/n nature ecology & evolution 8 Article https://doi.org/10.1038/s41559-022-01885-x Divergent evolutionary trajectories of bryophytes and tracheophytes from a complex common ancestor of land plants Received: 8 November 2021 Accepted: 12 August 2022 Published online: 29 September 2022 Check for updates Brogan J. Harris 16, James W. Clark 1,2,6, Dominik Schrempf³, Gergely J. Szöllősi ①3,4,5, Philip C. J. Donoghue 2, Alistair M. Hetherington 1 and Tom A. Williams 1,2 The origin of plants and their colonization of land fundamentally transformed the terrestrial environment. Here we elucidate the basis of this formative episode in Earth history through patterns of lineage, gene and genome evolution. We use new fossil calibrations, a relative clade age calibration (informed by horizontal gene transfer) and new phylogenomic methods for mapping gene family origins. Distinct rooting strategies resolve tracheophytes (vascular plants) and bryophytes (non-vascular plants) as monophyletic sister groups that diverged during the Cambrian, 515-494 million years ago. The embryophyte stem is characterized by a burst of gene innovation, while bryophytes subsequently experienced an equally dramatic episode of reductive genome evolution in which they lost genes associated with the elaboration of vasculature and the stomatal complex. Overall, our analyses reveal that extant tracheophytes and bryophytes are both highly derived from a more complex ancestral land plant. Understanding the origin of land plants requires tracing character evolution across a diversity of modern lineages. The origin and early evolution of land plants (embryophytes) consti- tuted a formative episode in Earth history, transforming the terrestrial landscape, the atmosphere and the carbon cycle¹². Along with bac- teria, algae, lichens and fungi³, land plants were fundamental to the creation of the earliest terrestrial ecosystems, and their subsequent diversification has resulted in more than 370,000 extant species*. Embryophytes form a monophyletic group nested within freshwater streptophyte algae and their move to land, while providing a new eco- logical niche, presented new challenges that required adaptation to water loss and growth against gravity6. Early innovations that evolved in response to these challenges include a thick waxy cuticle, stomata and a means of transporting water from the roots up vertically growing stems 2,5,7,8. Modern land plants comprise two main lineages, vascular plants (tracheophytes) and non-vascular plants (bryophytes), that have responded to these evolutionary challenges in different ways. The evolutionary origins of many gene families, including those of key transcription factors, have been shown to predate the coloniza- tion of land 9,10. However, studies of gene family evolution within land plants have typically been restricted to individual gene families or sets of genes that encode single traits 11-16. A lack of genome-scale data from non-flowering plants has also hindered efforts to reconstruct pat- terns of genome and gene content evolution more broadly across land plants, although this challenge has been mitigated by the publication of large transcriptomic datasets 18. Progress has also been made towards resolving the ambiguous phylogenetic relationships at the root of land plants 15,18-23. The bryophyte fossil record has also undergone a radical ¹School of Biological Sciences, University of Bristol, Bristol, UK. 2Bristol Palaeobiology Group, School of Earth Sciences, University of Bristol, Bristol, UK. ³Department of Biological Physics, Eötvös Loránd University, Budapest, Hungary. *MTA-ELTE 'Lendület' Evolutionary Genomics Research Group, Budapest, Hungary. 5 Institute of Evolution, Centre for Ecological Research, Budapest, Hungary. These authors contributed equally: Brogan J. Harris, James W. Clark. e-mail: tom.a.williams@bristol.ac.uk Nature Ecology & Evolution | Volume 6 | November 2022 | 1634-1643 1634 Article reinterpretation such that there are now many more records with the potential to constrain the timescale of early land plant evolution²4-26. Finally, new methods have been developed for timetree calibration based on the relative time constraints informed by horizontal gene transfer (HGT) events²7. Here we seek to exploit these advances in elucidating early land plant evolution. We first infer a rooted phylogeny of land plants using outgroup-free rooting methods and both concatenation and coales- cent approaches. We then