf venation with multiple vein orders and freely ending internal veinlets in the vascular aureoles (Esau 1953; Gifford and Foster 1989). On the basis of available experimental evidence (Foster 1952; Pray 1955; Poethig and Sussex 1985a,b; Hagemann and Gleissberg 1996; Dolan and Poethig 1998), this pattern is interpreted as an indication of diffuse leaf growth. Venation patterns suggestive of diffuse leaf growth, however, are not limited to the angiosperms. Species in two other groups that radiated in the Cretaceous and early Tertiary, the Gnetales and the fern clade that includes the polypods and dryopterids, also possess leaves of this type. As in the Paleozoic, Mesozoic innovations in leaf development were convergent. The simultaneously radiating angiosperms, gnetaleans, and polypods were the only groups to evolve this type of leaf, aside from a small leptosporangiate fern clade that includes Dipteris (fossil record dating to the Late Triassic [Collinson 1996]), the eusporangiate ferns Ophioglossum (order Ophioglossales represented in record only by Cenozoic specimens of another genus [Rothwell and Stockey 1989]) and Christensenia (order Marattiales has an extensive Paleozoic and later fossil record, but there are no known fossils of this genus), and a few Mesozoic specimens of unknown affinities (examples in Trivett and Pigg 1996).
Diffuse leaf growth may well provide structural or physiological advantages to vascular plants. For example, the pattern of highly reticulate venations, many vein orders, and internally ending veinlets that we associate with diffuse growth should provide redundancy of water transport and increased structural support for the typically larger laminae of angiosperms (Givnish 1979; Roth-Nebelsick et al. 2001). The capacity for developmental flexibility that this growth requires may also be advantageous. In addition to any particular functional value, convergent leaf form in these three groups might reflect initial radiations in similar environments. The center of diversity for polypod ferns is in the Tropics. Most basal angiosperm lineages (Feild et al. 2000) and Gnetum are also tropical understory plants. The convergent evolution of leaf development in these groups could reflect radiation in tropical understory environments with low light levels and high humidity.
The developmental hypotheses advanced in this paper are testable. Although living Sphenophyllum will never be available for direct scrutiny, the logic of using adult morphology as a proxy for developmental processes can be tested because the same arguments can be used to make developmental predictions about other laminar organs available for study in living plants. These include not only the leaves of unstudied ferns, angiosperms, cycads, Ginkgo, araucaurian conifers, and Gnetales, but also additional organs such as laminate floral parts and winged seeds. Venation patterns in these organs can differ substantially from those of leaves borne on the same plant. We predict that differing venation patterns will be found to reflect developmental mechanisms similar to those documented for leaves.
Sources used to compile morphological characters of Paleozoic leaf species and to determine their stratigraphic ranges and systematic affinities.
1. Abbott, M. L. 1958. The American species of Asterophyllites, Annularia, and Sphenophyllum. Bulletin of American Paleontology 38:298–389.
2. Alvarez Ramis, C., J. Doubinger, and R. Germer. 1978. Die sphenopteridischen Gewäsche des Saarkarbons. 1. Sphenopteris sensu stricto. Palaeontographica, Abteilung B 165:1–42.
3. ———. 1979. Die sphenopteridischen Gewäsche des Saarkarbons. 2. Alloiopteris und Palmatopteris. Palaeontographica, Abteilung B 170:126–150.
4. Andrews, H. N., T. L. Phillips, and N. W. Radforth. 1965. Paleobotanical studies in Arctic Canada, part I: Archaeopteris from Ellesmere Island. Canadian Journal of Botany 43:545–556.
5. Arnold, C. A. 1939. Observations on fossil plants from the Devonian of eastern North America. IV. Plant remains from the Catskill Delta deposits of northern Pennsylvania and southern New York. Contributions from the Museum of Paleontology, University of Michigan 5:271–314.
6. Batenburg, L. H. 1977. The Sphenophyllum species in the Carboniferous flora of Holz (Westphalian D, Saar Basin, Germany). Review of Palaeobotany and Palynology 24:69–99.
