Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology Syst. Biol. 71(2):301–319, 2022 © The Author(s) 2021. Published by Oxford University Press on behalf of the Society of Systematic Biologists. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. DOI:10.1093/sysbio/syab035 Advance Access publication May 13, 2021 A Comprehensive Phylogenomic Platform for Exploring the Angiosperm Tree of Life WILLIAM J. BAKER1,∗ , PAUL BAILEY1 , VANESSA BARBER1 , ABIGAIL BARKER1 , SIDONIE BELLOT1 , DAVID BISHOP1 , LAURA R. BOTIGUÉ1,2 , GRACE BREWER1 , TOM CARRUTHERS1 , JAMES J. CLARKSON1 , JEFFREY COOK1 , R OBYN S. COWAN1 , STEVEN DODSWORTH1,3 , NIROSHINI EPITAWALAGE1 , ELAINE FRANÇOSO1 , BERTA GALLEGO1 , MATTHEW G. JOHNSON4 , JAN T. KIM1,5 , KEVIN LEEMPOEL1 , OLIVIER MAURIN1 , CATHERINE MCGINNIE1 , LISA POKORNY1,6 , SHYAMALI R OY1 , MALCOLM STONE1 , EDUARDO TOLEDO1 , NORMAN J. WICKETT7 , ALEXANDRE R. ZUNTINI1 , WOLF L. EISERHARDT1,8 , PAUL J. KERSEY1 , ILIA J. LEITCH1 , AND FÉLIX FOREST1 1 Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AE, UK; 2 Centre for Research in Agricultural Genomics, Campus UAB, Edifici CRAG, Bellaterra Cerdanyola del Vallès, 08193 Barcelona, Spain; 3 School of Life Sciences, University of Bedfordshire, University Square, Luton LU1 3JU, UK; 4 Department of Biological Sciences, Texas Tech University, Lubbock, TX 79409, USA; 5 Department of Computer Science, School of Physics, Engineering and Computer Science, University of Hertfordshire, Hatfield, Hertfordshire AL10 9AB, UK; 6 Centre for Plant Biotechnology and Genomics (CBGP) UPM-INIA, 28223 Pozuelo de Alarcón (Madrid), Spain; 7 Plant Science and Conservation, Chicago Botanic Garden, 1000 Lake Cook Road, Glencoe, IL 60022, USA; and 8 Department of Biology, Aarhus University, 8000 Aarhus C, Denmark *Correspondence to be sent to: Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AE, UK; E-mail: w.baker@kew.org Wolf L. Eiserhardt, Paul J. Kersey, Ilia J. Leitch, and Félix Forest are joint senior authors. Received 23 February 2021; reviews returned 6 May 2021; accepted 8 May 2021 Associate Editor: Deren Eaton Abstract.—The tree of life is the fundamental biological roadmap for navigating the evolution and properties of life on Earth, and yet remains largely unknown. Even angiosperms (flowering plants) are fraught with data gaps, despite their critical role in sustaining terrestrial life. Today, high-throughput sequencing promises to significantly deepen our understanding of evolutionary relationships. Here, we describe a comprehensive phylogenomic platform for exploring the angiosperm tree of life, comprising a set of open tools and data based on the 353 nuclear genes targeted by the universal Angiosperms353 sequence capture probes. The primary goals of this article are to (i) document our methods, (ii) describe our first data release, and (iii) present a novel open data portal, the Kew Tree of Life Explorer (https://treeoflife.kew.org). We aim to generate novel target sequence capture data for all genera of flowering plants, exploiting natural history collections such as herbarium specimens, and augment it with mined public data. Our first data release, described here, is the most extensive nuclear phylogenomic data set for angiosperms to date, comprising 3099 samples validated by DNA barcode and phylogenetic tests, representing all 64 orders, 404 families (96%) and 2333 genera (17%). A “first pass” angiosperm tree of life was inferred from the data, which totaled 824,878 sequences, 489,086,049 base pairs, and 532,260 alignment columns, for interactive presentation in the Kew Tree of Life Explorer. This species tree was generated using methods that were rigorous, yet tractable at our scale of operation. Despite limitations pertaining to taxon and gene sampling, gene recovery, models of sequence evolution and paralogy, the tree strongly supports existing taxonomy, while challenging numerous hypothesized relationships among orders and placing many genera for the first time. The validated data set, species tree and all intermediates are openly accessible via the Kew Tree of Life Explorer and will be updated as further data become available. This major milestone toward a complete tree of life for all flowering plant species opens doors to a highly integrated future for angiosperm phylogenomics through the systematic sequencing of standardized nuclear markers. Our approach has the potential to serve as a much-needed bridge between the growing movement to sequence the genomes of all life on Earth and the vast phylogenomic potential of the world’s natural history collections. [Angiosperms; Angiosperms353; genomics; herbariomics; museomics; nuclear phylogenomics; open access; target sequence capture; tree of life.] Discovering the tree of life is among the most However, despite substantial progress, the evolutionary fundamental of the grand challenges in science today connections among Earth’s ca. 330,000 flowering plant (Hinchliff et al. 2015). The tree of life is the biological species (WCVP 2020) remain incompletely known. roadmap that allows us to discover, identify and The angiosperm research community were early classify life on Earth, to explore its properties, to and organized adopters of the molecular phylogenetic understand its origins and evolution, and to predict approach, resulting in numerous benchmark tree of how it will respond to future environmental change. life publications (e.g., Chase et al. 1993; Soltis et al. Of all eukaryotic lineages, the angiosperms (flowering 2008, 2011), and a community approach to phylogenetic plants) are among the most pressing priorities for tree classification (APG 1998; APG II 2003; APG III 2009; of life research. Angiosperms sustain the terrestrial APG IV 2016). Through this distributed effort, a wealth living world, including humanity, as primary producers, of DNA sequence data is now available in public ecosystem engineers, and earth system regulators. They repositories, covering ca. 107,000 (31%) of the ca. 350,000 hold potential solutions to global challenges, such as species of vascular plants (RBG Kew 2016; WCVP 2020), climate change, biodiversity loss, human health, food most of which are angiosperms (see also Cornwell security, and renewable energy (Antonelli et al. 2020). et al. 2019). However, the lack of sequence data for the In light of this, a phylogenetic framework with which to remaining 69% obstructs their accurate placement in navigate and interpret the species, trait and functional the tree of life. In addition, lack of complementarity diversity of angiosperms has never been more necessary. in gene sampling across public DNA sequence data 301 [09:18 27/1/2022 Sysbio-OP-SYSB210035.tex] Page: 301 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 302 SYSTEMATIC BIOLOGY VOL. 71 impedes phylogenetic synthesis (Hinchliff and Smith the HTS method most widely applied to herbarium 2014). For example, data from either one or both of DNA (Dodsworth et al. 2019). This approach (also rbcL and matK, the two most popular plastid genes for known as target enrichment, target capture, sequence phylogenetics, are available for only 54% of the ca. 107,000 capture, anchored hybrid enrichment) and its variations sequenced vascular plant species (RBG Kew 2016). (e.g., Hyb-Seq, which combines target sequence capture Comprehensive phylogenetic trees of flowering plants with genome skimming) use RNA or DNA probes to are in high demand (Hinchliff et al. 2015; Eiserhardt et al. enrich sequencing libraries for specifically targeted loci 2018), but currently can only be made “complete” using (Faircloth et al. 2012; Lemmon et al. 2012; Weitemier et al. proxies, such as taxonomic classification, to interpolate 2014). It is proving to be an increasingly cost-effective the unsequenced species (Smith and Brown 2018), means of isolating hundreds of loci for phylogenetic which may not accurately reflect relationships. Greater analysis from even centuries-old specimens (Brewer community-wide coordination of both taxon and gene et al. 2019), bringing comprehensive taxon sampling sampling would benefit phylogenetic data integration from herbarium collections within the reach of any immensely, creating numerous downstream scientific phylogenomic researcher (Hale et al. 2020). opportunities. Numerous target sequence probe sets have been High-throughput sequencing (HTS) now promises to developed for specific angiosperm groups (e.g., significantly deepen our understanding of evolutionary Annonaceae [Couvreur et al. 2019], Asteraceae [Mandel relationships among Earth’s species, including et al. 2014], Dioscorea [Soto Gomez et al. 2019], Euphorbia angiosperms (Li et al. 2019; Yang et al. 2020). For [Villaverde et al. 2018]). The design of these probe sets example, the One Thousand Plant Transcriptomes is informed by available genomic resources, as well as (1KP) initiative has brought an unprecedented scale criteria specific to the group of interest and research of data to bear on the plant tree of life (Wickett et al. questions. As a result, locus overlap between probe sets 2014; Gitzendanner et al. 2018; Leebens-Mack et al. tends to be minimal. Unlike the Sanger sequencing era, 2019). Nevertheless, with greatly increased data depth in which researchers converged on tractable genes such come trade-offs in taxon sampling; the pre-eminent as rbcL and matK, the lack of complementarity between HTS studies cited here account for less than 0.01% probe sets curtails prospects for data integration across of angiosperm species. Undeterred by this sampling broad taxonomic scales. In addition, development of gap, the Earth Biogenome Project has launched a custom probe sets is expensive, requiring considerable “moonshot for biology” by proposing to sequence and genomic resources and bioinformatic expertise. A characterize the genomes of all of Earth’s eukaryotic publicly available, universal probe set for angiosperms species over a 10-year period (Lewin et al. 2018). targeting a standard set of loci would resolve these Projects such as the 10,000 Plant Genomes Project issues (Buddenhagen et al. 2016; Chau et al. 2018). In (Cheng et al. 2018) and the Darwin Tree of Life Project response to this, we designed the Angiosperms353 (https://www.darwintreeoflife.org/) aim to contribute probe set (Johnson et al. 2019), drawing on 1KP to this goal by producing numerous chromosome-level transcriptome data from ca. 650 species across the genome assemblies across major lineages and regional angiosperms (Leebens-Mack et al. 2019). The probe set biotas. However, taxon sampling remains a significant targets 353 genes from 410 low-copy, protein-coding issue, due to the challenges of obtaining the high nuclear orthologs previously selected for phylogenetic molecular weight DNA required by these projects (for analysis across green plants (Leebens-Mack et al. 2019), long-read HTS) from samples that are both authentically enriching up to ca. 260 kbp from any flowering plant. identified and compliant with the spirit and letter of Angiosperms353 probes are an open data resource that the Nagoya Protocol (Secretariat of the Convention can be used without the expense of design or access to on Biological Diversity 2011). Despite its immense prior genomic data (Baker et al. 2021) and have already potential, the “whole genome” approach to discovering been successfully applied across different taxonomic the tree of life remains a future goal that will not be scales (e.g., Larridon et al. 2019; Murphy et al. 2020; achieved on a large taxonomic scale in the short term. Pérez-Escobar et al. 2021; Shee et al. 2021), including at Methodological compromises are required to accelerate the population level (Van Andel et al. 2019; Slimp et al. progress. 