THE EVOLUTION OF LEAVESThis update relates to a "theory" for the origin of leaves by Osborne et al. 2004, a long version of a 2001 paper in Nature. Osborne et al. worry about the fact that it took about 50 m.y. for vascular plants to evolve leaves. They suggest that the delay was caused by the inevitability that leaves, if they evolved earlier, would have fried.
Here is Carl Zimmer's blog on the paper, which is basically an elegant re-statement of their thesis.
Problems with the ideas of Osborne et al.As explained on p. 97, stomata on plant leaves take in carbon dioxide from the air to supply the photosynthesis that goes on inside the leaf spaces. Because they are open, stomata can lose water from the plant, in the form of water vapor that diffuses out as carbon dioxide diffuses in. The water vapor evaporates from the plant tissues inside the leaf. Living plants are often water-stressed, so they close the stomata when photosynthesis is not operating (at night, for example). But that is only a general observation, not a requirement.
Carbon dioxide levels were globally high in mid-Paleozoic times, say Osborne et al., so any proto-leaves would have needed comparatively few stomata to take in all the carbon dioxide they would have needed. As a result, proto-leaves would not have been evaporating much water. Evaporation cools a surface if fluid is evaporating from it (evaporation of sweat cools your skin if you sit in a breeze). A slow rate of evaporation would not cool a leaf very much. Osborne et al. seize on this biophysical fact to speculate that early vascular plants would have fried if they had evolved leaves: the added solar radiation shining on the leaf, combined with the lack of evaporative cooling, would have resulted in "lethal overheating," in their phrase.
Then 50 million years later, say Osborne et al., carbon dioxide levels and global temperatures dropped in the Late Devonian, making it finally possible for leaves to evolve without frying themselves. Leaves would now need many stomata to take in the carbon dioxide they needed, so they would have generated much more evaporative cooling; and they did. QED.
Obviously, I've simplified, but not much.
This construct depends absolutely on the assumption that leaves must close their stomata if and when they have taken in enough carbon dioxide for their photosynthesis, so they can no longer operate evapotranspiration for cooling. Any such closure would have had to relate to some reason connected with oversupply or overconcentration of carbon dioxide inside the leaf. This assumption is not discussed in Osborne et al. or in their previous paper in Nature.
What downside might there be to keeping the stomata open to promote evapotranspirative cooling? The only two possible results would be greater water loss, and greater intake of carbon dioxide.
- Water loss might be a problem for many plants living today, which often live in dry soils and dry air. But it was probably not important for early swamp-dwelling vascular plants.
- An oversupply of carbon dioxide might give mild acidification in the leaf spaces? However, carbon dioxide gets into the leaf spaces by diffusion, so, other things being equal, would never exceed atmospheric concentration. Even in the "high" carbon dioxide atmosphere of the Middle Devonian, these are not dangerous levels, even if the leaf cells continued to metabolize, thus producing carbon dioxide themselves. High carbon dioxide levels are apparently well tolerated by many plants today, so it is not much of a problem.
Therefore, on my simplistic assessment, the theory is based on assumptions so dubious that it cannot stand. I would have thought that the final invention of leaves and tree structure in the Late Devonian could have RESULTED in the carbon dioxide drop, rather than the reverse! Obviously it still took a long time for leaves to evolve, but I don't think we have a good answer yet. I'd be a poor botanist if all I did here was to criticize the idea of Osborne et al. without offering a better one: however, I am a poor botanist. Even so, I think I see a way through the problem, and I post it here for your enjoyment. I haven't done the necessary background work to polish it up, but I'll go with it until I do, or until you or other colleagues comment. Two colleagues who are plant physiologists have indicated that the remarks above are reasonable, but they are not to be held responsible for the rest of this essay, which asks:
Why Did Leaves Evolve?Leaves do more than photosynthesize. Essentially, they are extravagant expansions of the plant epidermis, increasing its surface area many times. Furthermore, they are typically studded with stomata, thus multiplying the gateways by which the leaves exchange gases with the atmosphere. The gases of most interest are carbon dioxide, oxygen, and water vapor, and. of course, all three are ingredients in the chemical reactions of photosynthesis. It's hardly surprising that Osborne et al. concentrated largely on photosynthesis as they worried about the origin of leaves.
