Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The multiple evolutionary origins of the eukaryotic N-glycosylation pathway

The multiple evolutionary origins of the eukaryotic N-glycosylation pathway Background: The N-glycosylation is an essential protein modification taking place in the membranes of the endoplasmic reticulum (ER) in eukaryotes and the plasma membranes in archaea. It shares mechanistic similarities based on the use of polyisoprenol lipid carriers with other glycosylation pathways involved in the synthesis of bacterial cell wall components (e.g. peptidoglycan and teichoic acids). Here, a phylogenomic analysis was carried out to examine the validity of rival hypotheses suggesting alternative archaeal or bacterial origins to the eukaryotic N-glycosylation pathway. Results: The comparison of several polyisoprenol-based glycosylation pathways from the three domains of life shows that most of the implicated proteins belong to a limited number of superfamilies. The N-glycosylation pathway enzymes are ancestral to the eukaryotes, but their origins are mixed: Alg7, Dpm and maybe also one gene of the glycosyltransferase 1 (GT1) superfamily and Stt3 have proteoarchaeal (TACK superphylum) origins; alg2/alg11 may have resulted from the duplication of the original GT1 gene; the lumen glycosyltransferases were probably co-opted and multiplied through several gene duplications during eukaryogenesis; Alg13/Alg14 are more similar to their bacterial homologues; and Alg1, Alg5 and a putative flippase have unknown origins. Conclusions: The origin of the eukaryotic N-glycosylation pathway is not unique and less straightforward than previously thought: some basic components likely have proteoarchaeal origins, but the pathway was extensively developed before the eukaryotic diversification through multiple gene duplications, protein co-options, neofunctionalizations and even possible horizontal gene transfers from bacteria. These results may have important implications for our understanding of the ER evolution and eukaryogenesis. Reviewers: This article was reviewed by Pr. Patrick Forterre and Dr. Sergei Mekhedov (nominated by Editorial Board member Michael Galperin). Keywords: N-glycosylation, Eukaryotes, Archaea, Bacteria, Eukaryogenesis, Prokaryotic cell walls, Polyisoprenol, Glycosyltransferase Background contributor to eukaryogenesis, for at least the mitochondria At the dawn of eukaryotes are known to have evolved from an alpha-proteobacterium All living organisms are traditionally classified into one that was engulfed prior to the last eukaryotic common of three domains of life, namely bacteria, archaea and ancestor (LECA) [1, 2]. The traditional three-domain tree eukaryotes. Eukaryotic cells have important cytological of life posits that eukaryotes and archaea are sister groups specificities, such as nuclei, organelles and other cellular [3–5] and, thus, it assumes that the mitochondrial ances- processes. As a result, the origin of eukaryotes from tor was incorporated in a pre-existing proto-eukaryotic former organisms is one of the most intriguing questions lineage [6, 7]. Other phylogenetic studies have suggested in biology. Endosymbiosis was certainly an important that the eukaryotic stem branched within the archaeal domain, close to the crenarchaea–“eocytes”–or to the Correspondence: j.lombard@exeter.ac.uk archaeal TACK superphylum [8–11]. Also, comparative National Evolutionary Synthesis Center, 2024 W. Main Street Suite A200, genomics have shown that many eukaryotic operational Durham, NC 27705, USA 2 (metabolic) genes are closely related to bacterial Biosciences, University of Exeter, Geoffrey Pope Building, Stocker Road, Exeter EX4 4QD, UK © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lombard Biology Direct (2016) 11:36 Page 2 of 31 homologues, while informational genes are more similar [42]. Archaeal glycans are also more diverse than their to archaeal ones [12]. These observations popularized the eukaryotic equivalents [40]. opinion according to which the eukaryotes are a chimeric N-glycosylation is less common in bacteria, where two lineage that resulted from the endosymbiosis of the N-glycosylation systems are known [43]. One of them bacterial ancestor of mitochondria within a bona fide independently tethers single monosaccharides to specific archaeon [13, 14] or a previous bacterium/archaeon con- amino acids using soluble cytoplasmic proteins. The sortium [15–17]. The recent accumulation of genomic other relies on a lipid-carrier-based mechanism similar data has now made possible to address these issues. The to the eukaryotic and archaeal pathways (Fig. 1b). debate remains open [18–20], but most authors now favor The latter is thought to be restricted to delta- and an archaeal origin of the eukaryotic stem [21–24]. epsilon-proteobacteria [43, 44]. The bacterial polypre- Following the success of comparative analyses to trace nol lipid carriers are unsaturated and have different back particular machineries to LECA [25–32], the origin names depending on their lengths [45]. For simplicity, and evolution of the eukaryotic N-glycosylation pathway they will all be referred to as bactoprenol phosphate will be studied here. (Bac-P). N-glycosylation in the three domains of life The evolution of polyisoprenol-based N-glycosylations More than half of all eukaryotic proteins are glycopro- Since polyisoprenol-based N-glycosylation pathways are teins, and 90 % of those are N-glycosylated [33]. Protein widespread in archaea and eukaryotes but uncommon in N-glycosylation modulates protein stability, solubility, bacteria, it has been speculated that this pathway origi- rigidity, orientation, interactivity, transport and signaling nated in archaea, was horizontally transferred to some [34]. It consists of the covalent attachment of an oligo- bacteria [46, 47] and inherited in eukaryotes from their saccharide to the nitrogen atom of specific asparagine putative archaeal-related progenitor [34, 48]. The poten- residues. In eukaryotes, the synthesis of the oligosac- tial archaeal origin of eukaryotic N-glycosylation recently charide core is mediated by a lipid carrier called brought this pathway into the spotlight of eukary- dolichol-phosphate (Dol-P) which is located in the ogenesis debates [14]. For instance, the “inside-out membranes of the endoplasmic reticulum (ER, Fig. 1a). hypothesis” recently suggested that the eukaryotic In opisthokonts (e.g. in humans and yeast), several nucleus evolved from ancestral “eocytes” (crenarchaeota) cytoplasmic-facing membrane-embedded glycosyltrans- that protruded membrane-bound blebs beyond their cell ferases (GTs) sequentially add monosaccharides from walls to increase the available exchange surface with the soluble nucleotide carriers to Dol-P molecules (Fig. 1b). bacterial ancestors of mitochondria. According to this The Dol-P-oligosaccharide is then “flipped” across the hypothesis, the location of the N-glycosylation pathway membrane and other membrane-embedded GTs resume in the eukaryotic ER membranes is a vestige of the the oligosaccharide decoration in the lumen side. The ancestral archaeal S-layer protein N-glycosylation [14]. monosaccharides used in the ER lumen are also attached The idea that eukaryotic N-glycosylation originated to Dol-P carriers. The final oligosaccharide is transferred from archaea is plausible, but alternative hypotheses en bloc to the acceptor protein located in the ER lumen have also been suggested. For instance, the eukaryotic by a N-oligosaccharyltransferase (N-OST). Soluble GTs N-glycosylation has been thought to be related to some located in the ER and Golgi lumen may subsequently kind of bacterial cell wall synthesis pathway [49–51]. modify the protein N-glycans [35], but these later modi- Indeed, bacterial O-glycosylation, tyrosine O-glycosylation, fications will not be studied in the present analysis. peptidoglycan synthesis, O-antigen LPS synthesis, teichoic In archaea, flagellins and cell wall S-layer proteins are acid synthesis, exopolysaccharide synthesis and, in a N-glycosylated using similar pathways located in their smaller scale, O- and C-mannosylation, use polyisoprenol cell membranes [36, 37]. Archaeal N-glycosylation lipid lipid carriers and a similar orientation across the carriers are traditionally called Dol-P, even though they membrane to protein N-glycosylation. Other observations, are more saturated than the eukaryotic ones (Fig. 1a, without being conclusive, call into question a simple [38–41]). Archaeal N-glycosylation pathways have a transposition of the archaeal N-glycosylation into eukary- topology similar to their eukaryotic counterparts, i.e. the otes: i) archaeal and eukaryotic Dol-P are different orientation of their reactions across the membrane is the molecules (Fig. 1a) [38–41] and they are probably same: archaeal N-glycans are synthesized in the cytoplas- synthesized using different biosynthesis pathways (data to mic face, then flipped to the outer side of cell mem- be published soon; Lombard, data not shown); and ii) branes and transferred to the protein (Fig. 1b, [37]). eukaryotic cytoplasmic GTs have homologues in Contrary to eukaryotes, however, the N-glycan chains prokaryotes but the precise evolutionary relationships are seldom modified in the outer side of archaeal mem- between eukaryotic and prokaryotic sequences are not branes, with the notable exception of Haloferax volcanii known yet [46]. Lombard Biology Direct (2016) 11:36 Page 3 of 31 Fig. 1 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 4 of 31 (See figure on previous page.) Fig. 1 Basic components of the polyisoprenol-based machineries. a Typical polyisoprenol lipid carriers in the three domains of life. b Topology (membrane orientation) of characteristic N-glycosylation pathways in the three domains of life and bacterial peptidoglycan biosynthesis. The lipid carrier is embedded in a cell membrane (ER membrane in eukaryotes, plasma membranes in archaea and bacteria), first facing the cytoplasmic side, then flipped to the opposite side (i.e. ER lumen in eukaryotes, periplasm in prokaryotes). Monosaccharides are attached one by one to the lipid carriers by specific glycosyltransferases, although in the eukaryotic N-glycosylation each kind of sugar is only represented once per compartment, for simplicity. The monosaccharides are nucleotide-activated in the cytoplasmic side or translocated to the ER lumen by separate lipid carriers In fact, the available evidence to support a bacterial or in some lineages [46]. In parasites, such losses may argu- archaeal origin of the eukaryotic N-glycosylation path- ably have been a convenient mechanism to synthesize way is limited, so a phylogenomic analysis was carried non-canonical N-glycans and escape the host defenses. out to tackle this issue. The phylogenetic study of the in- volved protein superfamilies suggests that this pathway Gene homology across polyisoprenol-based machineries was present in LECA and that its components have The main argument put forward to suggest bacterial or diverse origins. Some genes likely had archaeal ancestors– archaeal origins to the eukaryotic N-glycosylation especially from the TACK superphylum, to which I will pathway has been the use of polyisoprenol carriers and refer to as proteoarchaea [52]. Although this gives credit topological similarity, i.e. the eukaryotic pathway has the to hypotheses invoking a close relationship between eu- same functional orientation across the membrane than karyotes and proteoarchaea, many N-glycosylation genes several prokaryotic glycosylation pathways (Fig. 1b). specifically evolved in eukaryotes through gene duplica- Despite these similarities, the homology of the proteins tion, protein co-option and neofunctionalization. Two of implicated in these pathways had never been systematic- them could even have bacterial origins. Given the import- ally surveyed before. ance that some eukaryogenesis scenarios have conceded Most proteins involved in polyisoprenol-based pathways to the origins of the N-glycosylation pathway and its are glycosyltransferases (GTs). These enzymes are very com- presence in the ER membranes, these results may have mon and they are classified according to their structure, important consequences in debates about the origins of catalytic mechanism and amino acid sequences [55–58]. An eukaryotes and the ER. updated classification is available in the CAZy database (http://www.cazy.org/GlycosylTransferases.html) which, as Results and discussion of August 2016, contained 100 GT families. Nonetheless, Distribution of the N-glycosylation pathway proteins their extreme diversity, poor sequence conservation and across eukaryotes ability to easily change their substrate specificity make an The set of N-glycosylation GTs was reported to be very extensive in-depth evolutionary study of all GTs across variable among eukaryotes [46, 53], therefore raising the the three domains of life virtually impossible. question if this pathway was ancestral to all eukaryotes The pathways summarized in Table 2 were selected for or progressively developed during eukaryotic evolution this study because, according to the literature, they use [46, 48]. The recent accumulation of genomic data polyisoprenol lipid carriers and have similar topologies allows us to provide a more comprehensive perspective to the eukaryotic N-glycosylation pathway. The eukar- on the distribution of these genes across eukaryotes yotic GPI pathway was included notwithstanding being (Table 1, Additional file 1). With regard to the N-OST, polyisoprenol-independent because its mannosyltrans- the Stt3 catalytic unit is well conserved and most other ferases are the only known relatives of the eukaryotic subunits–except for those very short and arguably more N-glycosylation lumen GTs [59]. difficult to detect–are found in all major eukaryotic The proteins in the pathways from Table 2 were used to groups but haptophytes and cryptophytes. GTs and carry out all vs. all reciprocal BLASTp searches [60]. The putative flippase orthologues are found in all major results, together with some information from the litera- eukaryotic lineages. The cytoplasmic GTs are more ture and occasional HMMER confirmations, allowed the conserved than their lumenal counterparts. This is in construction of the putative homology groups reported in agreement with the observation that lumen GTs dele- Fig. 2 (see Methods). This dataset is limited to the path- tions have milder phenotypes than cytoplasmic GTs mu- ways characterized in model organisms and does not have tations [54]. The widespread distribution of the canonical a strong statistical value. Yet, it shows that most proteins N-glycosylation pathway proteins in eukaryotes and their from polyisoprenol-dependent pathways belong to a lim- consistent monophyly in protein family trees (see below) ited number of homology groups. For instance, the GTs support the ancestral presence of this pathway in LECA and colored in blue (GT group 1), red (GT group 2) and green the subsequent independent loss of some of these enzymes (HPT family), plus the khaki flippase group, are present in Lombard Biology Direct (2016) 11:36 Page 5 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes Alg7 Alg13 Alg14 Alg1 Alg2 Alg11 Rft1 Alg3 Alg9 Alg12 Alg6 Alg8 Alg10 Stt3 Ost1 Ost2 Ost3/6 Ost4 Ost5 Wbp1 Swp1 Dmp1 Alg5 Saccharomyces cerevisiae ++ + + ++ + ++ + ++ + + ++ ++ + ++ + + + Schizosaccharomyces pombe ++ + + ++ + ++ + ++ + + ++ + + + + + + Ustilago maydis ++ + + ++ + ++ + ++ + + ++ + + + + + Cryptococcus neoformans + + + ++ + + + ++ + + ++ + + + + Rhizophagus irregularis ++ + + ++ + ++ + ++++ + + ++ + + + + + Batrachochytrium dendrobatidis ++ + + + ++ + ++ + ++ + + ++ + + + + + Spizellomyces punctatus ++ + ++ + ++ ++ ++ + ++ + + + + + Allomyces macrogynus ++ ++ ++ ++ + ++ ++ ++ + ++ ++ ++ ++ ++ ++ ++ ++ ++ + ++ ++ Trichoplax adhaerens + +++ +++ + +++ + +++ + + + + + + + Amphimedon queenslandica + +++ + + + + + ++ + + + +++ + 5+ + + +++ + + + + + Mnemiopsis leidyi ++ + + ++ + + + + ++ ++ + + + + Hydra magnipapillata ++ + + ++ + + + + ++ ++ + + + + Hydra magnipapillata ++ + + ++ + ++ + ++ + +++ ++ ++ + + + + Drosophila melagonaster ++ + + ++ + ++ + ++ + ++ ++ + ++ + + + + Aplysia californica ++ + + ++ + ++ + ++ + +++ ++ + + + + + ++ Strongylocentrotus purpuratus + + + ++ + + ++ + + + + + + +++ + + ++ + + + +++ Mus musculus +++ + ++ +++ + ++ ++ ++ + ++ ++ ++ + + ++ + + Monosiga brevicollis + +++ + ++ + + + ++ + Salpingoeca rosetta + ++++ + ++ + + ++ + +++ + + + + + + + Capsaspora owczarzaki ++ + + ++ + ++ + +++ + ++ ++ + + + + ++ + Fonticula alba ++ + + ++ + ++ + ++ + + + + Thecamonas trahens ++ + + ++ + ++ + ++ + ++ ++ + + + + + Entamoeba histolytica + + + + ++ + + +++ ++ + +++ + Acanthamoeba castellanii + +++ + ++ + + + + ++ + + + + + Dictyostelium discoideum + +++ + ++ ++ + ++ + ++ + + + + + + + ++ Polysphondylium pallidum + +++ + ++ ++ + ++ + ++ + + ++ + + + + + Chlamydomonas reinhardtii + +++ + ++ + + ++ + + + + + + ++ Volvox carteri ++ + + ++ + ++ + ++ ++ + + + + + Chlorella variabilis + +++ + ++ + ++ + ++ + + + + + + + Asterochlorissp. ++ + + ++ ++ + ++ + ++ + + + + Bathycoccus prasinos + +++ + ++ + + + + + Micromonas pusilla ++ + + ++ ++ + ++ ++ + + Ostreococcus lucimarinus + +++ + ++ + + + + + + Lombard Biology Direct (2016) 11:36 Page 6 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes (Continued) Klebsormidium flaccidium ++ + + ++ + ++ + ++ + ++ ++ + + + + + Physcomitrella patens + + + + + ++ ++ + + ++ + + ++ +++ + ++ +++ + + + ++ +++ Selaginella moellendorffii 4+ + ++ + + + ++ + + + + + + ++ ++ + 4+ ++ ++ + 4+ Pinus taeda + +++ + + + + 5+ + ++ + + + + Amborella trichopoda ++ + + ++ + ++ + ++ + ++ ++ + + + + + + + Oryza sativa + ++++ + ++ + + ++ + +++ ++ + + + + + + + Aquilegia coerulea +++ ++ + +++ ++ +++ ++ + + ++ ++ + + ++ +++ ++ + ++ +++ +++ ++ ++ Prunus persica ++ ++ + ++ ++ + ++ + + + ++ + ++ ++ +++ + + ++ + + + Arabidopsis thaliana ++ + + ++ + + + + + ++ + + + ++ ++ +++ ++ ++ + ++ + ++ Calliarthron tuberculosum ++++ + ++ +++ ++ + + Chondrus crispus ++ + + + + + + + + + + + + + + + Cyanidioschyzon merolae ++ + + ++ + ++ + ++ + + ++ + + + Galdieria sulphuraria + + + + ++ + + ++ ++ ++ + + + + + + Porphyridium purpureum + + + + + + ++ + + + + +++ + + + + + + Pyropia yezoensis +++ +++ + + Cyanophora paradoxa + + ++ + ++ ++ + ++ + ++ + + ++ + Phaeodactylum tricornutum ++ + + ++ + ++ + ++ ++ + 4+ Blastocystis hominis ++ + + +++ + ++ + ++ + + ++ + + + + + Phytophthora parasitica ++ + + ++ + ++ + +++ ++ ++ ++ ++ + + + + Schizochytrium aggregatum ++ + + ++ + ++ + ++ + ++ + Aureococcus anophagefferens + +++ + + + + + ++ + + Plasmodium falciparum ++ + + + + + Babesia equi + +++ + + + Theileria annulata + ++ Toxoplasma gondii + + + ++ + ++ + + ++ + + ++ Gregarina niphandrodes ++ + ++ + + Ichthyophthirius multifiliis ++ + ++ + + + + + + + + Tetrahymena thermophila + + + ++ + + + ++ + + ++ + + + + Oxytricha trifallax ++ ++ + ++ + + + ++ + ++ + + + + Paramecium tetraurelia + + ++ + + + + + + + + ++ ++ ++ ++ + ++ ++ Symbiodinium minutum + + + ++ ++ +++ + 6+ + + Perkinsus marinus ++ + ++ + ++ ++ ++ + ++ + ++ ++ + ++ ++ + +++ ++ Bigelowiella natans + + + ++ + + + ++ + + + + ++ + + + + Lombard Biology Direct (2016) 11:36 Page 7 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes (Continued) Reticulomyxa filosa + + + + + + + + + + ++ ++ + + + + + + + Bodo saltans ++ + + ++ + ++ + ++ + 4+ + + Angomonas deanei +++ ++ + + ++ 5+ + Leishmania major + +++ + ++ + + 4+ + + Trypanosoma vivax + + + + + + + + + + +++ + Giardia lamblia ++ + + + Spironucleus salmonicida ++ + + Naegleria gruberi ++ + + ++ + ++ + ++ + + ++ + + + + + Trichomonas vaginalis + + + +++ + + + 5+ + + + + 5+ Emiliania huxleyi + + + ++ + ++ + + + + 4+ + 5+ Guillardia theta ++ + + ++ + ++ + ++ + +++ + + Failure in homologue detection may result from genome incomplete sequencing. (+) Presence of homologues in a genome (two or more (+) indicate as many paralogues). (F) on Alg14/Alg13 point out organisms in which the two genes are fused Lombard Biology Direct (2016) 11:36 Page 8 of 31 Table 2 Polyisoprenol-related pathways used in this analysis and respective literature N-glycosylation Eukaryotes (S. cerevisiae) KEGG: http://www.genome.jp/kegg/ GPI synthesis Eukaryotes (S. cerevisiae) KEGG: http://www.genome.jp/kegg/ O-mannosylation Eukaryotes (H. sapiens) KEGG: http://www.genome.jp/kegg/ [87] N-glycosylation Archaea (S. acidocaldarius)[45] N-glycosylation Archaea (M. maripaludis)[36, 40] N-glycosylation Archaea (M. voltae)[36, 71] N-glycosylation Archaea (A. fulgidus)[50] N-glycosylation Archaea (H. volcanii)[36, 40] N-glycosylation Archaea (H. volcanii, alternative) [113] N-glycosylation Bacteria (C. jejuni)[43] Peptidoglycan synthesis Bacteria (E. coli)[114] LPS synthesis (ABC transporter) Bacteria (E.coli)[115] LPS synthesis (wzy-dependent) Bacteria (S. enterica)[115] Wall teichoic acids Bacteria (B. subtilis)[116] O-glycosylation Bacteria (N. gonorrhoae)[117] O-glycosylation Bacteria (G. stearothermophilus)[118] Tyrosine O-glycosylation Bacteria (P. alvei)[101] Enterobacterial common antigen Bacteria (E. coli)[119] Capsule synthesis Bacteria (E. coli)[120] EPS - xanthan Bacteria (X. campestris)[121] EPS - succinoglycan Bacteria (R. meliloti)[122] EPS Bacteria (L. lactis)[73] pathways from the three domains of life, including the The eukaryotic N-glycosylation initiation complex has a eukaryotic N-glycosylation pathway. Some pathways from composite origin closely related organisms contain the same homology The canonical eukaryotic N-glycosylation pathway starts groups. This is for example the case of the O-glycosylation with the sequential link of two N-acetylglucosamines in G. stearothermophilus and Tyr. O-glycosylation in (GlcNAc) from soluble UDP-GlcNAc to Dol-P (Fig. 1, P. alvei (two Firmicutes);N-glycosylation in H. volcanii [61, 62]). The resulting GlcNAc dimer is named chito- and M. voltae (two Euryarchaeota); or O-antigen ABC- biose. The two GlcNAc transfers are consecutively carried dependent synthesis in E. coli and N-glycosylation in C. out by Alg7 and Alg13, which form a protein complex with jejuni (two Proteobacteria). It is more difficult to ascertain the membrane anchor Alg14 [51, 63, 64]. Alg7 belongs to clear parallels across pathways from distantly-related the polyprenol phosphate:N-acetylhexosamine-1-phosphate organisms. Nevertheless, some proteins of the eukaryotic transferase (HPT) family (dark green in Fig. 2, [65–67], N-glycosylation pathway belong to the same homology which also includes the bacterial MraY protein. Alg14 groups than proteins implicated in the peptidoglycan and Alg13 are respectively homologous to the N- and synthesis, LPS synthesis or putative N-glycosylation in C-terminal parts of bacterial MurG (purple in Fig. 2, [62]). A. fulgidus. The most exciting parallel is to be drawn MraY, MurG and other proteins form a complex involved between the eukaryotic and putative S. acidocaldarius in peptidoglycan synthesis (Fig. 2, [51, 68]). The resem- N-glycosylation pathways. This Sulfolobus pathway blances between the eukaryotic Alg7-Alg13-Alg14 and the has not been fully described yet [45], but it is the bacterial MraY-MurG complexes [69] could support the only prokaryotic mechanism studied here with homo- existence of a close evolutionary relationship between the logues of all eukaryotic cytoplasmic GTs and a cata- eukaryotic N-glycosylation and the bacterial peptidoglycan lytic N-OST subunit. This result is suggestive because synthesis pathways [49, 51]. Although most archaeal many eukaryogenesis scenarios assume a tight evolu- N-glycosylation initiation mechanisms are based on GTs tionary relationship between proteoarchaea and eu- from group 2 (red GTs in Fig. 2, [36]), HPT homologues karyotes [8–10, 14] but it must be balanced with the have been reported in several archaea [45, 70]. For phylogenetic analyses presented below. instance, the S. acidocaldarius genome contains both a Lombard Biology Direct (2016) 11:36 Page 9 of 31 Fig. 2 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 10 of 31 (See figure on previous page.) Fig. 2 Polyprenol-based glycosylation pathways (and GPI biosynthesis) colored according to detected homology groups. Horizontal lines represent ER membranes in the eukaryotes, plasma membranes in prokaryotes. Horizontal rectangles represent cytoplasmic glycosyltransferases (GTs) if they are below the membrane or lumen/periplasmic GTs if they are above the membrane. Vertical rectangles depict flippases or translocation mechanisms. Ovals represent the oligosaccharide transferases from the lipid carrier to the acceptor molecule. Diamonds portray proteins that are neither GTs nor translocases (e.g. acetyl or ethanolamine transferases in GPI biosynthesis). Extra shapes in the eukaryotic oligosaccharide transferases reflect the fact that these are complexes with many subunits. The cytoplasmic GTs depicted after a transfer to the acceptor molecule (e.g. Dpm1 and Alg5 in eukaryotic N-glycosylation) represent polyisoprenol-P-monosaccharide synthases tranfering single mannoses or glucoses to a lipid carrier to supply these sugars to the opposite side of the membrane. Proteins are colored according to the homology group to which they belong (as defined in Methods). Plain symbols represent proteins that were detected using the procedure described in Methods, whilst empty white shapes show more distant relationships that required bibliographic or extra analyses to be established. Empty transparent shapes represent the lack of detection of any homologues in the dataset HPT and a murG homologue, and the N-glycans in this family, one of which was lost in the euryarchaeal lineage. organism are also initiated with a chitobiose [45]. Despite Alternatively, the rampant paraphyly of this tree may be these promising similarities between the eukaryotic and indicative of some reconstruction artifacts due to the the putative S. acidocaldarius N-glycosylation pathways lack of phylogenetic signal or inappropriate phylogenetic (Fig. 2, [45]), the actual role of the archaeal HPT proteins reconstruction models. In that case, the crenarchaeal is debated [70, 71] and the MurG homologue remains gene could result from a phylum-specific duplication, uncharacterized to this date. and the basal branching of the proteoarchaeal group The phylogeny of each protein involved in the initiation could result from a long branch attraction artifact. complex was carried out. That of the HPT superfamily Regardless of the origin of the two proteoarchaeal para- shows two large groups (Bayesian Posterior Probability, logues, at least one representative of the HPT family BPP = 1), one mostly bacterial and the other including seems to have been present in both the last common most archaeal and eukaryotic sequences (Additional ancestor of euryarchaea and proteoarchaea and, thus, file 2). The phylogeny of the bacterial clade is discussed possibly in LACA. The eukaryotic Alg7 orthologues un- in detail the Additional file 2 but, in summary, it supports ambiguously branch within the paraphyletic crenarch- the idea that at least two proteins from the HPT aeal group–a group that also includes S. acidocaldarius superfamily were present in the Last Bacterial Common putative AglH (HPT) protein. Yet, there are no obvious Ancestor (LBCA). Here, we will center our attention on closest crenarchaeal relatives to the eukaryotic clade. the other cluster, for it contains the eukaryotic sequences. Thus, although the eukaryotic Alg7 genes likely have a A specific phylogenetic analysis only focusing on this proteoarchaeal origin, the precise identity of this ances- part of the tree was carried out (Fig. 3). It shows a tor remains uncertain. monophyletic group of eukaryotes (BPP = 1) in which se- Regarding the MurG/Alg14/Alg13 homologues, several quences cluster according to meaningful taxonomic occurrences of split or fused genes exist. For instance, the groups, such as paraphyletic groups of unikonta, viridi- heteromer EpsE/EpsF involved in the exopolysaccharide plantae or alveolates. A eukaryotic-only phylogeny re- (EPS) synthesis of firmicutes [72, 73] has been reported to vealed a better resolution and an overall topology in be split into the same protein domains as Alg14/Alg13 accordance with the main eukaryotic taxa (e.g. mono- [62]; in the contrary, excavates and some amoebozoa have phyly of unikonts, archaeplastida, excavates, etc. Data fused Alg14-Alg13 polypeptides (Table 1). In order to ac- not shown). This suggests that the lack of resolution count for these different architectures, independent phy- among eukaryotes in the Fig. 3 may be due to recon- logenies were constructed for each protein domain: struction artifacts owing to the extreme divergence be- catalytic (Alg13/C-terminal part of MurG) on the one side tween the prokaryotic and the eukaryotic sequences. and membrane anchor (Alg14/N-terminal part of MurG) The wide distribution of these genes in eukaryotes and on the other. In both phylogenies, MurG homologues the phylogenetic clustering of the eukaryotic sequences were widespread among bacteria and formed a highly according to their taxonomic groups support the pres- divergent clade (BPP = 0.99 in both cases, Fig. 4a, ence of Alg7 in LECA. Moreover, the eukaryotic clade Additional files 3 and 4). The bacterial MurG phylogeny is branches within a paraphyletic group of crenarchaea discussed in more detail in the Additional files 3 and 4 (BPP = 1) which itself branches in a paraphyletic group but, in short, MurG was likely already present in the of euryarchaea (BPP = 0.86). A proteoarchaeal clade LBCA. forms a sister group to the rest of archaeal and Phylogenies excluding the bacterial MurG homologues eukaryotic sequences (Fig. 3). The existence of the basal were carried out to more precisely tackle the evolution proteoarchaeal gene may suggest that the Last Archaeal of the closest prokaryotic relatives to the eukaryotes Common Ancestor (LACA) had two paralogues of this (Fig. 4b and Additional file 5). This includes archaeal Lombard Biology Direct (2016) 11:36 Page 11 of 31 Fig. 3 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 12 of 31 (See figure on previous page.) Fig. 3 Bayesian phylogeny of the HPT homologues. a Schematic phylogenetic tree including the bacterial sequences (see Additional file 2 for details). b Specific phylogeny of the archaeal/eukaryotic clade. The tree was reconstructed using 124 representative sequences and 203 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) fused sequences (referred to as MurG-like) and bacterial topology of the tree of life, the cenancestor most likely and archaeal split (EPS-like) sequences. The independent had a fused MurG-like protein. This is not the case of phylogenies of both the catalytic and membrane anchor the split proteins, of which the distribution among pro- domains show similar results to each other: three clades karyotes is too scarce to support their presence in the corresponding to 1) the archaeal fused MurG-like cenancestor. Since, contrary to bacteria, the eukaryotes homologues (BPP Alg13 = 0.94, BPP Alg14 = 0.67); 2) do not have other homologues of the Alg13/Alg14 pro- the bacterial and archaeal split EPS-like homologues teins, the most parsimonious hypothesis for the origin of (BPP Alg13 = 0.92); and 3) the eukaryotic proteins the split genes would be that epsE/epsF developed in the (BPP Alg13 = 0.7, BPP Alg14 = 0.87). The fused MurG-like bacterial lineage from the duplication and subsequent archaeal clade is divided into a crenarchaeal group and a split of their ancestral murG gene and were later ac- euryarchaeal group with some basal thaumarchaeota and quired in an early eukaryotic ancestor from an unidenti- bathyarchaeota. Despite the paraphyly of the proteoarch- fied bacterium. Once in the eukaryotic lineage, they aeal sequences (i.e. the crenarchaeal and thaumarchaeal evolved into the alg13/alg14 couple. sequences in this tree), the wide distribution of these pro- In summary, no evidence was found to support a teins in archaea and the monophyly of the crenarchaeal direct relationship between the bacterial peptidoglycan and euryarchaeal sequences supports the presence of this synthesis and the eukaryotic N-glycosylation pathway. gene in LACA. The internal topology of the eukaryotic The eukaryotic HPT was present in LECA and likely clade is very mixed and very poorly resolved