Multiple Origins of Heliozoa from Flagellate Ancestors

Home , Anaeromonadea, Apusozoa, Centrohelid, Conosa, Corticata, Cryptista

Molecular Phylogenetics and Evolution 93 (2015) 331–362

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier.com/locate/ympev

Multiple origins of Heliozoa from ﬂagellate ancestors: New cryptist subphylum Corbihelia, superclass Corbistoma, and monophyly of Haptista, Cryptista, Hacrobia and Chromista q ⇑ Thomas Cavalier-Smith , Ema E. Chao, Rhodri Lewis

Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK article info abstract

Article history: Heliozoan protists have radiating cell projections (axopodia) supported by microtubular axonemes Received 7 April 2015 nucleated by the centrosome and bearing granule-like extrusomes for catching prey. To clarify previously Revised 25 June 2015 confused heliozoan phylogeny we sequenced partial transcriptomes of two tiny naked heliozoa, the Accepted 10 July 2015 endohelean Microheliella maris and centrohelid Oxnerella marina, and the cercozoan pseudoheliozoan Available online 31 July 2015 Minimassisteria diva. Phylogenetic analysis of 187 genes confirms that all are chromists; but centrohelids (microtubules arranged as hexagons and triangles) are not sisters to Endohelea having axonemes in Keywords: transnuclear cytoplasmic channels (triangular or square microtubular arrays). Centrohelids are strongly Cell evolution sister to haptophytes (together phylum Haptista); we explain the common origins of their axopodia and Chromista Haptista haptonema. Microheliella is sister to new superclass Corbistoma (zooflagellate Telonemea and Heliozoan phylogeny Picomonadea, with asymmetric microfilamentous pharyngeal basket), showing that these axopodial Picomonas protists evolved independently from zooflagellate ancestors. We group Corbistoma and Endohelea as Telonema new cryptist subphylum Corbihelia with dense fibrillar interorganellar connections; endohelean axopodia and Telonema cortex are ultrastructurally related. Differently sampled trees clarify why corticate multigene eukaryote phylogeny is problematic: long-branch artefacts probably distort deep multigene phylogeny of corticates (Plantae, Chromista); basal radiations may be contradictorily reconstructed because of their extreme closeness and the Bayesian star-tree paradox. Haptista and Hacrobia are holophyletic, and Chromista probably are. Ó 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).

1. Introduction strengthen the cell cortex of some subgroups (Cavalier-Smith, 2003a,b; Cavalier-Smith and Chao, 2003a). Chromista comprise Chromista include all eukaryotes with chloroplasts of secondary two subkingdoms: Harosa (Heterokonta, Alveolata, Rhizaria) and symbiogenetic red algal origin and all having tubular ciliary hairs, Hacrobia (Haptophyta, Cryptista, Heliozoa) (Cavalier-Smith, and thus most marine algae (e.g. brown algae, diatoms, 2010a). Their deepest phylogeny remains partially unclear and dinoﬂagellates, haptophytes) (Cavalier-Smith, 1981, 1986, 1989, controversial (Burki et al., 2012a; Keeling, 2013) as chromists 2007, 2010a). Chromists also include all descendants from that apparently underwent rapid evolutionary radiation near the time momentous red-algal enslavement (Cavalier-Smith, 2003a, of the symbiogenetic origin of chloroplasts (Cavalier-Smith, 2013a; Keeling, 2009) that later lost one or both characters, thus 2013a), essentially contemporaneously with basal branching of embrace many ecologically and evolutionarily important sec- Plantae, with which some chromist lineages therefore often inter- ondary heterotrophs, notably malarial and related parasites, mingle even on gene-rich phylogenetic trees (e.g. Brown et al., Pseudofungi, ciliates, disparate ﬂagellates, and all axopodial pro- 2013; Burki et al., 2012a). A shared lateral gene transfer (LGT) from tists, e.g. Heliozoa (Yabuki et al., 2012), our focus here. Kingdoms bacteria into chloroplast DNA of haptophytes and cryptophytes Chromista and Plantae together form a clade known as corticates proved common ancestry of their chloroplasts (Rice and Palmer, because of their shared membrane-bounded alveoli that 2006), suggesting that Haptophyta are closer to cryptophytes than to heterokonts with which they were previously grouped as Chro- mophyta (Cavalier-Smith, 1981; after 1993 spelt Chromobiota). q This paper was edited by the Associate Editor J.B. Dacks. Some multigene trees show Hacrobia as a clade (Burki et al., ⇑ Corresponding author. 2009; Hackett et al., 2007; Patron et al., 2007; Janouškovec et al., E-mail address: [email protected] (T. Cavalier-Smith). http://dx.doi.org/10.1016/j.ympev.2015.07.004 1055-7903/Ó 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 332 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

2010), but others (Baurain et al., 2010; Burki et al., 2012a) show data for Minimassisteria, discovered contaminating the O. marina paraphyletic Hacrobia, which would imply that this LGT was culture used for the RNA extraction (Howe et al., 2011). We report ancestral for all chromists but lost by Harosa (Cavalier-Smith, the phylogenetic position of all three using 187 genes from 171 2007), and an extended Chromobiota (i.e. Harosa, Haptophyta, eukaryotes. Microheliella, Telonema, and the previously phylogenet- Heliozoa) weakly holophyletic, with Cryptista more distant as clas- ically hard to place micropinocytotic zooflagellate Picomonas (Not sically assumed (Cavalier-Smith, 1982, 1986, 1989, 2000), but et al., 2007; Seenivasan et al., 2013) form a novel ancient hacrobian grouping with or within Plantae rather than with Harosa. Multi- chromist clade, here made a new cryptist subphylum Corbihelia gene trees disagree concerning the closest relatives of centrohelid because all three (alone in eukaryotes) share characteristic dense heliozoa (classified with haptophytes in Haptista: Cavalier-Smith, microfilamentous connections between mitochondria, nucleus, 2003b) and zooflagellate Telonemea (Shalchian-Tabrizi et al., endomembranes, and cytoskeleton, whose evolutionary signifi- 2006), classified in Cryptista (Cavalier-Smith, 2007), as they can cance was only recently recognised (Yabuki et al., 2012, 2013). This switch positions (Burki et al., 2009, 2012a,b). rules out four of the five previous mutually contradictory indica- Also uncertain is whether Chromista are holophyletic (as on tions of Picomonas relationships, with cryptomonads (Seenivasan 127-gene trees (Burki et al., 2009) and mitochondrial protein et al., 2013; Not et al., 2007), glaucophytes (Burki et al., 2012a), or 42-gene trees (Derelle and Lang, 2012; Zhao et al., 2013)), as classi- green plus red algae (Brown et al., 2013) and Haptista (Yabuki cally argued (Cavalier-Smith, 1986, 1989) and the simplest interpre- et al., 2014). As Picomonas and Telonema are sisters, as seen by tation, or polyphyletic as some multigene trees suggest (Baurain Not et al. (2007), Burki et al. (2013), and Yabuki et al. (2014), and et al., 2010; Burki et al., 2012a,b); apparent polyphyly might be an share a unique asymmetric microfilamentous cytopharyngeal bas- artefact of poor basal resolution amongst short-branch Hacrobia ket (Klaveness et al., 2005; Seenivasan et al., 2013) not known in and Plantae and longer harosan branches wrongly attracting Harosa any other eukaryotes, we group them as corbihelian superclass towards outgroups (Cavalier-Smith, 2009). If Chromista were poly- Corbistoma. As 18S rDNA first weakly suggested (Cavalier-Smith phyletic, sharing former red algal chloroplasts with the same novel and Chao, 2003b), centrohelid heliozoa are strongly sisters of protein-import machinery (Cavalier-Smith, 1999; Stork et al., 2012) haptophytes, confirming the holophyly of Haptista, an infraking- might in theory be attributed to serial tertiary transfers of dom established for haptophytes plus centrohelids only descendants of the singly enslaved red algal chloroplast from an (Cavalier-Smith, 2003b), here reduced in rank to phylum with hap- early cryptophyte to Haptophyta and from them to Harosa after tophytes and Centroheliozoa subphyla. We discuss evolution of the the first photosynthetic chromist evolved that new machinery cytopharynx and microtubule-supported cell extensions (axopodia (Cavalier-Smith et al., 1994; Baurain et al., 2010). Conversely, if and haptonema), alternative feeding modes within Hacrobia, and of chromists are holophyletic and ancestrally photophagotrophic, a tubular ciliary hairs shared by most cryptists, including at least one single red algal enslavement produced the ancestral chromist, sub- telonemid (Cavalier-Smith, 2007; Klaveness et al., 2005; Yabuki sequent plastid losses yielding ancestrally heterotrophic chromist et al., 2014), concluding that ultrastructure, multigene trees, and groups like ciliates, Rhizaria, and Heliozoa (Cavalier-Smith, 1999, shared LGT, together complementarily and convincingly demon- 2007, 2010a; Cavalier-Smith and Scoble, 2013). strate the holophyly and ancestral photophagotrophy of Hacrobia. Heliozoa are a subset of the many chromists with radial cellular Because of suspicions (Leigh et al., 2008; Deschamps and projections (axopodia) supported by an axoneme of interlinked Moreira, 2009) that corticate multigene phylogenies may be seri- microtubules in a geometrically regular array. Axopodia carry ously distorted by the dual evolutionary origin of chromists, ances- granules (‘extrusomes’) that entrap prey and secrete immobilizing trally a eukaryote–eukaryote cellular chimaera with an internally and digestive enzymes. Axopodial protists are polyphyletic, com- enslaved red alga (Cavalier-Smith, 2003a, 2007, 2010a, 2013a; prising subphylum Radiozoa (within the rhizarian phylum Retaria), Keeling, 2009, 2013), we analysed a series of different taxon sam- rhizarian infraclass Phaeodaria of Cercozoa, phylum Heliozoa (with ples to investigate this in detail. We found some evidence of Haptophyta forming infrakingdom Haptista (Cavalier-Smith, taxon-sample dependent reciprocal distortion of chromist and 2003b), and a less speciose array of diverse pseudoheliozoa that plant multigene trees by attraction between red algae (Plantae) evolved independently within Cercozoa and Ochrophyta (an ances- and undetected nuclear gene paralogues of red algal origin in chro- trally and predominantly algal heterokont phylum that includes mists, so undetected paralogues might partly explain why the pseudoheliozoan pedinellids and actinophryids (Cavalier-Smith basal branching order of Plantae and Chromista is often contradic- and Scoble, 2013)). Axopodial skeletons and extrusomes of each tory (Brown et al., 2013; Burki et al., 2007, 2008, 2009, 2012a; group have characteristic ultrastructure, consistent with sequence Hackett et al., 2007). However, we conclude that long-branch arte- evidence assigning them to very different places in the chromist facts, the extremely rapid early radiation of corticate eukaryotes tree (reviewed by Cavalier-Smith and Chao (2012)). Most axopo- leading to misleadingly high support for contradictory branching dial protists are large, easily seen by light microscopy, and belong patterns on Bayesian trees because of the star-tree paradox to groups established in the nineteenth century. Recently we (Yang, 2007), and still insufficient data from some key lineages, discovered three superficially similar species of tiny axopodial are probably also important factors. Despite these limitations, protists that are not closely related: the 2–5 lm granofilosean our more extensively sampled trees allow us to conclude that cercozoan Minimassisteria diva (Howe et al., 2011); the naked Heliozoa sensu Yabuki et al. (2012) was polyphyletic, to improve (non-scaly) 7 lm centrohelid heliozoan Oxnerella marina hacrobian classification, and to give a new integrated picture of (Cavalier-Smith and Chao, 2012); and ultrastructurally novel early ultrastructural diversification of Hacrobia that fits both their 4 lm heliozoan Microheliella maris grouped in class Endohelea with contrasting feeding modes and ultrastructure and our improved the axopodial flagellate Heliomorpha because some axopodial multigene phylogeny. This reappraisal emphasises the coherent microtubules lie in cytoplasmic channels traversing the nucleus fundamental cell biology and phylogenetic unity of all Chromista. (Yabuki et al., 2012). To establish the position of Microheliella and centrohelids, branching deeply in Hacrobia on poorly resolved 18S and 28S rRNA 2. Materials and methods trees, sometimes together (Cavalier-Smith and Chao, 2003b, 2012), and clarify the phylogeny of Hacrobia (Okamoto et al., 2009; RNA was extracted from Oxnerella marina (Cavalier-Smith and Cavalier-Smith, 2010a), we did large-scale transcriptome sequenc- Chao, 2012) (culture including Minimassisteria diva strain CCAP ing on Microheliella and Oxnerella marina. This also gave extensive 1947/1) and Microheliella maris strain CCAP 1945/1 (Yabuki et al., T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 333

Fig. 1. RAxML PROTGAMMALGF tree for 171 eukaryotes based on 50, 854 amino acid positions in 187 protein-coding genes. Support values are not shown for bipartitions with 100% bootstrap support which also had maximal support (1.0) on a separate PhyloBayes CAT-GTR-C tree for the same alignment (the majority); bipartitions not found in CAT are shown by –, and ML support (left 100 pseudoreplicates) and posterior probability support for the consensus CAT tree (right) are shown for all non-maximally supported bipartitions. The number of amino acids included for each taxon follows its name; the three taxa we sequenced are marked by arrows. In ten genera sequences of closely related species are combined as speciﬁed in Supplementary Table S2. The tree is rooted between excavates and Euglenozoa in accord with Cavalier-Smith (2010a). For CAT two chains were run (4819 trees summed after burnin of 4618, maxdiff 1, meandiff 0.01545); the separate trees for each are shown as Figs. 2 and S1 as they did not fully converge for Hacrobia (showing them more mixed with Plantae in one than in the other) or for the positions of Mantamonas and Malawimonas within Sulcozoa or for Trichozoa (sister to Anaeromonadea in one, forming a metamonad clade as in ML but higher up, but to obazoa plus Amoebozoa, Mantamonas and Collodictyon in the other), though most of the topology was identical with maximal support in both chains to ML. Copromyxa, not included in previous amoebozoan multigene trees (Cavalier-Smith et al., 2015) as its sequences only later became available (Eme et al., 2014), is here conﬁrmed to belong in Tubulinea. Scale bars on Figs. 1–7 show the mean number of amino acid substitutions per site. 334 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

2012), cDNA libraries made, multiplex 454-pyrosequencing done, excluding Chromista, 63 chromists only, 45 Hacrobia plus Plantae; and 192 gene alignments (supplementary Table S1) constructed 38 Hacrobia plus Plantae excluding Rhodophyta; 17 Hacrobia only) for 220 eukaryotes as explained in detail previously for all 187 genes. The Hacrobia only alignment had 63% missing (Cavalier-Smith et al., 2014) and concatenated using SCaFoS data, so for Hacrobia only we also analysed a smaller alignment (Roure et al., 2007); RNA extraction methods for these species of 66 genes that excluded all genes missing in >50% of taxa; this are detailed in the electronic supplement. Because of paralogue had <39% missing data overall and gave the same topology with problems or insufficient data with five genes, we excluded them closely similar support. and carried out phylogenetic analysis on 187 genes. To increase gene representation in the hacrobian Picomonas com- 3. Results pared with Cavalier-Smith et al. (2014, 2015) we BLASTed our core alignment sequences against the genomic sequences for picobili- All trees strongly group Oxnerella with Polyplacocystis (wrongly phytes MS584-11, MS584-5, and MS584-22 (Yoon et al., 2011; named Raphidiophrys in early trees; see Cavalier-Smith and Chao http://dbdata.rutgers.edu/data/pico) to identify related genes; (2012)) forming a maximally supported centrohelid heliozoan positives were screened using single-gene trees and incorrect clade. By contrast Minimassisteria invariably groups within Rhi- paralogues and contaminant sequences (e.g. red algal) were zaria with maximal support, specifically within subphylum Filosa removed. This yielded too few authentic genes for strains 5 (none) of phylum Cercozoa with very strong support by CAT and LG ML; and 22 (only vatA) to be worth including, but greatly augmented its precise position as sister to Limnofila (expected from their data for MS584-11 (now represented by 74 of our 187 genes). Our classification within class Granofilosea and 18S rDNA trees: single-gene trees also revealed that 14 sequences for the uncultured Howe et al., 2011) was maximally supported on CAT trees but only phaeodarean Aulacantha (Sierra et al., 2013) are probably contami- weakly with LG (Fig. 1). This strong support for the expected con- nants from many protist groups (especially dinoflagellates and trasting positions of Oxnerella and Minimassisteria confirms that ciliates) or even animals; Sierra et al. did not use single-gene trees our phylogenetic separation of their sequences from the culture to exclude contaminants, a few of which we found in their other mixture using single-gene trees was accurate. sequences from environmental samples (e.g. Collozoum) and removed. 14 ’Aulacantha’ genes were omitted; four (rpl17, 18 and 30; rps 5) appeared to be contaminants of animal origin and 10 3.1. Polyphyly of Heliozoa (calm3, CALR, cfrg, rpl 6, rpl 13E, rpl 37A, Tub B, rps 14, rps 16, rps 6) of non-rhizarian protist origin, though most of these trees were Contrary to some earlier multigene trees, centrohelids never too poorly resolved to be certain about which group they came from. grouped with Cryptista; ML always placed centrohelids as We previously discovered that our Oxnerella data were actually a sister to Haptophyta with good support, a clade corresponding mixture of Oxnerella sequences and those for a granofilosean cerco- exactly to superphylum Haptista (Cavalier-Smith, 2003b). The zoan, an overlooked contaminant of the culture, purified and eukaryote-wide LG tree (Fig. 1) weakly groups Microheliella with a described as Minimassisteria diva (Howe et al., 2011). As previously weakly supported clade comprising the flagellates Telonema and explained (Cavalier-Smith and Chao, 2012), Minimassisteria over- Picomonas; we call the clade comprising these three genera grew and replaced Oxnerella over a few years and we fortunately Corbihelia as its unity has ultrastructural support (see discussion). extracted RNA for transcriptomics near the replacement midpoint, It was also a clade on seven other LG ML trees for different subsets getting partial transcriptomes from two phylogenetically distinct of the taxa shown on Fig. 1, and in all eight LG trees Telonema and axopodial protists in one analysis. Single-gene maximum likelihood Picomonas were sisters (we call this clade Corbistoma, as both have (ML) trees for all 187 genes unambiguously differentiated between a similar asymmetric pharyngeal basket (see discussion)). Support the two; sequences clearly grouping within Cercozoa were for both clades varied considerably with taxon sampling. Both are relabelled Minimassisteria diva. To increase taxon and gene found on some CAT trees (more often for Corbistoma) but not representation in chromists related to Minimassisteria we selected others; most CAT trees excluding the distant podiate and Eozoa orthologues from two protist transcriptomes publicly available outgroups converged, most showing Corbistoma and Corbihelia through the marine microbial eukaryote transcriptome sequencing clades; but those that included podiates usually did not converge project (MMETSP: http://marinemicroeukaryotes.org/resources; because of persistent inconsistencies within basal podiates, accessed January 2014; Keeling et al., 2014) by the same BLAST metamonads and/or Amoebozoa, and usually also in the Hacrobia/ and single-gene tree method: the foraminiferan Sorites sp. Plantae part of the tree, and often did not show Corbihelia as a clade (MMETSP0191) and the cercozoan Mataza sp. (MMETSP0087), iden- and sometimes not Corbistoma either. Fig. 2, a CAT tree for the same tified as a Mataza sp. (robustly close sister to Mataza hastifera but alignment as Fig. 1, exemplifies three major differences invariably sufficiently distinct to be a separate species) by our own compre- found on CAT trees compared with LG ML: (1) Corticates (Plantae hensive cercozoan 18S rDNA tree, not shown). Single-gene trees plus Chromista) are a maximally supported clade. (2) Apusomonads showed that a majority of selected ’Sorites’ sequences were actually are sisters to opisthokonts, so breviates branch one node more from a dinoflagellate, so the uncultured sample may have har- deeply making Apusozoa paraphyletic. (3) Metamonada are more boured an unnoticed parasitic dinoflagellate, which Fig. 1 shows is closely related to podiates than they are to jakobids, Percolozoa more closely related to peridinean dinoflagellates than to Oxyrrhis; and Euglenozoa. All three of these were as found by previous trees putative dinoflagellate parasite and genuinely foraminiferan Sorites using almost the same (Cavalier-Smith et al., 2014, 2015) or substan- sequences were included separately in our trees, as they greatly tially similar gene sets (Brown et al., 2013); for the reasons argued increase gene sampling for their groups, but numerous contami- previously the heterogeneous CAT trees are probably more accurate nants of the Sorites data from other protist groups were removed. for these nodes than the contradictory LG trees. 171 eukaryote-wide taxa were used for phylogenetic analysis Fig. 2 agrees with Fig. 1 in showing clade Haptista (centrohelids, by the best available heterogeneous amino-acid substitution haptophytes) and in Microheliella not grouping with centrohelids; model (PhyloBayes-MPI v.1b GTR-CAT-C-4rates) and for ML by i.e. both LG and CAT show that Heliozoa (centrohelids plus the RAxML-MPI v.7.2.8 PROTGAMMALGF (see supplementary mate- endohelians Microheliella and Heliomorpha) are polyphyletic. rial). This alignment had 53% missing data. We also ran trees for Though this chain grouped Microheliella strongly with Cryptista, seven taxonomically narrower alignments (164 eukaryotes exclud- not Corbistoma, the second chain (supplementary Fig. S1) placed ing Rhodophyta, 143 eukaryotes excluding Plantae, 109 eukaryotes it weakly with Telonema and centrohelids (PP 0.65). T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 335

