187-gene phylogeny of protozoan phylum Amoebozoa reveals a new class (Cutosea) of deep-branching, ultrastructurally unique, enveloped marine Lobosa and clarifies amoeba evolution.

Monophyly of protozoan phylum Amoebozoa, and subdivision into subphyla Conosa and Lobosa each with different cytoskeletons, are well established. However early diversification of non-ciliate lobose amoebae (Lobosa) is poorly understood. To clarify it we used recently available transcriptomes to construct a 187-gene amoebozoan tree for 30 species, the most comprehensive yet. This robustly places new genus Atrichosa (formerly lumped with Trichosphaerium) within lobosan class Tubulinea, not Discosea as previously supposed. We identified an earliest diverging lobosan clade comprising marine amoebae armoured by porose scaliform cell-envelopes, here made a novel class Cutosea with two pseudopodially distinct new families. Cutosea comprise Sapocribrum, ATCC PRA-29 misidentified as 'Pessonella', plus from other evidence Squamamoeba. We confirm that Acanthamoeba and ATCC 50982 misidentified as Stereomyxa ramosa are closely related. Discosea have a strongly supported major subclade comprising Thecamoebida plus Glycostylida (suborders Dactylopodina, Stygamoebina; Vannellina) phylogenetically distinct from Centramoebida. Stygamoeba is sister to Dactylopodina. Himatismenida are either sister to Centramoebida or deeper branching. Discosea usually appear holophyletic (rarely paraphyletic). Paramoeba transcriptomes include prokinetoplastid Perkinsela-like endosymbiont sequences. Cunea, misidentified as Mayorella, is closer to Paramoeba than Vexillifera within holophyletic Dactylopodina. Taxon-rich site-heterogeneous rDNA trees confirm cutosan distinctiveness, allow improved conosan taxonomy, and reveal previous dictyostelid tree misrooting.


Introduction
Amoebozoa are one of three radically different major groups of amoeboid eukaryotes that had independent evolutionary origins from different flagellate ancestors: each belongs in a different one of the three eukaryote supergroups now recognised -Amoebozoa in scotokaryotes, Rhizaria in corticates, and Percolozoa in Eozoa (Cavalier-Smith et al., 2015a). Abounding in soil and all aquatic habitats, as well as including parasites of animals, and both aerobes and anaerobes, Amoebozoa have over a thousand species. Their higher classification was confused for two centuries until electron microscopy revealed some key features, whose importance sequence phylogeny confirmed Smirnov et al., 2011). The purely amoeboid, entirely non-ciliate amoebozoan subphylum Lobosa (e.g. Amoeba, Acanthamoeba, Hartmannella; colloquially lobose amoebae) are the main focus of this paper. They are characterised by locomotion by generally non-eruptive, broadly lobed pseudopodia and by having often branched tubular mitochondrial cristae (Cavalier-Smith, 1998). Apart from Acanthamoeba, sometimes a human pathogen, Echinamoeba, and Vexillifera, Lobosa lack pointed subpseudopodia. Their sister subphylum Conosa is pseudopodially and mitochondrially more diverse, tending to have pointed pseudopodia or subpseudopodia and one or more cilia at least in some life cycle stages and a characteristic conose microtubular cytoskeleton. Conosa include cellular and plasmodial slime moulds, varipodid amoebae with branching pseudopodia, Multicilia, flagellates like Phalansterium, plus anaerobes like Entamoeba and Mastigamoeba with radically modified mitochondria. Multigene trees (Cavalier-Smith et al., 2015b) now clearly support the monophyly of Amoebozoa and this primary subdivision into two subphyla.
All Lobosa have smooth broad, rounded pseudopodia of highly varied forms, except for one remarkable recently discovered genus Sapocribrum whose tiny cells have filose pseudopodia (Lahr et al., 2015), which we show here to be unusually deep branching and of special evolutionary significance. Thread-like extensions can also be produced from a broader more typically lobose region in Parvamoeba (Cole et al., 2010;Kudryavtsev et al., 2011). Lobose amoebae with broad pseudopodia include those with shells or tests (testate amoebae, comprising probably unrelated orders -freshwater Arcellinida and marine Trichosida) and more diverse naked species called gymnamoebae (Page, 1987(Page, , 1988) that lack tests.
Largely in the past decade, naked amoebae (not as shapeless as often supposed) have been successfully grouped by their different characteristic pseudopodial forms during active locomotion (e.g. flattened, tubular, monopodial, multipodial, digitiform, fanshaped, conical) combined with ultrastructural variations in their cell surface coats (e.g. thick, thin, scales, glycostyles) into 12 distinctive orders, most of whose monophyly is well corroborated by 18S ribosomal DNA trees Smirnov et al., 2005Smirnov et al., , 2011. Four naked orders are grouped with arcellinids as class Tubulinea, united by tubular pseudopodia with monoaxial internal cytoplasmic flow. The other eight have flattened cells that by contrast exhibit multiaxial cytoplasmic flow or flow without a pronounced axis, and constitute class Discosea that at present also includes the unique and least well understood Trichosida.
Trichosida are phylogenetically obscure large multinucleate marine amoebae that uniquely have tests with numerous pores through which hair-like non-locomotory pseudopods emerge. This multiporose test and their having both locomotory lobopodia and thin (perhaps sensory) dactylopodia set Trichosida apart from all other Amoebozoa (Schuster, 1976;Angell, 1975Angell, , 1976, so much so they were sometimes thought to be related to Foraminifera instead (Loeblich and Tappan, 1964). The sole available 18S rDNA sequence from 'Trichosphaerium' sp. seemingly confirmed that Trichosida are amoebozoans, but is so divergent from all others that it grouped in Conosa with the long-branch myxogastrid slime moulds. Actin plus rDNA jointly contradictorily put it within Lobosa with Thecamoeba similis (Tekle et al., 2008), where trichosids have been tentatively classified , though a six-gene tree did put it in Conosa within slime moulds (Lahr et al., 2015). We argue here that this strain was misidentified and is not a Trichosphaerium, though is probably related, so establish a new genus Atrichosa for it.
Multigene trees support this subdivision of Lobosa into two monophyletic classes (Tubulinea and Discosea). However, even the broadest study to date using 17 Amoebozoa sampled relatively few orders; though resolution at the base of Discosea was weak because of a bush-like basal radiation and many orders being only singly represented by long unbroken branches, the trees hinted that its current classification into subclasses may be incorrect (Cavalier-Smith et al., 2015b). A parallel study provided partial transcriptome sequences for additional orders and two unclassified or unidentified strains (Grant and Katz, 2014); but their tree did not clarify amoebozoan relationships at all as it included only 13 Amoebozoa, used maximum likelihood only (combining 18S rDNA with 238 proteins, yet oddly using only 15,650 charactersroughly 60 amino acids per protein), and has some bizarre features compared with other published trees.
Therefore we now combine sequences from both studies and from Eme et al. (2014) and others now available in a more extensive analysis of 30 Amoebozoa using 187 concatenated protein genes (using 50,964 amino acids), now including most lobosan orders (8 of 11) plus five of the 16 conosan orders recognised here, with multiple representations of many of them. This more representative taxon sampling yields trees fully consistent with earlier ones (Cavalier-Smith et al., 2015b) but much stabler, yielding several new clear conclusions. This work was not straightforward since, in addition to some new transcriptomes being from then unidentified strains, it transpired that several were misidentified and three comprised obvious mixtures of two different eukaryotes. These mixtures (Table 1) were revealed by the 187 single-gene trees that we routinely carry out to check purity and paralogue uniformity. They showed that both Paramoeba transcriptomes contain numerous genes also from obligate mutually related prokinetoplastid endosymbionts (parasomes, related to Perkinsela; Young et al., 2014) that together form a robust clade within Euglenozoa as sister to Metakinetoplastina. The other mixed culture was Stygamoeba regulata, a marine amoeba that uniquely for Lobosa has flat mitochondrial cristae (Smirnov, 1995/6) so has been of unclear phylogenetic position but was recently grouped with Vermistella as a separate order Stygamoebida within subclass Flabellinia of Discosea . Its transcriptome was heavily contaminated by genes from Cunea ('Mayorella' sp.), perhaps why Grant and Katz (2014) did not include Stygamoeba on their multiprotein tree. By phylogenetically separating Stygamoeba and Cunea sequences using the single-gene trees we were able to include Stygamoeba on multiprotein trees for the first time and demonstrate that it is sister to Vannellida/Dactylopodida, Table 1 Amoeba names and strains with MMETSP transcriptomes (Keeling et al., 2014)  supporting all three taxa being now included in order Glycostylida . Misidentification of strain 'Mayorella sp.' (Grant and Katz, 2014) was confirmed by separate comprehensive 18S rDNA analyses of 227 Amoebozoa showing that it must be an undescribed species of Cunea (a paramoebid discosean recently discovered by Kudryavtsev and Pawlowski (2015)) that groups within order Dactylopodina as sister to other Paramoebidae, and is not a Mayorella. This rDNA analysis is the most comprehensive for Amoebozoa and first to use a site-heterogeneous nucleotide substitution model, generally accepted as superior (Lartillot et al., 2007;Brown et al., 2013) to site-homogeneous methods hitherto used for rDNA, and was also better in practice for particularly difficult groups where rDNA evolutionary rates vary dramatically as in Amoebozoa (Cavalier-Smith, 2015, where it yielded better trees for the even more challenging gregarines and Percolozoa). We critically compare the taxonomically more broadly sampled 18S rDNA trees with the genically far better sampled multiprotein trees.
Our most novel conclusion is that the two unidentified strains of Grant and Katz (2014), one now named Sapocribrum chincoteaguense (Lahr et al., 2015), form a clade distinct from both Tubulinea and Discosea, here made a third lobosan class, Cutosea, because of its unique test with scale-like substructure. Cutosea must also include Squamamoeba (Kudryavtsev and Pawlowski, 2012) and constitute a novel type of enveloped lobosan. Cutosea are sister to Tubulinea plus Discosea (the two latter here united as new superclass Glycopoda) and so the earliest lobosan lineage. Next in broad significance concerns Atrichosa (formerly 'Trichosphaerium' sp.) whose rDNA diverges so extremely from all other Amoebozoa that it could not be clearly placed in Lobosa or Conosa (Tekle et al., 2008;Lahr et al., 2015). Earlier it had been either provisionally excluded from Amoebozoa or if included thought to be related to Tubulinea  or after rDNA showed it to be amoebozoan guessed to belong in Discosea . Atrichosa is unambiguously a deep branching member of Tubulinea, probably closer to the Nolandina/ Amoebina (euamoebid) clade than to Echinamoebida. Thirdly, we show that Glycostylida are holophyletic and sister to Thecamoebida, casting strong doubt on existing classification of Discosea into subclasses . We discuss the evolutionary and taxonomic significance of these and other phylogenetic discoveries and present a revised higher-level classification for Amoebozoa.