even when inherited from proteoarchaea. The Alg13/Alg14 couple phylogenies are reconstructed only with eukaryotic se- is structurally more similar to their bacterial homologues quences (data not shown). This could translate a compli- and, thus, could have a bacterial origin. These results high- cated non-vertical evolutionary history for these genes but light the mixed origin of the eukaryotic N-glycosylation it most likely results from the small number of alignment initiation complex. Furthermore, since at least one repre- positions that was conserved in each phylogeny (116 and sentative of both the HPT and MurG families was present 97 respectively) and the fast evolution rate of these pro- in the respective common ancestors of both prokaryotic teins, especially the membrane anchor. Despite these diffi- domains, their presence may also be inferred in the cenan- culties, the wide distribution of Alg13/Alg14 orthologues cestor. This suggests that the cenancestral membranes across eukaryotes and their monophyly in all the phyloge- may have had the basic elements of a glycosylation ma- nies suggest that these proteins may have been ancestral chinery. Given the fact that most modern prokaryotic to eukaryotes. Finally, the split EPS-like prokaryotic se- polyisoprenol-based glycosylation pathways are involved quences are dominated by bacteria. The archaeal split in the synthesis of prokaryotic cell wall components, this EPS-like sequences are few and scattered among the bac- may suggest that the cenancestor was already bound by terial sequences, so they certainly result from HGTs. The some kind of glycosylated envelope, of which the specific EPS-like bacterial sequences form meaningful taxonomic nature is unknown. The implications of these results for groups, but their distribution across bacteria is not very the possible cenancestral cell walls will be discussed in a diverse. Thus, unlike MurG, the ancestral presence of paper to be published soon (Lombard, data not shown). EPS-like genes in the LBCA is uncertain. The relationship of the eukaryotic clades to their pro- Cytoplasmic mannosylation evolved from an archaeal karyotic relatives is more difficult to establish. The protein eukaryotic sequences form separate clades with regard After the chitobiose core, five mannoses are tethered to to the other domains of life, so these phylogenies could the growing glycan in the cytoplasmic face of the ER be discussed in terms of a three-domain of life topology. membrane (Fig. 1b). The first mannosylation is generally This line of thought would suggest that the eukaryotic ascribed to Alg1, the second and third to Alg2 and the Alg13/Alg14 proteins were inherited from the last com- two last to Alg11, but their activity depends on the forma- mon ancestor of extant organisms (i.e. the cenancestor tion of respective protein complexes: Alg1 homooligo- or LUCA) through a proto-eukaryotic lineage. Neverthe- mers, respective Alg2-Alg1 and Alg11-Alg1 heteromers less, if LBCA and LACA had fused MurG proteins, as [54]. All three proteins belong to the glycosyltransferase 1 this analysis supports, it follows that whatever the superfamily (GT1), even though Alg1 is classified as a Lombard Biology Direct (2016) 11:36 Page 13 of 31 Fig. 4 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 14 of 31 (See figure on previous page.) Fig. 4 Bayesian phylogeny of the Alg13/catalytic MurG domain. a Schematic phylogenetic tree including the bacterial MurG homologues (see Additional file 3 for details). b Specific phylogeny excluding the bacterial MurG clade. The tree is unrooted and was reconstructed using 132 representative sequences and 116 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) separate family in the CaZY database [54]. These both genes were inherited from LECA. The Alg2 se- eukaryotic genes were suggested to have archaeal ances- quences form a sister group to a clade made up of the tors [50], but early phylogenetic analyses were unable to eukaryotic Alg11 and most proteoarchaeal sequences. support this hypothesis. Thus, these proteins had also Several scenarios may explain this topology. One possi- been postulated to be eukaryotic-specific [46]. bility is that an ancestor of the proteoarchaea horizon- The GT1 superfamily is widely represented in most tally acquired one of these genes from eukaryotes early polyisoprenol-based pathways in the three domains of in proteoarchaeal evolution. Yet, since these homologues life (dark blue proteins in Fig. 2). The high number of are present in a wide diversity of archaea and form re- homologues of the GT1 enzymes detected in the three spective monophyletic groups of euryarchaea and pro- domains of life proved challenging to handle, so pre- teoarchaea, it seems likely that at least one of these liminary analyses were carried out to exclude the most proteins was present in LACA. In that case, interpreta- distant GT1 homologues from the analysis and concen- tions depend on the topology of the tree of life which is trate in the closest relatives to the eukaryotic sequences favored. In the context of a traditional tree of life, in (see Methods). These preparatory analyses showed that which archaea and eukaryotes are sister groups, the top- the Alg1 sequences formed a monophyletic group but ology could be explained by a hidden paralogy: the last their branching position within the superfamily tree was common ancestor of both domains had two paralogues, highly inconsistent and depended on the sequences the eukaryotic linage kept both of them and the pro- selected for the analysis (data not shown). As a result, teoarchaea and euryarchaea alternatively lost one of the closest prokaryotic relatives of the eukaryotic alg1 them. Nevertheless, the requirement of Alg1 to form could not be determined. The origin of this gene could functional heteromers with Alg2/Alg11 contradicts the not be assessed, and this may suggest that Alg1 was a ancestral paralogy of these proteins in archaea, as Alg1 eukaryotic innovation. A tentative explanation of the is a eukaryotic innovation. Alternatively, the ancestor of high Alg1 divergence–even when compared to the out- the eukaryotic (and bacterial) sequences that branch standing diversity of GT1 members–could be the role of within the archaeal clade could have been acquired from Alg1 in the formation of the mannosyltransferase com- archaea. Since Alg11 is more closely related to the pro- plexes [54]. teoarchaeal sequences, it seems more likely that the ori- Contrary to Alg1, Alg2 and Alg11 are closely related ginal eukaryotic GT1 gene had proteoarchaeal origins, a to each other and to some prokaryotic sequences. Their hypothesis that would also be congruent with the cur- evolution was studied together with their closest rently favored symbiogenic origin of eukaryotes [21–24]. prokaryotic relatives (see Methods). The outgroup of this In that case, the original eukaryotic GT1 gene would analysis is made up of some sequences from plastid- have been duplicated and acquired their alg2/alg11 func- bearing eukaryotes and cyanobacteria (Fig. 5). The first tions in eukaryotes. The fact that both eukaryotic clades sequences to diverge from the outgroup are some do not form relative sister groups within the proteoarch- Methanobacteriales and a group dominated by crenarch- aea could be a reconstruction artifact due to the muta- aea. The position of these sequences frequently changed tional saturation that resulted from the development of in the preliminary trees (data not shown) and their basal heteromers and new functions in eukaryotes. position in the Alg2/Alg11 tree calls into question their The characterization of the closest archaeal relatives to closeness to the proteins that interest us. Nevertheless, the eukaryotic Alg2/Alg11 proteins will be instrumental to the lack of characterization of these genes makes it diffi- the evaluation of these different evolutionary hypotheses. cult to make an informed decision to exclude them from ORF43 has been suggested to be involved in N-glycosylation the analysis. The rest of the tree is split in four clades: a in A. fulgidus but, in fact, the closest archaeal relatives to euryarchaeal clade also including several bacterial groups Alg2/Alg11 are uncharacterized GT1 proteins [50]. Agl16, (BPP = 0.98), a proteoarchaeal clade (BPP = 1) and two a GT1 protein implicated in N-glycosylation in S. eukaryotic clades respectively corresponding to Alg2 acidocaldarius [74] is not present among the closest (BPP = 0.95) and Alg11 (BPP = 1). The deep relationships relatives of Alg2/Alg11. Similarly, all the GT1 homologues within the eukaryotic clades are unresolved, but their from the Lokiarchaeum metagenome that has been monophyly and widespread distribution suggests that suggested as the closest archaeal relative to eukaryotes Lombard Biology Direct (2016) 11:36 Page 15 of 31 Fig. 5 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 16 of 31 (See figure on previous page.) Fig. 5 Bayesian phylogeny of the closest relatives to the eukaryotic Alg2/Alg11 homologues. The tree is unrooted and was reconstructed using 211 representative sequences and 194 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) [11] belong to other parts of the general (preliminary) Gene duplication and neofunctionalization events would GT1 superfamily tree. Until these archaeal proteins are explain the emergence of the plethora of specialized GTs characterized, the lack of functionally important Alg1 in known today [59]. archaea and currently favored symbiogenic scenarios for In the current analysis, the study of the lumen GTs the origin of life rather favor the proteoarchaeal origin of was extended to a much larger diversity of eukaryotes the eukaryotic alg2/alg11 GT1 ancestor. and found to be widespread in most eukaryotic lineages In summary, Alg2 and Alg11 could have evolved from (Table 1, Additional file 1). The phylogenies of each a proteoarchaeal GT1 ancestor and Alg1 has unknown superfamily support the monophyly of each lumen GT origins. Modern eukaryotic complexes based on Alg1, and, thus, its likely inheritance from LECA (data not Alg2 and Alg11 are eukaryotic innovations developed shown). This analysis also suggests a previously unre- prior to LECA. ported possible relationship between Alg10 and Gpi18 (data not shown). Nevertheless, the sequences from dif- The unclear origin of an uncertain flippase ferent superfamilies are so divergent that alignments The transport of polyisoprenol-linked glycans across the comprising several superfamilies are highly unreliable membrane is generally thought to be mediated by pro- (Guidance alignment score < 0.1), and so are the result- teins called flippases [75]. Rft1 was suggested to be the ing phylogenies. Here, particular attention was given to eukaryotic N-glycosylation flippase [76] but the actual the elucidation of the possible early origins of the activity of this protein has been called into question eukaryotic lumen GTs. As reported previously [46], basic [77–79]. Rft1 has homologues in the three domains of BLASTp searches did not detect prokaryotic homo- life (Fig. 2). The phylogeny of the Rft1 homologues logues, so HMM profiles were constructed for each shows that the eukaryotic sequences form a monophy- superfamily in order to look for them (see Methods letic group (BPP = 0.99) supportive of their inheritance [81]). The closest prokaryotic hits of each superfamily from LECA, but it is unable to clarify the relationship of were used as seeds for psi-BLAST searches [82] against the eukaryotic proteins with their prokaryotic relatives the bacterial and archaeal genomes. The vast majority of (Additional file 6). Thus, the origin of the eukaryotic the detected prokaryotic sequences had PMT domains Rft1 is unclear (see Additional file 6 for a more detailed (for Protein O-Mannosylation Transferases) or the un- discussion). known protein domain DUF2029 defined from mycobac- terial proteins (data not shown). The PMTs carry out The lumen GTs are related to the protein O-mannosyl protein O-mannosylation, which is the transfer of a man- transferases (PMTs) nose residue from Dol-P-Man to a protein in the lumen Once translocated to the lumen side of the ER mem- side of the ER membrane [83, 84]. O-mannosylation and brane, the Dol-P-linked glycan is mannosylated four N-glycosylation are highly intertwined, as inhibition of times and glucosylated three times to make up the ca- one enhances the other and both PMT and N-OST belong nonical opisthokont core glycan (Fig. 1b). These modifi- to the glycosyltransferase family C (GT-C) and may be cations are catalyzed by the mannosyltransferases Alg3, distantly related [85]. O-mannosylation has been described Alg9 and Alg12 and the glucosyltransferases Alg6, Alg8 in several bacteria including Mycobacteria [86, 87]–a and Alg10. These proteins appear to be unique to suggestive result given the detection of the mycobacterial- eukaryotes [46, 50], so their evolution has only been based unknown domain DUF2029 in the distant lumen studied in this domain [59]. The closest relatives of these GT prokaryotic homologues. O-mannosylation has not GTs are the eukaryotic mannosyltransferases involved in been described in archaea yet, but it is exciting to point the GPI biosynthesis (Fig. 2, [59]), which also takes place out that the protein Agl1, which was tentatively suggested in the lumen side of the ER membrane [80]. All the to be a GT in S. acidocaldarius [88], bears a PMT domain. lumen GTs have similar multispan transmembrane do- Incidentally, AglS is an archaeal protein which shares mains that have been compared to the topology of sugar with the mannosyltransferases Alg3, Alg9 and Alg12 a transporters [59]. Lumen GTs were classified into three Dol-P-mannose mannosyltransferase function [42]. AglS superfamilies according to the sugar they used, the sugar transfers a mannose to a S-layer protein N-glycan in H. linkage they catalyze and their common peptide motifs. volcanii and has also been suggested to be distantly related Lombard Biology Direct (2016) 11:36 Page 17 of 31 to the PMTs, but there is no obvious evolutionary rela- gene was vertically inherited in archaea and eukaryotes tionship between AglS and the eukaryotic lumen GTs. Fi- from their last common ancestor (in a three-domain tree nally, the phylogeny of the PMT homologues was carried of life perspective), or that the eukaryotes branch within out (Additional file 7). The eukaryotic PMTs (including a paraphyletic proteoarchaeal group (BPP = 0.88, as several paralogues) form a monophyletic clade (BPP = 1) shown in Fig. 6). This latter possibility is in agreement but their relationship to their prokaryotic relatives is un- with the latest and most accurate phylogenies and phylo- clear (see Additional file 7 for details). genomic studies, which support the closest relationship In summary, very distant sequence similarities were of the eukaryotic stem with the proteoarchaea [21–24]. detected between the ER lumen GTs and the PMTs. Despite this result, no specific proteoarchaeal lineage, PMTs use similar Dol-P-monosaccharide donors as the including Candidatus Lokiarchaeum [11], is specifically N-glycosylation lumen GTs (the same in the case of the related to the eukaryotic Stt3, so the identity of the pro- mannosyltransferases), carry out similar transfer reactions tearchaeal donor is unknown. This indicates a lack of and participate in functionally related mechanisms. As a resolution but it could also be explained by poor sam- result, the evolutionary relationship of these GTs with the pling among archaea or by the fact that the donor of this PMT enzymes seems significant. It suggests that the clos- sequence predated LECA and, therefore, it cannot ne- est ancestor of the lumen N-glycosyltransferases was not a cessarily be assigned to one of the modern proteoarch- sugar transporter [59] but a mannosyltransferase from the aeal lineages. GT-C superfamily. Little is known about the function or evolution of the non-catalytic N-OST subunits [47]. Ost1, Ost2, Ost3/6, Eukaryotic Stt3 was probably acquired from Swp1 and Wbp1 orthologues are widespread among proteoarchaea eukaryotes (Table 1, Additional file 1), but homologues of The N-OST that transfer the oligosaccharide from the the other subunits are more difficult to detect certainly lipid carrier to the acceptor protein have been genetically because of their small size (e.g. yeast Ost4 is 36 amino and biochemically described in eukaryotes [53, 89, 90], acids long). BLASTp did not find any sequence similar to bacteria [44, 47] and archaea [91–95]. The evolution of these subunits neither in bacteria nor in archaea, so dis- the catalytic Stt3/AglB subunit of the N-OST is by far the tant prokaryotic homologues were looked for using HMM best studied of the N-glycosylation enzymes. For instance, profiles. The few detected sequences participate in unre- many HGTs and duplications of aglB have been reported lated functions, so no promising candidates were found in archaea and a close relationship between these proteins (data not shown). Maybe when more will be known about and eukaryotic Stt3 has been suggested [46, 47, 96, 97]. the functions of the non-catalytic N-OST subunits, it will Yet, previous evolutionary studies were limited to one be possible to make more informed choices to study the domain of life, were unable to determine the specific origin of these eukaryotic genes. relationship of eukaryotes to archaea, or simply dis- regarded this issue. A dolichol-P-monosaccharide synthase could have an ar- A phylogeny of Stt3/AglB homologues from all three chaeal origin domains of life was carried out (Fig. 6). The eukaryotic Although strictly speaking they are not components sequences form a monophyletic clade (BPP = 0.92) which devoted to the N-glycosylation pathway, the two Dol-P- contains clear taxonomic groups but whose phylogeny is monosaccharide synthases supply mannose and glucose blurred by the presence of several paralogues. The rest to the lumen GTs (Fig. 1b). These enzymes belong to the of the tree is made up of archaeal sequences roughly glycosyltransferase group 2 (GT2, [75]), which also com- clustering according to the archaeal taxonomy and a few prises most GTs involved in the archaeal N-glycosylation bacterial homologues located within the euryarchaeal pathways (Fig. 2). Therefore, the origins of the eukaryotic clade. The widespread distribution of AglB homologues GT2 proteins were examined. In eukaryotes, Alg5 is a in archaea and the clustering of the sequences according Dol-P-glucose synthase [98] and Dpm is a Dol-P-mannose to the main archaeal taxonomic groups supports the an- synthase also providing mannose to GPI synthesis and O- cestral presence of this protein in LACA. Moreover, the mannosylation [99]. Two types of Dpm exist [75, 85, 100]: location of the bacterial sequences within the euryarch- yeast Dpm has a C-terminal transmembrane helix to aeal group supports the dominant opinion that this gene interact with membranes, whereas human Dpm1 lacks was transferred from archaea to bacteria [46–48] at least this structure and requires supplementary subunits twice, followed by a few subsequent transfers among (Dpm2/Dpm3). These differences were translated into two bacteria. The relationship between the archaeal and separate clades in previous Dpm phylogenies [46]. eukaryotic orthologues is more difficult to establish be- Similar to the GT1 superfamily, the GT2 superfamily cause the phylogeny is unrooted. Depending on where comprises many proteins (red proteins in Fig. 2) that we decide to root this phylogeny, it may suggest that the make its evolutionary study challenging. Only the closest Lombard Biology Direct (2016) 11:36 Page 18 of 31 Fig. 6 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 19 of 31 (See figure on previous page.) Fig. 6 Bayesian phylogeny of the catalytic N-OST subunit (Stt3/AglB/PglB) homologues. The tree is unrooted and was reconstructed using 163 representative sequences and 268 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange)and eukaryotes (purple) prokaryotic relatives to the eukaryotic Alg5/Dpm se- the bacterial polyisoprenol-based pathways involved in cell quences were kept for the analysis (see Methods). The wall synthesis [34, 48–50]. Yet, aside from a primordial resulting phylogeny is ambiguous (Additional file 8). The paper that had neither the data nor the intention to tackle prokaryotic sequences are mixed, but the (unrooted) this issue [46], this is the first attempt to provide a phylogeny seems to be split into two clades (BPP = 0.96): thorough examination of the relationship of the eukaryotic one dominated by bacteria and eukaryotic Alg5 sequences N-glycosylation pathway with its prokaryotic relatives. and another dominated by archaea and eukaryotic Dpm The proteins involved in polyisoprenol-based pathways sequences. The bacterial-dominated clade contains WsfH from the three domains of life, including the eukaryotic from P. alvei, which may supply periplasmic Bac-P-Glc to N-glycosylation pathway, belong to a small number of tyrosine O-glycosylation [101]. This is remarkable because gene families (Fig. 2). Only the putative N-glycosylation Alg5 are also Dol-P-Glc synthases, but it is balanced by pathway from S. acidocaldarius has an obvious parallel- the fact that AglD, a Dol-P-Man synthase from H.volcanii, ism with the eukaryotic N-glycosylation, as they are both is phylogenetically closer to the Alg5 clade. A few cre- made up from representatives of the same protein super- narchaeal sequences and a sequence from Candidatus families (Fig. 2). More detailed phylogenomic analyses, Lokiarchaeum are basal to the eukaryotic Alg5 clade, but however, show that some eukaryotic N-glycosylation the significance of these sequences is uncertain because genes may indeed have proteoarchaeal ancestors, but they are mixed to other euryarchaeal and bacterial only two of those belong to the S. acidocaldarius path- sequences. Thus, the origin of the eukaryotic alg5 way. This stresses the fact that simple homology is not gene is unclear. enough to draw robust evolutionary hypotheses about The archaeal-dominated clade includes H. volcanii the relationship among polyisoprenol-based pathways. AglJ, M. voltae AglK, A. fulgidus Orf39 and the euka- Not only the phylogenomic analyses are required to ryotic Dpm. As previously reported [46], the Dpm confirm evolutionary hypotheses, they also provide sequences are split into two groups depending on the promising prokaryotic targets for the dynamic field of presence of the C-terminal transmembrane helix. The prokaryotic glycosylation characterization [36, 40]. eukaryotic Dpm clades are intertwined with several pro- Table 1 shows that most proteins involved in the karyotic sequences, most notably the main proteoarch- eukaryotic N-glycosylation pathway are widespread in all aeal clade of the tree. Thus, it is tempting to suggest major eukaryotic taxa. The phylogenetic results of each that the eukaryotic sequences, as well as the bacterial superfamily implicated in this process have been sum- sequences in this cluster, have proteoarchaeal origins. marized in Fig. 7. Since many of these phylogenies are Since the Dpm1 type (devoid of a C-terminal transmem- unrooted, their interpretation sometimes depends on the brane helix) is more widespread in eukaryotes, it is tree of life topology that we favor, but they clearly plausible that it was the ancestral type in eukaryotes and support a more complicated scenario for the origin of the fused type emerged later. The characterization of the eukaryotic N-glycosylation pathway than previously some proteoarchaeal homologues should allow testing thought. All the canonical eukaryotic N-glycosylation genes this possibility. were present in LECA and inherited (or lost/modified) in In summary, the origins of the eukaryotic Alg5 are modern lineages, but the origins of the eukaryotic N- unclear, but Dpm may have a proteoarchaeal ancestor. glycosylation genes are diverse. Alg7, Dpm1 and also pos- sibly one GT1 and one Stt3 (N-OST catalytic subunit) have Conclusion proteoarchaeal ancestors. A duplication of the original N-glycosylation is a major protein modification in eukary- GT1 gene (blue family in Fig. 2) in the eukaryotic lineage otes that modulates the properties of many proteins. before LECA would account for genes alg2 and alg11.The Unraveling its origins is not only relevant to the particular stt3 gene has been duplicated in some eukaryotic lineages case of this essential pathway, but its location in the ER too. The Alg13/Alg14 couple is structurally most similar membranes may also have implications for the origins of to bacterial homologues, so it may have bacterial origins. the ER and the eukaryotes themselves. Previous hypoth- The lumen GTs could have resulted from the co-option of eses suggested a particular evolutionary link between this a PMT protein involved in O-mannosylation and subse- eukaryotic pathway and archaeal N-glycosylation or one of quent gene duplications. Finally, the other GT1 (Alg1), Lombard Biology Direct (2016) 11:36 Page 20 of 31 Fig. 7 Origins of the eukaryotic N-glycosylation proteins and presence of related superfamilies in the last common ancestors of each domain of life and the cenancestor (summary) GT2 (Alg5) and the putative flippase have unknown making of the eukaryotic N-glycosylation pathway. This origins and could be eukaryotic innovations. The origins is in agreement with currently favored hypotheses which inferred for the eukaryotic N-glycosylation genes will be assume a proteoarchaeal ancestor to the eukaryotic stem open to reevaluation when more information will be [21–24]. For instance, these results are congruent with available about their prokaryotic relatives that have been the “inside-out” hypothesis, which suggested that the pointed out in this work. presence of the eukaryotic N-glycosylation pathway in Overall, this is another example that in metabolism, as the ER membranes was a remnant of an eocyte S-layer in other functions, eukaryogenesis was achieved through N-glycosylation pathway ([14]). Yet, these results are also an intricate combination of pre-existing and new ele- compatible with other eukaryogenesis hypotheses, as ments. Although in a few cases (e.g. Alg13/Alg14, Alg2/ other eocyte or archaeal-stem hypotheses of eukaryogen- Alg11, Stt3) a proto-eukaryotic origin cannot be com- esis may also account for the proteoarchaeal origin of these pletely ruled out, the results presented here suggest a proteins [8, 10, 11] and the possibility that these genes strong influence of some proteoarchaeal genes in the were horizontally transferred from a proteoarchaeon to a Lombard Biology Direct (2016) 11:36 Page 21 of 31 non-proteoarchaeal eukaryotic stem cannot be excluded [3, guidance statistical values on the alignment were used to 13, 15–18]. Whatever the eukaryogenesis hypothesis, this make an informed choice to trim the sequences for the work shows that it is now time for works on eukaryotic or- final phylogenetic reconstructions, and the trimmed align- igins to study the early evolution of the eukaryotic mem- ments are reported in the Additional file 11. The final branes and membrane functions. phylogenies were constructed using the programs MrBayes 3.2.6 [108] and RaxML-HPC2 8.1.24 [109] im- Methods plemented in the CIPRES Science Gateway [110]. The The original sequence seeds used to build the homology MrBayes analysis used a LG substitution model [111] and groups in Fig. 2 were selected from the literature (Table 2) 4 categories for the Γ distribution of substitution rates. and the KEGG database (http://www.genome.jp/kegg/). Four chains of 1,000,000 generations were run, trees were All vs. all reciprocal BLASTp searches were carried out sampled every 100 generations and the first 25 % [60]. The homology groups were manually built from the generations were discarded as “burn in”.The RaxML inspection of the reciprocal BLASTp: hits with an e-value reconstructions used the LG + Γ model [112] and 4 rate inferior to 1e-05 were automatically considered to belong categories. A hundred bootstrap replicates were carried to the same homology group; proteins with two or more out with the same model to evaluate node support. The hits from the same homology group and e-values ranging Bayesian and maximum likelihood phylogenies gave between 0.01 and 1e-05 were also included in the hom- comparable results. The bootstrap support from ology group. Information from the literature was appealed RaxML of the main basal nodes was reported on the to confirm or suggest particular homologies that were not published Bayesian trees. unambiguously determined with simple reciprocal BLAST, In some cases, the basic BLASTp search did not e.g. in the case of the lumen GTs [59]. These cases of provide satisfactory results. When too many prokaryotic ambiguous homology appear as uncertain (empty white homologues were detected (e.g. GT1 and GT2 super- shapes in Fig. 2). families), a smaller prokaryotic genome list was used for The same sequence seeds were used for the phylogen- the preliminary analyses (Additional file 9). These earlier etic analyses. Local genomic databases were constructed results were curated and only the closest homologues to for each domain of life from a selection of representative the eukaryotic proteins of interest were kept. Supple- genomes (Additional file 9). BLASTp with default parame- mentary sequences from the regular prokaryotic local ters were carried out against these local databases. These genomic database were added later on to improve the results were complemented with supplementary BLASTs prokaryotic diversity of the analyses. At that step, how- against the same genomes carried out with alternative ever, only the closest prokaryotic relatives were kept, so seeds and manual searches in the orthologue databases recent paralogues may have been missed. On the con- EggNOG (http://eggnogdb.embl.de/#/app/home, [102]) trary, when very few similar sequences were detected in and OMA browser (http://omabrowser.org/oma/home/, a domain of life, the searches were extended to the non- [103]) to make sure that no orthologues were missed in redundant (nr) sequences annotated as belonging to that the Table 1. Sequences too long, too short or for which domain of life on GenBank (http://www.ncbi.nlm.nih. the protein domain under study was not detectable using gov/Genbank). If similar sequences could not be detected the Pfam database 28.0 (http://pfam.xfam.org/) were re- using these methods (e.g. lumen GTs in prokaryotes), moved from the rest of the analyses. Some exceptions HMMprofileswerebuilt from theMUSCLE alignments of were applied when necessary as, for example, the eur- the eukaryotic sequences and were used to look for distant yarchaeal MurG-like homologues, which were kept for the homologues using the hmmsearch tool implemented in the N-terminal analysis despite their sequences being too di- HMMER v3.1b2 webserver (http://hmmer.org/, [81]). Ho- vergent for Pfam to detect their respective protein domain mologues found in this way were validated as suitable in them. Preliminary alignments were carried out using candidates based on protein domain information from MUSCLE v3.8.31 [104], sequences were trimmed using Pfam or functional annotation from the literature. the program NET of the MUST package [105] and prelim- The validated prokaryotic homologues were then used inary phylogenies were reconstructed using FastTree to carry out BLASTp or psi-BLAST searches [82], v.2.1.7 [106]. These preparatory analyses were examined either against the regular list of genomes or the nr for the overall topology of the trees, and used to select sequences from GenBank, depending on the results. representative sequences for smaller datasets that could be used with more accurate phylogenetic reconstruction Reviewers’ comments methods. Once the representative sequences had been se- The author thanks the reviewers for their comments. lected, a new MUSCLE alignment was carried out on the When possible, the manuscript has been modified ac- Guidance server (http://guidance.tau.ac.il/ver2/, [107]). cording to their recommendations. The alignments are available in the Additional file 10. The Lombard Biology Direct (2016) 11:36 Page 22 of 31 First round of review Author’s response: This article originally was longer Referee 