estimate an updated timescale of land plant evolution incorporating densely sampled fossil calibrations that reflect a revised interpretation of the fossil record. We extend this analysis using gene transfer events to better calibrate the timescale of hornwort evolution, a poorly constrained region of the land plant tree. By build- ing on this dated phylogeny, we reconstruct the gene content evolution of bryophytes, tracheophytes and the ancestral embryophyte, reveal- ing how key genes, pathways and genomes diverged during early land plant evolution. Results Complementary rooting approaches support the monophyly of bryophytes A rooted phylogenetic framework is required to infer the nature of the ancestral embryophyte and to trace changes in gene content during the evolution of land plants. To that end, we compiled a com- prehensive dataset of the published genome and transcriptome data from embryophytes and their algal relatives, and we inferred species trees using concatenation (PhyloBayes and IQ-TREE) and coalescent (ASTRAL) approaches (Supplementary Information). When the tree was rooted with an algal outgroup, we recovered bryophyte mono- phyly and a root between bryophytes and tracheophytes with high support across all analyses (Extended Data Fig. 1), in agreement with recent work 15,18,20,22,23,28. However, rooting phylogenies with an out- group can influence the ingroup topology due to long-branch attraction (LBA)29-31, where distantly related or fast-evolving taxa artifactually branch with the outgroup. LBA resulting from the large evolutionary distance between land plants and their algal relatives has previously been suggested as a possible cause of the difficulty in resolving the land plant phylogeny³32. Indeed, outgroup-rooting analyses using different models 20,33, datasets and molecules (that is, chloroplast, mitochon- drial or nuclear sequences 22,28) have provided support for conflicting hypotheses about the earliest-branching lineages and the nature of the ancestral land plant. LBA is thus a known artefact when recovering the land plant phylogeny. To address the impact of LBA and complement traditional outgroup-rooting analyses, we used two outgroup-free rooting meth- ods-amalgamated likelihood estimation (ALE) and STRIDE 34,35-to infer root placement on a dataset of 24 high-quality embryophyte genomes without the inclusion of an algal outgroup (Fig. 1). ALE calculates gene family likelihoods for a given root position under a model of gene dupli- cation, transfer and loss (DTL)³4; support for candidate root positions can then be evaluated by comparing their summed gene family likeli- hoods. STRIDE first identifies putative gene duplications in unrooted gene trees that can act as synapomorphies for post-duplication clades. The root of the species tree is then estimated using a probabilistic model that accounts for conflict among the inferred duplications³5. Across 18,560 orthogroups, STRIDE recovered three most parsimo- nious roots: between bryophytes and tracheophytes, between liv- erworts and the remaining land plants and between hornworts and the remaining land plants (Fig. 1). Of these, the rooting on hornworts was assigned a 0.2% probability, on liverworts a 59.8% probability and between bryophytes and tracheophytes a 39.9% probability. To estimate root likelihoods using the ALE approach, we first used the divergence time estimates from the molecular clock analysis to convert branch lengths into units of geological time, allowing us to perform time-consistent reconciliations (that is, to prevent reconciliations in Nature Ecology & Evolution | Volume 6 | November 2022 | 1634-1643 https://doi.org/10.1038/s41559-022-01885-x which gene transfers occur into the past). We reconciled 18,560 gene families under the 12 rooted and dated embryophyte trees (Fig. 1a) and used an approximately unbiased (AU) test (Fig. 1b) to evaluate support for the tested root positions. The AU test rejected 9 of 12 roots (P<0.05; Fig. 2b and Supplementary Table 3), resulting in a credible set of three roots: the hornwort stem, the moss stem and a root between bryophytes and tracheophytes. These three credible roots are in close proximity on the tree, and root positions further from this region are rejected with increasing confidence (Fig. 1b and Supplementary Table 1). To evaluate the nature of the root signal for these three branches, we performed a family-filtering analysis in which families with high DTL rates were sequentially