Materials and methods
Starting from the 187 gene alignments asembled by Cavalier-Smith et al. (2015a,b) we manually (using macgde http://macgde. bio.cmich.edu/) added sequences of Copromyxa protea (family Hartmannellidae, order Euamoebida) from Eme et al. (2014) as well as many obtained by blasting against our alignments 11 Amoebozoa transcriptomes downloaded from the Marine Microbial Eukaryote Transcriptome database (http://marinemicroeukaryotes.org/resources; this website is now discontinued but the data slightly less conveniently are in http://data.imicrobe. us/project/view/104 as 'Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP)'). To check the purity of the downloaded transcriptomes we ran RAxML-MPI v.7.2.8 PROTGAM-MALGF trees (4 gamma rates) with 100 fast bootstrap resamplings for all 187 genes prior to concatenating them using SCaFoS (Roure et al., 2007). This showed that three MMETSP transcriptomes were mixtures of two different eukaryotes, whose proteins we separated phylogenetically using the 187 single-gene trees, so could include both in our analyses. As we judged three cultures to be misidentified (one now recognised to have been by the original authors who have named it Sapocribrum chincoteaguense (Lahr et al., 2015)), Table 1 summarises the original names in MMETSP and those used here and also the nature of the detected mixtures.
Sequences for Stygamoeba regulata (strain ATCC 50892; MMETSP0447) were heavily contaminated by sequences from 'Mayorella' sp. (MMETSP0417; ? strain ATCC 50980, unstated on website). This contamination was evident because for many genes our single-gene trees revealed two completely different sequences from the 'Stygamoeba' transcriptome, one identical to that for 'Mayorella' sp. and branching with Paramoeba within Dactylopodina and one greatly different and branching much more deeply. We removed all sequences identical to 'Mayorella' and treated the residue as authentic Stygamoeba regulata genes, enabling us to include this species on multigene trees. Nerad isolated both cultures from salt marsh sediment, Hog Island, eastern shore of Virginia, 2001. We suspect that both came from the same sample and only the 'Mayorella' culture was fully purified, the Stygamoeba one retaining unnoticed sparse 'Mayorella' contaminants that multiplied greatly before cultures were frozen. Consistently with that inference, the imicrobe and ATCC websites oddly give the same strain identifier (BSH-02190019) for both 'Mayorella' sp. and Stygamoeba regulata, but different 18S rDNA sequences, which implies that some ATCC information is inaccurate. The 18S rDNA sequence shown for Stygamoeba is the published one (Tekle et al., 2008). Our comprehensive new amoebozoan 18S tree shows that 'Mayorella' is certainly not a Mayorella sequence (see Section 3.1), but is evidently an undescribed species of the recently discovered paramoebid genus Cunea (Kudryavtsev and Pawlowski, 2015).
We found for most genes from the type strain of Paramoeba atlantica (CCAP 1560/9) and Paramoeba (=Neoparamoeba) aestuarina transcriptomes (MMETSP0151_2 and MMETSP0161_2 respectively) that there were two even more radically different sequences. These divergent second sequences grouped within Euglenozoa, as sister to metakinetoplastids (bodonids and trypanosomatids), invariably with 100% support for this Kinetoplastea clade, and if available for both amoebae grouped together as a maximally supported clade. Clearly these euglenozoan sequences came from the parasome, an obligate prokinetoplastid endosymbiont of Paramoeba related to Ichthyobodo and Perkinsela, a long recognised shared character for Paramoeba/Neoparamoeba (Young et al., 2014). As prokinetoplastids (Ichthyobodonidae) have not yet been placed on multigene trees we relabelled these prokinetoplastid sequences 'ichthyobodonid from Paramoeba atlantica' and 'ichthyobodonid from Paramoeba aestuarina' and included them in the multigene analysis as well as the two host Paramoeba sequences that invariably grouped within Lobosa, usually together. Because of the huge sequence distance between the taxa there was never even the slightest doubt which was a prokinetoplastid endosymbiont gene and which a host Paramoeba gene. As our gene selection protocol picked only the top hit against our alignments, sampling each member of the mixed transcriptomes was necessarily incomplete. That is probably why gene sampling was substantially lower for Stygamoeba than other MMETSP Amoebozoa, why those for Paramoeba somewhat lower, and those for the highly divergent parasome genes immensely lower. More parasome genes should therefore be findable in these transcriptomes.
Transcriptome MMETSP0437 from strain ATCC 50979 was labelled Sexangularia on the website; however, as Lahr et al. (2011) explained, this strain was misidentified and is not Sexangularia but a non-testate then unnamed species (Grant and Katz, 2014 called it Eukaryota sp. JRG-2011); in error their illustration was not of ATCC 50979 but of another unknown strain having finger-like pseudopodia, unlike ATCC 50979. Lahr et al. (2015) formally described ATCC 50979 as Sapocribrum chincoteaguense, so we use this name instead of wrong Sexangularia.
Transcriptome MMETSP0420 from strain ATCC PRA-29 was labelled Pessonella sp. on the websites and by Grant and Katz (2014), but as Kudryavtsev and Pawlowski (2013) explain, neither the only described Pessonella (Pussard, 1973) nor PRA-29 is characterised ultrastructurally, so it may not be a Pessonella; the rDNA tree of Kudryavtsev and Pawlowski (2013) strongly indicates that it is related to Squamamoeba, till now not assigned to a family or order, so following Lahr et al. (2015) we place 'Pessonella' in inverted commas to stress this uncertainty.
Multigene phylogenetic analysis used 50,964 amino acids and the best available site-heterogeneous amino-acid substitution model (PhyloBayes-MPI v.1b GTR-CAT-C-4rates: (Lartillot and Philippe, 2004;Lartillot et al., 2013) using two separate chains; trees were also run by maximum likelihood (ML) using the best substitution model (LG: Le and Gascuel, 2008) available for RAxML-MPI v.7.2.8 PROTGAMMALGF (Stamatakis et al., 2005) (4 gamma rates) and 100 fast bootstrap resamplings (Stamatakis et al., 2008), even though this model is site-homogeneous and thus less evolutionarily realistic than GTR-CAT-C (henceforth called CAT for brevity) (Lartillot and Philippe, 2004). To place the prokinetoplastids and root the amoebozoan subtree, we ran a eukaryotewide analysis for 121 eukaryotes (excluding chromists to avoid possible distortion by unrecognised red algal paralogues; see Deschamps and Moreira, 2009;Cavalier-Smith et al., 2015a). To avoid perturbing amoebozoan branching by distant outgroups and make convergence of PhyloBayes trees easier we ran trees restricted to the 30 Amoebozoa (plus 29-taxon trees omitting the very weakly sampled Vannella simplex to check that its large number of missing genes did not distort the tree).

Results and discussion
3.1. Culture MMETSP0417 is a novel Cunea species not 'Mayorella' sp Fig. 1, the first comprehensive site-heterogeneous 18S rDNA tree for Amoebozoa, shows that the strain yielding transcriptome MMETSP0417 and named 'Mayorella' sp. (Grant and Katz, 2014) has rDNA extremely similar to those of the closely related Cunea thuwala and C. profundata, being only about twice as divergent from them as they are from each other. Clearly it is a third undescribed Cunea species misidentified as Mayorella MMETSP0417 (they gave no microscopic evidence for its identity). The two known Mayorella branch entirely separately as sister to Dermamoeba plus Paradermamoeba and a related environmental rDNA lineage. Cunea is sister to the clade comprising Paramoeba, Korotnevella, and Pseudoparamoeba, and this clade (Paramoebidae) sister to Vexilliferidae (thus forming clade Dactylopodina) with moderate support (0.82) on this site-heterogeneous tree, in total harmony with all our multiprotein trees. This tree also shows that ATCC strain PRA-29, whose transcriptome is included in our analyses (as 'Pessonella', possibly misidentified) is robustly sister to Squamamoeba; this maximally supported clade appears the deepest branching in all Amoebozoa. However, this and all other deep branch positions within Amoebozoa are insignificantly supported and often contradictory between CAT and ML. This emphasises the necessity of multiprotein trees for deep phylogeny of Amoebozoa. We shall discuss other novel evolutionarily significant features of this tree (Section 3.6) after explaining the more robust multiprotein trees that are our central focus. For consistency, in all trees clade names are as in the taxonomy section (Section 3.7), a few new or differing slightly in circumscription from past usage.
3.2. Eukaryote-wide 187-protein trees reveal a novel major amoebozoan clade and that Paramoeba transcriptomes contain related Perkinsela-like endosymbionts The 121-taxon eukaryote-wide multiprotein tree (Fig. 2) shows in both CAT and LG trees a robust new lobosan clade (here called Cutosea) comprising Sapocribrum and PRA-29 probably misidentified as 'Pessonella'. Cutosea are the deepest branching Lobosa on the consensus CAT tree, sister to the previously known lobosan classes Discosea and Tubulinea, all three maximally supported clades by CAT. However Cutosea were in this position on only one of the two PhyloBayes chains; the other put them within Conosa as sister to Variosea (1.0) and LG ML placed them weakly (60%) as sister to Tubulinea. Near the base of scotokaryotes the position of the sulcozoan planomonads was also unstable and contradictory between the chains, so four nodes (red in Fig. 2) have weak support because of these two non-convergences in topology. Both conflicts persisted after running chains much longer (burnin 7767; 22,919 trees summed), though 0.98 support for excluding Cutosea from Discosea and Discosea/Tubulinea increased to 1 and 0.99. CAT topology in the rest of the tree was identical in both chains and congruent with other recent multigene studies except for non-grouping of the long-branch parabasalian Histomonas with the other metamonads with which it groups (thick arrow) with insignificant (21%) support on the LG tree. Fig. 2 confirms our inference from single-gene trees that the second highly divergent sequence found for most of the 187 genes in the Paramoeba atlantica and P. aestuarina transcriptomes is from the Ichthyobodo-related Perkinsela-like prokinetoplastid endosymbiont constituting the parasome. As Fig. 2 shows, both Perkinsela-like ichthyobodonid sequences group together with maximal support and this clade is within Euglenozoa as the maximally supported sister to metakinetoplastids, consistent with the strict congruence of the phylogeny of Perkinsela-like symbionts with their paramoebid hosts (Young et al., 2014).
This tree by both methods also shows that Amoebozoa are monophyletic for all 29 included species and divide into two clades, Lobosa and Conosa. With this very large taxon sample the two chains did not fully converge (maxdiff 1); the very distant outgroups might have distorted its internal phylogeny, so it is safest to use this tree simply to confirm the relatedness, distinctivess, and euglenozoan nature of the second gene set of the endosymbionts of Paramoeba, and to root the smaller Amoebozoa-only trees -the 30-taxon trees did converge.

Amoebozoa-only 187-protein trees confirm distinctiveness of Cutosea
The 29-taxon Amoebozoa-only tree ( Fig. 3) with the same amoebozoan taxa as Fig. 2 had exactly the same topology, but also did not converge (maxdiff 1), though the nature of the nonconvergence differed. In contrast to Fig. 2 the bipartition between Lobosa and Conosa was maximally supported by both chains and significantly supported by ML also. In both Cutosea are maximally . Two chains were run, which converged on the same topology (max diff. 0.222); 121, 286 trees were summed after removing the first 20% as burnin. Support values are posterior probabilities (left) and bootstrap percentages for 1000 resamplings for the corresponding ML GTR-C tree (right); black blobs signify maximal support by both methods (1.0, 100%). Taxa whose transcriptomes are in our multigene trees are in bold. Atrichosa algivora (='Trichosphaerium' sp.) was omitted as its 18S rDNA sequence is so divergent that it cannot be aligned for many of the 1470 nucleotide positions chosen for this analysis, making its position on previously published 18S rDNA trees largely meaningless. To reduce long-branch artefacts further, myxogastrid Mycetozoa (all with extremely longbranches) were omitted as were the longest branch vannellids (Clydonella, Ripella) and Archamoebae (Entamoeba, Pelomyxa); all 35 omitted longer-branch taxa except Atrichosa, Clydonella, and Ripella are present in a 300-taxon analysis with 262 Amoebozoa (Supplementary Fig. S1) that shows species names for all clades collapsed here to enable the tree to fit one page. Number of taxa in each collapsed lobosan branch is to the right of its name. supported as the deepest branch in Lobosa. The corresponding ML tree also excluded Cutosea from Discosea but with only weak (63%) support and placed it insignificantly (43%) as sister to Tubulinea. Both methods had maximal support for Sapocribrum and PRA-29 being sisters.
Most branches on the CAT tree had maximal (one only 0.99) support except for the base of Discosea, which was a maximally suported clade with Himatismenida sister to Centramoebida (0.76) in one chain but in the other was paraphyletic with Himatismenida most deeply and Tubulinea sister to Centramoebida, contrary to Fig. 2. Atrichosa was weakly sister to Nolandella/ Copromyxa (0.6; strongly 95% by ML) but maximally supported as a tubulinean by both methods. The monophyly of Glycostylida as here revised (Table 2), their internal branching order, and also their being sister to Stygamoeba were all maximally supported by CAT as in Fig. 2 and always maximally or well supported also by ML on both trees. Vannella simplex (most genes unsequenced) was excluded from Figs. 2 and 3, as excessive missing data can distort trees (Roure et al., 2013 2. PhyloBayes CAT-GTR-C consensus tree for 121 eukaryotes based on 50, 856 amino acid positions in 187 protein-coding genes. As most bipartitions were maximally supported by both methods, support values are not shown for those with maximal support both for this tree (posterior probability 1.0) and the corresponding LG ML tree for the same alignment (100%). The number of amino acids included for each taxon follows its name. Chromists were excluded because they are eukaryote-eukaryote chimaeras often having plastids of red algal origin, so unrecognised nuclear red algal genes of symbiogenetic origin present in chromist alignments could distort topology if both red algae and chromists are in the same tree (see Moreira, 2009 andCavalier-Smith et al., 2015a). As the text explains, the CAT tree did not fully converge after 12,318 cycles, because the positions of Cutosea and Planomonadida differed in the two chains, one differing slightly from this consensus tree, which lowered CAT support values for the four nodes in red; topology was identical for both chains in the rest of the tree; 12,558 trees summed after removing the first 6098 as burnin.
branch in Lobosa with maximal support by both CAT and ML (Fig. 4). Adding V. simplex did not change the branching order of Glycostylida or reduce CAT maximal support for its monophyly and branching order and being sister to Stenamoeba; and had minimal effects on the non-maximal ML support values. Atrichosa remained in the same position within Tubulinea with similar support. In fact including Vannella simplex made tree topology identical by CAT and ML, unlike Figs. 2 and 3, but with exactly the same topology as in Figs. 2 and 3 CAT trees. This congruence may be an example of breaking long branches being more important than excluding taxa with missing data (Roure et al., 2013). Figs. 2 and 3 ML disagreed with CAT, placing Cutosea weakly as sister to Discosea plus Tubulinea (Fig. 2 60%; Fig. 3 0.43% support). As CAT never showed that with any taxon sample, 18S rDNA never showed it, and there is no morphological reason to suspect such a relationship, we attribute this discrepancy to the sitehomogeneous LG model being less realistic than CAT, which gave a more consistent topology amongst taxon samples.
Taking Figs. 1-4 together we conclude that Sapocribrum, PRA-29, and Squamamoeba all belong to the same deepbranching lobosan clade, Cutosea. Unfortunately this cannot yet be directly shown on a single tree as no 18S rDNA sequence is available for Sapocribrum. Sapocribrum has filose pseudopodia unlike all Discosea, yet was nonetheless previously assumed to be in Discosea (Lahr et al., 2015); all our multiprotein trees strongly exclude Cutosea from Discosea, in keeping with their highly distinctive pseudopodia. Addition of genically sparse Vannella simplex does not distort the tree or affect the exclusion from the Thecamoebida/Glycostylida clade of Cutosea, which contrary to some published 18S rDNA trees never groups with Glycostylida, always branching three nodes lower.
In earlier 18S rDNA ML trees the Squamamoeba/PRA-29 clade was sister to Vexillifera with insignificant support (Kudryavtsev and Pawlowski, 2013). In homogeneous Bayesian rDNA trees a Squamamoeba/PRA-29/Vexillifera clade was absent (Kudryavtsev and Pawlowski, 2013). The original 18S rDNA grouping of PRA-29 with no support with Vexillifera minutissima (Tekle et al., 2008, wrongly labelled as Vanellidae [sic] and misgrouping with them) is an obvious long-branch artefact. So also was the placement of both a PRA-29/Sapocribrum moderately supported clade and 'Trichosphaerium' with V. minutissima within Dactylopodida (Lahr et al., 2011); actin separately grouped PRA-29 and Sapocribrum with moderate support in a completely unresolved position (Lahr et al., 2011). In a concatenated 18S rDNA and actin ML tree clade Cutosea was moderately supported but there was no support for its apparent position within an incorrectly paraphyletic Dactylopodida (Lahr et al., 2015). On an ostensibly four-gene tree devoid of glycostylids (in which PRA-29 was probably represented by only those two genes) PRA-29 was within Conosa (Tekle et al., 2008). The almost total lack of basal amoebozoan resolution of such 1-2 gene trees caused the phyletic distinctiveness of Cutosea to be previously entirely overlooked.
The 239-gene ML tree of Grant and Katz (2014)   . PhyloBayes GTR-CAT-C tree for 29 Amoebozoa using 187 genes (50, 964 amino acids), excluding genically poorly sampled Vannella simplex. Most bipartitions were maximally supported by both methods, so support values are not shown for those with maximal support both for this tree (posterior probability 1.0) and the LG ML tree for the same alignment (100%). The number of amino acids included for each taxon follows its name. The two chains did not converge because of conflict in the positions of Himatismenida and Tubulinea only (maxdiff 1; 46,919 trees summed after removing first 6090 as burnin; if burnin was 10,000 or 15,000 instead, topologies were identical and support values almost exactly the same). One chain with the same topology as this consensus tree had maximal support for holophyly of Discosea; the other put Tubulinea within Discosea with 0.99 support as sister to Centramoebida, displacing Himatismenida so it became sister to all other Discosea plus Tubulinea (0.98 support).