2021; Beck et al. 2021). The world’s natural history collections are a goldmine Here, we describe a large-scale effort to establish for genomic research (Buerki and Baker 2016), containing a new phylogenomic platform for exploring the tissues of almost all species of life on Earth known angiosperm tree of life, comprising a set of open to science. However, the condition of these tissues tools (Angiosperms353 probes, laboratory protocols, and the DNA therein varies widely, depending on analysis pipeline, data portal) and data (sequence data, age and preservation techniques, among other factors. assembled genes, alignments, gene trees, species tree). In the case of plants, herbarium specimens generally This platform, which directly addresses the challenges yield degraded DNA, which, though not useful for outlined above, is an outcome of the Plant and Fungal long-read HTS, is now being intensively exploited for Trees of Life project (PAFTOL; www.paftol.org) at the short-read HTS (Bakker et al. 2016; Brewer et al. 2019; Royal Botanic Gardens, Kew (RBG Kew 2015). As a Forrest et al. 2019; Alsos et al. 2020). In this context, step toward the ultimate goal of a complete species- target sequence capture is growing in popularity as level tree, we aim to gather DNA sequence data for the [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 302 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 303 Angiosperms353 genes from one species of all 13,862 angiosperm genera (WCVP 2020). This unprecedented data set of standard loci draws extensively on herbarium collections for comprehensive sampling, especially of genera that have not been sequenced before (Brewer et al. 2019). Extensive new data have been generated, analyzed and released into the public domain, along with corresponding phylogenetic inferences. By providing our data in open and accessible ways, including an interactive tree of life, we aim to foster a transparent and collaborative environment for future data re- use and synthesis. This article serves as the baseline reference for our platform, (i) documenting our methods, (ii) describing our first data release, comprising 17% of angiosperm genera, including initial insights on phylogenetic performance, and (iii) presenting a novel data portal, the Kew Tree of Life Explorer, through which our data and corresponding tree of life can be interrogated and downloaded. We conclude with reflections on the prospects for our approach, future development requirements and the role of open data for enhancing cross-community collaboration toward a complete tree of life. MATERIALS AND METHODS This section describes the workflow (Fig. 1) used by the PAFTOL project to generate our first data release (i.e., Data Release 1.0), which is publicly accessible through our open data portal, the Kew Tree of Life Explorer (https://treeoflife.kew.org), described below. The workflow consists of three main stages: (i) sample processing, encompassing sample selection and laboratory protocols for target sequence capture data generation (Fig. 2), (ii) data analysis, including target gene assembly, data mining, data validation and phylogenetic inference (Figs. 2 and 3), and (iii) data publication via the data portal (Fig. 4). The data accessible via the portal comprise raw data (unprocessed sequence reads) and results from “first pass” analyses (gene assemblies, alignments, gene trees, species tree). Though not exhaustive, these first explorations of the FIGURE 1. Summary workflow. Overview of steps taken by the data apply methods that are both rigorous and tractable PAFTOL project to generate Data Release 1.0 of the Kew Tree of Life Explorer (https://treeoflife.kew.org). The stages of the workflow are at our scale of operation. further elaborated in Figs. 2–4. Details of the first data release are also given in the data release notes in the portal via our secure FTP (http://sftp.kew.org/pub/treeoflife/) and are also archived at the Royal Botanic Gardens, Kew (RBGK) to the complete list of angiosperms within the World Research Repository (https://doi.org/10.34885/paftol). Checklist of Vascular Plants (WCVP 2020), which A new release note will be published in the same currently recognizes 13,862 accepted genera in 418 locations with each future data release and will detail families, aligned to the 64 orders of the APG IV any changes in methods used relative to the first release classification (APG IV 2016). We prioritized genera described here. that were not represented by published transcriptomic or genomic data in public sequence repositories (e.g., GenBank), and avoided genera that had already been sampled in large genomic initiatives such as the 1KP Sampling project (Leebens-Mack et al. 2019). The selection of We aimed to generate novel data from across the species within genera was made pragmatically, although angiosperms, using a stratified sampling approach of we prioritized the species of the generic type where one species per genus. Our sampling was standardized possible. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 303 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 304 SYSTEMATIC BIOLOGY VOL. 71 Sampling and DNA extraction Public data mining Tissue PULVERIZATION Target sequence 1KP Annotated Herbarium Mixer Mill MM400 capture transcriptomes genomes Silica-dried Sequence assembly Live plant DNA EXTRACTION Seed CTAB Coding Raw Assembled DNA PURIFICATION reads sequences transcripts (CDS) AMPure XP beads TRIMMING DNA Genomic DNA sample Trimmomatic DNA QC Trimmed Angiosperms353 Fragment size Fragment size read pairs > 350 bp < 350 bp RETRIEVAL BLASTN Library preparation SONICATION Covaris M220 Angiosperms353 Focused-ultrasonicator MAPPING TBLASTN LIBRARY PREPARATION WITH DUAL INDEXING NEBNext Ultra II + NEBNext Multiplex Oligos for Illumina Mapped sequences DNA Library library QC ASSEMBLY overlapSerial POOLING 20-24 selected libraries / pool Pooling and hybridisation RECOVERY FILTERING Library Phylogenetic analysis pool Recovered genes HYBRIDISATION Angiosperms353 probes ALIGNMENT Target- UPP enriched FILTERING & TRIMMING pool AMAS AMPLIFICATION OF TARGET-ENRICHED Aligned NEBNext Q5 HotStart HiFi PCR sequences Master Mix GENE TREE ESTIMATION Library pool IQ-TREE pool QC PURIFICATION Gene Agencourt AMPure trees XP Beads LONG BRANCH REMOVAL TreeShrink (run once) MULTIPLEXING 48-384 samples/pool DNA sequencing SPECIES TREE ESTIMATION Sequencing ASTRAL III pool SEQUENCING Illumina HiSeq/MiSeq Species tree FAMILY IDENTIFICATION Raw VALIDATION reads Preliminary tree only (See Fig.3) FIGURE 2. Sample processing and data analysis workflows. Sample processing (left): processes are indicated by bold headings with reagents and machines used given below; quality control (QC) checkpoints are indicated. Data analysis (right): pipeline products are shown in circles (available to download via the Kew Tree of Life Explorer, https://treeoflife.kew.org); processes are indicated by bold headings with programs used given below. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 304 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 305 FIGURE 3. Family identification validation workflow. Processes are indicated by bold headings. Embedded table (bottom right) indicates decisions made for each sample based on the two validation steps. Plant material was obtained from a variety of allowed us to record multiple sequence data sets sources (Fig. 2), primarily from the collections (fastq files) from one or several libraries, and one or of RBGK (herbarium, DNA bank, silica gel- several DNA extracts from a single specimen. Relevant dried tissue collection, living collection and voucher specimen information was also captured the Millennium Seed Bank, https://www. in the database (e.g., collector(s), collector number, kew.org/science/collections-and-resources/collections). herbarium acronym (following Index Herbariorum Additional material (tissue samples, extracted DNA) http://sweetgum.nybg.org/science/ih/), country of was generously provided by individuals in our origin, date of collection, specimen barcodes). Voucher collaborative networks (see Acknowledgements). To be data are available via our data portal (see below). Images selected, the material must have been (i) legally sourced of specimens sampled from the RBGK Herbarium are and made available for use in phylogenomic studies, in the process of being captured in RBGK’s online (ii) identified to species level, preferably by an expert herbarium catalogue (http://apps.kew.org/herbcat/) in the group, and (iii) ideally collected in the wild. As and, where available, are linked to the appropriate far as was practically achievable, we ensured that the records in the Kew Tree of Life Explorer. identity of each sample was substantiated by a voucher specimen deposited in a publicly accessible herbarium. All metadata were captured using a relational DNA Extraction database that allowed us to track processing of samples DNA was extracted from 40 mg of herbarium from the selection of material, through the library material, 20 mg of silica gel-dried material (Chase preparation pipeline to the completion of sequencing. and Hills 1991), or 100 mg of fresh material using a Data were recorded in four main tables (Specimen, modified CTAB extraction method (Doyle and Doyle Sample, Library, Sequencing). The database architecture 1987; Fig. 2). Plant tissue was pulverized using a [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 305 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 306 SYSTEMATIC BIOLOGY VOL. 71 MariaDB database Running on a Galera Cluster Python API Flask Kew secure FTP Running on CentOS 7 Azure Cloud server Using Nginx Frontend JavaScript Nuxt.js (Vue.js) for the trees Metadata Trees Species and Species and genes gene trees Species Gene tree alignments Kew secure FTP Recovered sequenes Kew Herbarium catalogue By gene and Specimen herbarium sheet sample European Nucleotide Archive (ENA) Manifests Sample raw reads Sample, gene and sequence UniProt Gene exemplar specimen FIGURE 4. Data publication workflow. Implementation of the Kew Tree of Life Explorer data portal is illustrated. Arrows indicate data flow from internal repository to public interface. Infrastructural components are shown in upper half; publicly available information is shown in lower half. External links available from the portal are listed in the lower left. Mixer Mill MM400 (Retsch GmbH, Germany). DNA All DNA samples were quality checked for extractions were purified by a magnetic bead clean-up concentration and degree of fragmentation. DNA using Agencourt AMPure XP beads (Beckman Coulter, concentration was measured using a Quantus (Promega, Indianapolis, IN, USA), according to the manufacturer’s Madison, WI, USA) or Qubit (Thermo Fisher Scientific, protocols. Samples obtained from the RBGK DNA bank Inchinnan, UK) fluorometer. DNA fragment size range (http://dnabank.science.kew.org/homepage.html) had was routinely assessed on a 1% agarose gel using been extracted using a modified CTAB method (Doyle ethidium bromide and visualized with a UVP Gel and Doyle 1987) followed by cesium chloride/ethidium Studio (AnalytikJena, Jena, Germany). For samples with bromide density gradient cleaning and dialysis. DNA a low DNA concentration (i.e., not visible on a gel), samples provided by external collaborators had been fragment sizes were assessed on a 4200 TapeStation extracted using a wide variety of extraction methods using Genomic DNA ScreenTape (Agilent Technologies, from living, silica gel-dried and herbarium material. Cheadle, UK). [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 306 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 307 Library Preparation “Angiosperms 353 v1”, Catalogue #308196) following Genomic DNA samples were diluted to 4 ng/L with the manufacturer’s protocol, version 4.0. Hybridization 10 mM Tris (pH 8.0). Those with an average fragment was typically performed at 65˚C for 24 h, with reactions size greater than 350 bp were sonicated to an average topped with 30 L of red Chill-out Liquid Wax (Bio-Rad, fragment size ca. 400 bp, using a Covaris M220 Focused- Hercules, CA, USA) to prevent evaporation. However, ultrasonicator (Covaris, Woburn, MA, USA) by adding for short libraries (i.e., <350 bp) the temperature was 50 L of diluted genomic DNA to a 130 L Covaris reduced to 60˚C, following the recommendations of microAFA tube. The sonication time was adjusted Arbor Biosciences. for each sample based on its average DNA fragment The target-enriched pools were amplified using the size (15–100 s, following the manufacturer’s protocols). KAPA HiFi 2X HotStart ReadyMix PCR Kit (Roche, Additional parameters used were peak incident power Basel, Switzerland) or NEBNext Q5 HotStart HiFi PCR to 50 W, duty factor to 10% and 200 cycles per burst. Master Mix (New England BioLabs, Ipswich, MA, USA) Libraries were prepared using the NEBNext Ultra II for 8–14 cycles. Amplified pools were then purified DNA Library Prep Kit (New England Biolabs, Ipswich, using Agencourt AMPure XP Beads (at 0.9X the sample MA, USA; Fig. 2). Size selection was not employed volume) and eluted in 15 L of 10 mM Tris (pH 8.0). for samples with highly degraded DNA. In the early Products were quantified with a Quantus fluorometer stages of the project, libraries were prepared following and re-amplified if the concentration was below 6 nM, the manufacturer’s protocols exactly, but the majority with three to six PCR cycles (see above). Final products were prepared using