Fig. 2. PhyloBayes CAT-GTR-C tree for 171 eukaryotes based on 50, 854 amino acid positions in 187 protein-coding genes. Support values are not shown for bipartitions with maximal support (1.0) for this tree (chain 1; 2385 trees summed after removing burnin of 3825 trees); those in red indicate bipartitions not present on a second tree (chain 2; Fig. S1) for the same alignment; because of these speciﬁc differences in topology maxdiff was 1. The number of amino acids included for each taxon follows its name; the three taxa we sequenced are marked by arrows. 336 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Non-convergence of Figs. 2 and S1 affected only five clades (which (Fig. 3); with CAT Hacrobia are a maximally supported clade in therefore lowered support values for the bipartitions shown in red one chain, whereas the other chain had all Hacrobia except on Fig. 2), all part of very close deep radiations: in Fig. S1 Trichozoa Telonema a maximally supported clade but Telonema maximally move to be sister to obazoa plus Collodictyon/Mantamonas,in supported as sister of Harosa (as weakly did the consensus tree, Amoebozoa Stenamoeba and Acanthamoeba/Cochliopodiidae move Fig. 3). In both chains Haptista was maximally supported (79% one node deeper and unite as a sister clade to Conosa, in corticates ML support), reinforcing evidence for polyphyly of Heliozoa. In haptophytes and core Cryptista (cryptomonads, Leucocrypta, and the one with holophyletic Hacrobia, Corbistoma was a maximally Palpitomonas; i.e. Cryptista other than Corbihelia) move into supported clade and Microheliella maximally supported sister to Plantae as paraphyletic sisters of red algae plus Picomonas. Instabil- core Cryptista. LG ML showed weakly supported Corbistoma ity of exact positions for all these taxa was noted previously and Corbihelia (Fig. 3). To circumvent the problem of CAT (Cavalier-Smith et al., 2014, 2015; Burki et al., 2009, 2012a,b). non-convergence and to prevent distortion by distant outgroups However, in contrast to Burki et al. (2012a) for a distinctly different we ran much smaller trees for 63 chromists only (Fig. 4). CAT gene set or Cavalier-Smith et al. (2014, 2015), all with much less now converges well and shows clade Hacrobia with maximal sup- sampling for Hacrobia and Harosa, Haptista are now closer to port and Corbistoma and Corbihelia as clades, grouping the latter Cryptista plus Plantae than to Harosa in all three eukaryote-wide with Cryptista sensu stricto with strong support (0.99). LG ML trees, the most extensively yet sampled. shows exactly the same topology for Hacrobia with strong support Figs. 2 and S1 agree in placing Picomonas strongly as sister to red for Corbistoma and Corbihelia. Hacrobia internal phylogeny is algae (not to Glaucophyta as in Burki et al., 2012a,b). It is evident that exactly the same and consistent with both CAT and ML when out- by CAT, which ought to be the better method, the three plant groups group Harosa are excluded, and also when genes missing in more (glaucophytes, green plants, red algae) and three hacrobian groups than half the taxa are excluded (Fig. 5). (Haptista, Corbihelia, core Cryptista) are part of a very rapid The CAT tree without chromists (Fig. 6) shows with maximal radiation with extremely close branches, and that the three support Rhodophyta and Viridiplantae as sisters and Glaucophyta corbihelids diverged so close to the base of this radiation that there the most divergent plant clade, which is generally accepted as are in effect eight closely radiating groups. Despite that closeness the correct topology for Plantae (Cavalier-Smith, 2007; Figs. 2 and S1 show contradictory positions with maximal support Deschamps and Moreira, 2009), and for which nuclear for Haptista and core Cryptista; they not only contradict each other (Deschamps and Moreira, 2009) and chloroplast multigene trees but also Burki et al. (2012a). We suggest that these contradictions (Shih et al., 2013) give congruent topology. However the corre- are probably examples of the star-tree paradox, where Bayesian sponding LG tree still shows the probably incorrect topology with methods can give spuriously high resolution to a star tree – one with red algae the deepest branch with 87% support. Therefore, a bias an instantaneous radiation having no real internal phylogenetic other than unrecognised red algal genes in chromists must explain signal (Yang, 2007). We do not claim that there is no historically why ML is less accurate than CAT in this case. Deschamps and meaningful signal at the base of Hacrobia/Plantae, merely that it is Moreira (2009) using only opisthokonts as outgroup did not see probably so slight as to be easily overridden by statistical and this conflict for their 143 nuclear genes; both ML and CAT gave systematic biases, and that the star-tree paradox makes it unwise the probably correct topology seen here only with CAT. However, to deduce a lot from high Bayesian support for a particular node in comparison with Figs. 2 and S1 shows that, even when 63 such extremely close radiations from just one taxon/gene sample. chromists are present, the three plant groups branch in the same It is wiser to look for congruence among different taxon and gene correct order by CAT, the difference being that one (Fig. 2) or three samples in an effort to distil signal from noise or bias. (Fig. S1) hacrobian lineages intrude into Plantae as successive sis- One possible source of bias is that Picomonas genomic ters of red algae. This stability of plant topology to the presence sequences were not from a culture but from an environmental or absence of chromists casts doubt on the idea that unrecognised single cell sample, so some contamination by red algal or other for- red algal genes in chromists are a primary cause of past conflicts in eign genes may exist; our single-gene trees did identify a few likely Plantae topology (Leigh et al., 2008; Deschamps and Moreira, contaminants including cyanobacterial and red algal that we 2009). Possibly a more important reason is that both chromist removed, but we cannot be sure that none remain to pull and plants are jointly almost a star radiation with basal branches Picomonas artefactually towards red algae. A second and arguably so close that almost anything can perturb their trees. However, more important potential bias in the same direction stems from as the strength of bias from putative red algal genes would be gene the fundamentally chimaeric nature of chromists; photosynthetic and taxon dependent we cannot eliminate the possibility that it is members at least are chimaeras of a red algal symbiont and a real phenomenon in some alignments. In a preliminary unknown biciliate host. Leigh et al. (2008) and Deschamps and eukaryote-wide CAT analysis with 188 genes and 164 taxa but Moreira (2009) postulated that some red algal genes may be pre- including markedly fewer corbihelian genes, red algae did branch sent undetected amongst chromist genes used for multigene phy- more deeply than other Plantae as weak sisters to Microheliella, logenies and thus distort corticate topology seriously. If true, such consistently with such a red algal gene bias that in theory could genes would tend to pull red algae away from other Plantae have been swamped by adding extra genes if relatively fewer were towards chromists, and chromists retaining more red algal genes of red algal origin; possibly however it was just a misleading con- towards or into Plantae. The degree of tree distortion would sequence of insufficient data for Corbihelia. Tests along the present depend strongly on how many genes were retained in each lineage lines would be a useful safeguard in future work. and included in a given alignment, so could vary greatly among taxon and gene samples. Deschamps and Moreira (2009) argued 3.2. Holophyly of Hacrobia and multigene tree distortions that to get the most accurate phylogeny for Plantae, all chromists should be omitted from the tree to avoid such artefacts; con- To test the idea of bias by undetected symbiogenetic red algal versely, we argue, if such putative red algal genes are indeed pre- genes in chromists further we ran trees omitting only red algae. A sent in chromist alignments, for the most accurate chromist trees small 38-taxon tree for Hacrobia plus Plantae but excluding red we should exclude all Plantae. We therefore performed reciprocal algae (Fig. 7) has a strong bipartition between Plantae and Hacrobia exclusion tests: first excluding Plantae, then chromists. by CAT and ML (0.99, 86%) showing that without Rhodophyta their Without Plantae, Hacrobia are a clade with 95% support by LG subgroups do not intermingle as they did on Figs. 2 and S1. The two ML and Chromista has higher support (93%) than podiates (71%) CAT chains fully converged (maxdiff 0.0385361) but grouped Cor- T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 337

Fig. 3. PhyloBayes CAT-GTR-C tree for 143 eukaryotes excluding Plantae based on 50, 854 amino acid positions in 187 protein-coding genes. The alignment is identical to that used for Figs. 1 and 2 except for excluding all Plantae. Support values are not shown for bipartitions which had maximal support (1.0) as well as 100% bootstrap support on separate RAxML PROTGAMMALGF trees for the same alignment (Figs. S2 and S3); bipartitions not found in ML are shown by –, and CAT consensus tree support (left) and ML support (right 100 pseudoreplicates) are shown for all non-maximally supported bipartitions. The number of amino acids included for each taxon follows its name; arrows mark the three taxa we sequenced. 13,051 trees were summed after removing the ﬁrst 4233 as burnin; however only one chain had exactly the same topology as the consensus tree; the other did not show the two partitions in red as Telonema moved to be sister to Picomonas (red arrow) making Corbistoma a clade (0.89, 59%) and the two Varisulca clades switched places, so maxdiff was 1 (meandiff 0.00735405); ML also showed a Corbihelia clade (61%). 338 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Fig. 4. PhyloBayes CAT-GTR-C tree for 63 chromists based on 50, 854 amino acid positions in 187 protein-coding genes. Bipartition support values are posterior probabilities (left), plus bootstrap percentages (100 pseudoreplicates, right or below) for a separate RAxML PROTGAMMALGF tree; no support values are shown when both were maximal (1.0 or 100%). The CAT trees converged strongly (two chains summed after burnin removal: maxdiff 0.038561). The number of amino acids included for each taxon in the two alignments follows its name; new taxa sequenced are in bold. The tree is rooted between Hacrobia and Harosa, both holophyletic in Fig. 1.

Fig. 5. PhyloBayes CAT-GTR-C tree for 17 Hacrobia using 187 genes (50, 854 amino acids). The first two bipartition support values are posterior probabilities (left; the second is for separate analysis of only 66 genes after excluding all genes absent in >50% of taxa, 13,422 amino acids included); the last two are bootstrap percentages (100 pseudoreplicates, below) for separate RAxML PROTGAMMALGF trees for 187 genes (third) or 66 genes (fourth); black blobs means all four were maximal (1.0 or 100%). Both CAT trees converged (two chains summed after burnin removal: maxdiff 0.108906 for 187 genes, 0.0570038 for 66 genes, i.e. good runs). The number of amino acids included for each taxon in the two alignments follows its name; new taxa sequenced are in bold. The tree is rooted between Haptista and Cryptista, both holophyletic in Fig. 1. bistoma with Haptista rather than other cryptists as did Fig. 4;ML and Plantae, and CAT no longer converged; the plant and hacrobian grouped Microheliella with Corbistoma as clade Corbihelia (42%), lineages became mixed up in three contradictory ways (Figs. 8, S4 but CAT put Microheliella strongly with other cryptists as in Figs. 2 and S5) suggesting that red algae introduce a conflicting signal or and 4. Fig. 7 is consistent with the holophyly of both Plantae and confusing bias; they did not group with Viridiplantae in any of Hacrobia, as are Figs. 1, 4 and 6. In marked contrast, when red algae the three trees as they should have done (and did robustly in are added neither method showed a bipartition between Hacrobia Fig. 6) but in contradictory places within Hacrobia (specifically T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 339

Fig. 6. PhyloBayes CAT-GTR-C tree for 109 eukaryotes excluding chromists based on 50, 854 amino acid positions in 187 protein-coding genes. The alignment is identical to that for Figs. 1 and 2 except for excluding all Chromista. Support values are not shown for bipartitions with 1.0 posterior probability support which also had maximal support (100%) on a separate RAxML PROTGAMMALGF tree for the same alignment (the majority); bipartitions not found in ML are shown by –; and posterior probability support for the consensus CAT tree (left) and ML support (right 100 pseudoreplicates) and are shown for all non-maximally supported bipartitions. The number of amino acids included for each taxon follows its name; the three taxa we sequenced are marked by arrows. Within Trichozoa, the very long stems of clades Parabasalia and Diplomonadida are shortened ﬁvefold. The two CAT chains did not converge (maxdiff 1) because of instability of two lineages only: one chain was exactly as this consensus tree, but in the other planomonads moved up two nodes to be sisters of obazoa (0.98) and Stenamoeba jumped to be weakly sister to Flamella/Filamoeba (0.53); these two changes meant that six clades were not present in that chain, so their support values in red are given for the consensus tree and in brackets for the chain where present. 340 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Fig. 7. PhyloBayes CAT-GTR-C tree for 38 corticates (17 Hacrobia and 21 Plantae excluding red algae) using 187 genes (50, 854 amino acids). Bipartition support values are posterior probabilities (left), plus bootstrap percentages (100 pseudoreplicates, right or below) for a separate RAxML PROTGAMMALGF tree; black blobs mean both were maximal (1.0 or 100%). The CAT trees converged strongly (two chains summed after burnin removal: maxdiff 0.038561). The number of amino acids included for each taxon in the two alignments follows its name; new taxa sequenced are in bold. The tree is rooted between Hacrobia and Plantae.

Fig. 8. RAxML PROTGAMMALGF tree for 45 corticates (17 Hacrobia and 28 Plantae) using 187 genes (50, 854 amino acids). Bipartition support values are bootstrap percentages (left, 100 pseudoreplicates) posterior probabilities (right or below) for a separate PhyloBayes CAT-GTR-C consensus tree; black blobs mean both were maximal (1.0 or 100%). As the CAT trees did not fully converge (maxdiff 1) trees for each chain are shown in supplementary material (Figs. S4 and S5). The number of amino acids included for each taxon follows its name; new taxa sequenced are in bold. The tree is rooted to give the same internal phylogeny of Plantae as in Fig. 7. T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 341 within Cryptista). With ML red algae were weakly (30%) sisters of 3.3. Improved phylogeny of Rhizaria Corbihelia (Fig. 8); in CAT chain 1 sister to Picomonas (0.77 Fig. S2); in chain 2 sister to Microheliella (0.82 Fig. S3). All three posi- All our trees group Minimassisteria and Limnofila together with tions make no biological sense and are probably wrong. moderate (LG) or strong (CAT) support; they represent the two As a further test we also ran trees omitting only red algae from most divergent subclades of the cercozoan class Granofilosea, the Fig. 1/2 eukaryote-wide alignment (Fig. 9 and S4). ML support which 18S rDNA did not always even show as a clade (Bass et al., for Corbihelia rises dramatically from 42% to 77%, and for Corbis- 2009; Howe et al., 2011). This is the first multigene evidence that toma from 48% to 76% (Fig. S6). Support for Haptista rose from this ultrastructurally coherent class is a clade. Our chromist-only 78% to 88%, but surprisingly with intermediate support levels tree, the most comprehensive to date, gives a more accurate phy- between Figs. 1 and 3 for its basal branches. The CAT trees did not logeny than a recent 36-gene tree for Rhizaria (Sierra et al., converge because in one chain Haptista are maximally with other 2013). In all our trees Radiozoa are maximally supported as holo- Hacrobia and Plantae as in Fig. 2 but in the other maximally with phyletic, in agreement with Burki et al. (2010), so the claim that Harosa; this might be another example of near star-phylogeny they are not a clade and Foraminifera branch within them was pre- spurious resolution. Both CAT trees had maximally supported mature (Sierra et al., 2013). Endomyxa were consistently a clade Haptista but a strongly supported Corbistoma (.96, .99) intruded with moderately good support as in one Burki et al. (2010) ML tree, into Plantae as fairly strongly supported sister to green plants. possibly not paraphyletic as in Sierra et al. (2013) and Burki et al. Leigh et al. (2008) showed that 15 ribosomal proteins collectively (2010) CAT and their ML tree without Reticulomyxa, another gave trees incongruent with the majority (35 proteins), wrongly instance of ML happening to give a more accurate tree than CAT attracting the chromist heterokont branch into Plantae as robust when undersampled. The phaeodarian Aulacantha groups as sister to red algae. They suggested these 15 discordant heterokont expected with Mataza in accord with their classification in class proteins may be of red algal origin, having replaced host proteins Thecofilosea (Cavalier-Smith and Chao, 2012). These marked in the ancestral chromist host-rhodophyte chimaera. Figs. 1 and 2 improvements in agreement with morphology, and for Cercozoa show no such attraction of heterokonts towards Plantae, suggesting with rRNA trees, probably stem from including more genes and that the attraction they observed was favoured by low taxon sam- taxa and excluding contaminants from the Sierra et al. environ- pling, and that this bias may be swamped by other genes or lineages mental sample data. As in Burki et al. (2012a,b) support for infrak- in our larger data set. As Leigh et al. had only one heterokont ingdom Halvaria is consistently maximal or almost so. sequence, an in silico chimaera of Phytophthora and Laminaria (Bapteste et al., 2002), we tested this by rerunning the Fig. 6 analysis 3.4. Resolution difficulties in outgroups after adding only Pythium to represent heterokonts. ML does group it within Plantae (99%) specifically as sister to red algae with moderate The non-convergence of the larger CAT trees including support (73% Fig. S5). This fits the Leigh et al. results; their hypothesis non-corticate outgroups (Figs. 2, S1, 3, 6) was not simply because of persisting red algal genes in heterokonts is one possible explana- lineages at the base of Hacrobia/Plantae intermingled in contradiction (others are discussed below). This result is significant also for tory ways described above, but also because of contradictory topol- the hypothesis that the long branches of most Harosa make ogy at the base of (a) the podiate/Malawimonas/Metamonada clade, multigene trees place them too deeply away from both Hacrobia and/or (b) Amoebozoa. One invariable CAT difference from ML was and Plantae and thus wrongly make Chromista appear polyphyletic that Metamonada always grouped with podiates with maximal (Cavalier-Smith, 2009). The oomycete clade comprising Pythium and support, and was never sister to podiates plus corticates as LG Phytophthora is much the shortest branch in Harosa (Figs. 1–3), only ML always more weakly showed; however, though always branch- about half the length of most harosan lineages. The fact that Pythium ing deeply within or close to podiates their exact placement varied is within Plantae and has the same branch length as plants in Fig. S5, with taxon sampling – sister clade to podiates (Fig. 2) or split into but is two nodes lower when numerous long-branch harosans are two separate clades, long-branch Trichozoa separating from included, supports that interpretation. short-branch Anaeromonada; the precise position of Trichozoa var- Using 143 genes and more taxa, Deschamps and Moreira (2009) ied when separate: most often (three times) to obazoa plus Manta- showed Plantae topology to alter in the presence or absence of monas and Collodictyon (Fig. S1, 171 taxa; Fig. 9,as chromist outgroups; they attributed that conflict to unrecognised Sulcozoa/Metamonada converged for the latter this may be the red algal paralogues in chromists pulling red algae too basally. most reliable position); once each as sister to obazoa (Fig. 6 con- The fact that in all our 171 and 164 taxon CAT trees the relative sensus and chain 1) or obazoa plus planomonads (Fig. 6 chain 2) branching order of the three groups of Plantae, even in the pres- or obazoa plus all three varisulcan groups (Fig. 3 one chain). The ence of intruding chromist lineages, is precisely the same as in only other differences between CAT and ML consistent in all taxon our 109 taxon tree and the 28 taxon tree of Deschamps and samples were (a) the incorrect grouping of Selaginella with Physco- Moreira (2009 Fig. 4), both without chromists, means that for our mitrella in land plants by ML and correct grouping with other tra- 187 gene set and much larger taxon samples there is no evidence cheophytes by more realistic CAT, and (b) paraphyly of Apusozoa in that chromists distort Plantae topology. However the presence of CAT only, both noted previously (Cavalier-Smith et al., 2014). Col- Plantae does change hacrobian topology on CAT trees, and the lodictyon position was slightly unstable: sister to Malawimonas in changes are different when red algae are present or not. Fig. 2 but weakly with Mantamonas in all other CAT trees (as also Comparison of Figs. 7 and 8 as well as Figs. 6 and S9 suggest that in Cavalier-Smith et al., 2014). Sulcozoan topology in Fig. 9 was in smaller taxon samples such distortion may occur. exactly as in Cavalier-Smith et al. (2014) despite the addition of Overall we conclude with high confidence that centrohelids are numerous Amoebozoa and of the very long branch Trichozoa not sisters to haptophytes (clade Haptista is now firmly supported) included in the most recent multigene trees; slight variation in and that they do not group with Microheliella, so Heliozoa sensu the positions of Collodictyon and Malawimonas shows they are Yabuki et al. (2012) is polyphyletic, and must be reclassified. With not stable to taxon sampling and that more complete sequences somewhat lower, but reasonable, confidence we conclude that are needed for them. Corbihelia and Corbistoma are both clades. For the first time we Though there was maximal support across most of the tree for have shown clearly that Telonema is related to other Cryptista both CAT chains for the same topology as for ML, in the two regions and provided good evidence for a relationship with Picomonas near the base of podiates and at the base of corticates branches are and Microheliella as a Corbihelia subclade. so closely spaced that they correspond to differences of only 1–3% 342 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Fig. 9. PhyloBayes CAT-GTR-C tree for 164 eukaryotes excluding red algae using 50, 854 amino acid positions in 187 protein-coding genes. The alignment is identical to that used for Fig. 1 except for excluding all red algae. Support values are not shown for bipartitions with maximal support (1.0) which also had maximal bootstrap support (100%) on a separate RAxML PROTGAMMALGF tree for the same alignment (the majority); bipartitions not found in ML are shown by –; posterior probability support (left) and ML support (right 100 pseudoreplicates) are shown for all non-maximally supported bipartitions. The number of amino acids included for each taxon follows its name; the three taxa we sequenced are marked by arrows. The two CAT chains converged on the same topology except for the position of Stenamoeba which on one was as shown and on the other was sister to Filamoeba/Flamella (probably incorrect: see Cavalier-Smith et al., 2015); this contradictory position (red arrow) had maximal support making maxdiff 1 and three support values (asterisk and red) 0.5. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.) in sequence, and may correspond with time intervals of 10–20 mil- simultaneous explosive eukaryote radiation at the base of both lion years, making unambiguous resolution unlikely given that corticates and scotokaryotes (podiates plus metamonads), in later changes must exceed them by more than 50-fold. These close marked contrast to the well spread out, easily resolved branches branches, especially clear on CAT trees (Figs. 2 and 9), imply a near throughout Eozoa. T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 343

4. Discussion nised similarities in cortical and cytopharyngeal ultrastructure between Picomonas and Telonema which are unique in eukaryotes 4.1. Polyphyletic origins of tiny naked axopodial protists now lead us to group them as a new zooflagellate superclass Cor- bistoma (see Section 4.3), so there is good ultrastructural support Our trees strongly confirm that Minimassisteria, superficially for that clade. Microheliella is more radically different as it lacks like Oxnerella except for retaining a weakly expressed biciliate cilia and has axopodia; however Heliomorpha is an ultrastructural phase (Cavalier-Smith and Chao, 2012; Howe et al., 2011), is a link between them as it has both cilia and axopodia. Microheliella, granofilosean cercozoan (Bass et al., 2009), whose slender Heliomorpha, Telonema, and Picomonas share a rare ultrastructural extrusome-bearing axopodia evolved separately from those of cen- character unknown in any other eukaryotes: the filogranular net- trohelid heliozoa and Microheliella. Centrohelid heliozoa and Micro- work that interconnects mitochondria, ER, and plasma membrane heliella, unlike Minimassisteria, lack cilia; our trees clearly show in Microheliella (Yabuki et al., 2012) and organelles generally, that both belong in subphylum Hacrobia, unlike Minimassisteria including cortical microtubules, in Telonema (Yabuki et al., 2013); that as a cercozoan is in Harosa. It was already evident that micro- and forms a filogranular sleeve around extremely long centrioles helids and centrohelids independently lost cilia (Yabuki et al., of Heliomorpha (Brugerolle and Mignot, 1984, under the invalid 2012), as Microheliella and the ultrastructurally related axopodial name Dimorpha). Extrusomes of Telonema subtilis and Heliomorpha flagellate Heliomorpha both have centrosome-nucleated axopodial are also extremely similar (Yabuki et al., 2013; Brugerolle and microtubules in trans-nuclear cytoplasmic channels (Brugerolle Mignot, 1984), very different from the scrolled ones of rollomonad and Mignot, 1984) so endohelean axopodia probably originated Cryptista, and distinctly different from other nearly isodiametric before microhelids lost cilia. extrusomes (loosely ‘kinetocysts’) from disparate groups (Bardele, All our ML and converged CAT trees now consistently show cen- 1972; Mylnikov and Mylnikov, 2008). Their central dense barrel trohelids as sister to haptophytes, in agreement with the earliest is distally squarely truncate (Heliomorpha) or gently rounded 18S rDNA distance trees (Cavalier-Smith and Chao, 2003b); this (Telonema), markedly different from the triangular pointed distal clade (Haptista) is strongly to maximally supported. The much portion of the partially bipartite centrohelid extrusomes (Bardele, stronger and now consistent support for clade Haptista than orig- 1975; Dürrschmidt and Patterson, 1987a). inally (Burki et al., 2009) probably stems from including a second The slenderer flattened extrusomes of Microheliella are very centrohelid and the greater sequence and taxon representation unlike those of T. subtilis; interestingly they resemble the flattened (e.g. addition of Palpitomonas) in our analyses. Robust grouping ’alveoli’ with dense contents in T. antarcticum; in comparing these of Oxnerella and Polyplacocystis is consistent with the idea that dense alveoli with cortical alveoli of alveolates Klaveness et al. Oxnerella became miniaturised and naked by losing the ancestral (2005) doubted their homology. Their irregular arrangement centrohelid silicified scales (Cavalier-Smith and Chao, 2012; (which differs from cell to cell; Klaveness personal communica- Cavalier-Smith and von der Heyden, 2007) and thus superficially tion) and morphological similarity to Microheliella’s extrusomes convergent in light microscope morphology with Minimassisteria. suggests that they may be extrusomes also. If so, the ancestor of Haptista clearly exclude Microheliella, which on ML trees always Corbihelia either had both types of extrusomes that were differen- groups instead with Telomema plus Picomonas with moderate to tially lost in different lineages or else an ancestral concentric type strong support and also consistently groups with these genera on was independently flattened in Microheliella and T. antarcticum, CAT trees that exclude Plantae; though this Corbihelia clade is itself becoming markedly smaller but unchanged in morphology during relatively weakly supported its grouping with other cryptists rather Microheliella cell miniaturisation. than Haptista is strongly supported in Fig. 1. The variable support When rare ultrastructural characters and sequence trees mutu- for Corbihelia simply reflects the very deep and close divergence ally reinforce their phylogenetic message it is usually correct. We of these three groups; it does not weaken the much stronger evi- therefore remove Endohelea from Heliozoa and group them with dence from the holophyly of Haptista that Microheliella not only lost Telonemea and Picomonas as new cryptist subphylum Corbihelia cilia but also evolved axopodia independently of centrohelids, as (Table 1) unified by sharing a filogranular network. A weak rela- Telonema and Picomonas both have two cilia and both lack axopo- tionship of Telonema with heterokonts on a 6-gene tree (Reeb dia. That refutes the idea of a common axopodial origin for Endohe- et al., 2009) was probably an undersampling artefact; grouping lea and Centrohelea (Yabuki et al., 2012), which was always with Harosa on a 258-gene tree (Burki et al., 2012a) may reflect dif- doubtful because of their marked ultrastructural differences. Poly- ferent taxon and gene samples or possibly tree distortion because phyletic origin of axopodia in unrelated protists of differing ultra- of the chimaeric nature of chromists (see Section 3.2). With our structure is a well established evolutionary phenomenon (Smith gene set when Microheliella, Picomonas, and Oxnerella were omitted and Patterson, 1986; Yabuki et al., 2012); this is yet another exam- using a much more limited corticate taxon sample, Telonema was ple, with strong sequence support for the first time. usually sister to the centrohelid Polyplacocystis and this clade to We conclude that three phylogenetically distant flagellate Haptophyta (Cavalier-Smith et al., 2014 Fig. 2), not to Harosa. ancestors of radically different ultrastructure (cercozoan, CAT 18S rDNA trees group Telonema weakly with Haptista and haptophyte-like, and Corbistoma-like) independently evolved min- Microheliella, not Cryptista, though not specifically with Microhe- ute, trophically similar, axopodial naked cells. liella (Cavalier-Smith and Chao, 2012). However, our chromist only trees (Fig. 4) clearly establish that Corbihelia are sisters not to Haptista but to Cryptomonada plus Leucocrypta and Palpitomonas, 4.2. New hacrobian subphylum Corbihelia the core cryptists. This robust cyptist affinity is consistent with earlier classification of Telonema in Cryptista (Cavalier-Smith, As Heliozoa (Endohelea, Centrohelea) is polyphyletic, its classes 2007 Fig. 3.6) because it has tubular ciliary hairs like Crypto- must be placed in separate phyla. When the distinctive ultrastruc- phyceae (later also discovered in the novel cryptist Palpitomonas ture of Microheliella was discovered, the only protist with clearly (Yabuki et al., 2010)) and abundant pellicle microtubules as in related ultrastructure was Heliomorpha, still not sequenced, so kathablepharids. All trees including Palpitomonas (Yabuki et al., both were grouped as Endohelea (Yabuki et al., 2012). Since then 2010) confirm it is a cryptist (Cavalier-Smith, 2013a; Yabuki Picomonas ultrastructure has been described, also seemingly et al., 2010, 2014); its deeper branching than Rollomonadia (Cryp- unique (Seenivasan et al., 2013), and Telonema ultrastructure stud- tomonada plus Leucocrypta) in Yabuki et al. (2014) and almost all ied in more detail (Yabuki et al., 2013). Several previously unrecog- of our trees fit its being in a separate subphylum (Cavalier-Smith 344 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Table 1 Revised classiﬁcation of kingdom Chromista Cavalier-Smith (1981) and its 9 phyla.