Table 2
Revised classification of phylum Amoebozoa and its two subphyla, seven classes and 28 orders.
maximally supported as a clade on the 151-gene LG ML tree of Katz and Grant (2015) but wrongly placed (no support given, therefore according to their figure legend <70%) as sister to the also weakly supported artefactual 'Trichosphaerium' [Atrichosa]/'Mayorella' [Cunea] sp. clade criticised above. On that ill-sampled tree neither Lobosa nor Conosa is a clade; Lobosa and Conosa are also mixed up on their supplementary non-converged 150-gene 235-taxon CAT tree where Cutosea was wrongly sister to Mycetozoa plus Archamoebae. The 207-gene ML tree of Katz and Grant (2015) was seriously distorted at the base of scotokaryotes by longbranch attraction of anaerobic protozoan clades to distant bacterial outgroups; weakly sampled Amoebozoa were wrongly shown as polyphyletic, branching in three different places; nonetheless Cutosea were a maximally supported clade, insignificantly (46%) sister to Filamoeba. Neither Grant and Katz (2014) nor Katz and Grant (2015) specified the genes used, but as the 150-gene analysis included 34,991 amino acids whereas the 239-gene analysis had only 15,650 the two sets were presumably very different despite assembly by the same pipeline. Taken together, our taxonomically much better sampled, genically better sampled (187 genes, 50,964 amino acids) and more thoroughly analysed trees strongly contradict all theirs by confidently showing Cutosea to be the most divergent lobosan lineage of all. Previously 18S rDNA trees grouped PRA-29 'Pessonella' sp. with Squamamoeba japonica with maximal support (Kudryavtsev and Pawlowski, 2013 Smirnov et al. (2011) incorrectly gave the family as Trichosidae Möbius, 1889, which is invalid as not directly derived from a genus. b It is unnecessary to provide new diagnoses for these three new suborders as they are simply reduced slightly in rank from the original orders, whose full diagnoses in the two cited papers are still correct for the suborders. c Position needs confirming by sequencing. d (Ptáčková et al., 2013). Its short ciliary transition zone suggests an affinity with clade B, e.g. M. simplex, its cone microtubules are more like those of clade A, e.g. M. punctachora, but it has a unique triangular microtubule nucleating centre below the centriole, suggesting it might be a third clade (transition zone differences previously suggested two distinct genera, one possibly being Mastigina: Cavalier-Smith, 2013). If M. aspera does not belong in clade A with Phreatamoeba balamuthi then Phreatamoeba could be a valid name for that whole clade, making it premature to have changed its name to Mastigamoeba (Simpson et al., 1997), especially as M. aspera is considered a junior synonym of Dinamoeba mirabilis, invalidating names Mastigamoeba and Mastigamoebidae retained here as an application for conservation is apparently being made (Ptáčková et al., 2013). f Hyalodiscus Hertwig and Lesser, 1874, a fan-shaped amoeba that Page (1988) thought possibly related to Flamella, was left in Discosea incertae sedis by Smirnov et al.
(2011) despite doubting (because of its unique rolling motion) that it was an amoebozoan, but is now reasonably classified in Vampyrellida in the rhizarian subphylum Endomyxa (Hess et al., 2012). Confusingly, in botanical nomenclature Hyalodiscus Ehrenberg, 1845 is a diatom. g As Cavalier-Smith (2013) noted, attributing this order to Frenzel, 1892 is probably incorrect. h As now transferred to Thecamoebida (see Section 5) and Vermistella does not group with Stygamoeba on Fig. 1 grouped this clade weakly with the himatismenid Cochliopodiidae, but ML put it as sister to the dactylopodid Vexillifera with no bootstrap support. Though no rDNA sequence is available for Sapocribrum and no proteins for Squamamoeba, an actin/18S rDNA ML tree with missing data grouped Sapocribrum as sister to PRA-29 'Pesssonella' sp. plus Squamamoeba (72% BS) and a six-gene ML tree (including only rDNA for Squamamoeba) put Squamamoeba as sister to Sapocribrum/PRA-29 'Pessonella' sp. with maximal support (Lahr et al., 2015). Thus Squamamoeba is a robust member of the Sapocribrum/PRA-29 cutosean clade and we can safely use ultrastructural features of both Sapocribrum and Squamamoeba for defining morphologically this newly recognised major lobosan group, even though such oligogenic trees did not clarify its position: the two-gene tree put it as sister to Vexillifera with no support and the six-gene tree with Cochliopodium also with no support, but neither tree had any basal resolution.

Phylogeny and probable holophyly of Discosea on Amoebozoaonly 187-protein trees
Irrespective of whether Vannella simplex is included (Fig. 4) or excluded (Fig. 3) Cunea is sister to Paramoeba and this clade sister to Vexillifera, forming a Dactylopodina clade with maximal support by both methods for every branch in the clade. Likewise Stygamoeba is sister to Dactylopodina with maximal support by CAT and 92/93% support by ML (1/96% in Fig. 2). This clade is sister to Vannellida with maximal CAT and moderate (71/67%) ML support; the joint clade corresponds to order Glycostylida as revised below and is sister to Stenamoeba (Thecamoebida) with maximal CAT and high (98/99%) ML support (100% in Fig. 2). Thus all our multigene trees strongly support Stenamoeba being sister to Glycostylida, contradicting the previous subdivision of Discosea into subclasses. Stenamoeba should have been sister to Centramoebida were Longamoebia (i.e. Dermamoebida, Thecamoebida and Centramoebida) a clade . We therefore do not use subclass Longamoebia in our revised classification (Table 2). Both methods equally strongly exclude Himatismenida from the Glycostylida clade showing that including them in a broadened Flabellinia  was incorrect so we do not use subclass Flabellinia either in Table 2. Multiprotein trees for more Discosea should enable a sounder subclass division.
Discosea themselves were a weakly supported clade on all ML trees; though maximally supported on two CAT trees (Figs. 2 and 4) in Fig. 3 that was only true for one chain, the other showing Discosea as paraphyletic as noted above. More likely than not Discosea are a clade, but this single discrepancy makes more extensive taxon sampling, especially of Dermamoebida, Parvamoeba, and Pellitina, vital to decide if Discosea are a clade or ancestral to Tubulinea. It is also uncertain whether Himatismenida are sister to Centramoebida (Figs. 2 and 4, and one Fig. 3 chain; we informally call this potential clade Centramoebia as both orders have centrosomes) or all other Discosea plus Tubulinea (only shown on the other Fig. 3 chain).
All trees agree in putting 'Stereomyxa ramosa' as sister to Acanthamoeba with maximal support, in agreement with the sparsely sampled ML trees of Grant and Katz (2014) and Katz and Grant (2015). If strain Chinc5 (ATCC 50982) really were 'Stereomyxa ramosa', that would support including both Stereomyxidae and Acanthamoebidae in Centramoebida (Rogerson and Patterson, 2002). Cavalier-Smith et al. (2004) and Smirnov et al. (2011) doubted that relationship so excluded Stereomyxidae (and Dictyosteliidae whose inclusion by Rogerson and Patterson was obviously mistaken). Table 2  not seen in Stereomyxa (Grell, 1966;Benwitz and Grell, 1971); this strain is probably misidentifed and actually an undescribed genus. Additional representation of three previously sequenced orders or suborders in Fig. 4 compared with Cavalier-Smith et al. (2015b) confirms their monophyly. Both Vannella group together with maximal support (suborder Vannellina), as do both Vexillifera, both Paramoeba, and both Filamoeba (order Varipodida, now with three species, remains maximally supported).