7. Beck, C. B. 1967. Eddya sullivanensis, gen. et sp. nov., a plant of gymnospermic morphology from the Upper Devonian of New York. Palaeontographica, Abteilung B 121:1–22.
8. Bell, W. A. 1938. Fossil flora of Sydney coalfield, Nova Scotia. Geological Survey of Canada Memoirs 215:1–334.
9. ———.1962. Flora of Pennsylvanian Pictou Group of New Brunswick. Geological Survey of Canada Bulletin 87:1–71.
10. Bertrand, P. 1930. Neuroptéridées: Bassin houiller de la Sarre et de la Lorraine. Études des gîtes minéraux de la France. I. Flore fossile, Fasc. 1:1–58. Paris.
11. ———. 1932. Aléthoptéridées: Bassin houiller de la Sarre et de la Lorraine. Études des gîtes minéraux de la France I. Flore fossile, Fasc. 2:1–40. Paris.
12. Boureau, E. 1964. Traité de paléobotanique, Tome III. Sphenophyta and noeggerathiophyta. Masson, Paris.
13. Boureau, E., and J. Doubinger. 1975. Traité de paléobotanique, Tome IV, Fasc. 2. Pteridophylla (première partie). Masson, Paris.
14. Buisine, M. 1961. Contribution à l'étude de la flore do terrain houiller: les Aléthoptéridées du Nord de la France. Études géologiques pour l'atlas de topographie souterraine. I. Flore fossile, Fasc. 4:1–317. Douriez-Bataille, Lille.
15. Carluccio, L. M., F. M. Hueber, and H. P. Banks. 1966. Archaeopteris macilenta, anatomy and morphology of its frond. American Journal of Botany 53:719–730.
16. Cleal, C. J., and B. A. Thomas. 1994. Plant fossils of the British coal measures. Palaeontological Association Field Guides to Fossils 6:1–222.
17. Corsin, P. 1932. Marioptéridées: Bassin houiller de la Sarre et de la Lorraine. Études des gîtes minéraux de la France. I. Flore fossile, Fasc 3:111–173. Paris.
18. ———. 1951. Pécoptéridées: bassin houiller de la Sarre et de la Lorraine. Études des gîtes minéraux de la France. I. Flore fossile, Fasc. 4:174–370. Paris.
19. Crookall, R. 1955–1976. Fossil plants of the Carboniferous rocks of Great Britain. Memoirs of the Geological Survey of Great Britain, Palaeontology 4:1–1004.
20. Daber, R. 1955. Pflanzengeographische Besonderheiten der Karbonflora des Zwickau-Lugauer Steinkohlenreviers: A, pteridophyllen (Farnlaubige Gewächse). Geologie 13:1–44.
21. ———. 1959. Die Mittel-Visé-Flora der Tiefbohrungen von Doberlug-Kirchain. Geologie 26:1–83.
22. Dalinval, A. 1960. Contribution à létude des Péricopteridées: les Pecopteris de Bassin houiller de Nord de la France. Études géologiques pour l'atlas de topographie souterraine. I. Flore fossile, Fasc. 3. Service Géologique des H.B.N.P.C., Lille.
23. Danze, J. 1956. Les fougères sphénoptéridiennes du Bassin houillier du Nord et du Pas-de-Calais. Études géologiques pour l'atlas de topographie souterrain. I. Flore fossile, Fasc.2. Service Géologique des H.B.N.P.C., Lille.
24. Danze-Corsin, P. 1953. Contribution à l'eétude des Marioptéridées: les Mariopteris du Nord de la France du Bassin houillier du Nord et du Pas-le-Calais. études géologiques pour l'atlas de topographie souterrain, I. Flore fossile, Fasc. 1. Service Géologique des H.B.N.P.C., Lille.
25. Darrah, W. C. 1969. A critical review of Upper Pennsylvanian floras of eastern United States with notes on the Mazon Creek flora of Illinois. W. C. Darrah, Gettysburg.
26. Dix, E. 1933. The succession of fossil plants in the millstone grit and the lower portion of the coal measures of the South Wales coalfield. Palaeontographica, Abteilung B 78:158–202.