1, Pr. Patrick Forterre, University Paris Sud, France and covered not only the origin of the eukaryotic N- Reviewer summary It’s an interesting and useful paper glycosylation pathway but also the putative presence of a that could be improved if fast evolving species were re- polyprenol-based glycosylation machinery in the cenan- moved from the phylogenetic analyses. cestor. In a previous submission, it was recommended to me to separate both subjects in order to make the mes- Reviewer comments sages clearer. As a result, I split the original article into In this paper, Jonathan Lombard has performed a rather this and another paper (reference [50]), where I discuss exhaustive phylogenomic analysis of N-glycosylation the implications for the last common ancestor of extant pathways in the three domains of life, focusing on the organisms (the cenancestor). In summary, the cenancestor most conserved proteins, that belong to universally dis- most likely had a biosynthetic pathway that could have tributed protein glycosyltransferase superfamilies. This is allowed the synthesis of a polyisoprenol-phosphate lipid an important work that could be useful for all scientists carrier similar to Bac-P in this organism. Together with interested in reconstructing the evolution of membranes the presence of glycosyltransferases described in the systems in the three domains. The paper is sometimes present article, I conclude that a glycosylation machinery difficult to read for people outside of the fields because, may have existed in the cenancestor. Since the over- for historical reasons, orthologous proteins have dif- whelming majority of the prokaryotic polyisoprenol-based ferent names in different domains. Jonathan Lombard glycosylation machineries are involved in the synthesis of could benefit from his analysis to propose, when pos- cell wall components, I suggest that the cenancestor sible, new names, such that orthologous proteins in dif- already had a mechanism to synthesize glycosylated cell ferent domains could be readily identified. walls. However, I could not unambiguously infer the pres- Author’s response: Thank you for this suggestion. I am ence of any oligosaccharide transferase (the proteins that aware of the fact that the number of proteins in this ana- ultimately select the final acceptor of the oligosaccha- lysis is difficult to handle, especially because it is difficult rides) in the cenancestor, so I think that the system may to see the connection among homologues that carry out have been less specific and developed into different cell different functions in different lineages. Unfortunately, walls glycosylation pathways later or just modified in the the field of polyisoprenol-related glycosylations is so vast modern lineages because of the strong selection pressures that I have not been able to figure out a systematic that act on the glycocalyx. As this is the subject of an- notation that would make sense at the scale of the three other article, I have not included this discussion in this domains. For instance, the HPT notation is already a article, but I now refer to these possibilities at the end of generic name given to a superfamily that comprises pro- the section “The eukaryotic N-glycosylation initiation teins such as Alg7, AglH MraY, WecA, TagO, etc. When complex has a composite origin”. possible, I have referred to glycosyltransferase superfam- Other proteins were probably present in the last com- ilies (GT1, GT2, etc.), but in cases such as MurG and mon ancestor of Archaea and Eukarya or originated in their homologues the renaming is difficult because of the Archaea if Eukarya emerge from within Archaea. Jonathan complicated protein domain architecture of the family Lombard has interpreted all his results on the basis of the and the fact that many of these proteins are uncharacter- second hypothesis (the archaeal ancestor scenario) for the ized. I strongly recommend using the Fig. 2 as a general origin of Eukaryotes. However, this hypothesis is mainly legend for the whole article, as it shows the relationships supported by concatenation of unresolved single trees among all the proteins discussed here. confused by the presence in the dataset of fast evolving Some of the proteins, such as HPT and MurG, indeed species (Nanoarchaea, Methanopyrus, Korarchaeon) (see appear universal and the topology of their trees (Figs. 3 Forterre 2015). This is problematic since even the best and 4) are rather consistent with those of classical uni- phylogenetic methods cannot prevent long branch versal proteins (Archaea and Eukarya closely related attraction in the presence of a very distant outgroup compared to Bacteria) (especially in the case of HPT). (Gouy et al., 2015), which is the case here with Bacteria. This suggests that these proteins are orthologous and Better analyses in which most fast evolving species have were already present in LUCA as indicated by the author been removed provided contradictory rootings. For in- in Fig. 7. Could Jonathan Lombard speculate about the stance, the analysis of Gribaldo and co-workers reject the role of these proteins in LUCA? Can we conclude that TACK (proteoarchaeal) superphylum (Rayman et al. 2015) LUCA already encoded glycosylated proteins at its sur- that the author favours here, grouping some Euryarchaea face? Is it possible to imagine a minimal glycosylation with “proteoarchaea”. The paper would be better if pathway that could have evolved into the complex path- Jonathan Lombard consider in parallel the Lake (eocyte/ ways observed in Archaea/Eukarya and that was partly proteoarchaea) scenario and the Woese’sscenarioin recruited in Bacteria for peptidoglycan biosynthesis? which Archaea and Eukarya are sister groups forming a Lombard Biology Direct (2016) 11:36 Page 23 of 31 clade (Akarya, sensu Forterre, 2015). In that case, it is is never mentioned in this paper. This is damaging and possible for instance that the Last Archaea and Eukarya can create some confusion. For instance in Fig. 5, the Common Ancestor (LAECA) contained paralogues of group “Proteoarchaea” includes only Crenarchaeal se- some proteins, followed by differential loss in some quences and two Nitrososphera (Thaumarchaeota) that archaeal phyla and in Eukarya. This could explain for in- branch within Crenarchaeota (a probable gene transfer). stance why Eukarya are more related to particular archaeal This group thus correspond to Crenarchaeota and not to phyla in the trees of Figs. 3, 5 and 6. Unfortunately, the “Proteoarchaea”. This important information is lost when phylogenie are difficult to interpret because Jonathan using the disputed phylum Proteoarchaea as an umbrella Lombard also uses many fast evolving species in his data- for diverse phyla. In general, I think dangerous and poten- set. For instance, Nanoarchaeaum and the Korarchaeon in tially misleading to talk about either the TACK or the Fig. 3, Methanopyrus kandleri in Fig. 4. I would like to see Proteoarchaea superphyla. These superphyla are only valid new phylogenies without these fast evolving organisms. if one root the archaeal tree between Euryarchaea and all It’s not necessary to do time consuming Bayesian analyses other archaea, as suggested by Petitjean et al. (2014). considering the small number of positions analyzed. These authors root the archaeal tree using Bacteria as out- PhyML would be sufficiently informative. group. Using the same strategy, Rayman et al. (2015) find Authors response: New phylogenies were carried out re- another root. It is in fact problematic to root the archaeal moving the Nanoarchaeum, Korarchaea and Methanopyrus tree using such a distant outgroup. Using Eukarya as a fast-evolving sequences. The Figs. 3 and 4b have been re- much closer outgroup, Brochier-Armanet et al. rooted the placed accordingly, but the other Figures of the paper did tree between Thaumarchaea and other Archaea (Brochier- not contain sequences from these organisms or their impact Armanet et al., 2008). We recently obtained the same was not obvious, so they were kept as they were. The overall rooting with a three domains topology of RNA polymer- topology of Fig. 3 did not change. Fig. 4b changed in that, in- ase after removing fast evolving sequences (manuscript in stead of branching within the bacterial EPS sequences, now preparation) so the question is quite open. the eukaryotic Alg13 homologues form a separate clade. Author’s response: in addition to removing the fast- I agree on the fact that previous versions of this manu- evolving sequences (see above), more thaumarchaea and script did not sufficiently discuss the phylogenies in the Bathyarchaea sequences were also included in the new context of the traditional three domains of life topology. phylogenies presented in Figs. 3 and 4a. This addition As these phylogenies are unrooted, a traditional proto- did not considerably change the topology of the resulting eukaryotic lineage cannot be ruled out from the inter- phylogenies, so the rest of phylogenies were kept un- pretation of the Alg13/Alg14, Alg2/Alg11 and Stt3 changed. Regarding to the term “Proteoarchaea”, the phylogenies. I have now discussed these possibilities in annotations in the trees are meant to provide an indica- the respective sections of the text. Yet, in order to provide tion of the taxonomic affiliation of the sequences used, as clear conclusions as possible in this already quite com- and not at all to discuss the internal phylogeny of each plicated article, I have decided to put forward the origins domain of life. For instance, there are sequences in these that I think are the most plausible in the abstract, final figures that are put under the same grouping despite be- conclusions and the summary in Fig. 7. This is based in ing paraphyletic, just to make the figures more readable. particular functional and structural arguments discussed The names chosen are such as the number of annotations for each gene, but also in the fact that the rooting of is both as few and precise as possible. For example, if a eukaryotes within archaea is now favored by most group contains mostly viridiplantae and a couple of authors (references [19–22]), including the work by rhodophytes, the whole group will be annotated as Raymann et al. 2015 cited by the reviewer. I have also “archaeplastida”, whereas if it only contains chloro- tried to forward in the conclusion the fact that the valid- phytes, it will be annotated as such. According to the ity of these hypotheses may change when more informa- same principle, if a group of sequences is only made up tion will be available about the prokaryotic relatives of of crenarchaea, it will be annotated as such, but if it also the eukaryotic proteins studied here. includes one thaumarchaea, it will be annotated as I am also surprised by the small number of “proteoarchaea”, as this is the first higher taxonomic Thaumarchaeota in the trees. I did some rapid BLAST term that includes all sequences in the group (I prefer searches and much more sequences could be retrieved. “proteoarchaea” to “TACK” as it is a name instead of an Some sequences of Bathyarchaea, that usually branch acronym and, therefore, there will always space for dis- with Caldiarchaeum and Thaumarchaea in protein trees cussion of what is in the name). That being said, I do not could also be added to see if one recover the usual think that the work by Raymann et al. 2015 necessarily monophyly of these three groups that, in my opinion, contradicts the proteoarchaeal monophily. That piece of should be all considered to be members of the phylum work found that the euryarchaea were paraphyletic, with “Thaumarchaeota”. Surprisingly, the name Thaumarchaeota some of them forming a sister group to a monophyletic Lombard Biology Direct (2016) 11:36 Page 24 of 31 group of proteoarchaea. On the base of that work, the Reviewer comments term that should be under scrutiny is “Euryarchaea”. I have only a few critical remarks. 1. This is the first The work by Brochier-Armanet et al. 2008 showed that thorough phylogenetic study of the families involved. Cenarchaeum symbiosum (a Thaumarchaea) could be While phylogenies of most of the mentioned families basal to all archaea when the eukaryotes were used as have been studied at large distances in the study an outgroup, but the choice of that outgroup is question- covering the majority of conserved protein families in able if the eukaryotes branch within the proteoarchaea, Eukaryotes (Yutin et al., 2008), these phylogenetic trees as some of the same authors suggest in the Raymann et have not been described or analyzed individually. The al. 2015 paper. In short, although the monophyly of both author presents a much more detailed study of the small the euryarchaea and proteoarchaea may be disputed, these number of families. Nevertheless, for a number of pro- groups are sufficiently well-established to be used in the tein families he has to conclude that “the closest pro- annotation of the figures in this paper, which does not in- karyotic group to the eukaryotic clade is unclear”.In tend to solve the phylogeny of archaea anyway. time with thousands new genomes sequenced and new A minor point; Jonathan Lombard use the term methods of phylogenetic analyses developed this may “Cenancestor” instead of LUCA for the Last Universal change. Therefore, I strongly believe that the author or Common ancestor. However, the “Cenancestor” can be other researches will return to the study of these protein the common ancestor of any group. If the author wants families in the future and at a very different level. There- to refer to the historical proposal by Fitch, he can con- fore, it might be helpful to add certain data to the sider that LUCA means the Last Universal CenAncestor! Supplementary material of this paper. I would insist on Author’s response: I have included the name of LUCA publishing multiple alignments of all proteins involved, in the paper, as the reviewer is right in that this term is both trimmed and untrimmed. The last suggestion is more generally known. Yet, the term “cenancestor” was prompted by the author’s argument that the topology of kept throughout the text. The definition of this term is eukaryotic clade…“most likely results from the small well stated in the paper, and I do not think that it is any number of alignment positions that was conserved in more ambiguous than the “Universal” in LUCA. each phylogeny (119 and 97 respectfully)…” Trimming Thank you for endorsing the publication of this work. of the initial alignments was done by the author auto- Gouy R, Baurain D, Philippe H. Rooting the tree of matically and it happens that the trimming parameters life: the phylogenetic jury is still out. Philos Trans R Soc used are the same for all the trees and therefore not op- Lond B Biol Sci. 2015 370(1678):20140329. doi:10.1098/ timal for some of them. Given the constantly changing rstb.2014.0329. and disappearing protein identifiers (and even gene Forterre P. The universal tree of life: an update. Front models) in the public databases, it would be essential to Microbiol. 2015 6:717. doi: 10.3389/fmicb.2015.00717. provide stable identifiers, such as GenBank CDS IDs or Raymann K, Brochier-Armanet C, Gribaldo S. The two- UniProt\UniParc IDs. domain tree of life is linked to a new root for the Archaea. Author’s response: The untrimmed and trimmed Proc Natl Acad Sci U S A. 2015 112(21):6670–5. alignments supporting all the phylogenies published in Petitjean C, Deschamps P, López-García P, Moreira D. this article have now been made available in Additional Rooting the domain archaea by phylogenomic analysis files 9 and 10 [10 and 11 after second revision], respect- supports the foundation of the new kingdom Proteoarch- ively. The IDs provided in these phylogenies are updated aeota. Genome Biol Evol. 2014 7(1):191–204. doi: 10.1093/ NCBI’s accession numbers, as currently recommended in gbe/evu274. the NCBI’s website. Although the accession numbers may indeed change over time, it has always been possible tra- Referee 2, Dr. Sergei Mekhedov, NCBI, NLM, NIH, USA cing back old accession numbers to their new IDs. If, for (nominated by Editorial Board member Michael Galperin) some reason, the identifiers were to become completely Reviewer summary inaccessible, the sequences will still be available on the I have no doubt that the paper by J. Lombard “The Additional file 9 [10, after second review]. multiple evolutionary origins of the eukaryotic N- I failed to find word “ortholog” in the text. This is not glycosylation pathway” deserves publication in Biology a big deal, but I believe that the author should have at Direct. For the first time the author puts a phylogenetic least checked the described protein families in some of analysis of N-glycosylation in the context of a metabolic the available databases of orthologous groups of pro- pathway. He performs the analyses at large evolutionary teins. I am convinced that his attempt to build his own distances comparing Eukaryotes, Archaea, and Bacteria homology groups by running BLAST all-against-all with and uses the state of the art methods. The main an unjustified expectation value cut-off resulted in in- conclusions cannot be called unexpected but are none- complete protein families for the list of Eukaryotes in theless valuable. Table 1. As an example, all question marks in Table 1 Lombard Biology Direct (2016) 11:36 Page 25 of 31 correspond to true orthologs of S. cerevisiae proteins CNC04720 [Cryptococcus neoformans var. which served as the basis for this table. The following neoformans JEC21] > tr|A0A015LNA6|A0A015LNA6_ proteins missed in Table 1 should be added. 9GLOM Swp1p [Rhizophagus irregularis DAOM 197198w] > gi|46099108|gb|EAK84341.1| hypothetical A. Missed orthologs of Ost4 (very short protein, protein UM03236.1 [Ustilago maydis 521] > tr| present in many plants): >tr|Q5KCD6|Q5KCD6_ I1G5D9|I1G5D9_AMPQE Uncharacterized protein CRYNJ Uncharacterized protein CNH00640 [Amphimedon queenslandica] > tr|Q7K110| [Cryptococcus neoformans var. neoformans JEC21] > Q7K110_DROME LD18774p [Drosophila tr|I1FYU0|I1FYU0_AMPQE Uncharacterized protein melanogaster] > sp|Q9DBG6|RPN2_MOUSE [Amphimedon queenslandica] > tr|A0A0D2X2S9| Dolichyl-diphosphooligosaccharide–protein A0A0D2X2S9_CAPO3 Uncharacterized protein glycosyltransferase subunit 2 [Mus musculus] > tr| [Capsaspora owczarzaki (strain ATCC 30864)] > tr| W4Y183|W4Y183_STRPU Uncharacterized protein L8GRU2|L8GRU2_ACACA Uncharacterized protein [Strongylocentrotus purpuratus] > tr|A0A0D2U678| [Acanthamoeba castellanii str. Neff] > sp|Q8L986| A0A0D2U678_CAPO3 Uncharacterized protein OST4B_ARATH Oligosaccaryltransferase [Arabidopsis [Capsaspora owczarzaki (strain ATCC 30864)] thaliana] > tr|J3LPF6|J3LPF6_ORYBR Uncharacterized > sp|Q54HG9|RPN2_DICDI Dolichyl- protein [Oryza brachyantha] > tr|A0A096R6U0| diphosphooligosaccharide–protein glycosyltransferase A0A096R6U0_MAIZE Uncharacterized protein subunit swp1 [Dictyostelium discoideum] > [Zea mays] > tr|A0A059A7I3|A0A059A7I3_EUCGR tr|D3BAP4|D3BAP4_POLPA Dolichyl- Uncharacterized protein [Eucalyptus grandis] > tr| diphosphooligosaccharide-protein glycotransferase G7KZM4|G7KZM4_MEDTR Oligosaccaryltransferase [Polysphondylium pallidum] > tr|L8H4E5|L8H4E5_ [Medicago truncatula] > tr|W5B6B2|W5B6B2_ ACACA Ribophorin ii (Rpn2) protein [Acanthamoeba WHEAT Uncharacterized protein [Triticum castellanii str. Neff] > tr|A8J449|A8J449_CHLRE aestivum] > sp|C7J4U3|OST4A_ORYSJ Dolichyl- Predicted protein [Chlamydomonas reinhardtii] > tr| diphosphooligosaccharide–protein glycosyltransferase D8TQJ2|D8TQJ2_VOLCA Putative uncharacterized subunit 4A [Oryza sativa subsp. Japonica] protein [Volvox carteri] > tr|E1ZHV6|E1ZHV6_CHLVA B. Missed orthologs of Swp1 (dolichyl- Putative uncharacterized protein [Chlorella variabilis] diphosphooligosaccharide-protein glycotransferase, in C. Missed ortholog of Alg12 > tr|M1V5M0| most lineages orthologs are substantially longer than in M1V5M0_CYAM1 Uncharacterized protein S. cerevisiae) > tr|C5LE16|C5LE16_PERM5 Putative CYME_CMM004C [Cyanidioschyzon merolae uncharacterized protein [Perkinsus marinus (strain strain 10D] ATCC 50983 / TXsc)] > tr|F2UPX0|F2UPX0_SALR5 D. Missed ortholog of Alg14 > gi|146144751|gb| Putative uncharacterized protein [Salpingoeca rosetta EAR99237.2| conserved hypothetical protein (strain ATCC 50818 / BSB-021)] > tr|B3SAA0| (macronuclear) [Tetrahymena thermophila SB210] B3SAA0_TRIAD Putative uncharacterized protein [Trichoplax adhaerens] > tr|Q22ZG2|Q22ZG2_TETTS Author’s response: The absence of the word “ortholog” Oligosaccharyltransferase subunit ribophorin is perfectly conscious. Orthologues are genes that have [Tetrahymena thermophila SB210] > tr|A0CJY2| evolved through speciation and, in principle, they are A0CJY2_PARTE Uncharacterized protein mostly expected to carry out the same or very similar [Paramecium tetraurelia] > tr|A0CN26|A0CN26_ functions in different species. The proteins involved in PARTE Uncharacterized protein [Paramecium eukaryotic N-glycosylation, however, belong to families tetraurelia] > tr|V9EKX7|V9EKX7_PHYPR with a high number of paralogues and probably xenolo- Uncharacterized protein [Phytophthora parasitica gues, so I preferred using the broader homology rather P1569] > sp|Q5N7W3|RPN2_ORYSJ Dolichyl- than orthology throughout the article. diphosphooligosaccharide–protein glycosyltransferase I think that there has been some kind of misunder- subunit 2 [Oryza sativa subsp. Japonica] > tr| standing with regard to Table 1 and the formation of A9S283|A9S283_PHYPA Predicted protein homology groups using all-vs-all BLAST, as both opera- [Physcomitrella patens subsp. Patens] > tions were completely independent from each other. On sp|Q93Z16|RPN2_ARATH dolichyl- the one hand, Table 1 is a typical presence/absence table. diphosphooligosaccharide–protein glycosyltransferase It was made using the well-described Saccaromyces subunit 2 [Arabidopsis thaliana] > sp|O74943| cerevisiae copy of each protein in the eukaryotic N- YJB6_SCHPO Uncharacterized protein C553.06 glycosylation pathway to look for homologues in all the [Schizosaccharomyces pombe 972 h-] > tr| eukaryotic genomes listed in the Additional file 8 [File 9 Q5KK05|Q5KK05_CRYNJ Uncharacterized protein after second review]. I have included these sequences in Lombard Biology Direct (2016) 11:36 Page 26 of 31 the updated version of Table 1, with the symbol “#”.Iam about the eukaryotic origins, there are so many different grateful to the reviewer for pointing me towards these genes in this pathway, each with a different story, that if I orthologues, that I had missed. mentioned the implications for each gene of all different On the other hand, the all-vs-all BLAST was used only eukaryogenesis hypotheses in the conclusion, the message of to build the homology groups shown in Fig. 2. In order to the article would become quite confusing (for reminder, do this, the sequences from proteins that had been de- the main message of this article is that the eukaryotic scribed in various glycosylation pathways (Table 2) were N-glycosylation pathway had mixed origins). Yet, since I used to build a small database, and the all-vs-all BLAST agree with the reviewer that the origin of eukaryotes is not was run only in that small database. The objective of set, I encourage the interested reader to refer to the that part of the study was to find out which proteins in- “Results and discussion” section, where I develop other volved in polyisoprenol-based glycosylations were related evolutionary explanations for each tree. to other proteins from other pathways. This provided a Considering the nomenclature problem, Jonathan uses visual description of the relationship among the proteins LACA, LECA, LBCA, but has still a problem with LUCA in these pathways (Fig. 2) and also determined the pro- (also the name is now mentioned). I don’tunderstand why teins that belonged to each superfamily, so they could be the U of LUCA is ambiguous? Universal, as in “universal jointly taken into account in the superfamily phylogenies. proteins” clearly means all organisms encoding ribosomes Thank you for endorsing the publication of this work. since LUCA is the Last Common Ancestor of organisms Reference Yutin, N., Makarova, K. S., Mekhedov, S. L., from the three domains that have been defined based on Wolf, Y. I., & Koonin, E. V. (2008). The Deep Archaeal ribosomal proteins (Forterre et al., 2014). Obviously, Roots of Eukaryotes. Molecular Biology and Evolution, universal do not refer to the Universe. In any case, at the 25(8), 1619–1630. moment – and probably for quite a long time - we are only concerned in studying life on Earth. If life can be Second round of review studied on planet X in the future, it will be easy to distin- Referee 1, Pr. Patrick Forterre, University Paris Sud, France guish the terrestrial LUCA from the X LUCA! Beside, the The revised version has been quite improved. It was a “L” of LUCA is very important since all the ancestors of very good idea to include both the untrimmed and LUCA were also our Cenancestors. The term Cenancestor trimmed alignments. Unfortunately, several nodes in the is therefore both ambiguous concerning the group of or- phylogenies are still not resolved after the removal of ganism under consideration but also the identity of the fast evolving species. This is probably because the small ancestor among all ancestors of the group. size of the proteins. It could be interesting to look in the Author’s response: I am very sorry, for my previous alignment before trimming to look for indels supporting response was clearly insufficient. When I said that the particular nodes. word “universal” could also be considered somehow Jonathan Lombard still favours the archaeal ancestor ambiguous, I was thinking to the fact that the term is scenario for the origin of eukaryotes to explain the data very broad. Take the example of viruses: Prof. Forterre because “most authors now favour an archaeal origin of has been a prominent actor in the debate to determine if the eukaryotic stem”. This is true but unfortunately, be- viruses are alive or not. I personally do not have a clear- cause this scenario is mostly based on very weak phylog- cut opinion on the issue, but viruses clearly are a part of enies (ref [8–11]) obtained from the concatenation of the “biological universe”. The same for prions and any individual trees that are nor resolved and contained other kind of molecular machines with various abilities. many fast evolving species (Ref [8–11]). There is only Was LUCA also the ancestor of some viruses? Maybe it one 2010 reference indicating that it’s controversial was, but in the present work I only studied the genomes (ref [18]). I would have added my two reviews in 2013 of extant cellular organisms, so I cannot say anything and 2015 in Archaea and Frontier in Microbiology in about a hypothetical ancestor of anything else. which biological arguments against the archaeal ancestor This does not mean that I completely discard the term scenario and against published phylogenies are critically “universal” as being ambiguous: what I was trying to discussed. illustrate was the fact that any term can be ambiguous if Author’s response: Thank you for your comments. not properly defined. And yet, in the present article the Although there are phylogenomic works based on larger sense of the term “cenancestor” is fully defined as “the datasets and excluding fast-evolving sequences which last common ancestor of extant organisms”. This defin- also support the internal branching of the eukaryotes ition includes the term “last”, which I agree is important, within archaea (e.g. Ref. [24]), I agree with Prof. Forterre and identifies the group under consideration (extant on the fact that the origin of the eukaryotic stem is not organisms). The term “organism” itself could be considered set yet. Thus, I have integrated in the manuscript the to be ambiguous, but we need to draw the line somewhere articles suggested by Prof. Forterre. Despite this uncertainty and I think that the reference to “extant organisms” is quite Lombard Biology Direct (2016) 11:36 Page 27 of 31 clear of the fact that I refer to cellular life as we know it. I these phyla may be closely related to each other and have am not hostile to the term “LUCA”, though, and that is been combined into a “superphylum” sometimes called why I included it in the revised version of the manuscript. TACK or proteoarchaea (Refs [10, 53]). In this particular If I use the term of cenancestor it is a matter of preference, case, the labeling as “proteoarchaea” is more informative, as the respective definitions of both terms are virtually as it shows that the sequences group together according to identical. known taxonomical clusters which would not be seen if Jonathan concludes the presence of proteins used for each sequence was labeled with the name of the particular the glycosylation of cell wall in LUCA. The term cell phyla of which sometimes they are the only representative. wall here is ambiguous since it often refers to the bacter- To take an apparently less controversial example, in the ial cell wall. In Archaea, glycosylation occurs at the level Additional file 2 most bacterial sequences are labeled with of the S-layer and in Eukaryotes at the level of the cyto- their phyla names, with the exceptions of the PVC plasmic membrane. This let open many options for the (planctomycetes/verrucomicrobia/chlamydiae) supergroup structure of the LUCA cell envelope. and the Bacteroidetes/Chlorobi supergroup. But in other Author’s response: I have now made sure that the term trees, when a group only contains planctomycetes or bac- “cell wall” is always qualified: “bacterial cell walls” refer teroidetes, it is only labeled with the name of the phylum, to components of the bacterial cell structures beyond the not the superphylum. I do not think that there is a problem plasma membrane (e.g. peptidoglycan, teichoic acids, of logic there, since the objective of these labels is not to exopolysaccharides, etc.); “archaeal cell walls” in prin- discuss what should be the definition of a microbial ciple refers to S-layer proteins, as relatively little is known phylum or superphylum, but just to show that the phylo- about how the other archeal external cell structures are genetic clustering is taxonomically significant. synthesized; “prokaryotic cell wall components” refer to Regarding Fig. 4, the original “proteoarchaeal” group both of the above; “cenancestral cell walls” are hypothet- was paraphyletic, so I changed the labeling completely. I ical, their nature is unknown and they are only men- hope that the new annotations will satisfy the concerns of tioned because the implications of this work may infer the reviewer. their existence. Cell walls will be further discussed in the Otherwise, it’s an important work that opens the way article to be published soon about the evolution of the poly- to future updated phylogenomic analyses of an import- isoprenol biosynthesis pathway. Eukaryotic cell walls are ant physiological pathway. out of the scope of this article as they are synthesized in Author’s response: thank you for your support, advice completely different ways, and they are never mentioned. and interesting debates. Jonathan Lombard still does not use the name Forterre P, Krupovic M, Prangishvili D. Cellular Thaumarchaea in the paper but still use in abundance domains and viral lineages. Trends Microbiol. 