RESEARCH ARTICLE

INVITED SPECIAL ARTICLE For the Special Issue: The Tree of Death: The Role of Fossils in Resolving the Overall Pattern of Plant Phylogeny

Building the monocot tree of death: Progress and challenges emerging from the macrofossil-­rich Zingiberales

Selena Y. Smith1,2,4,6 , William J. D. Iles1,3 , John C. Benedict1,4, and Chelsea D. Specht5

Manuscript received 1 November 2017; revision accepted 2 May PREMISE OF THE STUDY: Inclusion of fossils in phylogenetic analyses is necessary in order 2018. to construct a comprehensive “tree of death” and elucidate evolutionary history of taxa; 1 Department of Earth & Environmental Sciences, University of however, such incorporation of fossils in phylogenetic reconstruction is dependent on the Michigan, Ann Arbor, MI 48109, USA availability and interpretation of extensive morphological data. Here, the Zingiberales, whose 2 Museum of Paleontology, University of Michigan, Ann Arbor, familial relationships have been difficult to resolve with high support, are used as a case study MI 48109, USA to illustrate the importance of including fossil taxa in systematic studies. 3 Department of Integrative Biology and the University and Jepson Herbaria, University of California, Berkeley, CA 94720, USA METHODS: Eight fossil taxa and 43 extant Zingiberales were coded for 39 morphological seed 4 Program in the Environment, University of Michigan, Ann characters, and these data were concatenated with previously published molecular sequence Arbor, MI 48109, USA data for analysis in the program MrBayes. 5 School of Integrative Plant Sciences, Section of Plant Biology and the Bailey Hortorium, Cornell University, Ithaca, NY 14853, USA KEY RESULTS: Ensete oregonense is confirmed to be part of Musaceae, and the other 6 Author for correspondence (e-mail: [email protected]) seven fossils group with Zingiberaceae. There is strong support for Spirematospermum Citation: Smith, S. Y., W. J. D. Iles, J. C. Benedict, and C. D. friedrichii, Spirematospermum sp. ‘Goth’, S. wetzleri, and Striatornata sanantoniensis in Specht. 2018. Building the monocot tree of death: Progress and crown Zingiberaceae while “Musa” cardiosperma, Spirematospermum chandlerae, and challenges emerging from the macrofossil-­rich Zingiberales. Tricostatocarpon silvapinedae are best considered stem Zingiberaceae. Inclusion of fossils American Journal of Botany 105(8): 1389–1400. explains how different topologies from morphological and molecular data sets is due to doi:10.1002/ajb2.1123 shared plesiomorphic characters shared by Musaceae, Zingiberaceae, and Costaceae, and most of the fossils.

CONCLUSIONS: Inclusion of eight fossil taxa expands the Zingiberales tree and helps explain the difficulty in resolving relationships. Inclusion of fossils was possible in part due to a large morphological data set built using nondestructive microcomputed tomography data. Collaboration between paleo-­ and neobotanists and technology such as microcomputed tomography will help to build the tree of death and ultimately improve our understanding of the evolutionary history of monocots.

KEY WORDS anatomy; digital morphology; Ensete oregonense; Spirematospermum; Striatornata; Tricostatocarpon.

Monocot flowering plants represent ca. 22% of flowering plant from species such as bamboo or abaca; spices such as ginger, tur- species, encompassing a large diversity of morphology, habit, and meric, and saffron; and many ornamentals such as spring bulbs, ecologies. This group is economically important, including many of irises, and orchids. Monocots are also ecologically important, form- our staple food crops such as grains (maize, wheat, rice, sorghum), ing dominant components of grassland, savanna, fynbos, wetland, coconuts, plantains; pasture feed for animals; materials produced and seagrass ecosystems, as well as important parts of tropical forest

American Journal of Botany 105(8): 1389–1400, 2018; http://www.wileyonlinelibrary.com/journal/AJB © 2018 Botanical Society of America • 1389 1390 • American Journal of Botany

understories. At an ordinal level, the monocot phylogeny has been Crane, 1995; Smith, 2013). The lineages of monocots that are better relatively stable compared to other groups (APG, 1998, 2016), mak- represented tend to be those that are more lignified (e.g., palms) or ing them useful for broader studies. Most studies find monocots to grow in habitats that are near good depositional environments, such be ca. 135 Ma (e.g., Janssen and Bremer, 2004: 134 Ma; Magallón as quiet bodies of fresh water. et al., 2015: 135.7 Ma). Monocots represent a good model group First and foremost, the study and accurate naming of fossil taxa for elucidating the patterns and processes of evolution, and under- is a key component of building a reliable and accurate tree of death. standing their evolutionary history is fundamentally important to Taxonomy is a vital and dynamic process, and assigning a name and human nutrition and well-­being. rank to a fossil provides a taxonomic and phylogenetic framework Data from fossil taxa need to be included to obtain the most for the taxon in question; we recommend not using unnamed fos- comprehensive results when inferring phylogenetic relationships sils for dating phylogenies (e.g., Bell et al., 2010; Smith et al., 2010a; and investigating trait evolution, geographic histories, and other as- Zanne et al., 2014; Tank et al., 2015), as the lack of a name suggests pects of evolution for a lineage. In most cases, fossils are simply con- the need for careful evaluation of described morphology and/or sidered as constraints on the ages of nodes (e.g., Ho and Duchêne, ambiguity in phylogenetic placement based on characters analyzed. 2014; but see Ronquist et al., 2012a; Heath et al., 2014; Zhang et al., In addition, one must be cognizant of the framework within which 2016) within a molecular phylogeny. Either the clade(s) including taxa were named as this can influence where they are assumed to be the fossil(s) or even the entire tree is fixed or constrained: in topol- placed within a phylogenetic context. There has been a paradigm ogy, and inferred ages are dependent on the sequence data, calibra- shift in paleobotany regarding taxonomy. In the early 20th century, tion priors, and the model of rate variation used but not uncertainty attempts were made to place as many fossils as possible into extant in the tree or the fossil placement per se. In these cases, morpho- genera. Subsequently, thinking of fossils as extinct taxa that may logical data and fossils do not inform the topology of the inferred or may not be directly related to (or nested within) modern taxa phylogeny, but rather conform to placements dictated by the re- became more acceptable and encouraged. While some previously searcher. Fossil placements among and within lineages are therefore described fossils have been transferred to new extinct genera and not tested as part of the tree-building­ process. However, we know relationships to other fossils and to extant taxa were subsequently that present-­day diversity in all lineages is a result of complex in- re-­evaluated (e.g., “Viburnum” leaves from the Paleogene of North teractions on geological time scales of extinction, speciation, ecol- America and Asia that are now classified in the extinct genera ogy, morphology, and genetics (e.g., Barnosky, 2001; McElwain and Amersinia and Beringiaphyllum; Manchester et al., 1999), many Punyasena, 2007; Escapa and Pol, 2011; Green et al., 2011; Swenson, more have not been reinvestigated. Fossils assigned to extant genera 2011; Wiens, 2017). Incorporating morphological data from the should therefore be approached with particular caution, as the evi- fossil record is the only objective way of characterizing extinct line- dence for placing them in an extant genus (or even species) needs to ages and determining where they may fit in the evolutionary history be tested; however, these are very good candidates for further study. of a lineage. With fossils included as terminal units in the phyloge- Additionally, researchers should keep in mind that while presumed netic analysis, biogeographic patterns, trait evolution, and impact affinities and phylogenetic placement may change once re-evaluated,­ of environmental and ecological changes across lineages can be ex- the name of the taxon is often retained and reflects an incorrect amined with greater generality compared to studies based only on relationship. An example of this is the pollen genus Pandaniidites extant species with molecular sequence data. There are many cases Elsik, which was described from isolated grains that bore a resem- where the fossil record preserves morphological, spatiotemporal blance to Pandanus Parkinson (Pandanaceae), and given a name (e.g., Prasad et al., 2005, 2011; Smith et al., 2008, 2009b; Wilf and that reflected those observations, was classified as Pandanaceae. Escapa, 2015) and even climatic/environmental data (e.g., Wing Subsequently, the grains were found in situ in Pandaniidites that and Greenwood, 1993; Greenwood and Wing, 1995) that could not had aroid affinities, demonstrating they belong to Araceae and not be predicted, or would not be considered when only extant lineages Pandanaceae (Stockey et al., 1997): the name, however, using the are evaluated. Rather than only relying on the information present International Code of Nomenclature for algae, fungi, and plants, did in extant species to understand patterns of evolution and diversi- not change. Thus, usingPandaniidites to calibrate a Pandanaceae fication across the entire history of a clade, including fossils will node would be erroneous (Iles et al., 2015). improve our inferences by incorporating a temporal component to In addition, many fossil localities remain understudied or even studies of trait evolution, biogeographic patterns, and phylogenies. undiscovered. These have the potential to contribute significant in- formation, especially those from regions outside of the mid-­ and Reading the fossil record of monocots high-­latitude northern hemisphere, which is overrepresented in paleobotanical collections. Recent work, for example, has dramati- Numerous reviews regarding the fossil record of monocots (Doyle, cally improved the fossil record of Australia and New Zealand and 1973; Daghlian, 1981; Collinson et al., 1993; Herendeen and Crane, of many monocot lineages that are otherwise poorly known in the 1995; Gandolfo et al., 2000; Greenwood and Conran, 2000; Stockey, fossil record (see Conran et al., 2015 for a recent summary). As more 2006; Smith et al., 2010b; Friis et al., 2011; Smith, 2013) have dis- fossils are found, the potential to find multiple organs of a species cussed the challenges of working with monocot fossils, adding in organic connection grows, allowing for more accurate develop- complications to building a monocot “tree of death”, i.e., a tree of ment of a whole-­plant concept, and subsequently, for building more life that includes both extant and extinct taxa to resolve overall pat- complete trait matrices and reducing ambiguity in nomenclature. terns of phylogenetic relationships. Monocots generally have a low preservation potential because they are often small and herbaceous, The undeniable value of morphology lacking “woody” and highly lignified tissues, and have persistent se- nescent organs, so they either rot in place or never contribute to Morphology is key to being able to accurately place fossils in a phy- sediments and thus, do not enter the fossil record (Herendeen and logenetic context and to more broadly understand the evolutionary August 2018, Volume 105 • Smith et al.—Building the monocot tree of death • 1391