New genus Atrichosa is a tubulinean distinct from Trichosphaerium
Trichosphaerium is an unusual giant multinucleate marine amoeba with a thick test-like cuticle composed of calcite spicules (discovered by Schneider, 1878) that led some to place it in Foraminifera (Loeblich and Tappan, 1964) rather than Amoebozoa, or in Testacealobosa (Schuster, 1990, not in Arcellinida as Tekle et al. (2008) wrongly wrote). However unlike Foraminifera it has numerous lobose pseudopodia, leading others to include it in Lobosa , though Möbius (1889) placed it in its own group Trichosa. Cavalier-Smith et al. (2004) suggested that its lobopodia hinted that Trichosphaerium may belong in Tubulinea; but because of early reports that T. sieboldi had a complex multiphasic life cycle with a biciliate swarmer stage (Schaudinn, 1899) suggesting an affinity with Rhizaria, they provisionally excluded it from Amoebozoa until sequence evidence became available. Schaudinn also found similar giant amoebae lacking spicules; though rarely present, he thought they were stages in the life cycle of T. sieboldi variously called gamonts or sporonts, postulated to come from the spiculate form named schizonts and to regenerate spiculate schizonts in an alternation of generations.
Polne-Fuller (1987) isolated into axenic culture a non-spiculate trichosid with a thinner smooth test that fed on macroscopic seaweeds (Polne-Fuller et al., 1990). It could grow much larger (up to 1 mm) than non-spiculate forms associated with T. platyxyrum (Angell, 1976) and unlike them could undergo multiple fission into small amoebae with only few nuclei when less well fed and binary fission when well fed. The presence of both lobopodia and dactylopodia identify it as a trichosid, but similarity in pseudopodia and its large size and multinuclearity was insufficient reason to identify it as Trichosphaerium, which was done primarily because of Schaudinn's uncorroborated theory that naked and spicular trichosids are stages in one life cycle. Tekle et al. (2008) obtained an 18S rRNA sequence for Polne-Fuller's non-spiculate strain (ATCC 40318) included on our multiprotein trees (under our new name Atrichosa algivora), which they stated to be 'very similar to' an unpublished one of the 'smooth stage of Trichosphaerium sieboldi CCAP 1585/2' studied by Pawlowski and Fahrni (2007). Their 95% identity is enough to put them in the same family, but not close, and means they must be separate species; however, in the absence of experimental proof of the three-phase life cycle postulated by Schaudinn, there is no evidence for the Pawlowski strain being T. sieboldi rather than a separate genus. Though referred to only as dactylopodia by Tekle et al. (2008), most pseudopodia in their Fig. 1E are lobose, only one dactylopodium being evident. The Tekle et al. sequence is so divergent from all other Amoebozoa that it is hard to align and initially grouped with the equally longbranch myxogastrids, but as this seeming relationship disappeared when faster evolving nucleotides were removed Tekle et al. (2008) considered it a long-branch artefact.
That reasoning is shown to be correct by our firm evidence using 41,675 amino acids from 149 'Trichosphaerium' sp. genes that it is part of the Tubulinea clade, thus not a conosan. Our trees more weakly put it as sister to Copromyxa plus Nolandella, as on 18S rDNA trees are Arcellinida and Leptomyxida Kudryavtsev et al., 2014). We need transcriptome or genome sequences from these two orders to be sure that Atrichosa ('Trichosphaerium') is not specifically related to Testacealobosa which Schuster (1990) grouped with Trichosida in subclass Testacealobosa, not an entirely distinct clade as morphology makes likely.
We are unconvinced that strain ATCC 40318 is a Trichosphaerium, for reasons elaborated in the supplementary material. In essence there is no evidence that spicule-free 'Trichosphaerium' amoebae are life cycle stages of the original spiculate Trichosphaerium, rather than contaminants of non-clonal cultures. Biphasic alternation between spiculate 'schizonts' and smooth 'gamonts' (Schaudinn, 1899) is unproven; even Page (1983) who thought that the non-spiculate Pontifex maximus of Schaeffer (1926) was just a Trichosphaerium phase admitted that was uncertain. As strain ATCC 40318 entirely lacks spicules it was safest to make this strain of non-spiculate trichosid-like amoebae a new genus and species, Atrichosa algivora.
As, apart from absence of spicules Atrichosa are similar to Trichosphaerium, including having similar mitosis with intranuclear spindle and metaphase plate (compare Schuster, 1976;Angell, 1976), we assign Atrichosa to Trichosphaeriidae, but until sequences are available from a spiculate Trichosphaerium we cannot be sure they belong in the same family or even order. Nonetheless having both lobose pseudopodia and hairlike subpseudopodia emanating from them and passing through pores in a multiporose test is unique for Amoebozoa; these features are unlikely to be convergent and thus are synapomorphies for Trichosida. Sheehan and Banner (1973) thought that lobopodia were entirely within the test, but Angell (1975Angell ( , 1976 noted that they can penetrate though pores to contact the substratum; unless they do they cannot cause locomotion, which even Sheehan and Banner considered their function. Previously an actin tree did not group 'Trichosphaerium' with the arcellinid Arcella or with any Tubulinea, but placed it without statistical support in an apparently paraphyletic Discosea, and on a four-gene tree Trichosphaerium grouped strongly with PRA-35 (later renamed Parvamoeba monoura: Cole et al., 2010) (Tekle et al., 2008), which led Smirnov et al. (2011) to assume that it probably belonged in Flabellinia within Discosea despite lacking obvious discosean morphology. Though Parvamoeba is not on our multiprotein trees, Fig. 1 shows that by rDNA it does not group within Tubulinea as Atrichosa ('Trichosphaerium') robustly does on our protein trees, which therefore robustly contradict the grouping with Parvamoeba that is attributable to insufficient data plus the 18S rDNA long-branch bias. With six genes 'Trichosphaerium' again wrongly grouped with myxogastrids, not PRA-35 -still wrongly called 'Thecamoeba sp.' (Lahr et al., 2015). Another actin tree grouped 'Trichosphaerium' with the tubulinean Vermamoeba with 4% support but far away from the well-supported main Tubulinea clade (Lahr et al., 2011), confirming that actin sequences have too little information for reconstructing deep eukaryote phylogeny (see also Cavalier-Smith, 2015).
Bizarrely, the eukaryote-wide ML tree of Grant and Katz (2014) based on 238 proteins plus 18S rDNA grouped 'Trichosphaerium' with myxogastrids with 89% support. However, it was limited to just 13 Amoebozoa and only two other Tubulinea. That tree had another peculiar feature: placing breviates as sister to metamonads and grouping that presumably artefactual clade with Discoba, in complete contradiction to other multigene trees that using ML robustly place breviates as sister to apusomonads and this clade as sister to opisthokonts or if using PhyloBayes CAT place them as sister to opisthokonts plus apusomonads (Brown et al., 2013;Cavalier-Smith et al., 2014, 2015a. We once noted a similar Breviatea/Metamonada grouping when using many too few genes and amino acid positions (about a third of those in our published 187-192-gene papers: Cavalier-Smith et al., 2014, 2015a,b), and note that when Trichozoa are included in our multigene alignments they sometime do branch as sister to obazoa (opisthokonts, apusomonads, breviates) but only on CAT trees (Cavalier-Smith et al., 2015b). Contradictory features of the Grant and Katz tree might also have something to do with the unusual automated pipeline they used perhaps excluding too many informative characters, or even including shared LGTs between the anaerobic Breviatea and Metamonada. We especially do not understand why even though it uses 238 proteins (51 more than our 187) it included only 15, 560 characters (rDNA plus protein) little more than a third of the amino acids that we were able to align (50,964). Burki et al. (2012) using 258 genes included only slightly fewer amino acids than we did, so the Grant-Katz pipeline discarded vast amounts of reliable data. Grant and Katz did not use single-gene trees for quality control, essential for reliable multigene phylogeny. Visual checking such trees can reveal problems missed by automated scripts. As 'Trichosphaerium' would have to cross eight maximally supported nodes on our trees to group within Mycetozoa to be sister to myxogastrids, and has no morphological or ecological similarity to them, its position within Mycetozoa is incredible; the Grant and Katz tree must be considered unreliable.
A recent combined 150-protein LG and GTR + C rDNA ML tree using the same transcriptome data as we do (and the Grant/Katz pipeline) weakly placed 'Trichosphaerium' with 'Mayorella' [actually Cunea] sp.   Fig. 2 that does not show support values for that or any deep branching lobosan or scotokaryote nodes, yet incorrectly claimed to have resolved the eukaryote tree); their supplementary non-converged CAT tree for the proteins only (no support values at all) placed 'Trichosphaerium' deeper with no specific sister. As both trees included only four other Lobosa and no other Tubulinea they could not reveal that Atrichosa ('Trichosphaerium') is a tubulinean as we unambiguously show.
3.6. A 265-taxon amoebozoan 18S rDNA site-heterogeneous (CAT) tree is more accurate than ML Atrichosa was omitted from this tree (Fig. 1) because its 18S rDNA is so divergent and hard to align that its position is unreliable; including it risks distorting other branching patterns. Representatives of all other significant amoebozoan lineages are included except the also long-branch myxogastrids (included in Fig. S1 with a few other long-branch taxa in a 300-taxon tree). In several respects this first amoebozoan tree using an evolutionarily realistic site-heterogeneous model is more congruent with our multiprotein trees and morphologically based classification (Table 2) than is the algorithmically oversimplified sitehomogeneous ML tree. Here ML agrees with CAT PhyloBayes in grouping Cunea with other paramoebids (77%), unlike previous trees where MrBayes moderately supported this clade but ML support was <50% (Kudryavtsev and Pawlowski, 2015). Unlike the CAT rDNA tree and multiprotein trees, the ML rDNA tree for Fig. 1 alignment wrongly places Cutosea and Cochliopodium within Dactylopodina, and splits Ovalopodium away from Cochliopodium, making Cochliopodiidae wrongly appear polyphyletic in strong contradiction to multiprotein trees; Cochliopodiidae is correctly a clade (low 0.55 support) in Fig. 1.
Cutosea are a deeply divergent robust clade (0.98 100%) sister to Tubulinea plus Conosa, and do not group with any discosean lineage. However, the corresponding ML tree wrongly grouped Cutosea with Cochliopodiidae (18%) and this 'clade' (false, as totally contradicted by the 187-gene trees) was insignificantly (13%) sister to Vexilliferidae, as was Cutosea alone (no support) on a 151-taxon tree (Kudryavtsev and Pawlowski, 2013). This confirms that rDNA has insufficient information to place Cutosea; rDNA is consistent with 187-gene trees in Cutosea being very early diverging.
Like many earlier site-homogeneous Amoebozoa-wide trees (e.g. Cavalier-Smith et al., 2004;Smirnov et al., 2011;Lahr et al., 2011;Zadrobílková et al., 2015) Fig. 1 does not show Conosa, Lobosa and Variosea as clades, as Archamoebae group with Tubulinea (seen also by Zadrobílková et al., 2015) and Discosea appear paraphyletic (by CAT and ML, but no support). Though some published trees omitting many of the more problematic branches we included show clade Conosa (Kudryavtsev and Pawlowski, 2013;Kudryavtsev et al., 2014;Berney et al., 2015), our results confirm previous conclusions that rDNA has too little information to resolve basal amoebozoan branches Cavalier-Smith et al., 2015b), which site-heterogeneous trees cannot be expected to improve. Nonetheless Fig. 1 is more informative than most site-homogeneous trees for evolutionary important branches in Conosa, with important taxonomic implications for Variosea and Mycetozoa especially; as they are still severely undersampled by multiprotein trees, we summarise them below.
In Mycetozoa, Fig. 1 shows a robust clade comprising protosporangids and ceratiomyxids (0.99, 84%) -first found for 18S rDNA by Berney et al. (2015) and Zadrobílková et al. (2015) with strong support; and by Lahr et al. (2015) on 18S rDNA/actin trees (negligible support). With ML this clade is weakly (37%) sister to dictyostelids as in Berney et al. (2015) who also omitted Myxogastrea, and in Zadrobílková et al. (2015) and Lahr et al. (2015) who both included Myxogastrea which did not branch with this clade. With CAT the protosporangid/ceratiomyxid clade does not branch within Amoebozoa but is oddly sister to Apusozoa with insignificant support (0.49 Fig. 1). This curious position is not because we omitted myxogastrids, as when they are present Protosporangida/ Ceratiomyxa remain with Apusozoa, and myxogastrids are sister to Filamoebidae within Variosea (Fig. S1). Even when long-branch Ceratiomyxa (plus some other longer branches) are excluded in a 248-taxon tree, Protosporangium is oddly within Apusozoa (0.47, insignificant support). Though Protosporangium is not a particularly long branch its sequence is rather divergent from other Amoebozoa in places, which might cause an unidentified model incongruency with PhyloBayes alone, giving this unexpected position. Despite lack of agreement between ML and CAT, these trees are consistent with protosporangids alone among major protosteloid lineages lacking the V7 18S rRNA synapomorphy for Variosea (Berney et al., 2015) and with our revised classification excluding them from Variosea and placing them within mycetozoan subclass Exosporeae (Table 2). Neither our trees nor those of Berney et al. (2015) or Zadrobílková et al. (2015) support an earlier extremely weak grouping of protosporangids with protostelids (Shadwick et al., 2009) that this V7 synapomorphy contradicts; instead they are consistent with Lahr et al. (2011) who included only protosporangids, and found them strongly sister to dictyostelids, not Myxogastrea.
Traditionally plasmodial Ceratiomyxa was considered related to Myxogastrea not dictyostelids (e.g. Calkins, 1926), though put in a separate group Exosporeae, continued by Cavalier-Smith et al. (2004) in their own order. After discovering protostelids, Olive (1975;see also Hutner, 1985;Spiegel, 1990) included them in Protostelida. When first sequenced Ceratiomyxa branched immediately below Myxogastrea/dictyostelid bifurcation on 18S rDNA ML trees; branching order of these three clades using MrBayes was unresolved (Fiore-Donno et al., 2009); Ceratiomyxa did not group with Cavostelium, the only protosteloid then sequenced. As Ceratiomyxa grouped with Myxogastrea on an 18S rDNA/EF1-a tree (Fiore-Donno et al., 2009) later classifications kept the Exosporeae/ Myxogastria grouping (Cavalier-Smith, 2013;Ruggiero et al., 2015), though an 18S rDNA/actin tree put Exosporeae as sister to Dictyostelida (Lahr et al., 2013). Our study is the fourth to show Ceratiomyxa's relationship with Protosporangium; none contradicts it, so we added Protosporangida to Exosporeae (  Zadrobílková et al., 2015) as did our ML trees (whether or not Myxogastrea are included), and none with myxogastrids; our CAT trees including Myxogastrea did not group them with either Exosporeae or Dictyostelida or further clarify the branching order of these three mycetozoan groups. In the absence of sequence phylogeny or other evidence for a specific grouping of Exosporeae, we transfer thus expanded Exosporeae to Stelamoebea, grouping them with dictyostelids, which their unicellular fruiting body structure favours (Table 2). Following Cavalier-Smith (1993a,b), Cavalier-Smith et al. (2004), and Ruggiero et al. (2015) for Myxogastrea, we abandon the misleading synonym Myxomycetes Link, 1883 that wrongly implies they are fungi.
Except for excluding Angulamoeba and including Dictyostelida (neither significantly supported -merely 0.32 and 0.46 so unresolved), CAT shows Variosea as a clade (Fig. 1). The corresponding ML tree correctly included Angulamoeba in the Variosea clade but also included Dictyostelida (trivial 6% BS support). Our trees are therefore consistent with, but do not demonstrate, the idea that Variosea, redefined as those Conosa sharing a characteristic 18S rDNA V7 synapomorphy (Berney et al., 2015), are a clade distinct from Mycetozoa and Archamoebae and not paraphyletic. All variosean families, and all orders in Table 2 are clades in Fig. 1 except for morphologically coherent Ramamoebida that is a clade on the Variosea-only tree of Berney et al. (2015). For the first time, our trees show that Heliamoeba is sister to Filamoeba (0.9, 35%) so belongs in Filamoebidae. Telaepolella (Lahr et al., 2012) as Berney et al. (2015) found is robustly sister to Flamella with which it shares a fan-shaped locomotive form unique in Varipodida and almost so in Variosea (Soliformovum expulsum can be fan-shaped: Shadwick et al., 2009), the basis for new varipodid family Flamellidae distinct from Filamoebidae (Table 2). Acramoeba, groups with Soliformovum and Grellamoeba, so Acramoebidae is now transferred from Varipodida to Protostelida and Grellamoeba from Acramoebidae to Soliformovidae. This grouping is consistent with locomotory phase morphology and agrees with the site-homogeneous Amoebozoa-wide tree of Berney et al. (2015) but not their contradictory Variosea-only tree that showed a morphologically unreasonable grouping of Acramoeba with Multicilia. As the Soliformovidae/Acramoeba clade is weakly sister of Protosteliidae we retain Soliformovum in Protostelida, but former protosteloids that consistently do not group with Protosteliidae are placed in other orders (Table 2, see Section 3.7).
Basal phylogeny of cellular slime moulds (Dictyostelida) in site-heterogeneous rDNA trees (e.g. Fig. 1) strongly contradicts the earlier conclusion that the typically small-spored Group 1 dictyostelids are the most divergent branch (Schaap et al., 2006). Our most comprehensive 300-taxon CAT tree (Fig. S1) and corresponding ML tree both show the dictyostelid root between two large clades: Group 1/Group 2/Dictyostelium polycarpum and Groups 3/4/Dictyostelium polycephalum. Group 1 is not the most divergent, but a very long-branch sister to Group 2 protostelids (0.97; this clade is sister to Dictyostelium polycarpum with maximal support) or to Group 4 plus D. polycephalum (ML 69%; only 64% support for it branching below the D. polycephalum/Group 2 last common ancestor). Fig. 1 analyses omitted this potentially confusing long-branch clade.
An unexpected feature of Fig. 1 is that by CAT Thecamoeba does not group with Sappinia/Stenamoeba, but with Protacanthamoeba within Centramoebida with remarkably high support at two nodes (0.99, 0.97). However, the 300-taxon CAT tree places Thecamoeba instead weakly (0.4) as sister to the Sappinia/Stenamoeba clade (Fig. S1), showing Thecamoebida as a clade in agreement with site-homogeneous trees, so this topology, not the apparent polyphyly of Thecamoebida in Fig. 1 is probably correct. As both site-heterogeneous trees show a statistically supported Sappinia/ Stenamoeba clade, we establish a new family for it: Stenamoebidae Cavalier-Smith fam. n. Diagnosis: flattened uninucleate or binucleate, non-ciliate elongate, smoothly monopodial, lobose amoebae without prominent longitudinal folds (unlike Thecamoebidae); subspeudopodia absent; thin amorphous surface coat without scales or glycostyles, continuous and close to plasma membrane (Stenamoeba) or somewhat separated from membrane by discrete vertical elements and variable from continuous to a fragmented layer (Sappinia); aerobic; branched, tubular mitochondrial cristae. Type genus Stenamoeba Smirnov et al., 2007. Other genus Sappinia Dangeard, 1896.