Subkingdom 1. Hacrobia Okamoto et al. ex Cavalier-Smith, 2010a,b Phylum 1. Haptista Cavalier-Smith phyl. n.a (originally infrakingdom) Subphylum 1. Haptophytinab Cavalier-Smith subphyl. n.a (compositionally identical to phylum Haptophyta Hibberd ex Cavalier-Smith, 1986 (ancestrally photosynthetic biciliates with microtubule-supported haptonema ‘ﬁshing rod’ containing extensions of cortical alveoli) Class 1. Coccolithophyceae Casper, 1972 ex Rothmaler, 1951 (=Prymnesiophyceae Hibberd em. Cavalier-Smith et al., 1996b)(e.g.Emiliania, Isochrysis, Prymnesium) Class 2. Pavlovophyceae Cavalier-Smith, 1986 (e.g. Diacronema, Pavlova) Class 3. Rappephyceae Cavalier-Smith cl. n. (Diagnosis: rappemonads: plastid-bearing sisters of Coccolithophyceae plus Pavlovophyceae) Subphylum 2. Heliozoaa Cavalier-Smith subphyl. n. Class Centrohelea Kühn, 1926 ex Cavalier-Smith, 1993 (e.g. Oxnerella, Polyplacocystis, Acanthocystis, Raphidiophrys) (= unranked Centroheliozoa: Dürrschmidt and Patterson, 1987a,b) Phylum 2. Cryptista Cavalier-Smith, 1989 em. (lack cortical alveoli)c Subphylum 1. Rollomonadia Cavalier-Smith (2013a) Superclass 1. Cryptomonada Cavalier-Smith, 2004 (as subphylum) stat. n. Class 1. Cryptophyceaeb Fritsch in West and Fritsch, 1927 (e.g. Guillardia) Class 2. Goniomonadea Cavalier-Smith, 1993 (Goniomonas) Superclass 2. Leucocrypta Cavalier-Smith supercl. n.a (Diagnosis: as for subphylum Leucocrypta and class Leucocryptea in Cavalier-Smith (2004) p. 77) Class Leucocryptea Cavalier-Smith, 2004 Order Kathablepharida Cavalier-Smith, 1993 (e.g. Kathablepharis, Roombia) Subphylum 2. Palpitia Cavalier-Smith in Cavalier-Smith and Chao, 2012 Class Palpitea Cavalier-Smith in Cavalier-Smith and Chao, 2012 (Palpitomonas) Subphylum 3. Corbihelia Cavalier-Smith subphyl. n. (pharyngeal baskets or centrosome-nucleated radiating axopodia) Superclass 1. Endohelia Cavalier-Smith supercl. n. Class Endohelea Cavalier-Smith in Yabuki et al., 2012 Order 1. Microhelida Cavalier-Smith in Yabuki et al., 2012 (Microheliella) Order 2. Heliomonadida Cavalier-Smith, 1993 em. 2012 (Heliomorpha) Superclass 2. Corbistoma Cavalier-Smith supercl. n. Class 1. Picomonadea Seenivasan et al., 2013 Order Picomonadida Seenivasan et al., 2013 (Picomonas) Class 2. Telonemea Cavalier-Smith, 1993 Order Telonemida Cavalier-Smith, 1993 (Telonema, Lateronema) Subkingdom 2. Harosa Cavalier-Smith, 2010a,b Infrakingdom 1. Halvaria Cavalier-Smith, 2010a,b Superphylum 1. Heterokonta Cavalier-Smith, 1981 (=stramenopiles: tripartite anterior ciliary tubular hairs) Phyla Ochrophytab, Pseudofungi, Bigyra Superphylum 2. Alveolata Cavalier-Smith, 1991 stat. n. 2013 (cortical alveoli) Phyla Miozoa (Subphyla Protalveolata and Myzozoab (infraphyla Dinozoab and Apicomplexab)) and Ciliophora Infrakingdom 2. Rhizaria Cavalier-Smith, 2002a,b (Phyla Cercozoa and Retaria)

a Merely changed in rank (name and circumscription unaltered). b Taxa certainly ancestrally photosynthetic. c It is entirely inappropriate to call this phylum Cryptophyta as only one of its seven classes (Cryptophyceae) consists of algae; all others are entirely heterotrophic, phagotrophic, and non-algal; the vernacular term cryptophyte best applies to Cryptophyceae only and cryptist to the whole phylum. The only Hacrobia that should be subject to The International Code of Nomenclature for algae, fungi and plants are Cryptophyceae and Haptophytina; all others should be subject to the International Code of Zoological Nomenclature as should all non-photosynthetic Harosa.

and Chao, 2012; Cavalier-Smith, 2013a). Telonema and Microheliella Plantae, placing it as sister to Telonema with variable support, were a weakly supported clade on Hsp90 trees (Yabuki et al., 2012; sometimes strong; CAT trees mostly agree when red algae are Cavalier-Smith and Chao, 2012) and the EFL tree, consistently with excluded, some with strong support. Telonema and Picomonas both the present multigene trees. have a pair of lateral cilia with divergent centrioles, not parallel as Diagnosis of new cryptist subphylum Corbihelia in Rollomonadia, but are superficially very different because Cavalier-Smith: biciliates and/or axopodial heterotrophs with Telonema is relatively large with an immensely complex cortical dense filogranular network interconnecting organelles; skeleton and feeds on eukaryotes (Klaveness et al., 2005) whereas extrusomes ancestrally kinetocysts with shorter peripheral ring Picomonas is tiny with simplified centriolar roots compared with surrounding central dense core that is distally flattened or other hacrobian flagellates (Seenivasan et al., 2013). Nonetheless rounded, not pointed as in centrohelids; tubular mitochondrial confidence in this grouping together being correct (and artefactual cristae. Anterior centriole with a single anterior directed micro- nature of the relatively few trees including red algae that do tubular root. Etymology: contraction of Corbistoma and Endohelia, not show it) is greatly increased by the unique ultrastructural its two superclass, which also indicates that members have a similarities between Telonema and Picomonas we have found. basket-like cytopharynx or else heliozoan-like axopodia. Diagnosis At least two Telonema species are known ultrastructurally, of new superclass Endohelia Cavalier-Smith is as for its sole class Telonema antarcticum with tripartite tubular ciliary hairs, large Endohelea (Yabuki et al., 2012, p. 377). flattened cortical alveoli containing dense material, but no extrusomes (Klaveness et al., 2005; Shalchian-Tabrizi et al., 4.3. New cryptist superclass Corbistoma 2006), and T. subtilis with complex extrusomes, but no ciliary hairs or dense-content cortical alveoli (Shalchian-Tabrizi et al., 2006; Picomonadea were originally thought a new algal phylum (‘pico Yabuki et al., 2013). Both have a unique cell cortex supported by biliphytes’) related to Cryptista (Not et al., 2007), but instead are two dissimilar complex microtubular and microfibrillar multilay- heterotrophs feeding by pinocytosis not phagocytosis, put in a ered structures, together constituting a subcortical lamina; in separate phylum Picozoa (Seenivasan et al., 2013) as previous trees T. subtilis its two major parts are called the A2-layer and P-layer failed to reveal their affinities and their ultrastructure seemed (Yabuki et al., 2013). Though Yabuki et al. (2013) considered these unique offering no clues as to their relatives. Our LG ML trees layers unique among eukaryotes, only their extension over the strongly exclude Picomonas from core Cryptista, Haptista, and whole surface is unique; their substructure resembles the two T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 345 posterior centriolar roots of excavate protozoa (Simpson, 2003). cell miniaturisation and probably only the singlet root lost. Unlike The A2-layer is like the I-fibre-associated right root and P-layer like Telonemea, its asymmetric open basket-like cytopharyngeal skele- the left root with a C fibre, so we regard them as laterally hugely ton is purely microfibrillar and has no microtubules. Parallel thick extended hypertrophied homologues of these structures, both pre- microfibrils are cross-linked by finer ones. sent in the ancestral chromist as Cavalier-Smith (2013b) explained We postulate that the Picomonas cytopharynx and its microfib- previously. For comparisons with excavate outgroups the centrio- rils are homologuous with the fibrillar components of cytopharyn- lar face of the cell must be regarded as ventral; A-2 is therefore geal fibrillar element Fe of T. antarcticum where the fibrillar ventral right, consistent with this inferred homology. Telonemea components are more strongly developed and the microtubules have a narrow projection (proboscis) beside the cilia supported less obvious than in the A2 of T. subtilis. We suggest that the pha- by the two distinct concentric scroll-like skeletal structures. The gotrophic common ancestor of Picomonas and Telonemea evolved a larger outermost strongly curved ventral right scroll is the main cytopharynx with both microtubule and microfibrillar supporting A2-layer microtubule sheet (A2) with its orthogonal closely linked skeleton and that microtubules were lost when a Picomonas ances- four-component microfibrillar sheet that wraps round the projec- tor abandoned phagotrophy and was miniaturised. Because of the tion leaving the central apical region membrane unsupported positionally equivalent and similar asymmetric open basket shape and often with membrane invaginations (Yabuki et al., 2013; of their cytopharynx, a second synapomorphy additional to the Klaveness et al., 2005). We regard the proboscis as the cytostome left/right bipartition in ultrastructure for this clade, we call it Cor- and cytopharynx and suggest phagocytosis is initiated there; the bistoma. In view of this unusual shared ultrastructure, never found inner dorsal left scroll (A1 Yabuki et al., 2014; Fe (fibrous element) in any other eukaryotes, and their grouping together sometimes Klaveness et al., 2005) consists of a similar microtubular sheet that strongly on both CAT and LG ML trees we group Telonemea and on one side overlaps the A2 microtubules but its far side is free of Picomonadea as superclass Corbistoma. the I-fibre-like microfibrillar layer of A2 which extends from A2 well past the zone of microtubule overlap. The inner scroll Diagnosis of new superclass Corbistoma Cavalier-Smith: het- (A1/Fe) is more obviously distinct in T. antarcticum where it forms erotrophic biciliates with tubular mitochondrial cristae; cytos- the main skeleton of the cytopharynx and is clearly not part of the tome and basket-like cytopharynx supported by microfibrils subcortical lamina and not intrinsically multilayered. The P-layer is (with or without microtubules) associated with right centriolar an even larger scroll-like multilayer (microtubules inside, Pf sheet microtubular root or a hypertrophied derivative of it; cell body outside) that underlies the majority of the cell; it abuts A2 contin- partitioned into smaller cytopharyngeal zone and larger zone uously along a clearly defined spiral suture line (microtubule containing the nucleus, mitochondria, microbodies by a line organising boundary – MTOB of Yabuki et al., 2014) that passes running just behind the posterior centriole, each zone having through the centriole junction and thus divides the proboscis from a different set of ultrastructural components, those of major the main cell body at a slanting angle. zone held together by dense granular/microfibrillar structures; Picomonas cells are also divided into two distinct regions by an left and right posterior microtubular roots simple, without C indistinct cleft starting just behind the posterior centriole and I fibres (Picomonadea) or with hypertrophied right root (Seenivasan et al., 2013). The smaller region contains a cytophar- fibre into scroll-like sheets forming the proboscis part of a ynx and the larger the nucleus, mitochondrion, and microbodies bipartite cortical multilayered skeleton (Telonemea). Etymol- held together by dense fibrillar connections (less prominent than ogy: corbis L. basket; stoma Gk mouth. in Telonema), a two-zone arrangement just like Telonemea but no Labelling the cytopharyngeal region posterior in Picomonas and other eukaryotes. We regard these fibrillar connections between anterior in Telonemea, and the nucleus-containing regions anterior the extremely tightly packed Picomonas organelles and the dense and posterior respectively (Yabuki et al., 2014; Seenivasan et al., granules that connect mitochondrial membrane evaginations to 2013), hid these inferred homologies and does not reflect the abso- the posterior vacuole (Seenivasan et al., 2013 Fig. 3) as likely com- lute orientation of the cells based on centriole pair symmetry that ponents of a simplified filogranular network related to the much we used to infer them. This unique cell partitioning into two more obvious structures of telonemids and Endohelea; but as Pico- regions by the MTOB/cleft, which we regard as homologous, is monas cytosol is almost squeezed out of existence in this highly miaturised cell so filogranular structures are much less extensive really a left right bipartition related to the long axis of the ancestral than in these groups, and micrographs are also somewhat fuzzier excavate groove. That is superficially concealed by the very differ- making them indistinct, such equivalence is more open to question ent swimming modes the two groups took up after their inferred than for the other corbihelid genera. Unlike Telonemea, Picomonas common ancestor evolved the cytopharynx instead of the ancestral has no subpellicular lamina, but four conventional centriolar roots: excavate feeding groove fuelled by groove-associated posterior cil- one anterior dorsal root of two microtubules (Ar), a right posterior iary currents. The Picomonas posterior cilium remained pointing root (Pr2) of six that connects to the cytostome anterior, left poste- backwards when the groove disappeared and the cell became more rior root of two (Pr1), and a third posterior directed root of two rounded, the anterior cilium lengthened and its oppositely point- emanating from near the anterior posterior centriole junction ing laterally inserted cilia evolved a unique ’jump, drag, skedaddle’ (Ar2). Compared with an excavate/Colponema-like chromist ances- swimming mode (Seenivasan et al., 2013). Telonemea removed the tor we infer that Pr2 and 1 are homologues respectively of excavate posterior cilium from the groove with two contrasting results: (a) right and left posterior roots, Ar-1 homologues of the exca- to lie parallel to the anterior (Telonema subtilis) but retain its beat vate/Colponema dorsal anterior root. The only root of uncertain pattern, which thus propels the cell with its original posterior homology is Ar2, which is either a Colponema-like singlet root with pointing forwards, and altered the beat of the former anterior cil- a second added microtubule or more likely a homologue of the ium to conform; or (b) to project laterally (T. antarcticum). In view hacrobian rhizostyle/haptonema (discussion in Section 4.6). of these marked differences and the fact that each represents a sep- Compared with such an ancestor the cytostome developed in the arate major clade of the huge array of uncharacterised telonemean feeding groove, the other three microtubular roots were simplified, lineages that are as genetically divergent as the two haptophyte and posterior ones lost characteristic excavate fibrillar appendages, classes (Bråte et al., 2010), likely to have diverged over a hundred e.g. I fibre, which Telonemea by contrast retained and hypertro- million years ago (judged from the degree of their 18S rDNA diver- phied. The kinetid is therefore not radically different from the gence compared with that of other eukaryotes with a known fossil ancestral chromist pattern; no roots were added during Picomonas record), we transfer T. antarcticum to a new genus: 346 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Lateronema Cavalier-Smith. Diagnosis: biciliate phagotrophs; one or both cilia with short (0.5 lm or less) acronemes, not long (3 lm) ones on both like Telonema (Shalchian-Tabrizi et al., 2006); mostly swim with one cilium anterior and shorter pointing sideways, seldom (and slowly) with both pointing backwards like Telonema; irregularly arranged flattened cortical cisternae containing dense material; isodiametric extrusomes absent. Type species Lateronema antarctica comb. nov. Cavalier-Smith. Basionym Telonema antarcticum Thomsen, in Klaveness et al. (2005), p. 2603; we consider Lateronema femi- nine. Etymology: latero – L. side; nema Gk thread, because Fig. 11. Electron micrographs of tripartite tubular ciliary hairs still attached to the one cilium (thread) points sideways and none backwards dur- anterior cilium of Lateronema sp. In contrast to Fig. 10 the globular base of the hairs ing fast swimming. is partially obscured by the membrane through which they pass to join an outer ciliary microtubule doublet, but the funnel-like proboscis is shown more clearly Yabuki et al. (2013) doubted whether T. antarcticum has ciliary than in Fig. 10 beside the smooth posterior cilium. Sampled from the Weddell Sea, hairs because Fig. 3f of Klaveness et al. (2005) was of an uncultured Antarctic by Helge A. Thomsen who took the unpublished micrographs and gave permission for inclusion here; provided by Dag Klaveness. cell and their scanning electron micrograph (SEM) did not report hairs from the Oslofjiord culture; however that is unsurprising, as they did not make show cast transmission (T) EM preparations Fig. 12 illustrates Lateronema internal cell structure; its filogranular to look for them (Klaveness, personal communication) which are network in the ciliary region is closely similar to that of the endo- more reliable for that purpose. SEM preparation by contrast tends helean Heliomorpha as well as to Microheliella; more so than are the to collapse tubular hairs making them hard to see even if present partially distinct but in places connected adhesive fibres that are (e.g. Ectocarpus sperm Fu et al. (2014) Fig. 1C); Fig. 4B of more elaborate and distinctive for telonemeans. Shalchian-Tabrizi et al. (2006) shows that the less acronematic cilium that points forwards (therefore anterior cilium) has collapsed 4.4. Holophyly and evolution of Haptista structures very like those of the Ectocarpus anterior cilium (certainly having typical heterokont tubular hairs despite SEM not well Haptista grouped Haptophyta and centrohelid Heliozoa together displaying them), which we interpret as L. antarcticum collapsed as an infrakingdom (Cavalier-Smith, 2003b) based on the fact that tubular hairs; they are absent from the lateral smooth cilium. By both feed by microtubule-supported filiform appendages (the sin- SEM both T. subtilis cilia are smooth with 3 lm acronemes gle haptonema or multiple axopodia), and both often cover their (Shalchian-Tabrizi et al., 2006). We suspect that the hair-like struc- cell body with complex mineralised scales (siliceous in centrohe- ture in Fig. 2 k of Yabuki et al. (2013) is not a T. subtilis tubular hair, lids, usually calcareous but sometimes siliceous in haptophytes: but a simple hair or flaw in the section. Fig. 10 shows tripartite Cavalier-Smith and von der Heyden, 2007; Yoshida et al., 2006), tubular hairs in TEMs of an Antarctic telonemid which is probably as well as their grouping on some rRNA trees (Cavalier-Smith and an undescribed Lateronema species, as the hirsute cilium is several Chao, 2003b). Early multigene trees did not support that grouping, micrometres longer than the smooth one; Figs. 11 and 12A also as Telonema and the one centrohelid were positionally unstable prove that some telonemids have tubular hairs on the anterior cil- (Burki et al., 2009): the centrohelid grouped either with Telonema ium. As in heterokont chromists these are clearly tripartite, but (CAT) or haptophytes (ML; Telonema was with cryptists). With a their bases are globular as in the cryptist Palpitomonas (Yabuki substantially different gene set both methods made Telonema sister et al., 2010) not carrot-shaped as in most heterokonts (Harosa). to Harosa and the centrohelid sister to that clade plus haptophytes; no haptist clade was evident (Burki et al., 2012a,b); Picomonas was more distant, intruding into Plantae as sister to Glaucophyta. Addi- tion now of another centrohelid Oxnerella, and Microheliella, a distant relative of Telonema and Picomonas, reliably shows a robust haptist clade with strong support by both methods with almost all taxon samples, and thus strongly supports the holophyly of Hap- tista. Cultures and transcriptomes and ultrastructure are also vital for rappemonads, with plastid genes related to haptophytes (Kim et al., 2011). We do not know if these uncultured marine phototrophs are flagellates with a haptonema like haptophytes, more like centrohelids, or of different phenotype from either; but as the chloroplast 16S rDNA tree shows they diverged from classical haptophytes well before the split between their two classes rappemonads must be a fourth haptist class (Table 1). It is important to establish whether this split preceded or followed the now well established split from centrohelids; if it was earlier it would prove that centrohelids lost photosynthesis and probably also plastids (as nobody has made a molecular search for centrohelid plastid rRNA, one cannot strictly eliminate the unlikely possibility that they still have leucoplasts; none has ever been observed ultrastructurally but tiny reduced ones like apicoplasts might have been overlooked). Fig. 10. Elecrton micrographs of tripartite tubular ciliary hairs of Lateronema sp. In The plastid 16S rDNA tree of Kim et al. (2011) shows that rappe- this metal-shadowed cell from a whole mount the hairs fell off the anterior cilium monads separated from the Pavlovophyceae/Coccolithophyceae during drying so show the globular base clearly as well as terminal filaments; the clade much earlier than the last common ancestor of Crypto- characteristic proboscis identifies the cell as a telonemid. Sampled from the Weddell Sea, Antarctic by Helge A. Thomsen who took the unpublished micro- phyceae. This greater antiquity of plastid-bearing haptists makes graphs reproduced by his permission – provided by Dag Klaveness. it impossible that they acquired their chloroplasts from a crown T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 347