Stomata encourage evapotranspiration, where water vapor evaporates from the plant. Osborne et al. concentrated on the evaporative cooling that is an automatic result of transpiration, and their model focussed on the inability, as they see it, for the stomata to cool leaves in the high carbon dioxide atmosphere of the Middle Devonian.
But there is another aspect of evapotranspiration, one that may be much more significant for the origin of leaves. Evaporation (transpiration) from stomata high on the plant sets up a pressure gradient, a pump, that produces upward flow in the xylem of the plant. Fluid in the xylem flows upward from the roots to the leaves, carrying essential nutrients with it. Without such flow there can be no growth of the tissues high in the plant. The pressure of the pump is the same whether there are 2 stomata or 20,000, but the flow that is generated multiplies with the number of stomata. Your car battery produces 12 volts, but you need a bigger battery (more amps) to start a farm tractor than you do to start a Ford Escort.
Suppose an early plant has stomata in its stem, but has no leaves. The stomata can evaporate water vapor and set up a pressure gradient, but the flow of the nutrients is going to be low, so growth rates must be low.
Now suppose the plant evolves protoleaves. The increased stomata increase the nutrient budget delivered to the top of the plant, so it can grow higher and faster, increasing the stomata as more protoleaves are added, and so on. At the same time, photosynthesis increases, so that there is an increased downward flow of carbohydrates to the rest of the plant. This does not produce a plant that is vastly different in basic structure from its ancestor (yet), but such a plant might have a distinct advantage in an environment where faster growth yields higher reproductive success.
My most important point is that the addition of leaves by itself will not accomplish much. Leaves in higher plants are part of a system. The origin of leaves would have evolved side by side with increased capacity of xylem and phloem, to accommodate the increased reciprocal flow of nutrient-bearing fluids up and down the plant. There would be no point in evolving leaves if the roots were not gathering and transmitting nutrients from the soil efficiently. And the whole structure of the stem would have been expanded and strengthened to accommodate the transport system, and to bear the weight of the added leaves and their support branches.
So what we see as the "evolution of leaves" is just a part of the evolution of an integrated yet differentiated higher plant, with increasingly different functions operating in increasingly different anatomical regions. [It's not my fault that calculus has pirated these words to describe manipulations of abstract functions: I'm talking about real aspects of real plants here.]
The next level of question is evolutionary and ecological, and it goes back to the original conundrum which Osborne et al. failed to explain: why was it 50 million years between the evolution of the first vascular plants and the elaboration of those plants that Osborne et al. see mainly as "the evolution of leaves"?
I can offer some suggestions based on my story as outlined above. The first vascular plants were small (low to the ground) and apparently lived in tropical swamps. In such environments, evapotranspiration doesn't work very well (humans sweat profusely if they exercise in hot humid environments, but the sweat doesn't evaporate easily and the human quickly overheats). In plants, slow evapotranspiration means low pressure gradients in the vascular system, low nutrient flow, and overall a small energy budget. Plants cannot grow high, and sophisticated transport systems are not needed, so don't evolve: and that includes size, strength, and complexity of phloem and xylem; size, strength and complexity of roots; and leaves.
If I were to formulate a decent research project to analyse this further, I'd look as carefully at the root system as Osborne et al. have looked at evapotranspiration and cooling. Here's why.
If leaves didn't evolve, [and Osborne et al. don't have the right answer], it's because some other parameter in early plants didn't (yet) make leaves worthwhile. Why would increased photosynthesis not be a good thing? My suggestion is that extra nutrition from photosynthesis doesn't do any good if the growth of the plant is handicapped by lack of soil nutrients. In other words, the root system wasn't working up to modern standards.
As long as early plants were low to the ground, and small, their nutrient budget would have been small, and neither the system for getting nutrients out of the soil nor the system for pumping it round the plant would have been sophisticated. In fact, the ability to form leaves, and the trunk, branches, and twigs that would supported them, could not have been financed. It takes a lot of energy to build plant structures before leaves are budded, even in today's seedlings.