Revised higher-level taxonomy for Amoebozoa
In light of our new rDNA and multiprotein trees and also of many ultrastructural and sequence phylogenetic studies published since the last gymnamoeba classification was prepared for Smirnov et al. (2011), Table 2 presents an updated classification that treats Conosa in more detail than before, especially to integrate numerous new variosean genera (Berney et al., 2015). Diagnoses are below with comments clarifying some innovations: Diagnosis of new superclass Cutosa Cavalier-Smith, new class Cutosea Cavalier-Smith, and new order Squamocutida Cavalier-Smith: uninucleate amoebae bounded by a continuous thin, somewhat flexible, envelope (not closely attached to plasma membrane) having oval scale-like substructure within a denser matrix; one or many small pores penetrate the envelope, allowing pseudopodia to protrude for very slow, occasional locomotion; locomoting cells flattened, oval or round; cilia or centrosome absent; radiate floating forms absent. Etymology: cutis L. skin; squama L. scale; names chosen because cell envelopes are flexible like skin, but part of a mechanically continuous structure with scaliform substructure, morphologically and presumably mechanically separate from the cell body -like a recently moulted snake skin not yet sloughed off. Separate names are used for class and order to avoid confusion should cutoseans meriting a second order be found.
Squamamoebidae Cavalier-Smith fam. n. Diagnosis: as for Cutosea, plus: scale-like test substructures have central projecting filament; locomotion involves numerous tiny mammiliform pseudopodia fixed to substratum as cell moves forward; stationary forms can be strongly branched. Type genus Squamamoeba Kudryavtsev and Pawlowski, 2013. Comment: ATCC PRA-29 'Pessonella' sp. probably also belongs in this family if its bosses protrude through a multiporose test overlooked in the light microscope, but is excluded pending ultrastructure, needed to see if it is an undescribed Squamamoeba or new genus. Its statistically unsupported grouping with Squamamoeba rather than Sapocribrum on a 2-gene tree (Lahr et al., 2015) fits its greater morphological similarity, but its deep branching suggests PRA-29 may be an undescribed sister genus.
Sapocribridae Cavalier-Smith fam. n. Diagnosis: as for class, plus: scale-like substructures lack central projecting filament; one or sometimes two thin contractile filose pseudopodia longer than cell body; stationary forms rounded, unbranched. Type genus Saprocribrum Lahr et al., 2015. As Tubulinea and Discosea differ radically in surface structure from Cutosea and group together as a strongly supported clade we group them as new lobosan superclass Glycopoda.

New taxa in Tubulinea
Diagnosis of new genus Atrichosa Cavalier-Smith: Multinucleate algivorous marine amoebae with broad, short lobose pseudopodia; calcite spicules, cilia, and dactylopodia absent. Hair-like filopodia emanate from hyaline lobopodial zone and pass through pores in thin test that envelops most of the cell. Etymology: A Gk without; trich Gk hair, referring to the absence of spicules, unlike Trichosphaerium. Type species Atrichosa algivora Cavalier-Smith sp. n. Diagnosis as for genus, plus cells broadly oval, length 10-1000+ lm. Type culture ATCC 40318; type sequences EU273464-71; type illustration Fig. 1E of Tekle et al. (2008). Comment. This is the species whose transcriptome was sequenced by Katz and Grant (2015) under the name Trichosphaerium and shown in our multigene trees (Figs. 2-4). We considered the possibility that this is the same species or genus as Pontifex maximus (Schaeffer, 1926), but Shaeffer believed his amoeba was naked. If he was right they are not the same. If he was wrong and it was testate as Page (1983) assumed, it is possible that they are congeneric, but it is extremely hard to decide whether or not the Pontifex micrographs of Page or other observations/records of 'Pontifex' are the same species as ATCC 40318. The identity or otherwise of Atrichosa and Pontifex may be undecidable; only when a large number of morphologically similar strains are sequenced will it be possible to evaluate the phylogenetic diversity of such phenotypes. As for other poorly studied amoeboid groups like Variosea (Berney et al., 2015) it may be much greater than Page realised; nomenclatural clarity and unambiguity is probably best served by attaching a new name to a sequenced and microscopically defined strain with type culture available. Page (1983) thought P. maximus the same as Amoeba tentaculata (Gruber, 1881). Atrichosa is morphologically highly dissimilar from any true Amoeba, and if did belong to that genus it would have grouped with Copromyxa on our trees which it did not. One cannnot logically claim simultaneously that Amoeba is a more valid name for Pontifex (as did Page, 1983) and the same genus as Atrichosa.
Addition of Trichosida makes Tubulinea distinctly more heterogeneous, so it is now advisable to subdivide them into three new subclasses, separated at points of maximal phenotypic disparity: 1. Neolobosia Cavalier-Smith: Diagnosis: uninucleate or multinucleate aerobic amoebae; branched tubular mitochondrial cristae; locomotory shape monopodial or multipodial with cylindrical cross section pseudopodia, not strongly flattened; anastomoses or spine-like subpseudopodia absent. Etymology: Neo Gk new refers to this clade originating substantially later than the ancestral lobosan. Includes two new superorders: Eulobosia Cavalier-Smith (Diagnosis: amoebae without dactylopodia or tests of calcite spicules. Etymol: Eu-Gk well; lobose E. lobed, referring to the well-developed cylindrical lobe-like pseudopods of these classical lobose amoebae, comprising the naked order Euamoebida and testate Arcellinida) and Trichosia Cavalier-Smith (Diagnosis: giant multinucleate naked marine amoebae with fibrous multiporose test through which nonlocomotory filopodia protrude; locomotory lobopodia vary from a single broad major hyaline zone to numerous short mammiform lobes). Sole order Trichosida.
The very close branching of Nolandella and Copromyxa (much closer than any other two lobosan orders) and rather small phenotypic differences between them lead us to transfer Nolandellidae into Euamoebida as a new suborder and to establish a suborder for the traditional euamoebids: Suborder 1. Amoebina Cavalier-Smith subord. n. Diagnosis as for order Euamoebida (Smirnov et al., 2011, p. 565) Suborder 2. Nolandina Cavalier-Smith subord. n. Diagnosis as for order Nolandida (Smirnov et al., 2011, p. 565). Lahr et al. (2013) wrote 'Hartmannella abertawensis stands out as an immediate candidate to be transferred to Nolandella', unaware that Cavalier-Smith and Smirnov (in Smirnov et al., 2011, p. 561) had already done that and Cavalier-Smith had established order Nolandida for this taxon (Smirnov et al., 2011, p. 565) independently of synonymous Poseidonida of Lahr et al. (2011). 2. Leptomyxia Cavalier-Smith: Diagnosis: naked Lobosa; locomotive form varies from flattened expanded or reticulate, if slowly moving, to subcylindrical and monopodial especially in rapid movement; adhesive uroid; uni-to multi-nucleate; glycocalyx amorphous. Etymol: named after sole order Leptomxida.

New families and orders of Variosea
Ramamoebida Cavalier-Smith ord. n. Diagnosis: uninucleate, often elongated, sometimes reticulose, amoebae with branched filose pseudopodia or subspeudopodia that generally move too slowly for motility to be observable in the light microscope; phylogenetically more closely related to Cavostelium than to Filamoeba. Unlike Varipodida, cilia (in multiple kinetids) and a stalked spore-bearing fruiting stage sometimes present. Etymol: ramus L. branch, denoting their often branched, non-reticulose pseudopodia + amoeba (the cell body also is often branched). Includes the protosteloid Cavosteliidae and two new families with unstalked cysts, each a strongly supported rDNA clade: Ischnamoebidae Cavalier-Smith fam. n. Diagnosis: nonfruiting, non-ciliate amoebae with often branching filopodia; distinguished from Filamoebidae and Acramoebidae by narrow elongated or narrowly branched cell body and closer genetic relationship to Cavosteliidae; cysts round, oval or bean-shaped, unstalked. Type genus Ischnamoeba Geisen, Bass, and Berney in Berney et al., 2015. Other genus Darbyshirella Berney, Bass, andGeisen in Berney et al., 2015. Angulamoebidae Cavalier-Smith fam. n. Diagnosis: nonfruiting, amoebae with often branching filopodia, markedly less elongate cell bodies than Ischnamoebidae; distinguished from Acramoeba by cell body often being triply branched and by eating fungi not bacteria, and in at least one species by a multiciliate stage with more than one kinetid; cysts spherical to oval, unstalked. Type genus Angulamoeba Berney, Bass and Geisen in Berney et al., 2015. The sometimes biciliate (Ceratiomyxella only) protosteloid Schizoplasmodiidae (Nematostelium, Ceratiomyxella, Schizoplasmodium) also do not group with Protostelida on rDNA trees but weakly with the non-amoeboid uniciliate Phalansterium (Shadwick et al., 2009). Two highly reticulose, non-fruiting (i.e. non-protosteloid), extremely slow-moving amoebae (Arboramoeba, Dictyamoeba), having a novel reticulose morphotype (sensu Smirnov et al., 2011) for Amoebozoa, also group weakly with this clade (Berney et al., 2015). We establish new families within Phalansteriida for these previously unclassified genera: Arboramoebidae Cavalier-Smith fam. n. Diagnosis: non-ciliate, reticulose amoebae moving too slowly to see by real time microscopy; no obvious main cell body; protoplasmic networks grow to >600 lm, filopodia anastomosing when older; fine pointed pseudopodia predominantly at anterior front; double-walled cysts, not on stalks. Type genus Arboramoeba Geisen, Bass and Berney in Berney et al., 2015. Dictyamoebidae Cavalier-Smith fam. n. Diagnosis: non-ciliate, reticulose amoebae moving too slowly to see by real time microscopy; multiply branched main cell body, anastomosing, giant networks up to several mm; fine pointed pseudopodia near tips; unicellular cysts not on stalks. Type genus Dictyamoeba Berney, Bass, and Geisen, 2015. Our site-heterogeneous rDNA trees (Fig. 1) confirm that both genera, despite apparent absence of cilia (presumably independent losses), form a weakly supported clade with the non-amoeboid uniciliate Phalansterium plus the biciliate amoeboid, protosteloid Schizoplasmodiidae, as weakly suggested by Fig. 1 of Berney et al. (2015, but contradicted by their Fig. 2 that put Phalansterium slightly deeper than the five amoeboid genera, without support). We therefore transfer Schizoplasmodiidae from Protostelida to Phalansteriida and also add Arboramoeba and Dictyamoeba to Phalansteriida; as all five amoeboid genera now in Phalansteriida are reticulose this was likely the ancestral condition for Phalansteriida, lost by Phalansterium alone.
Our site-heterogeneous rDNA trees (Fig. 1) strongly confirm previous evidence (see Section 3.6) that Protosporangium plus Clastostelium are a clade entirely distinct from Protostelida sensu stricto, so we make this clade a new order: Protosporangida Cavalier-Smith. Diagnosis: non-ciliate amoebae with separate ciliate swarm cell phase with one or more unequally biciliate kinetids; fruiting stage with one or two spores borne on a single stalk; unlike Protostelida and other variosean protosteloids, lack the variosean-specific highly conserved nucleotide sequence motif with the consensus GGGUGAAG in the ascending stem, and UGGAUCCU in the descending stem of the unpaired region at the base of helix E43_2 in the V7 region of 18S rRNA.
Etymol: proto Gk first; sporangium E.; to emphasise that the singly stalked uni-or bispored sporangium is more primitive than other Mycetozoa.