histories of the lineages they represent. By using morphology and synchrotron-­based methods (collectively referred to as “μCT” here incorporating fossils directly into phylogenetic reconstruction, we for simplicity), has the potential to transform our approaches and can address key questions beyond the simple timing of divergences, help build the monocot tree of death much more efficiently. These such as: How have dispersal and vicariance influenced a lineage’s μCT techniques use X-­rays to image a sample based on differing distribution over time? What biotic and abiotic interactions have densities of materials, producing a series of 2D images (tomograms; driven trait evolution? Has morphological or ecological disparity roughly equivalent to digital greyscale versions of histological sec- increased, decreased, or remained constant in time? How does tions) that can be imported into various computer programs and tempo and mode of morphological and molecular trait evolution visualized in three dimensions (see e.g., Sutton, 2008; Smith et al., compare? How widespread are homologies within a group, and in 2009a; Davies et al., 2017 for more details). These 3D data sets can which cases are presumed synapomorphies representative of ple- be “sliced” in any orientation to produce new 2D sections and can be siomorphic characters? In some cases, the fossils will help resolve “segmented” or “digitally dissected” to highlight certain structures nodes; in others, they can at least provide insight as to why there is in 3D. The lack of morphoanatomical data is a significant barrier conflict among topologies. to building total evidence trees, but μCT opens up the possibility Monocot fossils are preserved in a variety of different ways, and of building large morphological data sets since it is so much more the amount of data we can obtain to help determine a fossil’s phy- rapid than traditional histology (minutes to hours vs. days). Further, logenetic placement relates to both the preservation type and the μCT offers several other advantages: it is useable on small and large organ(s) preserved. Some fossils will never be useful for calibrating objects; it enables visualization of structures with hard or mixed nodes on phylogenetic trees because they are not preserved well density tissues (especially common in many fruits and seeds) with- enough to display any distinctive characters for a particular lineage, out tearing or destruction of adjacent tissues; it is nondestructive, but others contain a wealth of information (see Iles et al., 2015). meaning museums and herbaria are more willing to have their spec- A major challenge (but also benefit) to including fossils in a phy- imens studied; and it produces digital data sets that can be shared logeny is having appropriate morphological data scored for both widely and publicly archived, enhancing researcher’s ability to work the fossil and any extant group with which it is being compared. collaboratively, to reinterpret data, and to find new features (e.g., Unfortunately, unlike other plant groups, the growth habits of many Smith et al., 2009a; Davies et al., 2017). For paleobotanists, μCT also monocots mean they do not often preserve as whole plants, but are enables us to understand what morphological interpretations might more likely to be represented in the fossil record by isolated organs. result merely from different planes of sections or variation in the The characters scored in the extant taxa will need to reflect charac- preservation of tissues (e.g., Smith et al., 2009a; Adams et al., 2016). ters represented in putative fossil relatives, and phylogenetic infor- For example, fossils are sometimes obliquely sectioned relative to mation will depend on which organ is represented. Reproductive the standard transverse and longitudinal sections in modern plants, structures (flowers, fruits, and seeds) tend to be more complex and which can affect a researcher’s ability to accurately and confidently have more phylogenetically informative characters compared to make comparisons with other taxa that have been preserved dif- vegetative features; good examples are the seeds of Zingiberales, ferently or may be three-dimensionally­ complex (e.g., Matsunaga which arguably have the most complex seeds among all angio- et al., 2018; morphology of Viracarpon fruits was disputed for more sperms (Benedict et al., 2015a,b, 2016). However, many monocot than 50 years, but resolved with the application of μCT). Because fossils represent vegetative stem or leaf organs. While these are accuracy of morphological interpretation can affect the scoring of often more cryptic in monocots, especially for gross morphology characters in a data matrix, using 3D data can play a positive role in (e.g., leaf shape, venation), there is still value in including as much assigning homologies and reconciling relationships. While mono- data on vegetative structures as possible, especially as these vege- cots have a reputation for having cryptic morphologies (e.g., Rudall, tative characters may be indicative of climatic or environmental 2000), more characterization of fossil and living monocot traits will conditions. undoubtedly lead to a better understanding of the subtleties of their One of the main challenges to building a monocot tree of death structure and enable recognition and identification of characters is the paucity of morphological data for both extant and extinct useful for phylogenetic analysis. monocots. Progress is being made, however, especially with the de- velopment of novel approaches. There is active research on phytolith Zingiberales: A case study in integrating fossils into the morphology, for example, as applied to the fossil record of grasses phylogeny and Cenozoic vegetation (e.g., Prasad et al., 2005, 2011; Strömberg, 2005, 2011; Miller et al., 2012; Strömberg et al., 2013), enhancing Evaluating the influences of different analytical choices in prepar- our understanding of variation in phytolith morphotypes across ex- ing a tree of death is necessary so that more robust hypotheses can tant monocots and determining the characters that can be used to be produced from the outset. Here, we use an extensive morpho- interpret affinities within the fossil record. For groups like Poaceae, logical data set of seed anatomy of Zingiberales to test the influ- the enhanced understanding of phytolith morphologies and their ence of different data partitions and taxon inclusion/exclusion on phylogenetic (or functional group) significance has greatly im- the phylogenetic placement of fossil taxa. Eight fossil taxa that are proved our ability to assign taxonomic affinities to fossils. Within Cretaceous to Pliocene in age are represented by permineralized some clades, unique phytolith morphotypes have been found (e.g., fruits and seeds with a wealth of morphological data (Chandler, Piperno, 2006; Prychid et al., 2004; Chen and Smith, 2013), which 1925; Jain, 1963; Friedrich and Koch, 1970; Koch and Friedrich, has led to the identification of the earliest unequivocal grass fossil, 1970; Goth, 1986; Knobloch and Mai, 1986; Friis, 1988; Manchester in the rice tribe Oryzeae from the latest Cretaceous (Maastrichtian) and Kress, 1993; Rodriguez-de­ la Rosa and Cevallos-Ferriz,­ 1994; of India (Prasad et al., 2005, 2011). Fischer et al., 2009). These include the taxon Spirematospermum X-­ray microcomputed tomography using either micro-­ Chandler, which has been the subject of much controversy over or nanoscale industrial computed tomography machines or its familial identity. Spirematospermum was originally considered 1392 • American Journal of Botany

to be a member of the family Zingiberaceae (Chandler, 1925; Rosa & Cevallos-Ferriz­ (material examined: ASU-­5002a) were Koch and Friedrich, 1970) but subsequently was thought to be both described by Rodriguez-­de la Rosa and Cevallos-­Ferriz (1994) more closely related to Musaceae (Manchester and Kress, 1993; from the Campanian of Mexico, while Musa cardiosperma Jain is Rodriguez-­de la Rosa and Cevallos-Ferriz,­ 1994; Fischer et al., from the Maastrichtian-­Danian Deccan Intertrappean Beds of 2009) with Fischer et al. (2009) including the taxon in Musaceae in India (Jain, 1963); despite the name, affinities to Musa are doubt- its own subfamily Pariemusinae, based on their study of S. wetzleri ful (material examined: BSIP 31654-­1 [holotype], 31654-­2, 31654-­ (Heer) Chandler. The availability now of extensively sampled seed 3; FLMNH UF18311-­62188, UF18311-­70295, UF18311-­70296, morphological data from across all eight families of the order and UF18311-­70298, UF18311-­9963, UF19279-­62144, UF19506-­70302; a large molecular data set with broad taxon sampling (Sass et al., CMNH PM1711, PM1765). Striatornata, Tricostatocarpon, and 2016) make it ideal for exploring the impact of including fossils Musa cardiosperma, plus Spire. wetzleri, are all preserved as both in resolving deeper nodes of the phylogeny and for testing the ef- fruits and seeds. Ensete oregonense (Musaceae) is represented by fect of different parameters and types and quality of data on the seeds in the Eocene Clarno Formation, Oregon (Manchester and placement of fossil lineages during phylogenetic reconstruction. Kress, 1993); we examined the holotype, FLMNH UF 6621. Fossils Understanding the influence of data and parameter choices can ascribed to Alpinia Roxb. (Mai and Walther, 1985, 1991; Mai, 1999) lead to developing best practices for building any lineage-­specific were shown to belong to different, non-Zingiberales­ taxa (Smith tree of death. et al., 2015) and are not considered further here.