Roots today operate an ion pump to take nutrients from the soil, put them into the root system, and keep them there. This process generates a pressure, called "root pressure", that can move the nutrients round the plant. In addition, many plants today have accessory colonies of fungi called "mycorrhizae" associated with the roots. Mycorrhizae concentrate nutrients from the soil and essentially drag them close to the roots, making the whole process cheaper for the plant: but the mycorrhizae in turn are "fed" by the plant. This is another "high-budget" operation, which costs more but gives a higher return.
I suggest, then, that the evolution of larger plants was initially fuelled by nutrient supply rather than photosynthesis. Once the nutrient supply was large enough and efficient enough, the plant could "afford" the investment in larger size (mainly height) that involved building stem (trunk), branches, and leaves to increase photosynthesis (interest income on the investment).
I would expect that Devonian root systems would show evolution toward an advanced state along with advances in all the other systems I have listed. There is evidence of fungus associated with plants as early as the Middle Devonian Rhynie Chert: it's not clear (to me) that they were performing as mycorrhizae yet, and I am not sure how we could tell that. The fungal association is a complex game between plant and fungus, and it may have begun as a parasitic association before the current mutualistic stand-off evolved. Gary Vermeij suggested to me that it might have been the evolution of the mycorrhizal symbiosis that triggered major land plant evolution late in the Devonian.
Why and when and how should this major change in Devonian plants have occurred? The answer has to be ecological. If plants were to colonize habitats out of the swamp, out on to marginally or seasonally drier areas, with better-drained soils, evapotranspiration would have been a more powerful agent able to drive fluids to greater heights. Plants could have grown higher, and would have evolved bigger pumps (leaves), better hydraulics for circulation (larger and more organized phloem and xylem), and better supplies of nutrients from below (from the roots) and of fuels from above (photosynthates from leaves). It's also possible that mycorrhizae dictate that their plants grow in soils that are not swampy (waterlogged).
As I have discussed, all this requires a larger, longer investment in plant growth. Plants would have become larger, more sophisticated, and longer-lived. There would have been a differentiation between adaptations of a seedling and adaptations of an adult (a much more pronounced life-cycle sequence). I expect there would have been evolutionary pressure to evolve a larger seed, to give a young plant a quick boost off the ground as it germinated. But my brain is tired, so I'll stop here.
I am arguing for an evolution of a whole-plant system rather than for an evolution of a piece of a plant (a leaf). Trying to think about the "origin of leaves" as an isolated problem is too simple, given that we are talking about a complex organism. It's like trying to talk about "the origin of the erect limbs of vertebrates" as a question all on its own, when it's really a question about stance, locomotion, energy, respiration, circulation, physiology, behavior, ecology, and evolution all at once. Changes in posture, locomotion, and respiration associated with changes in body temperature happened in synapsids leading to warm-blooded mammals, and in diapsids leading to warm-blooded dinosaurs and birds. [If you're a student who's just got this far in the book, you'll soon find out about this question: and it's a great story!]
Beerling, D. J. et al. 2001 Nature 410: 352-354.
Osborne, C. P., et al. 2004. Biophysical constraints on the origin of leaves inferred from the fossil record. PNAS 101: 10360-10362. Available on the Web.
First drafted by RC, October 5, 2004.
Updated December 27, 2004.
Links checked October 2, 2005.
Thanks to Terry Murphy, John Raven, and Gary Vermeij; but they fed me facts and principles, and any misuse of their help is my fault, not theirs.
Return to Chapter 8
Establishing the timescale of early land plant evolution is essential to testing hypotheses on the coevolution of land plants and Earth’s System. Here, we establish a timescale for early land plant evolution that integrates over competing hypotheses on bryophyte−tracheophyte relationships. We estimate land plants to have emerged in a middle Cambrian–Early Ordovocian interval, and vascular plants to have emerged in the Late Ordovician−Silurian. This timescale implies an early establishment of terrestrial ecosystems by land plants that is in close accord with recent estimates for the origin of terrestrial animal lineages. Biogeochemical models that are constrained by the fossil record of early land plants, or attempt to explain their impact, must consider a much earlier, middle Cambrian–Early Ordovician, origin.