Stygamoeba and the unity of order Glycostylida
Our trees provide the first strong evidence that Stygamoebidae are sister to Dactylopodida. Dactylopodida are not sisters of Vannellida as assumed when class Flabellinea was proposed for them . Instead Vannellida are sister to Stygamoebida/Dactylopodida. These three orders share the common feature of flattened shape with polyaxial flow of endoplasm and rather thick glycocalyx, but are sharply differentiated into three rather different phenotypes: Vannellida (fan-shaped, often with glycostyles forming the cell coat), Stygamoebida (flattened irregular long worm shape, with pyramidal glycostyles in Vermistella but none in Stygamoeba), and Dactylopodida [fingerlike pseudopodia (dactylopodia) formed from a frontal hyaline area and a cell coat of usually glycostyles (boat-shaped scales in Korotnevella only)]. Collectively they correspond closely (not quite exactly) with the earlier order Glycostylida ) assumed ancestrally to have had glycostyles. As ordinal rank should be reserved for taxa that differ rather substantially in phenotype from their closest relatives we have retained the now phylogenetically well substantiated, more comprehensive order Glycostylida emended by the exclusion of Multicilia (which is phylogenetically a variosean (Fig. 1); glycostyles that led to its former erroneous inclusion in Glycostylida presumably came from its vannellid food) and Mayorella, and inclusion of Stygamoebidae that were unknown when Glycostylida was established; the three glycostyle-bearing orders of Smirnov et al. (2011) become suborders (Table 2). By contrast subclass Flabellinia sensu Smirnov et al. (2011) is clearly not a clade as neither Himatismenida, nor Trichosida group with glycostylids on multiprotein trees, so we abandon Flabellinia as a taxon, though use it informally in Figs. 3 and 4 in a revised sense to label a robust, often flabellate, clade. Table 2 places the only other former flabellinian order, Pellitida, in Himatismenida as a suborder. Kudryavtsev and Pawlowski (2013) showed Pellita to be related to Goceviidae and so transferred Goceviidae from Himatismenida to Pellitida. Taxonomic simplicity is better favoured by instead transferring Pellita to Himatismenida, and grouping it with Goceviidae as new suborder Pellitina with the same composition as Pellitida sensu Kudryavtsev et al. (2014). Pellitina is now put in Centramoebida (see Section 5).
Absence of glycostyles from Stygamoeba may be secondary loss as it is in a few vannellids (Smirnov et al., 2007). However, in Fig. 1 Stygamoebidae is weakly paraphyletic with glycostyle-bearing Vermistella only being sister to other glycostylids (insignificantly different paraphyly in Fig. S1). Though our protein trees make it likely that this non-grouping of Stygamoeba is an incorrect consequence of low basal resolution by rDNA trees, if multiprotein trees confirm that Vermistella is even more closely related to other glycostylids than is Stygamoeba, absence of glycostyles in Stygamoeba could be the ancestral condition for the glycostylid clade and it would be appropriate to segregate Vermistella into a separate family (see Table 2 footnote h). In our trees the sister to Glycostylida is Stenamoeba (Thecamoebida, all with elongate monopodial cell bodies without dactylopodia), which indicates that their common ancestor had an undivided flattened cell shape without discrete finger-like pseudopods (dactylopodia) and that dactylopodia are a derived condition for suborder Dactylopodina alone. Differences in body form between Thecamoebida and Glycostylida are no greater than between the three glycostylid suborders, so they might reasonably be grouped one day as a superorder, whose common ancestor almost certainly had a flat cell body and no dactylopodia, these pseudopods evolving later in an ancestor of Dactylopodina.
The only discosean order not now represented on multigene trees is Dermamoebida comprising Dermamoebidae and Mayorellidae. Both have a very thick glycocalyx like Glycostylida with a complex substructure in marked contrast to Thecamoebida with a simple thin glycocalyx. Dermamoebida could be sisters to Glycostylida, even closer than Thecamoebida; on unresolved Figs. 1 and S1 all three are equidistant.

Distinctive features of Cutosea, the deepest branching Lobosa
We call this novel amoebozoan clade Cutosea, because Sapocribrum and Squamamoeba share a unique skin-like cuticle or envelope surrounding the whole cell except for rare small breaks or pores through which pseudopodia protrude (Kudryavtsev and Pawlowski, 2013;Lahr et al., 2015). The envelope comprises a dense matrix in which irregularly arranged oval structures about 150 nm (Sapocribrum) or 125 nm (Squamamoeba) in length are embedded. Though the embedded ovals were initially called a 'cell coat' of 'scales' (Kudryavtsev and Pawlowski, 2013;Lahr et al., 2015), the ovals are not discrete separable scales like true scales of Discosea (Cochliopodium, Dactylopodina) or chromists such as Paraphysomonadida, Thaumatomonadida or Haptista, but internal specialisations of a continuous, mechanically coherent-envelope. Consequently we refer to the ovals embedded in the cutosean envelope matrix not as scales but as scale-like (as on a crocodile or snake skin). In Sapocribrum and Squamamoeba the envelope is clearly separate from the plasma membrane, like the test of testate amoebae, not directly attached to it as is the carbohydrate-rich surface coat of animal cells or the much thicker but still flexible cell coats of many Discosea (Pellita, Dermamoebida -called cuticle in Mayorella, Glycostylida). Lahr et al. (2015) noted that the Sapocribrum 'coat' appears more plastic and deformable than a typical test and so refrained from calling it a test, which because of its clear mechanical distinctness from the plasma membrane might be thought more appropriate than cell-coat or cuticle. Its deformability during specimen drying might just be because it is rather thin and unmineralised like the Atrichosa test. Though tests of Arcellinida are uniporose (like that of Sapocribrum), those of Trichosida are multiporose like that of Squamamoeba; being multiporose is therefore no reason not to call the Squamamoeba enveloping structure a test. However, the locomotory mechanism of Squamamoeba (Kudyravtsev and Pavlowski, 2013) implies a plastic envelope that can deform as cytoplasm flows past the ventrally anchored mammiliform pseudopodia, not a rigid test. As testa originally meant shell, implying rigidity, we agree that the scalifom outermost layer of Cutosea is best not called a test, and refer to it as an envelope as, like an ordinary letter envelope, it is distinct from its contents and completely encloses them except for some tiny pores. Cutosea are a previously unrecognised third group of lobosan armoured, non-naked amoebae additional to the testate Arcellinida (uniporose) and Trichosida (multiporose).
Cutosean test structure is unique not only in Amoebozoa but for all protists. Squamamoeba locomotive forms have a hyaline anterior and produce small mammiliform pseudopodia as rounded bulges that fix to the substratum as the cell protoplasm slides past them, unlike most pseudopodia of Tubulinea, but probably rather like Atrichosa. These pseudopodia somewhat resemble those of the discosean Mayorella, which also has a dense membraneadherent cuticle that is thicker, but with no oval scale-like substructure. Unlike in Squamamoeba the Mayorella naked subpseudopodia do not protrude through cuticle breaks. Pellita also has multiple small subpseudopodia that emerge through cuticular breaks and mediate a similar locomotion, but they are smaller than in Squamamoeba and not mammiliform bulges; their coat substructure is multilayered with an outer layer of glycostyles and inner layer of small discrete subunits directly attached to the plasma membrane, and lacks scale-like ovals. Moreover Pellita has centrosomes (unlike Cutosea) and therefore fits better in Discosea like centrosome-bearing Himatismenida and Centramoebida. Sapocribrum has much thinner and longer pseudopodia, i.e. filopodia more like those of the conosan Filamoeba (but proportionally thinner) than any Lobosa except for the filiform subspeudopodia of Trichosphaeriidae. Usually there is only a single pseudopod, longer than the cell body, postulated to contract to pull the cell forward, but sometimes there are two joined together near their base that emerges though the single envelope pore (Lahr et al., 2015); a terminal web of subpseudopodia of unspecified form also mentioned might possibly help locomotion similarly to Squamamoeba.
Though no ultrastructure is available for ATCC PRA-29 'Pessonella' sp. the sole light micrograph shows tiny bulges resembling Squamamoeba locomotive pseudopodia (Tekle et al., 2008), suggesting that its ultrastructure may be closely similar. As Kudryavtsev and Pawlowski (2013) correctly indicate, there is no evidence that the almost entirely undescribed ATCC PRA-29, for which the only published morphology is a single phase contrast micrograph (Tekle et al., 2008) is a Pessonella. All three sequenced Cutosea (including PRA-29) are marine, move very slowly, immobile most of the time, and have no distinct floating form, in marked contrast to the sole described species of Pessonella (marginata: Pussard, 1973), a freshwater and compost amoeba with typical vannellid fan-shaped or lingulate locomotive form and markedly different floating form (like vannellids and many other Discosea) with blunt, cylindrical and often bent pseudopodia. Smirnov (http://amoeba.ifmo.ru/amecol/divers/pess.htm accessed 16 June 2015) has shown that a second undescribed freshwater Pessonella species has a glycostyle-like cell coat consistent with its current classification in Vannellidae  and lacks the discrete cutosean cell envelope. The fundamental ultrastructural similarity of Squamamoeba and Sapocribrum, unique in protists, and grouping of both with PRA-29 on oligogenic trees (Lahr et al., 2015), together with the firm exclusion by our 187-gene trees of Cutosea from Discosea, shows that including Squamamoeba in Dactylopodida (Kudryavtsev and Pawlowski, 2013) was incorrect, being based only on poorly resolved 18S rDNA trees, a flattened body form, and slight pseudopodial similarity to Mayorella subpseudopodia. Likewise including Sapocribrum in Flabellinia (Lahr et al., 2015) was incorrect. Given their unique multiporose flexible envelopes comprising linked scale-like plates (like ancient Chinese and mediaeval Japanese body armour) and deep divergence from all other Amoebozoa, we ranked Cutosea as a third lobosan class with single order, Squamocutida but two families to emphasise their marked pseudopodial and envelope differences.