removed and the likelihood re-evaluated. The rationale for this analysis is that the evolution of these families may be poorly described by the model, and so they may contribute misleading signals 36. In this case, the root order did not change after the removal of the high-DTL-rate families (Supplementary Fig. 1), suggesting broad support for these root positions from the data and analysis. Note that, in the ALE analysis, the moss and hornwort stems were accorded a higher summed gene family likelihood than was the branch separating bryophytes and tracheophytes, although the difference was not signifi- cant (hornwort stem log-likelihood,-824,522.9, P=0.624; moss stem log-likelihood, -824,606.5, P = 0.475; bryophyte stem log-likelihood, -824709.1, P = 0.277). In a secondary analysis, we also used ALE to compare support for these different root positions in a smaller dataset of 11 genomes that included algal outgroups; in this analysis, all roots were rejected except for a root between tracheophytes and bryophytes (Extended Data Fig. 2, P < 0.05). Finally, we constrained the topology of the tree inferred from the concatenated alignment to be in accordance with the three credible roots and computed the likelihood of sequence data along those trees. Trees with embryophyte roots constrained to hornworts and moss were significantly rejected (P<0.05, AU test; Supplementary Table 2). The agreement between three rooting methods using different sources of information (outgroup placement, gene duplications alone and DTL events more broadly) therefore provides the most compelling support for a root between bryophytes and tracheophytes from our analyses. Taking our analyses together with other recent work 15,20,22,23,28, suggests that a root between monophyletic tracheophytes and bryophytes is the best-supported hypothesis of land plant phylogeny. Bryophyte monophyly is therefore the default hypothesis with which to interpret land plant evolution. Combined fossil and genomic evidence, including an ancient HGT, calibrate the timescale of land plant evolution We estimated divergence times on the resolved land plant phylogeny (Fig. 2). We assembled a set of 68 fossil calibrations, representing every major lineage of land plant and notably sampling more bryophyte fossils than previous studies (Supplementary Methods). Despite this increased sampling, the fossil record of hornworts remains particularly sparse, and no fossils unambiguously calibrate the deepest branches within the clade. To ameliorate the limitations of the fossil record, we implemented a relative node age constraint based on the horizontal transfer of the chimaeric photoreceptor NEOCHROME from hornworts into ferns 37. To account for uncertainty in the timing of the gene transfer, we evaluated the impacts of several possible scenarios on our analy- ses (Extended Data Fig. 3). In the absence of direct fossil calibrations for hornworts, this gene transfer provides a relative constraint that ties the history of hornworts to that of ferns, for which more fossils are available. Our results are congruent with those of previous studies 38 but offer greater precision on many nodes and in some cases greater accuracy (Supplementary Fig. 2). This has been leveraged by a denser sampling of fossil calibrations, improved taxonomic sampling (especially among bryophytes), relative calibration of hornworts using the NEOCHROME HGT, and the ability to condition divergence times on a single topology. 1635 Article a Syntrichia Ceratodon Fontinalis Lineages Algae Bryophytes Physcomitrium 12 Hornworts Moss 8 Liverworts Lycophytes Anthoceros punctatus Anthoceros agrestis 11 9 Ferns 2 Gymnosperms Angiosperms Ceratophyllum Magnolia Cinnamomum Dioscorea Brachypodium Kalanchoe Arabidopsis Selaginella lepidophylla Selaginella moellendorfii Amborella Nymphaea Azolla Salvinia Ginkgo b https://doi.org/10.1038/s41559-022-01885-x Name Root Rank STRIDE AU Rejected Hornwort 9 1 0.2% 0.624 No Moss 12 2 0.0% 0.475 No Monophyletic bryophytes 8 3 39.9% 0.277 No Setaphyte 10 4 0.0% 0.001 Yes Between fern and gymnosperm 4 5 0.0% 9 x 10-50 Yes Lycophyte 7 6 0.0% 0.0001 Yes Liverwort 11 7 59.8% 0.002 Yes Fern 5 8 0.0% 0.055 Yes Angiosperm 2 9 0.0% 7 x 10-10 Yes Between lycophyte and fern 6 10 0.0% 1 x 10-130 Yes Gymnosperm 3 11 0.0% 3 × 10-46 Yes Arabidopsis 1 12 0.0% 3 x 10-9 Yes Gnetum Pinus Pseudotsuga с Monophyletic bryophytes Mosses Hornworts Fig. 1 Investigating the root of embryophytes using outgroup-free rooting. a, An