Establishing the timescale of early land plant evolution is essential for testing hypotheses on the coevolution of land plants and Earth’s System. The sparseness of early land plant megafossils and stratigraphic controls on their distribution make the fossil record an unreliable guide, leaving only the molecular clock. However, the application of molecular clock methodology is challenged by the current impasse in attempts to resolve the evolutionary relationships among the living bryophytes and tracheophytes. Here, we establish a timescale for early land plant evolution that integrates over topological uncertainty by exploring the impact of competing hypotheses on bryophyte−tracheophyte relationships, among other variables, on divergence time estimation. We codify 37 fossil calibrations for Viridiplantae following best practice. We apply these calibrations in a Bayesian relaxed molecular clock analysis of a phylogenomic dataset encompassing the diversity of Embryophyta and their relatives within Viridiplantae. Topology and dataset sizes have little impact on age estimates, with greater differences among alternative clock models and calibration strategies. For all analyses, a Cambrian origin of Embryophyta is recovered with highest probability. The estimated ages for crown tracheophytes range from Late Ordovician to late Silurian. This timescale implies an early establishment of terrestrial ecosystems by land plants that is in close accord with recent estimates for the origin of terrestrial animal lineages. Biogeochemical models that are constrained by the fossil record of early land plants, or attempt to explain their impact, must consider the implications of a much earlier, middle Cambrian–Early Ordovician, origin.
The establishment of plant life on land is one of the most significant evolutionary episodes in Earth history. Terrestrial colonization has been attributed to a series of major innovations in plant body plans, anatomy, and biochemistry that impacted increasingly upon global biogeochemical cycles through the Paleozoic. In some models, an increase in biomass over the continents, firstly by cryptogamic ground covers followed by larger vascular plants, enhanced rates of silicate weathering and carbon burial that drove major perturbations in the long-term carbon cycle (1, 2), resulting in substantial drops in atmospheric CO2 levels (3⇓⇓–6) (but see ref. 7) and increased oxygenesis (8). It also led to new habitats for animals (9) and fungi (10), major changes to soil types (11), and sediment stability that influenced river systems and landscapes (12). Attempts at testing these hypotheses on the coevolution of land plants (embryophytes) and the Earth System have been curtailed by a lack of consensus on the relationships among living plants, the timescale of their evolution, and the timing of origin of key body plan innovations (13). Although the megafossil record provides unequivocal evidence of plant life on land, the early fossil record is too sparse and biased by the nonuniformity of the rock record (13) to directly inform the timing and sequence of character acquisition in the assembly of plant body plans. Therefore, in attempting to derive a timescale for phytoterrestrialization of the planet, we have no recourse but to molecular clock methodology, employing the known fossil record to calibrate and constrain molecular evolution to time. Unfortunately, the relationships among the four principal lineages of land plants, namely, hornworts, liverworts, mosses, and tracheophytes, are unresolved, with almost every possible solution currently considered viable (14). In attempting to establish a robust timeline of land plant evolution, here we explore the impact of these conflicting phylogenetic hypotheses on divergence time estimates of key embryophyte clades.
Early morphology-based cladistic analyses of extant land plants suggested that the bryophytes are paraphyletic, but yielded conflicting topologies (15⇓–17). Molecular phylogenies have been no more certain, with some analyses supporting liverworts as the sister to all other land plants (18), with either mosses (19⇓–21) (Fig. 1F), hornworts (22⇓⇓⇓⇓–27) (Fig. 1E), or a moss−hornwort clade (28) (Fig. 1G) as the sister group to the vascular plants. Variants on these topologies have been suggested, such as a liverwort−moss clade as the sister group to the remaining land plants (29) (Fig. 1D). More recently, the debate has concentrated upon two hypotheses: hornworts as the sister to all other land plants (14, 30⇓⇓⇓–34) (Fig. 1B) or monophyletic bryophytes sister to the tracheophytes (14, 35, 36) (Fig. 1A). Transcriptome-level datasets support both topologies (14), but sequence heterogeneity makes inferring relationships among these early land plants difficult (36).
The seven alternative hypotheses considered in the dating analyses. (A) Monophyletic bryophytes; (B) liverwort–moss sister clade to tracheophytes; (C) mosses, liverworts, and hornworts as successive sister lineages to tracheophytes; (D) a moss–liverwort sister clade to other embryophytes; (E) hornworts, mosses, and liverworts as successive sister lineages to tracheophytes; (F) mosses, hornworts, and liverworts as successive sister lineages to tracheophytes; and (G) a moss–hornwort sister clade to tracheophytes.
