Morphological (and not anatomical or reproductive) features define early vascular plant phylogenetic relationships

Traditional evolutionary theory postulates that the evolution of reproductive features tends to be conservative compared to the diversification of anatomical and morphological features (Stebbins, 1951, 1970; Mayer, 1982; see also Wagner and Altenberg, 1996), possibly because selection is more “relaxed” on nonreproductive or nonspecialized traits, particularly for plants because of their obvious modular construction (see, for example, Armbruster and Wege, 2019). The perspective that reproductive and nonreproductive traits are modular and subject to different selection pressures is supported by evidence drawn from the phylogenetic analyses of extant plant lineages, which appear to corroborate the impression that convergent evolution and homoplasy are more frequent in internal structure and external appearance (Bateman et al., 1998; Cascales-Minana et al., 2010; Cascales-Minana, 2016). In fact, this approach to plant taxonomy dates back to the work of Carolus Linnaeus who predicated his taxonomy on the structure and arrangement of reproductive organs such as flowers (Linnaeus, 1753). Although the impression that vegetative characters are developmentally more “plastic” and less evolutionarily “conservative” than reproductive features is well engrained in the older literature (e.g., Cronquist, 1981; Stevens, 2001), it remains unclear whether this dyadic held true throughout the evolutionary history of plant life, particularly in the context of early vascular plants. Prior research has shown that the most ancient vascular plants evolved rapidly during the early to mid-Paleozoic (Knoll et al., 1979; Niklas et al., 1980). Indeed, so much so that this episode is comparable in its biological importance to the great Cambrian explosion of the metazoans.

Here, we evaluated this pivotal evolutionary period to determine the extent to which anatomical, morphological, or reproductive features support perceived phylogenetic relationships among early tracheophytes. This goal was approached using an extensively revised and updated character matrix for 37 fossil taxa and the extant genus Equisetum (Crepet and Niklas, 2018, 2019). The consensus tree topology emerging from this matrix was then “dissected” by comparing it to the tree topologies using only anatomical, morphological, or reproductive characters, and the three permutations of pairs of the three types of characters (e.g., anatomical and morphological characters). The extent to which each of the six tree topologies retrieved the topology of the original 54-character tree was used to assess the extent to which anatomical, morphological, or reproductive features contributed to early tracheophyte diversification.

Importantly, both the old and new character consensus tree topologies identify two major clades (i.e., a zosterophyllophyte-lycophyte clade and a trimerophyte-euphyllophyte clade) derived from a Cooksonia-rhyniophyte ancestral plexus. Although these topologies resonate well with previous assessments (e.g., see Banks, 1968, 1975; Stewart and Rothwell, 1993; Kenrick and Crane, 1997; Taylor et al., 2009), we make no claim that the topology of the 54-character consensus tree presented here reflects the real course of early tracheophyte evolution. We use this topology merely as a logical construct with which to evaluate the relative contributions the three different types of characters make to it. Another important caveat is that little is known about the gametophytes corresponding to the majority of the fossil plants used in our analysis, i.e., when dealing with the reproductive characters and character states used here, we are dealing predominantly with sporophytes.

MATERIALS AND METHODS

An extensively revised character matrix for 37 fossil taxa (and one extant taxon, Equisetum) was used for this study. This matrix differs in many important ways from the two previously published (Crepet and Niklas, 2018, 2019); e.g., one character was eliminated, and all but one additive character were rescored as non-additive. As before, the 37 fossils were selected using two stringent criteria: (1) each had to provide the best (iconic) representative of its lineage (e.g., Rhynia and Asteroxylon), and (2) each had to provide as many characters as possible to reduce the negative consequences of missing characters on resolving phylogenetic relationships (Nixon and Davis, 1991; Rothwell and Nixon, 2006). Thirty-one fossil taxa were from the Silurian–Devonian time interval; three taxa were selected from the Carboniferous to include important crown taxa (i.e., Lepidodendron, Lyginopteris, and Calamites), and three taxa were selected because they are the only representatives of evolutionarily significant lineages or they manifest unusual character states (i.e., Nothia, Elkinsia, and Eospermatopteris). The extant taxon Equisetum was included because recent analyses set it within a group traditionally considered to be ferns (e.g., Knie et al., 2015), whereas this genus has been aligned traditionally with the sphenopsid clade (e.g., a highly derived relative of taxa such as Calamites) (Stein et al., 1984; Elgorriaga et al., 2018). Importantly, the effects of missing characters were not totally avoided because the majority of the 37 fossil taxa lack anchorage systems and, in some cases, even reproductive organs (Taylor et al., 2009).