Phylogenetic data matrix MATERIALS AND METHODS Morphological data for extant taxa were coded using the characters Fossils from Benedict et al. (2015a,b, 2016) with modifications to remove uninformative characters and minimize character overlap, resulting Eight fossils were included in the analyses: four species in in 39 morphological characters (Appendix S1, see Supplemental Spirematospermum, and four other taxa placed in separate gen- Data with this article). Halopegia azurae (K.Schum.) K.Schum. era; only one of these is confidently placed to family (Ensete ore- (MO Boussengui-­Nongo et al. 18), Heliconia acuminata A.Rich. gonense Manchester & Kress). Specimens were examined directly (US Kress 91-­3242), Heliconia nutans Woodson (US n.c.), and in all cases, with data used in scoring coming from the literature Canna jaegeriana Urb. (US Kress 88-2245)­ were newly scored and and our own observations. Material examined is listed for each added to this study. Heliconia L. species were previously scored as taxon. Samples were observed from the following repositories (in- having opercula, but here are recoded as having no opercula, re- stitutional codes for specimen numbers indicated in parentheses): flecting developmental homologies (their operculum is derived Arizona State University (ASU); Bayerische Staatssammlung für from endocarp tissue, not the seed coat). Fossil taxon traits were Paläontologie und Geologie in Munich, Germany (SNSB-BSPG);­ coded using both published literature and our (S. Y. Smith and J. Birbal Sahni Institute of Palaeosciences, Lucknow, India (BSIP); C. Benedict) observations of specimens. Molecular data were se- Cleveland Museum of Natural History, Cleveland, Ohio, USA lected from Sass et al. (2016), with a subset of taxa chosen to match (CMNH); Florida Museum of Natural History, Gainesville, Florida, the species with available morphological data. Where both types USA (FLMNH UF); Museum für Naturkunde, Berlin, Germany of data were not available for a single species, molecular and mor- (NKM); Smithsonian National Museum of Natural History, phological characters were concatenated from congeneric species, Washington, D.C., USA (USNM). Spirematospermum chandlerae if only one representative of that genus was included (Calathea Friis (1998) is known from the late Campanian of North Carolina, G.Mey., Ischnosiphon Körn, Chamaecostus C.D.Specht & D.W.Stev., United States (Friis et al., 2011) (material examined: USNM 712537, Curcuma L., Elettariopsis Baker; and Distichochlamys M.F.Newman 712538). Additional Late Cretaceous taxa are Spirematospermum instead of Scaphochlamys Baker, which are sister genera). Our data friedrichii Knobloch & Mai (Knobloch and Mai, 1986) from the matrix is available in Appendix S2. Maastrichtian of Eisleben, Germany (material examined: NKM 8516A, 8516B, 8516C, 8516D, 8516E, 8516H [holotype]) and an Phylogenetic analyses undescribed species of Spirematospermum (Goth, 1986) from the late Campanian/early Maastrichtian of Kössen, Germany (hereaf- Several analyses were run to evaluate phylogenetic relationships, ter referred to as Spirematospermum sp. ‘Goth’; this taxon will be with results from the following presented here: (1) total evidence described in a subsequent paper; material examined: SNSB-BSPG­ (= molecular plus morphological data sets), all fossils; (2) total 1984 VII 15, 1984 VII 90). The species that has received the most evidence, single fossil (resulting in eight analyses); (3) morphol- attention is Spirematospermum wetzleri, as it is the most spatially ogy only matrix, all fossils. We ran total-­evidence analyses with and temporally widespread and well preserved (even with fruits) all the fossils and each fossil individually to evaluate differences at many localities; it forms up to 16% of Neogene European fos- in placement and support values. We ran the morphology-­only sil assemblages (Fischer et al., 2009). This species was described analysis to evaluate whether the morphology data set carried a from deposits across Eurasia and ranges in age from Eocene to different phylogenetic signal. Bayesian analyses of molecular and Pliocene, and we examined representatives from several Oligocene, morphological data were run in MrBayes v3.2.6 (Ronquist et al., Miocene, and Pliocene localities in Germany (material examined: 2012b). The molecular data are from Sass et al. (2016) and consist ASU-­5000a, b, c; SNSB-­BSPG 1970 I, 1977 I, 1972 XXI 79, 1972 of an alignment of 137,748 bp of coding regions from the plas- XXI 19; FLMNH UF19083-­049162). Striatornata sanantonien- tid and nuclear compartments. The nuclear data consist of 308 sis Rodriguez-­de la Rosa & Cevallos-­Ferriz (material examined: markers from across the genome. We partitioned the nuclear and ASU-­5001a) and Tricostatocarpon silvapinedae Rodriguez-­de la plastid components by codon position resulting in six molecular August 2018, Volume 105 • Smith et al.—Building the monocot tree of death • 1393