half of the recommended volumes were assessed using the TapeStation to determine the throughout to reduce costs. All DNA fragments were distribution of fragment sizes. The target-enriched pools indexed using NEBNext Multiplex Oligos for Illumina were normalized to 6 nM (using 10 nM Tris, pH 8.0) (Dual Index Primer sets 1 and 2, New England Biolabs, and multiplexed for sequencing, with the number of Ipswich, MA, USA). target-enriched pools combined in each sequencing pool The distribution of fragment sizes in each library varying from 2 to 20 (comprising a total of 48–384 was assessed with a 4200 TapeStation using standard samples) depending on the sequencing platform and D1000 tapes. Library concentration was measured using service provider requirements. a Quantus fluorometer. If the library concentration was less than 10 nM, up to eight additional PCR cycles were performed, following the NEBNext Ultra DNA Sequencing II Library Prep Kit protocol with IS5_reamp.P5 and Initially, DNA sequencing was performed on an IS6_reamp.P7 primers (Meyer and Kircher 2010). Library Illumina MiSeq at RBGK with version 3 chemistry quality assessment was then repeated. (Illumina, San Diego, CA, USA) and ran for 600 cycles to generate 2 × 300 bp paired-end reads. Subsequently, DNA sequencing was outsourced (Macrogen, Seoul, Pooling and Hybridization South Korea, or Genewiz, Takeley, UK) and performed on an Illumina HiSeq producing 2 × 150 bp paired- Prior to hybridization (Fig. 2), all libraries were end reads. Raw reads were deposited in the European normalized to 10 nM, using 10 mM Tris (pH 8.0) and then Nucleotide Archive under an umbrella project (accession combined into pools of 20–24 libraries, each containing number PRJEB35285) and can be accessed from the 10 L (0.1 pmol) of each normalized library (i.e., a total of individual sample records in the Kew Tree of Life ca. 600–700 ng DNA in each pool, assuming an average Explorer. fragment size of ca. 450 bp). To ensure even sequencing across all samples in a pool, species for pooling were selected to minimize the range of DNA fragment sizes and ensure a narrow taxonomic breadth. The latter Sequence Assembly criterion was needed because samples that are more Coding sequences were recovered from target- closely related to the taxa used to construct the probe enriched sequence data using our pipeline recoverSeqs set tend to preferentially hybridize. This can lead to an (accessible from our GitHub repository https:// over-representation of their sequences in the DNA data github.com/RBGKew/KewTreeOfLife, pypaftol if appropriate care is not taken when selecting species for “paftools” submodule) to retrieve sequences the sequencing pool. In rare cases, such as smaller pools orthologous to the Angiosperms353 target gene set (ca. 10 libraries) of short fragment (i.e., <300 bp) libraries, (Johnson et al. 2019; https://github.com/mossmatters/ it was necessary to recalculate the standard volume of Angiosperms353). This target set contained multiple normalized libraries to be added to ensure that the final reference sequences per gene, thereby covering a large pool contained ca. 500 ng of DNA. phylogenetic breadth to facilitate read recovery across The pooled libraries were dried in a SpinVac angiosperms. (Eppendorf, Dusseldorf, Germany), resuspended in 8 The process comprised four main stages (Fig. 2), L of 10 mM Tris (pH 8.0) and enriched by hybridizing applied to each sample: (i) sequence reads were with the Angiosperms353 probe kit (Johnson et al. 2019; trimmed using Trimmomatic (Bolger et al. 2014) Arbor Biosciences myBaits Target Sequence Capture Kit, with the following parameters: ILLUMINACLIP: [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 307 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 308 SYSTEMATIC BIOLOGY VOL. 71 : 2:30:10:2:true, LEADING: 10, extract the genomic or transcriptomic sequences TRAILING: 10, SLIDINGWINDOW: 4:20, MINLEN: 40, corresponding to the 353 genes. BLASTN relies on with the adaptor fasta file formatted for palindrome sequence identity (>70%) and the transcript or CDS trimming, (ii) trimmed read pairs were mapped to with the highest identity is considered to be the ortholog the Angiosperms353 target genes with TBLASTN. of a given target. Because initial recovery of genes from A representative reference sequence for each gene 1KP transcripts using the standard Angiosperms353 was then selected by identifying the sequence with target gene set (Johnson et al. 2019) was unsatisfactory, the largest number of mapped reads. (iii) This we used an expanded Angiosperms353 target set to representative gene was used as the reference for improve matching and retrieval of genes. The expanded assembling the gene-specific reads using an overlap- data set is a reduced version of the 1KP alignments based assembly algorithm (–assembler overlapSerial (Leebens-Mack et al. 2019) produced by Johnson option). OverlapSerial was developed specifically et al. (2019) for the design of the Angiosperms353 for this project (see our GitHub repository) with probe set from which non-angiosperm sequences had the aim of improving gene recovery, in terms of been removed and gap-only sites trimmed. The original gene length and number, relative to the widely used expanded target set is available from https://datadryad. HybPiper (Johnson et al. 2016) and was used as org/stash/dataset/doi:10.5061/dryad.s3h9r6j and a follows. First, the reads were aligned to and ordered reformatted version from our GitHub. As with the along the reference sequence based on a minimum novel target sequence capture assemblies, data were alignment size of 50 bases (–windowSizeReference removed from downstream analyses if the sum of the option) with a minimum sequence identity of gene lengths fell below 20% of the median value across 70% (–relIdentityThresholdReference option). all samples. Consecutive reads ordered along the reference sequence were aligned in a pair-wise manner to find read overlaps. If an overlap of at least 30 bases (–windowSizeReadOverlap option) and 90% sequence Family Identification Validation identity (–relIdentityThresholdReadOverlap option) To verify the family identification of our processed was found, the aligned reads were used to construct a samples, we implemented two validation steps, consensus contig with ambiguous bases represented by which were run in parallel (Fig. 3). The two steps “N”. This last parameter resulted in one or more sets consisted of (i) DNA barcode validation, which utilized of aligned reads with 90% sequence identity, each set nuclear ribosomal and plastid barcodes for DNA- being merged into a single contig. In the final stage, based identification, and (ii) phylogenetic validation, the exonerate protein2genome program was used to which checked the placement of each sample in a identify the exon-intron structure within each contig. preliminary tree relative to its expected position based One or more contigs were chosen that best represented on its initial family assignment. Identification checks the structure of the exon(s) in the reference gene chosen below the family level were not conducted due to the in step (ii). If the exons existed in multiple contigs, those incompleteness of adequate reference resources for contigs were joined together to form the recovered gene DNA barcode validation and sparseness of sampling coding sequence. for phylogenetic validation at the genus or species level. Target gene recovery success was assessed for each For barcode validation of target sequence capture sample by calculating the number of genes recovered data (Fig. 3), plastomes and ribosomal DNA were and the sum of the recovered gene lengths. Samples were recovered from raw reads using GetOrganelle (Jin et al. removed from downstream analyses if the sum of the 2020) and subsequently queried against databases of recovered gene lengths fell below 20% of the median reference plant barcodes using BLASTN (Camacho value across all samples. et al. 2009). For 1KP samples, transcriptome assemblies were directly used as queries in BLASTN. Note that we considered the family identity of annotated genomes to be correct and hence a barcode validation was Public Data Mining unnecessary. Six individual barcode reference databases In addition to newly generated target sequence were built from the NCBI nucleotide and BOLD capture data, the Angiosperms353 genes were mined databases (https://www.ncbi.nlm.nih.gov/nuccore; from publicly available genomic data (Fig. 2). For Data https://www.boldsystems.org/, accessed on 29 Release 1.0, we focused on mining data from the 1KP October 2020), one for the whole plastome, and the Initiative (Carpenter et al. 2019; Leebens-Mack et al. remaining five for specific loci (nuclear ribosomal 18S, 2019) and published genomes with gene annotations as well as plastid rbcL, matK, trnL, and trnH-psbA). As (https://plants.ensembl.org/), although other data for samples, the taxonomy of reference sequences was sources (e.g., the Sequence Read Archive) will be standardized to WCVP (WCVP 2020). BLAST results data-mined for future releases. The genes were retrieved were further filtered with a minimum identity >95% from assembled transcript sequences (1KP) or coding and a minimum coverage of reference locus 90% sequences (CDS; genomes) using paftools retrievetargets (except for whole plastomes, for which only a filtering from our pipeline, which uses BLASTN to identify and based on minimum length was applied). [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 308 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 309 Tests could only be completed if a sample’s given after expert review were made based on implausible family was present in the barcode databases and if tree placement (e.g., wrong higher clade) or sample at least one BLAST match remained after filtering. misidentification (e.g., match to another family in the Thus, zero to six barcode tests were conducted per barcode validation). sample. A sample passed an individual test if the All assembled Angiosperms353 gene data from all first ranked BLAST match (ranked by percentage of samples validated for inclusion form the basis of Data identity) confirmed its original family identification Release 1.0. These were made publicly available via the and failed otherwise. The final result of the barcode Kew Tree of Life Explorer. validation following the six individual barcode tests were determined as follows: (i) Confirmed, if one or Phylogeny Estimation more barcode tests matched the family identification of a sample; (ii) Rejected, if more than half of the barcode tests We inferred a phylogenetic tree from all validated gave the same incorrect family identification (requires at data (Data Release 1.0) for presentation in an interactive least two barcode tests); (iii) Inconclusive (otherwise). format in the Kew Tree of Life Explorer. This species tree Further details of the barcode validation methods was estimated from gene trees using the multi-species can be found in Supplementary Material available on coalescent summary method implemented in ASTRAL- Dryad at http://dx.doi.org/10.5061/dryad.ns1rn8ps7. III (Zhang et al. 2018). In addition to the angiosperm The scripts and lists of NCBI and BOLD accessions samples, ten samples representing seven gymnosperm used in barcode databases are available on our GitHub families from the 1KP initiative were mined for repository. Angiosperms353 orthologs (using retrievetargets, as To conduct phylogenetic validation (Fig. 3), a described above) and included in all analyses as preliminary phylogenetic tree was built using outgroup taxa. Our phylogenomic pipeline, available the complete, unvalidated data set, following the from our GitHub repository, is summarized below and phylogenetic methods described below. We then illustrated in Fig. 2. For each gene, DNA sequences were aligned with UPP assessed which nodes best represented each order 4.3.12 (Nguyen et al. 2015). At the start of the alignment and family in the tree. For every node in the tree, two process a set of 1000 sequences were selected for an metrics were calculated for all families and orders: initial backbone tree. Option -M was set to “-1” so that (i) the proportion of samples belonging to a given sequences could be selected within 25% of the median order/family that are descendants of the node, and (ii) full-length sequence. Filtering and trimming of the the proportion of samples descending from the node alignment were performed with AMAS (Borowiec 2016) that belong to the order/family. The two metrics were as follows. Sequences with insufficient coverage (<60%) then multiplied to produce an overall taxon concordance across well occupied columns of each gene alignment score. For each family and order, the highest scoring were removed. Well occupied columns were defined node was subsequently considered to best represent as those with more than 70% of positions occupied. the taxon in the tree (allowing the identification of Then, alignment columns with <0.3% occupancy were outlying samples). A node with a score of 1 for a given removed to remove very rare or unique insertions. order/family is the crown node (most recent common Finally, sequences with a total length of less than 80 bases ancestral node) of that taxon, which is monophyletic were removed, and genes with <30 overlapping bases (at in the tree. See Supplementary Figure S1 available on the 70% threshold mentioned above) were excluded. Dryad for an illustration. Gene trees were estimated with IQ-TREE 2.0.5 (Minh The family identification of each sample was et al. 2020) inferring branch support using the ultrafast determined as (i) Confirmed: if identified as belonging bootstrap method (option -B; Hoang et al. 2017) with to a family whose best scoring node had a taxon the maximum number of iterations set to 1,000 (option concordance score >0.5 and found as a descendant -nm) and using a single model of evolution (option -m of this node in the tree, (ii) Rejected: if identified as GTR+F+R). The use of a single model without testing belonging to a family whose best scoring node had many models of evolution was a pragmatic choice, a taxon concordance score >0.5 but not found as a following Abadi et al. (2019). TreeShrink 1.3.4 (Mai and descendant of this node, or (iii) Inconclusive: if identified Mirarab 2018) was used to remove abnormally long as belonging to a family whose best scoring node had branches from gene trees using default settings, except a taxon concordance score 0.5. Note that for families option -b, which was set to 20. The alignment and represented in the tree by a single sample, the validation gene tree estimation steps were then repeated on the was performed with respect to their orders. If the order samples retained by TreeShrink. Before reconstructing was represented by a single sample, the validation result the species tree using ASTRAL-III, nodes in the gene was coded as inconclusive. trees with bootstrap support values less than 30% The outputs of the phylogenetic and DNA barcode were collapsed using nw_ed from Newick Utilities 1.6.0 validation were combined to identify samples for (Junier and Zdobnov 2010). This value was deduced from automatic inclusion and exclusion from the final data set, interpreting Figure 1 in Hoang et al. (2017), adjusting the and samples for which a decision on inclusion/exclusion standard bootstrap threshold of 10% (recommended for was subject to expert review (Fig. 3). Exclusions ASTRAL-III), to 30% for the ultrafast bootstrap. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 309 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 310 SYSTEMATIC BIOLOGY VOL. 71 All gene alignments, gene trees and the ASTRAL-III been sequenced before. Data for the remainder were species tree are available for download from secure FTP mined from public sources (689 1KP transcriptomes, 61 and the Kew Tree of Life Explorer (also from Dryad). In annotated genomes). The majority of target sequence addition, the species tree is available to browse through capture data were generated from the RBGK collections an interactive tree viewer implemented within the Kew as follows: DNA Bank (43%), herbarium (28%), silica Tree of Life Explorer (see also Supplementary Fig. S2 gel-dried tissue collection (8%), living collection (2%), available on Dryad). and Millennium Seed Bank (0.3%). The remaining 19% of samples included in this study were provided by various collaborators of the PAFTOL project, either as DNA samples or as dried tissue (see Data Portal Implementation Acknowledgements). To disseminate results, a data portal (the Kew Tree of Sequence recovery from all 2522 target sequence Life Explorer; https://treeoflife.kew.org) was designed capture samples (prior to any quality controls) is and implemented (Fig. 4) with a layered architecture that visualized in Figure 5. Eighty-four target sequence comprised: (i) a MariaDB running on a Galera multi- capture samples and eleven 1KP transcriptomes were master cluster as a database management system; (ii) an removed from downstream analyses because the sum of API written in Python using the Flask framework and the gene lengths did not meet the quality threshold of 20% SQLAlchemy library; (iii) a front-end written using the of the median value across all samples. Vue.js framework and Nuxt.js for the tabular data (used to provide access to gene and specimen data) and content pages; (iv) a tree visualization module developed from the open source application PhyD3 (Kreft et al. 2017) Family Identification Validation using D3.js (Bostock 2012) for data visualization; and (v) deployment on a Linux (CentOS 7) server using Nginx The remaining 3177 samples (Table 1) were processed as web server and load balancer. through our sample family identification validation The data, with appropriate metadata and pipeline (Fig. 3, Supplementary Tables S1 and S2 documentation, are available for public download available on Dryad). Of these, 3064 (97%) were over secure FTP (http://sftp.kew.org/pub/treeoflife/) automatically cleared for inclusion and 67 were and the Kew Tree of Life Explorer under a Creative automatically excluded (Supplementary Table S1 Commons Attribution 4.0 International (CC BY 4.0) available on Dryad). The remaining 46 samples were license. When superseded by new releases, archived held for expert review, after which 35 were cleared for earlier releases will remain accessible via secure FTP. inclusion and 11 were excluded due to implausible tree placements. The majority of excluded samples (64 out of 78) were from the novel target sequence capture data, although 14 were 1KP transcriptomes, highlighting the RESULTS risk of sample misidentification in even the most highly curated data sets. Further details regarding the results Initial Data Set obtained during the family identification validation The initial data set prior to processing and analysis by DNA barcoding can be found in Supplementary comprised data from 3272 angiosperm samples, Material available on Dryad. representing 413 families of angiosperms (99%) and The final validated data set for Data Release 1.0 2428 genera (18%; Table 1). We generated novel target consisted of 3099 angiosperm samples (Table 1), only sequence capture data for 2522 of these samples, 5% fewer than were present in the initial data set. These which included 104 angiosperm genera that have never samples represent all 64 orders, 404 families (96%; 212 TABLE 1. Total number of angiosperm samples included at three stages of data release preparation Data source Initial dataset Preliminary tree pre-validation Final tree and Data Release 1.0 Target sequence 2522 2438 2374 capture data (304/1988/2397) (297/1947/2340) (292/1903/2280) 1KP transcriptomes 689 678 664 (254/544/682) (250/530/677) (245/517/663) Annotated genomes 61 61 61 (23/43/59) (23/43/59) (23/43/59) Total 3272 3177 3099 (413/2428/3079) (410/2388/3028) (404/2333/2956) Note: The first column represents all samples available in the initial dataset. The second column indicates samples included in our preliminary tree, prior to family identification validation, but after removal of samples for which the sum of the gene lengths fell below 20% of the median value across all samples. The third column provides numbers for the samples made public in the Kew Tree of Life Explorer, Data Release 1.0, and included in our final phylogenetic tree. Numbers of angiosperm families, genera, and species in each data subset are provided in brackets (as families/genera/species). [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 310 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 311 a) b) 200,000 Sum of gene lengths (bp) 15 20 10 10 100,000 5 0 0 1,000,000 2,000,000 3,000,000 4,000,000 0 100 200 300 Number of mapped reads Number of recovered genes FIGURE 5. Density plots of target sequence recovery from our raw data. Data are presented prior to any filtering, illustrating relationships of sum of gene lengths (bp) to (a) the number of mapped reads and (b) the number of recovered genes. Darker shades indicate greater density of data points. Black (upper and righthand) dotted lines indicate medians of variables and red (lower and righthand) dotted lines indicate the threshold used to remove samples from downstream analyses, set as 20% of the median value across all samples. represented by >1 sample), 2333 genera (17%), and 2956 in Table 2, likely reflects the variation in structure and species (0.01%). metabolite composition of the starting tissue, which is known to impede DNA extraction from various species and its downstream manipulation. This variation is Data Release 1.0: Sequence Quality and Gene Recovery one of the challenges in dealing with samples from a broad taxonomic range such as across the evolutionary Nine statistics were used to assess the sequence diversity of angiosperms. Variation in gene recovery quality across the 3099 samples of Data Release 1.0 across orders is visualized in Supplementary Figure S3 (Table 2). For the 2374 target sequence capture samples, available on Dryad. the mean percentage of on-target reads was 8%, the mean read depth per sample across all recovered genes was 90× with a median value of 38× and the mean Phylogenetic Results percentage length of recovered genes per sample was The final phylogenetic tree as inferred from Data 62%. The number of genes and the sum length of gene Release 1.0 is publicly available in interactive form via sequence recovered per sample were tightly associated the Kew Tree of Life Explorer. In the current release, as expected, varying continuously across the data set the tree is annotated with local posterior probabilities up to the full set of Angiosperms353 genes and a (LPPs, as given by ASTRAL-III) as indicators of branch total gene length of 256.9 kbp, close to the maximum support. Other measures of support (e.g., quartet scores) expected length of 260 kbp for recovering genes with can be found within tree files accessible via the RBGK this target gene set (Fig. 5). The total length of sequence secure FTP. For completeness, the tree is also available recovered from target sequence capture data was shorter in various formats, including Newick (Supplementary than for samples mined for Angiosperms353 genes Fig. S2 available on Dryad). from 1KP transcriptomes or annotated genomes data As a result of filtering and trimming steps during (Table 2). The reason for the shorter length of the alignment, six genes in Data Release 1.0 were recovered genes is that some exons were omitted during excluded from downstream phylogenetic analysis due the process of refining 1KP alignments to select gene to insufficient overlap between sequences. All statistics instances for the design of the Angiosperms353 probes provided below refer to the remaining data set. Thus, (Johnson et al. 2019). These missing exons were however the species tree is based on 347 gene alignments present in the expanded Angiosperms353 target set and totaling 824,878 sequences, 489,086,049 base pairs, and were therefore retrieved during data mining from 1KP 532,260 alignment columns. Of these, 509,987 columns transcriptomes and annotated genomes. The variation (96%) are variable and 475,181 columns (89%) are in performance of target enrichment across different parsimony informative. The proportion of gaps across samples, illustrated by the measures of variability shown all alignments is 61.6% and the median number of [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 311 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 312 SYSTEMATIC BIOLOGY VOL. 71 TABLE 2. Target sequence capture and gene recovery statistics by sample or gene for Data Release 1.0, including the results of mining of genes from the 1KP and annotated genome datasets Median Mean SD Minimum Maximum Raw reads per sample 1.757 × 106 2.822 × 106 3.076 × 106 1.676 × 104 4.054 × 107 Trimmed reads per sample 1.585 × 106 2.549 × 106 2.791 × 106 1.391 × 104 3.605 × 107 Percentage of reads on-target per samplea 5.676 8.020 7.704 0.005 50.953 Read depth per sampleb 38 90 105 5 2243 Read depth per genec 38 97 37 27 226 Recovered genes per sample Target sequence capture data 338 330 24 148 353 1KP transcriptomes 341 328 44 30 353 