2014, the controversial term Proteoarchaea (a putative super- 22(10):554–558. phylum) that look like Proteobacteria (a phylum). I still have problem with this choice. There is a lack of logic. Referee 2, Dr. Sergei Mekhedov, NCBI, NLM, NIH, USA In several figure, group are indicated sometimes by the (nominated by Editorial Board member Michael Galperin) name of their superphylum and other time by the name I cannot understand why the author avoids not just of their phylum (for instance in Fig. 2 [sic, Figure 3?], using the word “ortholog” but also the analysis of ortho- crenarchaeota are sometime named proteoarchaea, logous relationships in the protein families involved in sometimes crenarchae). In Fig. 4, thaumarchaea are N-glycosylation. He writes in his response that there are not indicated despite the fact that they branch with many paralogs in these families. I disagree: these ortho- Euryarchaea and not with crenarchaea. logous groups in Eukaryotes do not have many paralogs. Author’s response: The objective of these labels is to Just look at Table 1! Most columns have only one show when the sequences in these trees group according symbol (+ or # or ?). Every one of these single symbols to reasonable taxonomic clusters and when they do not. that I checked corresponds to a true ortholog of S.cere- This is best done using a mix of names from superphyla, visiae protein(s) used as the basis for Table 1. And if the phyla or even names from lower taxonomic ranks (e.g. in author suspects that some are xenologs (which I could eukaryotes). not identify having checked a part of Table 1) isn’t it the In Fig. 3 there is a group of sequences from crenarchaeota task of this publication to identify them? which is labeled as crenarchaea, and a group of crenarch- Author's response: I am sorry, in my response to the aeota, thaumarchaeota, aigarchaeota, bathyarchaeota first review I did not realize that the reviewer was mainly and lokiarchaeota which is annotated as “proteoarchaea”. concerned with the relationships among the eukaryotic In the latter case, 3 of the 5 suggested “phyla” only contain sequences. I agree with Dr. Mekhedov on the fact that the one sequence, because little genomic data is available for eukaryotes sequences are most likely orthologues. The these groups. Yet, several works have suggested that all of picture is less clear, though, when the evolutionary Lombard Biology Direct (2016) 11:36 Page 28 of 31 relationships are considered across the three domains of glycosylation pathway, whereas the other studies the evo- life. For instance, Fig. 2 shows multiple paralogues of the lution of the polyisoprenol biosynthesis pathways (Dol-P same families in prokaryotes, and some phylogenies & Bac-P) in the three domains of life. Both subjects are support the existence of HGTs. Moreover, many of this distantly related, but these are two independent pieces of prokaryotic homologues carry out completely different work. The other article was mentioned earlier in relation functions. In the new version of the manuscript, I have kept to Prof. Forterre’s request to discuss possible glycosylation the term “homologue” to refer to comparisons among differ- mechanisms in LUCA. But there is nothing in the present ent domains of life and I now use the term “orthologue” to paper that directly relies on the other article. In order to refer to the eukaryotic genes. avoid any misunderstanding, I have removed that refer- I do not like the expression “distant/uncertain hom- ence and I now refer to it as follows: “The implications of ology” in the legend of Table 1. “Uncertain” homologs, these results [meaning the results in this manuscript] for in my opinion, should not be published. the possible cenancestral cell walls will be discussed in a Author’s response: I agree that the wording was not paper to be published soon (Lombard, data not shown).” ideal. The relevance of these sequences was considered as The data presented in Table 1 are not complete and I uncertain because, despite being similar, they were ex- can only guess why the author missed obvious paralogs. tremely divergent, or short, or long or had any other spe- I would like to emphasize that the first step in phylogen- cificity that would put into question the fact that they etic analysis is collection of ALL relevant sequences. The were actual orthologues carrying out the same function. author failed to do this. I do not think that it is referee’s Still, they are good candidates that it would be worth job to find all missed proteins. I checked first three col- testing experimentally, so I think that they should be re- umns for missed paralogs. My general conclusion about ported somehow. In the current version of the article I why the author failed to collect ALL homologs: he used have been less conservative and I have removed that cat- conventional unidirectional BLAST which simply does egory (now all sequences appear with a “+” in Table 1). not detect all homologs. He does not give any cutoff There are 23 columns in Table 1 each corresponding values for BLAST. Unidirectional BLAST is not enough to a protein family involved in N-glycosylation. The to identify all homologs in 2016! The author discarded paper provides analyses of five phylogenetic trees. Of the sequences that are too long which is also a signifi- course, the author might say that the other relevant trees cant mistake. He avoided this mistake in the case of are published in reference [50]. But it is not accessible. fusions in Alg13 and Alg14 groups (because these fu- This may be appropriate in a draft, but not after final sions have been already documented in the literature), review. The author has either remove this reference or but why does he think that homologs in other families make sure that it is available. are always very similar in size? They are not! Neither are Author’sresponse: The objective of this paper was to look orthologs at large distances. This is the reason for his for the possible prokaryotic origins of the eukaryotic genes omission of MOST of the proteins in Swp1 group. involved in the N-glycosylation pathway. There are 23 Obvious paralogs in Alg13 group (one of them missed eukaryotic proteins considered in Table 1, but only 10 of in Table 1): gi|283436186|ref|NP_080523.2| putative them have valuable homologues among prokaryotes. Owing bifunctional UDP-N-acetylglucosamine transferase and to the fact that some of these proteins are paralogues of each deubiquitinase ALG13 [Mus musculus]; gi|260166652| other (Alg2/Alg11, Dpm/Alg5), the 7 phylogenies provided ref|NP_796104.2| glycosyltransferase 28 domain contain- in this paper track the evolution of 9 of these 10 proteins. ing 1-like [Mus musculus]; gi|18405716|ref|NP_565950.1| The remaining protein is Alg1, which is so divergent from UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1- the rest of the large GT1 superfamily that it was impossible phosphate transferase [Arabidopsis thaliana]; gi|15230258| to determine which were its closest prokaryotic relatives. A ref|NP_191281.1| Glycosyl transferase family 4 protein supplementary phylogeny is provided (the one of the PMTs), [Arabidopsis thaliana]. because of its possible relationship to the eukaryotic-specific Author’sresponse: I now provide a completely revised ver- lumen glycosyltransferases. From my point of view, these sion of Table 1. This was made using my previous BLAST phylogenies cover all the proteins required to discuss the results and completed with supplementary BLASTs that prokaryotic origins of the eukaryotic N-glycosylation path- were carried out using alternative seeds that in some cases way. There are no assumptions made in this article about revealed a few more eukaryotic orthologues that had been the origins of the eukaryotic N-glycosylation pathway based missed. Psi-BLASTs were also carried out but the results did on work published anywhere else. not provide any additional information. Finally, the There seems to be a confusion regarding the influence EggNOG and OMA databases were checked out for any on the present article of an article currently under review other orthologue that I may have missed. The newly added (“Reference [50]” in the discussion with the reviewers). Additional file 1 contains the accession numbers of all The present paper traces the origins of the eukaryotic N- sequences that were reported in Table 1. Lombard Biology Direct (2016) 11:36 Page 29 of 31 Additional files 2. Gabaldón T, Huynen MA. From endosymbiont to host-controlled organelle: the hijacking of mitochondrial protein synthesis and metabolism. PLoS Comp Biol. 2007;3:e219. Additional file 1: Accession numbers for sequences in Table 1. (XLS 63 kb) 3. Woese CR, Kandler O, Wheelis ML. Towards a natural system of organisms: Additional file 2: Bayesian phylogeny of the HPT. (PDF 167 kb) proposal for the domains Archaea, Bacteria, and Eucarya. Proc Natl Acad Sci U S A. 1990;87:4576–9. Additional file 3: Full Bayesian phylogeny of the Alg13 and C-terminal 4. Iwabe N, Kuma K, Hasegawa M, Osawa S, Miyata T. Evolutionary relationship catalytic domain of the MurG homologues. (PDF 139 kb) of archaebacteria, eubacteria, and eukaryotes inferred from phylogenetic Additional file 4: Full Bayesian phylogeny of the Alg14 and N-terminal trees of duplicated genes. Proc Natl Acad Sci U S A. 1989;86:9355–9. membrane domain of the MurG homologues. (PDF 139 kb) 5. Gogarten JP, Kibak H, Dittrich P, Taiz L, Bowman EJ, Bowman BJ, Manolson MF, Additional file 5: Bayesian phylogeny of the Alg14 and N-terminal Poole RJ, Date T, Oshima T. Evolution of the vacuolar H + −ATPase: implications membrane domain of the MurG homologues, excluding bacterial MurG for the origin of eukaryotes. Proc Natl Acad Sci U S A. 1989;86:6661–5. sequences. (PDF 134 kb) 6. Cavalier-Smith T. The phagotrophic origin of eukaryotes and phylogenetic Additional file 6: Bayesian phylogeny of the Rft1 homologues. classification of Protozoa. Int J Syst Evol Microbiol. 2002;52:297–354. (PDF 213 kb) 7. Devos DP, Reynaud EG. Intermediate steps. Science. 2010;330:1187–8. 8. Lake JA, Henderson E, Oakes M, Clark MW. Eocytes: a new ribosome Additional file 7: Bayesian phylogeny of the PMT homologues. structure indicates a kingdom with a close relationship to eukaryotes. (PDF 139 kb) Proc Natl Acad Sci U S A. 1984;81:3786–90. Additional file 8: Bayesian phylogeny of the closest relatives to the 9. Cox CJ, Foster PG, Hirt RP, Harris SR, Embley TM. The archaebacterial origin eukaryotic Alg5/Dpm homologues (GT2 superfamily). (PDF 181 kb) of eukaryotes. Proc Natl Acad Sci U S A. 2008;105:20356–61. Additional file 9: List of the genomes used in this analysis. (PDF 106 kb) 10. Guy L, Ettema TJG. The archaeal “TACK” superphylum and the origin of eukaryotes. Trends Microbiol. 2011;19:580–7. Additional file 10: Untrimmed alignments. (TXT 2684 kb) 11. Spang A, Saw JH, Jørgensen SL, Zaremba-Niedzwiedzka K, Martijn J, Lind AE, Additional file 11: Trimmed alignments. (TXT 397 kb) van Eijk R, Schleper C, Guy L, Ettema TJG. Complex archaea that bridge the gap between prokaryotes and eukaryotes. Nature. 2015;521:173–9. 12. Rivera MC, Jain R, Moore JE, Lake JA. Genomic evidence for two functionally Abbreviations distinct gene classes. Proc Natl Acad Sci U S A. 1998;95(May):6239–44. Bac-P, bactoprenol-phosphate; Dol-P, dolichol-phosphate; EPS, exopolysaccharide; 13. Martin W, Müller M. The hydrogen hypothesis for the first eukaryote. Nature. GlcNAc, N-acetyl-glucosamine; GT, glycosyltranferase; GT1, glycosyltransferase 1998;392:37–41. superfamily 1; GT2, glycosyltransferase superfamily 2; HGT, horizontal gene 14. Baum DA, Baum B. An inside-out origin for the eukaryotic cell. BMC Biol. transfer; HMM, hidden Markov model; HPT, polyprenol phosphate:N- 2014;12:76. acetylhexosamine-1-phosphate transferase superfamily; LACA, Last Archaeal 15. Gupta RS, Golding GB. The origin of the eukaryotic cell. Trends Biochem Sci. Common Ancestor; LBCA, Last Bacterial Common Ancestor; LECA, Last 1996;21:166–71. Eukaryotic Common Ancestor; LPS, lipopolysaccharide; PMT, protein 16. Moreira D, Lopez-Garcia P. Symbiosis between methanogenic archaea and O-mannosyltransferase; TACK, Thaumarchaeota, Aigarchaeota, Crenarchaeota, delta-proteobacteria as the origin of eukaryotes: the syntrophic hypothesis. Korarchaeota J Mol Evol. 1998;47:517–30. 17. Margulis L, Dolan MF, Guerrero R. The chimeric eukaryote: origin of the Acknowledgements nucleus from the karyomastigont in amitochondriate protists. Proc Natl I thank U. Gophna and D. Moreira for the stimulating conversations that Acad Sci U S A. 2000;97:6954–9. crystallized in the development of this project and T. A. Richards for his 18. Gribaldo S, Poole AM, Daubin V, Forterre P, Brochier-Armanet C. The origin advice at later steps. I also thank D. Moreira, T. A. Richards, C. Petitjean, F. of eukaryotes and their relationship with the Archaea: are we at a Savory, R. Bamford, V. Attah three anonymous reviewers of a previous phylogenomic impasse? Nat Rev Microbiol. 2010;8:743–52. version, P. Forterre and S. Mekhedov for their comments on the manuscript. 19. Forterre P. The common ancestor of archaea and eukarya was not an archaeon. Archaea. 2013;2013. http://www.ncbi.nlm.nih.gov/pubmed/?term=The+common Funding +ancestor+of+archaea+and+eukarya+was+not+an+archaeon This work was supported by the National Science Foundation through the 20. Forterre P. The universal tree of life: an update. Front Microbiol. 2015;6:717. National Evolutionary Synthesis Center (NESCent, grant number NSF #EF-0905606). 21. Williams TA, Foster PG, Nye TMW, Cox CJ, Embley TM. A congruent phylogenomic signal places eukaryotes within the Archaea. Proc Biol Sci. 2012;279:4870–9. Availability of data and materials 22. McInerney JO, O’Connell MJ, Pisani D. The hybrid nature of the Eukaryota The untrimmed and trimmed alignments supporting the conclusions of this and a consilient view of life on Earth. Nat Rev Microbiol. 2014;12:449–55. article are respectively included in the Additional files 10 and 11. 23. Koonin EV. Archaeal ancestors of eukaryotes: not so elusive any more. BMC Biol. 2015;13:84. 24. Raymann K, Brochier-Armanet C, Gribaldo S. The two-domain tree of life is Author’s contributions linked to a new root for the Archaea. Proc Natl Acad Sci U S A. 2015;112:6670–5. JL designed the study, carried out the analyses and prepared the manuscript. 25. Eme L. Phylogénomique des structures multiprotéiques eucaryotes impliquées dans le cycle cellulaire et contribution à la phylogénie des Competing interests eucaryotes. 2011. The author declares no competing interests. 26. Eme L, Moreira D, Talla E, Brochier-Armanet C. A complex cell division machinery was present in the last common ancestor of eukaryotes. Consent for publication PLoS One. 2009;4:e5021. Not applicable. 27. Bapteste E, Charlebois RL, MacLeod D, Brochier C. The two tempos of nuclear pore complex evolution: highly adapting proteins in an ancient Ethics approval and consent to participate frozen structure. Genome Biol. 2005;6:R85. Not applicable. 28. Gabaldón T, Snel B, Van Zimmeren F, Hemrika W, Tabak H, Huynen MA. Origin and evolution of the peroxisomal proteome. Biol Direct. 2006;1. Received: 11 April 2016 Accepted: 26 July 2016 29. Dacks JB, Field MC. Evolution of the eukaryotic membrane-trafficking system: origin, tempo and mode. J Cell Sci. 2007;120(Pt 17):2977–85. 30. Collins L. Complex spliceosomal organization ancestral to extant eukaryotes. References Mol Biol Evol. 2005;22:1053–66. 1. Embley TM, Martin W. Eukaryotic evolution, changes and challenges. Nature. 31. Richards TA, Cavalier-Smith T. Myosin domain evolution and the primary 2006;440:623–30. divergence of eukaryotes. Nature. 2005;436:1113–8. Lombard Biology Direct (2016) 11:36 Page 30 of 31 32. Ramesh MA, Malik S, Logsdon JM. A phylogenomic inventory of meiotic 59. Oriol R, Martinez-duncker I, Chantret I, Mollicone R, Codogno P. Common genes: evidence for sex in Giardia and an early eukaryotic origin of meiosis. origin and evolution of glycosyltransferases using Dol-P-monosaccharides as Curr Biol. 2005;15:185–91. donor substrate. Mol Biol Evol. 2002;19:1451–63. 33. Apweiler R, Hermjakob H, Sharon N. On the frequency of protein 60. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment glycosylation, as deduced from analysis of the SWISS-PROT database. search tool. J Mol Biol. 1990;215:403–10. Biochim Biophys Acta. 1999;1473:4–8. 61. Lehrman MA. Biosynthesis of N-acetylglucosamine-P-P-dolichol, the 34. Helenius A, Aebi M. Roles of N-linked glycans in the endoplasmic reticulum. committed step of asparagjne-linked oligosaccharide assembly. Annu Rev Biochem. 2004;73:1019–49. Glycobiology. 1991;1:553–62. 35. Schiller B, Hykollari A, Yan S, Paschinger K, Wilson IBH. Complicated N-linked 62. Chantret I, Dancourt J, Barbat A, Moore SEH. Two proteins homologous to glycans in simple organisms. Biol Chem. 2012;393:661–73. the N- and C-terminal domains of the bacterial glycosyltransferase MurG are 36. Eichler J. Extreme sweetness: protein glycosylation in archaea. Nat Rev required for the second step of dolichyl-linked oligosaccharide synthesis in Microbiol. 2013;11:151–6. Saccharomyces cerevisiae. J Biol Chem. 2005;280:9236–42. 37. Albers S-V, Meyer BH. The archaeal cell envelope. Nat Rev Microbiol. 63. Gao X-D, Tachikawa H, Sato T, Jigami Y, Dean N. Alg14 recruits Alg13 to the 2011;9:414–26. cytoplasmic face of the endoplasmic reticulum to form a novel bipartite 38. Guan Z, Naparstek S, Kaminski L, Konrad Z, Eichler J. Distinct glycan-charged UDP-N-acetylglucosamine transferase required for the second step of phosphodolichol carriers are required for the assembly of the pentasaccharide N-linked glycosylation. J Biol Chem. 2005;280:36254–62. N-linked to the Haloferax volcanii S-layer glycoprotein. Mol Microbiol. 64. Lu J, Takahashi T, Ohoka A, Nakajima KI, Hashimoto R, Miura N, Tachikawa H, 2010;78:1294–303. Gao X-D. Alg14 organizes the formation of a multiglycosyltransferase complex involved in initiation of lipid-linked oligosaccharide biosynthesis. 39. Guan Z, Meyer BH, Albers S-V, Eichler J. The thermoacidophilic archaeon Glycobiology. 2011;22:504–16. Sulfolobus acidocaldarius contains an unsually short, highly reduced dolichyl phosphate. Biochim Biophys Acta. 2011;1811:607–16. 65. Lehrman MA. A family of UDP-GIcNAc/MurNAc: polyisoprenol-P GlcNAc/ 40. Jarrell KF, Ding Y, Meyer BH, Albers S-V, Kaminski L, Eichler J. N-linked MurNAc-1-P transferases. Glycobiology. 1994;4:768–71. glycosylation in Archaea: a structural, functional, and genetic analysis. 66. Dal Nogare AR, Dan N, Lehrman MA. Conserved sequences in enzymes of Microbiol Mol Biol Rev. 2014;78:304–41. the UDP-GlcNAc/MurNAc family are essential in hamster UDP-GlcNAc: 41. Chang MM, Imperiali B, Eichler J, Guan Z. N-Linked glycans are assembled dolichol-P GlcNAc-1-P transferase. Glycobiology. 1998;8:625–32. on highly reduced dolichol phosphate carriers in the hyperthermophilic 67. Anderson MS, Eveland SS, Price NPJ. Conserved cytoplasmic motifs that archaea Pyrococcus furiosus. Plos One. 2015;10:e0130482. distinguish sub-groups of the polyprenol phosphate : N-acetylhexosamine-1- 42. Cohen-Rosenzweig C, Yurist-Doutsch S, Eichler J. AglS, a novel component phosphate transferase family. FEMS Microbiol Lett. 2000;191:169–75. of the Haloferax volcanii N-glycosylation pathway, is a dolichol phosphate- 68. Bouhss A, Trunkfield AE, Bugg TDH, Mengin-Lecreulx D. The biosynthesis of mannose mannosyltransferase. J Bacteriol. 2012;194:6909–16. peptidoglycan lipid-linked intermediates. FEMS Microbiol Rev. 2008;32:208–33. 43. Nothaft H, Szymanski CM. Protein glycosylation in bacteria: sweeter than 69. Mohammadi T, Karczmarek A, Crouvoisier M, Bouhss A, Mengin-Lecreulx D, ever. Nat Rev Microbiol. 2010;8:765–78. den Blaauwen T. The essential peptidoglycan glycosyltransferase MurG 44. Jervis AJ, Langdon R, Hitchen P, Lawson AJ, Wood A, Fothergill JL, Morris HR, forms a complex with proteins involved in lateral envelope growth as well Dell A, Wren B, Linton D. Characterization of N-linked protein glycosylation in as with proteins involved in cell division in Escherichia coli. Mol Microbiol. Helicobacter pullorum. J Bacteriol. 2010;192:5228–36. 2007;65:1106–21. 45. Meyer BH, Albers S-V. Hot and sweet: protein glycosylation in 70. Shams-Eldin H, Chaban B, Niehus S, Schwarz RT, Jarrell KF. Identification Crenarchaeota. Biochem Soc Trans. 2013;41:384–92. of the archaeal alg7 gene homolog (encoding N-acetylglucosamine-1- 46. Samuelson J, Banerjee S, Magnelli P, Cui J, Kelleher DJ, Gilmore R, Robbins PW. phosphate transferase) of the N-linked glycosylation system by cross- The diversity of dolichol-linked precursors to Asn-linked glycans likely results domain complementation in Saccharomyces cerevisiae. J Bacteriol. from secondary loss of sets of glycosyltransferases. Proc Natl Acad Sci U S A. 2008;190:2217–20. 2005;102:1548–53. 71. Larkin A, Chang M, Whitworth G, Imperiali B. Biochemical evidence for an 47. Maita N, Nyirenda J, Igura M, Kamishikiryo J, Kohda D. Comparative alternate pathway in N-linked glycoprotein biosynthesis. Nat Chem Biol. structural biology of eubacterial and archaeal oligosaccharyltransferases. 2013;9:367–73. J Biol Chem. 2010;285:4941–50. 72. Kolkman MAB, Van der Zeijst BAM, Nuijten PJM. Functional analysis 48. Schwarz F, Aebi M. Mechanisms and principles of N-linked protein of glycosyltransferases encoded by the capsular polysaccharide glycosylation. Curr Opin Struct Biol. 2011;21:576–82. biosynthesis locus of Streptococcus pneumoniae serotype 14. J Biol 49. Bugg T, Brandish P. From peptidoglycan to glycoproteins: common Chem. 1997;272:19502–8. features of lipid-linked oligosaccharide biosynthesis. FEMS Microbiol Lett. 73. De Vuyst L, De Vin F, Vaningelgem F, Degeest B. Recent developments in 1994;119:255–62. the biosynthesis and applications of heteropolysaccharides from lactic acid 50. Burda P, Aebi M. The dolichol pathway of N -linked glycosylation. Biochim bacteria. Inter Diary J. 2001;11:687–707. Biophys Acta. 1999;1426:239–57. 74. Meyer BH, Peyfoon E, Dietrich C, Hitchen P, Panico M, Morris HR, Dell A, 51. Noffz C, Keppler-Ross S, Dean N. Hetero-oligomeric interactions between Albers S-V. Agl16, a thermophilic glycosyltransferase mediating the last step early glycosyltransferases of the dolichol cycle. Glycobiology. 2009;19:472–8. of N-Glycan biosynthesis in the thermoacidophilic crenarchaeon Sulfolobus acidocaldarius. J Bacteriol. 2013;195:2177–86. 52. Petitjean C, Deschamps P, López-García P, Moreira D. Rooting the domain archaea by phylogenomic analysis supports the foundation of the new 75. Maeda Y, Kinoshita T. Dolichol-phosphate mannose synthase: Structure, kingdom Proteoarchaeota. Genome Biol Evol. 2015;7:191–204. function and regulation. Biochim Biophys Acta. 2008;1780:861–8. 53. Kelleher DJ, Banerjee S, Cura AJ, Samuelson J, Gilmore R. Dolichol-linked 76. Helenius J, Ng DTW, Marolda CL, Walter P, Valvano MA, Aebi M. oligosaccharide selection by the oligosaccharyltransferase in protist and Translocation of lipid-linked oligosaccharides across the ER membrane fungal organisms. J Cell Biol. 2007;177:29–37. requires Rft1 protein. Nature. 2002;415:447–50. 54. Gao X-D, Nishikawa A, Dean N. Physical interactions between the Alg1, 77. Frank CG, Sanyal S, Rush JS, Waechter CJ, Menon AK. Does Rft1 flip an Alg2, and Alg11 mannosyltransferases of the endoplasmic reticulum. N-glycan lipid precursor? Nature. 2008;454:E3–4. discussion E4–5. Glycobiology. 2004;14:559–70. 78. Rush JS, Gao N, Lehrman MA, Matveev S, Waechter CJ. Suppression of Rft1 55. Ullman C, Perkins S. A classification of nucleotide-diphospho-sugar expression does not impair the transbilayer movement of Man5GlcNAc2-P-P- glycosyltransferases based on amino acid sequence similarities. Biochem J. dolichol in sealed microsomes from yeast. J Biol Chem. 2009;284:19835–42. 1997;326(Pt 3):929–39. 79. Jelk J, Gao N, Serricchio M, Signorell A, Schmidt RS, Bangs JD, 56. Liu J, Mushegian A. Three monophyletic superfamilies account for the Acosta-Serrano A, Lehrman MA, Bütikofer P, Menon AK. Glycoprotein majority of the known glycosyltransferases. Prot Sci. 2003;12:1418–31. biosynthesis in a eukaryote lacking the membrane protein Rft1. 57. Coutinho PM, Deleury E, Davies GJ, Henrissat B. An evolving hierarchical J Biol Chem. 2013;288:20616–23. family classification for glycosyltransferases. J Mol Biol. 2003;328:307–17. 80. Eisenhaber B, Maurer-Stroh S, Novatchkova M, Schneider G, Eisenhaber F. 58. Lairson LL, Henrissat B, Davies GJ, Withers SG. Glycosyltransferases: Enzymes and auxiliary factors for GPI lipid anchor biosynthesis and structures, functions, and mechanisms. Annu Rev Biochem. 2008;77:521–55. post-translational transfer to proteins. Bioessays. 2003;25:367–85. Lombard Biology Direct (2016) 11:36 Page 31 of 31 81. Finn RD, Clements J, Arndt W, Miller BL, Wheeler TJ, Schreiber F, Bateman A, 103. Altenhoff AM, Škunca N, Glover N, Train C-M, Sueki A, Piližota I, Gori K, Eddy SR. HMMER web server: 2015 update. Nucleic Acids Res. 2015;43:W30–8. Tomiczek B, Müller S, Redestig H, Gonnet GH, Dessimoz C. The OMA orthology 82. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ. database in 2015: function predictions, better plant support, synteny view and Gapped BLAST and PSI-BLAST : a new generation of protein database other improvements. Nucleic Acids Res. 2015;2015(43):D240–9. search programs. Nucleic Acids Res. 1997;25:3389–402. 104. Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32:1792–7. 83. Lussier M, Sdicu A, Bussey H. The KTR and MNN1 mannosyltransferase 105. Philippe H. MUST, a computer package of management utilities for families of Saccharomyces cerevisiae. J Biol Chem. 1995;270:2770–5. sequences and trees. Nucleic Acids Res. 1993;21:5264–72. 84. Gentzsch M, Tanner W. The PMT gene family: protein O-glycosylation in 106. Price MN, Dehal PS, Arkin AP. FastTree 2–approximately maximum-likelihood Saccharomyces cerevisiae is vital. EMBO J. 1996;15:5752–9. trees for large alignments. PLoS One. 2010;5:e9490. 85. Loibl M, Strahl S. Protein O-mannosylation: what we have learned from 107. Penn O, Privman E, Ashkenazy H, Landan G, Graur D, Pupko T. GUIDANCE: baker’s yeast. Biochim Biophys Acta. 2013;1833:2438–46. a web server for assessing alignment confidence scores. Nucleic Acids Res. 86. VanderVen BC, Harder JD, Crick DC, Belisle JT. Export-mediated assembly 2010;38(Web Server issue):W23–8. of mycobacterial glycoproteins parallels eukaryotic pathways. Science. 108. Ronquist F, Teslenko M, van der Mark P, Ayres DL, Darling A, Höhna S, 2005;309:941–3. Larget B, Liu L, Suchard MA, Huelsenbeck JP. MrBayes 3.2: efficient Bayesian 87. Lommel M, Strahl S. Protein O-mannosylation: conserved from bacteria to phylogenetic inference and model choice across a large model space. humans. Glycobiology. 2009;19:816–28. Syst Biol. 2012;61:539–42. 88. Meyer BH, Zolghadr B, Peyfoon E, Pabst M, Panico M, Morris HR, Haslam SM, 109. Stamatakis A. RAxML version 8: a tool for phylogenetic analysis and Messner P, Schäffer C, Dell A, Albers S-V. Sulfoquinovose synthase - an post-analysis of large phylogenies. Bioinformatics. 2014;30:1312–3. important enzyme in the N-glycosylation pathway of Sulfolobus acidocaldarius. 110. Miller MA, Pfeiffer W, Schwartz T. Creating the CIPRES Science Gateway for Mol Microbiol. 2011;82:1150–63. Inference of Large Phylogenetic Trees. In: Proceedings of the Gateway 89. Yan Q, Lennarz WJ. Stt3p is directly involved in the glycosylation process. Computing Environments Workshop (GCE). New Orleans; 2010:1–8. http:// J Biol Chem. 2002;277:47692–700. www.phylo.org/sub_sections/portal/cite.php 90. Koiwa H, Li F, Mccully MG, Mendoza I, Koizumi N, Manabe Y, Nakagawa Y, 111. Le SQ, Gascuel O. An improved general amino acid replacement matrix. Zhu J, Rus A, Pardo JM, Bressan RA, Hasegawa PM. The STT3a subunit Mol Biol Evol. 2008;25:1307–20. isoform of the Arabidopsis oligosaccharyltransferase controls adaptive 112. Le SQ, Dang CC, Gascuel O. Modeling protein evolution with several responses to salt/osmotic stress. Plant Cell. 2010;15(October 2003):2273–84. amino acid replacement matrices depending on site rates. Mol Biol Evol. 91. Chaban B, Voisin S, Kelly J, Logan SM, Jarrell KF. Identification of genes 2012;29:2921–36. involved in the biosynthesis and attachment of Methanococcus voltae 113. Kaminski L, Guan Z, Yurist-doutsch S. Two distinct N-glycosylation pathways N-linked glycans: insight into N-linked glycosylation pathways in Archaea. process the Haloferax volcanii S-layer glycoprotein upon changes in Mol Microbiol. 2006;61:259–68. environmental salinity. mBio. 2013;4:1–9. 92. Igura M, Maita N, Obita T, Kamishikiryo J, Maenaka K, Kohda D. Purification, 114. Typas A, Banzhaf M, Gross CA, Vollmer W. From the regulation of crystallization and preliminary X-ray diffraction studies of the soluble peptidoglycan synthesis to bacterial growth and morphology. Nat Rev domain of the oligosaccharyltransferase STT3 subunit from the thermophilic Microbiol. 2011;10:123–36. archaeon Pyrococcus furiosus. Acta Crystallogr Sect F: Struct Biol Cryst 115. Raetz CRH, Whitfield C. Lipopolysaccharide endotoxins. Annu Rev Biochem. Commun. 2007;63:798–801. 2002;71:635–700. 93. VanDyke DJ, Wu J, Logan SM, Kelly JF, Mizuno S, Aizawa S-I, Jarrell KF. 116. Brown S, Santa Maria JP, Walker S. Wall teichoic acids of gram-positive Identification of genes involved in the assembly and attachment of a novel bacteria. Annu Rev Microbiol. 2013;67:313–36. flagellin N-linked tetrasaccharide important for motility in the archaeon 117. Vik Å, Erik F, Haug J, Bilsborough S, Schneider A, Egge-jacobsen W, Koomey M. Methanococcus maripaludis. Mol Microbiol. 2009;72:633–44. Broad spectrum O-linked protein glycosylation in the human pathogen 94. Matsumoto S, Shimada A, Nyirenda J, Igura M, Kawano Y, Kohda D. Crystal Neisseria gonorrhoeae. Proc Natl Acad Sci U S A. 2009;106:4447–52. structures of an archaeal oligosaccharyltransferase provide insights into the 118. Messner P, Steiner K, Zarschler K, Schäffer C. S-layer nanoglycobiology of catalytic cycle of N-linked protein glycosylation. Proc Natl Acad Sci U S A. bacteria. Carbohydr Res. 2008;343:1934–51. 2013;110:17868–73. 119. Rick PD, Barr K, Sankaran K, Kajimura J, Rush JS, Waechter CJ. Evidence that 95. Meyer BH, Albers S-V. AglB, catalyzing the oligosaccharyl transferase step of the wzxE gene of Escherichia coli K-12 encodes a protein involved in the the archaeal N-glycosylation process, is essential in the thermoacidophilic transbilayer movement of a trisaccharide-lipid intermediate in the assembly crenarchaeon Sulfolobus acidocaldarius. MicrobiologyOpen. 2014;3:531–43. of enterobacterial common antigen. J Biol Chem. 2003;278:16534–42. 96. Magidovich H, Eichler J. Glycosyltransferases and oligosaccharyltransferases 120. Whitfield C. Biosynthesis and assembly of capsular polysaccharides in in Archaea: putative components of the N-glycosylation pathway in the Escherichia coli. Annu Rev Biochem. 2006;75:39–68. third domain of life. FEMS Microbiol Lett. 2009;300:122–30. 121. Vorhölter F-J, Schneiker S, Goesmann A, Krause L, Bekel T, Kaiser O, Linke B, 97. Kaminski L, Lurie-Weinberger MN, Allers T, Gophna U, Eichler J. Phylogenetic- Patschkowski T, Rückert C, Schmid J, Sidhu VK, Sieber V, Tauch A, Watt SA, and genome-derived insight into the evolution of N-glycosylation in Archaea. Weisshaar B, Becker A, Niehaus K, Pühler A. The genome of Xanthomonas Mol Phylogenet Evol. 2013;68:327–39. campestris pv. campestris B100 and its use for the reconstruction of 98. Heesen S, Lehle L, Weissmann A, Aebi M. Isolation of the ALG5 locus metabolic pathways involved in xanthan biosynthesis. J Biotechnol. encoding the UDP-glucose : dolichyl-phosphate glucosyltransferase from 2008;134:33–45. Saccharomyces cerevisiae. Eur J Biochem. 1994;224:71–9. 122. Reuber TL, Walker GC. Biosynthesis of succinoglycan, a symbiotically 99. Orlean P, Albright C, Robbins PW. Cloning and sequencing of the yeast important exopolysaccharide of Rhizobium keliioti. Cell. 1993;74:269–80. gene for dolichol phosphate mannose synthase, an essential protein. J Biol Chem. 1988;263:17499–507. 100. Colussi PA, Taron CH, Mack JC, Orlean P. Human and Saccharomyces cerevisiae dolichol phosphate mannose synthases represent two classes of the enzyme, but both function in Schizosaccharomyces pombe. Proc Natl Acad Sci U S A. 1997;94(July):7873–8. 101. Zarschler K, Janesch B, Pabst M, Altmann F, Messner P, Schäffer C. Protein tyrosine O-glycosylation–a rather unexplored prokaryotic glycosylation system. Glycobiology. 2010;20:787–98. 102. Huerta-Cepas J, Szklarczyk D, Forslund K, Cook H, Heller D, Walter MC, Rattei T, Mende DR, Sunagawa S, Kuhn M, Jensen LJ, von Mering C, Bork P. eggNOG 4.5: a hierarchical orthology framework with improved functional annotations for eukaryotic, prokaryotic and viral sequences. Nucleic Acids Res. 2016;44:D286–293. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Biology Direct Springer Journals

The multiple evolutionary origins of the eukaryotic N-glycosylation pathway

Biology Direct , Volume 11 (1) – Aug 4, 2016

Loading next page...
 
/lp/springer-journals/the-multiple-evolutionary-origins-of-the-eukaryotic-n-glycosylation-Kw5r25sgGW

References (122)

Publisher
Springer Journals
Copyright
Copyright © 2016 by The Author(s).