Fig. 12. Electron micrographs of ciliary hairs and cortex of sectioned Lateronema sp. Sampled from the Weddell Sea, Antarctic by Helge A. Thomsen who fixed the cells in glutaraldehyde and osmium tetroxide and stained sections by standard protocols, and provided the unpublished micrographs and gave publication permission. (A) Section through anterior cilium (AC) and its centriole showing endohelid-like filogranular material (F) and adhesive fibres (AF) connecting organelles and three hairs on the cilium (arrows). (B) Section showing the junction (J) between A- and P-layers and a naked region of the cell surface (arrow) near the cytostome where the two layers may be able to move apart to ingest large eukaryotic algal prey Klaveness (personal communication) observed in Telonema. PC = posterior cilium; J = junction between cortical A- and P-layers; G = Golgi Apparatus; M = mitochondrion with tubular cristae; N = nucleus. 348 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 cryptophyte as assumed by some scenarios that deny a photosyn- Corbihelia as sister to other cryptists; the ML tree is identical. This thetic common ancestry for chromists and postulate plastid strongly supports holophyly of Cryptista sensu Cavalier-Smith transfer by tertiary symbiogenesis instead (Baurain et al., 2010; (2007) if we also include Microheliella, which could not be classified Cavalier-Smith et al., 1994); the evidence against the tertiary trans- satisfactorily as its ultrastructure was then unknown, and Picomo- fer interpretation is discussed in detail in Section 4.6. From scale nas, then neither discovered nor named. Holophyly of thus fossils Prymnesiophyceae are at least 225 My old (Cavalier-Smith, expanded Cryptista (Table 1) and its basal branching order need fur- 2006a); the ancestrally photosynthetic rappemonad/haptophyte ther critical testing by sequencing transcriptomes for extra species lineage must go back well over 400 My. When first separated from to break up the long branches of Microheliella, Telonema, and Picomo- heterokont chrysophytes, haptophytes were considered one class nas, which should stabilise trees further (Roure et al., 2013), and for by phycologists or an order by protozoologists (Hibberd, 1971; a rappemonad as a deep-branching outgroup, as would obtaining Green and Jordan, 2002); when formally made a phylum completer transcriptomes or genome sequences. Environmental (Cavalier-Smith, 1986; name introduced but not validly published DNA sequences show that telonemids and picomonads are deep by Hibberd, 1972) they were so highly ranked just because their clades with numerous lineages available for culturing (Bråte et al., closest relatives were unknown. Now we know they are related to 2010; Not et al., 2007; Seenivasan et al., 2013; Shalchian-Tabrizi rappemonads and centrohelids it is best to reduce Haptista in rank et al., 2007; Yoon et al., 2011); there is at least another undescribed to phylum and group haptophytes with rappemonads as a plastid- microhelid (Yabuki et al., 2012), but more important is to obtain iferous subphylum (Table 1): Heliomorpha to break the endohelean branch more deeply. The presence of indistinguishable tripartite ciliary tubular hairs Diagnosis of new phylum Haptista Cavalier-Smith (under in the cryptists Palpitomonas and Lateronema (with globular base ICZN): ancestrally photosynthetic scale-bearing biciliates with and single terminal filament) argue that such hairs was present in isodiametric extrusomes with concentric internal structure, the common ancestor of all Cryptista, and lost independently by Tel- feeding by one or more filiform microtubule-supported projec- onema, Picomonas, Endohelea, Goniomonadea, and Leucocrypta. tions, whose bases are ancestrally associated with fibrillar Leucocrypta evolved scales to cover their cell and ciliary mem- material (centrohelid centrosome, haptonemal ’root’ in branes (Lee and Kugrens, 1991) that unavoidably entailed loss of haptophytes). the ancestral cryptist tubular ciliary hairs. Such loss is comprehen- Diagnosis of new subphylum Haptophytina Cavalier-Smith sible and evolutionarily easy if the mode of swimming and/or feed- (under ICN): Unicellular biciliate chromists lacking tubular cil- ing changes, as happened to them all. The so called bipartite hairs of iary hairs; typically with fucoxanthin-containing chloroplasts Cryptophyceae are not strictly bipartite as they have a very short surrounded by periplastid membrane within rough endoplas- slightly differentiated curved base (Cavalier-Smith, 1999) that our mic reticulum; haptonema or vestiges of it usually present. Phy- trees indicate must have evolved secondarily from the ancestral logenetically includes all plastid-bearing hacrobia more closely globular type. Multiple losses of tripartite hairs also occurred in het- related to Coccolithophyceae than to Cryptophyceae. erokonts (Cavalier-Smith and Scoble, 2013), as did secondary sim- Diagnosis of new subphylum Heliozoa Cavalier-Smith: plification of basal regions (independently in Proteromonas and non-ciliate phagoheterotrophs with axopodia supported by Oikomonas: Cavalier-Smith et al., 1995/6) so it is unsurprising that hexagonally arranged microtubules (usually linked by triangle both also occurred in cryptists when swimming modes changed. array microtubules in regular tessellation) nucleated by a single Centriolar roots are a major feature of the body plans of flagellate centrosome; axopodia bear concentric core extrusomes; central eukaryotes (Cavalier-Smith, 2000, 2013b; Moestrup, 2000; capsule absent, unlike Radiozoa. Sole class Centrohelea. Simpson, 2003) and their distribution is often congruent with sequence trees. The unusually complex roots of cryptophytes The haptophyte haptonema is nucleated on the right of the pos- (Roberts et al., 1981; Roberts, 1984), when it was wrongly assumed terior centriole and in some (e.g. Diacronema; Green and Hibberd, that the ancestral condition for eukaryotes was two roots per centri- 1977) is basally attached to a large fibrillar structure, the ’haptone- ole (Moestrup, 2000) were hard to interpret, but it became easier mal root’. We postulate that haptonemal roots are homologous once we recognised that Chromista and Plantae had an excavate with the centrohelid centrosome, present in the last common ancestor (Cavalier-Smith and Karpov, 2012; Cavalier-Smith, ancestor of Haptista, and were reduced in derived haptophytes 2013b) with one anterior and three posterior roots. We can now also with reduced haptonema no longer used for feeding (e.g. Pavlova: see similarities between corbihelian and cryptomonad roots that Green, 1973, 1980). Possibly some basal linkers between haptone- support their inclusion in one phylum. There is only a single mal microtubules are related to those of centrohelid axonemes. four-microtubule anteriorly directed root (R3) in Cryptomonas, However, the haptonema differs from axopodia in lacking extru- clearly the ancestral condition for Hacrobia, independently lost in somes and in having a smooth longitudinal membrane-bounded Pavlovophyceae and Palpitomonas, and supplemented by a second tubule continuous with one of the smooth cortical alveoli underly- derived root only in Coccolithophyceae. The cryptomonad right ing most of the haptophyte cell surface (Parke et al. 1971). posterior root (R2) is clearly homologous with the right posterior root of Heliomorpha; in cryptomonads it is associated with a taper- 4.5. Holophyly and expansion of phylum Cryptista ing cross-striated root not found in excavates and therefore derived; Heliomorpha has an ultrastructurally indistinguishable striated root Inclusion of Telonemea within Cryptista because of the shared that is associated with its posterior right root and the axopodium presence of tubular ciliary hairs (Cavalier-Smith, 2007) was called lying parallel to it (Brugerolle and Mignot, 1984). We suggest that into question by the instability of the first transcriptome-based a similar striated root evolved in the ancestral cryptist and that a trees (using 127 genes, 29,235 amino acids) (Burki et al., 2009); cross-striated structure in the Lateronema cytopharynx (Fig. 3c of although ML placed Telonema as sister to Cryptophyceae as pre- Klaveness et al., 2005) might be derived from it. Cryptophyte root dicted, CAT contradictorily grouped it with the sole centrohelid. Cr of two microtubules arises positionally like the excavate ventral Moreover, neither Telonema nor Picomonas grouped with other singlet root between the two centrioles and is just a singlet at its cryptists using a different set of proteins (50,293 amino acids) extreme base; contrary to Moestrup (2000) it is not R4, for which (Burki et al., 2012a). Our CAT analysis for chromists only (using there are no genuine homologues in Hacrobia. The left posterior root 50,854 amino acids) groups Telonema with Picomonas, this with four microtubules in Cryptomonas (4r of Roberts, 1984)is corbistome clade with Microheliella, and with 0.99 support shows equivalent to R1 of other corticates and excavates. T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 349

Thus all five excavate roots persisted in Cryptophyceae but no I transformed into other structures by any cryptists that lack it fibre remained and the right root is unsplit: the three posterior (e.g. Palpitomonas). Hacrobia is strongly supported by both roots all rotated forwards with the ventral centriole after the methods as a clade that is distinct from both Harosa (Fig. 2) and ancestral feeding groove was lost. The cryptophyte rhizostyle is a Plantae (Fig. 7) when non-corticate out groups are excluded. bundle of two slightly curved parallel microtubule rows (six micro- Eukaryote-wide trees sometimes show Hacrobia as a weekly sup- tubules per arc) situated immediately to the right of the kinetid ported clade (Fig. 1) and sometimes as weakly paraphyletic (e.g. exactly like the haptonema base (typically four to six microtubules Figs. 2, 9), but none showed Cryptista as branching more deeply per arc) and is present also in Goniomonas (Kim and Archibald, than all other chromists. These findings collectively for the first 2013), the sister of Cryptophyceae, and is thus a synapomorphy time definitively disprove an old idea that Cryptista are the earliest for Cryptomonada. In Leucocrypta (kathablepharids), the sisters diverging chromists (Cavalier-Smith, 1986), and show that the of cryptomonads, we now for the first time interpret the micro- chromist tree is congruent with the idea that haptophyte and cryp- tubular skeleton of their previously seemingly unique cytopharynx tist chloroplasts decended vertically without tertiary transfers from (Clay and Kugrens, 1999 Fig. 19) as having evolved by multiplying a common photosynthetic hacrobian ancestor that acquired a bac- the longitudinal rhizostyle many times to form a ring around the terial gene by LGT (Rice and Palmer, 2006). This strongly confirms cytostome: each of over a dozen rhizostyle-like units composing the logical basis for establishing Hacrobia (Okamoto et al., 2009; the ring consists of two arcs (outer with 10, inner with six micro- Cavalier-Smith, 2010a) and adds the conclusion that centrohelids tubules in Kathablepharis). Thus it is beyond reasonable doubt that and Corbihelia also belong in Hacrobia, previously ultrastructurally cryptist subphylum Rollomonadia ancestrally had a cytoskeletal clear for telonemids (Cavalier-Smith, 2007) but suggested weakly at structure comprising two short nested microtubule arcs. best, and inconsistently, for the other groups by earlier contradic- Are the two microtubule arcs related to endohelian axopodia tory trees (e.g. Burki et al., 2009, 2012a; Cavalier-Smith and Chao, and/or the corbistome pharyngeal skeleton? The position of the 2012; Zhao et al., 2013). Interestingly with increased taxon and Heliomorpha centrosome to the right of the kinetid like the rhi- gene sampling CAT consistently places centrohelids with hapto- zostyle/haptonema suggests that its square array axopodial axo- phytes and usually also Telonema with cryptophytes as originally neme could have evolved relatively simply from an ancestral found by ML (Burki et al., 2009) and contradicts their CAT trees, double arc. We suggest it did and that the Microheliella triangular but agrees with earlier classification (Cavalier-Smith, 2003b, axopodial axoneme did so also; divergent evolution of square 2007); this cautions against assuming that CAT trees are always and triangular array axopodia from a double arc ancestor is more superior to LG ML trees and shows that with too few related taxa likely than direct conversion of a fully formed square or triangular they can sometimes be less accurate. Sparsely sampled trees should pattern into the other. We therefore predict that Heliomorpha gene not be used prematurely to reject classifications based on ultra- sequences will show Heliomorpha as sister to microhelids, diverg- structure that have often proved more reliable than early overcon- ing soon after corbistomes and Endohelea split. fidently interpreted sequence trees. Sections 4.8–9 discuss the Section 4.3 noted that it is unclear whether Picomonas Ar2 root reasons for contradictory results for trees that include Plantae; here evolved by simplifying the rhizostyle/haptonema double arc, or by we outline the evolutionary significance of hacrobian holophyly. adding a second microtubule to the excavate ventral singlet as pos- The uniquely hacrobian double microtubule arc skeleton in tulated for cryptomonads. Both its position and the pervasiveness conjunction with our improved nuclear multigene trees and the of simplification during picomonad miniaturisation favour an ori- shared bacteria-to-chloroplast LGT (Rice and Palmer, 2006) gin from the double arc. If correct, then there is no evidence for a provides three entirely independent individually compelling and singlet root in Corbihelia, so it may have been lost after they collectively incontrovertible lines of evidence for the holophyly diverged from core Cryptista. However if Ar2 were a modified sin- of Hacrobia. Importantly, two of them relate to non-chloroplast glet root this addition of a microtubule might be a cryptist characters, thus decisively refuting the idea of an independent synapomorphy. history of hacrobian plastids compared with the rest of the cell Heliomorpha has a single anterior microtubular root of about six by tertiary lateral chloroplast transfer (Baurain et al., 2010). microtubules and single right posterior root of about nine, both We also suggest that centrohelid axonemes evolved from the associated with microfibrillar material (Brugerolle and Mignot, same haptonema-like ancestor by rearranging their basal connec- 1984; their reconstruction shows the posterior centriole docked tors to form a hexagonal/triangular lattice. Thus all three different into the base of the anterior one, probably the ancestral condition axonemal patterns of hacrobian axopodiate chromists may have for excavates and their least altered descendants, which is consis- evolved from the same double arc microtubule ancestor. tent with most of their micrograph labelling, but their Fig. 12 The position of the Heliomorpha centrosome (Brugerolle and seems to label centrioles the other way round, possibly in error). Mignot, 1984) and the centrosome-like haptonema ’root’ to the The root beside label Fa in Fig. 3d of Klaveness et al. (2005) may right of the posterior centriole is similar. Could it be that if centro- be the homologous anterior root of Telonemea (see also our helid centrosomes and the haptonema root are homologous, then Fig. 12). Though one Heliomorpha axopodial axoneme runs parallel both also are homologous with endohelean centrosomes, in which to the right side of the posterior centriole, it seems that the major case the idea that the endohelean and centrohelean centrosomes fibrillar attachment of the centriole is to the base of the anterior are homologous would be correct? That question is hard to answer centriole and root. By contrast the haptonemal root is associated now, partly because the root pattern of Telonemea is not thor- with the posterior centriole and its right posterior root, suggesting oughly studied and there is some ambiguity in interpretation of that these structures may not be homologous, which is consistent Heliomorpha roots (and we lack sequence evidence to corroborate with our decisive multigene tree evidence for independent origin its ultrastructurally-based taxonomic position). of centrosome-attached axopodia in Microheliella and centrohelids. Thus Hacrobia are united by three extremely different characters: chlorophyll c containing chloroplasts located inside a 4.6. Monophyly and evolution of Hacrobia periplastid membrane inside the nuclear envelope lumen with a derived protein-targeting machinery shared by Harosa (Myzozoa, We argue that the double arc microtubule bundle shared by hap- Ochrophyta: Cavalier-Smith, 2013a); chloroplasts with a unique tophytes, cryptomonads, and Leucocrypta evolved in the last com- shared lateral gene LGT from bacteria (Rice and Palmer, 2006); mon ancestor of Hacrobia and is another important previously and the double arc microtubular root, a previously unrecognised unrecognised ultrastructural synapomorphy for Hacrobia, lost or ultrastructural synapomorphy. 350 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Though it is theoretically possible that photosynthetic haptists be in Heliomorpha). If Ar2 and the axopodial axonemes are all rhi- got their plastids by tertiary transfer from a stem cryptist zostyle/haptonemal homologues, then Palpitomonas would be the (Cavalier-Smith et al., 1994) there is no reason from tree topology only hacrobian that completely lost this structure. to postulate such a complex event. All methods strongly support If, contrary to that interpretation, homology of endohelean, the holophyly of hacrobian, harosan, and chromist plastids in centrohelid, and haptonemal centrosomes were to be established 38-chloroplast-gene trees (Janouškovec et al., 2010); their chro- by proteomic studies of them all, this would show instead that mist topology is exactly as in Fig. 3, so there is no general incongru- Hacrobia ancestrally shared this fundamental feature of their cell ence between nuclear and chloroplast gene trees for chromists. The body plan and provide molecular tools for seeking vestiges of such common presence of a gene in haptophyte and cryptophyte plastid a structure in other cryptists which do not obviously have one – DNA that was laterally transferred from a bacterium (Rice and though the centriole region ultrastructure of Telonemea is so Palmer, 2006) is far better explained by lateral transfer into a pho- poorly known that one might be found by serial section ultrastruc- tosynthetic common ancestor of all hacrobia and by five losses of ture. If that were to be established, a common ancestral plastids early in their evolution than by a purely hypothetical centrosome might have been a predisposing factor for the indepen- and entirely unnecessary tertiary transfer for which there is not a dent addition of axonemes with non-homologous microtubule sliver of direct evidence (unlike the sole demonstrated tertiary arrangements. transfer of a haptophyte plastid into one small dinoflagellate lin- Basal branching within Hacrobia is ancient, seemingly eage, really a mechanistically easier, plastid replacement; see much older than Animalia (550 Mya) or crown Rhodophyta Cavalier-Smith, 2013a). Independent plastid losses by centrohelids, (>600 Mya); the base of Hacrobia on Fig. 1 is about three times dee- Corbihelia, goniomonads, Leucocryptea, and Palpitomonas would per than Coccolithophyceae (225 Mya from scale fossils: have been mechanistically easy if they occurred before the host Cavalier-Smith, 2006a). Of course if Hacrobia actually were ances- became dependent on non-photosynthetic plastid functions and trally photosynthetic with a plastid of red algal origin they cannot no disadvantage if they evolved alternative ways of heterotrophic actually be older than red algae; in our view the long bare stems at feeding, as they undoubtedly did. Tertiary symbiogenesis placing the base of Rhodophyta and Viridiplantae in Figs. 1 and 2 are tem- a cryptophyte chloroplast in a heterotrophic host would be ultra- porally misleading and represent transient grossly elevated evolu- structurally extremely complicated, as it would have to simultane- tionary rates in protein sequence not a long elapsed time, exactly ously transfer genes from the symbiont nucleus and nucleomorph as we first postulated for similarly grossly exaggerated bare stems to the host and radically reconstruct membrane topology; no in rDNA (Cavalier-Smith et al., 1996a; later for protein trees also, example is known. The only known tertiary symbiogenesis Cavalier-Smith, 2002a); the necessity of locating the fast and slow involved a haptophyte symbiont that did not retain or construct evolving regions on trees for proper interpretations of chloroplast chromist-like membrane topology around the acquired chloroplast origins was emphasised even earlier (Cavalier-Smith, 1980, pp. (Bergholtz et al., 2005), proving that just acquiring foreign genes by 896–899). If both stems were shortened almost to zero this would tertiary symbiogenesis is not enough to make ultrastructure mimic make these two long branches almost equal to that of the third that shared by Haptophytina, Cryptista, and Ochrophyta, which the plant group Glaucophyta, whose branch length is more concordant tertiary transfer interpretation of the identity of membrane topol- with that of other relatively short-branch eukaryotes, which would ogy in Hacrobia and harosan Ochrophyta would require for any then put the base of crown Rhodophyta marginally deeper than validity. that of Hacrobia in Fig. 1 and thus eliminate this apparent temporal As no Hacrobia clearly display a split right root as do excavates, contradiction. It is likely that such transient hyperacceleration of the simplest explanation for it being seemingly unsplit is that in stem lineage substitution rates, quickly reversed by a subsequent the ancestral hacrobian the nucleating centre for the outer branch slow down to more typical rates, also accounts for the even longer became completely split from the inner part when the groove was bare stems in Fig. 1 at the base of Retaria, Radiozoa, Foraminifera, lost and underwent lateral duplication to form the rhizostyle/ Diplomonadida, and Parabasalia. In the case of Radiozoa and Fora- haptonema double arc ancestor and allowing it to be oriented dif- minifera with an excellent fossil record we can be quite certain ferently from the inner branch (e.g. remain pointing backwards in that transient hyperacceleration is the cause of their ultra-long Cryptomonada when the other posterior roots and centrioles reori- branches rather than uniform long-persisting more moderate ented forwards, but to project radially out of the cell in Haptista). acceleration, as ’correcting’ the latter alone would not place them The fibre connecting the Heliomorpha centrosome to the right ante- early enough compared with the fossil dates for the first animals rior root is ultrastructurally remarkably like an I fibre (Fig. 11 of and plants. This is most obvious in Fig. 2 where the base of the Brugerolle and Mignot, 1984). Though positionally wrong as it is >500 My old Foraminifera wrongly appears to be millions of years always on the right posterior root in excavates (Simpson, 2003), in the future if branch lengths were assumed accurately to reflect the right anterior root of cryptomonads becomes the right poste- time! One must take into account such transient hyperacceleration rior root every cell cycle (Perasso et al., 1992), so it would require in a minority of lineages when mapping sequence trees onto the only earlier timing of its assembly to put it on the anterior root fossil record that gives the only objective indication of elapsed instead, so it might (like the hypertrophied I fibre of Telonemea) time. have been inherited from excavates, in accordance with our Hacrobia have probably been a distinct chromist branch for argument that this structure remained ancestrally in Corbihelia. >675 My, divergence between Corbihelia and Haptista being that A singlet root could have been overlooked in haptophytes by old. That fits a recent estimate that Chromista are >700 My old phycologists expecting to find only two posterior roots; it should (Cavalier-Smith, 2006a). Like Harosa and Cryptista, they ultimately be actively sought by serial sectioning in them and Telonemea – evolved from a Colponema-like ancestor with an excavate Fig. 4A of Green and Hibberd (1977) shows a microtubule exactly centriolar root pattern (one anterior and three posterior roots) between the Diacronema centrioles where the excavate singlet and cortical alveoli (Cavalier-Smith, 2013b; Cavalier-Smith and starts beside a row of 6 microtubules (left root R2: r1 in his figure). Karpov, 2012). Of the nine hacrobian classes containing flagellates, If this is the excavate singlet root (and if the slightly offset micro- six retained the single anterior-directed root but Pavlovophyceae tubule next to it does not belong with it but is the first one on the lost it, so has no anterior roots, and Coccolithophyceae evolved a right root), this would suggest that Hacrobia ancestrally retained a second anterior root that converges distally with the ancestral singlet root and that it may have been lost in Picomonas (or in one, uniquely in Coccolithophyceae and a synapomorphy of the Corbihelia generally if it is absent in Telonemea as it appears to class (Moestrup, 2000; Yoshida et al., 2006). T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 351