Annotated genomes 346 341 13 287 353 Recovered genes lengths across all samplesd (bp) Target sequence capture data 387 477 347 48 3564 1KP transcriptomes 717 803 466 50 4689 Annotated genomes 972 1,136 642 45 8601 Sum of recovered gene lengths per sample (bp) Target sequence capture data 1.613 × 105 1.576 × 105 4.355 × 105 3.433 × 104 2.569 × 105 1KP transcriptomes 2.753 × 105 2.627 × 105 6.659 × 105 6.498 × 105 3.674 × 105 Annotated genomes 3.901 × 105 3.876 × 105 1.868 × 104 3.217 × 105 4.273 × 105 Percentage length per recovered genee across all samples Target sequence capture data 63 62 16 27 96 1KP transcriptomes 88 85 10 44 100 Percentage length of recovered genese per sample Target sequence capture data 63 62 14 20 95 1KP transcriptomes 88 84 13 16 100 Note: SD = standard deviation. a Across all recovered genes. b At bases with 4× depth across all recovered genes, calculated by Samtools depth program. c At bases with 4× depth across all samples, calculated by Samtools depth program. d See Supplementary Figure S7 available on Dryad. e Percentage length calculated against each representative target gene. The upper five rows apply to target sequence capture data only. genes per sample is 284 (mean: 265.3, standard deviation Fig. S5c available on Dryad), while the effective number (SD): 64.3, min: 22, max: 347; Supplementary Table S3 of gene trees informing nodes shows the opposite available on Dryad). The median number of samples trend (Supplementary Fig. S5e available on Dryad). per gene alignment is 2421 (mean: 2377.2, SD: 359) Local posterior probabilities show a tendency to be and median alignment length is 1259 (mean: 1533.9, lower (1st quartile) at the deepest taxonomic level SD: 985.7; Table 3). The resulting gene trees are highly (Supplementary Fig. S5a available on Dryad). Major resolved, with a median support across all nodes groups (i.e., monocots, asterids and rosids) show (ultrafast bootstrap) of 98% (mean: 87.8%, SD: 18.560) similar distributions of both local posterior probabilities across all nodes in all gene trees (Supplementary Fig. S4 (Supplementary Fig. S5b available on Dryad) and quartet available on Dryad). Only 1.3% of all nodes in all gene support values (Supplementary Fig. S5d available on trees are very poorly supported (ultrafast bootstrap Dryad), despite the fact that the effective number of gene <30%; Supplementary Fig. S4 available on Dryad) and trees supporting nodes is more variable in monocots thus collapsed prior to species tree inference. Further (Supplementary Fig. S5f available on Dryad), which is statistics for individual gene alignments and gene trees the result of the lower recovery rates for some orders are reported in Table 3 and Supplementary Table S3 in this group such as Alismatales, Commelinales, and available on Dryad. Liliales (Supplementary Fig. S3 available on Dryad). The species tree accommodates 82% of the quartet Discounting taxa represented by a single sample relationships in the gene trees (ASTRAL normalized (193 families, one order), 96% of testable families and quartet score of 0.82). The majority (76.8%) of nodes in the 83% of testable orders were resolved as monophyletic species tree were well-supported (LPP 95%, cf. Sayyari in the species tree. Most of the samples of non- and Mirarab 2016), and only seven nodes were informed monophyletic families and orders could be assigned to by too few gene trees (i.e., <20) to evaluate support. a clade that represents the family or order well, despite Comparing node support in the species tree at different lacking some samples and/or containing some outlier taxonomic levels (Supplementary Fig. S5 available on samples from other taxa (“concordant taxa” where taxon Dryad), median quartet support is progressively higher concordance score >0.5, see Materials and Methods for toward shallower taxonomic levels (Supplementary details). Only five families (Francoaceae, Hernandiaceae, [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 312 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 313 TABLE 3. Properties of the 347 gene alignments and gene trees underpinning the species tree included in the Kew Tree of Life Explorer Data Release 1.0 Median Mean SD Minimum Maximum Number of samples 2421 2377.2 358.8 491 3014 % of total samplesa 77.9 76.5 11.5 15.8 96.9 Alignment length 1259.0 1533.9 985.7 250 8119 % gapsb 58.9 57.9 11.3 14.4 85.8 Variable sites 1224 1469.7 940.6 240 7873 % variable sites 96.6 96.0 2.5 81.5 100 Parsimony informative sites 1137 1369.4 859.3 233 6792 % parsimony informative sites 90.7 90.0 4.20 69.1 98.9 % nodes in gene trees above 30% UFBSc 98.9 98.5 1.3 90.7 99.9 Mean supportc of all nodes 88.1 87.8 2.7 78.9 94.3 Median supportc of all nodes 98.0 97.6 1.8 90.0 100 SD = standard deviation. a Percentage of samples in species tree present in alignment/gene tree. b Percentage of empty cells in each alignment. c UFBS: ultrafast bootstrap. Phyllanthaceae, Pontederiaceae, and Schlegeliaceae, to taxon, specimen, sequence, alignment and tree represented by 11 samples) and two orders (Bruniales data, with associated metadata for the current data and Icacinales, represented by six samples) were so release in accordance with the Toronto guidelines on dispersed that this was not possible (“discordant taxa” pre-publication data sharing (Toronto International where taxon concordance score  0.5). At the family Data Release Workshop Authors 2009). Users can level, 2893 samples were resolved in the expected family, browse by species, gene or interactive phylogenetic two samples were resolved in an unexpected position, tree. The species interface permits searches by order, and 204 samples were not testable because they belonged family, genus, or species, and provides voucher to a discordant family or a family represented by a specimen metadata (including links to online specimen single sample. At the order level, 3060 samples were images, where available), simple sequence metrics, resolved in the expected order, 32 samples were resolved access to assembled genes and raw data. The gene in an unexpected position, and seven samples were not interface documents all Angiosperms353 genes and testable (see Supplementary Tables S4–S6 available on associated metrics, links to gene identities in UniProt Dryad for lists of specimens from singly represented (https://www.uniprot.org/) and provides access to taxa, poorly resolved taxa, and outliers to well-resolved assembled genes across taxa. The tree of life interface taxa, respectively). Placements of all but five genera and enables browsing and taxon searching of the species tree seven families were consistent with the WCVP/APG inferred from the current release data set, as well as tree IV taxonomic hierarchy of genera, families and orders. downloads (as PNG or Newick) and zooming into user- Concordance with existing taxonomy was lower at the defined subtrees. All processed data (assembled genes, genus level, with only 74% of testable genera resolving alignments, gene trees, species trees) and archived as monophyletic and 47 genera (represented by 130 releases are available from RBGK’s secure FTP site samples) being discordant; these numbers partly reflect (http://sftp.kew.org/pub/treeoflife/), whereas raw the deliberate inclusion of multiple samples from genera sequence reads are deposited within the European suspected a priori to be potentially non-monophyletic. Nucleotide Archive (project number PRJEB35285) for In addition to resolving most genera, families and integration within the Sequence Read Archive. orders as monophyletic, our tree supports more than half (58%) of the relationships among orders presented by the Angiosperm Phylogeny Group (APG IV 2016; DISCUSSION Supplementary Fig. S6 available on Dryad). Congruence The new phylogenomic platform described here is with APG IV varies among major clades, being notably a major milestone toward a comprehensive tree of high in magnoliids (100% of APG IV relationships life for all flowering plant species. The sequencing of supported) and monocots (80%), while being a standardized nuclear marker set of this scale for substantially lower in eudicots (47%), especially in rosids so many taxa is unprecedented, opening doors to a (33%). Nodes in our tree that are congruent with APG highly integrated future for angiosperm phylogenetics IV ordinal relationships are slightly better supported on in the genomic era. Much like a “next generation” average (mean LPP 0.98, median 1) than nodes that are rbcL, which underpinned so many Sanger sequencing- incongruent with APG IV (mean LPP 0.75, median 0.94). based plant phylogenetic studies, the Angiosperms353 genes offer opportunities for continuous synthesis of HTS data across angiosperms. The foundational data Tree of Life Explorer set presented here can be re-used or extended for tree The Kew Tree of Life Explorer of life research at almost any taxonomic scale (Johnson (https://treeoflife.kew.org) provides open access et al. 2019; Larridon et al. 2019; Van Andel et al. 2019; [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 313 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 314 SYSTEMATIC BIOLOGY VOL. 71 Murphy et al. 2020; Pérez-Escobar et al. 2021; Shee value to new data being generated with these methods. A et al. 2021; Slimp et al. 2021; Beck et al. 2021).This is stronger understanding of the sequencing requirements the first phylogenetic project to gather novel HTS data (e.g., coverage) for gene recovery from such data could across angiosperms with a stratified taxon sampling at guide new data generation so that Angiosperms353 the genus level. Our sampling strategy systematically genes can be retrieved routinely as a by-product of other and comprehensively represents both the diversity of research. angiosperms and their deep-time diversification. As We took several open data measures to encourage genus-level sampling becomes increasingly complete— community uptake, in both the design of our tools and a target that is well within reach—this backbone will the sharing of our data. The Angiosperms353 probe set substantially increase our ability to study the dynamics itself was designed to be a transparent, “off-the-shelf” of plant diversity over time and revisit long-standing toolkit that is open, inexpensive and accessible to all, questions in systematics (Magallón et al. 2018; Sauquet especially researchers discouraged by the complexity and Magallón 2018; Soltis et al. 2019). Importantly, and cost of custom probe design (Johnson et al. 2019). it will also sharpen the focus on truly intractable Our sequence data for Angiosperms353 genes are openly phylogenetic problems (Yang et al. 2020; Zhao et al. 2020), available via the Kew Tree of Life Explorer and the encouraging the exploration of the biological drivers of Sequence Read Archive, as a public foundation data these phenomena. set shared according to pre-publication best practice Our approach has already led to a burst of community (Toronto International Data Release Workshop Authors engagement. More than a dozen studies utilizing 2009). The Explorer offers enhanced transparency and Angiosperms353 probes are already published (e.g., accessibility by allowing users to navigate the data via a Larridon et al. 2019; Howard et al. 2020; Murphy phylogenetic snapshot of the current release, along with et al. 2020; Pérez-Escobar et al. 2021; Shee et al. metadata (e.g., specimen data) and intermediate data 2021; Slimp et al. 2021; McLay et al. 2021), and (e.g., gene assemblies, alignments, gene trees). Thanks two journal special issues focused on the probe set to these resources, cross-community collaboration via are in preparation (Baker et al. 2021) arising from Angiosperms353 is gaining momentum. a recent symposium (Lagomarsino and Jabaily Our tree, which is based on the most extensive nuclear 2020). The probe set has also been adopted by phylogenomic data set in flowering plants to date, is the Genomics for Australian Plants consortium strongly supported, credible and highly congruent with (https://www.genomicsforaustralianplants.com/), existing taxonomy and many hypothesized relationships which aims to sequence all Australian angiosperm among orders (APG IV 2016; Supplementary Fig. S6 genera, coordinating with the PAFTOL project to available on Dryad). The data confirm both the optimize collective taxonomic coverage. A subset of effectiveness of Angiosperms353 probes across all the Angiosperms353 genes is now accessible for non- major angiosperm clades and the ability of the angiosperm land plants thanks to a probe set developed genes to resolve relationships across taxonomic