Here we attempt to establish a timescale of early land plant evolution that integrates over the contested topological relationships among bryophytes and tracheophytes. To achieve this, we constructed 37 fossil calibrations with minimum and soft maximum constraints, following best practice (37). This requires that calibrations are established on the basis of (i) a specific fossil specimen reposited in a publicly accessible collection, (ii) an apomorphy-based justification of clade assignment, (iii) reconciliation of morphological and molecular phylogenetic context of clade assignment, (iv) geographic and stratigraphic provenance, and (v) justification of geochronological age interpretation. Thus defined, these calibrations were combined with existing genetic data (14) in a Bayesian relaxed molecular clock analysis in which we also explored the impact of genetic dataset size and competing calibration strategies, as well as alternative substitution models, on divergence time estimates (Table 1). We find that topology and dataset size have minimal impact on age estimates, but slightly more variance in clade age estimates occurred when using alternative calibration strategies. We conclude that embryophytes emerged within a middle Cambrian to Early Ordovician interval and, regardless of topology, all four major lineages of land plants had diverged by the late Silurian. These dates are older than those used in the latest biogeochemical models (6, 8), and thus our results have implications for simulations of atmospheric chemistry and climate during the Paleozoic.
The competing hypotheses of relationships among bryophytes and tracheophytes all produce congruent age estimates across the phylogeny (Fig. 2 and Tables 2 and 3). Age estimates of key nodes (Embryophyta, Tracheophyta) are very similar regardless of the underlying topology (Fig. 2 and Tables 2 and 3). At the full range of uncertainty across topologies, the 95% highest posterior density (HPD) of ages for the embryophyte node ranges from the mid-Cambrian (Series 2; 515.2 Ma) to Early Ordovician (473.5 Ma) (Table 2), with the bulk of the distributions in the Cambrian (Fig. 2). There is a slightly higher variance in the estimated age of tracheophytes between the different topologies, but there is overlap in all of the 95% HPD age ranges (Fig. 2 and Tables 2 and 3). Estimates for the age of crown tracheophytes range from Late Ordovician (Katian; 450.8 Ma) to the latest Silurian (419.3 Ma).
Age estimates for the seven topologies used in analyses, highlighting the 95% HPD age uncertainty for embryophytes and tracheophytes. Age estimates are shown for (A) monophyletic bryophytes, (B) hornworts−sister, (C) hornworts−liverworts−mosses, (D) liverworts−mosses−sister, (E) liverworts−mosses−hornworts, (F) liverworts−hornworts−mosses, and (G) liverworts−sister.
The two main hypotheses of early land plant relationships (monophyletic bryophytes and hornworts-sister) give congruent estimates for all nodes across the tree (Fig. 3 and Table 3). For example, the age estimates based on the two topologies are similar for Viridiplantae (972.4 Ma to 669.9 Ma), Streptophyta (890.9 Ma to 629.1 Ma), and Angiospermae (246.6 Ma to 195.4 Ma).
Detailed phylogenies showing the congruent age estimates produced using the monophyletic (A) and hornworts−sister (B) topologies.
The 95% HPD age estimates for of embryophytes and tracheophytes from divergence time analyses using the seven alternative topologies
The 95% HPD age estimates for named nodes in the analyses using the two main topologies of early land plants (monophyletic, hornworts−sister)
Infinite site plots describe the relationship between clade age and uncertainty (95% HPD of clade age estimates). As the volume of sequence data increases, it is anticipated that clade age estimates should converge on a straight line, with residual dispersion reflecting uncertainty in calibrations that cannot be overcome by additional sequence data (38). We explored the impact of dataset size based on the monophyletic bryophytes topology, trimming the original dataset (1.7 million nucleotides) based on taxon completeness by 50%, 75%, 99%, and 99.9%. As expected, the resulting infinite sites plots reveal greater uncertainty (<R2) associated with the smallest datasets (Fig. 4) and greatest disparity between the smallest and largest datasets (SI Appendix, Fig. S5). However, these differences are small, and, generally, the infinite sites plots indicate that the clade age estimates are effectively insensitive to three orders of difference in the number of nucleotides used in the analysis.