unrooted maximum likelihood tree was inferred from an alignment of 24 species and 249 single-copy orthogroups under the LG + C60+ G4 + F model 69. Twelve candidate root positions for embryophytes were investigated using both ALE and STRIDE. For the ALE analysis, the unrooted tree was rooted in each of the 12 positions and scaled to geological time on the basis of the results of the divergence time analysis, and 18,560 gene clusters were reconciled using the ALEml algorithm88. The green circles highlight supported roots following the ALE analysis, while the red circles denote supported nodes in the STRIDE analysis. b, The likelihood of the 12 embryophyte roots was assessed with an AU test. The AU test significantly rejected 9 of the 12 roots, with roots on hornworts, moss and monophyletic bryophytes (root positions 9, 12 and 8, respectively) comprising the credible set. c, Phylogenetic trees constrained to the credible roots were inferred in IQ-TREE69 under the LG + C60+ G + F model. An AU test was used to evaluate the likelihood of each of the constrained trees⁹0, with the root resulting in monophyletic bryophytes being the only one not to be significantly rejected. The role and influence of fossil calibrations in molecular clock studies, especially maximum age calibrations, remain controver- sial 23,39,40. While the fossil record is an incomplete representation of past diversity, our analyses account for this uncertainty in the form of soft minima and maxima. Morris et al.38 inferred a relatively young age for the embryophyte crown ancestor (515-470 million years ago (Ma)), making use of a maximum age constraint based on the absence of embryophyte spores in strata for which fossilization conditions were such that spores of non-embryophyte algae have been preserved. Hedges et al.39 and Su et al. 2³ argued against the suitability of this maxi- mum age constraint on the basis that calibrations derived from fossil absences are unreliable and that the middle Cambrian maximum age exerts too great an influence on the posterior estimate 8,41. To assess the sensitivity of our approach to the effect of maximum age calibra- tions, we repeated the clock analyses with less informative maximum age calibrations (Supplementary Methods). Removing the maximum age constraint on the embryophyte node produced highly similar estimates to when the maximum is employed (Extended Data Fig. 4). Relaxing all maxima did result in more ancient estimates for the origin of embryophytes, although still considerably younger than recent studies23, extending the possible origin for land plants back to the Edi- acaran (540-597 Ma; Extended Data Fig. 4). The older ages estimated Nature Ecology & Evolution | Volume 6 | November 2022 | 1634-1643 in Su et al. 23 seem to reflect, in part, differences in the phylogenetic assignment of certain fossils (Supplementary Methods), such as the putative algae Proterocladus antiquus and the liverwort Ricardiothallus devonicus, rather than a dependence on the maximum age calibration. Our results reject the possibility that land plants originated during the Neoproterozoic, instead supporting an origin of the land plant crown group during the mid-late Cambrian, 515-493 Ma, with crown tracheophytes and crown bryophytes originating 452-447 Ma (Late Ordovician) and 500-473 Ma (late Cambrian to Early Ordovician), respectively. Within bryophytes, the divergence between Setaphyta (mosses + liverworts) and hornworts occurred by 479-450 Ma (Ordovi- cian), with the radiation of crown mosses by 420-364 Ma (latest Silurian to Late Devonian) and crown liverworts 440-412 Ma (early Silurian to Early Devonian). Among tracheophytes, the crown ancestor of lyco- phytes is dated to the middle Silurian to Early Devonian, 431-411 Ma, coincident with that of euphyllophytes 432-414 Ma. The calibration of hornwort diversification using the NEOCHROME HGT had a substantial impact on inferences of stem and crown group age. In the absence of fossil calibrations on deep nodes, hornworts are characterized by an ancient stem lineage and the youngest crown line- age among land plants 38,42. The effect of the relative age constraint is to make the crown group older (294-214 Ma; Fig. 2) and thus shorten the 1636 Bryophytes Hor. Mosses Liverworts Lyco. Article https://doi.org/10.1038/s41559-022-01885-x Phragmoplastida Embryophyta Ferns Tracheophytes Euphyllophytes Spermatophytes Tu. Cr. Ed. Ca. Or. Si. De. Ca. Pe. Tr. Ju. Cr. Pa. N. 860 720 635 541 485 