4.7. Holophyly of corticates two clades have only weak support using neomuran genes (e.g. Figs. 1–3 and S1–S8; and Brown et al., 2013; Burki et al., 2009, Corticates (i.e. kingdoms Plantae and Chromista) are a maxi- 2012a,b; Cavalier-Smith et al., 2014, 2015), but are each mally supported clade on all our eukaryote-wide CAT trees and holophyletic on the only multigene tree for genes of strongly supported on all recently published site-heterogeneous a-proteobacterial origin that included all their major subgroups multigene trees, but most published site-homogeneous (ML) trees (Zhao et al., 2013; 84 taxa, 42 genes, 11, 384 characters: 0.99 sup- using genes of neomuran origin show them as a clade that is rela- port for the chromist clade). For the reasons just explained for the tively weakly supported (Brown et al., 2013; Burki et al., 2012a,b; corticate clade, we argue that the holophyly shown by mitochon- Cavalier-Smith et al., 2014, 2015). (Neomuran origin genes are drial function genes is historically accurate and that the difficulty those inherited from the last common ancestor of neomura, i.e. of showing this by the neomuran origin genes is an artefact of eukaryotes plus archaebacteria, sometimes wrongly called archae- (a) the harosan long branches pulling them too deeply and by extra bacterial genes: Cavalier-Smith, 2014.) Multigene trees using divergence exacerbating (b) the poor resolution caused by near nuclear genes of eubacterial origin, however, show maximal or saturation of these genes. With Harosa too low in the tree near maximal support for corticate holophyly by both homoge- (Cavalier-Smith, 2009), we see an artefactual residual neous and heterogeneous algorithms with many fewer genes (He short-branch Hacrobia/Plantae clade. The fact that when chromists et al., 2014, 37 eubacterial genes: 1.0, 98%; Derelle et al., 2015, are represented on the tree only by the shortest branch harosan, 37 eubacterial genes 1.0 100%, 39 a-proteobacterial 1.0, 98%). We Pythium, it actually branches within Plantae (Fig. S6), not below attribute the higher corticate support using ML for nuclear genes Plantae/Hacrobia or even below Plantae/Hacrobia/podiates as in of eubacterial origin partly to this gene set being relatively less other eukaryote-wide trees, strongly supports the idea that the mutationally saturated (Derelle et al., 2015) than genes of neomu- long branches of the majority of Harosa wrongly pull all Harosa ran origin used in the majority of multigene trees (Nosenko et al., away from Plantae towards long-branch outgroups of corticates. 2013) including ours. This makes constructing correct topology The assumption that trees based on similar amounts of infor- intrinsically harder for the latter, so it is unsurprising that the less mation from nuclear, chloroplast, and mitochondrial genomes evolutionarily realistic LG ML performs less well than CAT. An 83– should resolve monophyly and branching order of Chromista 84 taxa 42-gene study of mitochondrial function genes from both equally well (Baurain et al., 2010) is wrong. The assertion that this nuclear and mitochondrial genomes (14 not in the Derelle et al. idea was predicted by the chromalveolate theory (Cavalier-Smith, a-proteobacterial set) clearly showed corticate holophyly by CAT 1999) is false and their claim to have disproved chromist mono- (0.99) but with LG ML Cryptista alone or a Hacrobia clade phyly fundamentally mistaken. The original paper arguing for (depending on inclusion or not of Collodictyon) grouped very monophyly of Chromista and of Plantae (Cavalier-Smith, 1982) weakly (31–41%) with Plantae instead of Harosa (Derelle and did not make that unreasonable prediction; in fact it argued that Lang, 2012). For this gene set CAT seems more accurate than LGF chromists originated by symbiogenetic enslavement of an early ML, possibly partly because many mitochondria-coded genes have eukaryote alga by a phagoheterotrophic biciliate host so early in only a weak phylogenetic signal (Derelle and Lang, 2012). chloroplast evolution that it could have happened almost immedi- A second problem with the neomuran set noted previously is ately following the origin of the common (then hypothetical) the relatively long branches for Harosa (Cavalier-Smith, 2009); protein-import machinery of Plantae. If that were true one would for all three harosan subgroups (alveolates, heterokonts, rhizari- not expect any sequence trees to be able to resolve their mono- ans) they are substantially longer than for Hacrobia and Plantae. phyly unless some drastic, never subsequently altered, changes A systematic error pulling Harosa too far down the tree by distant occurred during their earliest diversification; nonetheless it was outgroups is therefore much more likely than in the eubacterial suggested that cytochrome c trees might help test the ideas – origin gene set where Heterokonta and Rhizaria have branches nobody then was using full rDNA sequences for phylogeny. The fact similar in length to Hacrobia/Plantae, with only alveolates having that we can resolve chromist and plant monophyly on some rather long branches (but not relatively as discordant as for multigene trees and a consistent branching order for both king- neomuran genes). Comparison of Figs. 1–3 with those of doms shows that their divergence was not as rapid as it might have Cavalier-Smith et al. (2014, 2015) using an essentially similar gene been. Nonetheless evidence from all trees of a relatively rapid basal set strongly supports this interpretation. The longer-branch haro- radiation of all major corticate subgroups is incontrovertible; all san taxa included here to increase chromist taxon sampling were major changes in chloroplast and cell structure did take place in deliberately omitted from the 173-, 188-, and 192-gene trees of the very earliest phase of corticate evolution; not a uniformist Cavalier-Smith et al. (2014, 2015) (many of which included all gradual evolution spread though time, but a relatively sudden early 187 genes used here) specifically to reduce the severity of this quantum evolution of cell characters that changed little afterwards likely long-branch artefact. Not surprisingly therefore, in the was the rule (as in other megaevolutionary events: Cavalier-Smith, eukaryote-wide LG tree (Fig. 1) and in that omitting Rhodophyta 2002a,b, 2006a). (Fig. 2), Harosa branch more deeply than before, below the com- The idea that chromists originated very early in plant evolution mon ancestor of podiates and short-branch corticates making cor- might be thought to be contradicted by multigene chloroplast trees ticates wrongly appear polyphyletic (weakly 36%, 40%), whereas in rooted on cyanobacteria that do not show chromists emerging at the comparable trees excluding the longest-branch Harosa the very base of the plant radiation but within Rhodophyta as (Cavalier-Smith et al., 2014, 2015) corticates are holophyletic (also the second branch from their base (e.g. Shih et al., 2013 where weakly). By contrast in the corresponding CAT trees corticates are chromists emerge 5–8% above the base of crown Rhodophyta). always holophyletic with maximal support. This shows that the However such chloroplast and cyanobacterial trees are grossly long branches of Harosa in such trees do tend to introduce non-clock like, far more so that the nuclear gene trees analysed artefacts, as earlier argued and most strongly seen in ML trees. here but in quite similar ways qualitatively, and so must be very carefully and critically interpreted when trying to map them onto 4.8. Holophyly of Plantae and Chromista: contradictions between gene objective geological time. Chloroplast protein trees also have much sets longer red algal/chromist and viridiplant branches than glaucophytes (by 2–4 times) or most cyanobacteria, both red algae and Exactly the same problem is seen for Chromista and Plantae, green plants exhibit long bare stems that are probably caused by whose constituent lineages are often intermingled or if shown as greatly accelerated evolutionary rates in their stem lineages 352 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 exactly as explained above for both these lineages for nuclear others for accurately resolving a particular phylogenetic problem. genes. Thus it is likely that whatever population genetic bottleneck There is no biochemical or evolutionary reason to expect a nuclear or other evolutionary quirk caused these hyperaccelerations set of largely proteasomal or 80S ribosomal (or translation-related) affected both nuclear and chloroplast genes (though chloroplast genes (92 of 108 genes (91%) in Baurain et al. (2010); 104 of our genes, especially in green plants, show relatively less slowing present 187 genes; 56%, and therefore less functionally biased) to down after the acceleration – perhaps because being fewer the have the same resolution for every phylogenetic question as strength of stabilising/purifying selection is weaker than for the chloroplast genes. nuclear gene selection used here). Though it is tempting to relate Furthermore it is well known that mitochondrial and organelle hyperacceleration to the megaevolutionary changes associated genomes can evolve at very different average rates from nuclear with the origins of plastids, the trees of Shih et al. (2013) show that genes and that all three genomes can be subject to wildly idiosyn- a similar transient hyperacceleration yielding a hugely long bare cratic evolutionary changes that bear no relation to the deeply stem occurred in the stem lineage of cyanobacterial clade 1 (which unrealistic but common mathematical assumption of statistical includes the aberrantly pigmented Prochlorococcus and numerous uniformity across lineages (e.g. the bizarre minicircles of dinoflag- typical normal coccoid Synechococcus as well as the cyanobac- ellate plastid genomes (Zhang et al., 1999); bizarre events in evo- terium enslaved by the cercozoan testate amoeba Paulinella as a lution of euglenozoan and dinoflagellate mitochondria (Lukeš chromatophore). This makes it likely that hyperacceleratory pro- et al., 2009); radical differences in evolutionary mode of mitochon- duction of long bare stems for some clades is related to relatively drial DNA of green plants that led some to postulate mistakenly rare features of population genetic history that temporarily grossly two independent origins of their mitochondria (Gray, 1989); or unbalance or distort the predominant historical pattern of muta- the grossly accelerated evolution of microsporidial nuclear gen- tion–selection equilibria, but are not necessarily associated with omes that led Vossbrinck et al. (1987) to propose erroneously that megaevolutionary events of exceptional magnitude. Such accelera- they were the earliest diverging eukaryotes. The originator of the tions may often be associated with megaevolutionary events, but chromist/chromalveolate ideas (Cavalier-Smith, 1981, 1982, can also sometimes (possibly more rarely) arise in their apparent 1999) did not make the biochemically and evolutionarily unrea- absence, as exemplified by cyanobacterial clade C1. Our purpose sonable prediction that nuclear, chloroplast, and mitochondrial in emphasising this gross departure from more common and sub- genomes should resolve monophyly and branching order of Chro- stantially more clock-like (only in a very loose sense of ’clock’) mista equally. The fact that they do not was already long known; molecular evolution is to stress that the length of these bare stems this idea was thus falsified prior to Baurain et al. mistakenly linking on chloroplast trees (and that at the base of Plantae) probably it to the chromalveolate theory; their erroneous claim to have greatly exaggerates the time elapsed since chloroplasts originated. refuted the holophyly of chromists rested on the truth of that false If one shortened them substantially to allow for that this would assumption, even more than on the actual phylogeny of chromists. place chromist emergence very close to the base of both The fact that about 40 mitochondrial and nuclear genes with Rhodophyta and Plantae as a whole and therefore group the origins mitochondrial functions show chromist monophyly strongly with and diversification of both plastids (Plantae) and chromists so clo- CAT, whereas 100–200 mainly proteasomal/ribosomal genes do sely together in time (likely substantially less than 1% of the total not, has nothing to do with their separate genomic origin, but age of Plantae) that poor resolution of basal corticates on nuclear results from different evolutionary constraints on function that gene trees is entirely to be expected. Such relatively rare episodic makes mitochondrial proteins better molecular markers for basal events are probably more important in some cases of molecular corticate phylogeny as they are less substitutionally saturated in evolution than seems to be realised by those who suppose that this part of the tree. It would be absurd to consider this different unique evolutionary events can be adequately modeled by statisti- support level evidence for a tertiary lateral transfer of mitochon- cal averages. dria, as Baurain et al. argued for the similar case of chloroplasts How easy it is to resolve a major rapid radiation like that of on no greater justification. Moreover, though all chloroplast gene Chromista or Plantae or animals or scotokaryotes by gene sequence trees show chromist monophyly more strongly than neomuran trees depends on (a) how many close branches there are in a given nuclear genes (probably partly just because of many fewer distinct time interval; (b) how close they are relative to the evolutionary outgroups), chloroplast trees show the same conflict as nuclear rates of the specific genes being used – this influences how many genes between showing holophyletic Chromobiota (Yoon et al., mutations occur that might be inherited subsequently as ancestral 2002) versus holophyletic Hacrobia (Janouškovec et al., 2010). characters linking close relatives; and (c) how many of these Thus it is not the case that they are invariably better resolving than potentially molecular synapomorphies are destroyed by subse- nuclear neomuran genes. Oddly, Baurain et al. assert that their quent changes, which will depend on later evolutionary rates and totally unreliable method of discordant support levels offers a good the nature of subsequent changes that may differ considerably general test of phylogenetic incongruence, but overlook the fact among descendant lineages. Factors (b) and (c) are necessarily dif- that exactly the same incongruence exists for Plantae as for Chro- ferent between radically different gene sets and lineages. It has mista. Illogically they do not see this as evidence for the polyphyly been known since the early days of molecular phylogeny that no or paraphyly of Plantae, presumably because they believe that con- universal molecular clock exists for all genes, and the figures in clusion is incorrect despite many trees mixing up the three plant Shih et al. (2013) show how grossly unclocklike in the long term groups with chromists and support for plant monophyly being are all the chloroplast genes used for phylogeny – even they under- weak; by not doing so they applied inconsistent standards to the play that problem as they omit the orders of magnitude faster two kingdoms. evolving aberrant dinoflagellate plastid minicircle genes (Zhang Not only can equal resolvability for a given clade across gene et al., 1999). Even the most ardent believers in the flawed idea of sets not be justifiably assumed, neither can uniformity across lin- a molecular clock and neutral evolution never supposed that all eages and time. One well known example of non-uniformity over genes evolve similarly, and accepted that different genes evolved time is heterotachy where site-specific rates vary over time at rates that differ over orders of magnitude (Kimura, 1963). It (Philippe and Lopez, 2001; Lopez et al. 2003) which though some- was also long known that these radically different rates were times modeled was not allowed for in recent eukaryote-wide related to differences in function, as biochemists realised (King multigene trees discussed here including ours. Less widely appre- and Jukes, 1969). It has therefore been generally accepted by phy- ciated is that qualitative features of amino acid substitutions vary logeneticists that some gene sequences are more suitable than over time, a form of heterogeneity called heteropecilly (Roure and T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 353

Philippe, 2011). Both heterotachy and heteropecilly can in princi- independently (b) by our demonstration of the higher evolutionary ple either increase or decrease resolution in a radiation depending rate of Harosa pulling them artefactually away from Hacrobia. on their distribution in time and across lineages. The biggest The likely transient hyperaccelerations in substitution rate of increase in resolution for a clade would occur if the rate of substi- their stem lineages simply explains why it is so easy to resolve tutions dramatically increased and novel types occurred in its stem Viridiplantae and Rhodophyta as clades separately, and together but then decreased to zero so that the shared changes were never with the temporally more realistic very short stem (during which overridden by subsequent changes after the lineage radiated. Such shared novel changes in plastid protein targeting evolved: drastic non-uniformist events do occur in evolution, for example Cavalier-Smith, 2007), also explains why it is very hard to resolve when radically new paralogues like a- and b-tubulin diverged from their joint relationship (so that ML often contradicts CAT for the a common ancestor or EF1-a evolved in the ancestral eukaryote or latter only). Merely shortening the red and green stems to equal ancestral neomuran respectively. As explained previously these the small red plus green joint stem would make the shortest red hugely stretched stems of clades on sequence trees in ways that algal and green plant lineages no longer than the overall much led many clock-believers to misinterpret them in relation to time shorter glaucophyte branch, so would abolish most of the big (Cavalier-Smith, 2002a) and to misinterpret rooting by paralogue branch length difference between them and glaucophytes. On this trees (Cavalier-Smith, 2002a, 2006b); but such gross departures interpretation of the tree disproportions, the origin of crown from statistical regularity also allowed elongation factors to give Viridiplantae and Rhodophyta was much closer to the basal plant incontrovertible proof that neomura are a clade which cannot be radiation than their stem lengths suggest, consistent with the view established by normal enzymes that evolve more uniformly that the primary radiation of plants into three lineages and the through time (Cavalier-Smith, 2002a). Different genes certainly novel pigmentation and thylakoid stacking of Viridiplantae would do not evolve in the same way at all times and in all lineages; in have been extremely rapid consequence of the symbiogenetic ori- the case of protein synthesis elongation factors there is now exper- gin of chloroplasts allowing the colonisation of zones of different imental evidence (Cacan et al., 2013) for the heterotachy/quantum light regimes where contrasting accessory pigments would be evolutionary effects postulated during the functional divergence of adaptive (Cavalier-Smith, 1982, 2007, 2013a). their paralogues by Cavalier-Smith (2002a). Temporary highly ele- The mistaken claim to have invalidated chromist monophyly vated rates and altered patterns of substitution are not just associ- (Baurain et al., 2010) has unfortunately been uncritically much ated with gene duplication but can occur in a single molecule cited by others, even though their central argument from differen- strongly affected by concomitant changes in interacting molecules tial bootstrap support against a photosynthetic common ancestor as argued in detail for ribosomal RNA (Cavalier-Smith et al., 1996a; of chromists is fallacious, with flaws we therefore had to explain Cavalier-Smith, 2002a). In such cases it is easy to see why they above in detail. Petersen et al. (2014) repeated that erroneous diverge from the naive and often incorrect assumption of uniform claim and showed a diagrammatic scheme invoking three or four rates and patterns. When there are no identifiable direct functional tertiary symbiogeneses that is too imprecise to be a scientific shifts for a particular molecule displaying heterotachy it may be a hypothesis; it did not distinguish which of four supposed indepen- result of multiple interactions between different molecules shifting dent symbioses were assumed to be secondary and which tertiary in ways that collectively could be functionally equivalent (Philippe or specify specific donors or recipients for any tertiary transfers, et al., 2003). making it phylogenetically untestable, nor show any awareness To explain why multigene sequence trees can still resolve the of the cell biological difficulties of such transfers, whose postula- relationship between red algae and cryptomonad nucleomorphs tion is totally unnecessary. They did not realise that absence of (enslaved red algal nuclei: Douglas et al., 2001) despite substitu- the translocation proteins associated with the periplastid mem- tions being at least eight time faster in nucleomorphs than in red brane (Der 1, Cdc48) in dinoflagellates that they noted was pre- algae and the nucleomorph branch being 20 times longer than dicted by the thesis that dinoflagellates alone amongst chromists the stem uniting it with red algae (Baurain et al., 2010), one must lost the periplastid membrane (Cavalier-Smith, 1999, 2013a). Their suppose the occurrence of many irreversible changes in ribosomal Der1 tree well fits a single red algal secondary symbiogenesis and and proteasomal proteins specifically in this stem. A clock-like uni- no tertiary transfers (Cavalier-Smith, 1999), contrary to their form pattern with no frozen, effectively irreversible substitutions scheme, and is fully consistent with the chromalveolate theory in the stem would necessarily have entailed later reversals or dis- (Cavalier-Smith, 1999) that they wrongly stigmatise as ’defunct’. tracting overwriting of the shared changes during subsequent They omitted to mention that their Der 1 tree shows alveolates highly elevated evolutionary rates, and cannot be the correct evo- as sisters/basal to heterokonts on the periplastid paralogue branch lutionary model for the red algal stem. Very likely such drastic and so specifically refutes one of the three tertiary transfers postu- quasi-irreversible change occurred in the stem of Viridiplantae lated by Baurain et al. (2010), which would expect alveolate par- and Rhodophyta also, stretching these stems as argued above and alogues to be nested shallowly within heterokonts. The Cdc48 making the origin of the nucleomorph appear longer after their tree, though showing alveolates nesting deeply within hetero- divergence than it really was (Figs. 6 and 9). By contrast the com- konts, lacks bootstrap support and thus is compatible with their mon stem of Viridiplantae plus Rhodophyta is 11 times shorter being sisters; more importantly it rules out the idea that alveolate than the red algal stem and 9 times shorter than the viridiplant plastids were acquired substantially more recently than hetero- stem. We suggest that this shortness does not reflect a shorter time kont plastids as claimed by Baurain et al. (2010), and thus better interval, but merely the absence of the independently temporally fits the chromalveolate thesis that ancestral Halvaria had chromo- elevated aberrant evolutionary rate and pattern that must have phyte plastids and that they were lost by presently aplastidic sub- occurred in the red stem and probably did also in the green stem, groups. Both trees also clearly contradict the idea that haptophytes as already noted in Section 4.6. Baurain et al. (2010) asserted that if and Halvaria acquired their plastids substantially later than Cryp- trees can resolve the red algal/nucleomorph clade a similar num- tophyceae and the origin of Chromista as Baurain et al. (2010) pos- ber of amino acids should also resolve the chromist radiation tulated, so their idea of late serial tertiary transfers is now made equally strongly. That naive expectation is invalidated by (a) the defunct, and the chromalveolate theory vindicated, by direct clear evidence for heterogeneity in the basal radiation of Plantae phylogenetic studies of the key periplastid membrane associated that probably makes the red algal/nucleomorph clade radically chromist synapomorphies predicted by the chromalveolate theory. more easily resolvable compared with many clades radiating Periplastid translocation protein trees of Felsner et al. (2011) rein- around that time than a homogeneous model would expect, and force these conclusions; the Cdc48 tree unlike Petersen et al. 354 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