in parallel (Breinholt et al. 2021), inviting the prospect of scales (Supplementary Fig. S5 available on Dryad). data integration across all land plants. Angiosperms353 Variable sequence recovery notwithstanding (Table 2, genes (as distinct from the Angiosperms353 probes) are Supplementary Fig. S3 available on Dryad), most nodes also being leveraged as components of custom-designed in our tree are underpinned by large numbers of gene probe sets (e.g., Jantzen et al. 2020; Ogutcen et al. trees (Supplementary Fig. S5e available on Dryad), 2021). This approach gives all the integrative benefits of allowing the species tree to be inferred with confidence Angiosperms353, while permitting (i) the tailoring of (Supplementary Fig. S5a available on Dryad) despite Angiosperms353 probes to a taxonomic group by using gene tree conflict (Supplementary Fig. S5c available on more specific target data to increase gene recovery, Dryad). However, even the most strongly supported and (ii) the inclusion of additional loci pertinent to phylogenetic hypotheses must be viewed with caution the research in question. Angiosperms353 probes as they may be biased by model misspecification have also been directly combined with an existing and wrong assumptions. Moreover, our “first pass” custom probe set (Nikolov et al. 2019) as a “probe analyses based on a set of standard methods may not cocktail” in a single hybridization, capturing both sets suit this data set perfectly (see below). Nevertheless, of targets simultaneously with remarkable efficiency our findings are rendered credible by their high (Hendriks et al. 2021). These possibilities render the concordance with taxonomy, an independent point of invidious choice between specific and universal probe reference that has been extensively ground-truthed sets increasingly irrelevant (Kadlec et al. 2017). by pre-phylogenomic DNA data, especially plastid Although target sequence capture is the most cost- loci. Agreement with existing family circumscriptions effective way to retrieve the Angiosperms353 genes at is particularly striking. In contrast, congruence with the current time, the opportunity to mine the genes previously hypothesized relationships among orders from other kinds of HTS data (e.g., shotgun sequence (APG IV 2016) is much lower (Supplementary Fig. S6 data, RNA sequence data) should not be overlooked. available on Dryad). Some of these earlier hypothesized This represents a further opportunity for community ordinal relationships derive from relatively weak engagement, both via mining of public data in the evidence (bootstrap/jackknife >50%; APG IV 2016), Sequence Read Archive, for example, and by adding which may partly explain this disagreement. However, [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 314 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 315 it may also be due to phylogenetic conflict between bases across all gene alignments (0.01%; Supplementary nuclear and plastid genomes, as the established ordinal Table S3 available on Dryad) suggests that gene assembly relationships rest primarily on evidence from plastid was not strongly impacted by divergent gene copies, loci, substantiated more recently by plastid genomes (Li such as paralogs. However, some paralogy may have et al. 2019). It is hardly surprising, then, that a large- gone unnoticed due to the pervasiveness of gene and scale nuclear analysis presents strongly supported, genome duplication in plants (Li and Barker 2020). alternative relationships (Supplementary Fig. S6 Overall, we expect that the occasional presence of available on Dryad). The conundrum remains that these paralogs in our current analysis would more likely lead incongruences are visible at the ordinal backbone, but to inflated estimates of gene tree incongruence, and not the family level. A more comprehensive exploration thus result in reduced support values, than significant of these relationships, the underlying phylogenetic topological biases (Yan et al. 2020). Thus, we consider signal and their systematic implications is currently our tree relatively conservative while acknowledging underway. that we are not yet exploiting the full potential of The analyses presented here are primarily intended as our data. Although a rigorous analysis of paralogy in a window onto the information content of our current Angiosperms353 genes was not tractable for this data data release and are not a complete exploration of release, we look forward to deeper insights emerging the data. Thus, downstream application of the current as community-wide engagement with Angiosperms353 species tree comes with caveats. We used current, widely grows. accepted methods in a pipeline that can be re-run in a semi-automated fashion whenever we release new data. As a consequence, not all possible analysis options and effects have been explored. We anticipate that users of PROSPECTS our data will probe it more rigorously and will tailor both In the immediate future, we will deliver a further data sampling and phylogenomic analyses to their specific release through which we expect to reach the milestone questions. For example, users may leverage our data by of sampling 50% of all angiosperm genera. This enriching a subset with denser sampling of their own to target will be achieved through substantial novel data address more focused evolutionary questions. A further production by PAFTOL and collaborators, augmented by exemplar use case could be deeper re-analysis of our data data mined from public sources. In-depth phylogenetic from raw sequence reads to investigate gene history and analyses of our data and their evolutionary implications conflict. are also underway. Important limitations in our analysis relate to (i) Beyond this point, we see three priority areas in which taxon sampling, (ii) gene selection (ii) gene recovery, future platform developments might be concentrated, (iii) models of sequence evolution, and (iv) paralogy. resources permitting. Firstly, taxon sampling to the Taxon sampling for intermediate data releases is biased genus level must be completed. Our original target of by the current state of progress toward our systematic sampling all angiosperm genera remains, but the mode sampling strategy. This will be addressed in future data of reaching this is likely to evolve. We anticipate an releases and can be adjusted by users of our data. In acceleration in production of Angiosperms353 data by addition, potential phylogenetic biases attributable to the broader community. The completion of generic-level the function or other properties of the Angiosperms353 sampling will require both the integration of community genes remain poorly understood and require further data in the broader angiosperm tree of life as well investigation. Gene recovery relied upon the standard as strategic investment in filling inevitable data gaps Angiosperms353 target file (Johnson et al. 2019), which, for orphan groups. Secondly, numerous opportunities by its universal nature, can yield patchy results. for refinement exist across our methods. For example, However, it has recently been reported that tailoring insights from our data might permit the optimization of target sequences to specific taxonomic groups can the Angiosperms353 probes to improve gene capture. improve recovery (McLay et al. 2021); this will be Efficiency of gene assembly from sequence data can tested in future data releases. Moreover, we are yet also be improved bioinformatically (McLay et al. 2021). to exploit intronic data captured in the “splash zone” However, as costs of sequencing decline, target sequence adjacent to our target exons. By necessity, our “first capture in vitro may no longer be necessary, the target pass” phylogenetic analysis does not explore the fast- genes simply being mined from sufficiently deeply evolving spectrum of methodological options available sequenced genomes. Thirdly, for the full integrative for phylogenomic analysis. For example, we rely on a potential of Angiosperms353 genes to be achieved, simple standard model of sequence evolution, but more infrastructure for aggregating and sharing this coherent sophisticated models accounting for codon positions body of data must be improved. While the Kew Tree or amino acids may improve phylogenetic inference. of Life Explorer provides a proof-of-concept, it is the Potential paralogy is not addressed by our current public data repositories (e.g., NCBI, ENA) that offer the pipeline. The genes underpinning our analysis were greatest prospects of a mechanism to achieve this. To carefully chosen to represent single-copy genes across fully parallel the earlier success of public repositories flowering plants (Johnson et al. 2019; Leebens-Mack for facilitating single-gene phylogenetic trees (e.g., rbcL, et al. 2019). The very low proportion of ambiguous matK), new tools are needed to assist with efficient [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 315 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 316 SYSTEMATIC BIOLOGY VOL. 71 upload and annotation of target capture loci and scripts) are available from the Dryad Digital Repository associated metadata. (https://doi.org/10.5061/dryad.ns1rn8ps7). Even with a completed genus-level angiosperm tree of life well within reach, the monumental task of sampling all species remains. The scale of this challenge is 24-fold SUPPLEMENTARY MATERIAL greater than the genus-level tree toward which we are currently working. However, with sufficient investment, Data available from the Dryad Digital Repository: increased efficiencies and community engagement, such http://dx.doi.org/10.5061/dryad.ns1rn8ps7. an ambition could potentially be realized. Collections- based institutions are poised to play a critical role in this endeavor through increasingly routine molecular FUNDING characterization of their specimens, perhaps as part of digitization programs and are already facilitating the This work was supported by grants from the Calleva growing trend toward species-complete sampling in Foundation and the Sackler Trust to the Plant and phylogenomic studies (e.g., Loiseau et al. 2019; Murphy Fungal Trees of Life project at the Royal Botanic Gardens, et al. 2020; Kuhnhäuser et al. 2021). Our platform Kew. Additional support was received from the Garfield demonstrates how large-scale phylogenomic projects can Weston Foundation, as part of the Global Tree Seed Bank capitalize on natural history collections to achieve a Programme. much more complete sampling than hitherto possible. The growing movement to sequence the genomes of all life on Earth, inspired by the Earth Biogenome Project ACKNOWLEDGEMENTS (Lewin et al. 2018), significantly boosts the prospects for Numerous people have supported this work completing the tree of life for all species but is hampered through collaboration, sharing expertise, providing by the focus on “gold standard” whole genomes samples, supporting acquisition of samples from RBGK requiring the highest quality input DNA. Our platform collections, and assisting with laboratory work, data offers the opportunity to bridge the gap between the collection, specimen digitization and computational ambition of these projects and the vast phylogenomic infrastructure. A full list of acknowledgements is potential of natural history collections. However, as life given in Supplementary Material. Three anonymous on Earth becomes increasingly imperilled, we cannot reviewers, Deren Eaton and Bryan Carstens provided afford to wait. To meet the urgent demand for best invaluable feedback on the manuscript that materially estimates of the tree of life, we must dynamically improved the resulting paper. Particular thanks go to integrate phylogenetic information as it is generated, Kathy Willis, former Director of Science at the Royal providing synthetic trees of life to the broadest Botanic Gardens, Kew, for inspiring the establishment community of potential users (Eiserhardt et al. 2018). Our of the PAFTOL project. platform facilitates this crucial synthesis by providing a cross-cutting data set and directing the community toward universal markers that seem set to play a central role in completing an integrated angiosperm tree of life. REFERENCES Abadi S., Azouri D., Pupko T., Mayrose I. 2019. Model selection may not be a mandatory step for phylogeny reconstruction. Nat. Commun. 10:934. Alsos I.G., Lavergne S., Merkel M.K., Boleda M., Lammers Y., Alberti A., DATA AVAILABILITY Pouchon C., Denoeud F., Pitelkova I., Puşcaş M., Roquet C., Hurdu All data generated in this study are publicly B.