Subject
Life Sciences; Life Sciences, general
eISSN
1745-6150
DOI
10.1186/s13062-016-0137-2
pmid
27492357
Publisher site
See Article on Publisher Site

Abstract

Background: The N-glycosylation is an essential protein modification taking place in the membranes of the endoplasmic reticulum (ER) in eukaryotes and the plasma membranes in archaea. It shares mechanistic similarities based on the use of polyisoprenol lipid carriers with other glycosylation pathways involved in the synthesis of bacterial cell wall components (e.g. peptidoglycan and teichoic acids). Here, a phylogenomic analysis was carried out to examine the validity of rival hypotheses suggesting alternative archaeal or bacterial origins to the eukaryotic N-glycosylation pathway. Results: The comparison of several polyisoprenol-based glycosylation pathways from the three domains of life shows that most of the implicated proteins belong to a limited number of superfamilies. The N-glycosylation pathway enzymes are ancestral to the eukaryotes, but their origins are mixed: Alg7, Dpm and maybe also one gene of the glycosyltransferase 1 (GT1) superfamily and Stt3 have proteoarchaeal (TACK superphylum) origins; alg2/alg11 may have resulted from the duplication of the original GT1 gene; the lumen glycosyltransferases were probably co-opted and multiplied through several gene duplications during eukaryogenesis; Alg13/Alg14 are more similar to their bacterial homologues; and Alg1, Alg5 and a putative flippase have unknown origins. Conclusions: The origin of the eukaryotic N-glycosylation pathway is not unique and less straightforward than previously thought: some basic components likely have proteoarchaeal origins, but the pathway was extensively developed before the eukaryotic diversification through multiple gene duplications, protein co-options, neofunctionalizations and even possible horizontal gene transfers from bacteria. These results may have important implications for our understanding of the ER evolution and eukaryogenesis. Reviewers: This article was reviewed by Pr. Patrick Forterre and Dr. Sergei Mekhedov (nominated by Editorial Board member Michael Galperin). Keywords: N-glycosylation, Eukaryotes, Archaea, Bacteria, Eukaryogenesis, Prokaryotic cell walls, Polyisoprenol, Glycosyltransferase Background contributor to eukaryogenesis, for at least the mitochondria At the dawn of eukaryotes are known to have evolved from an alpha-proteobacterium All living organisms are traditionally classified into one that was engulfed prior to the last eukaryotic common of three domains of life, namely bacteria, archaea and ancestor (LECA) [1, 2]. The traditional three-domain tree eukaryotes. Eukaryotic cells have important cytological of life posits that eukaryotes and archaea are sister groups specificities, such as nuclei, organelles and other cellular [3–5] and, thus, it assumes that the mitochondrial ances- processes. As a result, the origin of eukaryotes from tor was incorporated in a pre-existing proto-eukaryotic former organisms is one of the most intriguing questions lineage [6, 7]. Other phylogenetic studies have suggested in biology. Endosymbiosis was certainly an important that the eukaryotic stem branched within the archaeal domain, close to the crenarchaea–“eocytes”–or to the Correspondence: j.lombard@exeter.ac.uk archaeal TACK superphylum [8–11]. Also, comparative National Evolutionary Synthesis Center, 2024 W. Main Street Suite A200, genomics have shown that many eukaryotic operational Durham, NC 27705, USA 2 (metabolic) genes are closely related to bacterial Biosciences, University of Exeter, Geoffrey Pope Building, Stocker Road, Exeter EX4 4QD, UK © 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lombard Biology Direct (2016) 11:36 Page 2 of 31 homologues, while informational genes are more similar [42]. Archaeal glycans are also more diverse than their to archaeal ones [12]. These observations popularized the eukaryotic equivalents [40]. opinion according to which the eukaryotes are a chimeric N-glycosylation is less common in bacteria, where two lineage that resulted from the endosymbiosis of the N-glycosylation systems are known [43]. One of them bacterial ancestor of mitochondria within a bona fide independently tethers single monosaccharides to specific archaeon [13, 14] or a previous bacterium/archaeon con- amino acids using soluble cytoplasmic proteins. The sortium [15–17]. The recent accumulation of genomic other relies on a lipid-carrier-based mechanism similar data has now made possible to address these issues. The to the eukaryotic and archaeal pathways (Fig. 1b). debate remains open [18–20], but most authors now favor The latter is thought to be restricted to delta- and an archaeal origin of the eukaryotic stem [21–24]. epsilon-proteobacteria [43, 44]. The bacterial polypre- Following the success of comparative analyses to trace nol lipid carriers are unsaturated and have different back particular machineries to LECA [25–32], the origin names depending on their lengths [45]. For simplicity, and evolution of the eukaryotic N-glycosylation pathway they will all be referred to as bactoprenol phosphate will be studied here. (Bac-P). N-glycosylation in the three domains of life The evolution of polyisoprenol-based N-glycosylations More than half of all eukaryotic proteins are glycopro- Since polyisoprenol-based N-glycosylation pathways are teins, and 90 % of those are N-glycosylated [33]. Protein widespread in archaea and eukaryotes but uncommon in N-glycosylation modulates protein stability, solubility, bacteria, it has been speculated that this pathway origi- rigidity, orientation, interactivity, transport and signaling nated in archaea, was horizontally transferred to some [34]. It consists of the covalent attachment of an oligo- bacteria [46, 47] and inherited in eukaryotes from their saccharide to the nitrogen atom of specific asparagine putative archaeal-related progenitor [34, 48]. The poten- residues. In eukaryotes, the synthesis of the oligosac- tial archaeal origin of eukaryotic N-glycosylation recently charide core is mediated by a lipid carrier called brought this pathway into the spotlight of eukary- dolichol-phosphate (Dol-P) which is located in the ogenesis debates [14]. For instance, the “inside-out membranes of the endoplasmic reticulum (ER, Fig. 1a). hypothesis” recently suggested that the eukaryotic In opisthokonts (e.g. in humans and yeast), several nucleus evolved from ancestral “eocytes” (crenarchaeota) cytoplasmic-facing membrane-embedded glycosyltrans- that protruded membrane-bound blebs beyond their cell ferases (GTs) sequentially add monosaccharides from walls to increase the available exchange surface with the soluble nucleotide carriers to Dol-P molecules (Fig. 1b). bacterial ancestors of mitochondria. According to this The Dol-P-oligosaccharide is then “flipped” across the hypothesis, the location of the N-glycosylation pathway membrane and other membrane-embedded GTs resume in the eukaryotic ER membranes is a vestige of the the oligosaccharide decoration in the lumen side. The ancestral archaeal S-layer protein N-glycosylation [14]. monosaccharides used in the ER lumen are also attached The idea that eukaryotic N-glycosylation originated to Dol-P carriers. The final oligosaccharide is transferred from archaea is plausible, but alternative hypotheses en bloc to the acceptor protein located in the ER lumen have also been suggested. For instance, the eukaryotic by a N-oligosaccharyltransferase (N-OST). Soluble GTs N-glycosylation has been thought to be related to some located in the ER and Golgi lumen may subsequently kind of bacterial cell wall synthesis pathway [49–51]. modify the protein N-glycans [35], but these later modi- Indeed, bacterial O-glycosylation, tyrosine O-glycosylation, fications will not be studied in the present analysis. peptidoglycan synthesis, O-antigen LPS synthesis, teichoic In archaea, flagellins and cell wall S-layer proteins are acid synthesis, exopolysaccharide synthesis and, in a N-glycosylated using similar pathways located in their smaller scale, O- and C-mannosylation, use polyisoprenol cell membranes [36, 37]. Archaeal N-glycosylation lipid lipid carriers and a similar orientation across the carriers are traditionally called Dol-P, even though they membrane to protein N-glycosylation. Other observations, are more saturated than the eukaryotic ones (Fig. 1a, without being conclusive, call into question a simple [38–41]). Archaeal N-glycosylation pathways have a transposition of the archaeal N-glycosylation into eukary- topology similar to their eukaryotic counterparts, i.e. the otes: i) archaeal and eukaryotic Dol-P are different orientation of their reactions across the membrane is the molecules (Fig. 1a) [38–41] and they are probably same: archaeal N-glycans are synthesized in the cytoplas- synthesized using different biosynthesis pathways (data to mic face, then flipped to the outer side of cell mem- be published soon; Lombard, data not shown); and ii) branes and transferred to the protein (Fig. 1b, [37]). eukaryotic cytoplasmic GTs have homologues in Contrary to eukaryotes, however, the N-glycan chains prokaryotes but the precise evolutionary relationships are seldom modified in the outer side of archaeal mem- between eukaryotic and prokaryotic sequences are not branes, with the notable exception of Haloferax volcanii known yet [46]. Lombard Biology Direct (2016) 11:36 Page 3 of 31 Fig. 1 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 4 of 31 (See figure on previous page.) Fig. 1 Basic components of the polyisoprenol-based machineries. a Typical polyisoprenol lipid carriers in the three domains of life. b Topology (membrane orientation) of characteristic N-glycosylation pathways in the three domains of life and bacterial peptidoglycan biosynthesis. The lipid carrier is embedded in a cell membrane (ER membrane in eukaryotes, plasma membranes in archaea and bacteria), first facing the cytoplasmic side, then flipped to the opposite side (i.e. ER lumen in eukaryotes, periplasm in prokaryotes). Monosaccharides are attached one by one to the lipid carriers by specific glycosyltransferases, although in the eukaryotic N-glycosylation each kind of sugar is only represented once per compartment, for simplicity. The monosaccharides are nucleotide-activated in the cytoplasmic side or translocated to the ER lumen by separate lipid carriers In fact, the available evidence to support a bacterial or in some lineages [46]. In parasites, such losses may argu- archaeal origin of the eukaryotic N-glycosylation path- ably have been a convenient mechanism to synthesize way is limited, so a phylogenomic analysis was carried non-canonical N-glycans and escape the host defenses. out to tackle this issue. The phylogenetic study of the in- volved protein superfamilies suggests that this pathway Gene homology across polyisoprenol-based machineries was present in LECA and that its components have The main argument put forward to suggest bacterial or diverse origins. Some genes likely had archaeal ancestors– archaeal origins to the eukaryotic N-glycosylation especially from the TACK superphylum, to which I will pathway has been the use of polyisoprenol carriers and refer to as proteoarchaea [52]. Although this gives credit topological similarity, i.e. the eukaryotic pathway has the to hypotheses invoking a close relationship between eu- same functional orientation across the membrane than karyotes and proteoarchaea, many N-glycosylation genes several prokaryotic glycosylation pathways (Fig. 1b). specifically evolved in eukaryotes through gene duplica- Despite these similarities, the homology of the proteins tion, protein co-option and neofunctionalization. Two of implicated in these pathways had never been systematic- them could even have bacterial origins. Given the import- ally surveyed before. ance that some eukaryogenesis scenarios have conceded Most proteins involved in polyisoprenol-based pathways to the origins of the N-glycosylation pathway and its are glycosyltransferases (GTs). These enzymes are very com- presence in the ER membranes, these results may have mon and they are classified according to their structure, important consequences in debates about the origins of catalytic mechanism and amino acid sequences [55–58]. An eukaryotes and the ER. updated classification is available in the CAZy database (http://www.cazy.org/GlycosylTransferases.html) which, as Results and discussion of August 2016, contained 100 GT families. Nonetheless, Distribution of the N-glycosylation pathway proteins their extreme diversity, poor sequence conservation and across eukaryotes ability to easily change their substrate specificity make an The set of N-glycosylation GTs was reported to be very extensive in-depth evolutionary study of all GTs across variable among eukaryotes [46, 53], therefore raising the the three domains of life virtually impossible. question if this pathway was ancestral to all eukaryotes The pathways summarized in Table 2 were selected for or progressively developed during eukaryotic evolution this study because, according to the literature, they use [46, 48]. The recent accumulation of genomic data polyisoprenol lipid carriers and have similar topologies allows us to provide a more comprehensive perspective to the eukaryotic N-glycosylation pathway. The eukar- on the distribution of these genes across eukaryotes yotic GPI pathway was included notwithstanding being (Table 1, Additional file 1). With regard to the N-OST, polyisoprenol-independent because its mannosyltrans- the Stt3 catalytic unit is well conserved and most other ferases are the only known relatives of the eukaryotic subunits–except for those very short and arguably more N-glycosylation lumen GTs [59]. difficult to detect–are found in all major eukaryotic The proteins in the pathways from Table 2 were used to groups but haptophytes and cryptophytes. GTs and carry out all vs. all reciprocal BLASTp searches [60]. The putative flippase orthologues are found in all major results, together with some information from the litera- eukaryotic lineages. The cytoplasmic GTs are more ture and occasional HMMER confirmations, allowed the conserved than their lumenal counterparts. This is in construction of the putative homology groups reported in agreement with the observation that lumen GTs dele- Fig. 2 (see Methods). This dataset is limited to the path- tions have milder phenotypes than cytoplasmic GTs mu- ways characterized in model organisms and does not have tations [54]. The widespread distribution of the canonical a strong statistical value. Yet, it shows that most proteins N-glycosylation pathway proteins in eukaryotes and their from polyisoprenol-dependent pathways belong to a lim- consistent monophyly in protein family trees (see below) ited number of homology groups. For instance, the GTs support the ancestral presence of this pathway in LECA and colored in blue (GT group 1), red (GT group 2) and green the subsequent independent loss of some of these enzymes (HPT family), plus the khaki flippase group, are present in Lombard Biology Direct (2016) 11:36 Page 5 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes Alg7 Alg13 Alg14 Alg1 Alg2 Alg11 Rft1 Alg3 Alg9 Alg12 Alg6 Alg8 Alg10 Stt3 Ost1 Ost2 Ost3/6 Ost4 Ost5 Wbp1 Swp1 Dmp1 Alg5 Saccharomyces cerevisiae ++ + + ++ + ++ + ++ + + ++ ++ + ++ + + + Schizosaccharomyces pombe ++ + + ++ + ++ + ++ + + ++ + + + + + + Ustilago maydis ++ + + ++ + ++ + ++ + + ++ + + + + + Cryptococcus neoformans + + + ++ + + + ++ + + ++ + + + + Rhizophagus irregularis ++ + + ++ + ++ + ++++ + + ++ + + + + + Batrachochytrium dendrobatidis ++ + + + ++ + ++ + ++ + + ++ + + + + + Spizellomyces punctatus ++ + ++ + ++ ++ ++ + ++ + + + + + Allomyces macrogynus ++ ++ ++ ++ + ++ ++ ++ + ++ ++ ++ ++ ++ ++ ++ ++ ++ + ++ ++ Trichoplax adhaerens + +++ +++ + +++ + +++ + + + + + + + Amphimedon queenslandica + +++ + + + + + ++ + + + +++ + 5+ + + +++ + + + + + Mnemiopsis leidyi ++ + + ++ + + + + ++ ++ + + + + Hydra magnipapillata ++ + + ++ + + + + ++ ++ + + + + Hydra magnipapillata ++ + + ++ + ++ + ++ + +++ ++ ++ + + + + Drosophila melagonaster ++ + + ++ + ++ + ++ + ++ ++ + ++ + + + + Aplysia californica ++ + + ++ + ++ + ++ + +++ ++ + + + + + ++ Strongylocentrotus purpuratus + + + ++ + + ++ + + + + + + +++ + + ++ + + + +++ Mus musculus +++ + ++ +++ + ++ ++ ++ + ++ ++ ++ + + ++ + + Monosiga brevicollis + +++ + ++ + + + ++ + Salpingoeca rosetta + ++++ + ++ + + ++ + +++ + + + + + + + Capsaspora owczarzaki ++ + + ++ + ++ + +++ + ++ ++ + + + + ++ + Fonticula alba ++ + + ++ + ++ + ++ + + + + Thecamonas trahens ++ + + ++ + ++ + ++ + ++ ++ + + + + + Entamoeba histolytica + + + + ++ + + +++ ++ + +++ + Acanthamoeba castellanii + +++ + ++ + + + + ++ + + + + + Dictyostelium discoideum + +++ + ++ ++ + ++ + ++ + + + + + + + ++ Polysphondylium pallidum + +++ + ++ ++ + ++ + ++ + + ++ + + + + + Chlamydomonas reinhardtii + +++ + ++ + + ++ + + + + + + ++ Volvox carteri ++ + + ++ + ++ + ++ ++ + + + + + Chlorella variabilis + +++ + ++ + ++ + ++ + + + + + + + Asterochlorissp. ++ + + ++ ++ + ++ + ++ + + + + Bathycoccus prasinos + +++ + ++ + + + + + Micromonas pusilla ++ + + ++ ++ + ++ ++ + + Ostreococcus lucimarinus + +++ + ++ + + + + + + Lombard Biology Direct (2016) 11:36 Page 6 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes (Continued) Klebsormidium flaccidium ++ + + ++ + ++ + ++ + ++ ++ + + + + + Physcomitrella patens + + + + + ++ ++ + + ++ + + ++ +++ + ++ +++ + + + ++ +++ Selaginella moellendorffii 4+ + ++ + + + ++ + + + + + + ++ ++ + 4+ ++ ++ + 4+ Pinus taeda + +++ + + + + 5+ + ++ + + + + Amborella trichopoda ++ + + ++ + ++ + ++ + ++ ++ + + + + + + + Oryza sativa + ++++ + ++ + + ++ + +++ ++ + + + + + + + Aquilegia coerulea +++ ++ + +++ ++ +++ ++ + + ++ ++ + + ++ +++ ++ + ++ +++ +++ ++ ++ Prunus persica ++ ++ + ++ ++ + ++ + + + ++ + ++ ++ +++ + + ++ + + + Arabidopsis thaliana ++ + + ++ + + + + + ++ + + + ++ ++ +++ ++ ++ + ++ + ++ Calliarthron tuberculosum ++++ + ++ +++ ++ + + Chondrus crispus ++ + + + + + + + + + + + + + + + Cyanidioschyzon merolae ++ + + ++ + ++ + ++ + + ++ + + + Galdieria sulphuraria + + + + ++ + + ++ ++ ++ + + + + + + Porphyridium purpureum + + + + + + ++ + + + + +++ + + + + + + Pyropia yezoensis +++ +++ + + Cyanophora paradoxa + + ++ + ++ ++ + ++ + ++ + + ++ + Phaeodactylum tricornutum ++ + + ++ + ++ + ++ ++ + 4+ Blastocystis hominis ++ + + +++ + ++ + ++ + + ++ + + + + + Phytophthora parasitica ++ + + ++ + ++ + +++ ++ ++ ++ ++ + + + + Schizochytrium aggregatum ++ + + ++ + ++ + ++ + ++ + Aureococcus anophagefferens + +++ + + + + + ++ + + Plasmodium falciparum ++ + + + + + Babesia equi + +++ + + + Theileria annulata + ++ Toxoplasma gondii + + + ++ + ++ + + ++ + + ++ Gregarina niphandrodes ++ + ++ + + Ichthyophthirius multifiliis ++ + ++ + + + + + + + + Tetrahymena thermophila + + + ++ + + + ++ + + ++ + + + + Oxytricha trifallax ++ ++ + ++ + + + ++ + ++ + + + + Paramecium tetraurelia + + ++ + + + + + + + + ++ ++ ++ ++ + ++ ++ Symbiodinium minutum + + + ++ ++ +++ + 6+ + + Perkinsus marinus ++ + ++ + ++ ++ ++ + ++ + ++ ++ + ++ ++ + +++ ++ Bigelowiella natans + + + ++ + + + ++ + + + + ++ + + + + Lombard Biology Direct (2016) 11:36 Page 7 of 31 Table 1 Presence of detectable homologues of the N-glycosylation proteins in a diversity of eukaryotes (Continued) Reticulomyxa filosa + + + + + + + + + + ++ ++ + + + + + + + Bodo saltans ++ + + ++ + ++ + ++ + 4+ + + Angomonas deanei +++ ++ + + ++ 5+ + Leishmania major + +++ + ++ + + 4+ + + Trypanosoma vivax + + + + + + + + + + +++ + Giardia lamblia ++ + + + Spironucleus salmonicida ++ + + Naegleria gruberi ++ + + ++ + ++ + ++ + + ++ + + + + + Trichomonas vaginalis + + + +++ + + + 5+ + + + + 5+ Emiliania huxleyi + + + ++ + ++ + + + + 4+ + 5+ Guillardia theta ++ + + ++ + ++ + ++ + +++ + + Failure in homologue detection may result from genome incomplete sequencing. (+) Presence of homologues in a genome (two or more (+) indicate as many paralogues). (F) on Alg14/Alg13 point out organisms in which the two genes are fused Lombard Biology Direct (2016) 11:36 Page 8 of 31 Table 2 Polyisoprenol-related pathways used in this analysis and respective literature N-glycosylation Eukaryotes (S. cerevisiae) KEGG: http://www.genome.jp/kegg/ GPI synthesis Eukaryotes (S. cerevisiae) KEGG: http://www.genome.jp/kegg/ O-mannosylation Eukaryotes (H. sapiens) KEGG: http://www.genome.jp/kegg/ [87] N-glycosylation Archaea (S. acidocaldarius)[45] N-glycosylation Archaea (M. maripaludis)[36, 40] N-glycosylation Archaea (M. voltae)[36, 71] N-glycosylation Archaea (A. fulgidus)[50] N-glycosylation Archaea (H. volcanii)[36, 40] N-glycosylation Archaea (H. volcanii, alternative) [113] N-glycosylation Bacteria (C. jejuni)[43] Peptidoglycan synthesis Bacteria (E. coli)[114] LPS synthesis (ABC transporter) Bacteria (E.coli)[115] LPS synthesis (wzy-dependent) Bacteria (S. enterica)[115] Wall teichoic acids Bacteria (B. subtilis)[116] O-glycosylation Bacteria (N. gonorrhoae)[117] O-glycosylation Bacteria (G. stearothermophilus)[118] Tyrosine O-glycosylation Bacteria (P. alvei)[101] Enterobacterial common antigen Bacteria (E. coli)[119] Capsule synthesis Bacteria (E. coli)[120] EPS - xanthan Bacteria (X. campestris)[121] EPS - succinoglycan Bacteria (R. meliloti)[122] EPS Bacteria (L. lactis)[73] pathways from the three domains of life, including the The eukaryotic N-glycosylation initiation complex has a eukaryotic N-glycosylation pathway. Some pathways from composite origin closely related organisms contain the same homology The canonical eukaryotic N-glycosylation pathway starts groups. This is for example the case of the O-glycosylation with the sequential link of two N-acetylglucosamines in G. stearothermophilus and Tyr. O-glycosylation in (GlcNAc) from soluble UDP-GlcNAc to Dol-P (Fig. 1, P. alvei (two Firmicutes);N-glycosylation in H. volcanii [61, 62]). The resulting GlcNAc dimer is named chito- and M. voltae (two Euryarchaeota); or O-antigen ABC- biose. The two GlcNAc transfers are consecutively carried dependent synthesis in E. coli and N-glycosylation in C. out by Alg7 and Alg13, which form a protein complex with jejuni (two Proteobacteria). It is more difficult to ascertain the membrane anchor Alg14 [51, 63, 64]. Alg7 belongs to clear parallels across pathways from distantly-related the polyprenol phosphate:N-acetylhexosamine-1-phosphate organisms. Nevertheless, some proteins of the eukaryotic transferase (HPT) family (dark green in Fig. 2, [65–67], N-glycosylation pathway belong to the same homology which also includes the bacterial MraY protein. Alg14 groups than proteins implicated in the peptidoglycan and Alg13 are respectively homologous to the N- and synthesis, LPS synthesis or putative N-glycosylation in C-terminal parts of bacterial MurG (purple in Fig. 2, [62]). A. fulgidus. The most exciting parallel is to be drawn MraY, MurG and other proteins form a complex involved between the eukaryotic and putative S. acidocaldarius in peptidoglycan synthesis (Fig. 2, [51, 68]). The resem- N-glycosylation pathways. This Sulfolobus pathway blances between the eukaryotic Alg7-Alg13-Alg14 and the has not been fully described yet [45], but it is the bacterial MraY-MurG complexes [69] could support the only prokaryotic mechanism studied here with homo- existence of a close evolutionary relationship between the logues of all eukaryotic cytoplasmic GTs and a cata- eukaryotic N-glycosylation and the bacterial peptidoglycan lytic N-OST subunit. This result is suggestive because synthesis pathways [49, 51]. Although most archaeal many eukaryogenesis scenarios assume a tight evolu- N-glycosylation initiation mechanisms are based on GTs tionary relationship between proteoarchaea and eu- from group 2 (red GTs in Fig. 2, [36]), HPT homologues karyotes [8–10, 14] but it must be balanced with the have been reported in several archaea [45, 70]. For phylogenetic analyses presented below. instance, the S. acidocaldarius genome contains both a Lombard Biology Direct (2016) 11:36 Page 9 of 31 Fig. 2 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 10 of 31 (See figure on previous page.) Fig. 2 Polyprenol-based glycosylation pathways (and GPI biosynthesis) colored according to detected homology groups. Horizontal lines represent ER membranes in the eukaryotes, plasma membranes in prokaryotes. Horizontal rectangles represent cytoplasmic glycosyltransferases (GTs) if they are below the membrane or lumen/periplasmic GTs if they are above the membrane. Vertical rectangles depict flippases or translocation mechanisms. Ovals represent the oligosaccharide transferases from the lipid carrier to the acceptor molecule. Diamonds portray proteins that are neither GTs nor translocases (e.g. acetyl or ethanolamine transferases in GPI biosynthesis). Extra shapes in the eukaryotic oligosaccharide transferases reflect the fact that these are complexes with many subunits. The cytoplasmic GTs depicted after a transfer to the acceptor molecule (e.g. Dpm1 and Alg5 in eukaryotic N-glycosylation) represent polyisoprenol-P-monosaccharide synthases tranfering single mannoses or glucoses to a lipid carrier to supply these sugars to the opposite side of the membrane. Proteins are colored according to the homology group to which they belong (as defined in Methods). Plain symbols represent proteins that were detected using the procedure described in Methods, whilst empty white shapes show more distant relationships that required bibliographic or extra analyses to be established. Empty transparent shapes represent the lack of detection of any homologues in the dataset HPT and a murG homologue, and the N-glycans in this family, one of which was lost in the euryarchaeal lineage. organism are also initiated with a chitobiose [45]. Despite Alternatively, the rampant paraphyly of this tree may be these promising similarities between the eukaryotic and indicative of some reconstruction artifacts due to the the putative S. acidocaldarius N-glycosylation pathways lack of phylogenetic signal or inappropriate phylogenetic (Fig. 2, [45]), the actual role of the archaeal HPT proteins reconstruction models. In that case, the crenarchaeal is debated [70, 71] and the MurG homologue remains gene could result from a phylum-specific duplication, uncharacterized to this date. and the basal branching of the proteoarchaeal group The phylogeny of each protein involved in the initiation could result from a long branch attraction artifact. complex was carried out. That of the HPT superfamily Regardless of the origin of the two proteoarchaeal para- shows two large groups (Bayesian Posterior Probability, logues, at least one representative of the HPT family BPP = 1), one mostly bacterial and the other including seems to have been present in both the last common most archaeal and eukaryotic sequences (Additional ancestor of euryarchaea and proteoarchaea and, thus, file 2). The phylogeny of the bacterial clade is discussed possibly in LACA. The eukaryotic Alg7 orthologues un- in detail the Additional file 2 but, in summary, it supports ambiguously branch within the paraphyletic crenarch- the idea that at least two proteins from the HPT aeal group–a group that also includes S. acidocaldarius superfamily were present in the Last Bacterial Common putative AglH (HPT) protein. Yet, there are no obvious Ancestor (LBCA). Here, we will center our attention on closest crenarchaeal relatives to the eukaryotic clade. the other cluster, for it contains the eukaryotic sequences. Thus, although the eukaryotic Alg7 genes likely have a A specific phylogenetic analysis only focusing on this proteoarchaeal origin, the precise identity of this ances- part of the tree was carried out (Fig. 3). It shows a tor remains uncertain. monophyletic group of eukaryotes (BPP = 1) in which se- Regarding the MurG/Alg14/Alg13 homologues, several quences cluster according to meaningful taxonomic occurrences of split or fused genes exist. For instance, the groups, such as paraphyletic groups of unikonta, viridi- heteromer EpsE/EpsF involved in the exopolysaccharide plantae or alveolates. A eukaryotic-only phylogeny re- (EPS) synthesis of firmicutes [72, 73] has been reported to vealed a better resolution and an overall topology in be split into the same protein domains as Alg14/Alg13 accordance with the main eukaryotic taxa (e.g. mono- [62]; in the contrary, excavates and some amoebozoa have phyly of unikonts, archaeplastida, excavates, etc. Data fused Alg14-Alg13 polypeptides (Table 1). In order to ac- not shown). This suggests that the lack of resolution count for these different architectures, independent phy- among eukaryotes in the Fig. 3 may be due to recon- logenies were constructed for each protein domain: struction artifacts owing to the extreme divergence be- catalytic (Alg13/C-terminal part of MurG) on the one side tween the prokaryotic and the eukaryotic sequences. and membrane anchor (Alg14/N-terminal part of MurG) The wide distribution of these genes in eukaryotes and on the other. In both phylogenies, MurG homologues the phylogenetic clustering of the eukaryotic sequences were widespread among bacteria and formed a highly according to their taxonomic groups support the pres- divergent clade (BPP = 0.99 in both cases, Fig. 4a, ence of Alg7 in LECA. Moreover, the eukaryotic clade Additional files 3 and 4). The bacterial MurG phylogeny is branches within a paraphyletic group of crenarchaea discussed in more detail in the Additional files 3 and 4 (BPP = 1) which itself branches in a paraphyletic group but, in short, MurG was likely already present in the of euryarchaea (BPP = 0.86). A proteoarchaeal clade LBCA. forms a sister group to the rest of archaeal and Phylogenies excluding the bacterial MurG homologues eukaryotic sequences (Fig. 3). The existence of the basal were carried out to more precisely tackle the evolution proteoarchaeal gene may suggest that the Last Archaeal of the closest prokaryotic relatives to the eukaryotes Common Ancestor (LACA) had two paralogues of this (Fig. 4b and Additional file 5). This includes archaeal Lombard Biology Direct (2016) 11:36 Page 11 of 31 Fig. 3 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 12 of 31 (See figure on previous page.) Fig. 3 Bayesian phylogeny of the HPT homologues. a Schematic phylogenetic tree including the bacterial sequences (see Additional file 2 for details). b Specific phylogeny of the archaeal/eukaryotic clade. The tree was reconstructed using 124 representative sequences and 203 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) fused sequences (referred to as MurG-like) and bacterial topology of the tree of life, the cenancestor most likely and archaeal split (EPS-like) sequences. The independent had a fused MurG-like protein. This is not the case of phylogenies of both the catalytic and membrane anchor the split proteins, of which the distribution among pro- domains show similar results to each other: three clades karyotes is too scarce to support their presence in the corresponding to 1) the archaeal fused MurG-like cenancestor. Since, contrary to bacteria, the eukaryotes homologues (BPP Alg13 = 0.94, BPP Alg14 = 0.67); 2) do not have other homologues of the Alg13/Alg14 pro- the bacterial and archaeal split EPS-like homologues teins, the most parsimonious hypothesis for the origin of (BPP Alg13 = 0.92); and 3) the eukaryotic proteins the split genes would be that epsE/epsF developed in the (BPP Alg13 = 0.7, BPP Alg14 = 0.87). The fused MurG-like bacterial lineage from the duplication and subsequent archaeal clade is divided into a crenarchaeal group and a split of their ancestral murG gene and were later ac- euryarchaeal group with some basal thaumarchaeota and quired in an early eukaryotic ancestor from an unidenti- bathyarchaeota. Despite the paraphyly of the proteoarch- fied bacterium. Once in the eukaryotic lineage, they aeal sequences (i.e. the crenarchaeal and thaumarchaeal evolved into the alg13/alg14 couple. sequences in this tree), the wide distribution of these pro- In summary, no evidence was found to support a teins in archaea and the monophyly of the crenarchaeal direct relationship between the bacterial peptidoglycan and euryarchaeal sequences supports the presence of this synthesis and the eukaryotic N-glycosylation pathway. gene in LACA. The internal topology of the eukaryotic The eukaryotic HPT was present in LECA and likely clade is very mixed and very poorly resolved even