27. Doubinger, J. 1956. Contribution à l'étude des Flores autuno-stephaniennes. Mémoires de la Société Géologique de France 75.
28. Doubinger, J., and R. Germer. 1971. Neue pecopteridenfunde im Saarkarbon. Palaeontographica, Abteilung B 133:72–88.
29. ———. 1971. Die Gattung Odontopteris im Saarkarbon. Palaeontographica, Abteilung B 136:129–147.
30. Doubinger, J., P. Vetter. J. Langiaux, J. Galtier, J. Broutin. 1995. La flore fossile du bassin houiller de Saint-Étienne. Mémoires du Muséum National d'Histoire Naturelle. 164.
31. Gastaldo, R. A., and L. C. Matten. 1978. Studies on North American pecopterids. I. Pecopteris vera n. sp. from the Middle Pennsylvanian of southern Illinois. Palaeontographica, Abteilung B 165:43–52.
32. Gensel, P. G. 1988. On Neuropteris Brongniart and Cardiopteridum Nathorst from the Early Carboniferous Price Formation in southwestern Virginia, U.S.A. Review of Palaeobotany and Palynology 54:105–119.
33. Gensel, P. G., and H. N. Andrews. 1984. Plant life in the Devonian. Praeger, New York.
34. Gillespie, W. M., and H. W. Pfefferkorn. 1986. Taeniopterid lamina on Phasmatocycas megasporophylls (Cycadales) from the Lower Permian of Kansas, U.S.A. Review of Palaeobotany and Palynology 49:99–116.
35. Gillespie, W. M., J. A. Clendening, and H. W. Pfefferkorn. 1978. Plant fossils of West Virginia. West Virginia Geological and Economic Survey Educational Series ED-3A.
36. Gothan, W., and W. Remy. 1957. Steinkohlenpflanzen. Gluckauf Verlag, Essen.
37. Guthorl, P. 1953. Querschnitt durch des Saar-LothringischeKarbon. Palaeontographica, Abteilung B 94:139–191.
38. Hill, S. A., S. E. Scheckler, and J. F. Basinger. 1997. Ellesmeris sphenopteroides, Gen. et sp. nov., a new zygopterid fern from the Upper Devonian (Frasnian) of Ellesmere, N.W.T. Arctic Canada. American Journal of Botany 84:85–103.
39. Høeg, O. A. 1931, 1935, 1945. Notes on the Devonian flora of western Norway. Kgl Norske Videnskabers Selskabs Skrifter 6:1–18, 15:1–18, 25:183–192.
40. ———. 1942. The Downtownian and Devonian flora of Spitzbergen. Norges Svalbad-og Ishaws-Unserskelser 83:1–229.
41. Kerp, J. H. F. 1988. Aspects of Permian of palaeobotany and palynology. X. The west and central European species of the genus Autunia Krasser emend. Kerp (Peltaspermaceae) and the form-genus Callipteris Brongniart 1849. Review of Palaeobotany and Palynology 54:249–360.
42. Kidston, R. 1923–1925. Fossil plants of the Carboniferous rocks of Great Britain. Memoirs of the Geological Survey of Great Britain, Palaeontology 2:1–681.
43. Knauss, M. J., and W. M. Gillespie. 2001. Genselia compacta (Jongmans et al.) Knaus et Gillespie comb. nov.: new insights into possible developmental pathways of early photosynthetic units. Palaeontographica, Abteilung B 256:69–94.
44. Krings, M., and H. Kerp. 1998. Epidermal anatomy of Barthelopteris germarii from the Upper Carboniferous and Lower Permian of France and Germany. American Journal of Botany 85:553–562.
45. Laveine, J.-P. 1967. Les Neuroptéridées du Nord de la France. Études géologiques pour l'atlas de topographie souterrain. I. Flore fossile, Fasc. 5. Service Géologique des H.B.N.P.C., Lille.
46. Laveine, J.-P., R. Coquel, and S. Loboziak. 1977. Phylogénie générale des Calliptéridiacées (Pteridospermopsida). Geobios 10:757–847.