Phylogenetic analyses: morphology only partitions. Each partition was evaluated under the GTR+Γ4 model of sequence evolution. For the morphological data, we used a Bayesian morphological analysis resolved five families as mono- symmetric model of character state change and a Γ parameter 4 phyletic although support varied (82–100% PP), but neither to allow site-­to-site­ heterogeneity, and ascertainment bias was Zingiberaceae nor Strelitziaceae were resolved (although the lat- set to parsimony informative. Topology and branch lengths were ter formed a clade with its sister group, Lowiaceae, with 84% PP; linked across all six molecular partitions and the morphological Fig. 3). Relationships within extant Zingiberaceae were completely partition. Two runs consisting of four chains each were run in unresolved, a single species, Etlingera elatior (Jack) R.M.Sm., was parallel, the “temperature” was set between 0.01–0.03 because resolved as the sister group to Costaceae, with weak support. The this range resulted in optimal acceptance ratios. Analyses were fossil Ensete oregonense was well supported in Musaceae (100% run for at least 2 million generations (sampling every 500) and PP). The other fossil taxa formed a polytomy with members of until the standard deviation of split frequencies fell below 0.01. Zingiberaceae, of which some formed species pairs with varying The morphology-only­ analysis was run under the same setting as support (52–89% PP). A moderately supported bipartition (95% the total-evidence­ analysis above except that only the morpho- PP) separates the families Cannaceae, Heliconiaceae, Lowiaceae, logical data were considered, and it was run for 5 million genera- Marantaceae, and Strelitziaceae from Costaceae, Musaceae, tions. Our data matrix is available in the online supplemental data Zingiberaceae, and the fossils (Fig. 3). All phylogenetic trees are (Appendix S2). We consider posterior probabilities (PP) over 98% available in Appendix S4. to be strong support, 95–98% to be moderate support, and below 95% to be weak support. Character evolution Character evolution Several characters were found that supported the placement of fos- We used maximum parsimony (MP) to map traits and infer sils (excluding Ensete oregonense) within Zingiberaceae and as a character evolution under the total-evidence­ all-­fossils and the clade including Musa cardiosperma, Spire. chandlerae, Spire. wet- morphology-­only all-fossils­ topologies. Maximum parsimony re- zleri, Striatornata sanantoniensis, and Tricostatocarpon silvapine- constructions (with opt = deltran) and most parsimonious recon- dae, which we will refer to as the ‘Spirematospermum plus’ clade. struction sets were inferred with PAUP* version 4.0a159 (Swofford, Contorted seeds from tight packing in fruit (character 5) was found 2002). The NEXUS file to carry out these analyses is available in the to be shared by all members of the ‘Spirematospermum plus’ clade Appendix S3. and, was lacking in the two fossil members excluded from this clade, Spire. sp. ‘Goth’ and Spire. friedrichii. A similar case was seen in seed coat tapering at the micropyle (character 7) and chalazal mesotestal RESULTS proliferation of cells (character 22): all members except Musa cardio- sperma in the ‘Spirematospermum plus’ clade had these features, but Phylogenetic analyses: total evidence they were lacking in Spire. sp. ‘Goth’ and Spire. friedrichii. The chala- zal chamber (character 23) was found to be a mixture of Amomum-­ Bayesian total-­evidence analyses completed between 2 and 34 mil- type and Alpinia-­type chambers in the ‘Spirematospermum plus’ lion generations. Non-­topology parameter estimated sample sizes clade but is lacking in Spire. sp. ‘Goth’ and Spire. friedrichii. Another were nearly always over 200 indicating stable parameter estimates. character that separated the ‘Spirematospermum plus’ clade from The total-­evidence all-­fossils analysis generally showed most fami- Spire. sp. ‘Goth’ and Spire. friedrichii was the relative thickness of lies receiving moderate (one family) to strong (five families) support the mechanical layer (character 25). Spirematospermum sp. ‘Goth’ with the exception of Zingiberaceae with 93% PP (Fig. 1) (Lowiaceae and Spire. friedrichii had seed coats less than 100 μm wide at the was represented by a single individual.). Within Zingiberaceae, the thinnest point of the seed, where the ‘Spirematospermum plus’ clade subfamily Zingiberoideae was resolved with weak support (84% had seeds greater than 100 μm wide to more than 200 μm wide at PP), and support for relationships within the clade varied from the thinnest point of the seed. Endotestal thickness (character 33) 84–97% PP. The subfamily Alpinioideae was not resolved, forming was also found to be mostly uniform in the ‘Spirematospermum plus’ a polytomy at the base of the family with Zingiberoideae and all the clade with the exception of Tricostatocarpon silvapinedae where fossils except Ensete oregonense. All species of Spirematospermum it was difficult to discern this character from the fossil material. were placed within Zingiberaceae; Spirematospermum sp. ‘Goth’ Endotestal thickness was also scored as missing for Spire. sp. ‘Goth’ and Spire. friedrichii as isolated species arising from the fam- and Spire. friedrichii because it was too difficult to discern from in ily polytomy, and Spire. chandlerae and Spire. wetzleri forming a the fossil material. A uniform exotesta (character 29) was found in clade with Musa cardiosperma, Striatornata sanantoniensis, and all members of the ‘Spirematospermum plus’ clade, Spire. s p . ‘G o t h’, Tricostatocarpon silvapinedae with poor support (63% PP) (also and Spire. friedrichii, and was one of the only characters that united arising from the family polytomy). Ensete oregonense was mod- all taxa considered to be Spirematospermum. One exception is Musa erately supported in Musaceae (97% PP) but its placement within cardiosperma, where the exotesta was highly fragmented and dif- the family is unresolved. The total-­evidence single-fossil­ analyses ficult to discern whether the exotesta was uniform or not. Finally, resemble the all-­fossils analysis (Fig. 2), except that more branches a micropylar collar (character 16) was lacking in all fossils except are resolved (e.g., Zingiberaceae subfamily Alpinioideae) and that for Ensete oregonense. Several of these character reconstructions are Zingiberaceae typically also had higher support (96–100% PP com- shown on the total evidence all fossils tree (Fig. 1). pared to 93% PP), with the exception of the Musa cardiosperma Several other morphological characters were found that sup- analysis where Zingiberaceae had 74% PP and the Spire. chandlerae port a close relationship between Costaceae, Musaceae, and analysis with 91% PP. Zingiberaceae. A chalazal indentation of the seed coat (character 10) 1394 • American Journal of Botany

†Ensete oregonense 22 Musa coccinea 97 61 Musa balbisiana 7 85 Musa acuminata Musaceae Musella lasiocarpa 88 5 Ensete ventricosum 93 Ensete superbum 7 Heliconia nutans Heliconia acuminata Heliconiaceae Orchidantha maxillarioides Lowiaceae 7 Ravenala madagascariensis Strelitzia nicolai Strelitziaceae Strelitzia reginae Canna iridiflora Canna indica Cannaceae Thaumatococcus daniellii Donax canniformis Goeppertia roseopicta 25 Ischnosiphon helenae Marantaceae Marantachloa leucantha Stromanthe stromanthoides 25 Halopegia azurea 96 Costus pulveralentus 7 Costus dubius Cheilocostus speciosus 5 Tapeinochilos ananassae Costaceae Chamaecostus subsessilis 7 Dimerocostus strobilaceus Monocostus uniflorus Siphonochilus kirkii Siphonochiloideae 95 Siamanthus siliquosus 93 7 25 25 Riedelia sp. Alpinia zerumbet 7 97 5 Etlingera elatior 99 22 Alpinioideae Alpinia purpurata 7 93 25 Elettariopsis stenosiphon 51 Aframomum angustifolium 7 Zingiberaceae 54 22 Renealmia alpinia 5 7 25 †Spirematospermum sp. ‘Goth’ 7 †Spirematospermum friedrichii 7 †Spirematospermum chandlerae 66 †Striatornata sanantoniensis †Tricostatocarpon silvapinedae 5 22 92 †Spirematospermum wetzleri †Musa cardiosperma 7 22 Globba winitti 84 22 5 Curcuma longa Hedychium coronarium 87 Zingiberoideae Scaphochlamys sp. 91 92 97 Zingiber officinale 0.02 Zingiber spectabile FIGURE 1. Bayesian total-­evidence all-fossil­ phylogenetic analysis of molecular and morphological data. Tree is rooted with Musaceae as indicated by Sass et al. (2016). Fossils are indicated by a dagger symbol and bold type. Numbers adjacent to nodes represent the Bayesian posterior probability expressed in percentages; nodes without numbers received full support, 100% posterior probability. The scale is in expected number of state changes (molecular or morphological) per site. Thick ticks indicate maximum parsimony character change reconstruction Using deltran optimization for se- lected characters; thin ticks indicate reversals. See text for further details. August 2018, Volume 105 • Smith et al.—Building the monocot tree of death • 1395

†Ensete oregonense †Musa cardiosperma †Striatornata sanantoniensis †Tricostatocarpon silvapinedae †Ensete oregonense Musaceae Musaceae Musa Musaceae coccinea Heliconiaceae 58 Musa balbisiana Musacea Heliconiaceae Heliconiaceae Lowiaceae 80 Musa acuminata 92 Strelitziaceae Lowiaceae Lowiaceae

Musella e lasiocarpa 97 Cannaceae 79 Ensete Strelitziaceae Strelitziaceae Marantaceae ventricosum 86 Ensete 98 Costaceae superbum Cannaceae Cannaceae 98 Siphonochiloideae Heliconiaceae 97 82 Marantaceae Marantaceae 96 Zingiberoideae Lowiaceae Zingiberaceae Siamanthus siliquosus Strelitziaceae 96 75 90 Costaceae Costaceae Riedelia sp. Cannaceae Alpinia Siphonochiloideae Siphonochiloideae zerumbet 54 98 Marantaceae 86 Etlingera Alpinioideae elatior Costaceae 74 †Musa †Striatornata Alpinia cardiosperma Zingiberaceae sanantoniensis purpurata Zingiberaceae 53 Elettariopsis Siphonochiloideae 76 68 stenosiphon 96 Zingiberoideae Zingiberoideae †Tricostatocarpon 57 ZingiberaceaeZingiberoideae 51 silvapinedae 61 68 Aframomum Alpinioideae angustifolium Alpinioideae Alpinioideae 54 Renealmia alpinia

†Spirematospermum friedrichii †Spirematospermum sp. ‘Goth’ †Spirematospermum chandlerae †Spirematospermum wetzleri

Musaceae Musaceae Musaceae Musaceae Heliconiaceae Heliconiaceae 99 Heliconiaceae Heliconiaceae Lowiaceae Lowiaceae 99 Strelitziaceae Strelitziaceae Lowiaceae Lowiaceae Cannaceae Cannaceae 98 Strelitziaceae Strelitziaceae Marantaceae Marantaceae

95 Costaceae 99 Costaceae Cannaceae Cannaceae Siphonochiloideae Siphonochiloideae 93 99 96 97 Zingiberoideae Zingiberoideae Marantaceae Marantaceae 91 98 Zingiberaceae Siamanthus Zingiberaceae Siamanthus siliquosus siliquosus 88 95 Costaceae Costaceae Riedelia Riedelia sp. sp. Alpinia Alpinia zerumbet zerumbet Siphonochiloideae Siphonochiloideae 99 83 Etlingera 92 Etlingera Alpinioideae elatior Alpinioideae elatior 99 †Spirematospermum †Spirematospermum Alpinia Alpinia Zingiberaceae friedrichii Zingiberaceae sp. ‘Goth’ purpurata purpurata 83 92 52 53 Elettariopsis Elettariopsis 69 82 stenosiphon stenosiphon Zingiberoideae Zingiberoideae Aframomum Renealmia 58 67 82 angustifolium 92 alpinia Aframomum 93 89 Renealmia 91 Alpinioideae Alpinioideae 70 alpinia angustifolium †Spirematospermum 87 †Spirematospermum chandlerae wetzleri FIGURE 2. Bayesian total-­evidence single-­fossil phylogenetic analysis of molecular and morphological data showing individual fossil placement. Trees are rooted with Musaceae as indicated by Sass et al. (2016). Each tree represents the analysis of a single fossil as indicated by the dagger symbol and bold type. Numbers adjacent to nodes represent the Bayesian posterior probability; nodes without a number received full support, 100% poste- rior probability. Clades removed from the vicinity of the fossil placement were collapsed for clarity.