Infinite site plots showing the effects of including more sequence data on the precision of age estimates. All ages are plotted using the monophyletic bryophytes topology with (A) datasets including all sites, and datasets trimmed so sequences are complete for (B) 50%, (C) 75%, (D) 95%, and (E) 99.9% of taxa.
Across all alternative dating strategies, the age estimate for crown Embryophyta ranges from 583.1 Ma to 470.0 Ma (Fig. 5 and Table 4), which is larger than the range across the different topologies (515.2 Ma to 473.5 Ma). The greatest variance is seen when the embryophyte constraint is removed, resulting in older age estimates in the hornworts–sister topology, with an age distribution that stretches into the Proterozoic (to the middle Ediacaran), compared with the bulk of the distributions that fall within the Cambrian for all other age estimates (Fig. 5).
The estimated ages of embryophyte and tracheophyte divergence is more variable due to differences in modeling compared with differences in dataset size or topology. Using the monophyletic topology, the impact on age estimation was tested by using alternative strategies to model substitution rates, age constraints, and by excluding outgroups. An asterisk (*) denotes analysis performed on hornworts−sister topology.
We employed different parametric distributions (uniform, Cauchy, skew-t) to express the prior probability of divergence timing relative to the minimum and soft maximum constraints. This often has a dramatic impact on divergence time estimates (39⇓–41); however, different prior distributions have minimal impact on age estimates for embryophytes. The largest difference is seen with the younger age estimates produced using the skew-t distribution (Fig. 5), but both the skew-t and Cauchy models produce younger mean estimates for embryophytes compared with the uniform distribution (Fig. 5). Similarly, there is a younger estimated age for tracheophytes with the skew-t and Cauchy models compared with the uniform distribution (Fig. 5). The age of the tracheophyte node ranges from 472.2 Ma to 422.4 Ma across all alternative dating strategies.
95% HPD age estimates for embryophytes and tracheophytes in analyses after removing all nonembryophyte lineages, employing a correlated clock model, and applying different strategies for the shape of prior node age constraints (uniform unless stated)
Our results demonstrate that divergence time analyses of early land plant evolution are largely insensitive to tree topology and dataset size; however, they show some sensitivity to calibration strategy and, in particular, the calibration on crown Embryophyta. This clearly demonstrates the informative nature of the calibration on crown Embryophyta, which is comparatively narrow in its temporal range (515.5 Ma to 469.0 Ma). The soft maximum constraint on the age of this clade is based on the maximum age of the oldest-possible nonmarine palynomorphs, encompassing all possible total-group embryophyte records (SI Appendix). Land plant spores are encountered commonly among marine palynomorph assemblages, and they have the same fossilization and sampling potential as acritarchs. However, the oldest-possible embryophyte records are preceded stratigraphically by thick sequences bearing only marine palynomorphs. These marine palynomorphs demonstrate that the conditions required for preserving embryophyte remains obtained and, thus, the absence of land plant spores constitutes evidence that embryophytes were not present at this time (42). Thus, we discount the results of the divergence time analyses in which the embryophyte calibration is not employed. Similarly, the skew-t and Cauchy distributions, which reflect a nonuniform probability of divergence timing between the minimum and maximum constraints, suggest younger clade ages. However, these nonuniform distributions are unduly informative, since we have no insight or additional evidence that might inform the probability of the time of divergence between minimum and maximum constraints. Hence, we reject the ensuing results in favor of those based on a uniform distribution which reflects equal probability of divergence timing between minimum and maximum constraints. Since the remaining sources of uncertainty have little impact, a holistic timescale encompassing all relevant uncertainties is, effectively, that represented in Fig. 2. It is difficult to foresee how higher precision can be achieved while also maintaining accuracy. We have shown that additional sequence data and topological uncertainty have little material impact, both perhaps as a consequence of the short temporal succession of clade divergences among early embryophytes and attendant issues such as incomplete lineage sorting. Improved taxon sampling among liverworts and hornworts (especially) is likely to yield more precise estimates for divergences among bryophytes on some topologies, as would improved sampling of their fossil record—which our analyses predict to extend deep into the Lower Paleozoic.