With the exception of Equisetum, extant taxa were omitted from analyses because their highly derived character states are not found in fossils in the time interval examined here and because this would result in numerous missing characters (for the fossil taxa) that would alter the phylogenetic interpretations of fossil organisms (Nixon and Davis, 1991). Moreover, analyses using larger data sets containing extant taxa are likely to place fossil organisms in topologies largely defined by the characters associated with extant organisms. Thus, the placement of fossil organisms on tree topologies is biased simply because of the sheer number of derived characters. In addition, the primary objective here was to explore the possibility that matrices composed exclusively of fossil organisms might provide new insights into the early evolution of vascular plant complexity.

Appendices 1 and 2 provide the 54-character matrix and the revised scoring of the character states used to determine the phylogenetic relationships among the 38 taxa. The statistics for the consensus trees reported in this study are provided in the appropriate figure legends.

The extent to which some characters were rescored can be assessed by comparing the topology of the 54-character tree presented here (Fig. 1) with that presented in our previous publications (Crepet and Niklas, 2018, 2019, see fig. 1 in both cases). However, as in previous analyses, all characters were scored using traditional definitions and criteria, e.g., (1) whether a taxon had monopodial branching (here defined as branching systems generated by terminal meristems capable of producing lateral branches that are developmentally and physiologically subordinate, e.g., Calamites and Equisetum) versus pseudomonopodial branching (here defined as branching systems produced by the unequal division of terminal meristems giving the appearance of a main axis bearing lateral branches, e.g., Rhynia and Psilophyton) and (2) whether secondary growth or a siphonostele was present (Esau, 1965; Bierhorst, 1971; Beck et al., 1982; Schmidt, 1982).

The extent to which the new consensus tree topology using all 54 characters (Fig. 1) depended on vegetative versus reproductive characters was evaluated by generating six consensus trees: one based using only reproductive features (18 characters; 47 character states), another based solely on anatomical features (14 characters; 54 character states), another using only morphological features (22 characters; 57 character states); a fourth based on morphological and anatomical characters (36 characters; 111 character states); and the consensus trees for reproductive and morphological characters and for reproductive and anatomical characters (Figs. 2-4).

All matrixes were extensively analyzed; over 150 iterations were computed using Winclada and maximum parsimony phylogenetic analytical protocols (heuristic search; maximum number of trees = 5,000,000; number of replications = 400) were performed using the program NONA (Goloboff, 1999), a variant of the program Winclada (Asado, version 1.1 beta, Cornell University; Nixon, 1999). Characters were all unweighted. Winclada treats inapplicable characters (-) and unknown or unresolved characters (?) in the same manner, i.e., scoring a character as “-” or “?” has no effect on tree topology. With the exception of character 47, all characters where scored as non-additive. The scoring of character 47 (i.e., branching planarity; see Appendix 2) assumes that unequal branching is a prerequisite for the evolution of pseudomonopodial and monopodial branching, i.e., “overtopping” prefaced subsequent developmental modifications giving rise to more elaborate branching systems. We concede that saltational evolution is possible, but, in our view, this is less parsimonious. All of the tree topologies presented here were generated using Nixon Ratchet (1999); when, using the previous character matrix, its application resulted in the same set of the parsimonious trees identified in our previous analyses.

RESULTS

The consensus tree of the revised 54-character matrix identified a zosterophyllophyte-lycophyte clade and a trimerophyte-euphyllophyte clade (with a Calamites-Archeocalamites-Equisetum late-divergent triad) emerging from a Cooksonia-rhyniophyte plexus of fossil plants, regardless of whether trees were rooted with Cooksonia or Aglaophyton (Fig. 1). This topology contained four polytomies, one within the zosterophyllophyte-lycophyte clade, which failed to resolve the relationships among Clwydia, Colpodexylon, and a Leclercqia-Lepidodendron dyad, and three within the trimerophyte-euphyllophyte clade, one of which was extensive. Additional analyses (see Appendix 3) indicated that the extensive trimerophyte-euphyllophyte polytomy is essentially resolved when Oocampsa is removed from the matrix. This taxon had many missing characters states (e.g., its anatomy is currently unknown). The 54-character matrix failed to group Aneurophyton with a Rellimia-Tetraxylopteris dyad (see Toledo et al., 2018 for similar findings). In addition, the seed plants Elkinsia and Lyginopteris appear on separate branches within the trimerophyte-euphyllophyte clade. The latter was grouped in an unresolved polytomy composed of Lyginopteris, sphenopsids, and Equisetum.