when inherited from proteoarchaea. The Alg13/Alg14 couple phylogenies are reconstructed only with eukaryotic se- is structurally more similar to their bacterial homologues quences (data not shown). This could translate a compli- and, thus, could have a bacterial origin. These results high- cated non-vertical evolutionary history for these genes but light the mixed origin of the eukaryotic N-glycosylation it most likely results from the small number of alignment initiation complex. Furthermore, since at least one repre- positions that was conserved in each phylogeny (116 and sentative of both the HPT and MurG families was present 97 respectively) and the fast evolution rate of these pro- in the respective common ancestors of both prokaryotic teins, especially the membrane anchor. Despite these diffi- domains, their presence may also be inferred in the cenan- culties, the wide distribution of Alg13/Alg14 orthologues cestor. This suggests that the cenancestral membranes across eukaryotes and their monophyly in all the phyloge- may have had the basic elements of a glycosylation ma- nies suggest that these proteins may have been ancestral chinery. Given the fact that most modern prokaryotic to eukaryotes. Finally, the split EPS-like prokaryotic se- polyisoprenol-based glycosylation pathways are involved quences are dominated by bacteria. The archaeal split in the synthesis of prokaryotic cell wall components, this EPS-like sequences are few and scattered among the bac- may suggest that the cenancestor was already bound by terial sequences, so they certainly result from HGTs. The some kind of glycosylated envelope, of which the specific EPS-like bacterial sequences form meaningful taxonomic nature is unknown. The implications of these results for groups, but their distribution across bacteria is not very the possible cenancestral cell walls will be discussed in a diverse. Thus, unlike MurG, the ancestral presence of paper to be published soon (Lombard, data not shown). EPS-like genes in the LBCA is uncertain. The relationship of the eukaryotic clades to their pro- Cytoplasmic mannosylation evolved from an archaeal karyotic relatives is more difficult to establish. The protein eukaryotic sequences form separate clades with regard After the chitobiose core, five mannoses are tethered to to the other domains of life, so these phylogenies could the growing glycan in the cytoplasmic face of the ER be discussed in terms of a three-domain of life topology. membrane (Fig. 1b). The first mannosylation is generally This line of thought would suggest that the eukaryotic ascribed to Alg1, the second and third to Alg2 and the Alg13/Alg14 proteins were inherited from the last com- two last to Alg11, but their activity depends on the forma- mon ancestor of extant organisms (i.e. the cenancestor tion of respective protein complexes: Alg1 homooligo- or LUCA) through a proto-eukaryotic lineage. Neverthe- mers, respective Alg2-Alg1 and Alg11-Alg1 heteromers less, if LBCA and LACA had fused MurG proteins, as [54]. All three proteins belong to the glycosyltransferase 1 this analysis supports, it follows that whatever the superfamily (GT1), even though Alg1 is classified as a Lombard Biology Direct (2016) 11:36 Page 13 of 31 Fig. 4 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 14 of 31 (See figure on previous page.) Fig. 4 Bayesian phylogeny of the Alg13/catalytic MurG domain. a Schematic phylogenetic tree including the bacterial MurG homologues (see Additional file 3 for details). b Specific phylogeny excluding the bacterial MurG clade. The tree is unrooted and was reconstructed using 132 representative sequences and 116 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) separate family in the CaZY database [54]. These both genes were inherited from LECA. The Alg2 se- eukaryotic genes were suggested to have archaeal ances- quences form a sister group to a clade made up of the tors [50], but early phylogenetic analyses were unable to eukaryotic Alg11 and most proteoarchaeal sequences. support this hypothesis. Thus, these proteins had also Several scenarios may explain this topology. One possi- been postulated to be eukaryotic-specific [46]. bility is that an ancestor of the proteoarchaea horizon- The GT1 superfamily is widely represented in most tally acquired one of these genes from eukaryotes early polyisoprenol-based pathways in the three domains of in proteoarchaeal evolution. Yet, since these homologues life (dark blue proteins in Fig. 2). The high number of are present in a wide diversity of archaea and form re- homologues of the GT1 enzymes detected in the three spective monophyletic groups of euryarchaea and pro- domains of life proved challenging to handle, so pre- teoarchaea, it seems likely that at least one of these liminary analyses were carried out to exclude the most proteins was present in LACA. In that case, interpreta- distant GT1 homologues from the analysis and concen- tions depend on the topology of the tree of life which is trate in the closest relatives to the eukaryotic sequences favored. In the context of a traditional tree of life, in (see Methods). These preparatory analyses showed that which archaea and eukaryotes are sister groups, the top- the Alg1 sequences formed a monophyletic group but ology could be explained by a hidden paralogy: the last their branching position within the superfamily tree was common ancestor of both domains had two paralogues, highly inconsistent and depended on the sequences the eukaryotic linage kept both of them and the pro- selected for the analysis (data not shown). As a result, teoarchaea and euryarchaea alternatively lost one of the closest prokaryotic relatives of the eukaryotic alg1 them. Nevertheless, the requirement of Alg1 to form could not be determined. The origin of this gene could functional heteromers with Alg2/Alg11 contradicts the not be assessed, and this may suggest that Alg1 was a ancestral paralogy of these proteins in archaea, as Alg1 eukaryotic innovation. A tentative explanation of the is a eukaryotic innovation. Alternatively, the ancestor of high Alg1 divergence–even when compared to the out- the eukaryotic (and bacterial) sequences that branch standing diversity of GT1 members–could be the role of within the archaeal clade could have been acquired from Alg1 in the formation of the mannosyltransferase com- archaea. Since Alg11 is more closely related to the pro- plexes [54]. teoarchaeal sequences, it seems more likely that the ori- Contrary to Alg1, Alg2 and Alg11 are closely related ginal eukaryotic GT1 gene had proteoarchaeal origins, a to each other and to some prokaryotic sequences. Their hypothesis that would also be congruent with the cur- evolution was studied together with their closest rently favored symbiogenic origin of eukaryotes [21–24]. prokaryotic relatives (see Methods). The outgroup of this In that case, the original eukaryotic GT1 gene would analysis is made up of some sequences from plastid- have been duplicated and acquired their alg2/alg11 func- bearing eukaryotes and cyanobacteria (Fig. 5). The first tions in eukaryotes. The fact that both eukaryotic clades sequences to diverge from the outgroup are some do not form relative sister groups within the proteoarch- Methanobacteriales and a group dominated by crenarch- aea could be a reconstruction artifact due to the muta- aea. The position of these sequences frequently changed tional saturation that resulted from the development of in the preliminary trees (data not shown) and their basal heteromers and new functions in eukaryotes. position in the Alg2/Alg11 tree calls into question their The characterization of the closest archaeal relatives to closeness to the proteins that interest us. Nevertheless, the eukaryotic Alg2/Alg11 proteins will be instrumental to the lack of characterization of these genes makes it diffi- the evaluation of these different evolutionary hypotheses. cult to make an informed decision to exclude them from ORF43 has been suggested to be involved in N-glycosylation the analysis. The rest of the tree is split in four clades: a in A. fulgidus but, in fact, the closest archaeal relatives to euryarchaeal clade also including several bacterial groups Alg2/Alg11 are uncharacterized GT1 proteins [50]. Agl16, (BPP = 0.98), a proteoarchaeal clade (BPP = 1) and two a GT1 protein implicated in N-glycosylation in S. eukaryotic clades respectively corresponding to Alg2 acidocaldarius [74] is not present among the closest (BPP = 0.95) and Alg11 (BPP = 1). The deep relationships relatives of Alg2/Alg11. Similarly, all the GT1 homologues within the eukaryotic clades are unresolved, but their from the Lokiarchaeum metagenome that has been monophyly and widespread distribution suggests that suggested as the closest archaeal relative to eukaryotes Lombard Biology Direct (2016) 11:36 Page 15 of 31 Fig. 5 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 16 of 31 (See figure on previous page.) Fig. 5 Bayesian phylogeny of the closest relatives to the eukaryotic Alg2/Alg11 homologues. The tree is unrooted and was reconstructed using 211 representative sequences and 194 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange) and eukaryotes (purple) [11] belong to other parts of the general (preliminary) Gene duplication and neofunctionalization events would GT1 superfamily tree. Until these archaeal proteins are explain the emergence of the plethora of specialized GTs characterized, the lack of functionally important Alg1 in known today [59]. archaea and currently favored symbiogenic scenarios for In the current analysis, the study of the lumen GTs the origin of life rather favor the proteoarchaeal origin of was extended to a much larger diversity of eukaryotes the eukaryotic alg2/alg11 GT1 ancestor. and found to be widespread in most eukaryotic lineages In summary, Alg2 and Alg11 could have evolved from (Table 1, Additional file 1). The phylogenies of each a proteoarchaeal GT1 ancestor and Alg1 has unknown superfamily support the monophyly of each lumen GT origins. Modern eukaryotic complexes based on Alg1, and, thus, its likely inheritance from LECA (data not Alg2 and Alg11 are eukaryotic innovations developed shown). This analysis also suggests a previously unre- prior to LECA. ported possible relationship between Alg10 and Gpi18 (data not shown). Nevertheless, the sequences from dif- The unclear origin of an uncertain flippase ferent superfamilies are so divergent that alignments The transport of polyisoprenol-linked glycans across the comprising several superfamilies are highly unreliable membrane is generally thought to be mediated by pro- (Guidance alignment score < 0.1), and so are the result- teins called flippases [75]. Rft1 was suggested to be the ing phylogenies. Here, particular attention was given to eukaryotic N-glycosylation flippase [76] but the actual the elucidation of the possible early origins of the activity of this protein has been called into question eukaryotic lumen GTs. As reported previously [46], basic [77–79]. Rft1 has homologues in the three domains of BLASTp searches did not detect prokaryotic homo- life (Fig. 2). The phylogeny of the Rft1 homologues logues, so HMM profiles were constructed for each shows that the eukaryotic sequences form a monophy- superfamily in order to look for them (see Methods letic group (BPP = 0.99) supportive of their inheritance [81]). The closest prokaryotic hits of each superfamily from LECA, but it is unable to clarify the relationship of were used as seeds for psi-BLAST searches [82] against the eukaryotic proteins with their prokaryotic relatives the bacterial and archaeal genomes. The vast majority of (Additional file 6). Thus, the origin of the eukaryotic the detected prokaryotic sequences had PMT domains Rft1 is unclear (see Additional file 6 for a more detailed (for Protein O-Mannosylation Transferases) or the un- discussion). known protein domain DUF2029 defined from mycobac- terial proteins (data not shown). The PMTs carry out The lumen GTs are related to the protein O-mannosyl protein O-mannosylation, which is the transfer of a man- transferases (PMTs) nose residue from Dol-P-Man to a protein in the lumen Once translocated to the lumen side of the ER mem- side of the ER membrane [83, 84]. O-mannosylation and brane, the Dol-P-linked glycan is mannosylated four N-glycosylation are highly intertwined, as inhibition of times and glucosylated three times to make up the ca- one enhances the other and both PMT and N-OST belong nonical opisthokont core glycan (Fig. 1b). These modifi- to the glycosyltransferase family C (GT-C) and may be cations are catalyzed by the mannosyltransferases Alg3, distantly related [85]. O-mannosylation has been described Alg9 and Alg12 and the glucosyltransferases Alg6, Alg8 in several bacteria including Mycobacteria [86, 87]–a and Alg10. These proteins appear to be unique to suggestive result given the detection of the mycobacterial- eukaryotes [46, 50], so their evolution has only been based unknown domain DUF2029 in the distant lumen studied in this domain [59]. The closest relatives of these GT prokaryotic homologues. O-mannosylation has not GTs are the eukaryotic mannosyltransferases involved in been described in archaea yet, but it is exciting to point the GPI biosynthesis (Fig. 2, [59]), which also takes place out that the protein Agl1, which was tentatively suggested in the lumen side of the ER membrane [80]. All the to be a GT in S. acidocaldarius [88], bears a PMT domain. lumen GTs have similar multispan transmembrane do- Incidentally, AglS is an archaeal protein which shares mains that have been compared to the topology of sugar with the mannosyltransferases Alg3, Alg9 and Alg12 a transporters [59]. Lumen GTs were classified into three Dol-P-mannose mannosyltransferase function [42]. AglS superfamilies according to the sugar they used, the sugar transfers a mannose to a S-layer protein N-glycan in H. linkage they catalyze and their common peptide motifs. volcanii and has also been suggested to be distantly related Lombard Biology Direct (2016) 11:36 Page 17 of 31 to the PMTs, but there is no obvious evolutionary rela- gene was vertically inherited in archaea and eukaryotes tionship between AglS and the eukaryotic lumen GTs. Fi- from their last common ancestor (in a three-domain tree nally, the phylogeny of the PMT homologues was carried of life perspective), or that the eukaryotes branch within out (Additional file 7). The eukaryotic PMTs (including a paraphyletic proteoarchaeal group (BPP = 0.88, as several paralogues) form a monophyletic clade (BPP = 1) shown in Fig. 6). This latter possibility is in agreement but their relationship to their prokaryotic relatives is un- with the latest and most accurate phylogenies and phylo- clear (see Additional file 7 for details). genomic studies, which support the closest relationship In summary, very distant sequence similarities were of the eukaryotic stem with the proteoarchaea [21–24]. detected between the ER lumen GTs and the PMTs. Despite this result, no specific proteoarchaeal lineage, PMTs use similar Dol-P-monosaccharide donors as the including Candidatus Lokiarchaeum [11], is specifically N-glycosylation lumen GTs (the same in the case of the related to the eukaryotic Stt3, so the identity of the pro- mannosyltransferases), carry out similar transfer reactions tearchaeal donor is unknown. This indicates a lack of and participate in functionally related mechanisms. As a resolution but it could also be explained by poor sam- result, the evolutionary relationship of these GTs with the pling among archaea or by the fact that the donor of this PMT enzymes seems significant. It suggests that the clos- sequence predated LECA and, therefore, it cannot ne- est ancestor of the lumen N-glycosyltransferases was not a cessarily be assigned to one of the modern proteoarch- sugar transporter [59] but a mannosyltransferase from the aeal lineages. GT-C superfamily. Little is known about the function or evolution of the non-catalytic N-OST subunits [47]. Ost1, Ost2, Ost3/6, Eukaryotic Stt3 was probably acquired from Swp1 and Wbp1 orthologues are widespread among proteoarchaea eukaryotes (Table 1, Additional file 1), but homologues of The N-OST that transfer the oligosaccharide from the the other subunits are more difficult to detect certainly lipid carrier to the acceptor protein have been genetically because of their small size (e.g. yeast Ost4 is 36 amino and biochemically described in eukaryotes [53, 89, 90], acids long). BLASTp did not find any sequence similar to bacteria [44, 47] and archaea [91–95]. The evolution of these subunits neither in bacteria nor in archaea, so dis- the catalytic Stt3/AglB subunit of the N-OST is by far the tant prokaryotic homologues were looked for using HMM best studied of the N-glycosylation enzymes. For instance, profiles. The few detected sequences participate in unre- many HGTs and duplications of aglB have been reported lated functions, so no promising candidates were found in archaea and a close relationship between these proteins (data not shown). Maybe when more will be known about and eukaryotic Stt3 has been suggested [46, 47, 96, 97]. the functions of the non-catalytic N-OST subunits, it will Yet, previous evolutionary studies were limited to one be possible to make more informed choices to study the domain of life, were unable to determine the specific origin of these eukaryotic genes. relationship of eukaryotes to archaea, or simply dis- regarded this issue. A dolichol-P-monosaccharide synthase could have an ar- A phylogeny of Stt3/AglB homologues from all three chaeal origin domains of life was carried out (Fig. 6). The eukaryotic Although strictly speaking they are not components sequences form a monophyletic clade (BPP = 0.92) which devoted to the N-glycosylation pathway, the two Dol-P- contains clear taxonomic groups but whose phylogeny is monosaccharide synthases supply mannose and glucose blurred by the presence of several paralogues. The rest to the lumen GTs (Fig. 1b). These enzymes belong to the of the tree is made up of archaeal sequences roughly glycosyltransferase group 2 (GT2, [75]), which also com- clustering according to the archaeal taxonomy and a few prises most GTs involved in the archaeal N-glycosylation bacterial homologues located within the euryarchaeal pathways (Fig. 2). Therefore, the origins of the eukaryotic clade. The widespread distribution of AglB homologues GT2 proteins were examined. In eukaryotes, Alg5 is a in archaea and the clustering of the sequences according Dol-P-glucose synthase [98] and Dpm is a Dol-P-mannose to the main archaeal taxonomic groups supports the an- synthase also providing mannose to GPI synthesis and O- cestral presence of this protein in LACA. Moreover, the mannosylation [99]. Two types of Dpm exist [75, 85, 100]: location of the bacterial sequences within the euryarch- yeast Dpm has a C-terminal transmembrane helix to aeal group supports the dominant opinion that this gene interact with membranes, whereas human Dpm1 lacks was transferred from archaea to bacteria [46–48] at least this structure and requires supplementary subunits twice, followed by a few subsequent transfers among (Dpm2/Dpm3). These differences were translated into two bacteria. The relationship between the archaeal and separate clades in previous Dpm phylogenies [46]. eukaryotic orthologues is more difficult to establish be- Similar to the GT1 superfamily, the GT2 superfamily cause the phylogeny is unrooted. Depending on where comprises many proteins (red proteins in Fig. 2) that we decide to root this phylogeny, it may suggest that the make its evolutionary study challenging. Only the closest Lombard Biology Direct (2016) 11:36 Page 18 of 31 Fig. 6 (See legend on next page.) Lombard Biology Direct (2016) 11:36 Page 19 of 31 (See figure on previous page.) Fig. 6 Bayesian phylogeny of the catalytic N-OST subunit (Stt3/AglB/PglB) homologues. The tree is unrooted and was reconstructed using 163 representative sequences and 268 conserved sites. Multifurcations correspond to branches with Bayesian posterior probabilities <0.5. Numbers at nodes indicate Bayesian posterior probabilities higher than 0.5. Bootstrap values from maximum likelihood analyses are reported on basal and major nodes. Colors on leaves represent the affiliation of sequences to a domain of life: archaea (blue), bacteria (orange)and eukaryotes (purple) prokaryotic relatives to the eukaryotic Alg5/Dpm se- the bacterial polyisoprenol-based pathways involved in cell quences were kept for the analysis (see Methods). The wall synthesis [34, 48–50]. Yet, aside from a primordial resulting phylogeny is ambiguous (Additional file 8). The paper that had neither the data nor the intention to tackle prokaryotic sequences are mixed, but the (unrooted) this issue [46], this is the first attempt to provide a phylogeny seems to be split into two clades (BPP = 0.96): thorough examination of the relationship of the eukaryotic one dominated by bacteria and eukaryotic Alg5 sequences N-glycosylation pathway with its prokaryotic relatives. and another dominated by archaea and eukaryotic Dpm The proteins involved in polyisoprenol-based pathways sequences. The bacterial-dominated clade contains WsfH from the three domains of life, including the eukaryotic from P. alvei, which may supply periplasmic Bac-P-Glc to N-glycosylation pathway, belong to a small number of tyrosine O-glycosylation [101]. This is remarkable because gene families (Fig. 2). Only the putative N-glycosylation Alg5 are also Dol-P-Glc synthases, but it is balanced by pathway from S. acidocaldarius has an obvious parallel- the fact that AglD, a Dol-P-Man synthase from H.volcanii, ism with the eukaryotic N-glycosylation, as they are both is phylogenetically closer to the Alg5 clade. A few cre- made up from representatives of the same protein super- narchaeal sequences and a sequence from Candidatus families (Fig. 2). More detailed phylogenomic analyses, Lokiarchaeum are basal to the eukaryotic Alg5 clade, but however, show that some eukaryotic N-glycosylation the significance of these sequences is uncertain because genes may indeed have proteoarchaeal ancestors, but they are mixed to other euryarchaeal and bacterial only two of those belong to the S. acidocaldarius path- sequences. Thus, the origin of the eukaryotic alg5 way. This stresses the fact that simple homology is not gene is unclear. enough to draw robust evolutionary hypotheses about The archaeal-dominated clade includes H. volcanii the relationship among polyisoprenol-based pathways. AglJ, M. voltae AglK, A. fulgidus Orf39 and the euka- Not only the phylogenomic analyses are required to ryotic Dpm. As previously reported [46], the Dpm confirm evolutionary hypotheses, they also provide sequences are split into two groups depending on the promising prokaryotic targets for the dynamic field of presence of the C-terminal transmembrane helix. The prokaryotic glycosylation characterization [36, 40]. eukaryotic Dpm clades are intertwined with several pro- Table 1 shows that most proteins involved in the karyotic sequences, most notably the main proteoarch- eukaryotic N-glycosylation pathway are widespread in all aeal clade of the tree. Thus, it is tempting to suggest major eukaryotic taxa. The phylogenetic results of each that the eukaryotic sequences, as well as the bacterial superfamily implicated in this process have been sum- sequences in this cluster, have proteoarchaeal origins. marized in Fig. 7. Since many of these phylogenies are Since the Dpm1 type (devoid of a C-terminal transmem- unrooted, their interpretation sometimes depends on the brane helix) is more widespread in eukaryotes, it is tree of life topology that we favor, but they clearly plausible that it was the ancestral type in eukaryotes and support a more complicated scenario for the origin of the fused type emerged later. The characterization of the eukaryotic N-glycosylation pathway than previously some proteoarchaeal homologues should allow testing thought. All the canonical eukaryotic N-glycosylation genes this possibility. were present in LECA and inherited (or lost/modified) in In summary, the origins of the eukaryotic Alg5 are modern lineages, but the origins of the eukaryotic N- unclear, but Dpm may have a proteoarchaeal ancestor. glycosylation genes are diverse. Alg7, Dpm1 and also pos- sibly one GT1 and one Stt3 (N-OST catalytic subunit) have Conclusion proteoarchaeal ancestors. A duplication of the original N-glycosylation is a major protein modification in eukary- GT1 gene (blue family in Fig. 2) in the eukaryotic lineage otes that modulates the properties of many proteins. before LECA would account for genes alg2 and alg11.The Unraveling its origins is not only relevant to the particular stt3 gene has been duplicated in some eukaryotic lineages case of this essential pathway, but its location in the ER too. The Alg13/Alg14 couple is structurally most similar membranes may also have implications for the origins of to bacterial homologues, so it may have bacterial origins. the ER and the eukaryotes themselves. Previous hypoth- The lumen GTs could have resulted from the co-option of eses suggested a particular evolutionary link between this a PMT protein involved in O-mannosylation and subse- eukaryotic pathway and archaeal N-glycosylation or one of quent gene duplications. Finally, the other GT1 (Alg1), Lombard Biology Direct (2016) 11:36 Page 20 of 31 Fig. 7 Origins of the eukaryotic N-glycosylation proteins and presence of related superfamilies in the last common ancestors of each domain of life and the cenancestor (summary) GT2 (Alg5) and the putative flippase have unknown making of the eukaryotic N-glycosylation pathway. This origins and could be eukaryotic innovations. The origins is in agreement with currently favored hypotheses which inferred for the eukaryotic N-glycosylation genes will be assume a proteoarchaeal ancestor to the eukaryotic stem open to reevaluation when more information will be [21–24]. For instance, these results are congruent with available about their prokaryotic relatives that have been the “inside-out” hypothesis, which suggested that the pointed out in this work. presence of the eukaryotic N-glycosylation pathway in Overall, this is another example that in metabolism, as the ER membranes was a remnant of an eocyte S-layer in other functions, eukaryogenesis was achieved through N-glycosylation pathway ([14]). Yet, these results are also an intricate combination of pre-existing and new ele- compatible with other eukaryogenesis hypotheses, as ments. Although in a few cases (e.g. Alg13/Alg14, Alg2/ other eocyte or archaeal-stem hypotheses of eukaryogen- Alg11, Stt3) a proto-eukaryotic origin cannot be com- esis may also account for the proteoarchaeal origin of these pletely ruled out, the results presented here suggest a proteins [8, 10, 11] and the possibility that these genes strong influence of some proteoarchaeal genes in the were horizontally transferred from a proteoarchaeon to a Lombard Biology Direct (2016) 11:36 Page 21 of 31 non-proteoarchaeal eukaryotic stem cannot be excluded [3, guidance statistical values on the alignment were used to 13, 15–18]. Whatever the eukaryogenesis hypothesis, this make an informed choice to trim the sequences for the work shows that it is now time for works on eukaryotic or- final phylogenetic reconstructions, and the trimmed align- igins to study the early evolution of the eukaryotic mem- ments are reported in the Additional file 11. The final branes and membrane functions. phylogenies were constructed using the programs MrBayes 3.2.6 [108] and RaxML-HPC2 8.1.24 [109] im- Methods plemented in the CIPRES Science Gateway [110]. The The original sequence seeds used to build the homology MrBayes analysis used a LG substitution model [111] and groups in Fig. 2 were selected from the literature (Table 2) 4 categories for the Γ distribution of substitution rates. and the KEGG database (http://www.genome.jp/kegg/). Four chains of 1,000,000 generations were run, trees were All vs. all reciprocal BLASTp searches were carried out sampled every 100 generations and the first 25 % [60]. The homology groups were manually built from the generations were discarded as “burn in”.The RaxML inspection of the reciprocal BLASTp: hits with an e-value reconstructions used the LG + Γ model [112] and 4 rate inferior to 1e-05 were automatically considered to belong categories. A hundred bootstrap replicates were carried to the same homology group; proteins with two or more out with the same model to evaluate node support. The hits from the same homology group and e-values ranging Bayesian and maximum likelihood phylogenies gave between 0.01 and 1e-05 were also included in the hom- comparable results. The bootstrap support from ology group. Information from the literature was appealed RaxML of the main basal nodes was reported on the to confirm or suggest particular homologies that were not published Bayesian trees. unambiguously determined with simple reciprocal BLAST, In some cases, the basic BLASTp search did not e.g. in the case of the lumen GTs [59]. These cases of provide satisfactory results. When too many prokaryotic ambiguous homology appear as uncertain (empty white homologues were detected (e.g. GT1 and GT2 super- shapes in Fig. 2). families), a smaller prokaryotic genome list was used for The same sequence seeds were used for the phylogen- the preliminary analyses (Additional file 9). These earlier etic analyses. Local genomic databases were constructed results were curated and only the closest homologues to for each domain of life from a selection of representative the eukaryotic proteins of interest were kept. Supple- genomes (Additional file 9). BLASTp with default parame- mentary sequences from the regular prokaryotic local ters were carried out against these local databases. These genomic database were added later on to improve the results were complemented with supplementary BLASTs prokaryotic diversity of the analyses. At that step, how- against the same genomes carried out with alternative ever, only the closest prokaryotic relatives were kept, so seeds and manual searches in the orthologue databases recent paralogues may have been missed. On the con- EggNOG (http://eggnogdb.embl.de/#/app/home, [102]) trary, when very few similar sequences were detected in and OMA browser (http://omabrowser.org/oma/home/, a domain of life, the searches were extended to the non- [103]) to make sure that no orthologues were missed in redundant (nr) sequences annotated as belonging to that the Table 1. Sequences too long, too short or for which domain of life on GenBank (http://www.ncbi.nlm.nih. the protein domain under study was not detectable using gov/Genbank). If similar sequences could not be detected the Pfam database 28.0 (http://pfam.xfam.org/) were re- using these methods (e.g. lumen GTs in prokaryotes), moved from the rest of the analyses. Some exceptions HMMprofileswerebuilt from theMUSCLE alignments of were applied when necessary as, for example, the eur- the eukaryotic sequences and were used to look for distant yarchaeal MurG-like homologues, which were kept for the homologues using the hmmsearch tool implemented in the N-terminal analysis despite their sequences being too di- HMMER v3.1b2 webserver (http://hmmer.org/, [81]). Ho- vergent for Pfam to detect their respective protein domain mologues found in this way were validated as suitable in them. Preliminary alignments were carried out using candidates based on protein domain information from MUSCLE v3.8.31 [104], sequences were trimmed using Pfam or functional annotation from the literature. the program NET of the MUST package [105] and prelim- The validated prokaryotic homologues were then used inary phylogenies were reconstructed using FastTree to carry out BLASTp or psi-BLAST searches [82], v.2.1.7 [106]. These preparatory analyses were examined either against the regular list of genomes or the nr for the overall topology of the trees, and used to select sequences from GenBank, depending on the results. representative sequences for smaller datasets that could be used with more accurate phylogenetic reconstruction Reviewers’ comments methods. Once the representative sequences had been se- The author thanks the reviewers for their comments. lected, a new MUSCLE alignment was carried out on the