It is possible that a Total Evidence approach (43), integrating living and fossil species, both morphological and molecular data and evolutionary models, will leverage some increased precision. Perhaps more importantly, such an approach might provide a means of more precisely dating the origin of land plant body plan innovations (e.g., stomata, leaves, rooting systems) that have been considered influential in the evolution of the Earth System (44). In the interim, our evolutionary timescale achieves precision while also integrating all of the principal sources of uncertainty, providing a framework for inferring plant evolutionary history, the veracity of its fossil record, and the impact of phytoterrestrialization on the evolution of global biogeochemical cycles.
The Origin of the Embryophytes and Tracheophytes.
Considering the 95% HPDs of divergence times across all topologies, the origin of crown embryophytes is dated to 515.1 Ma to 470.0 Ma (middle Cambrian–Early Ordovician). However, all of the mean estimated ages are resolved within the Phanerozoic across all alternative topologies and dating strategies, and the majority are dated to around 500 Ma (middle Cambrian Series 2). Only one analysis has a 95% HPD that stretches into the Proterozoic. The full span of age estimates for the crown tracheophyte node is 472.2 Ma to 419.3 Ma (Floian, Early Ordovician to the late Silurian). Only one analysis has a 95% HPD that stretches to the Early Ordovician, with those using a uniform prior resulting in estimated mean ages close to the Ordovician−Silurian boundary (∼444 Ma). The span of the tracheophyte stem lineage ranges across all analyses from 25.1 My to 60.0 My; these intervals are shorter for the paraphyletic topology than the monophyletic bryophytes topology (35.5 My and 51.6 My, respectively) (SI Appendix, Fig. S6).
Impacts of Alternative Topologies and Dating Strategies on Divergence Time Estimates.
The impact of analytical uncertainty on the estimated age of Embryophyta is minimized by the use of carefully selected temporal information from the fossil record. Differences in topology had a minimal impact on divergence time estimates for Embryophyta (Fig. 5 and Table 2). For each topology, the posterior age estimates conform largely to the specified calibration constraints on clade age (∼511 Ma to 469 Ma). Potential differences in age estimates for embryophytes only appear when the specified age constraint for this node is removed. On the hornworts–sister topology, age estimates for Embryophyta extend into the Proterozoic without the embryophyte calibration, whereas the monophyletic bryophytes topology yields congruent age estimates with or without the user-applied embryophyte age constraint (Fig. 5). Thus, topology can influence the estimated ages for nodes, but only when we ignore germane evidence from the fossil record. Therefore, the use of well-researched and justified fossil constraints, when incorporated alongside tests of model uncertainty, adds confidence in the conclusion of an Early Phanerozoic origin for embryophytes.
There are only minor differences across topologies for the estimated age of tracheophytes, as all trees produce comparable mean estimates (Table 2). One topology, hornworts−liverworts−mosses, produces a younger age from the 95% HPD interval (419 Ma) compared with all other trees (430 Ma), but this younger age is anomalous (i.e., slightly younger than the minimum derived directly from fossil evidence at 420.7 Ma) and has little overall support; the bulk of the posterior age of tracheophytes for the hornworts−liverworts−mosses tree is above 430 Ma.
Comparisons with the Fossil Record.
The first unequivocal embryophyte body fossil taxon, Cooksonia cf. pertoni, appears in the Wenlock [minimum age of 426.9 Ma (45)]. The first account of crown tracheophyte body fossils is shortly after, in the Ludlow [minimum age of 420.7 Ma (46)], followed by an apparent explosion of diversity in the Early Devonian (13). Our mean age estimates are older for both nodes, by 40 My for the embryophytes and 20 My for crown tracheophytes. However, in both cases, this is a consequence of a dearth of continental lithofacies before the late Silurian−Early Devonian (47). The earliest known fossils of embryophyte affinity are permanently fused tetrahedral tetrad cryptospores [sensu stricto Steemans (48), Wellman (49)] that have a long history of occurrences within marine deposits (13) from the Middle Ordovician [Dapingian; 469 Ma (50)]. Cryptospores of unclear affinity from the Cambrian [sensu stricto Strother (51)], while not considered unequivocally embryophyte, informed our soft maximum constraint (515.5 Ma). Our middle Cambrian−Early Ordovician estimate for the origin of crown embryophytes is compatible with an embryophyte interpretation; however, our results do not suggest that they reflect a protracted cryptic earlier evolutionary history. Likewise, the dispersed record of trilete spores that first appear in the Katian (Late Ordovician) (52), followed by an explosion of diversity in the Silurian (13), indicates an earlier origin for tracheophytes that is congruent with our estimates.