47. Leary, R. L., and H. W. Pfefferkorn. 1977. An Early Pennsylvanian flora with Megalopteris and Noeggerathiales from west-central Illinois. Illinois State Geological Survey Circular 500:1–46.
48. Leclercq, S., and P. M. Bonamo. 1971. A study of the fructification of Milleria (Protopteridium) Thomsonii Lang from the Middle Devonian of Belgium. Palaeontographica, Abteilung B 136:83–114.
49. Mamay, S. H. 1955. Acrangiophyllum, a new genus of Pennsylvanian Pteropsida based on fertile foliage. American Journal of Botany 42:177–183.
50. ———. 1989. Evolsonia, a new genus of Gigantopteridaceae from the Lower Permian Vale Formation, north-central Texas. American Journal of Botany 76:1299–1311.
51. Mapes, G., and J. T. Schabilion. 1979. A new species of Acitheca (Marattiales) from the Middle Pennsylvanian of Oklahoma. Journal of Palaeontology 53:685–694.
52. Matten, L. C. 1981. Svalbardia banksii sp. nov. from the Upper Devonian (Frasnian) of New York state. American Journal of Botany 68:1383–1391.
53. Obrhel, J. V. 1961. Die flora der Srbsko-Schicten (Givet) des mittelböhmischen Devons. Sborník Ustredniho Ustavu Geologického 26:7–46.
54. Phillips, T. L., H. N. Andrews, and P. G. Gensel. 1972. Two heterosporous species of Archaeopteris from the Upper Devonian of West Virginia. Palaeontographica, Abteilung B 139:47–71.
55. Read, C. B., and S. H. Mamay. 1964. Upper Paleozoic floral zones and floral provinces of the United States. U.S. Geological Survey Professional Paper 454-K:1–35.
56. Remy, R., and W. Remy. 1959. Pflanzenfossilien. Akademie, Berlin.
57. ———. 1977. Die floren des Erdaltertums. Glückauf, Essen.
58. ———. 1978. Die flora des Perms im Trompia-Tal und die Grenze Saxon-Thuring in den Alpen. Argumenta Paleobotanica 5:57–90.
59. Scheckler, S. E. 1976. Ontogeny of progymnosperms. I. Shoots of Upper Devonian Aneurophytales. Canadian Journal of Botany 54:202–219.
60. Scheckler, S. E., and H. P. Banks. 1971. Proteokalon a new genus of progymnosperms from the Devonian of New York state and its bearing on phylogenetic trends in the group. American Journal of Botany 58:874–884.
61. ———. 1986. Geology, floristics, and paleoecology of Late Devonian coal swamps from Appalachian North America. Annales de la Société Géologique de Belgique 109:209–222.
62. Schweitzer, H.-J. 1973. Die Mittledevon-flora von Lindlar (Rheinland), Part 3. Filicinae—Calamophyton primaevum Kräusel und Weyland. Palaeontographica, Abteilung B 140:17–150.
63. ———. 1986. The land flora of the English and German Zechstein sequences. In G. M. Harwood and D. B. Smith, eds. The English Zechstein and related topics. Geological Society of London Special Publication 22:31–54.
64. Skog, J. E., and P. G. Gensel. 1980. A fertile species of Triphyllopteris from the early Carboniferous (Mississippian) of southwestern Virginia. American Journal of Botany 67:440–451.
65. Stockmans, F. 1948. Végétaux du Dévonien supérieur de la Belgique. Mémoires du Musée Royal d'Histoire Naturelle de Belgique 110:1–85.
66. Stockmans, F., and Y. Willière. 1953. Végétaux namuriens de la Belgique. Association pour l'Étude de la Paléontologie et Stratigraphie Houillères 13:1–382.
67. ———. 1955. Végétaux namuriens de la Belgique. II. Assise de Chokier, Zone de Bioul. Association pour l'Étude de la Paléontologie et Stratigraphie Houillères 23:1–35.
68. ———. 1956. Végétaux de la Zone d'Oupeye a Sarolay (Argenteau). Association pour l'Étude de la Paléontologie et Stratigraphie Houillères 25: Plates A, B.