was found in Musaceae, Costaceae, and one Zingiberaceae DISCUSSION (Etlingera elatior) but was absent in the other five families. A coni- cal or cylindrical-shaped­ micropylar region (character 12), endotes- Total evidence and the resolution of fossil identity tal gap (character 34), ­chalazal pigment group (character 35), and raphe canal (character 38)­ were also found to be shared by mem- Approaches that integrate molecular sequence data from extant bers of Costaceae, Musaceae, and Zingiberaceae and was not found taxa with morphological data from extant and extinct taxa can help in any of the other families. The presence of a hilar rim (character our understanding of character evolution and resolve fossil identity 19) was found in Musaceae, some Zingiberaceae, and Lowiaceae, in the context of living diversity (e.g., Nylander et al., 2004). The but not in any of the other five families. Another feature only seen total-­evidence approach adopted here uses a large molecular align- among Musaceae and Zingiberaceae taxa was highly differentiated ment (+100k bp) of the Zingiberales (Sass et al., 2016) with recent seed coats, where the mesotesta is composed of two to three differ- work characterizing and describing Zingiberales seed morphology ent types of cells (character 32). To show the strong support for a (Benedict et al., 2015a,b, 2016) to place newly scored Zingiberales clade consisting of Costaceae, Musaceae, and Zingiberaceae several fossils. of these characters are mapped on to the morphology only all fos- Bayesian total-­evidence analyses considering fossils singly or sils tree (Fig. 3). The character reconstructions for all 39 characters as a whole were consistent in fossil placement to the family level using deltran optimization (shown in Figs. 1 and 3) and the most within Zingiberales. The Eocene fossil Ensete oregonense is strongly parsimonious reconstruction sets are available in Appendix S5. supported as belonging to Musaceae in the single-­fossil analysis 1396 • American Journal of Botany

Musa balbisiana Musa acuminata Musa coccinea †Ensete oregonense Musaceae Ensete superbum 58 Ensete ventricosum Musella lasiocarpa †Tricostatocarpon silvapinedae 89 †Spirematospermum wetzleri 34 †Musa cardiosperma 57 †Spirematospermum sp. ‘Goth’ †Spirematospermum friedrichii 52 †Spirematospermum chandlerae †Striatornata sanantoniensis 38 Siphonochilus kirkii Siphonochiloideae Curcuma longa 38 Hedychium coronarium 38 Zingiber officinale Zingiberoideae 38 Zingiber spectabile Globba winitti Scaphochlamys sp. 38 Aframomum angustifolium Zingiberaceae Siamanthus siliquosus 12 38 Riedelia sp. 38 Alpinia zerumbet 12 Alpinia purpurata Alpinioideae 12 38 Elettariopsis stenosiphon 12 Renealmia alpinia 12 Etlingera elatior 67 Cheilocostus speciosus 12 38 Costus dubius 12 96 Tapeinochilos ananassae 53 12 38 Monocostus uniflorus Costaceae Costus pulveralentus 56 Chamaecostus subsessilis 12 Dimerocostus strobilaceus Heliconia nutans Heliconia acuminata Heliconiaceae 51 Orchidantha maxillarioides Lowiaceae 84 Ravenala madagascariensis 71 Strelitzia nicolai Strelitziaceae 95 Strelitzia reginae 12 34 35 38 Canna iridiflora Canna indica Cannaceae Thaumatococcus daniellii Goeppertia roseopicta 82 Stromanthe stromanthoides 91 Donax canniformis Marantaceae Marantachloa leucantha 55 Ischnosiphon helenae 0.08 Halopegia azurea FIGURE 3. Bayesian morphology-­only all-­fossil phylogenetic analysis. Tree is rooted with Musaceae as indicated by Sass et al. (2016). Fossils are indi- cated by a dagger symbol and bold type. Numbers adjacent to nodes represent the Bayesian posterior probability; nodes without numbers received full support, 100% posterior probability. The scale is in expected number of morphological state changes per site. Thick ticks indicated maximum parsimony character change reconstruction using deltran optimization for selected characters, thin ticks indicate reversals. See text for further details. August 2018, Volume 105 • Smith et al.—Building the monocot tree of death • 1397

(100% PP) and moderately supported in the all-fossils­ analysis and “banana group” (Musaceae, Lowiaceae, Heliconiaceae, (97% PP), confirming initial fossil placement (Manchester and Strelitziaceae). Eliminating fossil taxa from consideration in both Kress, 1993), although no specific relationship within the fam- the total-­evidence and morphology-­only analysis (see Appendix ily is resolved. While Ensete oregonense is not found in a clade S4) strengthens this conflict because the total-­evidence-­supported with other Ensete, our sampling of Musaceae is not as broad as bipartition is now 100% PP (the tree showing identical family rela- it might be, and the fossil is found in a polytomy with Musa and tionships to Sass et al., 2016). It is currently unclear how to interpret Ensete. Thus, it is premature to think Ensete oregonense might not this conflict. One possibility is to believe that the morphological data be in Ensete. In contrast, M. cardiosperma from the Cretaceous– are saying something true about phylogenetic relationships among Paleogene boundary has no association with Musaceae and instead the families and fossils of Zingiberales that is being masked or mis- is placed in Zingiberaceae, although with poor support (76% PP). lead by the molecular data. However, the molecular data come from The remaining fossils (Spirematospermum spp., Striatornata san- two separate genomic compartments (plastid and nuclear), each antoniensis, and Tricostatocarpon silvapinedae) also group with consisting of tens of thousands of base pairs that are congruent as to Zingiberaceae, with support varying from poor to moderate in family relationships (see Sass et al., 2016). Previous molecular anal- Spire. chandlerae and T. silvapinedae (91% and 96% PP, respec- yses also tend to recognize that Costaceae and Zingiberaceae form a tively) to strong in Spire. friedrichii, Spire. sp. ‘Goth’, Spire. wetzleri, strongly supported clade with Cannaceae and Marantaceae (ginger and Striatornata sanantoniensis (98–100% PP). Three fossils,Spire. group). In addition, morphological analysis based on different sets chandlerae, Spire. wetzleri, and T. silvapinedae are found deeply of vegetative and reproductive characters never group Musaceae nested within subfamily Alpinioideae and closely associated with with Costaceae and Zingiberaceae to the exclusion of Cannaceae Aframomum and Renealmia (Fig. 2), although strong support is and Marantaceae (Dahlgren and Rasmussen, 1983; Kress, 1990; lacking for these placements. The other fossils are in a clade with Kress et al., 2001). subfamilies Alpinioideae and Zingiberoideae, again with low sup- The remaining option is that family relationships as recon- port. Furthermore, when fossils are analyzed together, several of structed by the morphological data reflect instead something about these fossils (Spire. chandlerae, Spire. wetzleri, Striatornata sanan- the evolution of seed morphology (the basis for the morphological toniensis, T. silvapinedae, and M. cardiosperma) form a clade, albeit matrix considered here). This interpretation is supported by the with poor support (63% PP) (Fig. 1). The dissolution of subfamily fact that at least eight characters unite Musaceae, Costaceae, and Alpinioideae in this latter analysis is probably caused by the com- Zingiberaceae (or Musaceae+Zingiberaceae) vs. the other fam- bined effects of low phylogenetic signal in the morphology data in ilies and include features like the presence of a chalazal pigment that part of the tree (see Fig. 3) and multiple (fossil) samples with group, raphe canal, endotestal gap, and chalazal chamber. These no molecular data (i.e., clustered taxa with missing data) to reliably features are commonly seen in the fossil taxa as well, suggesting inform their placement. that these features should be interpreted as plesiomorphic (or pos- Individual and combined analysis of fossil taxa support place- sibly convergent), present in Musaceae (found as sister to all other ment of Ensete oregonense in Musaceae, and Spire. friedrichii, Zingiberales), Zingiberaceae, and sometimes Costaceae (which is Spire. sp. ‘Goth’, Spire. wetzleri, and Striatornata sanantonionensis sister to Zingiberaceae; Kress et al., 2001; Kress and Specht, 2006; in Zingiberaceae with strong support (98–100% PP) (Figs. 1–3). Sass et al., 2016), and lost/simplified through time in many of the The remaining fossils, Musa cardiosperma, Spire. chandlerae, and other lineages. Tricostatocarpon silvapinedae, are also best placed in Zingiberaceae; however, support is considerably weaker (74–96% PP) and, there- Core challenges to building the monocot tree of death fore, ultimately these may represent stem lineage Zingiberaceae. Several of the taxa that can be considered crown Zingiberaceae As with any large group, the primary hindrance to building a due to their strongly supported placement within the family are monocot tree of death—from which broader evolutionary, ecolog- Campanian (or early Maastrichtian) in age; thus, Zingiberales were ical, and morphological patterns can be drawn—is the lack of data. already diverse (both morphologically and taxonomically) by this Monocots are considered to have essentially an invisible radiation time. during much of the Cretaceous (Crepet, 2008), as evidenced by the lack of fossil record for many monocot lineages. Some groups of Morphological and molecular incongruence monocots have a good fossil record in part because they inhabit environments with a high preservation potential, such as aquatic The Bayesian analysis of morphology and fossils resolves sev- and semi-­aquatic alismatids (Stockey, 2006; Smith, 2013), or grow eral families with high support (Cannaceae, Heliconiaceae, and abundantly and have organs resistant to decay (e.g., palms, gingers). Musaceae with 100% PP), and several other families with low to Other monocots, though, will never have a great fossil record; for moderate support (Costaceae and Marantaceae, with 96% PP and example, mycoheterotrophs or achlorophyllous taxa that are gen- 82% PP, respectively) and also identifies several family groups erally small forest-dwellers­ (although there are charcoalified flow- (Heliconiaceae, Lowiaceae, and Strelitziaceae, with 51% PP). An un- ers from the Turonian attributed to Triuridaceae; Gandolfo et al., usual feature of the morphology tree is the separation of Costaceae, 2002). However, as more work is done to study both old and new Musaceae, Zingiberaceae, and fossils from the remaining families localities, it is likely that at least some of these taxonomic biases can with moderate support (95% PP) (Fig. 3). This bipartition is not seen be addressed. For example, orchids (family Orchidaceae) would fall in the total-­evidence tree that shows Costaceae and Zingiberaceae into this category of plants with poor preservation potential, but (and all fossils except Ensete oregonense) together in a clade with multiple fossils have now been placed with reasonable confidence Cannaceae and Marantaceae (96% PP) (Fig. 2), as is consistent in this family (Ramírez et al., 2007; Conran et al., 2009; Poinar, with previous analyses of Zingiberales that recognized a “gin- 2016a,b; Poinar and Rasmussen, 2017), demonstrating that with ger group” (Zingiberaceae, Costaceae, Cannaceae, Marantaceae) perserverance (and the right search image) more fossils for even 1398 • American Journal of Botany