The main challenge in testing our divergence time estimates for the bryophyte lineages is their very poor representation in the rock record (13). Nevertheless, our results establish a predictive temporal framework for the stratigraphic intervals in which to prospect for fossils implied by the ghost lineages in our evolutionary timescale. Regardless of the topology, we date the first and second divergences within the bryophytes between 496.5 Ma and 456.2 Ma (late Cambrian–Late Ordovician) and 478.7 Ma and 438.0 Ma (Early Ordovician–early Silurian), respectively. The oldest credible candidate bryophyte fossil is the Pragian (Early Devonian) Riccardiothallus devonicus (53), although the security of its classification is limited by preservation of only gross morphology. The mismatch between the estimated ages and unequivocal fossil finds is contributed to by their low fossilization potential, principally because bryophytes do not biosynthesize lignin. When body fossils occur, they are often too poorly preserved to allow recognition of synapomorphies. However, some extant bryophytes produce permanent tetrads and dyads (54, 55) similar to the cryptospores. The wall ultrastructure of cryptospores, known from as early as the Middle Ordovician, is similar to the multilaminate walls observed in permanent tetrads produced by extant liverworts, such as Sphaerocarpos (56). The presence of liverwort-like spores in the Middle Ordovician is not incongruent with the estimated dates of divergence of the liverworts across all topologies in our analyses. Sporangia described from the Late Ordovician of Oman are significant fragments of plant anatomy recovered from very rare instances of nonmarine Ordovician rocks (57). The spore masses contain either dyads or tetrads, the former displaying multilaminate walls, and most specimens preserve at least a partial covering, making it very difficult to argue that they are anything but land plant sporangia (57, 58). Unfortunately, our understanding of the parent plants of cryptospores, the cryptophytes, is restricted to much later charcoal Lagerstätten in the Pridoli and Lochkovian (59, 60). These fossil plants possess a combination of both bryophytic and tracheophytic characters, and thus their taxonomic position is currently unclear (60). The confirmation of the main synapomorphy for the tracheophytes, the presence of vascular tissues, is particularly difficult to demonstrate, due to the minute size and fused nature of these fossils.
Theories on the process of terrestrialization have long argued for a close temporal relationship between the emergence of land plants and terrestrial animals, particularly arthropods, substantiated by their approximately concurrent first fossil occurrence in terrestrial facies (61, 62). However, this is likely an instance of pseudocongruence, with lineages of differing antiquity exhibiting coeval stratigraphic first occurrences because of secular variation in the preservation of Lower Paleozoic terrestrial facies (40). Thus, a shift from dominantly marine to terrestrial facies results in a telescoped first stratigraphic appearance of disparate terrestrial lineages (63). The results of our divergence time analyses indicate a much earlier (∼70 My to 80 My) origin of land plants, but, surprisingly, this remains congruent with the latest divergence time estimates for three or four independent transitions to terrestrialization among arthropod lineages (hexapods, arachnids, and, perhaps, twice among myriapods) (64). Thus, although our results corroborate the view that the early fossil records of terrestrial arthropods and land plants are temporally misleading, they also corroborate the hypothesis of a close temporal relationship between the emergence of land plants and terrestrial animals, with plants creating habitats suitable for terrestrial arthropods.
Comparisons with Previous Studies.
Previous analyses indicate either a Proterozoic (mainly Cryogenian) (65⇓–67) or Phanerozoic (68⇓–70) origin of the embryophytes. Of the latter, dates range from the Early Ordovician [∼474 Ma to 477 Ma (69, 70)] to early Silurian [435 Ma to 425 Ma (68)]. The majority of our results are congruent with a Phanerozoic origin, but with older estimated ages (middle Cambrian; Fig. 5
Summary of the analyses performed employing the seven alternative hypotheses, removal of the embryophyte constraints, and trimming dataset size