When possible, the manuscript has been modified ac- Guidance server (http://guidance.tau.ac.il/ver2/, [107]). cording to their recommendations. The alignments are available in the Additional file 10. The Lombard Biology Direct (2016) 11:36 Page 22 of 31 First round of review Author’s response: This article originally was longer Referee 1, Pr. Patrick Forterre, University Paris Sud, France and covered not only the origin of the eukaryotic N- Reviewer summary It’s an interesting and useful paper glycosylation pathway but also the putative presence of a that could be improved if fast evolving species were re- polyprenol-based glycosylation machinery in the cenan- moved from the phylogenetic analyses. cestor. In a previous submission, it was recommended to me to separate both subjects in order to make the mes- Reviewer comments sages clearer. As a result, I split the original article into In this paper, Jonathan Lombard has performed a rather this and another paper (reference [50]), where I discuss exhaustive phylogenomic analysis of N-glycosylation the implications for the last common ancestor of extant pathways in the three domains of life, focusing on the organisms (the cenancestor). In summary, the cenancestor most conserved proteins, that belong to universally dis- most likely had a biosynthetic pathway that could have tributed protein glycosyltransferase superfamilies. This is allowed the synthesis of a polyisoprenol-phosphate lipid an important work that could be useful for all scientists carrier similar to Bac-P in this organism. Together with interested in reconstructing the evolution of membranes the presence of glycosyltransferases described in the systems in the three domains. The paper is sometimes present article, I conclude that a glycosylation machinery difficult to read for people outside of the fields because, may have existed in the cenancestor. Since the over- for historical reasons, orthologous proteins have dif- whelming majority of the prokaryotic polyisoprenol-based ferent names in different domains. Jonathan Lombard glycosylation machineries are involved in the synthesis of could benefit from his analysis to propose, when pos- cell wall components, I suggest that the cenancestor sible, new names, such that orthologous proteins in dif- already had a mechanism to synthesize glycosylated cell ferent domains could be readily identified. walls. However, I could not unambiguously infer the pres- Author’s response: Thank you for this suggestion. I am ence of any oligosaccharide transferase (the proteins that aware of the fact that the number of proteins in this ana- ultimately select the final acceptor of the oligosaccha- lysis is difficult to handle, especially because it is difficult rides) in the cenancestor, so I think that the system may to see the connection among homologues that carry out have been less specific and developed into different cell different functions in different lineages. Unfortunately, walls glycosylation pathways later or just modified in the the field of polyisoprenol-related glycosylations is so vast modern lineages because of the strong selection pressures that I have not been able to figure out a systematic that act on the glycocalyx. As this is the subject of an- notation that would make sense at the scale of the three other article, I have not included this discussion in this domains. For instance, the HPT notation is already a article, but I now refer to these possibilities at the end of generic name given to a superfamily that comprises pro- the section “The eukaryotic N-glycosylation initiation teins such as Alg7, AglH MraY, WecA, TagO, etc. When complex has a composite origin”. possible, I have referred to glycosyltransferase superfam- Other proteins were probably present in the last com- ilies (GT1, GT2, etc.), but in cases such as MurG and mon ancestor of Archaea and Eukarya or originated in their homologues the renaming is difficult because of the Archaea if Eukarya emerge from within Archaea. Jonathan complicated protein domain architecture of the family Lombard has interpreted all his results on the basis of the and the fact that many of these proteins are uncharacter- second hypothesis (the archaeal ancestor scenario) for the ized. I strongly recommend using the Fig. 2 as a general origin of Eukaryotes. However, this hypothesis is mainly legend for the whole article, as it shows the relationships supported by concatenation of unresolved single trees among all the proteins discussed here. confused by the presence in the dataset of fast evolving Some of the proteins, such as HPT and MurG, indeed species (Nanoarchaea, Methanopyrus, Korarchaeon) (see appear universal and the topology of their trees (Figs. 3 Forterre 2015). This is problematic since even the best and 4) are rather consistent with those of classical uni- phylogenetic methods cannot prevent long branch versal proteins (Archaea and Eukarya closely related attraction in the presence of a very distant outgroup compared to Bacteria) (especially in the case of HPT). (Gouy et al., 2015), which is the case here with Bacteria. This suggests that these proteins are orthologous and Better analyses in which most fast evolving species have were already present in LUCA as indicated by the author been removed provided contradictory rootings. For in- in Fig. 7. Could Jonathan Lombard speculate about the stance, the analysis of Gribaldo and co-workers reject the role of these proteins in LUCA? Can we conclude that TACK (proteoarchaeal) superphylum (Rayman et al. 2015) LUCA already encoded glycosylated proteins at its sur- that the author favours here, grouping some Euryarchaea face? Is it possible to imagine a minimal glycosylation with “proteoarchaea”. The paper would be better if pathway that could have evolved into the complex path- Jonathan Lombard consider in parallel the Lake (eocyte/ ways observed in Archaea/Eukarya and that was partly proteoarchaea) scenario and the Woese’sscenarioin recruited in Bacteria for peptidoglycan biosynthesis? which Archaea and Eukarya are sister groups forming a Lombard Biology Direct (2016) 11:36 Page 23 of 31 clade (Akarya, sensu Forterre, 2015). In that case, it is is never mentioned in this paper. This is damaging and possible for instance that the Last Archaea and Eukarya can create some confusion. For instance in Fig. 5, the Common Ancestor (LAECA) contained paralogues of group “Proteoarchaea” includes only Crenarchaeal se- some proteins, followed by differential loss in some quences and two Nitrososphera (Thaumarchaeota) that archaeal phyla and in Eukarya. This could explain for in- branch within Crenarchaeota (a probable gene transfer). stance why Eukarya are more related to particular archaeal This group thus correspond to Crenarchaeota and not to phyla in the trees of Figs. 3, 5 and 6. Unfortunately, the “Proteoarchaea”. This important information is lost when phylogenie are difficult to interpret because Jonathan using the disputed phylum Proteoarchaea as an umbrella Lombard also uses many fast evolving species in his data- for diverse phyla. In general, I think dangerous and poten- set. For instance, Nanoarchaeaum and the Korarchaeon in tially misleading to talk about either the TACK or the Fig. 3, Methanopyrus kandleri in Fig. 4. I would like to see Proteoarchaea superphyla. These superphyla are only valid new phylogenies without these fast evolving organisms. if one root the archaeal tree between Euryarchaea and all It’s not necessary to do time consuming Bayesian analyses other archaea, as suggested by Petitjean et al. (2014). considering the small number of positions analyzed. These authors root the archaeal tree using Bacteria as out- PhyML would be sufficiently informative. group. Using the same strategy, Rayman et al. (2015) find Authors response: New phylogenies were carried out re- another root. It is in fact problematic to root the archaeal moving the Nanoarchaeum, Korarchaea and Methanopyrus tree using such a distant outgroup. Using Eukarya as a fast-evolving sequences. The Figs. 3 and 4b have been re- much closer outgroup, Brochier-Armanet et al. rooted the placed accordingly, but the other Figures of the paper did tree between Thaumarchaea and other Archaea (Brochier- not contain sequences from these organisms or their impact Armanet et al., 2008). We recently obtained the same was not obvious, so they were kept as they were. The overall rooting with a three domains topology of RNA polymer- topology of Fig. 3 did not change. Fig. 4b changed in that, in- ase after removing fast evolving sequences (manuscript in stead of branching within the bacterial EPS sequences, now preparation) so the question is quite open. the eukaryotic Alg13 homologues form a separate clade. Author’s response: in addition to removing the fast- I agree on the fact that previous versions of this manu- evolving sequences (see above), more thaumarchaea and script did not sufficiently discuss the phylogenies in the Bathyarchaea sequences were also included in the new context of the traditional three domains of life topology. phylogenies presented in Figs. 3 and 4a. This addition As these phylogenies are unrooted, a traditional proto- did not considerably change the topology of the resulting eukaryotic lineage cannot be ruled out from the inter- phylogenies, so the rest of phylogenies were kept un- pretation of the Alg13/Alg14, Alg2/Alg11 and Stt3 changed. Regarding to the term “Proteoarchaea”, the phylogenies. I have now discussed these possibilities in annotations in the trees are meant to provide an indica- the respective sections of the text. Yet, in order to provide tion of the taxonomic affiliation of the sequences used, as clear conclusions as possible in this already quite com- and not at all to discuss the internal phylogeny of each plicated article, I have decided to put forward the origins domain of life. For instance, there are sequences in these that I think are the most plausible in the abstract, final figures that are put under the same grouping despite be- conclusions and the summary in Fig. 7. This is based in ing paraphyletic, just to make the figures more readable. particular functional and structural arguments discussed The names chosen are such as the number of annotations for each gene, but also in the fact that the rooting of is both as few and precise as possible. For example, if a eukaryotes within archaea is now favored by most group contains mostly viridiplantae and a couple of authors (references [19–22]), including the work by rhodophytes, the whole group will be annotated as Raymann et al. 2015 cited by the reviewer. I have also “archaeplastida”, whereas if it only contains chloro- tried to forward in the conclusion the fact that the valid- phytes, it will be annotated as such. According to the ity of these hypotheses may change when more informa- same principle, if a group of sequences is only made up tion will be available about the prokaryotic relatives of of crenarchaea, it will be annotated as such, but if it also the eukaryotic proteins studied here. includes one thaumarchaea, it will be annotated as I am also surprised by the small number of “proteoarchaea”, as this is the first higher taxonomic Thaumarchaeota in the trees. I did some rapid BLAST term that includes all sequences in the group (I prefer searches and much more sequences could be retrieved. “proteoarchaea” to “TACK” as it is a name instead of an Some sequences of Bathyarchaea, that usually branch acronym and, therefore, there will always space for dis- with Caldiarchaeum and Thaumarchaea in protein trees cussion of what is in the name). That being said, I do not could also be added to see if one recover the usual think that the work by Raymann et al. 2015 necessarily monophyly of these three groups that, in my opinion, contradicts the proteoarchaeal monophily. That piece of should be all considered to be members of the phylum work found that the euryarchaea were paraphyletic, with “Thaumarchaeota”. Surprisingly, the name Thaumarchaeota some of them forming a sister group to a monophyletic Lombard Biology Direct (2016) 11:36 Page 24 of 31 group of proteoarchaea. On the base of that work, the Reviewer comments term that should be under scrutiny is “Euryarchaea”. I have only a few critical remarks. 1. This is the first The work by Brochier-Armanet et al. 2008 showed that thorough phylogenetic study of the families involved. Cenarchaeum symbiosum (a Thaumarchaea) could be While phylogenies of most of the mentioned families basal to all archaea when the eukaryotes were used as have been studied at large distances in the study an outgroup, but the choice of that outgroup is question- covering the majority of conserved protein families in able if the eukaryotes branch within the proteoarchaea, Eukaryotes (Yutin et al., 2008), these phylogenetic trees as some of the same authors suggest in the Raymann et have not been described or analyzed individually. The al. 2015 paper. In short, although the monophyly of both author presents a much more detailed study of the small the euryarchaea and proteoarchaea may be disputed, these number of families. Nevertheless, for a number of pro- groups are sufficiently well-established to be used in the tein families he has to conclude that “the closest pro- annotation of the figures in this paper, which does not in- karyotic group to the eukaryotic clade is unclear”.In tend to solve the phylogeny of archaea anyway. time with thousands new genomes sequenced and new A minor point; Jonathan Lombard use the term methods of phylogenetic analyses developed this may “Cenancestor” instead of LUCA for the Last Universal change. Therefore, I strongly believe that the author or Common ancestor. However, the “Cenancestor” can be other researches will return to the study of these protein the common ancestor of any group. If the author wants families in the future and at a very different level. There- to refer to the historical proposal by Fitch, he can con- fore, it might be helpful to add certain data to the sider that LUCA means the Last Universal CenAncestor! Supplementary material of this paper. I would insist on Author’s response: I have included the name of LUCA publishing multiple alignments of all proteins involved, in the paper, as the reviewer is right in that this term is both trimmed and untrimmed. The last suggestion is more generally known. Yet, the term “cenancestor” was prompted by the author’s argument that the topology of kept throughout the text. The definition of this term is eukaryotic clade…“most likely results from the small well stated in the paper, and I do not think that it is any number of alignment positions that was conserved in more ambiguous than the “Universal” in LUCA. each phylogeny (119 and 97 respectfully)…” Trimming Thank you for endorsing the publication of this work. of the initial alignments was done by the author auto- Gouy R, Baurain D, Philippe H. Rooting the tree of matically and it happens that the trimming parameters life: the phylogenetic jury is still out. Philos Trans R Soc used are the same for all the trees and therefore not op- Lond B Biol Sci. 2015 370(1678):20140329. doi:10.1098/ timal for some of them. Given the constantly changing rstb.2014.0329. and disappearing protein identifiers (and even gene Forterre P. The universal tree of life: an update. Front models) in the public databases, it would be essential to Microbiol. 2015 6:717. doi: 10.3389/fmicb.2015.00717. provide stable identifiers, such as GenBank CDS IDs or Raymann K, Brochier-Armanet C, Gribaldo S. The two- UniProt\UniParc IDs. domain tree of life is linked to a new root for the Archaea. Author’s response: The untrimmed and trimmed Proc Natl Acad Sci U S A. 2015 112(21):6670–5. alignments supporting all the phylogenies published in Petitjean C, Deschamps P, López-García P, Moreira D. this article have now been made available in Additional Rooting the domain archaea by phylogenomic analysis files 9 and 10 [10 and 11 after second revision], respect- supports the foundation of the new kingdom Proteoarch- ively. The IDs provided in these phylogenies are updated aeota. Genome Biol Evol. 2014 7(1):191–204. doi: 10.1093/ NCBI’s accession numbers, as currently recommended in gbe/evu274. the NCBI’s website. Although the accession numbers may indeed change over time, it has always been possible tra- Referee 2, Dr. Sergei Mekhedov, NCBI, NLM, NIH, USA cing back old accession numbers to their new IDs. If, for (nominated by Editorial Board member Michael Galperin) some reason, the identifiers were to become completely Reviewer summary inaccessible, the sequences will still be available on the I have no doubt that the paper by J. Lombard “The Additional file 9 [10, after second review]. multiple evolutionary origins of the eukaryotic N- I failed to find word “ortholog” in the text. This is not glycosylation pathway” deserves publication in Biology a big deal, but I believe that the author should have at Direct. For the first time the author puts a phylogenetic least checked the described protein families in some of analysis of N-glycosylation in the context of a metabolic the available databases of orthologous groups of pro- pathway. He performs the analyses at large evolutionary teins. I am convinced that his attempt to build his own distances comparing Eukaryotes, Archaea, and Bacteria homology groups by running BLAST all-against-all with and uses the state of the art methods. The main an unjustified expectation value cut-off resulted in in- conclusions cannot be called unexpected but are none- complete protein families for the list of Eukaryotes in theless valuable. Table 1. As an example, all question marks in Table 1 Lombard Biology Direct (2016) 11:36 Page 25 of 31 correspond to true orthologs of S. cerevisiae proteins CNC04720 [Cryptococcus neoformans var. which served as the basis for this table. The following neoformans JEC21] > tr|A0A015LNA6|A0A015LNA6_ proteins missed in Table 1 should be added. 9GLOM Swp1p [Rhizophagus irregularis DAOM 197198w] > gi|46099108|gb|EAK84341.1| hypothetical A. Missed orthologs of Ost4 (very short protein, protein UM03236.1 [Ustilago maydis 521] > tr| present in many plants): >tr|Q5KCD6|Q5KCD6_ I1G5D9|I1G5D9_AMPQE Uncharacterized protein CRYNJ Uncharacterized protein CNH00640 [Amphimedon queenslandica] > tr|Q7K110| [Cryptococcus neoformans var. neoformans JEC21] > Q7K110_DROME LD18774p [Drosophila tr|I1FYU0|I1FYU0_AMPQE Uncharacterized protein melanogaster] > sp|Q9DBG6|RPN2_MOUSE [Amphimedon queenslandica] > tr|A0A0D2X2S9| Dolichyl-diphosphooligosaccharide–protein A0A0D2X2S9_CAPO3 Uncharacterized protein glycosyltransferase subunit 2 [Mus musculus] > tr| [Capsaspora owczarzaki (strain ATCC 30864)] > tr| W4Y183|W4Y183_STRPU Uncharacterized protein L8GRU2|L8GRU2_ACACA Uncharacterized protein [Strongylocentrotus purpuratus] > tr|A0A0D2U678| [Acanthamoeba castellanii str. Neff] > sp|Q8L986| A0A0D2U678_CAPO3 Uncharacterized protein OST4B_ARATH Oligosaccaryltransferase [Arabidopsis [Capsaspora owczarzaki (strain ATCC 30864)] thaliana] > tr|J3LPF6|J3LPF6_ORYBR Uncharacterized > sp|Q54HG9|RPN2_DICDI Dolichyl- protein [Oryza brachyantha] > tr|A0A096R6U0| diphosphooligosaccharide–protein glycosyltransferase A0A096R6U0_MAIZE Uncharacterized protein subunit swp1 [Dictyostelium discoideum] > [Zea mays] > tr|A0A059A7I3|A0A059A7I3_EUCGR tr|D3BAP4|D3BAP4_POLPA Dolichyl- Uncharacterized protein [Eucalyptus grandis] > tr| diphosphooligosaccharide-protein glycotransferase G7KZM4|G7KZM4_MEDTR Oligosaccaryltransferase [Polysphondylium pallidum] > tr|L8H4E5|L8H4E5_ [Medicago truncatula] > tr|W5B6B2|W5B6B2_ ACACA Ribophorin ii (Rpn2) protein [Acanthamoeba WHEAT Uncharacterized protein [Triticum castellanii str. Neff] > tr|A8J449|A8J449_CHLRE aestivum] > sp|C7J4U3|OST4A_ORYSJ Dolichyl- Predicted protein [Chlamydomonas reinhardtii] > tr| diphosphooligosaccharide–protein glycosyltransferase D8TQJ2|D8TQJ2_VOLCA Putative uncharacterized subunit 4A [Oryza sativa subsp. Japonica] protein [Volvox carteri] > tr|E1ZHV6|E1ZHV6_CHLVA B. Missed orthologs of Swp1 (dolichyl- Putative uncharacterized protein [Chlorella variabilis] diphosphooligosaccharide-protein glycotransferase, in C. Missed ortholog of Alg12 > tr|M1V5M0| most lineages orthologs are substantially longer than in M1V5M0_CYAM1 Uncharacterized protein S. cerevisiae) > tr|C5LE16|C5LE16_PERM5 Putative CYME_CMM004C [Cyanidioschyzon merolae uncharacterized protein [Perkinsus marinus (strain strain 10D] ATCC 50983 / TXsc)] > tr|F2UPX0|F2UPX0_SALR5 D. Missed ortholog of Alg14 > gi|146144751|gb| Putative uncharacterized protein [Salpingoeca rosetta EAR99237.2| conserved hypothetical protein (strain ATCC 50818 / BSB-021)] > tr|B3SAA0| (macronuclear) [Tetrahymena thermophila SB210] B3SAA0_TRIAD Putative uncharacterized protein [Trichoplax adhaerens] > tr|Q22ZG2|Q22ZG2_TETTS Author’s response: The absence of the word “ortholog” Oligosaccharyltransferase subunit ribophorin is perfectly conscious. Orthologues are genes that have [Tetrahymena thermophila SB210] > tr|A0CJY2| evolved through speciation and, in principle, they are A0CJY2_PARTE Uncharacterized protein mostly expected to carry out the same or very similar [Paramecium tetraurelia] > tr|A0CN26|A0CN26_ functions in different species. The proteins involved in PARTE Uncharacterized protein [Paramecium eukaryotic N-glycosylation, however, belong to families tetraurelia] > tr|V9EKX7|V9EKX7_PHYPR with a high number of paralogues and probably xenolo- Uncharacterized protein [Phytophthora parasitica gues, so I preferred using the broader homology rather P1569] > sp|Q5N7W3|RPN2_ORYSJ Dolichyl- than orthology throughout the article. diphosphooligosaccharide–protein glycosyltransferase I think that there has been some kind of misunder- subunit 2 [Oryza sativa subsp. Japonica] > tr| standing with regard to Table 1 and the formation of A9S283|A9S283_PHYPA Predicted protein homology groups using all-vs-all BLAST, as both opera- [Physcomitrella patens subsp. Patens] > tions were completely independent from each other. On sp|Q93Z16|RPN2_ARATH dolichyl- the one hand, Table 1 is a typical presence/absence table. diphosphooligosaccharide–protein glycosyltransferase It was made using the well-described Saccaromyces subunit 2 [Arabidopsis thaliana] > sp|O74943| cerevisiae copy of each protein in the eukaryotic N- YJB6_SCHPO Uncharacterized protein C553.06 glycosylation pathway to look for homologues in all the [Schizosaccharomyces pombe 972 h-] > tr| eukaryotic genomes listed in the Additional file 8 [File 9 Q5KK05|Q5KK05_CRYNJ Uncharacterized protein after second review]. I have included these sequences in Lombard Biology Direct (2016) 11:36 Page 26 of 31 the updated version of Table 1, with the symbol “#”.Iam about the eukaryotic origins, there are so many different grateful to the reviewer for pointing me towards these genes in this pathway, each with a different story, that if I orthologues, that I had missed. mentioned the implications for each gene of all different On the other hand, the all-vs-all BLAST was used only eukaryogenesis hypotheses in the conclusion, the message of to build the homology groups shown in Fig. 2. In order to the article would become quite confusing (for reminder, do this, the sequences from proteins that had been de- the main message of this article is that the eukaryotic scribed in various glycosylation pathways (Table 2) were N-glycosylation pathway had mixed origins). Yet, since I used to build a small database, and the all-vs-all BLAST agree with the reviewer that the origin of eukaryotes is not was run only in that small database. The objective of set, I encourage the interested reader to refer to the that part of the study was to find out which proteins in- “Results and discussion” section, where I develop other volved in polyisoprenol-based glycosylations were related evolutionary explanations for each tree. to other proteins from other pathways. This provided a Considering the nomenclature problem, Jonathan uses visual description of the relationship among the proteins LACA, LECA, LBCA, but has still a problem with LUCA in these pathways (Fig. 2) and also determined the pro- (also the name is now mentioned). I don’tunderstand why teins that belonged to each superfamily, so they could be the U of LUCA is ambiguous? Universal, as in “universal jointly taken into account in the superfamily phylogenies. proteins” clearly means all organisms encoding ribosomes Thank you for endorsing the publication of this work. since LUCA is the Last Common Ancestor of organisms Reference Yutin, N., Makarova, K. S., Mekhedov, S. L., from the three domains that have been defined based on Wolf, Y. I., & Koonin, E. V. (2008). The Deep Archaeal ribosomal proteins (Forterre et al., 2014). Obviously, Roots of Eukaryotes. Molecular Biology and Evolution, universal do not refer to the Universe. In any case, at the 25(8), 1619–1630. moment – and probably for quite a long time - we are only concerned in studying life on Earth. If life can be Second round of review studied on planet X in the future, it will be easy to distin- Referee 1, Pr. Patrick Forterre, University Paris Sud, France guish the terrestrial LUCA from the X LUCA! Beside, the The revised version has been quite improved. It was a “L” of LUCA is very important since all the ancestors of very good idea to include both the untrimmed and LUCA were also our Cenancestors. The term Cenancestor trimmed alignments. Unfortunately, several nodes in the is therefore both ambiguous concerning the group of or- phylogenies are still not resolved after the removal of ganism under consideration but also the identity of the fast evolving species. This is probably because the small ancestor among all ancestors of the group. size of the proteins. It could be interesting to look in the Author’s response: I am very sorry, for my previous alignment before trimming to look for indels supporting response was clearly insufficient. When I said that the particular nodes. word “universal” could also be considered somehow Jonathan Lombard still favours the archaeal ancestor ambiguous, I was thinking to the fact that the term is scenario for the origin of eukaryotes to explain the data very broad. Take the example of viruses: Prof. Forterre because “most authors now favour an archaeal origin of has been a prominent actor in the debate to determine if the eukaryotic stem”. This is true but unfortunately, be- viruses are alive or not. I personally do not have a clear- cause this scenario is mostly based on very weak phylog- cut opinion on the issue, but viruses clearly are a part of enies (ref [8–11]) obtained from the concatenation of the “biological universe”. The same for prions and any individual trees that are nor resolved and contained other kind of molecular machines with various abilities. many fast evolving species (Ref [8–11]). There is only Was LUCA also the ancestor of some viruses? Maybe it one 2010 reference indicating that it’s controversial was, but in the present work I only studied the genomes (ref [18]). I would have added my two reviews in 2013 of extant cellular organisms, so I cannot say anything and 2015 in Archaea and Frontier in Microbiology in about a hypothetical ancestor of anything else. which biological arguments against the archaeal ancestor This does not mean that I completely discard the term scenario and against published phylogenies are critically “universal” as being ambiguous: what I was trying to discussed. illustrate was the fact that any term can be ambiguous if Author’s response: Thank you for your comments. not properly defined. And yet, in the present article the Although there are phylogenomic works based on larger sense of the term “cenancestor” is fully defined as “the datasets and excluding fast-evolving sequences which last common ancestor of extant organisms”. This defin- also support the internal branching of the eukaryotes ition includes the term “last”, which I agree is important, within archaea (e.g. Ref. [24]), I agree with Prof. Forterre and identifies the group under consideration (extant on the fact that the origin of the eukaryotic stem is not organisms). The term “organism” itself could be considered set yet. Thus, I have integrated in the manuscript the to be ambiguous, but we need to draw the line somewhere articles suggested by Prof. Forterre. Despite this uncertainty and I think that the reference to “extant organisms” is quite Lombard Biology Direct (2016) 11:36 Page 27 of 31 clear of the fact that I refer to cellular life as we know it. I these phyla may be closely related to each other and have am not hostile to the term “LUCA”, though, and that is been combined into a “superphylum” sometimes called why I included it in the revised version of the manuscript. TACK or proteoarchaea (Refs [10, 53]). In this particular If I use the term of cenancestor it is a matter of preference, case, the labeling as “proteoarchaea” is more informative, as the respective definitions of both terms are virtually as it shows that the sequences group together according to identical. known taxonomical clusters which would not be seen if Jonathan concludes the presence of proteins used for each sequence was labeled with the name of the particular the glycosylation of cell wall in LUCA. The term cell phyla of which sometimes they are the only representative. wall here is ambiguous since it often refers to the bacter- To take an apparently less controversial example, in the ial cell wall. In Archaea, glycosylation occurs at the level Additional file 2 most bacterial sequences are labeled with of the S-layer and in Eukaryotes at the level of the cyto- their phyla names, with the exceptions of the PVC plasmic membrane. This let open many options for the (planctomycetes/verrucomicrobia/chlamydiae) supergroup structure of the LUCA cell envelope. and the Bacteroidetes/Chlorobi supergroup. But in other Author’s response: I have now made sure that the term trees, when a group only contains planctomycetes or bac- “cell wall” is always qualified: “bacterial cell walls” refer teroidetes, it is only labeled with the name of the phylum, to components of the bacterial cell structures beyond the not the superphylum. I do not think that there is a problem plasma membrane (e.g. peptidoglycan, teichoic acids, of logic there, since the objective of these labels is not to exopolysaccharides, etc.); “archaeal cell walls” in prin- discuss what should be the definition of a microbial ciple refers to S-layer proteins, as relatively little is known phylum or superphylum, but just to show that the phylo- about how the other archeal external cell structures are genetic clustering is taxonomically significant. synthesized; “prokaryotic cell wall components” refer to Regarding Fig. 4, the original “proteoarchaeal” group both of the above; “cenancestral cell walls” are hypothet- was paraphyletic, so I changed the labeling completely. I ical, their nature is unknown and they are only men- hope that the new annotations will satisfy the concerns of tioned because the implications of this work may infer the reviewer. their existence. Cell walls will be further discussed in the Otherwise, it’s an important work that opens the way article to be published soon about the evolution of the poly- to future updated phylogenomic analyses of an import- isoprenol biosynthesis pathway. Eukaryotic cell walls are ant physiological pathway. out of the scope of this article as they are synthesized in Author’s response: thank you for your support, advice completely different ways, and they are never mentioned. and interesting debates. Jonathan Lombard still does not use the name Forterre P, Krupovic M, Prangishvili D. Cellular Thaumarchaea in the paper but still use in abundance domains and viral lineages. Trends Microbiol. 