poorly represented clades can be found. In addition, the early-­ likely stem Zingiberaceae. Reclassification of “Musa” cardiosperma, divergent lineages cannot be ignored. Although not addressed here, which is here placed within Zingiberaceae, seems justified based on there are putative monocot fossils that may represent early lineages these results, and this taxon is currently being revised by coauthors and need further study to confirm (or deny) their affinities (see e.g., S. Y. Smith and J. C. Benedict. The high level of extinct diversity, Smith, 2013 for discussion), and applying newer methodologies including Cretaceous fossils, in a more-­derived family suggests that such as μCT can help provide evidence for both extinct crown taxa family-­level diversification occurred rapidly within Zingiberales. To and more enigmatic stem taxa (e.g., Viracarpon fossils; Matsunaga undertake this study and include fossil taxa in the Zingiberales phy- et al., 2018). These kinds of enigmatic fossils are important to place logeny, morphological data were needed. Microcomputed tomogra- evolutionarily and provide critical insights into disparity and diver- phy proved to be effective for collecting this data and is promising sity within monocots. for facilitating similar studies in the future. Incorporation of fossil Incorporation of morphological data into phylogenetic analy- taxa in phylogenetic analyses provides context to interpretation of ses—permitting the inclusion of fossils—is a vital step for recon- the results and emphasizes the utility and importance of including structing phylogeny. Here we used morphological data from seeds, morphological and fossil data in building not only a tree of life, but both extant and fossil, to expand the Zingiberales phylogeny. Our a tree of death as well, which ultimately is necessary to better inform analysis illustrates that with further such studies that develop novel the evolutionary history of a group. character matrices including phylogenetically informative traits coded for extant and fossil taxa, building a monocot tree of death is possible. While morphology and careful homology assessment are ACKNOWLEDGEMENTS critical for incorporating fossils into the tree of death, challenges re- main in the lack of available morphological data of sufficient quality, The authors thank Steven Manchester (University of Florida) for depth, and breadth to be able to place fossil taxa; but this is a tracta- access to fossils of Ensete oregonense and “Musa” cardiosperma; ble problem, if people and resources are available. Increasing use Sergio Cevallos-­Ferriz for access to material of Striatornata of μCT offers a tool that is complementary to traditional histology and Tricostatocarpon; and Margaret Collinson (Royal Holloway and morphological characterization to expand our sampling and University of London), W. Friedrich (Aarhus University), Else build a morphological data set, with the advantage of being able to Marie Friis (Swedish Museum of Natural History), Kurt Goth, make a rich data set available simultaneously to multiple research- Michael Krings (Bayerische Staatssammlung für Paläontologie und ers who can mine it for informative data (e.g., following open digital Geologie, Munich), Stephan Schultka (Museum für Naturkunde, morphology guidelines; Davies et al., 2017). Further, paleobotanists Berlin), Rashmi Srivastava (Birbal Sahni Institute of Palaeosciences, and neobotanists need to work together to build these data sets and Lucknow), and Scott Wing (Smithsonian National Museum the tree of death. The tree of life can never be complete without in- of Natural History) for access to Spirematospermum material. cluding past life, but we cannot include extinct taxa without synthe- Comments from Jeff Saarela (Canadian Museum of Nature) and an sizing knowledge of the morphologies for extant species. This task anonymous reviewer, as well as the editors, greatly improved the can be daunting, especially for a group such as monocots that has paper, and we thank them for their efforts. This work was supported a reputation of having “cryptic” morphological characters. While by the U. S. National Science Foundation grants DEB 1257080 fruits and seeds are rich in systematically informative characters, (S.Y.S.) and 1257701 (C.D.S.). vegetative structures including stems and leaves are understudied and likely include more useful features than has been acknowledged SUPPORTING INFORMATION to date. Work on developing better ontologies, and even leveraging resources for crowdsourcing and citizen science (e.g., O’Leary et al., 2018), could help to build the necessary data sets for constructing a Additional Supporting Information may be found online in the comprehensive tree of death. supporting information tab for this article.

LITERATURE CITED CONCLUSIONS

Adams, N. F., M. E. Collinson, S. Y. Smith, M. K. Bamford, F. Forest, P. Malakasi, Eight extinct species represented by fossil seeds record the presence F. Marone, and D. Sykes. 2016. X-rays­ and virtual taphonomy resolve the of past lineages and morphologies within Zingiberales. Extinct spe- first Cissus (Vitaceae) macrofossils from Africa as early-­diverging members cies record different combinations of characters than seen in extant of the genus. American Journal of Botany 103: 1657–1677. taxa. What may seem like incongruence between morphological Angiosperm Phylogeny Group [APG]. 1998. An ordinal classification of the and molecular data sets in inferring phylogenetic relationships is families of flowering plants. Annals of the Missouri Botanical Garden 85: better understood with incorporation of fossil data. The fact that 531–553. the familial affinities of Spirematospermum have been debated for Angiosperm Phylogeny Group [APG]. 2016. An update of the Angiosperm over two decades (largely based on study of Spire. wetzleri) but Phylogeny Group classification for the orders and families of flowering not quantitatively tested points to the need to incorporate fossils plants: APG IV. Botanical Journal of the Linnean Society 181: 1–20. Barnosky, A. D. 2001. Distinguishing the effects of the Red Queen and Court in phylogenetic analyses not only to place the fossils, but also to Jester on Miocene mammal evolution in the northern Rocky Mountains. better understand morphological evolution, homoplasy, and ap- Journal of Vertebrate Paleontology 21: 172–185. parent conflicts in the data. Here, we demonstrated that all zin- Bell, C. D., D. E. Soltis, and P. S. Soltis. 2010. The age and diversification of the giberalean fossil seeds belong to Zingiberaceae, except Ensete angiosperms re-­visited. American Journal of Botany 97: 1296–1303. oregonense, which is in Musaceae. While some taxa appear to be Benedict, J. C., S. Y. Smith, M. E. Collinson, J. Leong-Škorničková, C. D. Specht, within crown Zingiberaceae (close to Alpinioideae), others are J. L. Fife, F. Marone, et al. 2015a. Evolutionary significance of seed structure August 2018, Volume 105 • Smith et al.—Building the monocot tree of death • 1399