-I., Thuiller W., Zimmermann N.E., Hollingsworth P.M., Coissac E. 2020. The treasure vault can be opened: Large-scale genome released under a Creative Commons Attribution 4.0 skimming works well using herbarium and silica gel dried material. International (CC BY 4.0) license and the Toronto Plants. 9:432. guidelines on pre-publication data sharing (Toronto Antonelli A., Fry C., Smith R.J., Simmonds M.S.J., Kersey P.J., Pritchard International Data Release Workshop Authors 2009). H.W., Abbo M.S., Acedo C., Adams J., Ainsworth A.M., Allkin B., Annecke W., Bachman S.P., Bacon K., Bárrios S., Barstow C., The data are accessible via the Kew Tree of Life Battison A., Bell E., Bensusan K., Bidartondo M.I., Blackhall-Miles Explorer (https://treeoflife.kew.org) and our secure FTP R.J., Borrell J.S., Brearley F.Q., Breman E., Brewer R.F.A., Brodie (http://sftp.kew.org/pub/treeoflife/). Raw sequence J., Cámara-Leret R., Campostrini Forzza R., Cannon P., Carine M., reads are deposited in the European Nucleotide Carretero J., Cavagnaro T.R., Cazar M.E., Chapman T., Cheek M., Archive (https://www.ebi.ac.uk/ena/browser/home) Clubbe C., Cockel C., Collemare J., Cooper A., Copeland A.I., Corcoran M., Couch C., Cowell C., Crous P., da Silva M., Dalle under umbrella project PRJEB35285. Scripts and G., Das D., David J.C., Davies L., Davies N., De Canha M.N., other files relating to our phylogenomic pipeline de Lirio E.J., Demissew S., Diazgranados M., Dickie J., Dines T., are available at our GitHub (https://github.com/ Douglas B., Dröge G., Dulloo M.E., Fang R., Farlow A., Farrar K., RBGKew/KewTreeOfLife). Supplementary materials Fay M.F., Felix J., Forest F., Forrest L.L., Fulcher T., Gafforov Y., Gardiner L.M., Gâteblé G., Gaya E., Geslin B., Gonçalves S.C., Gore cited in this paper plus Data Release 1.0 data sets C.J.N., Govaerts R., Gowda B., Grace O.M., Grall A., Haelewaters duplicated from our secure FTP (assembled genes, D., Halley J.M., Hamilton M.A., Hazra A., Heller T., Hollingsworth gene alignments, gene trees, species tree, examples of P.M., Holstein N., Howes M.J.R., Hughes M., Hunter D., Hutchinson [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 316 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 317 N., Hyde K., Iganci J., Jones M., Kelly L.J., Kirk P., Koch H., Grisai- Buerki S., Baker W.J. 2016. Collections-based research in the genomic Greilhuber I., Lall N., Langat M.K., Leaman D.J., Leão T.C., Lee era. Biol. J. Linn. Soc. 117:5–10. M.A., Leitch I.J., Leon C., Lettice E., Lewis G.P., Li L., Lindon H., Liu Camacho C., Coulouris G., Avagyan V., Ma N., Papadopoulos J., Bealer J.S., Liu U., Llewellyn T., Looney B., Lovett J.C., Luczaj L., Lulekal K., Madden T.L. 2009. BLAST+: Architecture and applications. BMC E., Maggassouba S., Malécot V., Martin C., Masera O.R., Mattana Bioinformatics 10:421. E., Maxted N., Mba C., McGinn K.J., Metheringham C., Miles S., Carpenter E.J., Matasci N., Ayyampalayam S., Wu S., Sun J., Yu J., Miller J., Milliken W., Moat J., Moore P.G.P., Morim M.P., Mueller Jimenez Vieira F.R., Bowler C., Dorrell R.G., Gitzendanner M.A., G.M., Muminjanov H., Negrão R., Nic Lughadha E., Nicholson N., Li L., Du W., K. Ullrich K., Wickett N.J., Barkmann T.J., Barker M.S., Niskanen T., Nono Womdim R., Noorani A., Obreza M., O’Donnell Leebens-Mack J.H., Wong G.K.-S. 2019. Access to rna-sequencing K., O’Hanlon R., Onana J.M., Ondo I., Padulosi S., Paton A., Pearce data from 1,173 plant species: the 1000 Plant Transcriptomes T., Pérez Escobar O.A., Pieroni A., Pironon S., Prescott T.A.K., Qi Initiative (1KP). GigaScience. 8:giz126. Y.D., Qin H., Quave C.L., Rajaovelona L., Razanajatovo H., Reich Chase M.W., Hills H.H. 1991. Silica gel: An ideal material for field P.B., Rianawati E., Rich T.C.G., Richards S.L., Rivers M.C., Ross A., preservation of leaf samples for DNA studies. Taxon. 40:215–220. Rumsey F., Ryan M., Ryan P., Sagala S., Sanchez M.D., Sharrock S., Chase M.W., Soltis D.E., Olmstead R.G., Morgan D., Les D.H., Mishler Shrestha K.K., Sim J., Sirakaya A., Sjöman H., Smidt E.C., Smith B.D., Duvall M.R., Price R.A., Hills H.G., Qiu Y.L., Kron K.A., Rettig D., Smith P., Smith S.R., Sofo A., Spence N., Stanworth A., Stara K., J.H., Conti E., Palmer J.D., Manhart J.R., Sytsma K.J., Michaels H.J., Stevenson P.C., Stroh P., Suz L.M., Tambam B.B., Tatsis E.C., Taylor Kress W.J., Karol K.G., Clark W.D., Hedren M., Gaut B.S., Jansen I., Thiers B., Thormann I., Vaglica V., Vásquez-Londoño C., Victor R.K., Kim K.J., Wimpee C.F., Smith J.F., Furnier G.R., Strauss S.H., J., Viruel J., Walker B.E., Walker K., Walsh A., Way M., Wilbraham Xiang Q.Y., Plunkett G.M., Soltis P.S., Swensen S.M., Williams S.E., J., Wilkin P., Wilkinson T., Williams C., Winterton D., Wong K.M., Gadek P.A., Quinn C.J., Eguiarte L.E., Golenberg E., Learn G.H., Woodfield-Pascoe N., Woodman J., Wyatt L., Wynberg R., Zhang Graham S.W., Barrett S.C.H., Dayanandan S., Albert V.A. 1993. B.G. 2020. State of the world’s plants and fungi 2020. Kew: Royal Phylogenetics of seed plants—an analysis of nucleotide sequences Botanic Gardens. from the plastid gene rbcL. Ann. Missouri Bot. Gard. 80:528–580. APG. 1998. An ordinal classification for the families of flowering plants. Chau J.H., Rahfeldt W.A., Olmstead R.G. 2018. Comparison of taxon- Ann. Missouri Bot. Gard. 85:531–553. specific versus general locus sets for targeted sequence capture in APG II. 2003. An update of the Angiosperm Phylogeny Group plant phylogenomics. Appl. Plant. Sci. 6:e1032. classification for the orders and families of flowering plants: Apg Cheng S., Melkonian M., Smith S.A., Brockington S., Archibald J.M., II. Bot. J. Linn. Soc. 141:399–436. Delaux P.-M., Li F.-W., Melkonian B., Mavrodiev E.V., Sun W., Fu APG III. 2009. An update of the Angiosperm Phylogeny Group Y., Yang H., Soltis D.E., Graham S.W., Soltis P.S., Liu X., Xu X., classification for the orders and families of flowering plants: Apg Wong G.K.-S. 2018. 10kp: A phylodiverse genome sequencing plan. III. Bot. J. Linn. Soc. 161:105–121. GigaScience. 7:giy013. APG IV. 2016. An update of the Angiosperm Phylogeny Group Cornwell W.K., Pearse W.D., Dalrymple R.L., Zanne A.E. 2019. What we classification for the orders and families of flowering plants: Apg (don’t) know about global plant diversity. Ecography. 42:1819–1831. IV. Bot. J. Linn. Soc. 181:1–20. Couvreur T.L.P., Helmstetter A.J., Koenen E.J.M., Bethune K., Brandão Baker W.J., Dodsworth S., Forest F., Graham S.W., Johnson R.D., Little S.A., Sauquet H., Erkens R.H.J. 2019. Phylogenomics M.G., McDonnell A., Pokorny L., Tate J.A., Wicke S., Wickett of the major tropical plant family Annonaceae using targeted N.J. Forthcoming 2021. Exploring Angiosperms353: An open, enrichment of nuclear genes. Front. Plant Sci. 9:1941. community toolkit for collaborative phylogenomic research on Dodsworth S., Pokorny L., Johnson M.G., Kim J.T., Maurin O., flowering plants. Amer. J. Bot. Wickett N.J., Forest F., Baker W.J. 2019. Hyb-Seq for flowering plant Bakker F.T., Lei D., Yu J., Mohammadin S., Wei Z., van de Kerke S., systematics. Trends Plant Sci. 24:887–891. Gravendeel B., Nieuwenhuis M., Staats M., Alquezar-Planas D.E., Doyle J.J., Doyle J.L. 1987. A rapid DNA isolation procedure from small Holmer R. 2016. Herbarium genomics: Plastome sequence assembly quantities of fresh leaf tissue. Phytochem. Bull. 19:11–15. from a range of herbarium specimens using an iterative organelle Eiserhardt W.L., Antonelli A., Bennett D.J., Botigué L.R., Burleigh genome assembly pipeline. Biol. J. Linn. Soc. 117:33–43. J.G., Dodsworth S., Enquist B.J., Forest F., Kim J.T., Kozlov A.M., Beck J.B., Markley M.L., Zielke M.G., Thomas J.R., Hale H.J., Williams Leitch I.J., Maitner B.S., Mirarab S., Piel W.H., Pérez-Escobar O.A., L.D., Johnson M.G. 2021. Is Palmer’s elm leaf goldenrod real? Pokorny L., Rahbek C., Sandel B., Smith S.A., Stamatakis A., Vos The Angiosperms353 kit provides within-species signal in Solidago R.A., Warnow T., Baker W.J. 2018. A roadmap for global synthesis ulmifolia s.L. bioRxiv:2021.2001.2007.425781. of the plant tree of life. Am. J. Bot. 105:614–622. Bolger A.M., Lohse M., Usadel B. 2014. Trimmomatic: A Faircloth B.C., McCormack J.E., Crawford N.G., Harvey M.G., flexible trimmer for Illumina sequence data. Bioinformatics. Brumfield R.T., Glenn T.C. 2012. Ultraconserved elements anchor 30:2114–2120. thousands of genetic markers spanning multiple evolutionary Borowiec M.L. 2016. AMAS: A fast tool for alignment manipulation timescales. Syst. Biol. 61:717–726. and computing of summary statistics. PeerJ. 4:e1660. Forrest L.L., Hart M.L., Hughes M., Wilson H.P., Chung K.-F., Tseng Bostock M. 2012. D3.Js—data-driven documents. http://d3js.Org/. Y.-H., Kidner C.A. 2019. The limits of Hyb-Seq for herbarium Breinholt J.W., Carey S.B., Tiley G.P., Davis E.C., Endara L., McDaniel specimens: Impact of preservation techniques. Front. Ecol. Evol. S.F., Neves L.G., Sessa E.B., von Konrat M., Chantanaorrapint S., 7:439. Fawcett S., Ickert-Bond S.M., Labiak P.H., Larraín J., Lehnert M., Gitzendanner M.A., Soltis P.S., Wong G.K.-S., Ruhfel B.R., Soltis D.E. Lewis L.R., Nagalingum N.S., Patel N., Rensing S.A., Testo W., 2018. Plastid phylogenomic analysis of green plants: a billion years Vasco A., Villarreal J.C., Williams E.W., Burleigh J.G. 2021. A target of evolutionary history. Am. J. Bot. 105:291–301. enrichment probe set for resolving the flagellate land plant tree of Hale H., Gardner E.M., Viruel J., Pokorny L., Johnson M.G. life. Appl. Plant. Sci. 9:e11406. 2020. Strategies for reducing per-sample costs in target capture Brewer G.E., Clarkson J.J., Maurin O., Zuntini A.R., Barber V., Bellot sequencing for phylogenomics and population genomics in plants. S., Biggs N., Cowan R.S., Davies N.M.J., Dodsworth S., Edwards Appl. Plant. Sci. 8:e11337. S.L., Eiserhardt W.L., Epitawalage N., Frisby S., Grall A., Kersey Hendriks K., Mandáková T., Hay N.M., Ly E., Hooft van Huysduynen P.J., Pokorny L., Leitch I.J., Forest F., Baker W.J. 2019. Factors A., Tamrakar R., Thomas S.K., Toro-Núñez O., Pires J.C., affecting targeted sequencing of 353 nuclear genes from herbarium Nikolov L.A., Koch M.A., Windham M.D., Lysak M.A., Forest F., specimens spanning the diversity of angiosperms. Front. Plant Sci. Mummenhoff K., Baker W.J., Lens F., Bailey C.D. Forthcoming 2021. 10:1102. The best of both worlds: Combining lineage specific and universal Buddenhagen C., Lemmon A.R., Lemmon E.M., Bruhl J., Cappa J., bait sets in target enrichment hybridization reactions. Appl. Plant. Clement W.L., Donoghue M.J., Edwards E.J., Hipp A.L., Kortyna Sci. M. 2016. Anchored phylogenomics of angiosperms I: Assessing the Hinchliff C.E., Smith S.A. 2014. Some limitations of public sequence robustness of phylogenetic estimates. bioRxiv:086298. data for phylogenetic inference (in plants). PLoS One. 9:e98986. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 317 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 318 SYSTEMATIC BIOLOGY VOL. 71 Hinchliff C.E., Smith S.A., Allman J.F., Burleigh J.G., Chaudhary R., Y., Ralph P., Real D., Riggins C.W., Ruhsam M., Sage R.F., Sakai A.K., Coghill L.M., Crandall K.A., Deng J., Drew B.T., Gazis R., Gude Scascitella M., Schilling E.E., Schlösser E.-M., Sederoff H., Servick K., Hibbett D.S., Katz L.A., Laughinghouse H.D., McTavish E.J., S., Sessa E.B., Shaw A.J., Shaw S.W., Sigel E.M., Skema C., Smith Midford P.E., Owen C.L., Ree R.H., Rees J.A., Soltis D.E., Williams A.G., Smithson A., Stewart C.N., Stinchcombe J.R., Szövényi P., Tate T., Cranston K.A. 2015. Synthesis of phylogeny and taxonomy into J.A., Tiebel H., Trapnell D., Villegente M., Wang C.-N., Weller S.G., a comprehensive tree of life. Proc. Natl. Acad. Sci. USA. 112:12764. Wenzel M., Weststrand S., Westwood J.H., Whigham D.F., Wu S., Hoang D.T., Chernomor O., von Haeseler A., Minh B.Q., Vinh L.S. Wulff A.S., Yang Y., Zhu D., Zhuang C., Zuidof J., Chase M.W., 2017. UFBoot2: Improving the ultrafast bootstrap approximation. Pires J.C., Rothfels C.J., Yu J., Chen C., Chen L., Cheng S., Li J., Li Mol. Biol. Evol. 35:518–522. R., Li X., Lu H., Ou Y., Sun X., Tan X., Tang J., Tian Z., Wang F., Howard C.C., Crowl A.A., Harvey T.S., Cellinese N. 2020. Wang J., Wei X., Xu X., Yan Z., Yang F., Zhong X., Zhou F., Zhu Y., Peeling back the layers: The complex dynamics shaping the Zhang Y., Ayyampalayam S., Barkman T.J., Nguyen N.-p., Matasci evolution of the Ledebouriinae (Scilloideae, Asparagaceae). N., Nelson D.R., Sayyari E., Wafula E.K., Walls R.L., Warnow T., An bioRxiv:2020.2011.2002.365718. H., Arrigo N., Baniaga A.E., Galuska S., Jorgensen S.A., Kidder T.I., Jantzen J.R., Amarasinghe P., Folk R.A., Reginato M., Michelangeli F.A., Kong H., Lu-Irving P., Marx H.E., Qi X., Reardon C.R., Sutherland Soltis D.E., Cellinese N., Soltis P.S. 2020. A two-tier bioinformatic B.L., Tiley G.P., Welles S.R., Yu R., Zhan S., Gramzow L., Theißen pipeline to develop probes for target capture of nuclear loci with G., Wong G.K.