2014, the controversial term Proteoarchaea (a putative super- 22(10):554–558. phylum) that look like Proteobacteria (a phylum). I still have problem with this choice. There is a lack of logic. Referee 2, Dr. Sergei Mekhedov, NCBI, NLM, NIH, USA In several figure, group are indicated sometimes by the (nominated by Editorial Board member Michael Galperin) name of their superphylum and other time by the name I cannot understand why the author avoids not just of their phylum (for instance in Fig. 2 [sic, Figure 3?], using the word “ortholog” but also the analysis of ortho- crenarchaeota are sometime named proteoarchaea, logous relationships in the protein families involved in sometimes crenarchae). In Fig. 4, thaumarchaea are N-glycosylation. He writes in his response that there are not indicated despite the fact that they branch with many paralogs in these families. I disagree: these ortho- Euryarchaea and not with crenarchaea. logous groups in Eukaryotes do not have many paralogs. Author’s response: The objective of these labels is to Just look at Table 1! Most columns have only one show when the sequences in these trees group according symbol (+ or # or ?). Every one of these single symbols to reasonable taxonomic clusters and when they do not. that I checked corresponds to a true ortholog of S.cere- This is best done using a mix of names from superphyla, visiae protein(s) used as the basis for Table 1. And if the phyla or even names from lower taxonomic ranks (e.g. in author suspects that some are xenologs (which I could eukaryotes). not identify having checked a part of Table 1) isn’t it the In Fig. 3 there is a group of sequences from crenarchaeota task of this publication to identify them? which is labeled as crenarchaea, and a group of crenarch- Author's response: I am sorry, in my response to the aeota, thaumarchaeota, aigarchaeota, bathyarchaeota first review I did not realize that the reviewer was mainly and lokiarchaeota which is annotated as “proteoarchaea”. concerned with the relationships among the eukaryotic In the latter case, 3 of the 5 suggested “phyla” only contain sequences. I agree with Dr. Mekhedov on the fact that the one sequence, because little genomic data is available for eukaryotes sequences are most likely orthologues. The these groups. Yet, several works have suggested that all of picture is less clear, though, when the evolutionary Lombard Biology Direct (2016) 11:36 Page 28 of 31 relationships are considered across the three domains of glycosylation pathway, whereas the other studies the evo- life. For instance, Fig. 2 shows multiple paralogues of the lution of the polyisoprenol biosynthesis pathways (Dol-P same families in prokaryotes, and some phylogenies & Bac-P) in the three domains of life. Both subjects are support the existence of HGTs. Moreover, many of this distantly related, but these are two independent pieces of prokaryotic homologues carry out completely different work. The other article was mentioned earlier in relation functions. In the new version of the manuscript, I have kept to Prof. Forterre’s request to discuss possible glycosylation the term “homologue” to refer to comparisons among differ- mechanisms in LUCA. But there is nothing in the present ent domains of life and I now use the term “orthologue” to paper that directly relies on the other article. In order to refer to the eukaryotic genes. avoid any misunderstanding, I have removed that refer- I do not like the expression “distant/uncertain hom- ence and I now refer to it as follows: “The implications of ology” in the legend of Table 1. “Uncertain” homologs, these results [meaning the results in this manuscript] for in my opinion, should not be published. the possible cenancestral cell walls will be discussed in a Author’s response: I agree that the wording was not paper to be published soon (Lombard, data not shown).” ideal. The relevance of these sequences was considered as The data presented in Table 1 are not complete and I uncertain because, despite being similar, they were ex- can only guess why the author missed obvious paralogs. tremely divergent, or short, or long or had any other spe- I would like to emphasize that the first step in phylogen- cificity that would put into question the fact that they etic analysis is collection of ALL relevant sequences. The were actual orthologues carrying out the same function. author failed to do this. I do not think that it is referee’s Still, they are good candidates that it would be worth job to find all missed proteins. I checked first three col- testing experimentally, so I think that they should be re- umns for missed paralogs. My general conclusion about ported somehow. In the current version of the article I why the author failed to collect ALL homologs: he used have been less conservative and I have removed that cat- conventional unidirectional BLAST which simply does egory (now all sequences appear with a “+” in Table 1). not detect all homologs. He does not give any cutoff There are 23 columns in Table 1 each corresponding values for BLAST. Unidirectional BLAST is not enough to a protein family involved in N-glycosylation. The to identify all homologs in 2016! The author discarded paper provides analyses of five phylogenetic trees. Of the sequences that are too long which is also a signifi- course, the author might say that the other relevant trees cant mistake. He avoided this mistake in the case of are published in reference [50]. But it is not accessible. fusions in Alg13 and Alg14 groups (because these fu- This may be appropriate in a draft, but not after final sions have been already documented in the literature), review. The author has either remove this reference or but why does he think that homologs in other families make sure that it is available. are always very similar in size? They are not! Neither are Author’sresponse: The objective of this paper was to look orthologs at large distances. This is the reason for his for the possible prokaryotic origins of the eukaryotic genes omission of MOST of the proteins in Swp1 group. involved in the N-glycosylation pathway. There are 23 Obvious paralogs in Alg13 group (one of them missed eukaryotic proteins considered in Table 1, but only 10 of in Table 1): gi|283436186|ref|NP_080523.2| putative them have valuable homologues among prokaryotes. Owing bifunctional UDP-N-acetylglucosamine transferase and to the fact that some of these proteins are paralogues of each deubiquitinase ALG13 [Mus musculus]; gi|260166652| other (Alg2/Alg11, Dpm/Alg5), the 7 phylogenies provided ref|NP_796104.2| glycosyltransferase 28 domain contain- in this paper track the evolution of 9 of these 10 proteins. ing 1-like [Mus musculus]; gi|18405716|ref|NP_565950.1| The remaining protein is Alg1, which is so divergent from UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1- the rest of the large GT1 superfamily that it was impossible phosphate transferase [Arabidopsis thaliana]; gi|15230258| to determine which were its closest prokaryotic relatives. A ref|NP_191281.1| Glycosyl transferase family 4 protein supplementary phylogeny is provided (the one of the PMTs), [Arabidopsis thaliana]. because of its possible relationship to the eukaryotic-specific Author’sresponse: I now provide a completely revised ver- lumen glycosyltransferases. From my point of view, these sion of Table 1. This was made using my previous BLAST phylogenies cover all the proteins required to discuss the results and completed with supplementary BLASTs that prokaryotic origins of the eukaryotic N-glycosylation path- were carried out using alternative seeds that in some cases way. There are no assumptions made in this article about revealed a few more eukaryotic orthologues that had been the origins of the eukaryotic N-glycosylation pathway based missed. Psi-BLASTs were also carried out but the results did on work published anywhere else. not provide any additional information. Finally, the There seems to be a confusion regarding the influence EggNOG and OMA databases were checked out for any on the present article of an article currently under review other orthologue that I may have missed. The newly added (“Reference [50]” in the discussion with the reviewers). Additional file 1 contains the accession numbers of all The present paper traces the origins of the eukaryotic N- sequences that were reported in Table 1. Lombard Biology Direct (2016) 11:36 Page 29 of 31 Additional files 2. Gabaldón T, Huynen MA. From endosymbiont to host-controlled organelle: the hijacking of mitochondrial protein synthesis and metabolism. PLoS Comp Biol. 2007;3:e219. Additional file 1: Accession numbers for sequences in Table 1. (XLS 63 kb) 3. Woese CR, Kandler O, Wheelis ML. Towards a natural system of organisms: Additional file 2: Bayesian phylogeny of the HPT. (PDF 167 kb) proposal for the domains Archaea, Bacteria, and Eucarya. Proc Natl Acad Sci U S A. 1990;87:4576–9. Additional file 3: Full Bayesian phylogeny of the Alg13 and C-terminal 4. Iwabe N, Kuma K, Hasegawa M, Osawa S, Miyata T. Evolutionary relationship catalytic domain of the MurG homologues. (PDF 139 kb) of archaebacteria, eubacteria, and eukaryotes inferred from phylogenetic Additional file 4: Full Bayesian phylogeny of the Alg14 and N-terminal trees of duplicated genes. Proc Natl Acad Sci U S A. 1989;86:9355–9. membrane domain of the MurG homologues. (PDF 139 kb) 5. Gogarten JP, Kibak H, Dittrich P, Taiz L, Bowman EJ, Bowman BJ, Manolson MF, Additional file 5: Bayesian phylogeny of the Alg14 and N-terminal Poole RJ, Date T, Oshima T. Evolution of the vacuolar H + −ATPase: implications membrane domain of the MurG homologues, excluding bacterial MurG for the origin of eukaryotes. Proc Natl Acad Sci U S A. 1989;86:6661–5. sequences. (PDF 134 kb) 6. Cavalier-Smith T. The phagotrophic origin of eukaryotes and phylogenetic Additional file 6: Bayesian phylogeny of the Rft1 homologues. classification of Protozoa. Int J Syst Evol Microbiol. 2002;52:297–354. (PDF 213 kb) 7. Devos DP, Reynaud EG. Intermediate steps. Science. 2010;330:1187–8. 8. Lake JA, Henderson E, Oakes M, Clark MW. Eocytes: a new ribosome Additional file 7: Bayesian phylogeny of the PMT homologues. structure indicates a kingdom with a close relationship to eukaryotes. (PDF 139 kb) Proc Natl Acad Sci U S A. 1984;81:3786–90. Additional file 8: Bayesian phylogeny of the closest relatives to the 9. Cox CJ, Foster PG, Hirt RP, Harris SR, Embley TM. The archaebacterial origin eukaryotic Alg5/Dpm homologues (GT2 superfamily). (PDF 181 kb) of eukaryotes. Proc Natl Acad Sci U S A. 2008;105:20356–61. Additional file 9: List of the genomes used in this analysis. (PDF 106 kb) 10. Guy L, Ettema TJG. The archaeal “TACK” superphylum and the origin of eukaryotes. Trends Microbiol. 2011;19:580–7. Additional file 10: Untrimmed alignments. (TXT 2684 kb) 11. Spang A, Saw JH, Jørgensen SL, Zaremba-Niedzwiedzka K, Martijn J, Lind AE, Additional file 11: Trimmed alignments. (TXT 397 kb) van Eijk R, Schleper C, Guy L, Ettema TJG. Complex archaea that bridge the gap between prokaryotes and eukaryotes. Nature. 2015;521:173–9. 12. Rivera MC, Jain R, Moore JE, Lake JA. Genomic evidence for two functionally Abbreviations distinct gene classes. Proc Natl Acad Sci U S A. 1998;95(May):6239–44. Bac-P, bactoprenol-phosphate; Dol-P, dolichol-phosphate; EPS, exopolysaccharide; 13. Martin W, Müller M. The hydrogen hypothesis for the first eukaryote. Nature. GlcNAc, N-acetyl-glucosamine; GT, glycosyltranferase; GT1, glycosyltransferase 1998;392:37–41. superfamily 1; GT2, glycosyltransferase superfamily 2; HGT, horizontal gene 14. Baum DA, Baum B. An inside-out origin for the eukaryotic cell. BMC Biol. transfer; HMM, hidden Markov model; HPT, polyprenol phosphate:N- 2014;12:76. acetylhexosamine-1-phosphate transferase superfamily; LACA, Last Archaeal 15. Gupta RS, Golding GB. The origin of the eukaryotic cell. Trends Biochem Sci. Common Ancestor; LBCA, Last Bacterial Common Ancestor; LECA, Last 1996;21:166–71. Eukaryotic Common Ancestor; LPS, lipopolysaccharide; PMT, protein 16. Moreira D, Lopez-Garcia P. Symbiosis between methanogenic archaea and O-mannosyltransferase; TACK, Thaumarchaeota, Aigarchaeota, Crenarchaeota, delta-proteobacteria as the origin of eukaryotes: the syntrophic hypothesis. Korarchaeota J Mol Evol. 1998;47:517–30. 17. Margulis L, Dolan MF, Guerrero R. The chimeric eukaryote: origin of the Acknowledgements nucleus from the karyomastigont in amitochondriate protists. Proc Natl I thank U. Gophna and D. Moreira for the stimulating conversations that Acad Sci U S A. 2000;97:6954–9. crystallized in the development of this project and T. A. Richards for his 18. Gribaldo S, Poole AM, Daubin V, Forterre P, Brochier-Armanet C. The origin advice at later steps. I also thank D. Moreira, T. A. Richards, C. Petitjean, F. of eukaryotes and their relationship with the Archaea: are we at a Savory, R. Bamford, V. Attah three anonymous reviewers of a previous phylogenomic impasse? Nat Rev Microbiol. 2010;8:743–52. version, P. Forterre and S. Mekhedov for their comments on the manuscript. 19. Forterre P. The common ancestor of archaea and eukarya was not an archaeon. Archaea. 2013;2013. http://www.ncbi.nlm.nih.gov/pubmed/?term=The+common Funding +ancestor+of+archaea+and+eukarya+was+not+an+archaeon This work was supported by the National Science Foundation through the 20. Forterre P. The universal tree of life: an update. Front Microbiol. 2015;6:717. National Evolutionary Synthesis Center (NESCent, grant number NSF #EF-0905606). 21. Williams TA, Foster PG, Nye TMW, Cox CJ, Embley TM. A congruent phylogenomic signal places eukaryotes within the Archaea. Proc Biol Sci. 2012;279:4870–9. Availability of data and materials 22. McInerney JO, O’Connell MJ, Pisani D. The hybrid nature of the Eukaryota The untrimmed and trimmed alignments supporting the conclusions of this and a consilient view of life on Earth. Nat Rev Microbiol. 2014;12:449–55. article are respectively included in the Additional files 10 and 11. 23. Koonin EV. Archaeal ancestors of eukaryotes: not so elusive any more. BMC Biol. 2015;13:84. 24. Raymann K, Brochier-Armanet C, Gribaldo S. The two-domain tree of life is Author’s contributions linked to a new root for the Archaea. Proc Natl Acad Sci U S A. 2015;112:6670–5. JL designed the study, carried out the analyses and prepared the manuscript. 25. Eme L. Phylogénomique des structures multiprotéiques eucaryotes impliquées dans le cycle cellulaire et contribution à la phylogénie des Competing interests eucaryotes. 2011. The author declares no competing interests. 26. Eme L, Moreira D, Talla E, Brochier-Armanet C. A complex cell division machinery was present in the last common ancestor of eukaryotes. Consent for publication PLoS One. 2009;4:e5021. Not applicable. 27. Bapteste E, Charlebois RL, MacLeod D, Brochier C. The two tempos of nuclear pore complex evolution: highly adapting proteins in an ancient Ethics approval and consent to participate frozen structure. Genome Biol. 2005;6:R85. Not applicable. 28. Gabaldón T, Snel B, Van Zimmeren F, Hemrika W, Tabak H, Huynen MA. Origin and evolution of the peroxisomal proteome. Biol Direct. 2006;1. Received: 11 April 2016 Accepted: 26 July 2016 29. Dacks JB, Field MC. Evolution of the eukaryotic membrane-trafficking system: origin, tempo and mode. J Cell Sci. 2007;120(Pt 17):2977–85. 30. Collins L. Complex spliceosomal organization ancestral to extant eukaryotes. References Mol Biol Evol. 2005;22:1053–66. 1. Embley TM, Martin W. Eukaryotic evolution, changes and challenges. Nature. 31. Richards TA, Cavalier-Smith T. Myosin domain evolution and the primary 2006;440:623–30. divergence of eukaryotes. Nature. 2005;436:1113–8. Lombard Biology Direct (2016) 11:36 Page 30 of 31 32. Ramesh MA, Malik S, Logsdon JM. A phylogenomic inventory of meiotic 59. Oriol R, Martinez-duncker I, Chantret I, Mollicone R, Codogno P. Common genes: evidence for sex in Giardia and an early eukaryotic origin of meiosis. origin and evolution of glycosyltransferases using Dol-P-monosaccharides as Curr Biol. 2005;15:185–91. donor substrate. Mol Biol Evol. 2002;19:1451–63. 33. Apweiler R, Hermjakob H, Sharon N. On the frequency of protein 60. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment glycosylation, as deduced from analysis of the SWISS-PROT database. search tool. J Mol Biol. 1990;215:403–10. Biochim Biophys Acta. 1999;1473:4–8. 61. Lehrman MA. Biosynthesis of N-acetylglucosamine-P-P-dolichol, the 34. Helenius A, Aebi M. Roles of N-linked glycans in the endoplasmic reticulum. committed step of asparagjne-linked oligosaccharide assembly. Annu Rev Biochem. 2004;73:1019–49. Glycobiology. 1991;1:553–62. 35. Schiller B, Hykollari A, Yan S, Paschinger K, Wilson IBH. Complicated N-linked 62. Chantret I, Dancourt J, Barbat A, Moore SEH. Two proteins homologous to glycans in simple organisms. Biol Chem. 2012;393:661–73. the N- and C-terminal domains of the bacterial glycosyltransferase MurG are 36. Eichler J. Extreme sweetness: protein glycosylation in archaea. Nat Rev required for the second step of dolichyl-linked oligosaccharide synthesis in Microbiol. 2013;11:151–6. Saccharomyces cerevisiae. J Biol Chem. 2005;280:9236–42. 37. Albers S-V, Meyer BH. The archaeal cell envelope. Nat Rev Microbiol. 63. Gao X-D, Tachikawa H, Sato T, Jigami Y, Dean N. Alg14 recruits Alg13 to the 2011;9:414–26. cytoplasmic face of the endoplasmic reticulum to form a novel bipartite 38. Guan Z, Naparstek S, Kaminski L, Konrad Z, Eichler J. Distinct glycan-charged UDP-N-acetylglucosamine transferase required for the second step of phosphodolichol carriers are required for the assembly of the pentasaccharide N-linked glycosylation. J Biol Chem. 2005;280:36254–62. N-linked to the Haloferax volcanii S-layer glycoprotein. Mol Microbiol. 64. Lu J, Takahashi T, Ohoka A, Nakajima KI, Hashimoto R, Miura N, Tachikawa H, 2010;78:1294–303. Gao X-D. Alg14 organizes the formation of a multiglycosyltransferase complex involved in initiation of lipid-linked oligosaccharide biosynthesis. 39. Guan Z, Meyer BH, Albers S-V, Eichler J. The thermoacidophilic archaeon Glycobiology. 2011;22:504–16. Sulfolobus acidocaldarius contains an unsually short, highly reduced dolichyl phosphate. Biochim Biophys Acta. 2011;1811:607–16. 65. Lehrman MA. A family of UDP-GIcNAc/MurNAc: polyisoprenol-P GlcNAc/ 40. Jarrell KF, Ding Y, Meyer BH, Albers S-V, Kaminski L, Eichler J. N-linked MurNAc-1-P transferases. Glycobiology. 1994;4:768–71. glycosylation in Archaea: a structural, functional, and genetic analysis. 66. Dal Nogare AR, Dan N, Lehrman MA. Conserved sequences in enzymes of Microbiol Mol Biol Rev. 2014;78:304–41. the UDP-GlcNAc/MurNAc family are essential in hamster UDP-GlcNAc: 41. Chang MM, Imperiali B, Eichler J, Guan Z. N-Linked glycans are assembled dolichol-P GlcNAc-1-P transferase. Glycobiology. 1998;8:625–32. on highly reduced dolichol phosphate carriers in the hyperthermophilic 67. Anderson MS, Eveland SS, Price NPJ. Conserved cytoplasmic motifs that archaea Pyrococcus furiosus. Plos One. 2015;10:e0130482. distinguish sub-groups of the polyprenol phosphate : N-acetylhexosamine-1- 42. Cohen-Rosenzweig C, Yurist-Doutsch S, Eichler J. AglS, a novel component phosphate transferase family. FEMS Microbiol Lett. 2000;191:169–75. of the Haloferax volcanii N-glycosylation pathway, is a dolichol phosphate- 68. Bouhss A, Trunkfield AE, Bugg TDH, Mengin-Lecreulx D. The biosynthesis of mannose mannosyltransferase. J Bacteriol. 2012;194:6909–16. peptidoglycan lipid-linked intermediates. FEMS Microbiol Rev. 2008;32:208–33. 43. Nothaft H, Szymanski CM. Protein glycosylation in bacteria: sweeter than 69. Mohammadi T, Karczmarek A, Crouvoisier M, Bouhss A, Mengin-Lecreulx D, ever. Nat Rev Microbiol. 2010;8:765–78. den Blaauwen T. The essential peptidoglycan glycosyltransferase MurG 44. Jervis AJ, Langdon R, Hitchen P, Lawson AJ, Wood A, Fothergill JL, Morris HR, forms a complex with proteins involved in lateral envelope growth as well Dell A, Wren B, Linton D. Characterization of N-linked protein glycosylation in as with proteins involved in cell division in Escherichia coli. Mol Microbiol. Helicobacter pullorum. J Bacteriol. 2010;192:5228–36. 2007;65:1106–21. 45. Meyer BH, Albers S-V. Hot and sweet: protein glycosylation in 70. Shams-Eldin H, Chaban B, Niehus S, Schwarz RT, Jarrell KF. Identification Crenarchaeota. Biochem Soc Trans. 2013;41:384–92. of the archaeal alg7 gene homolog (encoding N-acetylglucosamine-1- 46. Samuelson J, Banerjee S, Magnelli P, Cui J, Kelleher DJ, Gilmore R, Robbins PW. phosphate transferase) of the N-linked glycosylation system by cross- The diversity of dolichol-linked precursors to Asn-linked glycans likely results domain complementation in Saccharomyces cerevisiae. J Bacteriol. from secondary loss of sets of glycosyltransferases. Proc Natl Acad Sci U S A. 2008;190:2217–20. 2005;102:1548–53. 71. Larkin A, Chang M, Whitworth G, Imperiali B. Biochemical evidence for an 47. Maita N, Nyirenda J, Igura M, Kamishikiryo J, Kohda D. Comparative alternate pathway in N-linked glycoprotein biosynthesis. Nat Chem Biol. structural biology of eubacterial and archaeal oligosaccharyltransferases. 2013;9:367–73. J Biol Chem. 2010;285:4941–50. 72. Kolkman MAB, Van der Zeijst BAM, Nuijten PJM. Functional analysis 48. Schwarz F, Aebi M. Mechanisms and principles of N-linked protein of glycosyltransferases encoded by the capsular polysaccharide glycosylation. Curr Opin Struct Biol. 2011;21:576–82. biosynthesis locus of Streptococcus pneumoniae serotype 14. J Biol 49. Bugg T, Brandish P. From peptidoglycan to glycoproteins: common Chem. 1997;272:19502–8. features of lipid-linked oligosaccharide biosynthesis. FEMS Microbiol Lett. 73. De Vuyst L, De Vin F, Vaningelgem F, Degeest B. Recent developments in 1994;119:255–62. the biosynthesis and applications of heteropolysaccharides from lactic acid 50. Burda P, Aebi M. The dolichol pathway of N -linked glycosylation. Biochim bacteria. Inter Diary J. 2001;11:687–707. Biophys Acta. 1999;1426:239–57. 74. Meyer BH, Peyfoon E, Dietrich C, Hitchen P, Panico M, Morris HR, Dell A, 51. Noffz C, Keppler-Ross S, Dean N. Hetero-oligomeric interactions between Albers S-V. Agl16, a thermophilic glycosyltransferase mediating the last step early glycosyltransferases of the dolichol cycle. Glycobiology. 2009;19:472–8. of N-Glycan biosynthesis in the thermoacidophilic crenarchaeon Sulfolobus acidocaldarius. J Bacteriol. 2013;195:2177–86. 52. Petitjean C, Deschamps P, López-García P, Moreira D. Rooting the domain archaea by phylogenomic analysis supports the foundation of the new 75. Maeda Y, Kinoshita T. Dolichol-phosphate mannose synthase: Structure, kingdom Proteoarchaeota. Genome Biol Evol. 2015;7:191–204. function and regulation. Biochim Biophys Acta. 2008;1780:861–8. 53. Kelleher DJ, Banerjee S, Cura AJ, Samuelson J, Gilmore R. Dolichol-linked 76. Helenius J, Ng DTW, Marolda CL, Walter P, Valvano MA, Aebi M. oligosaccharide selection by the oligosaccharyltransferase in protist and Translocation of lipid-linked oligosaccharides across the ER membrane fungal organisms. J Cell Biol. 2007;177:29–37. requires Rft1 protein. Nature. 2002;415:447–50. 54. Gao X-D, Nishikawa A, Dean N. Physical interactions between the Alg1, 77. Frank CG, Sanyal S, Rush JS, Waechter CJ, Menon AK. Does Rft1 flip an Alg2, and Alg11 mannosyltransferases of the endoplasmic reticulum. N-glycan lipid precursor? Nature. 2008;454:E3–4. discussion E4–5. Glycobiology. 2004;14:559–70. 78. Rush JS, Gao N, Lehrman MA, Matveev S, Waechter CJ. Suppression of Rft1 55. Ullman C, Perkins S. A classification of nucleotide-diphospho-sugar expression does not impair the transbilayer movement of Man5GlcNAc2-P-P- glycosyltransferases based on amino acid sequence similarities. Biochem J. dolichol in sealed microsomes from yeast. J Biol Chem. 2009;284:19835–42. 1997;326(Pt 3):929–39. 79. Jelk J, Gao N, Serricchio M, Signorell A, Schmidt RS, Bangs JD, 56. Liu J, Mushegian A. Three monophyletic superfamilies account for the Acosta-Serrano A, Lehrman MA, Bütikofer P, Menon AK. Glycoprotein majority of the known glycosyltransferases. Prot Sci. 2003;12:1418–31. biosynthesis in a eukaryote lacking the membrane protein Rft1. 57. Coutinho PM, Deleury E, Davies GJ, Henrissat B. An evolving hierarchical J Biol Chem. 2013;288:20616–23. family classification for glycosyltransferases. J Mol Biol. 2003;328:307–17. 80. Eisenhaber B, Maurer-Stroh S, Novatchkova M, Schneider G, Eisenhaber F. 58. Lairson LL, Henrissat B, Davies GJ, Withers SG. Glycosyltransferases: Enzymes and auxiliary factors for GPI lipid anchor biosynthesis and structures, functions, and mechanisms. Annu Rev Biochem. 2008;77:521–55. post-translational transfer to proteins. Bioessays. 2003;25:367–85. Lombard Biology Direct (2016) 11:36 Page 31 of 31 81. Finn RD, Clements J, Arndt W, Miller BL, Wheeler TJ, Schreiber F, Bateman A, 103. Altenhoff AM, Škunca N, Glover N, Train C-M, Sueki A, Piližota I, Gori K, Eddy SR. HMMER web server: 2015 update. Nucleic Acids Res. 2015;43:W30–8. Tomiczek B, Müller S, Redestig H, Gonnet GH, Dessimoz C. The OMA orthology 82. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ. database in 2015: function predictions, better plant support, synteny view and Gapped BLAST and PSI-BLAST : a new generation of protein database other improvements. Nucleic Acids Res. 2015;2015(43):D240–9. search programs. Nucleic Acids Res. 1997;25:3389–402. 104. Edgar RC. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 2004;32:1792–7. 83. Lussier M, Sdicu A, Bussey H. The KTR and MNN1 mannosyltransferase 105. Philippe H. MUST, a computer package of management utilities for families of Saccharomyces cerevisiae. J Biol Chem. 1995;270:2770–5. sequences and trees. Nucleic Acids Res. 1993;21:5264–72. 84. Gentzsch M, Tanner W. The PMT gene family: protein O-glycosylation in 106. Price MN, Dehal PS, Arkin AP. FastTree 2–approximately maximum-likelihood Saccharomyces cerevisiae is vital. EMBO J. 1996;15:5752–9. trees for large alignments. PLoS One. 2010;5:e9490. 85. Loibl M, Strahl S. Protein O-mannosylation: what we have learned from 107. Penn O, Privman E, Ashkenazy H, Landan G, Graur D, Pupko T. GUIDANCE: baker’s yeast. Biochim Biophys Acta. 2013;1833:2438–46. a web server for assessing alignment confidence scores. Nucleic Acids Res. 86. VanderVen BC, Harder JD, Crick DC, Belisle JT. Export-mediated assembly 2010;38(Web Server issue):W23–8. of mycobacterial glycoproteins parallels eukaryotic pathways. Science. 108. Ronquist F, Teslenko M, van der Mark P, Ayres DL, Darling A, Höhna S, 2005;309:941–3. Larget B, Liu L, Suchard MA, Huelsenbeck JP. MrBayes 3.2: efficient Bayesian 87. Lommel M, Strahl S. Protein O-mannosylation: conserved from bacteria to phylogenetic inference and model choice across a large model space. humans. Glycobiology. 2009;19:816–28. Syst Biol. 2012;61:539–42. 88. Meyer BH, Zolghadr B, Peyfoon E, Pabst M, Panico M, Morris HR, Haslam SM, 109. Stamatakis A. RAxML version 8: a tool for phylogenetic analysis and Messner P, Schäffer C, Dell A, Albers S-V. Sulfoquinovose synthase - an post-analysis of large phylogenies. Bioinformatics. 2014;30:1312–3. important enzyme in the N-glycosylation pathway of Sulfolobus acidocaldarius. 110. Miller MA, Pfeiffer W, Schwartz T. Creating the CIPRES Science Gateway for Mol Microbiol. 2011;82:1150–63. Inference of Large Phylogenetic Trees. In: Proceedings of the Gateway 89. Yan Q, Lennarz WJ. Stt3p is directly involved in the glycosylation process. Computing Environments Workshop (GCE). New Orleans; 2010:1–8. http:// J Biol Chem. 2002;277:47692–700. www.phylo.org/sub_sections/portal/cite.php 90. Koiwa H, Li F, Mccully MG, Mendoza I, Koizumi N, Manabe Y, Nakagawa Y, 111. Le SQ, Gascuel O. An improved general amino acid replacement matrix. Zhu J, Rus A, Pardo JM, Bressan RA, Hasegawa PM. The STT3a subunit Mol Biol Evol. 2008;25:1307–20. isoform of the Arabidopsis oligosaccharyltransferase controls adaptive 112. Le SQ, Dang CC, Gascuel O. Modeling protein evolution with several responses to salt/osmotic stress. Plant Cell. 2010;15(October 2003):2273–84. amino acid replacement matrices depending on site rates. Mol Biol Evol. 91. Chaban B, Voisin S, Kelly J, Logan SM, Jarrell KF. Identification of genes 2012;29:2921–36. involved in the biosynthesis and attachment of Methanococcus voltae 113. Kaminski L, Guan Z, Yurist-doutsch S. Two distinct N-glycosylation pathways N-linked glycans: insight into N-linked glycosylation pathways in Archaea. process the Haloferax volcanii S-layer glycoprotein upon changes in Mol Microbiol. 2006;61:259–68. environmental salinity. mBio. 2013;4:1–9. 92. Igura M, Maita N, Obita T, Kamishikiryo J, Maenaka K, Kohda D. Purification, 114. Typas A, Banzhaf M, Gross CA, Vollmer W. From the regulation of crystallization and preliminary X-ray diffraction studies of the soluble peptidoglycan synthesis to bacterial growth and morphology. Nat Rev domain of the oligosaccharyltransferase STT3 subunit from the thermophilic Microbiol. 2011;10:123–36. archaeon Pyrococcus furiosus. Acta Crystallogr Sect F: Struct Biol Cryst 115. Raetz CRH, Whitfield C. Lipopolysaccharide endotoxins. Annu Rev Biochem. Commun. 2007;63:798–801. 2002;71:635–700. 93. VanDyke DJ, Wu J, Logan SM, Kelly JF, Mizuno S, Aizawa S-I, Jarrell KF. 116. Brown S, Santa Maria JP, Walker S. Wall teichoic acids of gram-positive Identification of genes involved in the assembly and attachment of a novel bacteria. Annu Rev Microbiol. 2013;67:313–36. flagellin N-linked tetrasaccharide important for motility in the archaeon 117. Vik Å, Erik F, Haug J, Bilsborough S, Schneider A, Egge-jacobsen W, Koomey M. Methanococcus maripaludis. Mol Microbiol. 2009;72:633–44. Broad spectrum O-linked protein glycosylation in the human pathogen 94. Matsumoto S, Shimada A, Nyirenda J, Igura M, Kawano Y, Kohda D. Crystal Neisseria gonorrhoeae. Proc Natl Acad Sci U S A. 2009;106:4447–52. structures of an archaeal oligosaccharyltransferase provide insights into the 118. Messner P, Steiner K, Zarschler K, Schäffer C. S-layer nanoglycobiology of catalytic cycle of N-linked protein glycosylation. Proc Natl Acad Sci U S A. bacteria. Carbohydr Res. 2008;343:1934–51. 2013;110:17868–73. 119. Rick PD, Barr K, Sankaran K, Kajimura J, Rush JS, Waechter CJ. Evidence that 95. Meyer BH, Albers S-V. AglB, catalyzing the oligosaccharyl transferase step of the wzxE gene of Escherichia coli K-12 encodes a protein involved in the the archaeal N-glycosylation process, is essential in the thermoacidophilic transbilayer movement of a trisaccharide-lipid intermediate in the assembly crenarchaeon Sulfolobus acidocaldarius. MicrobiologyOpen. 2014;3:531–43. of enterobacterial common antigen. J Biol Chem. 2003;278:16534–42. 96. Magidovich H, Eichler J. Glycosyltransferases and oligosaccharyltransferases 120. Whitfield C. Biosynthesis and assembly of capsular polysaccharides in in Archaea: putative components of the N-glycosylation pathway in the Escherichia coli. Annu Rev Biochem. 2006;75:39–68. third domain of life. FEMS Microbiol Lett. 2009;300:122–30. 121. Vorhölter F-J, Schneiker S, Goesmann A, Krause L, Bekel T, Kaiser O, Linke B, 97. Kaminski L, Lurie-Weinberger MN, Allers T, Gophna U, Eichler J. Phylogenetic- Patschkowski T, Rückert C, Schmid J, Sidhu VK, Sieber V, Tauch A, Watt SA, and genome-derived insight into the evolution of N-glycosylation in Archaea. Weisshaar B, Becker A, Niehaus K, Pühler A. The genome of Xanthomonas Mol Phylogenet Evol. 2013;68:327–39. campestris pv. campestris B100 and its use for the reconstruction of 98. Heesen S, Lehle L, Weissmann A, Aebi M. Isolation of the ALG5 locus metabolic pathways involved in xanthan biosynthesis. J Biotechnol. encoding the UDP-glucose : dolichyl-phosphate glucosyltransferase from 2008;134:33–45. Saccharomyces cerevisiae. Eur J Biochem. 1994;224:71–9. 122. Reuber TL, Walker GC. Biosynthesis of succinoglycan, a symbiotically 99. Orlean P, Albright C, Robbins PW. Cloning and sequencing of the yeast important exopolysaccharide of Rhizobium keliioti. Cell. 1993;74:269–80. gene for dolichol phosphate mannose synthase, an essential protein. J Biol Chem. 1988;263:17499–507. 100. Colussi PA, Taron CH, Mack JC, Orlean P. Human and Saccharomyces cerevisiae dolichol phosphate mannose synthases represent two classes of the enzyme, but both function in Schizosaccharomyces pombe. Proc Natl Acad Sci U S A. 1997;94(July):7873–8. 101. Zarschler K, Janesch B, Pabst M, Altmann F, Messner P, Schäffer C. Protein tyrosine O-glycosylation–a rather unexplored prokaryotic glycosylation system. Glycobiology. 2010;20:787–98. 102. Huerta-Cepas J, Szklarczyk D, Forslund K, Cook H, Heller D, Walter MC, Rattei T, Mende DR, Sunagawa S, Kuhn M, Jensen LJ, von Mering C, Bork P. eggNOG 4.5: a hierarchical orthology framework with improved functional annotations for eukaryotic, prokaryotic and viral sequences. Nucleic Acids Res. 2016;44:D286–293.

Journal

Biology DirectSpringer Journals

Published: Aug 4, 2016

There are no references for this article.