443 419 359 299 252 201 145 66 23 0 Time (millions of years before present) Fig. 2 The timescale of land plant evolution. Divergence times in millions of years as inferred using a molecular clock model, 68 fossil calibrations and an HGT. The inference that the common ancestor of embryophytes lived during the Cambrian is robust to the choice of maximum age constraints (Supplementary Methods). The divergence times of hornworts are constrained by an HGT into polypod ferns, with the result that the hornwort crown is inferred to have diverged during the Permian-Triassic. The nodes are positioned on the mean age, and the bars represent the 95% highest posterior density. length of the stem, with divergence times within the crown group all moving older. We repeated the analysis with alternative placements for the relative time constraint, with the age of crown hornworts becoming increasingly ancient when the transfer was placed into the ancestor of more inclusive clades, Cyatheales + Polypodiales (258-419 Ma) or before the divergence of Gleicheniales from the Cyatheales + Polypo- diales clade (331-445 Ma), respectively (these scenarios are illustrated in Extended Data Fig. 3). All of these estimates considerably predate the earliest unequivocal fossils assigned to hornworts. However, given the scarcity of hornwort fossils, it seems likely that this clade is older than a literal reading of the fossil record might suggest. Gene content of the embryophyte common ancestor We used gene-tree/species-tree reconciliation to estimate the gene content of the embryophyte common ancestor (Supplementary Tables 3-5). We used the genome dataset from the ALE rooting analysis with the addition of five algal genomes, to better place the origin of families that predate the origin of embryophytes (Supplementary Fig. 3). The tree was dated following the same methodology as the larger dating analysis while using an applicable subset of calibrations, allowing the use of a dated reconciliation algorithm (ALEml) to improve the estima- tion of DTL events (Supplementary Fig. 4). The analysis of ancestral gene content highlighted considerable gene gain along the ancestral embryophyte branch (Fig. 3a and Supple- mentary Table 3). A substantial number of duplications defined this tran- sition, with fewer transfers and losses observed. Our analysis suggests that the common ancestor of embryophytes and Zygnematales had more of the building blocks of plant complexity than extant Zygnematales, which have undergone a loss of 1,442 gene families since their diver- gence, the largest loss observed on the tree (Fig. 3a). Functional charac- terization of the genes lost in the Zygnematales using the KEGG database Nature Ecology & Evolution | Volume 6 | November 2022 | 1634-1643 identified gene families involved in the production of cytoskeletons, exosomes and phenylpropanoid synthesis (Supplementary Table 6). Exosomes and complex cytoskeletons are essential for multicellular organisms to function 43,44, and the inferred loss of these gene families is consistent with the hypothesis that the body plan of the algal ancestor of embryophytes was multicellular³, rather than possessing the single-cell or filamentous architecture observed in extant Zygnematales. The more complex cytoskeleton could be associated with increased rigidity, helping overcome the gravitational and evaporative pressures associ- ated with the transition to land. Interestingly, phenylpropanoids are associated with protection against UV irradiance 45 and homiohydry³, suggesting that the common ancestor may have been better adapted to a terrestrial environment than extant Zygnematales. We also observed greater gene loss along the bryophyte stem line- age (Fig. 3a and Supplementary Tables 3, 7 and 8), with the rate of gene loss (in terms of gene families per year) substantially greater than in all other major clades (Fig. 3b). It is important to note that inferences of gene loss from large-scale analyses are sensitive to the approach used to cluster sequences and define gene families; current approaches are not consummate. We therefore sought to evaluate the robustness of our conclusions using a range of sensitivity analyses (Supplemen- tary Figs. 5-8). These suggested that, while the number of inferred gene losses on the bryophyte stem varies, it remains an event of major gene loss under all conditions tested. We also observed considerable losses along the tracheophyte stem, countered by a greater number of