The topology of the 54-character consensus tree degenerated into an extensive polytomy when only reproductive characters were used (Fig. 2A). Only six taxa were sorted into two polytomies, i.e., an Elkinsia-Lyginopteris- Stauropteris triad and an Oocampsa-Rellimia-Tetraxylopteris triad. The topology of the 54-character tree was retrieved to some degree when anatomical and morphological characters were used in combination (Fig. 2B). Nevertheless, substantive differences between the topology of the 54-character consensus tree (Fig. 1) and that generated by vegetative characters were observed (Fig. 2B); e.g., the zosterophyllophyte-lycophyte polytomy became more extensive and basal, whereas the most extensive trimerophyte-euphyllophyte polytomy became more resolved. The failure of reproductive characters to resolve any discernable phylogenetic signal was not attributable to the smaller number of reproductive characters compared to the combined number of anatomical and morphological characters (18 vs. 36 characters, and 47 vs. 111 character states, respectively).

The topology of the consensus trees using only morphological characters (Fig. 3A) retained the Calamites-Archaeocalamites-Equisetum triad and the trimerophyte-euphyllophyte clade, but identified the latter as emerging from a zosterophyllophyte-lycophyte constellation of taxa. It also failed to group the taxa within the Cooksonia-rhyniophyte plexus. The consensus tree using only anatomical characters was a complete polytomy (Fig. 3B). The topology of the 54-character tree also collapsed when reproductive and anatomical characters were analyzed together (Fig. 4A). However, it was retrieved in part when reproductive and morphological characters were combined, e.g., the trimerophyte-euphyllophyte clade and Calamites-Archeocalamites-Equisetum triad reemerged (Fig. 4B).

These analyses were interpreted to indicate that morphological characters were primarily responsible for the phylogenetic relationships hypothesized when using the new 54-character matrix.

DISCUSSION

The Silurian–Devonian time period cast a long shadow over subsequent vascular plant evolution (DiMichele et al., 2001; Stein et al., 2007; Toledo et al., 2018). Many of the anatomical, morphological, and reproductive features characterizing the sporophytes of modern-day plant lineages, such as the distinctive leaf morphology and stem anatomy of contemporary lycophytes as well as ferns, Equisetum, and other euphyllophytes, evolved during this period of plant evolution (summarized by Taylor et al., 2009). It was also a time when the seed plants made their first appearance in the fossil record (Rothwell et al., 1989). Indeed, by the end of the Devonian, the phenotypic sine qua non of almost every major tracheophyte lineage made its debut (Banks, 1975; Gensel and Andrews, 1984; Kenrick and Crane, 1997; Bateman et al., 1998; Elgorriaga et al., 2018). What remains largely unknown, however, is the relative extent to which different phenotypic features contributed to what is arguably the most rapid period in tracheophyte diversification.

Our analyses indicate that reproductive and anatomical sporophytic innovations played a less significant role during the Silurian-Devonian period of tracheophyte diversification than morphological features. Well before the seed plant habit made its first appearance by the end of the Devonian, our analyses show that very few character states distinguish among the reproductive or anatomical characteristics of the vascular land plants. The majority of reproductive organs consisted of simple isolated, aggregated, or fused sporangia, positioned either terminally or laterally on otherwise simple axes (Tomescu, 2009; Chomicki et al., 2017; Bonacorsi and Leslie, 2019). Even when features such as dimorphic spores (heterospory) and cone-like aggregates of sporangia (strobili) are included as character states, the number of reproductive features is less than that of morphological features, even when considering the best preserved and most completely preserved early tracheophytes (see Appendix 1). Indeed, in contrast to the relatively low diversity of reproductive features, the sporophytes of vascular plants by the end of the Devonian achieved a truly remarkable morphological diversity and well-defined organographic distinctions among roots, stems, and leaves (Gensel and Berry, 2001; Tomescu, 2009; Corvez et al., 2012; Crepet and Niklas, 2019). These vegetative innovations provided opportunities to occupy new niches (e.g., lianas and vines) and elevate photosynthetic and reproductive organs above potential competitors for space and nutrients.