(2014) strongly supports chromist monophyly and shows no nest- Hacrobia, uninformatively labelling the clade ’’’orphans’’ clade 1’ – ing of alveolates within heterokonts or of any of the four chromo- inappropriate, misleading, populist jargon: parents of all clades phyte groups within each other as the Baurain et al. (2010) are dead, making all orphans. Compared with recent CAT trees scenario would predict; their split tree even groups the hacrobians (Brown et al., 2013; Cavalier-Smith et al., 2014), it probably (Emiliania, Guillardia); their Uba1 tree is less well resolved but wrongly shows Apusozoa as holophyletic not paraphyletic and its consistent. position of Metamonada is misleading (both errors usual for LG Many seem not to like the chromist idea merely because it ML) and very oddly has the same wrong topology for Euglenozoa involves many chloroplast losses and groups phototrophs and as did some early 18S rRNA trees; their exaggerated assertion that heterotrophs contrary to 19th century ideas of plant/animal dis- its internal phylogeny of major clades is ’generally concordant with tinctness abandoned half a century ago by more critical protistolo- previous phylogenies’ is wrong in other respects also, e.g. placing gists. Chloroplast losses occur much more frequently than their Acetabularia as the deepest branching green plant. Their Fig. 2 LG origin (once) or than secondary symbiogenesis (three times) or ter- ML tree using 207 proteins (names not stated and only 17,220 tiary symbiogenesis (once) (Cavalier-Smith, 1993, 2013a), making amino acids) and no rRNA is better in that Cryptista (including subjective antipathy to the assumption of multiple losses within Telonema) are holophyletic (44% support) but that and the chromists an evolutionarily unwarranted reason for rejecting chro- holophyly of Hacrobia also (39%) are less meaningful than our mist monophyly – repeatedly pointed out for decades; many still stronger results as Picomonas and centrohelids were omitted. Their seem in the thrall of Margulis (1970, 1981) who assumed 30 chloro- rather misleading 442-taxon Fig. 2 ML tree included prokaryotes, plast origins and no losses. Present data are compatible with which are so distant and hard to align for that many genes, holophyletic Chromista, as shown decisively on 42-protein trees wrongly rooted eukaryotes within trichozoan metamonads, and (Zhao et al., 2013); no detailed coherent alternative has been artefactually broke Amoebozoa into three clades. It had only 83% proposed (Keeling, 2013). support for Harosa (and the wrong basal harosan topology, thereby Another contradiction among gene sets exists between CAT contradicting with 72% support their Fig. 1 and our and other trees for nuclear genes of eubacterial or a-proteobacterial origin recent multigene trees) and Hacrobia was within Plantae (sister (Derelle et al., 2015) and mixed nuclear/mitochondrial genes of to Viridiplantae with 65% support), thus contradicting their a-proteobacterial origin (Zhao et al., 2013): the latter shows chro- Fig. 1. Support for their probably erroneous Plantae/Hacrobia clade mist monophyly (0.99), whereas the former put Cryptista sister to was negligible (35%). Neither their claim that their trees were Plantae (1, 0.95). This difference might be because Derelle et al. better sampled for protists than other studies, nor their title’s unlike Zhao et al. omitted key hacrobian and plant taxa (Hapto- erroneous assertion that their analyses ’resolved’ the eukaryote phyta, Glaucophyta) so the real closest relatives of the included tree, was justifiable. In fact, their analyses showed nothing not Plantae and Hacrobia were absent, or it might reflect artefactual already known about eukaryote phylogeny, perpetuated some attraction of Cryptista towards red algae in this gene set as in incorrect interpretations, and were less technically thorough than our 187-gene set. other studies. In particular neither of their self-contradictory trees Katz and Grant (2015) mistakenly claimed that ‘‘Chromalveo- provides evidence against chromist or hacrobian holophyly, which lata’’ (= Chromista sensu Cavalier-Smith, 2010a) was ’now falsified’ for no good reason they assumed to be wrong. by Parfrey et al. (2010). However, though Parfrey’s 16-gene tree (ML only) placed classical Hacrobia within Plantae rather than as 4.9. Holophyly of Plantae and Chromista: distortion by symbiogenetic sister to (strongly supported) Harosa and bizarrely placed centro- gene transfer? helids as sister to Archamoebea (wrongly shown outside Amoebo- zoa), no bootstrap support was shown for either position, and thus Our and all other multigene trees consistently imply rapid, it did not even resolve the monophyly of Amoebozoa, which is now near-simultaneous basal radiation of Plantae and Chromista, a sit- well established and much easier to show than for Hacrobia or uation where Bayesian trees can give spurious resolution (Yang, Chromista (Cavalier-Smith et al., 2015). Therefore the poorly 2007). Holophyly of Plantae is now generally agreed despite earlier resolved and systematically misleading oligogenic tree of Parfrey controversy (Keeling, 2013). Though Plantae are holophyletic on et al. (2010), wrongly claiming to be well-resolved, certainly did some recent multigene trees (Burki et al., 2007, 2008, 2009; Zhao not ’falsify’ the holophyly of Chromista as now circumscribed. et al., 2012), many others contradictorily intersperse plant and It is not possible to compare the 151-gene 232-taxon tree chromist lineages (Baurain et al., 2010; Brown et al., 2013; Burki (36,346 characters) of Katz and Grant (2015, Fig. 1) in detail with et al., 2012a) because of intrusion of one or more hacrobian chro- those discussed above, as the authors included rDNA as well as mist taxa into Plantae (Cryptista, Picomonas, Haptophyta). Our 150 proteins (unlike studies from other research groups) and failed eukaryote-wide trees confirm that any of these hacrobia plus one to specify which protein genes were used, so we do not know or both Corbihelia (Microheliella, Telonema) may intrude into Plan- whether they were of neomuran or eubacterial origin or a mixture, tae, but which do so depends strongly but erratically on taxon and and used only the evolutionarily less realistic, often misleading, LG gene sampling. Differently sampled corticate-only CAT trees show ML method. On that tree Hacrobia were holophyletic (no support separate Plantae and Chromista or interspersion. given) but had largely wrong internal topology (Roombia with We attribute these conflicts to two main factors: first the extre- Picomonas, wrongly called picobiliphyte, not cryptomonads; Cryp- mely close branching of all six basal corticate groups (Glaucophyta, tista seemingly paraphyletic; all with no support given). Hacrobia Rhodophyta, Viridiplantae, Harosa, Cryptista, Haptista) consistent were sister to Plantae (not within them, thus contradicting with explosive early algal radiation after chloroplasts originated, Parfrey et al., 2010 and their own Fig. 2) and this ’clade’ was sister as repeatedly argued: Cavalier-Smith (1982, 1986, 1999, 2007, to Harosa, but support values were not given for either; Plantae 2013a; and 1992 explained in detail why one should not expect and corticates were maximally supported. Their alignment also sequence trees to resolve chromist monophyly easily). Secondly omitted both centrohelids (their automated pipeline failed to find the likely presence in alignments of chromist protein paralogues the long available Polyplacocystis mRNA data, presumably because of unrecognised red algal origin, likely to pull red algae towards it was programmed to search only a protein database) and Palpito- chromists (Deschamps and Moreira, 2009) and conversely chro- monas, and thus was of limited value for checking holophyly of mist subgroups with unidentified red algal paralogues into Plantae Hacrobia sensu Cavalier-Smith (2010a); despite showing (Leigh et al., 2008). Which chromists are pulled into Plantae will hacrobian monophyly, their tree tendentiously avoided the name depend on two things: (a) which by chance replaced the most host T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 355 proteins by red algal paralogues when the redundancy necessarily be valuable further tests of the internal phylogeny of Chromista existing in their ancestral red algal/host chimaera was reduced by and Plantae. Other gene sets also need to be developed and studied independent differential gene loss as daughter lineages diverged; similarly, but it will not be adequate merely to add more genes and and (b) which genes with contradictory phylogenetic signals hap- ignore the serious technical problem posed by the fundamentally pen to be represented in different lineages in a given alignment. eukaryote–eukaryote chimaeric nature of Chromista. As the true signal from host genes is inherently weak because of Shih et al. (2013) by using BLAST to detect cyanobacteria- the explosive radiation of these six groups and trivially later close related genes found prima facie evidence that the hacrobian radiation of the five basal haptist lineages (i.e. 11 groups Emiliania may have >5000 genes of cyanobacterial origin and the near-simultaneously diverging), relatively few unrecognised red heterokont harosan Ectocarpus 2000 and two diatoms 1600 – algal paralogues could seriously distort corticate multigene trees. similar numbers to those found in glaucophytes and green algae. Corticate basal topology is therefore expected to vary greatly with They assumed that reflects transfer of thousands of red algal genes taxon and gene sample, as we and others observe. into chromists during their symbiogenetic origin. Unfortunately Given the shortness of so many included genes and explosive- their study did not include cryptomonads. Nor did they identify ness of corticate basal radiation single-gene trees lack any basal the genes involved, to see if there are a thousand or more shared corticate resolution, making it effectively impossible to detect all by all chromist algae (which would strengthen their conclusion) red algal paralogues persisting in chromist genomes using or largely distinct sets in each taxon (which would weaken it), or single-gene trees. We noticed that for some included genes cryp- run single-gene trees to check conclusions from BLAST – a rather tists or other hacrobia group with red algae, with no significant crude tool that could easily overestimate the numbers of trans- support, but too weakly to distinguish symbiogenetic replacement ferred genes. Assuming for now its partial validity, their results sug- of host by red algal genes from normal erratic groupings through gest that harosan genes of red algal origin are either fewer or more statistical noise; we found only about three cryptomonad genes divergent than hacrobian ones and less likely to be identified by for which red algal origins were so strongly supported that we their BLAST protocol. The long branches of Harosa discussed above removed them. Inevitably our alignments and those for all pub- make it likely that greater divergence is the main cause. This would lished eukaryote-wide multigene trees must include some also explain why they detected many fewer cyanobacteria-like unrecognised red-algal paralogues, wrongly assumed to be of host genes in the sole included red alga Cyanidioschyzon than in the glau- origin. The efficacy of the program Leigh et al. (2008) used to iden- cophyte Cyanophora and three green plants – its branch is much tify 15 putatively red-algal paralogues in a 22-species (7 corticates) longer on all multigene trees – and even why among the greens 60-gene alignment is untested on larger alignments (computation- they found more for the shorter-branch Physcomitrella and ally unfeasible for our far larger one), but it could probably not reli- Arabidopsis than longer-branch Micromonas. ably detect all. At present the best practical way to circumvent this Katz and Grant (2015) claimed that their automated pipeline problem and obtain reliable plant or chromist multigene trees is to identified 122 genes of cyanobacterial origin amongst the 1554 exclude all chromists from studies of Plantae (first done by conserved eukaryotic genes it selected; 12 such genes were Deschamps and Moreira, 2009) and all Plantae from Chromista, included in their Figs. 1 and 2 criticised above, despite their sug- first done here (Figs. 1 and 2, S1) – all other studies have totally gestion that such genes (not named) could be phylogenetically ignored this problem, a major reason why chromist phylogeny misleading. Removing them greatly reduced support for holophyly was unresolved and their monophyly doubted. of Plantae and caused glaucophytes to become sisters of Viridiplan- Fig. 6 of Petersen et al. (2014) illustrates an arguably clear tae (probably wrong: Deschamps and Moreira, 2009; example of retention of a red algal paralogue of Cdc48: the ances- Cavalier-Smith, 2007), but had only trivial effects on the rest of tral chromist paralogue recruited for protein transfer across the the tree. It is unlikely that their method correctly identified periplastid membrane strongly (99%) groups with red algae, sug- cyanobacterial origin genes, as massive numbers of single-gene gesting that there was no gene duplication of a host Cdc48 but that trees generated and screened by pipelines can be dominated by instead host and symbiont genes were both simply retained, the false positives and thus be highly misleading in identifying host protein being kept for chromist trans-ER translocation and sequence taxonomic affinities (see trenchant critiques in Woehle the red algal one adapted for translocation across the periplastid et al., 2011; Burki et al., 2012b). membrane. The branching near the base of the red algal tree indi- The apparently greater divergence of harosan putatively red cates that the symbiogenesis happened soon after the origin of red algal origin genes suggested by the Shih et al. (2013) results may algae as previous evidence and our evolutionary arguments indi- explain why our trees found more evidence for serious attraction cated, not substantially later as Baurain et al. (2010) assumed. problems between red algae and all hacrobian lineages, than for The Cdc48 tree of Felsner et al. (2011) also supports a red algal ori- Harosa. The fact that including only the shortest branch harosan gin, but unlike Petersen et al. (2014) strongly supports chromist Pythium causes LG ML to place it as sister to red algae suggests that monophyly; their Uba1 tree also suggests retention of the red algal a red algal attraction may be present but is quantitatively paralogue for periplastid membrane translocation. swamped by the greater divergence of longer branch Harosa hav- We cannot overemphasise that any red-algal tree distortion ing overwritten and swamped out the signal, so that long-branch artefact by unrecognised endosymbiont genes would be gene- exclusion dominates quantitatively over red algal attraction. If as and lineage-specific, which explains why published multigene has long been argued (Cavalier-Smith, 1982, 2007, 2013a) red trees with different gene sets have been so contradictory and con- algae diverged very early in Plantae, and diverged into subphyla fusing for corticates, but broadly remarkably consistent for the rest Eurhodophytina, sister to chromist chloroplasts, and Cyanidio- of the eukaryote tree. This gene and lineage specificity made it phytina almost immediately thereafter, then it follows that chro- hard to understand the reasons for their inconsistency until we mists probably also originated only just after the origin of red ran trees with exactly the same gene sets after systematically algae, when red algal genes would not have diverged much from excluding Plantae (Fig. 3) or Chromista (Fig. 6) or Rhodophyta those of green algae and glaucophytes. Differences between their (Fig. 9). Similar systematic exclusion studies are needed for other genes would have been so slight that hardly any genes acquired gene sets to see whether they also are subject to red algal attrac- by chromists by symbiogenetic gene transfer would group on tion effects, e.g. that of Derelle et al. (2015) when augmented by single-gene trees with high statistical support with red algae, Corbihelia, Haptista, and Glaucophyta. Plantae and Chromista making it almost impossible to detect and reliably eliminate them exclusion tests for this apparently less saturated gene set would all from data sets; for example the criterion of 70% bootstrap sup- 356 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 port for incongruent branching to detect wrong paralogues used by modification that not only Cryptophyceae, Ochrophyta, and hapto- Derelle et al. (in press) would almost never be effective, and similar phytes, but also alveolates (first defined by Cavalier-Smith, 1991) problems apply to the method of Leigh et al. (2008). Were it easy to stemmed from this same secondary symbiogenesis (Cavalier- demonstrate the holophyly of Plantae and Chromista, that would Smith, 1999). The endosymbiont was certainly an early red alga, refute the temporal close linking argued by Cavalier-Smith for not an even older immediate precursor of red plus green algae as the origins of chloroplasts and all major plant and chromist lin- originally envisaged (Cavalier-Smith, 1982). Just as Cavalier- eages. The difficulty of basally resolving corticate topology con- Smith (1999) argued, the chromist bipartite chloroplast-targeting firms the correctness of this general perspective of extremely sequence comprises an amino-terminal part for dual use as a signal rapid radiation of all algae in a relatively short timespan. sequence and for recognition by the periplastid membrane’s The above discussion followed Leigh et al. (2008) and import machinery, now known to have evolved from ER Deschamps and Moreira (2009) in assuming that the observed protein-export machinery (Stork et al., 2012), and a subterminal attraction of some chromist lineages to Rhodophyta is an artefact transit sequence recognised by Toc. Acceptance of the holophyly of persisting endosymbiotic red algal genes. However, that of the basically phagophototrophic Chromista has been delayed assumed that Plantae are really a clade as first argued by by the fact that several early diverging lineages lost plastids, unlike Cavalier-Smith (1981, 1982). However, it is in theory possible that in Plantae that earlier than chromistan algae became dependent on Chromista are a clade but Plantae are not and that this attraction to non-photosynthetic plastid functions like fatty and amino acid red algae is a genuine phylogenetic signal but obscured for Harosa synthesis (and almost none retained phagotrophy). by their often long branches. If the host of the ancestral symbio- The idea that tubular ciliary hairs were an ancestral character genesis were historically a sister not of Plantae as a whole, as we also for chromists (Cavalier-Smith, 1981, 1986) is now more consider most likely, but solely of red algae that had not yet lost strongly supported than before because of their discovery in Palpit- cilia, phagotrophy or cortical alveoli, and the endosymbiont was omonas and Lateronema, which branch rather early, and probably (as chloroplast trees suggest) a sister to Rhodophytina, then both correct. Molecular characterisation of cryptist tubular hairs host and symbiont would have been very closely related and both remains essential to corroborate this, but for now the original def- photosynthetic. This might have greatly helped the original estab- inition of chromists as having chloroplasts inside the ER and/or lishment of the secondary symbiogenesis mechanistically and is an tubular mastigonemes (Cavalier-Smith, 1981) seems vindicated; alternative explanation to symbiogenetic gene transfer to the it was always envisaged that some protists might have lost both apparent red algal signal in some chromists (a third possibility is defining characters and might therefore need eventually to be convergent fortuitous similarities having no phylogenetic mean- transferred in the light of sequence evidence (strongly advocated ing). We do not see how this can be ruled out by current sequence from multiple sources to test the system by Cavalier-Smith, phylogeny evidence, but have not favoured it in the past as if the 1981) from Protozoa to Chromista (explicity by reference to the host were photosynthetic it is not obvious that there would have Actinophryida (Cavalier-Smith, 1986, pp. 332–333), now in Ochro- been a selective advantage for the ancestral secondary symbiogen- phyta (Cavalier-Smith and Scoble, 2013) though they evolved from esis that created the first chromist chloroplast; therefore though raphidophytes not pedinellids as some had suggested). Other physically possible it is evolutionarily implausible. heliozoans were transferred to Chromista (Cavalier-Smith, Our 187-gene, eukaryote-wide CAT tree excluding chromists 2003a); so later were Alveolata and Rhizaria (Cavalier-Smith, (Fig. 6) robustly shows the accepted branching order between 2010a), as a result of multigene tree evidence for holophyly of the three plant groups, as for 19 plastid-targeted proteins Harosa (Burki et al., 2007), only weakly and intermittently seen (Reyes-Prieto and Bhattacharya, 2007), in 143-nuclear-gene trees on single-gene trees (e.g. Cavalier-Smith and Chao, 2003a). if chromists are excluded (Deschamps and Moreira, 2009), and However, as also predicted (Cavalier-Smith, 1986), sequence evi- 42 mitochondrial function genes from two genomes (Zhao et al., dence has never made any transfers necessary from Chromista to 2013). Our eukaryote-wide trees minus Plantae quite strongly Protozoa, as Chromista was originally well defined by two inde- show holophyly of Hacrobia – unlikely to be an artefact of missing pendent distinctive ultrastructural characters. data (Roure et al., 2013), though the fact that the fraction of miss- It is now clearer than before that the ancestor of all Cryptista had ing data is proportionally higher in Hacrobia than average for the tripartite tubular ciliary hairs on the anterior cilium (on posterior whole eukaryotic alignment remains a limitation for their accurate also only in Cryptophyceae), as did the ancestor of all Heterokonta placement on multigene trees. Within Hacrobia missing data are even though both groups have some species that secondarily lost probably not seriously misleading, as the tree is topologically them. In heterokonts there is also one case of movement of these identical and only minimally differs in a minority of support values hairs from cilium to cell body – Proteromonas (Brugerolle and when we exclude genes missing from >50% taxa (Fig. 5). Joyon, 1975; Brugerolle and Bardele, 1988). The aberrant chromist Though we identified two major artefacts that systematically Reticulosphaera socialis has tubular hairs on at least one cilium as distort corticate multigene trees (Harosa long branches, unde- well as on the cell body, the only known example (Grell et al., tected red algal paralogues in chromists) and consider them major 1990). However these hairs are thicker than the tubular hairs of causes of conflict in this part of the eukaryote tree (combined with other chromists, lack a terminal filament, and are not known to very rapid basal corticate radiation making resolution inherently have a distinctive base, so are unipartite. R. socialis was assumed challenging even if not worsened by them), we do not exclude to be a heterokont because of these hairs, but R. japonica the only the possibility of additional still unrecognised sources of system- other described species was found by rDNA sequencing to be an atic error. Nonetheless, no reason remains from multigene trees aberrant haptophyte Cavalier-Smith et al. (1999), but has not been to doubt the holophyly of both kingdoms Plantae and Chromista. studied ultrastructurally. These two species could (despite their Plantae are characterised by having chloroplasts that evolved in a names and broadly similar general light microscope morphology) single primary symbiogenesis by the evolution of a common belong to different chromist phyla, so we do not know whether R. protein-import machinery (the Toc/Tic multiprotein translocons socialis is a haptophyte or a heterokont; without protein or gene recognising transit sequences) exactly as Cavalier-Smith (1982) sequences it is impossible to say whether its hairs are related to argued. Likewise Chromista evolved by a single secondary symbio- heterokont or to cryptist tubular hairs and were dramatically mod- genesis involving a different protein-import machinery across the ified when partially transferred onto the cell body or purely conver- periplastid membrane and peri-chloroplast rough endoplasmic gent structures. We also do not know if R. japonica has tubular hairs reticulum, as Cavalier-Smith (1982) argued – with the later and so should not assume their presence in any Haptista. T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 357