69. ———. 1965. Documentos paleobotanicos para el estudio de la geología Hullera del Noroeste de España. Mémoires du Institut Royal des Sciences Naturelles de Belgique, 2e série, Fasc. 79.
70. Wagner, R. H. 1968. Upper Westphalian and Stephanian species of Alethopteris. Uitgevers-Maatschappij, Maastricht.
71. Walton, J. 1926. Contributions to the knowledge of Lower Carboniferous plants. Part 1, On the genus Rhacopteris. Part 2, On the morphology of Sphenopteris Teiliana, Kidston, and its bearing on the position of the fructification on the frond of some Lower Carboniferous plants. Philosophical Transactions of the Royal Society of London B 215:201–224.
72. White, D. 1929. Flora of the Hermit Shale. Carnegie Institution of Washington Publication 405:1–221.
73. Zodrow, E. L. 1989. Revision of Silesian sphenophyll biostratigraphy of Canada. Review of Palaeobotany and Palynology 58:301–332.
Characters used for descriptions of leaf morphology. Inapplicable or missing characters were coded as “?”.
1. Lamina: 0, three-dimensional; 1, planar.
2. Lamina: 0, lobed; 1, not lobed; 2, variable.
3. Entire lamina: 0, simple; 1, sinuous margin; 2, variable.
4. Margin: 0, smooth; 1, margin responds to each vein ending; 2, variable.
5. Location of maximum width: 0, proximal; 1, central; 2, distal; 3, variable.
6. Maximum width: 0, singular; 1, maintained; 2, variable (for characters 5 and 6, linear or lobed =? ).
7. Attachment: 0, narrow; 1, broad; 2, variable.
8. Narrow attachment type: 0, not stalked; 1, stalked; 2, variable.
9. Insertion: 0, perpendicular to somewhat angled (>60°); 1, severely angled (<60°); 2, variable; 3, skew (Sphenophyllum; Cordaites).
10. Attachment proximal: 0, straight; 1, constricted; 2, decurrent; 3, variable.
11. Attachment distal: 0, straight; 1, constricted; 2, decurrent; 3, variable.
12. Shape of pinnule body: 0, symmetric; 1, not symmetric; 2, variable.
13. Not-symmetric pinnules are: 0, falcate or otherwise regular but asymmetric; 1, irregular; 2, variable.
14. Lamina (or lobe) tip: 0, regular; 1, irregular edge; 2, variable (linear, lobed leaves are 1).
15. Regular lamina tip: 0, acute; 1, rounded; 2, variably rounded or acute; 3, wedge.
16. Wedge leaf distal margin: 0, curved; 1, straight; 2, acute; 3, irregular; 4, variable.
17. Maximum number of veins in laminar segment: 0, one; 1, more than one; 2, variable.
18. Branching of veins within lamina: 0, absent; 1, present; 2, variable.
19. Number of distinct vein orders besides any midvein present: number.
20. Relations of vein orders: 0, strictly hierarchical; 1, not ; 2, variable (=? if no midvein).
21. Minimum number of branchings from base or midvein to margin: number (from midvein = 1).
22. Maximum number of branchings from base or midvein to margin: number.
23. Branching of laterals: 0, just dichotomous; 1, also subdichotomous; 2, also pinnate.
24. Location of vein branching along proximal-distal axis: 0, only dispersed; 1, restricted at vein level; 2, restricted at whole lamina level; 3, restriction at both vein and whole lamina level (restrictions in linear leaves are considered at whole lamina level = 2).