Evolutionary diversification and origin of Lobosa
Our evidence for early divergence of Cutosea illuminates the ancestral state for Lobosa. Their distinctive perforated dense envelope somewhat resembles the thick amorphous pellicle of Pellita (in Discosea) in that breaks in it are necessary for locomotion and also the multiperforated flexible theca of the tubulinean Trichosphaeriidae, but differs in not being integrated with the cell membrane. Thus one deep branching clade in both the other lobosan classes has a perforated flexible envelope-like morphology, structurally and functionally similar but not identical to that of Cutosea. Did this morphotype evolve independently three times in slightly different ways from ancestral naked amoebae? Or did the ancestral lobosan have a perforated envelope morphology like that of Cutosea and tubulinean and discosean 'gymnamoebae' evolve nakedness independently by modifying it into an even more flexible cell coat? We think it somewhat more likely than not that the common ancestor of Lobosa had cells enveloped by a thick coherent envelope, theca or pellicle that needed occasional breaks to allow pseudopodial extrusion. These coherent envelopes evidently rather limit active amoeboid locomotion; such amoebae are much less speciose than other Lobosa that have liberated their pseudopodia from this rather rigid constraint, which we postulate occurred in contrasting ways in five independently evolving lobosan subgroups.
First, Cochliopodiidae and Goceviidae restricted the thick layer to the dorsal surface as a tectum (probably independently, judging from rDNA trees: Kudryavtsev et al., 2014) and thus could develop the whole ventral surface for locomotion in a limpet-like manner, retaining the tectum as a dorsal protective shell. What the tectum protects against is an important question; though it might offer some protection against predation by large protists, we are sceptical of this and suggest that it may mainly protect against viruses invading from the supernatant medium through the plasma membrane and/or from ultraviolet radiation. UV-protection is important in many shallow aquatic habitats and in uppermost soil layers. The first author noted that almost all protozoa he found living on a sandy beach in the very clear waters of Lake Baikal (Siberia) had marked light-aversive behaviour, hiding below sand grains and only occasionally very briefly emerging; he once watched a Baikal beach Gocevia (Pellitina) in a Petri dish crawling among sand grains carrying a layer of tiny dense granules on its back -as it progressed these accidentally fell off, but remained coherent -presumably by mucilage, and the amoeba turned round, crawled under the granule layer and once it was reinstated on its back proceeded onwards with its dorsal cargo. Such behaviour arguably indicates that the granules are important to and controlled by it. We suggest that they and other dorsal structures like the large scales of Cochliopodium may offer UV-protection amongst other benefits.
Secondly Glycostylida/Dermamoebida evolved a more flexible but still thick glycocalyx, possibly ancestrally composed of glycostyles which allow the coat to stay with the plasma membrane during pseudopodial motion, unlike in Pellita and Cutosea and to evolve fan-like, irregularly vermiform or dactylopodial locomotory forms. Thirdly, by contrast, Stenamoeba the sister group to Glycodermia achieved flexibility by retaining surface coherence by greatly thinning the ancestrally thick glycocalyx so that the cell surface can be easily deformed (often wrinkled) during monopodial locomotion.
In contrast to those three ways of achieving flexibility during multiaxial endoplasmic flow in Discosia, Tubulinea evolved tubular near-cylindrical pseudopodia with a thick cortical gel and central uniaxial flow. Many larger Tubulinea have a thick amorphous glycocalyx (not so obviously multilayered as in many Flabellinia). Its thickness might be the ancestral condition for Tubulinea and derived from that of the earliest flattened common ancestor of Tubulinea and Discosea; if so the real novelty for Tubulinea was greater flexibility allowed by novel tubular pseudopodia. Were Tubulinea ancestrally monopodial like Echinamoebidae, the most divergent order, so the multipodial condition evolved independently in Amoebina and Leptomyxida? Or are unipodial forms derived from multipodial ones? Given that Cutosea include both multipodial and predominantly unipodial genera, and Discosea are predominantly unipodial with only a few (e.g. Dactylopodina and Pellitina) essentially multipodial, it is hard to decide whether the ancestral state for Amoebozoa was multipodial or unipodial. A reasonable scenario, given that the sulcozoan ancestors of Amoebozoa (see Cavalier-Smith, 2013 andCavalier-Smith et al., 2014) have very irregular pseudopodia variable in numbers, is that this may also have been true of the first Amoebozoa, and unipodial Lobosa evolved by multiple divergent simplifications. As Echinamoebida are the most divergent Tubulinea, the pointed subpseudopodia of Echinamoeba only within Tubulinea might be a relic of an ancestral condition for Amoebozoa, as those of Conosa were probably ancestrally pointed. Whether Leptomyxida, which likely diverged before the common ancestor of the most typical Tubulinea (subclass Neolobosia), and which adopt a more flattened, sometimes anastomosing, morphology unlike other Tubulinea, are secondarily atypical or retained a flattened form from ancestral Lobosa is uncertain. Multinuclearity clearly evolved at least twice in Tubulinea -in Trichosida and in some giant Amoebina.
The above evolutionary synthesis emphasises that tubular pseudopodia of Amoeba proteus, the prototype for early ideas about amoeboid locomotion (e.g. Mast, 1926;Pantin, 1923) are not the archetypal lobosan state, still less that of Amoebozoa as a whole. As the sulcozoan ancestor of Amoebozoa had a dorsal submembrane theca, ventral pseudopodia and posterior ciliary gliding (Cavalier-Smith, 2013), all three were lost during or immediately prior to the evolutionary transition to the ancestral lobosan. It would not be surprising if (as suggested above) the ancestral lobosan evolved an alternative extracellular flexible envelope to replace the lost stabilising function of the sulcozoan dorsal intracellular pellicular thickening, allowing retention of the ancestral sulcozoan ventral multiple pseudopodial condition during the radical changes that made Amoebozoa. The above interpretation of an enveloped lobosan common ancestor makes the transition from sulcozoan to amoebozoan body plan much more gradual and less dramatic, therefore more evolutionarily comprehensible, than previously (Cavalier-Smith, 2013); it is much more plausible than would be the conversion of a sulcozoan in one step to a classical neolobosan amoeba. We need information about comparative chemistry and biogenesis of these varied envelopes and their distribution in more Cutosea and other ill-studied early lineages before we can judge whether this envelope-first scenario for early lobosan evolution is really preferable to the only slightly less plausible standard naked-amoeba-first assumption implying polyphyletic origins of envelopes/coats. Whichever is correct it is entirely clear that Amoeba is not a primitive organism, contrary to a centuries-old misconception, but a highly derived specialised one. After a century and a half, the idea that eukaryotes and cells generally evolved from an amoeboid ancestor (Haeckel, 1866), which Cavalier-Smith (1987 dubbed the 'moneran myth' is finally defunct. Both prokaryotes and eukaryotes ancestrally had rigid cell surfaces and the transition between them was mediated by loss of the eubacterial cell wall and consequential origin of phagocytosis, cilia, and a largely rigid cell surface supported by cross-linked actin microfilaments and microtubules (Cavalier-Smith, 2014). Cavalier-Smith (1981) argued that amoeboid locomotion is clearly advanced not primitive, rejecting Haeckel's idea of the first eukaryote being an amoeba as the Zoological Myth, just as Margulis (1970) called the also now falsified idea that it was an alga the Botanical Myth. We are now left with semi-rigid biciliate phagotrophic zooflagellate reality as the only well-corroborated idea of our first eukaryotic ancestor (Cavalier-Smith, 2000, 2013, 2014.

Variosean diversity and basal evolution of Conosa
The ancestral conosan would have had two cilia and multiple ventral pointed pseudopodia or subspeudopodia when it evolved from the sulcozoan ancestor of Amoebozoa by losing dorsal ciliary gliding, as Cavalier-Smith (2013) explained. For a much better understanding of their origins we need multigene trees for all lineages of Variosea and Mycetozoa, as previous 18S rDNA trees suggested that protosteloids (non-social amoebae whose resting cysts are born on stalks and thus are called spores by analogy with those of slime mould fruiting bodies, and which were formerly lumped in class Protostelea) and Variosea are phylogenetically partially overlapping (Shadwick et al., 2009). Prior to the present paper Variosea comprised four morphologically distinct orders, three ciliated: uniciliate Phalansteriida (Phalansterium, Rhizomonas and three other largely sedentary and non-amoeboid genera: Cavalier-Smith, 2013; Cavalier-Smith and Scoble, 2013); Holomastigida, (the only minimally amoeboid Multicilia); and putatively multiciliate but not highly mobile Artodiscida (no sequences available). The non-ciliate Varipodida are unfortunately the only variosean order with available transcriptomes, making it very hard to establish their relationships and to decide whether Variosea are the paraphyletic ancestors of all Conosa or the earliest diverging clade or clarify phenotypic evolutionary patterns within the class.
Though no transcriptomes are available for 'Protostelea', the recent discovery of several new genera of non-fruiting Variosea, their placement on rDNA trees, and the recognition that some protosteloid lineages share an apparent molecular synapomorphy in the V7 region of 18S rDNA with them (Berney et al., 2015) enabled us to realign the boundary between Variosea and Mycetozoa, radically revise Protostelida, and establish new conosan orders Ramamoebida and Protosporangida, yielding a conosan classification more consistent with sequence phylogeny and morphological diversity. For the first time this assigns the plethora of recently discovered variosean genera (Berney et al., 2015) to families and orders. We followed Berney et al. (2015) in regarding all protosteloids possessing the variosean V7 synapomorphy as members of class Variosea. Table 2 formally transfers them into class Variosea, but their deep genetic diversity compared with Varipodida made it inappropriate to include them in that order.
Previously all conosan protosteloids were in order Protostelida (Olive, 1967) arranged in four morphologically and genetically distinct families (Cavalier-Smith, 2013). We removed from Protostelida all protosteloids except the two families whose amoebae have a simple non-branched and non-reticulose cell structure (Protosteliidae, Soliformovidae), which form a weakly supported clade on some 18S rDNA trees (Berney et al., 2015; and our distance trees, not shown), but not on others (Shadwick et al., 2009). Cavosteliidae, a strongly supported rDNA clade with sculptured spore walls and often a ciliated stage, do not group on rDNA trees with Protostelida sensu stricto (Shadwick et al., 2009;Berney et al., 2015); instead they either group weakly with Dictyostelida (Shadwick et al., 2009) or form a weakly supported clade (Ramamoebida) with three recently established variosean genera of amoebae with branched pointed pseudopodia that move too slowly for motility to be observable in the light microscope (like Acramoeba) and which are not known to form aerial spores (thus not protosteloids) (Berney et al., 2015). One of these non-fruiting genera (Angulamoeba) has branched cells sometimes with cilia in multiple kinetids as in Cavostelium. Though the filose subspeudopodia of Cavosteliidae are typically shorter than those of the non-fruiting amoebae of this putative clade, they are sufficiently similar to be grouped in a new, mostly soil-dwelling, presently entirely non-marine, order Ramamoebida.
Protosporangids are robustly sister with near maximum support to Ceratiomyxa in published site-homogeneous rDNA trees and in our Fig. 1 site-heterogeneous tree. We therefore placed Protosporangida within subclass Exosporeae of Mycetozoa, the probable sister to dictyostelids. Exosporeae as thus expanded all have non-ciliate trophic cells and short duration swarm cells with one or more biciliate kinetids formed just after germination of smooth-walled stalked spores.
The new phylogeny and taxonomic revision of Variosea emphasises that cilia have been lost independently by Varipodida and several members of other orders. One large order (Phalansteriida) retains the ancestral single kinetid (reduced to one cilium in Phalansterium but not Ceratiomyxella), but four orders (e.g. Protostelida) comprise or include families that multiplied kinetids per cell. Thus kinetid multiplication is not restricted to Multicilia as once thought  but is a pervasive evolutionary phenomenon in Variosea. Multigene trees including all these orders are necessary to establish the currently unclear basal branching order of Variosea and determine how many kinetid multiplications have occurred; present evidence suggests multiplication was independent in mycetozoan protosporangids and Variosea.