Both the tubular hairs and the presence of autofluorescent car- et al. (2014) CAT trees). Until this contradiction is convincingly otenoids in the posterior cilium are unique chromist characters resolved it is best to treat Neolouka as a separate phylum. The unrelated to chloroplasts that point to the holophyly of chromists new clade name Opimoda of Derelle et al. (2015) would be (Cavalier-Smith, 1992); in theory the ciliary carotenoids could have an unnecessary and destabilising synonym of podiates been transferred from haptophytes to heterokonts by a hypotheti- (Cavalier-Smith, 2013b) if their tree topology placing Malawimonas cal tertiary symbiogenesis (though far fetched), but the hairs could within Sulcozoa were correct, and having no advantages should not as no Haptista have tripartite hairs. not replace podiates. However, it is premature to judge the precise position of Malawimonas within basal scotokaryotes, especially 4.10. Supergroup names from trees like those of Derelle et al. (2015) that necessarily exclude the anaerobic Metamonada, some of which can group with Confusion can arise from arbitrary or unnecessary name Malawimonas on multigene trees (e.g. Fig. 3; and Brown et al., changes, which should therefore be avoided, though some changes 2013; Cavalier-Smith et al., 2015). As CAT trees are considered may be needed for clarity or to reduce past confusions. The clade more accurate than site-homogeneous trees for the neomuran comprising Chromista and Plantae was originally called photokary- gene set (see Brown et al., 2013), we now also remove Metamon- otes (Cavalier-Smith, 1999) but following a criticism that this ada from Loukozoa (reinstating their phylum rank, Anaeromonada misleadingly implied a common photosynthetic ancestry of both and Trichozoa being subphyla again: Cavalier-Smith, 2003c) and groups (which would be untrue if they are sisters as most evidence similarly transfer them from protozoan subkingdom Eozoa to indicates, but true if chromists were nested within Plantae as a few subkingdom Neozoa (Cavalier-Smith, 1997), here redefined as trees have suggested) it was changed to corticates because superphylum Sarcomastigota (i.e. phyla Sulcozoa, Amoebozoa, both kingdoms have cortical alveoli in one or more phyla Choanozoa, Microsporidia) plus Metamonada and Neolouka. This (Cavalier-Smith, 2003a; see also 2003b; Cavalier-Smith and Chao, revision of protozoan classification makes taxonomic composition 2003a). However Adl et al. (2012) introduced an entirely unneces- of subkingdom Eozoa (introduced by Cavalier-Smith, 1997) the sary, and less euphonious third synonym with no intuitive mean- same as later Discoba (Hampl et al., 2009), i.e. Euglenozoa, Percolo- ing – Diaphoretickes, which is destabilising and should not be zoa, and Loukozoa (Eolouka only), which if the eukaryote root is used. To stabilise nomenclature we now formally establish superk- between Euglenozoa and Percolozoa/Eolouka (Figs. 1–3, 6 and 9) ingdom Corticata to embrace kingdoms Plantae and Chromista (i.e. is paraphyletic (not a clade) but is the deepest branching eukaryote to mean the same as the vernacular term ’corticates’ so either can clade on a 72-gene tree rooted on archaebacteria (Raymann et al., be used as is appropriate): 2015, Fig. 1) on which excavates were unfortunately too poorly represented to confirm where the Eozoa/Neozoa boundary lies. Diagnosis of new superkingdom Corticata Cavalier-Smith: Scotokaryotes, introduced to name the sister clade to corticates Eukaryotes ancestrally with a ventral feeding groove and (then photokaryotes) (Cavalier-Smith, 1999), originally was com- cytoskeleton comprising a single dorsal centriolar microtubular positionally equivalent to podiates, recently preferred as a more root and dorsal fan, posterior split right root with I fibre and left euphonious name referring to a putative synapomorphy root with C fibre, and intermediate single root, but distin- (Cavalier-Smith, 2013b), but we here made scotokaryotes distinc- guished from excavates by cortical alveoli and often having tive by broadening its composition to refer to the wider clade com- ciliary hairs not associated with a latticed paraxonemal rod; prising podiates, Neolouka and Metamonada, now shown by CAT phylogenetically comprises all descendants of the common trees (Figs. 2, 3, 6 and 9); it is apposite in that sense as scotos (dark ancestor of Arabidopsis and Paramecium. Etymology: cortex Gk) refers to the fact that the clade is the only one of the three L. bark, rind. Lankester (1878, 1885) divided Protozoa into Gym- major supergroups that includes no phototrophs, in contrast to nomyxa (essentially Amoebozoa plus Rhizaria and Labyrinthu- corticates and Eozoa. Trees for eubacterial origin genes (e.g. Zhao lea) and Corticata (essentially Alveolata plus the heterokont et al., 2013; Derelle et al., 2015) cannot be used to decide whether Paraphysomonas and plant Volvox, all now in Corticata, as well or not scotokaryotes are a clade as Metamonada lack most of these as choanoflagellates and Euglena). Superkingdom Corticata genes. refines Lankester’s group by excluding Euglenozoa and The nomenclaturally destabilising name Archaeplastida intro- choanoflagellates, but including eukaryotes that (a) later duced by Adl et al. (2005) as an exact synonym for the longstand- acquired chloroplasts and lost cortical alveoli if they also ing and more widely accepted Plantae sensu Cavalier-Smith (1981) evolved cell walls, or (b) became secondarily amoeboid and so was completely unnecessary and less euphonious, so its use should lost cortical alveoli. be strongly discouraged. Plantae was probably first made a kingdom that included green plants (including green algae) and If the eukaryote root is between Euglenozoa and all other red algae by Haeckel (1866), though he wrongly also included eukaryotes (Cavalier-Smith, 2010a; Lasek-Nesselquist and cyanobacteria, Fungi, and chromist brown algae; it was unwise to Gogarten, 2013), the sister clade to Corticata is podiates (ML trees) attempt to replace that venerable and widely used name, as it had or scotokaryotes (podiates plus Metamonada and Neolouka: CAT been correctly refined by excluding them long ago and has been in trees). Here we formally remove Malawimonas (subphylum continuous use in the precise sense used here for 34 years. Neolouka) from Loukozoa, making Neolouka a new phylum (Diag- Although we have reasonably well established holophyly for nosis as for subphylum Neolouka: Cavalier-Smith, 2013b, p. 122) Corticata, Chromista, and Hacrobia, established new subphylum within protozoan subkingdom Neozoa, not in Eozoa as formerly. Corbihelia, the basal branching order of Plantae, and the mono- Cavalier-Smith (2013b) argued that Malawimonas should be placed phyly of podiates, there are three major remaining open questions within Sulcozoa if multigene trees confirmed that relationship. Our concerning eukaryote deep phylogeny. Firstly, whether the root is present and recent trees (Cavalier-Smith et al., 2014, 2015) using indeed between Euglenozoa and all other eukaryotes, as we argue, genes of neomuran origin decisively show that Malawimonas is or instead between scotokaryotes and Corticata/Eozoa as some more closely related to Sulcozoa than to jakobids, as two different trees suggest (Derelle at al., 2015; if this were true Eozoa/Discoba gene sets of eubacterial origin also strongly do (Derelle et al., would be a clade or else between Eozoa and Corticata/scotokary- 2015). Some of our CAT trees (e.g. Figs. 1 and 2) actually place otes as others suggest (Raymann et al. 2015)). Secondly, whether Malawimonas within Sulcozoa but others (Figs. 3 and 9) place them Metamonada is a clade as our ML and some CAT trees indicate or and some or all metamonads below podiates (as in Cavalier-Smith is polyphyletic with separate clades Anaeromonada and Trichozoa 358 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 as other CAT trees suggest (though this would be evolutionarily figured in Butterfield (2015, Fig. 2) are only 750 My old; his acceptable if cilia multiplied and anaerobiosis evolved indepen- assertion that Cavalier-Smith (2013b) accepted acanthomorphic dently; Fig. 6 had maximal support for metamonad polyphyly acritarchs over twice as old as having ’eukaryotic affiliations’ is (both chains), but with other taxon samples the two CAT chains incorrect; it was also wrong to call those in his Fig. 4 ’unambiguous disagreed, one chain showing polyphyly and the other holophyly); eukaryotic fossils’, though sensible to have given up his earlier sequencing transcriptomes of short branch Trichozoa should dis- view that they are fungi. They are highly ambiguous: Tappania tinguish between metamonad holophyly (most likely) and poly- plana might be actinomycete-like, albeit of rare/unusual form, as phyly. Thirdly, at the base of scotokaryotes the branch order of suggested previously (Cavalier-Smith, 2006a), or some extinct Malawimonas, metamonad, and varisulcan taxa remains contradic- bacterial group; though Cavalier-Smith (2013b) recognised that tory on CAT trees with different taxon samples; note that Shuiyousphaeridium ’might be’ stem eukaryotes, as might a-proteobacterial and eubacterial gene sets (Derelle et al., 2015) \Dictyosphaera and Gigantosphaeridium, the alternative that they agree with our present neomuran gene sets in placing Malawi- are unusually complex prokaryotes (Cavalier-Smith, 2006a) monas in scotokaryotes, but contradict our taxonomically much remains perfectly reasonable and cannot strictly be ruled out; better sampled trees by placing them within podiates not as close claims that they are green algae with plastids are groundless and sisters of podiates – their two data sets are also mutually contra- the idea that Shuiyousphaeridium and Dictyosphaera are phases in dictory with respect to Malawimonas’s exact position and that of a single sexual green algal heteromorphic life cycle even wilder Collodictyon, the a-proteobacterial set making them paraphyletic speculation (Agic´ et al., 2015). There is no convincing evidence that relatives of Amoebozoa, and the eubacterial set a sister clade to these early acritarchs are even crown eukaryotes – that assump- Amoebozoa plus Apusozoa and opisthokonts (as with neomuran tion seems refuted by dated multigene analyses (Eme et al., genes), but neither can help with metamonads as they lack aerobic 2014); their estimates for the age of red algae (thus a maximum mitochondrial functions. age for chromists) range hugely from 636 to 1206 My depending on assumptions, the most realistic favouring the youngest dates, 4.11. Corticate and scotokaryote big-bang radiations favouring the view that Bangiomorpha is probably not a red alga (Cavalier-Smith, 2006a). Butterfield (2015) admits that ’the oldest In addition to a few important differences in topology between clearly indigenous steranes that can be convincingly allied with our ML and CAT trees there is a striking difference in the propor- eukaryotes’ are only 750 My old. If Plantae arose 800 ± 50 My tions of the basal stems and later branches within corticates and ago and Chromista 750 ± 50 My (Cavalier-Smith, 2013b), trigger- scotokaryotes. Basal branches appear to be relatively more spread ing immediately the scotokaryote/corticate big bang radiations, the out with ML (e.g. Fig. 1) than with CAT where the basal radiations proportions of our CAT trees (e.g. Fig. 2) make it possible that of both supergroups appears to be explosive, taking place so crown eukaryotes arose only 0.9–1.2 Gy ago. Cross-calibrated phy- rapidly that resolution is hard. Given the greater evolutionary real- logenetic dating of duplicated paralogues (Shih and Matzke, 2013) ism of the site-heterogeneous CAT model it is likely that this very gives similar dates for the mitochondrial symbiosis (i.e. stem rapid radiation at the base of both clades was a real historical eukaryotes): ATPase a-subunit 1.249 Gy; b-subunit 1.176 Gy; event. The simplest explanation for the near simultaneity of both Elongation factor Tu 1.196 Gy. But for the origin of chloroplasts radiations is that the symbiogenetic origin of chloroplasts and Plantae dates are more contradictory: ATPase a-subunit (Cavalier-Smith, 2013a) stimulated not only the primary 3-fold 1.055 Gy; b-subunit 0.857 Gy; Tu 1.188 Gy. Though error margins radiation of Plantae directly but also the near simultaneous pri- are wide, these results imply that crown eukaryotes are no older mary 4-fold radiation of Chromista following the red algal sec- than 1.2 Gy and support the unpopular but carefully argued view ondary enslavement but also the basal radiation of scotokaryotes that many palaeontologists have been too ready to accept doubtful because of the novel foods and niches provided by these photosyn- evidence for substantially earlier origins of eukaryotes, plants, and thetic radiations and the associated sudden increase in ecosystem chromists (Cavalier-Smith, 2002a, 2006a, 2010b). complexity (Cavalier-Smith, 2002b). By contrast megadiversity was much lower and radiation of Eozoa less extensive and more 5. Major conclusions gradual in the absence of eukaryote algae; Euglenophyceae, chromist chloraracheans (e.g. Bigelowiella) and a small peridinean 1. Haptista (centrohelid heliozoa, haptophytes, rappemonads) are dinoflagellate lineage acquired green algal chloroplasts by sec- a clade ancestrally characterised by large flattened mineralised ondary enslavement substantially after the origins of their parent body scales and one or many microtubule-supported filiform phyla and did not have as great an evolutionary impact as chromo- cell extensions. phytes (Cavalier-Smith, 2013a). 2. Endohelea (non-ciliate Microheliella and biciliate Heliomorpha) The most thorough Bayesian multiprotein molecular ’clock’ are sister to a new cryptist superclass Corbistoma (Telonemea, study yet strongly supports the idea of a massive early radiation Picomonas; united by a unique cytopharyngeal basket and the of scotokaryotes and corticates (Eme et al., 2014); the date sug- unique division of the cell into two distinct zones, one for inges- gested by their most realistic Bayesian lognormal ’clock’ assumption and one for the other main organelles). The new cryptist tions for the eukaryote root is about 1 Gy ago, agreeing with subphylum Corbihelia characterised by a unique filogranular estimates by direct critical interpretation of the fossil record by network interconnecting nuclei, mitochondria and other orga- Cavalier-Smith (2002b, 2006a, 2013b, 2014) of 0.85–1.2 Gy ago. nelles comprises Endohelia plus Corbistoma. Eme et al. tested three assumptions about the root position, but 3. Cryptista ancestrally had tubular ciliary hairs (because the new did not include a root between Euglenozoa and all other eukary- telonemean genus Lateronema has tripartite hairs with bulbous otes as we favour (Cavalier-Smith, 2010a, 2013b, 2014) and used base and its divergence from Palpitomonas with indistinguish- for the present trees. Such a root might give a date closer to able hairs is the deepest in Cryptista). 1.1–1.2 Gy (for reasons see Cavalier-Smith, 2013b), but each 4. Endohelia evolved axopodia independently of centrohelids but improvement in ’clock’ models (there is in fact no clock!) reduces inherited similar extrusomes from a common ancestor. the estimated root date, making it closer to what the fossil record 5. Hacrobia (phyla Haptista, Cryptista) are a clade, ancestrally directly reveals, so even more realistic models could reduce it with tripartite ciliary hairs if these are homologous with those further. The only truly unambiguous crown eukaryote fossils of Heterokonta. We discovered a new striking cytoskeletal T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 359

synapomorphy for Hacrobia that reinforces the already strong Appendix A. Supplementary material evidence from our multigene trees for their holophyly and rules out the possibility that any of them belong in Plantae as some Supplementary data associated with this article can be found, in trees have misleadingly implied. the online version, at http://dx.doi.org/10.1016/j.ympev.2015.07. 6. Chromista are holophyletic with Hacrobia and Harosa sisters, as 004. shown robustly by multigene trees for 42 proteins of a-proteobacterial origin acquired during the mitochondrial References symbiogenesis. On multigene tree trees for proteins of host neomuran origins this is hard to demonstrate for three reasons: (a) Adl, S.M., Simpson, A.G., Farmer, M.A., Andersen, R.A., Anderson, O.R., Barta, J.R., a red-algal gene distortion; (b) the rapid evolution that has Bowser, S.S., Brugerolle, G., Fensome, R.A., Fredericq, S., James, T.Y., Karpov, S., Kugrens, P., Krug, J., Lane, C.E., Lewis, L.A., Lodge, J., Lynn, D.H., Mann, D.G., made Harosa long-branches tends to pull them too far down McCourt, R.M., Mendoza, L., Moestrup, O., Mozley-Standridge, S.E., Nerad, T.A., the tree away from shorter branch Hacrobia and Plantae; and Shearer, C.A., Smirnov, A.V., Spiegel, F.W., Taylor, M.F., 2005. The new higher (c) many genes from this set have also generally evolved faster level classification of eukaryotes with emphasis on the taxonomy of protists. J. Eukaryot. Microbiol. 52, 399–451. and become almost substitutionally saturated for regions that Adl, S.M., Simpson, A.G., Lane, C.E., Lukes, J., Bass, D., Bowser, S.S., Brown, M.W., diverged during the numerous close bifurcations in this rapid Burki, F., Dunthorn, M., Hampl, V., Heiss, A., Hoppenrath, M., Lara, E., Le Gall, L., corticate basal radiation, making it intrinsically hard to resolve Lynn, D.H., McManus, H., Mitchell, E.A., Mozley-Stanridge, S.E., Parfrey, L.W., Pawlowski, J., Rueckert, S., Shadwick, R.S., Schoch, C.L., Smirnov, A., Spiegel, even without the sources of systematic bias we have revealed. F.W., 2012. The revised classification of eukaryotes. J. Eukaryot. Microbiol. 59, 7. Superkingdom Corticata (comprising sister kingdoms 429–493. Chromista and Plantae) is a clade. Two early predictions Agic´, H., Moczydłowska, M., Yin, L.-M., 2015. Affinity, life cycle, and intracellular complexity of organic-walled microfossils from the Mesoproterozoic of Shanxi, (Cavalier-Smith, 1982) have been confirmed with no com- China. J. Paleontol. 89, 28–50. pelling contrary evidence: (a) Chloroplasts of Plantae originated Bapteste, E., Brinkmann, H., Lee, J.A., Moore, D.V., Sensen, C.W., Gordon, P., Duruflé, from a cyanobacterium by the origin of a shared novel L., Gaasterland, T., Lopez, P., Müller, M., Philippe, H., 2002. The analysis of 100 genes supports the grouping of three highly divergent amoebae: Dictyostelium, protein-import machinery in a biciliate host with cortical Entamoeba, and Mastigamoeba. Proc. Natl. Acad. Sci. USA 99, 1414–1419. alveoli, non-tubular ciliary hairs, and a multilayered centriolar Bardele, C.F., 1972. Cell cycle, morphogenesis, and ultrastructure in the root (now recognised as the left posterior root with C fibre of pseudoheliozoan Clathrulina elegans. Z. Zellforsch. Mikrosk. Anat. 130, 219–242. excavate origins). (b) Chloroplasts of Chromista originated by Bardele, C.F., 1975. The fine structure of the centrohelidian heliozoan Heterophrys marina. Cell Tissue Res. 161, 85–102. a single secondary enslavement of a red alga by the origin of dif- Bass, D., Chao, E.E., Nikolaev, S., Yubuki, A., Ishida, K., Berney, C., et al., 2009. ferent common protein-import machinery at the same time as Phylogeny of novel naked filose and reticulose Cercozoa: Granofilosea cl. n. and tripartite tubular ciliary hairs and that all chromists without Proteomyxidea revised. Protist 160, 75–109. Baurain, D., Brinkmann, H., Petersen, J., Rodríguez-Ezpeleta, N., Stechmann, A., plastids or tubular hairs lost them. Demoulin, V., et al., 2010. Phylogenomic evidence for separate acquisition of 8. As chromists are evolutionary chimaeras of a red alga and a plastids in cryptophytes, haptophytes, and stramenopiles. Mol. Biol. Evol. 27, non-photosynthetic host undetected genes of red algal origin 1698–1709. Bergholtz, T., Daugbjerg, N., Moestrup, Ø., 2005. On the identity of Karlodinium seriously distort multigene phylogeny of corticates (for both veneficum and description of Karlodinium armiger sp. nov. (Dinophyceae), based Chromista and Plantae) and tend to pull red algae away from on light and electron microscopy, nuclear-encoded LSU rDNA, and pigment other Plantae towards Hacrobia causing plant and hacrobian composition. J. Phycol. 42, 170–193. Bråte, J., Klaveness, D., Rygh, T., Jakobsen, K.S., Shalchian-Tabrizi, K., 2010. groups to intermingle. These distortions are gene and lineage Telonemia-specific environmental 18S rDNA PCR reveals unknown diversity specific and so are not consistent across taxon and gene and multiple marine–freshwater colonizations. BMC Microbiol. 10, 168. sampling, thus accounting for many published contradictory Brown, M.W., Sharpe, S.C., Silberman, J.D., Heiss, A.A., Lang, B.F., Simpson, A.G., et al., 2013. Phylogenomics demonstrates that breviate flagellates are related to corticate multigene trees where these differed. opisthokonts and apusomonads. Proc. Roy. Soc. B 280, 20131755. 9. Arguments against chromist holophyly based on bipartition Brugerolle, G., Bardele, C.F., 1988. Cortical cytoskeleton of the flagellate support values are fallacious; had they been valid they would Proteromonas lacertae: interrelation between microtubules, membrane and argue equally against plant holophyly. somatonemes. Protoplasma 142, 46–54. Brugerolle, G., Joyon, L., 1975. Étude cytologique ultrastructurales des genres Proteromonas et Karotomopha (Zoomastigophorea, Proteromonadida Grassé Data accessibility 1952). Protistologica 11, 531–546. Brugerolle, G., Mignot, J.-P., 1984. Les caractéristiques ultrastructurales de l’hélioflagellé Dimorpha mutans Gruber (Sarcodina – Actinopoda) et leur Sequences are deposited in GenBank under two BioProject intérêt phylétique. Protistologica 20, 97–112. numbers: PRJNA73901 (Microheliella); PRJNA222678 (Oxnerella Burki, F., Corradi, N., Sierra, R., Pawlowski, J., Meyer, G.R., Abbott, C.L., et al., 2013. and Minimassisteria). Phylogenomics of the intracellular parasite Mikrocytos mackini reveals evidence for a mitosome in Rhizaria. Curr. Biol. 23, 1541–1547. Burki, F., Inagaki, Y., Bråte, J., Archibald, J.M., Keeling, P.J., Cavalier-Smith, T., et al., Acknowledgments 2009. Large-scale phylogenomic analyses reveal that two enigmatic protist lineages, Telonemia and Centroheliozoa, are related to photosynthetic chromalveolates. Genome Biol. Evol. 1, 231–238. We thank NERC for Grant NE/E004156/1 and for an earlier Burki, F., Kudryavtsev, A., Matz, M.V., Aglyamova, G.V., Bulman, S., Fiers, M., et al., Professorial Fellowship for TCS. RL thanks NERC for a research 2010. Evolution of Rhizaria: new insights from phylogenomic analysis of studentship. We thank Cédric Berney, Anna Maria Fiore-Donno, uncultivated protists. BMC Evol. Biol. 10, 377. Burki, F., Flegontov, P., Obornik, M., Cihlar, J., Pain, A., Lukes, J., Keeling, P.J., 2012a. and Elizabeth Snell for help editing parts of a much smaller precur- Re-evaluating the green versus red signal in eukaryotes with secondary plastid sor of our alignment; and Matthew Brown, Fabien Burki, Vladimir of red algal origin. Genome Biol. Evol. 4, 626–635. Hampl, Naira Rodríguez-Espeleta, Laura Eme, and Kamran Burki, F., Okamoto, N., Pombert, J.F., Keeling, P.J., 2012b. The evolutionary history of Shalchian-Tabrizi for providing gene alignments or protein haptophytes and cryptophytes: phylogenomic evidence for separate origins. Proc. Roy. Soc. B 279, 2246–2254. sequences. We thank Helge A. Thomsen for permission to use his Burki, F., Shalchian-Tabrizi, K., Minge, M., Skjaeveland, A., Nikolaev, S.I., Jakobsen, unpublished micrographs of Lateronema, and Dag Klaveness for K.S., et al., 2007. Phylogenomics reshuffles the eukaryotic supergroups. PLoS providing them and his own unpublished micrographs of Telonema One 2, e790. Burki, F., Shalchian-Tabrizi, K., Pawlowski, J., 2008. Phylogenomics reveals a new and for comments on telonemid swimming and feeding behaviour. ’megagroup’ including most photosynthetic eukaryotes. Biol. Lett. 4, 366–369. We thank Oxford University Advanced Research Computing (ARC: Butterfield, N.J., 2015. Frontiers in paleontology: early evolution of the Eukaryota. http://dx.doi.org/10.5281/zenodo.22558) for access to their clus- Palaeontology 58, 5–17. Cacan, E., Kratzer, J.T., Cole, M.F., Gaucher, E.A., 2013. Interchanging functionality ters (including the IRIDIS HPC facility of the Centre for Innovation) among homologous elongation factors using signatures of heterotachy. J. Mol. and efficient advice. Evol. 76, 4–12. 360 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