duplications (Supplementary Table 9). This suggests a period of genomic upheaval on both sides of the embryophyte phylogeny. Gene Ontology (GO) term functional annotation of the gene families lost in bryophytes reveals reductions in shoot and root development from the ancestral embryophyte (Supplementary Table 7 and Extended Data Fig. 5). To investigate the evolution of genes underlying morphological 1637 events Number of DTL b Article a 35,000 30,000 25,000 20,000 4,705 Event 6,561 Duplication 7,508 Transfer Loss 500 7,043 https://doi.org/10.1038/s41559-022-01885-x Chlorokybus Klebsormidium Chara. braunii Spirogloea Mesotaenium Anthoceros punctatus Anthoceros agrestis 12,981 M. polymorpha rud. 10,646 M. polymorpha 9,597 8,485 9,712 Physcomitrium Fontinalis 14,601 14,857 Syntrichia 14,980 Ceratodon 10,473 Selaginella moellendorfii 11,287 11,927 14,376 11,755 13,244 Selaginella lepidophylla Salvinia Azolla Ginkgo Gnetum 15,547 15,273 Pseudotsuga 19,282 Pinus 15,000 с 15,109 Embryophyte Liverwort 15,521 10,000 507 6,462 1,031 5,553 5,000 6,642 5,618 2,616 0 Zygnemataceae Embryophyta Bryophyta Tracheophyta Angiosperm 16,752 Amborella Nymphaea Ceratophyllum Magnolia 15,764 15,719 Cinnamomum Kalanchoe 15,186 Arabidopsis 15,881 Dioscorea 14,435 Brachypodium Fig. 3 | Gene content reconstruction of the ancestral embryophyte. a, Ancestral gene content was inferred for the internal branches of the embryophyte tree. A maximum likelihood tree was inferred from an alignment of 30 species of plants and algae, comprising 185 single-copy orthologues and 71,855 sites, under the LG + C60+ G4 + F model in IQ-TREE69, and rooted in accordance with our previous phylogenetic analysis. A timescale for the tree was then calculated using a subset of 18 applicable fossil calibrations in MCMCtree. We reconciled 20,822 gene family clusters, inferred using Markov clustering 87, against the rooted dated species tree using the ALEml algorithm 88. The summed copy number of each gene family (under each branch) was determined using custom Python code (branchwise_number_of_events.py). Branches with reduced copies from the ancestral node are coloured in red. The numbers of DTL events are represented by purple, blue and red circles, respectively. The sizes of the circles are proportional to the summed number of events (the scale is indicated by the grey circle). b, The number of DTL events scaled by time for four clade- defining branches in the embryophyte tree. c, The number of shared gene families between the ancestral embryophyte, liverwort and angiosperm. The ancestral embryophyte shares more gene families with the ancestral angiosperm than with the ancestral liverwort. differences between tracheophytes and bryophytes, we evaluated the evolutionary history of gene families containing key Arabidopsis genes for vasculature and stomata (Supplementary Table 10). Gene families associated with both vasculature and stomatal function exhibited lineage-specific loss in bryophytes (Supplementary Figs. 9 and 10). Specifically, four orthologous gene families that are involved in the determination of the Arabidopsis body plan, containing WOX4, SPCH/ MUTE/FAMA, AP2 and ARR, were inferred to be lost on the bryophyte stem (Supplementary Table 10). To investigate these inferred losses in more detail, we manually curated sequence sets and inferred phyloge- netic trees for these families (Supplementary Methods and Extended Data Fig. 6). These analyses of individual gene families corroborated the pattern of loss along the branch leading to bryophytes. The loss of these orthologous gene families strengthens the hypothesis that Nature Ecology & Evolution | Volume 6 | November 2022 | 1634-1643 ancestral embryophytes had a more complex vasculature system than that of extant bryophytes³. Overall, the loss of gene families (Fig. 3) and the change in GO term frequencies (Extended Data Fig. 5) suggest a widespread reduction in complexity in bryophytes, and the ancestral embryophyte being more complex than previously envisaged. Indeed, gene loss defines the bryophytes early in their evolutionary history, but large numbers of duplication and transfer events are observed follow- ing the divergence of the setaphytes and hornworts (Supplementary Table 3), with (for example) extant mosses boasting a similar gene copy number to tracheophytes (Fig. 3). Discussion We have presented a time-scaled phylogeny for embryophytes, which confirms the growing body of evidence that bryophytes form 1638