in Alpinioideae (Zingiberaceae). Botanical Journal of the Linnean Society Greenwood, D. R., and S. L. Wing. 1995. Eocene continental climates and latitu- 178: 441–466. dinal temperature gradients. Geology 23: 1044–1048. Benedict, J. C., S. Y. Smith, M. E. Collinson, J. Leong-Škorničková, C. D. Specht, Heath, T. A., J. P. Huelsenbeck, and T. Stadler. 2014. The fossilized birth–death F. Marone, X. Xiao, and D. Y. Parkinson. 2015b. Seed morphology and anat- process for coherent calibration of divergence-­time estimates. Proceedings of omy and its utility in recognizing subfamilies and tribes of Zingiberaceae. the National Academy of Sciences, USA 111: E2957–E2966. American Journal of Botany 102: 1814–1841. Herendeen, P. S., and P. R. Crane. 1995. The fossil history of monocotyle- Benedict, J. C., S. Y. Smith, C. D. Specht, M. E. Collinson, J. Leong–Škorničková, dons. In P. J. Rudall, P. J. Cribb, D. F. Cutler, and C. J. Humphries [eds.], D. Y. Parkinson, and F. Marone. 2016. Species diversity driven by morpho- Monocotyledons: systematics and evolution, 1–21. Royal Botanic Gardens, logical and ecological disparity: a case study of comparative seed mor- Kew, UK. phology and anatomy across a large monocot order. AoB Plants 8(1): Ho, S. Y. W., and S. Duchêne. 2014. Molecular-clock­ methods for estimating plw063. evolutionary rates and timescales. Molecular Ecology 23: 5947–5965. Chandler, M. E. J. 1925. The Upper Eocene flora of Hordle, Hants. Printed for the Iles, W. J. D., S. Y. Smith, M. A. Gandolfo, and S. W. Graham. 2015. Monocot fos- Palaeontographical Society, London, UK. sils suitable for molecular dating analyses. Botanical Journal of the Linnean Collinson, M. E., M. C. Boulter, and P. L. Holmes. 1993. Magnoliophyta Society 178: 346–374. (Angiospermae). In M. J. Benton [ed.], The fossil record 2, 810–841. Jain, R. K. 1963. Studies in Musaceae. I. Musa cardiosperma sp. nov., a fossil Chapman & Hall, London, UK. banana fruit from the Deccan Intertrappean Series, India. Palaeobotanist 12: Conran, J. G., J. M. Bannister, and D. E. Lee. 2009. Earliest orchid macrofos- 45–54. sils: Early Miocene Dendrobium and Earina (Orchidaceae: Epidendroideae) Janssen, T., and K. Bremer. 2004. The age of major monocot groups inferred from from New Zealand. American Journal of Botany 96: 466–474. 800+ rbcL sequences. Botanical Journal of the Linnean Society 146: 385–389. Conran, J. G., J. M. Bannister, D. E. Lee, R. J. Carpenter, E. M. Kennedy, T. Knobloch, E., and D. H. Mai. 1986. Monographie der Früchte und Samen in Reichgelt, and R. E. Fordyce. 2015. An update of monocot macrofossil data der Kreide von Mitteleuropa. Rozpravy Ústředního ústavu Geologického 47: from New Zealand and Australia. Botanical Journal of the Linnean Society 1–219. 178: 394–420. Koch, B. E., and W. L. Friedrich. 1970. Geologisch-Paäobotanische­ Untersuchung Crepet, W. L. 2008. The fossil record of angiosperms: requiem or renaissance? der Miozänen Braunkohlen bei Fasterholt in Jütland, Dänemark. Bulletin of Annals of the Missouri Botanical Garden 95: 3–33. the Geological Society of Denmark 20: 169–191. Daghlian, C. P. 1981. A review of the fossil record of monocotyledons. Botanical Kress, W. J. 1990. The phylogeny and classification of Zingiberales. Annals of the Review 47: 517–555. Missouri Botanical Garden 77: 698–721. Dahlgren, R. M. T., and F. N. Rasmussen. 1983. Monocot evolution: Characters Kress, W. J., L. M. Prince, W. J. Hahn, and E. A. Zimmer. 2001. Unravelling the and phylogenetic estimation. Evolutionary Biology 16: 255–395. ­evolutionary radiation of the families of the Zingiberales using morphologi- Davies, T. G., I. A. Rahman, S. Lautenschlager, J. A. Cunningham, R. J. Asher, cal and molecular evidence. Systematic Biology 50: 926–944. P. M. Barrett, K. T. Bates, et al. 2017. Open data and digital morphology. Kress, W. J., and C. D. Specht. 2006. The evolutionary and biogeographic ori- Proceedings of the Royal Society, B, Biological Sciences 284: 20170194. gin and diversification of the tropical monocot order Zingiberales. Aliso 22: Doyle, J. A. 1973. Fossil evidence on early evolution of the monocotyledons. 619–630. Quarterly Review of Biology 48: 399–413. Magallón, S., S. Gómez-Acevedo, L. L. Sánchez-Reyes, and T. Hernández- Escapa, I. H., and D. Pol. 2011. Dealing with incompleteness: new advances for Hernández. 2015. A metacalibrated time-­tree documents the early rise of the use of fossils in phylogenetic analysis. Palaios 26: 121–124. flowering plant phylogenetic diversity. New Phytologist 207: 437–453. Fischer, T. C., R. Butzmann, B. Meller, T. Rattei, M. Newman, and D. Mai, D. H. 1999. Die Untermiozänen Floren aus der Spremberger Folge und Holscher. 2009. The morphology, systematic position and inferred biol- dem 2 Flözhorizont in der Lausitz. Teil I. Farnpflanzen, Koniferen und ogy of Spirematospermum – an extinct genus of Zingiberales. Review of Monokotyledonen. Palaeontographica Abteilung B 250: 1–76. Palaeobotany and Palynology 157: 391–426. Mai, D. H., and H. Walther. 1985. Die obereozänen Floren des Weisselster-­ Friedrich, W. L., and B. E. Koch. 1970. Comparison of fruits and seeds of fos- Beckens und seiner Randgebiete. Abhandlungen des Staatlichen Museums sil Spirematospermum (Zingiberaceae) with those of living Cenolophon. für Mineralogie und Geologie zu Dresden 33: 1–260. Bulletin of the Geological Society of Denmark 20: 192–195. Mai, D. H., and H. Walther. 1991. Die oligozänen und untermiozänen Floren Friis, E. M. 1988. Spirematospermum chandlerae sp. nov., an extinct species of Nordwest-­ Sachsens unde des Bitterfelder Raumes. Abhandlungen des Zingiberaceae from the North American Cretaceous. Tertiary Research 9: Staatlichen Museums für Mineralogie und Geologie zu Dresden 38: 1–230. 7–12. Manchester, S. R., P. R. Crane, and L. B. Golovneva. 1999. An extinct genus with Friis, E. M., P. R. Crane, and K. R. Pedersen. 2011. Early flower and angiosperm affinities to extantDavidia and Camptotheca (Cornales) from the Paleocene evolution. Cambridge University Press, Cambridge, UK. of North America and Eastern Asia. International Journal of Plant Sciences Gandolfo, M. A., K. C. Nixon, and W. L. Crepet. 2000. Monocotyledons: a re- 160: 188–207. view of their Early Cretaceous record. In K. L. Wilson and D. A. Morrson Manchester, S. R., and W. J. Kress. 1993. Ensete oregonense sp. nov. from the [eds.], Monocots: systematics and evolution, 44–51. CSIRO, Collingwood, Eocene of western North America and its phylogeographic significance. Australia. American Journal of Botany 80: 1264–1272. Gandolfo, M. A., K. C. Nixon, and W. L. Crepet. 2002. Triuridaceae fossil flowers Matsunaga, K. K. S., S. Y. Smith, S. R. Manchester, D. Kapgate, D. Ramteke, A. from the Upper Cretaceous of New Jersey. American Journal of Botany 89: Garbout, and H. Villarraga-Gómez. 2018. Reinvestigating an enigmatic 1940–1957. Late Cretaceous monocot: morphology, taxonomy, and biogeography Goth, K. 1986. Erster Nachweis von Spirematospermum-Samen aus der of Viracarpon. PeerJ 6: e4580. https://doi.org/10.7717/peerj.4580. Oberkreide von Kössen in Tirol. Courier Forschungsinstitut Senckenberg 86: McElwain, J. C., and S. W. Punyasena. 2007. Mass extinction events and the 171–175. plant fossil record. Trends in Ecology and Evolution 22: 548–557. https://doi. Green, W. A., G. Hunt, S. L. Wing, and W. A. DiMichele. 2011. Does extinction org/10.1016/j.tree.2007.09.003. wield an axe or pruning shears? How interactions between phylogeny and Miller, L., S. Y. Smith, N. D. Sheldon, and C. A. E. Strömberg. 2012. Eocene ecology affect patterns of extinction. Paleobiology 37: 72–91. vegetation dynamics inferred from a high-resolution­ phytolith record. GSA Greenwood, D. R., and J. G. Conran. 2000. The Australian Cretaceous and Bulletin 124: 1577–1589. Tertiary monocot fossil record. In K. L. Wilson and D. A. Morrson [eds.], Nylander, J. A. A., F. Ronquist, J. P. Huelsenbeck, and J. Nieves-Aldrey. 2004. Monocots: systematics and evolution, 52–59. CSIRO, Collingwood, Bayesian phylogenetic analysis of combined data. Systematic Biology 53: Australia. 47–67. 1400 • American Journal of Botany