Cavalier-Smith, T., 1980. Cell compartmentation and the origin of eukaryote Cavalier-Smith, T., von der Heyden, S., 2007. Molecular phylogeny, scale membranous organelles. In: Schwemmler, W., Schenk, H.E.A. (Eds.), evolution and taxonomy of centrohelid Heliozoa. Mol. Phylogenet. Evol. 44, Endocytobiology: Endosymbiosis and Cell Biology, a Synthesis of Recent 1186–1203. Research. de Gruyter, Berlin, pp. 893–916. Cavalier-Smith, T., Allsopp, M.T., Chao, E.E., 1994. Chimeric conundra: are Cavalier-Smith, T., 1981. Eukaryote kingdoms: seven or nine? BioSystems 14, 461– nucleomorphs and chromists monophyletic or polyphyletic? Proc. Natl. Acad. 481. Sci. USA 91, 11368–11372. Cavalier-Smith, T., 1982. The origins of plastids. Biol. J. Linn. Soc. 17, 289–306. Cavalier-Smith, T., Allsopp, M.T.E.P., Chao, E.E., Boury-Esnault, N., Vacelet, J., 1996a. Cavalier-Smith, T., 1986. The kingdom Chromista: origin and systematics. In: Sponge phylogeny, animal monophyly and the origin of the nervous system: Round, F.E., Chapman, D.J. (Eds.), Progress in Phycological Research. Biopress, 18S rRNA evidence. Can. J. Zool. 74, 2031–2045. Bristol, pp. 309–347. Cavalier-Smith, T., Allsopp, M.T.E.P., Häuber, M.M., Gothe, G., Chao, E.E., Couch, J., Cavalier-Smith, T., 1989. The kingdom Chromista. In: Green, J.C., Leadbeater, B.S.C., Maier, U.-G., 1996b. Chromobiote phylogeny: the enigmatic alga Diver, W.L. (Eds.), The Chromophyte Algae: Problems and Perspectives. Reticulosphaera japonensis is an aberrant haptophyte, not a heterokont. Eur. J. Clarendon Press, Oxford, pp. 381–407. Phycol. 31, 255–263. Cavalier-Smith, T., 1991. Cell diversification in heterotrophic flagellates. In: Cavalier-Smith, T., Chao, E.E., Snell, E.A., Berney, C., Fiore-Donno, A.M., Lewis, R., Patterson, D.J., Larsen, J. (Eds.), The Biology of Free-living Heterotrophic 2014. Multigene eukaryote phylogeny reveals the likely protozoan ancestors of Flagellates. Clarendon Press, Oxford, pp. 113–131. opisthokonts (animals, fungi, choanozoans) and Amoebozoa. Mol. Phylogenet. Cavalier-Smith, T., 1992. The number of symbiotic origins of organelles. BioSystems Evol. 81, 71–85. 28, 91–106 (discussion 107–108). Cavalier-Smith, T., Chao, E.E., Thompson, C.E., Hourihane, S.L., 1995/1996. Cavalier-Smith, T., 1993. Kingdom Protozoa and its 18 phyla. Microbiol. Rev. 57, Oikomonas, a distinctive zooflagellate related to chrysomonads. Arch. 953–994. Protistenkd. 146, 273–279. Cavalier-Smith, T., 1997. Amoeboflagellates and mitochondrial cristae in eukaryote Cavalier-Smith, T., Fiore-Donno, A.M., Chao, E.E., Kudryavtsev, A., Berney, C., Snell, evolution: megasystematics of the new protozoan subkingdoms Eozoa and E.A., Lewis, R., 2015. Multigene phylogeny resolves deep branching of Neozoa. Arch. Protistenkd. 147, 237–258. Amoebozoa. Mol. Phylogenet. Evol. 215, 293–304. http://dx.doi.org/10.1016/ Cavalier-Smith, T., 1999. Principles of protein and lipid targeting in secondary j.ympev.2014.1008.1011. symbiogenesis: euglenoid, dinoflagellate, and sporozoan plastid origins and the Clay, B., Kugrens, P., 1999. Systematics of the enigmatic kathablepharids, including eukaryotic family tree. J. Eukaryot. Microbiol. 46, 347–366. EM characterization of the type species, Kathablepharis phoenikoston, and new Cavalier-Smith, T., 2000. Flagellate megaevolution: the basis for eukaryote observations on K. remigera comb. nov. Protist 150, 43–59. diversification. In: Green, J.C., Leadbeater, B.S.C. (Eds.), The Flagellates. Taylor Derelle, R., Lang, B.F., 2012. Rooting the eukaryotic tree with mitochondrial and and Francis, London, pp. 361–390. bacterial proteins. Mol. Biol. Evol. 29, 1277–1289. Cavalier-Smith, T., 2002a. The neomuran origin of archaebacteria, the negibacterial Derelle, R., Torruella, G., Klimeš, V., Brinkmann, H., Kim, E., Vlcˇek, C., Lang, B.F., Eliaš, root of the universal tree and bacterial megaclassification. Int. J. Syst. Evol. M., 2015. Bacterial proteins pinpoint a single eukaryotic root. Proc. Natl. Acad. Microbiol. 52, 7–76. Sci. USA 112, E693–E699. http://dx.doi.org/10.1073/pnas.1420657112. Cavalier-Smith, T., 2002b. The phagotrophic origin of eukaryotes and phylogenetic Deschamps, P., Moreira, D., 2009. Signal conflicts in the phylogeny of the primary classification of Protozoa. Int. J. Syst. Evol. Microbiol. 52, 297–354. photosynthetic eukaryotes. Mol. Biol. Evol. 261, 2745–2753. Cavalier-Smith, T., 2003a. Genomic reduction and evolution of novel genetic Douglas, S., Zauner, S., Fraunholz, M., Beaton, M., Penny, S., Deng, L.T., Wu, X., Reith, membranes and protein-targeting machinery in eukaryote–eukaryote M., Cavalier-Smith, T., Maier, U.G., 2001. The highly reduced genome of an chimaeras (meta-algae). Philos. Trans. Roy. Soc. Lond. B 358, 109–134. enslaved algal nucleus. Nature 410, 1091–1096. Cavalier-Smith, T., 2003b. Protist phylogeny and the high-level classification of Dürrschmidt, M., Patterson, D.J., 1987a. A light and electron microscopic study of a Protozoa. Eur. J. Protistol. 39, 338–348. new species of centroheliozoon, Chlamydaster fimbriatus. Tissue Cell 19, 365–376. Cavalier-Smith, T., 2003c. The excavate protozoan phyla Metamonada Grassé Dürrschmidt, M., Patterson, D.J., 1987b. On the organization of the Heliozoa emend. (Anaeromonadea, Parabasalia, Carpediemonas, Eopharyngia) and Raphidiophrys ambigua Penard and R. pallida Schulze. Ann. Sci. Nat., Zool., Paris Loukozoa emend. (Jakobea, Malawimonas): their evolutionary affinities and 13 serie, 8, 135–155. new higher taxa. Int. J. Syst. Evol. Microbiol. 53, 1741–1758. Eme, L., Sharpe, S.C., Brown, M.W., Roger, A.J., 2014. On the age of eukaryotes: Cavalier-Smith, T., 2004. Chromalveolate diversity and cell megaevolution: evaluating evidence from fossils and molecular clocks. Cold Spring Harbor interplay of membranes, genomes and cytoskeleton. In: Hirt, R.P., Horner, D.S. Perspect. Biol. 6. http://dx.doi.org/10.1101/cshperspect.a016139. (Eds.), Organelles, Genomes and Eukaryote Phylogeny. CRC Press, London, pp. Felsner, G., Sommer, M.S., Gruenheit, N., Hempel, F., Moog, D., Zauner, S., Martin, W., 75–108. Maier, U.G., 2011. ERAD components in organisms with complex red plastids Cavalier-Smith, T., 2006a. Cell evolution and Earth history: stasis and revolution. suggest recruitment of a preexisting protein transport pathway for the Philos. Trans. Roy. Soc. B 361, 969–1006. periplastid membrane. Genome Biol. Evol. 3, 140–150. Cavalier-Smith, T., 2006b. Rooting the tree of life by transition analysis. Biol. Direct Fu, G., Nagasato, C., Oka, S., Cock, J.M., Motomura, T., 2014. Proteomics analysis 1, 19. of heterogeneous flagella in brown algae (stramenopiles). Protist 165, 662– Cavalier-Smith, T., 2007. Evolution and relationships of algae: major branches of the 675. tree of life. In: Brodie, J., Lewis, J. (Eds.), Unravelling the Algae. CRC Press, Boca Gray, M.W., 1989. The evolutionary origins of organelles. Trends Genet. 5, 294–299. Raton, London, pp. 21–55. Green, J.C., 1973. Studies in the fine structure and taxonomy of flagellates in the Cavalier-Smith, T., 2009. Megaphylogeny, cell body plans, adaptive zones: causes genus Pavlova. II. A freshwater representative, Pavlova granifera (Mack) comb. and timing of eukaryote basal radiations. J. Eukaryot. Microbiol. 56, 26–33. nov. Br. Phycol. J. 8, 1–12. Cavalier-Smith, T., 2010a. Kingdoms Protozoa and Chromista and the eozoan root of Green, J.C., 1980. The fine structure of Pavlova pinguis Green and a preliminary the eukaryotic tree. Biol. Lett. 6, 342–345. survey of the order Pavlovales. Br. Phycol. J. 15, 151–191. Cavalier-Smith, T., 2010b. Deep phylogeny, ancestral groups, and the four ages of Green, J.C., Hibberd, D.J., 1977. The ultrastructure and taxonomy of Diacronema life. Philos. Trans. Roy. Soc. B 365, 111–132. vlkianum (Prymnesiophyceae) with special reference to the haptonema and Cavalier-Smith, T., 2013a. Symbiogenesis: mechanisms, evolutionary consequences, flagellar apparatus. J. Mar. Biol. Assoc. UK 57, 1125–1136. and systematic implications. Annu. Rev. Ecol. Evol. Syst. 44, 145–172. Green, J.C., Jordan, R.W., 2002. Order Prymnesiida. In: Lee, J.J., Leedale, G., Bradbury, Cavalier-Smith, T., 2013b. Early evolution of eukaryote feeding modes, cell P. (Eds.), The Illustrated Guide to the Protozoa, second ed. Society of structural diversity, and classification of the protozoan phyla Loukozoa, Protozoologists, pp. 1268–1302. Sulcozoa, and Choanozoa. Eur. J. Protistol. 49, 115–178. Grell, K.G., Heini, A., Schüller, S., 1990. The ultrastructure of Reticulosphaera socialis Cavalier-Smith, T., 2014. The neomuran revolution and phagotrophic origin of Grell (Heterokontophyta). Eur. J. Protistol. 26, 37–54. eukaryotes and cilia in the light of intracellular coevolution and a revised tree of Hackett, J.D., Yoon, H.S., Li, S., Reyes-Prieto, A., Rummele, S.E., Bhattacharya, D., life. In: Keeling, P.J., Koonin, E.V. (Eds.), The Origin and Evolution of Eukaryotes. 2007. Phylogenomic analysis supports the monophyly of cryptophytes and Cold Spring Harbor Perspectives Biol, Cold Spring Harbor. http://dx.doi.org/ haptophytes and the association of Rhizaria with chromalveolates. Mol. Biol. 10.1101/cshperspect.a016006. Evol. 24, 1702–1713. Cavalier-Smith, T., Chao, E.E., 2003a. Phylogeny of Choanozoa, Apusozoa, and other Haeckel, E., 1866. Generelle Morphologie der Organismen. Reimer, Berlin. Protozoa and early eukaryote megaevolution. J. Mol. Evol. 56, 540–563. Hampl, V., Hug, L., Leigh, J.W., Dacks, J.B., Lang, B.F., Simpson, A.G., Roger, A.J., 2009. Cavalier-Smith, T., Chao, E.E., 2003b. Molecular phylogeny of centrohelid heliozoa, a Phylogenomic analyses support the monophyly of Excavata and resolve novel lineage of bikont eukaryotes that arose by ciliary loss. J. Mol. Evol. 56, relationships among eukaryotic ‘‘supergroups’’. Proc. Natl. Acad. Sci. USA 106, 387–396. 3859–3864. Cavalier-Smith, T., Chao, E.E., 2012. Oxnerella micra sp. n. (Oxnerellidae fam. n.), a He, D., Fiz-Palacios, O., Fu, C.-J., Fehling, J., Tsai, C.-C., Baldauf, S.L., 2014. An tiny naked centrohelid, and the diversity and evolution of Heliozoa. Protist 163, alternative root for the eukaryote tree of life. Curr. Biol. 24, 465–470. http:// 574–601. dx.doi.org/10.1016/j.cub.2014.01.036. Cavalier-Smith, T., Karpov, S.A., 2012. Paracercomonas kinetid ultrastructure, origins Hibberd, D.J., 1971. The ultrastructure and taxonomy of the Chrysophyceae of the body plan of Cercomonadida, and cytoskeleton evolution in Cercozoa. and Prymnesiophyceae (Haptophyceae): a survey with some new Protist 163, 47–75. observations on the ultrastructure of the Chrysophyceae. Biol. J. Linn. Soc. 72, Cavalier-Smith, T., Scoble, J.M., 2013. Phylogeny of Heterokonta: Incisomonas 55–80. marina, a uniciliate gliding opalozoan related to Solenicola (Nanomonadea), Hibberd, D.J., 1972. Chrysophyta: definition and interpretation. Br. Phycol. J. 7, 281. and evidence that Actinophryida evolved from raphidophytes. Eur. J. Protistol. Howe, A.T., Bass, D., Scoble, J.M., Lewis, R., Vickerman, K., Arndt, H., et al., 2011. 49, 328–353. Novel cultured protists identify deep-branching environmental DNA clades of T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362 361

Cercozoa: new genera Tremula, Micrometopion, Minimassisteria, Nudifila, Petersen, J., Ludewig, A.K., Michael, V., Bunk, B., Jarek, M., Baurain, D., Brinkmann, H., Peregrinia. Protist 162, 332–372. 2014. Chromera velia, endosymbioses and the rhodoplex hypothesis – plastid Janouškovec, J., Horák, A., Oborník, M., Lukeš, J., Keeling, P.J., 2010. A common red evolution in cryptophytes, alveolates, stramenopiles, and haptophytes (CASH algal origin of the apicomplexan, dinoflagellate, and heterokont plastids. Proc. lineages). Genome Biol. Evol. 6, 666–684. Natl. Acad. Sci. USA 107, 10949–10954. Philippe, H., Lopez, P., 2001. On the conservation of protein sequences in evolution. Katz, L.A., Grant, J.R., 2015. Taxon-rich phylogenomic analyses resolve the Trends Biochem. Sci. 26, 414–416. eukaryotic tree of life and reveal the power of subsampling by sites. Syst. Philippe, H., Casane, D., Gribaldo, S., Lopez, P., Meunier, J., 2003. Heterotachy and Biol. 64, 406–415. functional shift in protein evolution. IUBMB Life 55, 257–265. Keeling, P.J., 2009. Chromalveolates and the evolution of plastids by secondary Raymann, K., Brochier-Armanet, C., Gribaldo, S., 2015. The two domain tree of endosymbiosis. J. Eukaryot. Microbiol. 56, 1–8. life is linked to a new root for the Archaea. Proc. Natl. Acad. Sci. USA 112, Keeling, P.J., 2013. The number, speed, and impact of plastid endosymbioses in 6670–6675. eukaryotic evolution. Annu. Rev. Plant Biol. 64, 583–607. Reeb, V.C., Peglar, M.T., Yoon, H.S., Bai, J.R., Wu, M., Shiu, P., et al., 2009. Keeling, P.J., Burki, F., Wilcox, H.M., Allam, B., Allen, E.E., Amaral-Zettler, L.A., Interrelationships of chromalveolates within a broadly sampled tree of Armbrust, E.V., Archibald, J.M., Bharti, A.K., Bell, C.J., Beszteri, B., Bidle, K.D., photosynthetic protists. Mol. Phylogenet. Evol. 53, 202–211. Cameron, C.T., Campbell, L., Caron, D.A., Cattolico, R.A., Collier, J.L., Coyne, K., Reyes-Prieto, A., Bhattacharya, D., 2007. Phylogeny of nuclear-encoded plastid- Davy, S.K., Deschamps, P., Dyhrman, S.T., Edvardsen, B., Gates, R.D., Gobler, C.J., targeted proteins supports an early divergence of glaucophytes within Plantae. Greenwood, S.J., Guida, S.M., Jacobi, J.L., Jakobsen, K.S., James, E.R., Jenkins, B., Mol. Biol. Evol. 24, 2358–2361. John, U., Johnson, M.D., Juhl, A.R., Lovejoy, C., Lynn, D.H., Marchetti, A., Rice, D.W., Palmer, J.D., 2006. An exceptional horizontal gene transfer in plastids: McManus, G., Nedelcu, A.M., Menden-Deuer, S., Miceli, C., Mock, T., gene replacement by a distant bacterial paralog and evidence that haptophyte Montresor, M., Moran, M.A., Murray, S., Nadathur, G., Nagai, S., Ngam, P.B., and cryptophyte plastids are sisters. BMC Biol. 4, 31. Palenik, B., Pawlowski, J., Petroni, G., Piganeau, G., Posewitz, M.C., Rengefors, K., Roberts, K.R., 1984. Structure and significance of the cryptomonad flagellar Romano, G., Rumpho, M.E., Rynearson, T., Schilling, K.B., Schroeder, D.C., apparatus. I. Cryptomonas ovata (Cryptophyta). J. Phycol. 20, 590–599. Simpson, A.G.B., Slamovits, C.H., Smith, D.R., Smith, G.J., Smith, S.R., Sosik, Roberts, K.R., Stewart, K.D., Mattox, K.R., 1981. The flagellar apparatus of Chilomonas H.M., Stief, P., Theriot, E., Twary, S.N., Umale, P.E., Vaulot, D., Wawrik, B., paramecium (Cryptophyceae) and its comparison with certain zooflagellates. J. Wheeler, G.L., Wilson, W.H., Xu, Y., Zingone, A., Worden, A.Z., 2014. The marine Phycol. 17. microbial eukaryote transcriptome sequencing project (MMETSP): illuminating Roure, B., Philippe, H., 2011. Site-specific time heterogeneity of the substitution the functional diversity of eukaryotic life in the oceans through transcriptome process and its impact on phylogenetic inference. BMC Evol. Biol. 11, 17. sequencing. PLoS Biol. 12, e1001889. Roure, B., Baurain, D., Philippe, H., 2013. Impact of missing data on Kim, E., Archibald, J.M., 2013. Ultrastructure and molecular phylogeny of the phylogenies inferred from empirical phylogenomic data sets. Mol. Biol. Evol. cryptomonad Goniomonas avonlea sp. nov. Protist 164, 160–182. 30, 197–214. Kim, E., Harrison, J.W., Sudek, S., Jones, M.D., Wilcox, H.M., Richards, T.A., et al., Roure, B., Rodríguez-Ezpeleta, N., Philippe, H., 2007. SCaFoS: a tool for selection, 2011. Newly identified and diverse plastid-bearing branch on the eukaryotic concatenation and fusion of sequences for phylogenomics. BMC Evol. Biol. 7 tree of life. Proc. Natl. Acad. Sci. USA 108, 1496–1500. (Suppl. 1), S2. Kimura, M., 1963. The Neutral Theory of Molecular Evolution. Cambridge University Seenivasan, R., Sausen, N., Medlin, L.K., Melkonian, M., 2013. Picomonas judraskeda Press, Cambridge. gen. et sp. nov.: the first identified member of the Picozoa phylum nov., a King, J.L., Jukes, T.H., 1969. Non-Darwinian evolution. Science 164, 788–798. widespread group of picoeukaryotes, formerly known as ’picobiliphytes’. PLoS Klaveness, D., Shalchian-Tabrizi, K., Thomsen, H.A., Eikrem, W., Jakobsen, K.S., 2005. One 8, e59565. Telonema antarcticum sp. nov., a common marine phagotrophic flagellate. Int. J. Shalchian-Tabrizi, K., Eikrem, W., Klaveness, D., Vaulot, D., Minge, M.A., Le Gall, F., Syst. Evol. Microbiol. 55, 2595–2604. et al., 2006. Telonemia, a new protist phylum with affinity to chromist lineages. Lankester, E.R., 1878. Preface to Gegenbaur’s Elements of Anatomy. In: English Proc. Roy. Soc. B 273, 1833–1842. Translation of the 2nd edition. Macmillan, London. Shalchian-Tabrizi, K., Kauserud, H., Massana, R., Klaveness, D., Jakobsen, K.S., 2007. Lankester, E.R., 1885. Protozoa. In: Baynes, T.S. (Ed.), The Encyclopedia Britannica, Analysis of environmental 18S ribosomal RNA sequences reveals unknown 9th ed. J. M. Stoddard Co. Ltd, Philadelphia, pp. 831–865. diversity of the cosmopolitan phylum Telonemia. Protist 158, 173–180. Lasek-Nesselquist, E., Gogarten, J.P., 2013. The effects of model choice and Shih, P.M., Matzke, N.J., 2013. Primary endosymbiosis events date to the later mitigating bias on the ribosomal tree of life. Mol. Phylogenet. Evol. 69, Proterozoic with cross-calibrated phylogenetic dating of duplicated ATPase 17–38. proteins. Proc. Natl. Acad. Sci. USA 110, 12355–12360. Lee, R.E., Kugrens, P., 1991. Katablepharis ovalis, a colourless flagellate with Shih, P.M., Wu, D., Latifi, A., Axen, S.D., Fewer, D.P., Talla, E., Calteau, A., Cai, F., interesting cytological characteristics. J. Phycol. 27, 505–513. Tandeau de Marsac, N., Rippka, R., Herdman, M., Sivonen, K., Coursin, T., Laurent, Leigh, J., Susko, E., Baumgartner, M., Roger, A.J., 2008. Testing congruence in T., Goodwin, L., Nolan, M., Davenport, K.W., Han, C.S., Rubin, E.M., Eisen, J.A., phylogenetic analysis. Syst. Biol. 57, 104–115. Woyke, T., Gugger, M., Kerfeld, C.A., 2013. Improving the coverage of the Lopez, P., Casane, D., Philippe, H., 2003. Heterotachy, an important process of cyanobacterial phylum using diversity-driven genome sequencing. Proc. Natl. protein evolution. Mol. Biol. Evol. 19, 1–7. Acad. Sci. USA 110, 1053–1058. Lukeš, J., Leander, B.S., Keeling, P.J., 2009. Cascades of convergent evolution: the Sierra, R., Matz, M.V., Aglyamova, G., Pillet, L., Decelle, J., Not, F., de Vargas, C., corresponding evolutionary histories of euglenozoans and dinoflagellates. Proc. Pawlowski, J., 2013. Deep relationships of Rhizaria revealed by phylogenomics: Natl. Acad. Sci. USA 106 (Suppl. 1), 9963–9970. a farewell to Haeckel’s Radiolaria. Mol. Phylogenet. Evol. 67, 53–59. Margulis, L., 1970. Origin of Eukaryotic Cells. Yale University Press, New Haven. Simpson, A.G.B., 2003. Cytoskeletal organization, phylogenetic affinities and Margulis, L., 1981. Symbiosis in Cell Evolution. Freeman, San Francisco. systematics in the contentious taxon Excavata (Eukaryota). Int. J. Syst. Evol. Moestrup, Ø., 2000. The flagellate cytoskeleton: introduction of a general Microbiol. 53, 1759–1777. terminology for microtubular roots in protists. In: Leadbeater, B.S., Green, J.C. Smith, R., Patterson, D.J., 1986. An analysis of heliozoan interrelationships: an (Eds.), The Flagellates: Unity, Diversity and Evolution. Taylor & Francis, London, example of the potentials and limitations of ultrastructural approaches to the pp. 69–94. study of protistan phylogeny. Proc. Roy. Soc. Lond. B 227, 325–366. Mylnikov, A.P., Mylnikov, A.A., 2008. The structure of extrusive organelles in Stork, S., Moog, D., Przyborski, J.M., Wilhelmi, I., Zauner, S., Maier, U.G., 2012. alveolate and other heterotrophic flagellates. Tsitologia 50, 406–412. Distribution of the SELMA translocon in secondary plastids of red algal origin Nosenko, T., Schreiber, F., Adamska, M., Adamski, M., Eitel, M., Hammel, J., and predicted uncoupling of ubiquitin-dependent translocation from Maldonado, M., Muller, W.E., Nickel, M., Schierwater, B., Vacelet, J., Wiens, M., degradation. Eukaryot. Cell 11, 1472–1481. Worheide, G., 2013. Deep metazoan phylogeny: when different genes tell Vossbrinck, C.R., Maddox, J.V., Friedman, S., Debrunner-Vossbrinck, B.A., Woese, different stories. Mol. Phylogenet. Evol. 67, 223–233. C.R., 1987. Ribosomal RNA sequence suggests microsporidia are extremely Not, F., Valentin, K., Romari, K., Lovejoy, C., Massana, R., Tobe, K., et al., 2007. ancient eukaryotes. Nature 326, 411–414. Picobiliphytes: a marine picoplanktonic algal group with unknown affinities to Woehle, C., Dagan, T., Martin, W.F., Gould, S.B., 2011. Red and problematic green other eukaryotes. Science 315, 253–255. phylogenetic signals among thousands of nuclear genes from the Okamoto, N., Chantangsi, C., Horak, A., Leander, B.S., Keeling, P.J., 2009. Molecular photosynthetic and apicomplexa-related Chromera velia. Genome Biol. Evol. 3, phylogeny and description of the novel katablepharid Roombia truncata gen. et 1220–1230. sp. nov., and establishment of the Hacrobia taxon nov. PLoS One 4, e7080. Yabuki, A., Chao, E.E., Ishida, K., Cavalier-Smith, T., 2012. Microheliella maris Parfrey, L.W., Grant, J., Tekle, Y.I., Lasek-Nesselquist, E., Morrison, H.G., Sogin, M.L., (Microhelida ord. n.), an ultrastructurally highly distinctive new axopodial Patterson, D.J., Katz, L.A., 2010. Broadly sampled multigene analyses yield a protist species and genus, and the unity of phylum Heliozoa. Protist 163, 356– well-resolved eukaryotic tree of life. Syst. Biol. 59, 518–533. 388. Parke, M., Green, J.C., Manton, I., 1971. Observations on the fine structure of zoids of Yabuki, A., Eikrem, W., Takishita, K., Patterson, D.J., 2013. Fine structure of Telonema the genus Phaeocystis [Haptophyceae]. J. Mar. Biol. Assoc. UK 51, 927–941. subtilis Griessmann, 1913: a flagellate with a unique cytoskeletal structure Patron, N.J., Inagaki, Y., Keeling, P.J., 2007. Multiple gene phylogenies support the among eukaryotes. Protist 164, 556–569. monophyly of cryptomonad and haptophyte host lineages. Curr. Biol. 17, Yabuki, A., Inagaki, Y., Ishida, K., 2010. Palpitomonas bilix gen. et sp. nov.: a novel 887–891. deep-branching heterotroph possibly related to Archaeplastida or Hacrobia. Perasso, L., Hill, D.R.A., Wetherbee, R., 1992. Transformation and development of the Protist 161, 523–528. flagellar apparatus of Cryptomonas ovata (Cryptophyceae) during cell division. Yabuki, A., Kamikawa, A., Ishikawa, S.A., Kolisko, M., Kim, E., Tanabe, A.S., Kume, K., Protoplasma 170, 53–67. Ishida, K., Inagaki, Y., 2014. Palpitomonas bilix represents a basal cryptist 362 T. Cavalier-Smith et al. / Molecular Phylogenetics and Evolution 93 (2015) 331–362

lineage: insight into the character evolution in Cryptista. Sci. Rep. 4, 4641. of Hyalolithus neolepis gen. et sp. nov. (Prymnesiophyceae, Haptophyta). Protist http://dx.doi.org/10.1038/srep04641. 157, 213–234. Yang, Z., 2007. Fair-balance paradox, star-tree paradox, and Bayesian phylogenetics. Zhao, S., Burki, F., Bråte, J., Keeling, P.J., Klaveness, D., Shalchian-Tabrizi, K., 2012. Mol. Biol. Evol. 24, 1639–1655. Collodictyon – an ancient lineage in the tree of eukaryotes. Mol. Biol. Evol. 29, Yoon, H.S., Hackett, J.D., Pinto, G., Bhattacharya, D., 2002. The single, ancient origin 1557–1568. of chromist plastids. Proc. Natl. Acad. Sci. USA 99, 15507–15512. Zhao, S., Shalchian-Tabrizi, K., Klaveness, D., 2013. Sulcozoa revealed as a Yoon, H.S., Price, D.C., Stepanauskas, R., Rajah, V.D., Sieracki, M.E., Wilson, W.H., paraphyletic group in mitochondrial phylogenomics. Mol. Phylogenet. Evol. et al., 2011. Single-cell genomics reveals organismal interactions in 69, 462–468. uncultivated marine protists. Science 332, 714–717. Zhang, Z., Green, B.R., Cavalier-Smith, T., 1999. Single gene circles in dinoﬂagellate Yoshida, M., Noël, M.H., Nakayama, T., Naganuma, T., Inouye, I., 2006. A chloroplast genomes. Nature 400, 155–159. haptophyte bearing siliceous scales: ultrastructure and phylogenetic position