Our analyses show that the failure of reproductive or anatomical features to resolve the phylogenetic relationships identified by the consensus 54-character tree is not a consequence of a small number of reproductive or anatomical features compared to a larger number of morphological features. For example, the topology of the consensus tree based exclusively on morphological characters complies reasonably well with that of the consensus tree using all 54 characters (compare Fig. 3A with Fig. 1). Yet, even when reproductive and anatomical characters are combined in an analysis, they fail to resolve any phylogenetic signatures (Fig. 4A). The opposite would be expected if phenotypic diversity was a simple matter of character-number or permutation. Consider that the number of different phenotypes that can appear as a result of n character permutations is given by the formula _nP_k = n!/(n – k)!, where k is the number of characters required to distinguish among phenotypes. Given that the combined number of anatomical and reproductive characters is n = 32, the number of different phenotypes defined by just three different anatomical character permutations equals _nP_k = 32!/(32 – 3)! = 29,760. In contrast, the number of different taxa distinguishable on the basis of three morphology character permutations equals _nP_k = 22!/(22 – 3)! = 9240. Clearly, the choice of three characters is arbitrary, and we do not suggest that the evolution of new taxa is the simple result of character permutation. It also might be argued that what matters is not the number of possible character permutations, but rather the number of useful synapomorphies relative to plesiomorphy and homoplasy, i.e., the issue is not to “distinguish among taxa”, but rather to assemble taxa on a correct tree. In response, our calculations show that the ability to distinguish among taxa should increase on average as the number of characters differentiating among taxa increases. Surely, the construction of a “correct tree” must surely depend on the ability to distinguish among taxa.

We note that anatomical characters taken in isolation (Fig. 3B) are completely ineffective at resolving the phylogenetic relationships among the fossil taxa selected for this study. Yet, these 14 characters retrieve elements of the phylogenetic relationship hypothesized in the 54-character tree when combined with morphological features (Fig. 2B). Among present-day plants, anatomical and morphological characters are generally highly correlated by virtue of the developmental regulatory systems giving rise to them (Chomicki et al., 2017). The analyses presented here indicate that anatomical-morphological correlations appear to hold true for the fossil taxa used in this study. For example, ancient fossil plants with simple primary vascular anatomies tend to be small and leafless, or possess comparatively simple leaf-like appendages, whereas ancient plant axes with more complex primary vasculatures (such as a tubular siphonostele) generally have larger and more complex leaf-like appendages (Tomescu, 2009; Corvez et al., 2012). There is every reason to believe that the evolution of the organographic distinctions among roots, stems, and leaves during the Silurian–Devonian time period resulted in similar anatomical-morphological interdependencies as those seen in present-day plants.

Why should the reproductive and anatomical features of the early sporophytes be so less useful than morphological features at resolving the phylogeny of ancient tracehophytes? We offer two observations in response to this question: (1) The sporophytes of early tracheophytes were likely reproductive bet-hedgers—they probably relied as much (or perhaps more) on asexual reproduction as on sexual reproduction (see Tiffney and Niklas, 1985; Niklas and Cobb, 2017), and (2) anatomically different hydraulic systems can be equally effective at delivering nutrients within comparatively simple plant body plans, particularly in the absence of secondary vascular tissues (see Niklas, 2013). Asexual propagation by means of extensive horizontal (rhizomatous) growth and fragmentation over a suitable substrate would have provided a competitive advantage in close-encounter vegetative confrontations, whereas the elevation of any suitable spore-producing structure above the surrounding vegetation would have favored long-distance spore dispersal and subsequent colonization of new sites (Tiffney and Niklas, 1985). In terms of hydraulics, most of the fossil taxa considered here have relatively simple vascular systems constructed out of primary tissues. From a nutrient-delivery perspective, we speculate that the efficiency of these systems was relatively insensitive to the shape of the stele or the location of protoxylem. For example, a centrarch haplostele is arguably as efficient at transporting water longitudinally as an exarch actinostele. Indeed, far more important are the number and diameter of their liquid-transport cells. If true, these speculations provide a reasonable explanation for the trends seen in consensus tree topologies, both with and without reproductive and anatomical characters.

Once again, it is essential to stress that every phylogenetic analysis is a hypothesis (sensu Stein et al., 1984; Rothwell and Nixon, 2006) and that the phylogenetic relationships illustrated in Fig. 1 are no exception. This topology is not presented as an attempt to formalize the phylogeny of early vascular land plants. The discovery of new taxa or the reassessment of the character states of previously well-studied fossils could alter, perhaps radically, our perception of early tracheophyte evolution (compare Fig. 1 with that in Appendix 3). Nevertheless, the topology of the 54-character tree is robust, even when critical characters are intentionally recoded (Crepet and Niklas, 2019). Based on this topology, the early Paleozoic appears to have been a period during which sporophyte phenotypic complexity and size evolved rapidly and largely as a consequence of morphological innovations and, with few exceptions, to a much lesser degree reproductive or anatomical invention.