O’Leary, M. A., K. Alphonse, A. H. Mariangeles, D. Cavaliere, A. Cirranello, T. widespread Cyperaceae: Volkeria messelensis gen. et sp. nov., a fossil Dietterich, M. Julius, et al. 2018. Crowds replicate performance of scientists mapanioid sedge from the Eocene of Europe. American Journal of Botany scoring phylogenetic matrices of phenotypes. Systematic Biology 67: 49–60. 96: 1506–1518. https://doi.org/10.1093/sysbio/syx052. Smith, S. Y., M. E. Collinson, P. J. Rudall, and D. A. Simpson. 2010b. Cretaceous Piperno, D. R. 2006. Phytoliths: A comprehensive guide for archaeologists and and Paleogene fossil record of Poales: review and current research. In O. paleoecologists. AltaMira Press, UK. Seberg, G. Petersen, A. S. Barfod, and J. I. Davis [eds.], Diversity, phylogeny Poinar, G. Jr. 2016a. Beetles with orchid pollinaria in Dominican and Mexican and evolution in the monocotyledons, 333–356. Aarhus University Press, amber. American Entomologist 62: 172–177. Aarhus, Denmark. Poinar, G. Jr. 2016b. Orchid pollinaria (Orchidaceae) attached to stingless bees Smith, S. Y., M. E. Collinson, J. C. Benedict, J. Leong-Škorničková, F. Marone, (Hymenoptera: Apidae) in Dominican amber. Neues Jahrbuch für Geologie and D. Parkinson. 2015. Revision of putative Alpinia (Zingiberaceae) fos- und Paläontologie – Abhandlungen 279: 287–293. sils from the Paleogene and Neogene of western Europe. Palaeontographica Poinar, G. Jr., and F. N. Rasmussen. 2017. Orchids from the past, with a new spe- Abteilung B 293: 101–123. cies in Baltic amber. Botanical Journal of the Linnean Society 183: 327–333. Stockey, R. A. 2006. The fossil record of basal monocots. In J. T. Columbus, E. A. Prasad, V., C. A. E. Strömberg, H. Alimohammadian, and A. Sahni. 2005. Friar, J. M. Porter, L. M. Prince, and M. G. Simpson [eds.], Monocots: com- Dinosaur coprolites and the early evolution of grasses and grazers. Science parative biology and evolution (excluding Poales), 91–106. Rancho Santa 310: 1177–1180. Ana Botanic Garden, Claremont, CA, USA. Prasad, V., C. A. E. Strömberg, A. D. Leaché, B. Samant, R. Patnaik, L. Tang, D. Stockey, R. A., G. L. Hoffman, and G. W. Rothwell. 1997. The fossil mono- M. Mohabey, et al. 2011. Late Cretaceous origin of the rice tribe provides cot Limnobiophyllum scutatum: resolving the phylogeny of Lemnaceae. evidence for early diversification in Poaceae. Nature Communications 2: 480. American Journal of Botany 84: 355–368. Prychid, C. J., P. J. Rudall, and M. Gregory. 2004. Systematics and biology of Strömberg, C. A. E. 2005. Decoupled taxonomic radiation and ecological expan- silica bodies in monocotyledons. Botanical Review 69: 377–440. sion of open-habitat­ grasses in the Cenozoic of North America. Proceedings Ramírez, S. R., B. Gravendeel, R. B. Singer, C. R. Marshall, and N. E. Pierce. 2007. of the National Academy of Sciences, USA 102: 11980–11984. Dating the origin of the Orchidaceae from a fossil orchid with its pollinator. Strömberg, C. A. E. 2011. Evolution of grasses and grassland ecosystems. Annual Nature 448: 1042–1045. Reviews of Earth and Planetary Sciences 39: 517–544. Rodriguez-de la Rosa, R. A., and S. R. S. Cevallos-Ferriz. 1994. Upper Cretaceous Strömberg, C. A. E., R. E. Dunn, R. H. Madden, M. J. Kohn, and A. A. Carlini. zingiberalean fruits with in situ seeds from southeastern Coahuila, Mexico. 2013. Decoupling the spread of grasslands from the evolution of grazer-­type International Journal of Plant Sciences 155: 786–805. herbivores in South America. Nature Communications 4: 1478. https://doi. Ronquist, F., S. Klopfstein, L. Vilhelmsen, S. Schulmeister, D. L. Murray, and A. org/10.1038/ncomms2508. P. Rasnitsyn. 2012a. A total-­evidence approach to dating with fossils, applied Sutton, M. D. 2008. Tomographic techniques for the study of exceptionally pre- to the early radiation of the Hymenoptera. Systematic Biology 61: 973–999. served fossils. Proceedings of the Royal Society, B, Biological Sciences 275: Ronquist, F., M. Teslenko, P. van Der Mark, D. L. Ayres, A. Darling, S. Höhna, B. 1587–1593. Larget, et al. 2012b. Mrbayes 3.2: efficient Bayesian phylogenetic inference Swenson, N. G. 2011. The role of evolutionary processes in producing biodiver- and model choice across a large model space. Systematic Biology 61: 539–542. sity patterns, and the interrelationships between taxonomic, functional and Rudall, P. J. 2000. ‘Cryptic’ characters in monocotyledons: homology and coding. phylogenetic biodiversity. American Journal of Botany 98: 472–480. In R. Scotland and R. T. Pennington [eds.], Homology and systematics: coding Swofford, D. L. 2002. PAUP*: phylogenetic analysis using parsimony (*and other characters for phylogenetic analysis, 114–123. Taylor and Francis, London, UK. methods), version 4.0a159. Sinauer, Sunderland, MA, USA. Sass, C., W. J. D. Iles, C. Barrett, S. Y. Smith, and C. D. Specht. 2016. Revisiting Tank, D. C., J. M. Eastman, M. W. Pennell, P. S. Soltis, D. E. Soltis, C. E. Hinchliff, the Zingiberales: Using multiplexed exon capture to resolve ancient and re- J. W. Brown, et al. 2015. Nested radiations and the pulse of angiosperm di- cent phylogenetic splits in a charismatic plant lineage. PeerJ 4: e1584. versification: increased diversification rates often follow whole genome du- Smith, S. A., J. M. Beaulieu, and M. J. Donoghue. 2010a. An uncorrelated relaxed-­ plications. New Phytologist 207: 454–467. clock analysis suggests an earlier origin for flowering plants. Proceedings of Wiens, J. J. 2017. What explains patterns of biodiversity across the tree of life? the National Academy of Sciences, USA 107: 5897–5902. BioEssays 39: 1600128. Smith, S. Y. 2013. The fossil record of non-commelinid monocotyledons. In Wilf, P., and I. H. Escapa. 2015. Green Web or megabiased clock? Plant fos- P. Wilkin and S. J. Mayo [eds.], Early events in monocot evolution, 29–59. sils from Gondwanan Patagonia speak on evolutionary radiations. New Cambridge University Press, Cambridge, UK. Phytologist 207: 283–290. Smith, S. Y., M. E. Collinson, and P. J. Rudall. 2008. Fossil Cyclanthus Wing, S. L., and D. R. Greenwood. 1993. Fossils and fossil climates: The case for (Cyclanthanceae, Pandanales) from the Eocene of Germany and England. equable continental interiors in the Eocene. Philosophical Transactions of the American Journal of Botany 95: 688–699. Royal Society of London, B, Biological Sciences 341: 243–252. Smith, S. Y., M. E. Collinson, D. A. Simpson, P. J. Rudall, F. Marone, and M. Zanne, A. E., D. C. Tank, W. K. Cornwell, J. M. Eastman, S. A. Smith, R. G. Stampanoni. 2009a. Virtual taphonomy using synchrotron tomographic mi- FitzJohn, D. J. McGlinn, et al. 2014. Three keys to the radiation of angio- croscopy reveals cryptic features and internal structure of modern and fossil sperms into freezing environments. Nature 506: 89–92. plants. Proceedings of the National Academy of Sciences, USA 106: 12013–12018. Zhang, C., T. Stadler, S. Klopfstein, T. A. Heath, and F. Ronquist. 2016. Total-­ Smith, S. Y., M. E. Collinson, D. A. Simpson, P. J. Rudall, F. Marone, and M. evidence dating under the fossilized birth–death process. Systematic Biology Stampanoni. 2009b. Elucidating the affinities and habitat of ancient, 65: 228–249.