-S., One Thousand Plant Transcriptomes I. 2019. One applications in Melastomataceae. Appl. Plant. Sci. 8:e11345. thousand plant transcriptomes and the phylogenomics of green Jin J.-J., Yu W.-B., Yang J.-B., Song Y., dePamphilis C.W., Yi T.-S., Li D.-Z. plants. Nature. 574:679–685. 2020. GetOrganelle: A fast and versatile toolkit for accurate de novo Lemmon A.R., Emme S.A., Lemmon E.M. 2012. Anchored hybrid assembly of organelle genomes. Genome Biol. 21:241. enrichment for massively high-throughput phylogenomics. Syst. Johnson M.G., Gardner E.M., Liu Y., Medina R., Goffinet B., Shaw Biol. 61:727–744. A.J., Zerega N.J.C., Wickett N.J. 2016. Hybpiper: Extracting coding Lewin H.A., Robinson G.E., Kress W.J., Baker W.J., Coddington J., sequence and introns for phylogenetics from high-throughput Crandall K.A., Durbin R., Edwards S.V., Forest F., Gilbert M.T.P., sequencing reads using target enrichment. Appl. Plant. Sci. Goldstein M.M., Grigoriev I.V., Hackett K.J., Haussler D., Jarvis E.D., 4:1600016. Johnson W.E., Patrinos A., Richards S., Castilla-Rubio J.C., van Sluys Johnson M.G., Pokorny L., Dodsworth S., Botigue L.R., Cowan R.S., M.-A., Soltis P.S., Xu X., Yang H., Zhang G. 2018. Earth Biogenome Devault A., Eiserhardt W.L., Epitawalage N., Forest F., Kim J.T., Project: Sequencing life for the future of life. Proc. Natl. Acad. Sci. Leebens-Mack J.H., Leitch I.J., Maurin O., Soltis D.E., Soltis P.S., USA. 115:4325–4333. Wong G.K., Baker W.J., Wickett N.J. 2019. A universal probe set for Li H.-T., Yi T.-S., Gao L.-M., Ma P.-F., Zhang T., Yang J.-B., Gitzendanner targeted sequencing of 353 nuclear genes from any flowering plant M.A., Fritsch P.W., Cai J., Luo Y., Wang H., van der Bank M., designed using k-medoids clustering. Syst. Biol. 68:594–606. Zhang S.-D., Wang Q.-F., Wang J., Zhang Z.-R., Fu C.-N., Yang J., Junier T., Zdobnov E.M. 2010. The newick utilities: High-throughput Hollingsworth P.M., Chase M.W., Soltis D.E., Soltis P.S., Li D.-Z. phylogenetic tree processing in the Unix shell. Bioinformatics. 2019. Origin of angiosperms and the puzzle of the Jurassic gap. Nat. 26:1669–1670. Plants. 5:461–470. Kadlec M., Bellstedt D.U., Le Maitre N.C., Pirie M.D. 2017. Targeted Li Z., Barker M.S. 2020. Inferring putative ancient whole-genome NGS for species level phylogenomics:“Made to measure” or “one duplications in the 1000 Plants (1KP) Initiative: access to gene family size fits all”? PeerJ. 5:e3569. phylogenies and age distributions. GigaScience. 9:giaa004. Kreft Ł., Botzki A., Coppens F., Vandepoele K., Van Bel M. 2017. Phyd3: Loiseau O., Olivares I., Paris M., de La Harpe M., Weigand A., A phylogenetic tree viewer with extended phyloXML support for Koubinova D., Rolland J., Bacon C.D., Balslev H., Borchsenius functional genomics data visualization. Bioinformatics. 33:2946– F. 2019. Targeted capture of hundreds of nuclear genes unravels 2947. phylogenetic relationships of the diverse neotropical palm tribe Kuhnhäuser B.G., Bellot S., Couvreur T.L.P., Dransfield J., Henderson Geonomateae. Front. Plant Sci. 10:864. A., Schley R., Chomicki G., Eiserhardt W.L., Hiscock S.J., Baker W.J. Magallón S., Sánchez-Reyes L.L., Gómez-Acevedo S.L. 2018. Thirty 2021. A robust phylogenomic framework for the calamoid palms. clues to the exceptional diversification of flowering plants. Ann. Mol. Phylogenet. Evol. 157:107067. Bot. 123:491–503. Lagomarsino L.P., Jabaily R.S. 2020. Virtual Botany Conference 2020 Mai U., Mirarab S. 2018. TreeShrink: Fast and accurate detection of symposium—Angiosperms353: A new essential tool for plant outlier long branches in collections of phylogenetic trees. BMC systematics. http://2020.botanyconference.org/engine/search/ Genomics. 19:272. index.php?func=detail&aid=941. Mandel J.R., Dikow R.B., Funk V.A., Masalia R.R., Staton S.E., Kozik Larridon I., Villaverde T., Zuntini A.R., Pokorny L., Brewer G.E., A., Michelmore R.W., Rieseberg L.H., Burke J.M. 2014. A target Epitawalage N., Fairlie I., Hahn M., Kim J., Maguilla E., Maurin enrichment method for gathering phylogenetic information from O., Xanthos M., Hipp A.L., Forest F., Baker W.J. 2019. Tackling rapid hundreds of loci: an example from the Compositae. Appl. Plant. radiations with targeted sequencing. Front. Plant Sci. 10:1655. Sci. 2:1300085. Leebens-Mack J.H., Barker M.S., Carpenter E.J., Deyholos M.K., McLay T.G.B., Gunn B.F., Ning W., Tate J.A., Nauheimer L., Joyce E.M., Gitzendanner M.A., Graham S.W., Grosse I., Li Z., Melkonian M., Simpson L., Schmidt-Lebuhn A.N., Baker W.J., Forest F., Jackson C.J. Mirarab S., Porsch M., Quint M., Rensing S.A., Soltis D.E., Soltis Forthcoming 2021. New targets acquired: Improving locus recovery P.S., Stevenson D.W., Ullrich K.K., Wickett N.J., DeGironimo L., from the Angiosperms353 probe set. Appl. Plant. Sci. Edger P.P., Jordon-Thaden I.E., Joya S., Liu T., Melkonian B., Miles Meyer M., Kircher M. 2010. Illumina sequencing library preparation N.W., Pokorny L., Quigley C., Thomas P., Villarreal J.C., Augustin for highly multiplexed target capture and sequencing. Cold Spring M.M., Barrett M.D., Baucom R.S., Beerling D.J., Benstein R.M., Harbor Protocols. 2010:pdb.prot5448. Biffin E., Brockington S.F., Burge D.O., Burris J.N., Burris K.P., Minh B.Q., Schmidt H.A., Chernomor O., Schrempf D., Woodhams Burtet-Sarramegna V., Caicedo A.L., Cannon S.B., Çebi Z., Chang M.D., Von Haeseler A., Lanfear R. 2020. Iq-tree 2: New models and Y., Chater C., Cheeseman J.M., Chen T., Clarke N.D., Clayton H., efficient methods for phylogenetic inference in the genomic era. Mol. Covshoff S., Crandall-Stotler B.J., Cross H., dePamphilis C.W., Der Biol. Evol. 37:1530–1534. J.P., Determann R., Dickson R.C., Di Stilio V.S., Ellis S., Fast E., Feja Murphy B., Forest F., Barraclough T., Rosindell J., Bellot S., Cowan N., Field K.J., Filatov D.A., Finnegan P.M., Floyd S.K., Fogliani B., R., Golos M., Jebb M., Cheek M. 2020. A phylogenomic analysis of García N., Gâteblé G., Godden G.T., Goh F., Greiner S., Harkess A., Nepenthes (Nepenthaceae). Mol. Phylogenet. Evol. 144:106668. Heaney J.M., Helliwell K.E., Heyduk K., Hibberd J.M., Hodel R.G.J., Nguyen N.-P.D., Mirarab S., Kumar K., Warnow T. 2015. Ultra-large Hollingsworth P.M., Johnson M.T.J., Jost R., Joyce B., Kapralov M.V., alignments using phylogeny-aware profiles. Genome Biol. 16:124. Kazamia E., Kellogg E.A., Koch M.A., Von Konrat M., Könyves K., Nikolov L.A., Shushkov P., Nevado B., Gan X., Al-Shehbaz I.A., Filatov Kutchan T.M., Lam V., Larsson A., Leitch A.R., Lentz R., Li F.-W., D., Bailey C.D., Tsiantis M. 2019. Resolving the backbone of the Lowe A.J., Ludwig M., Manos P.S., Mavrodiev E., McCormick M.K., Brassicaceae phylogeny for investigating trait diversity. New Phytol. McKain M., McLellan T., McNeal J.R., Miller R.E., Nelson M.N., Peng 222:1638–1651. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 318 301–319 Copyedited by: YS MANUSCRIPT CATEGORY: Systematic Biology 2022 BAKER ET AL.—A PHYLOGENOMIC PLATFORM FOR ANGIOSPERMS 319 Ogutcen E., Christe C., Nishii K., Salamin N., Möller M., Perret M. 2021. Soto Gomez M., Pokorny L., Kantar M.B., Forest F., Leitch I.J., Phylogenomics of Gesneriaceae using targeted capture of nuclear Gravendeel B., Wilkin P., Graham S.W., Viruel J. 2019. A customized genes. Mol. Phylogenet. Evol. 157:107068. nuclear target enrichment approach for developing a phylogenomic Pérez-Escobar O.A., Dodsworth S., Bogarín D., Bellot S., Balbuena baseline for Dioscorea yams (Dioscoreaceae). Appl. Plant. Sci. J.A., Schley R., Kikuchi I., Morris S.K., Epitawalage N., Cowan R., 7:e11254. Maurin O., Zuntini A., Arias T., Serna A., Gravendeel B., Torres M.F., Toronto International Data Release Workshop Authors. 2009. Nargar K., Chomicki G., Chase M.W., Leitch I.J., Forest F., Baker W.J. Prepublication data sharing. Nature 461:168–170. Forthcoming 2021. Hundreds of nuclear and plastid loci yield novel Van Andel T., Veltman M.A., Bertin A., Maat H., Polime T., Hille Ris insights into orchid relationships. Amer. J. Bot. Lambers D., Tjoe Awie J., De Boer H., Manzanilla V. 2019. Hidden RBG Kew. 2015. A global resource for plant and fungal knowledge. rice diversity in the Guianas. Front. Plant Sci. 10:1161. Science strategy 2015-2020. Kew: Royal Botanic Gardens. Villaverde T., Pokorny L., Olsson S., Rincón-Barrado M., Johnson RBG Kew. 2016. The State of the World’s Plants report-2016. Kew: Royal M.G., Gardner E.M., Wickett N.J., Molero J., Riina R., Sanmartín Botanic Gardens. I. 2018. Bridging the micro- and macroevolutionary levels in Sauquet H., Magallón S. 2018. Key questions and challenges in phylogenomics: Hyb-Seq solves relationships from populations to angiosperm macroevolution. New Phytol. 219:1170–1187. species and above. New Phytol. 220:636–650. Sayyari E., Mirarab S. 2016. Fast coalescent-based computation of local WCVP. 2020. World Checklist of Vascular Plants, version 2.0. branch support from quartet frequencies. Mol. Biol. Evol. 33:1654– Facilitated by the Royal Botanic Gardens, kew. Published 1668. on the internet; http://wcvp.science.kew.org/, retrieved Secretariat of the Convention on Biological Diversity. 2011. Nagoya 18 November 2020. protocol on access to genetic resources and the fair and equitable Weitemier K., Straub S.C.K., Cronn R.C., Fishbein M., Schmickl sharing of benefits arising from their utilization to the convention R., McDonnell A., Liston A. 2014. Hyb-Seq: Combining target on biological diversity. Montreal: United Nations Environment enrichment and genome skimming for plant phylogenomics. Appl. Programme. Plant. Sci. 2:1400042. Shee Z.Q., Frodin D.G., Cámara-Leret R., Pokorny L. Forthcoming 2021. Wickett N.J., Mirarab S., Nguyen N., Warnow T., Carpenter E., Matasci Reconstructing the complex evolutionary history of the Papuasian N., Ayyampalayam S., Barker M.S., Burleigh J.G., Gitzendanner Schefflera radiation through herbariomics. Front. Plant Sci. 11:258. M.A., Ruhfel B.R., Wafula E., Der J.P., Graham S.W., Mathews S., Slimp M., Williams L.D., Hale H., Johnson M.G. Forthcoming 2021. On Melkonian M., Soltis D.E., Soltis P.S., Miles N.W., Rothfels C.J., the potential of Angiosperms353 for population genomics. Appl. Pokorny L., Shaw A.J., DeGironimo L., Stevenson D.W., Surek B., Plant Sci. Villarreal J.C., Roure B., Philippe H., dePamphilis C.W., Chen T., Smith S.A., Brown J.W. 2018. Constructing a broadly inclusive seed Deyholos M.K., Baucom R.S., Kutchan T.M., Augustin M.M., Wang plant phylogeny. Amer. J. Bot. 105:302–314. J., Zhang Y., Tian Z., Yan Z., Wu X., Sun X., Wong G.K.-S., Leebens- Soltis D.E., Smith S.A., Cellinese N., Wurdack K.J., Tank D.C., Mack J. 2014. Phylotranscriptomic analysis of the origin and early Brockington S.F., Refulio-Rodriguez N.F., Walker J.B., Moore M.J., diversification of land plants. Proc. Natl. Acad. Sci. USA. 111:E4859. Carlsward B.S., Bell C.D., Latvis M., Crawley S., Black C., Diouf D., Yan Z., Du P., Hahn M.W., Nakhleh L. 2020. Species tree inference Xi Z., Rushworth C.A., Gitzendanner M.A., Sytsma K.J., Qiu Y.-L., under the multispecies coalescent on data with paralogs is accurate. Hilu K.W., Davis C.C., Sanderson M.J., Beaman R.S., Olmstead R.G., bioRxiv:498378. Judd W.S., Donoghue M.J., Soltis P.S. 2011. Angiosperm phylogeny: Yang L., Su D., Chang X., Foster C.S.P., Sun L., Huang C.-H., Zhou X., 17 genes, 640 taxa. Amer. J. Bot. 98:704–730. Zeng L., Ma H., Zhong B. 2020. Phylogenomic insights into deep Soltis D.E., Soltis P.S., Chase M.W., Mort M.E., Albach D.C., Zanis phylogeny of angiosperms based on broad nuclear gene sampling. M., Savolainen V., Hahn W.J., Hoot S.B., Fay M.F., Axtell M., Plant Commun. 1:100027. Swensen S.M., Prince L.M., Kress W.J., Nixon K.C., Farris J.S. 2008. Zhang C., Rabiee M., Sayyari E., Mirarab S. 2018. ASTRAL-III: Angiosperm phylogeny inferred from 18s rDNA, rbcL, and atpB Polynomial time species tree reconstruction from partially resolved sequences. Bot. J. Linn. Soc. 133:381–461. gene trees. BMC Bioinformatics 19:153. Soltis P.S., Folk R.A., Soltis D.E. 2019. Darwin review: angiosperm Zhao T., Xue J., Kao S.-m., Li Z., Zwaenepoel A., Schranz M.E., Van de phylogeny and evolutionary radiations. Proc. R. Soc. Lond. B Biol. Peer Y. 2020. Novel phylogeny of angiosperms inferred from whole- Sci. 286:20190099. genome microsynteny analysis. bioRxiv:2020.2001.2015.908376. [09:18 27/1/2022 Sysbio-wwwpaftol.org] Page: 319 301–319