Protosteloid polyphyly and the origin and diversification of Mycetozoa
The discovery of many new Variosea with unstalked cysts, some related to protosteloids with stalked cysts/spores (Berney et al., 2015), throws new light on the origins of protosteloids and Mycetozoa. Berney et al. suggested that an ancestor of Variosea could produce fruiting bodies, i.e. stalked cysts, implying that their stalks evolved once and were lost many times. That assumption would imply about five losses if Fig. 1 topology is correct. As no morphological evidence exists for homology between variosean 'spore' stalks, it is more likely that a stalk evolved polyphyletically four times in Variosea. Evolution of an extracellular stalk for a standard amoebozoan cyst is probably a simple secretion of additional extracellular material at one pole of the cell. We are sure that stalked fruiting bodies and aggregative multicellularity as a prelude to fruiting evolved polyphyletically in at least five very distant protist phyla: Amoebozoa, Ciliophora (Sorogena), Percolozoa (Acrasis), Choanozoa (Fonticula: Brown et al., 2009), Heterokonta (Sorodiplophrys) and Cercozoa (Guttulinopsis: Brown et al., 2012), showing that selection for aerial dispersion is widespread and probably does not need particularly rare mutations or complex development to evolve several times. Most Variosea are soil amoebae with many opportunities to invade subaerial habitats like branches and twigs where aerial dispersion is highly advantageous. Evolving a stalk alone is even simpler than aggregative multicellularity plus a fruiting body, so a polyphyletic origin of protosteloid stalks is highly likely. Thus there is no reason to retain the highly polyphyletic class Protostelea.
There is no convincing argument for the variosean ancestor having had aerial spore dispersal. The idea that all Amoebozoa ancestrally had fruiting bodies that were lost many times (Shadwick et al., 2009) is extremely unlikely. The isolated positions of Copromyxa in Tubulinea (Fig. 2) and the protosteloid Protosteliopsis in Vannellida (Fig. S1) make it virtually certain that both evolved from non-stalked ancestors and the protosteloid state evolved independently in Protosteliopsis and variosean protosteloids. The protosteloid phenotype also evolved twice independently in Discosea (Endostelium, and an acanthamoebid: Shadwick et al., 2009), so it probably arose about 10 times in Amoebozoa. We agree with Shadwick et al. (2009) that Eumycetozoa should no longer be used as a taxon name for all Amoebozoa with fruiting bodies; we abandoned it before then, correctly considering Copromyxa as not mycetozoan . The close relationship of the protosporangid protosteloids to Ceratiomyxa makes it unlikely that their common ancestor (the ancestral exosporean) was a protosteloid in the sense of having single stalked spores (cysts), but there is no convincing evidence from sequence trees that Exosporeae are related specifically to any variosean protosteloid lineages; exosporean stalks probably evolved independently from them. As the assumption that Amoebozoa ancestrally had fruiting bodies is devoid of evidence, we strongly oppose the suggestion that it could be appropriate to replace the name Amoebozoa by Eumetazoa or any other merely because stalked cysts are more widely present in Amoebozoa than previously suspected (Shadwick et al., 2009). Such pointless replacement would be highly confusing.
Fiore-Donno et al. (2009) called dictyostelids, myxogastrids, and Ceratiomyxa collectively Macromycetozoa, made a superclass by Cavalier-Smith (2013). However, the strong evidence that unicellular protosporangids are within this clade makes the name descriptively less appropriate. Moreover, it becomes an unnecessary synonym of Mycetozoa now Protostelea is abandoned, so Table 2 discontinues its use. That protosporangids are now included and acrasids excluded is no obstacle to continued use of the historic name Mycetozoa for this taxon and clade, whose last common ancestor probably did have aerial spore dispersion. Until multigene trees are available for additional variosean orders it will not be clear what are the closest relatives of Mycetozoa. Our siteheterogeneous rDNA trees do not help resolve the branching order amongst the four conosan classes; on past homogeneous rDNA trees the long-branch Myxogastrea often branched weakly with or near Archamoebae which contain two especially long branches (Entamoeba, Pelomyxa) (Shadwick et al., 2009;Kudryavtsev and Pawlowski, 2013), but when these are excluded (Fig. 1) there is no tendency for Archamoebae to group with shorter branch Mycetozoa -thus Fig. 1 neither contradicts nor supports the possible holophyly of Semiconosia (Variosea plus Mycetozoa), which previous multigene trees did not unambiguously resolve (Cavalier-Smith et al., 2015b).
Our rDNA trees are also important for mapping character evolution onto dictyostelid cellular slime mould phylogeny, as they convincingly disprove the idea that small-spored Group 1 Dictyostelium (e.g. D. antarcticum, D. stellarum) are the most divergent clade (Schaap et al., 2006). Schaap et al. (2006) based that on two-gene trees with codon second nucleotides of a-tubulin given twice the weight of codon 1 nucleotides (codon 3 nucleotides omitted) plus 18S rDNA nucleotides (their supplementary Fig. S1C) with only nine outgroup species: three Variosea; six Discosea. That tree did not even group Dictyostelida with Variosea as our multigene trees robustly show to be correct, but weakly (0.7, 54%) with Thecamoeba similis, which our Fig. 1 shows is like Dictyostelida an extremely long branch, and thus a bad outgroup choice. Most likely the long-branch Group 1 Dictyostelida were drawn by the Thecamoeba sequence to the base of the dictyostelid clade by long-branch attraction artefacts, just as all dictyostelids were wrongly pulled towards it. In fact, their Fig. S3 two-gene analyses where Thecamoeba was omitted and three Variosea, one tubulinean, and four green plants were outgroups (more balanced), and their homogenous Bayesian tree weighting both tubulin codon positions and rDNA nucleotides equally, found the same topology as we did (with 0.85 support). Curiously they did not mention that, possibly because their ML and LogDet trees showed the same topology when tubulin second codon positions were single weighted as with double weighting. Their two-gene trees using a-tubulin amino acids (generally accepted as superior to using nucleotides) also had the same topology as we found (whether weighted the same, twice or thrice as rDNA nucleotides) so it is a mystery why their text entirely ignored that topology.
Their single-gene a-tubulin tree rooted on one protostelid and three myxogastrid (Physarum) sequences yielded a third topology with Group 2 the deepest branch (no support shown, so <50%). If one roots that tree between Groups 1/2 and Groups 3/4, as our trees show to be correct, Group 2 has the longest branch in dictyostelids, which may account for it being drawn artefactually towards the base.
As explained elsewhere (Cavalier-Smith, 2015), a-tubulin is not a good marker for single-gene trees as it has a huge diversity in evolutionary rates in numerous groups, just as does 18S rDNA in Amoebozoa. It is not surprising that combining data from two genes with two opposite long-branch artefacts (Schaap et al., 2006) gave three contradictory topologies (not just two as they implied) depending on algorithms, weighting, and choice of outgroups and nucleotides versus amino acids. Their analyses used only 1374 rDNA nucleotides; ours used 1470 and many more outgroup sequences and thus is expected to give much more accurate dictyostelid rooting. Though our Fig. S1 included fewer dictyostelids than theirs, as many have extremely similar sequences, our own ML trees including all dictyostelid sequences from Schaap et al. (2006) had exactly the same topology: the root is reliably between groups 1/2 and groups 3/4. This conclusion is important because Romeralo et al. (2011) ignored the contradictory results, outgroup limitations, and caveats of Schaap et al. (2006), assuming their erroneous conclusion of early divergence of Group 1 to be correct. We conclude that the almost uniformly small spores of Group 1 is probably secondary reduction from medium-spored ancestors, probably not the ancestral dicytostelid state. As cell size usually correlates with genome size (Cavalier-Smith, 2005), this clade likely also underwent genome reduction compared with most Conosa, possibly stemming from selection for smaller spores for better aerial dispersion and/or ability to generate more spores from a given biomass. However, Romeralo et al. (2011) noted that two not directly related new Australian group 1 species (D. myxobasis and boomerasporum) have much larger spores than most dictyostelids, indicating secondary increases in spore size; one wonders if they have an unusual derived ecology. Romeralo et al. (2012) recognised that previous attempts to root dictyostelids were equivocal and the root position then still uncertain; our more firmly rerooting the tree does not alter the majority of their conclusions. Our CAT trees weakly group Dictyostelium polycephalum with Group 3 containing D. caveatum, and somewhat more strongly reject its deeper branching than both groups 3 and 4 by ML (with only 58% support for this deeper position on ours) which could also be a long-branch artefact, so it may be better to include it in group 3 and not treat it as a separate clade (Schaap et al., 2006;Romeralo et al., 2011Romeralo et al., , 2012. CAT confirms that Polysphondylium violaceum is sister to group 4, and that Polysphondylium is polyphyletic (two independent origins of whorled branches) and Dictyostelium paraphyletic (i.e. dictyostelids ancestrally lacked such branches and originated by evolving cellular stalks). That makes it highly likely that Acytostelium is polyphyletic (lost stalk cell differentiation twice), not paraphyletic as Romeralo et al. (2012) suggested (contradicting their reasonable conclusion that Dictyostelium is paraphyletic not polyphyletic).
3.13. Significance of Cutosa and low mobility Variosea for the origin of Conosa and Amoebozoa Most Variosea except Flamellidae (Flamella, Telaepolella) are much less mobile than Glycopoda. Filamoebidae, non-ciliate members of Protostelida and Ramamoebida, and all trophic Phalansteriida except non-amoeboid Phalansterium itself are essentially non-mobile cells with highly branched tree-like pseudopodia Dyková et al., 2005Dyková et al., , 2010Berney et al., 2015). Though some Filamoeba species have a motile form (Page, 1967) a branched immobile form appears to predominate (Page, 1967;Dyková et al., 2005). Flamellidae however, alone in Conosa, have fan-shaped active locomotion (Kudryavtsev et al., 2009) indistinguishable from that of Vannellidae. This locomotory morphotype therefore apparently evolved twice independently in Amoebozoa, and is thus a rare derived condition within Variosea. Because the many uniciliate forms in Phalansteriida (Cavalier-Smith, 2013 and Table 2) do not show significant amoeboid movement, their relationship to Amoebozoa was only realised after we sequenced Phalansterium rDNA . Thus as Cavalier-Smith (2013) noted, Conosa were probably ancestrally largely immobile, like Sulcozoa using pseudopods mainly for feeding not locomotion as in most Glycopoda. The more mobile Flamella, Mycetozoa, and certain Archamoebea (e.g. Entamoeba and Tricholimax, with eruptive pseudopodia that evolved independently of each other and of Percolozoa) probably independently evolved greater pseudopodia-based locomotion.
That Cutosa are the most divergent Lobosa suggests that ancestral Lobosa also were low mobility protists, effective rapid pseudopodial locomotion of Glycopoda evolving secondarily significantly after loss of posterior ciliary gliding, the sulcozoan groove, and dorsal pellicle during the origin of Amoebozoa (Cavalier-Smith, 2013). Our site-heterogeneous rDNA trees are consistent with this; all disagree with a site-homogeneous two-gene tree weakly suggesting that Cutosa may be sister to Vexillifera (Lahr et al., 2015); Fig. 1 clearly excludes them from holophyletic Dactylopodina (Fig. 1, posterior probability 0.81). Fig. 1 weakly placed them as sister to all other Amoebozoa, like our much stronger multigene trees showing they are not Discosea. However, unconverged Fig. S1 including very long branch Amoebozoa did not recover holophyletic Dactylopodina and weakly placed Cutosea within Discosea as sister to Cochliopodium similarly to a Mr Bayes tree , who excluded Parvamoeba which wrongly attracted Ovalopodium away from Cochliopodium in Fig. S1, and wrongly grouped with Sappinia in Lahr et al., 2015). As neither rDNA tree consistently resolves the deepest branchings within Amoebozoa these contradictory positions are all meaningless -there are too many extremely closely spaced and too variable length basal amoebozoan branches for single-gene trees to be reliable. Even a six-gene tree that also showed a Cochliopodium grouping was basally totally unresolved (Lahr et al., 2015), making further trees with many scores of proteins and many more taxa than here essential for testing and extending our conclusions.

Major conclusions
1. Trichosida are not Discosea as recently supposed, but belong in Tubulinea and are here grouped with orders Euamoebida and Arcellinida as subclass Neolobosia. The transcriptomesequenced strain is probably not Trichosphaerium, so we described it as a new non-spiculate trichosid species Atrichosa algivora. 2. Nolandella and Copromyxa branch so closely together that we include both in Euamoebida in new suborders Nolandina and Amoebina. 3. Better taxon sampling of Discosea including all orders but one largely establishes the branching order of the main subgroups congruently with their diverse pseudopodial forms and cell coat ultrastructure. Five multiprotein trees strongly favour holophyly of Discosea, but one suggested paraphyly. 4. Vexilliferidae and Paramoebidae are robustly sisters within holophyletic Dactylopodina. The Perkinsela-like ichthyobodonid endosymbionts of both Paramoeba (parasomes) together form a robust prokinetoplastid clade that is firmly sister to metakinetoplastids within Euglenozoa. 5. Dactylopodids, Stygamoeba, and vannellids are a robust discosean clade here treated as revised order Glycostylida (ancestrally with cell coat of glycostyles) with these three groups ranked as suborders of contrasting locomotory shapes. The Stygamoeba culture used for transcriptome sequencing was heavily contaminated by 'Mayorella' sp. (actually Cunea) sequences but we identified genes for each component of the mixed culture unambiguously using 187 single-gene trees. 'Mayorella' sp. is not a Mayorella but a third, undescribed species of the paramoebid genus Cunea and robustly sister to Paramoeba. 6. Thecamoebida represented by Stenamoeba are sister to Glycostylida. 7. 'Stereomyxa' (misidentified; should be new genus) and Acanthamoeba are robust sisters. 8. Sapocribrum and PRA-29 are robust sisters, forming new order Squamocutida and class Cutosea characterised by a pellicle with embedded scale-like oval structures and unusual very slow moving pseudopodia. PRA-29 was probably misidentified as 'Pessonella' sp. and should be a new genus. Cutosea include Squamamoeba and are sisters of Discosea plus Tubulinea and thus the most divergent Lobosa. 9. Recognition of Cutosea together with the overall robustness of the better sampled amoebozoan multigene tree enabled substantially novel interpretations of lobosan evolution, making ultrastructure, pseudopodial form, and phylogeny more congruent, and simplifying discosean taxonomy. 10. Our trees raise the possibility that ancestral Lobosa were enveloped and naked groups (gymnamoebae) multiply derived from them by losing the envelope. 11. Our site-heterogeneous rDNA trees confirm polyphyly of Protostelea and Protostelida, so Protostelea is abandoned and Protostelida restricted to a robust clade transferred to Variosea. Other variosean protosteloids group with five morphologically and genetically related non-fruiting recently described variosean genera not previously assigned to orders; some are transferred to Phalansteriida with related non-fruiting amoebae; Cavosteliidae is grouped with others as new reticulose order Ramamoebida. A protosteloid clade closely related to Ceratiomyxida is made a new order of Mycetozoa: Protosporangida within expanded subclass Exosporeae of revised class Stelamoebea.
12. Our rDNA trees show that the position previously assumed for the root of Dictyostelida was incorrect because of an overlooked long-branch artefact. Group 1 dictyostelids are not the most divergent clade but are extra-long-branch sisters of Group 2; the root lies between Groups 1/2/Dictyostelium polycarpum and Groups 3/4.

Note added in proof
During proof correction another multigene amoebozoan prepublication appeared online: Tekle et al. (2016).
It convincingly groups Parvamoeba with Himatismenida as Smirnov et al. (2011) argued, even within Cochliopodiidae; Gocevia fonbrunei with Centramoebida not Himatismenida (confirming their independent origin of a dorsal tectum); and Vermistella with Thecamoebida not Glycostylida; and shows monophyly of Vannellina and the robust Glycostylida/Thecamoebida clade we called Flabellinia. Unfortunately it omitted Stygamoeba and Copromyxa; did not realise that 'Mayorella' is actually Cunea so mistakenly concluded that Dactylopodida is not a clade and therefore wrongly moved Mayorella from Dermamoebida to Dactylopodida; used no site-heterogeneous method; and ran no Amoebozoa-only trees. Resolution on their LG ML tree (not the better LGF) was far too weak to resolve amoebozoan deep phylogeny (neither revealing the deep branching of Cutosea nor even recovering Conosa or Lobosa clades) or to justify their conclusion of non-monophyly of Discosea; apparent topological differences from our better resolved trees (notably exclusion of Himatismenida from Discosea: 41% 'support') are statistically insignificant. 'Trichosphaerium' was with Tubulinea as in our trees, but less strongly. Consistently with our Table 2 not listing Unda separately, U. shaefferi Sawyer, 1975, Trans. Am. Micros. Soc. 94, p. 319 unambiguously nests within Vannella; it is just another Vannella (V. schaefferi (Sawyer) Cavalier-Smith comb. n.), as morphology hinted (Page, 1983).