25. Vein level restriction of branching favors: 0, distal; 1, center; 2, proximal; 3, variable.
26. Lamina level restriction of branching favors: 0, distal; 1, center; 2, proximal; 3, variable.
27. Location of vein branching from origin to edge: 0, only dispersed; 1, restricted; 2, variable.
28. Location of vein branching from origin to edge favors: 0, origin; 1, edge; 2, variable.
29. Vein paths: 0, only divergent; 1, convergence present; 2, variable.
30. Vein convergence: 0, only passive; 1, strict; 2, variable.
31. Location of convergence: 0, dispersed; 1, restricted; 2, variable.
32. Location of convergence restricted to: 0, vein origin; 1, lamina edge; 2, variable.
33. Vein reticulations: 0, no; 1, irregular; 2, regular.
34. Minimum number of reticulations from base or midvein to margin: number.
35. Maximum number of reticulations from base or midvein to margin: number.
36. Location of reticulations: 0, dispersed; 1, restricted to origin; 2, restricted to edge.
37. Vein orders with reticulation: 0, only most distal; 1, more than most distal.
38. Enclosed space: 0, elongate perpendicular to margin; 1, elongate parallel to margin; 2, isodiametric; 3, irregular; 4 variable.
39. Lamina innervated from rachis: 0, once; 1, more than once; 2, variable.
40. Multiple lamina innervations: 0, all equivalent; 1, unequal strength; 2, variable.
41. Lamina innervations from base: 0, evenly spaced; 1, unevenly; 2, variable (centered single vein = 0).
42. Lamina innervations from midvein: 0, evenly spaced; 1, unevenly; 2, variable.
43. Innervation of lamina from base: 0, straight; 1, angled; 2, branches immediately; 3, irregular (with respect to main axis of lamina, a vein following the angled insertion of the lamina would be scored as straight).
44. Innervation of lamina from midvein: 0, straight; 1, angled; 2, branches immediately; 3, irregular.
45. Midvein: 0, no; 1, yes; 2, variable.
46. Midvein: 0, weak, not straight; 1, strong; 2, variable.
47. Midvein: 0, included in lamina; 1, distinct from lamina; 2, variable.
48. Midvein length: 0, as long as lamina; 1, closer to distal than other margins; 2, reaches all margins equally; 3, farther from distal; 4 variable.
49. Midvein of uniform thickness: 0, no; 1, yes; 2, variable.
50. Angle of midvein (or other innervation) insertion: 0, same as pinnule; 1, different; 2, variable.
51. “midvein” branches: 0, to both sides; 1, just 1, side; 2, variable.
52. Path of laterals to margin: 0, parallel; 1, not parallel; 2, variable (if linear =?).
53. Lateral vein paths: 0, straight; 1, curved; 2, variable but regular; 3, irregular paths.
54. Concavity of vein curvature: 0, concave up (distal); 1, concave down (proximal); 2, variable.
55. Location of vein endings: 0, all veins equally reach margin (or marginal vein); 1, some internal endings; 2, only internal endings; 3, no free endings.
56. Direction of vein paths: 0, only toward a margin; 1, internally directly veins (perpendicular to or independent of margin).
57. Marginal vein: 0, absent; 1, present.
58. Vein endings: 0, just on distal edge; 1, all margins but expanded base; 2, all margins;3, variable.
59. Vein density: 0, uniform; 1, increases; or 2, decreases toward margin; 3, irregular.
60. Veins within a lamina lobe: 0, always include all of the connected veins distal to the last shared dichotomy (i.e., always forming a monophyletic group of veins); 1, not always the case (i.e., also paraphyletic vein groupings within lobes).
61. Lobing: 0, about each vein; 1, vein groups; 2, both types present.
62. Angle of marginal intersection of veins: 0, ∼90°; 1, angled; 2, variable but consistent; 3, irregular.
63. Vein endings where present: 0, evenly spaced; 1, uneven but predictable; 2, irregular; 3, variable type (if linear or lobed about each vein, then 62 and 63 =? ).
We thank R. Lupia for providing software to create the dissimilarity matrix (available at http://geosci.uchicago.edu/paleo/csource); B. Craft for programming assistance; W. Stein and N. Sinha for thoughtful reviews; and T. Feild, M. Foote, M. Hoolbrook, E. Kellogg, E. Kramer, R. Lupia, C. Marshall, R. Moran, and M. Thompson for helpful discussions. Research leading to this paper was supported in part by a National Science Foundation predoctoral fellowship and by the NASA Astrobiology Institute.
- Accepted 15 August 2001.