The 5′ cap is a unique feature of eukaryotic messenger RNAs (mRNA) and eukaryotic viruses not found on eubacterial and archaeal RNAs [1]. The addition of a m⁷G cap structure, or the cap 0 nucleotide, occurs co-transcriptionally via three consecutive reactions executed by the capping enzymatic apparatus: (i) hydrolysis of the 5′-triphosphate of nascent pre-mRNAs to a diphosphate by RNA 5′ triphosphatase (TPase), (ii) capping of the diphosphate end with GMP by the RNA guanylyltransferase (GTase) and (iii) methylation of the GpppN cap by RNA guanine-7- methyltransferase (MTase). The RNA cap is involved in multiple cellular functions including splicing, nucleocytoplasmic export, mRNA turnover and translation initiation [2], [3].

In addition to the N7-methylated guanosine cap, the 1st and 2nd adjacent nucleotides may be 2′O-ribose methylated after transcription [4] forming cap 1 and cap 2 structures, respectively. In contrast to the cap 0, the role of cap 1 and cap 2 modified nucleotides is unclear. Their presence may reveal a phylogenetic trend with increasing levels of complexity in multicellular eukaryotes [4], [5]. mRNAs in multiple species of yeast have been shown to contain only a cap 0 whereas the adjacent 5′ nucleotides of most multicellular eukaryotic mRNAs are further modified to form cap 1 and/or cap 2 structures [4], [6]. Thus far trypanosomes are the only protists to have their mRNA cap structures examined and these were found to contain a hypermodified cap 4 structure m⁷Gpppm₂⁶AmpAmpCmpm³Um [7]. The unconventional method of providing mRNAs with cap modification via trans-splicing of a splice leader RNA at their 5′ ends confers this unique cap structure to trypanosome mRNAs [8], [9]. Recently, enzymes involved in cap 4 formation have been identified [10]–[12], although the role of these different cap nucleotide modifications remains to be elucidated [11], [13].

The N7-methylated guanosine cap (cap 0) is a structural consensus in all eukaryotic mRNAs unlike the different levels of modification of subsequent nucleotides observed in different organisms. The type and structural organization of the guanosine cap enzymatic apparatus that confers the cap 0 to the 5′end of mRNAs has been examined in a wide variety of eukaryotes [14]–[18], including the divergent protists Giardia [19] and trypanosomes [20]–[22]. An interesting evolutionary scenario has emerged [1], which contradicts [23] current models derived from multiple eukaryotic phylogenetic analyses [24]. This scenario dictates a multicellular (eg. metazoa and plants) versus a microbial eukaryotic pattern (e.g. fungi and microbial eukaryotes). The majority of investigated microbial eukaryotic species possess a three-component capping system (TPase, GTase and MTase) while metazoan and plants encode a two-component system with the fusion of TPase and GTase polypeptides and a separate MTase [1].

In contrast to structural conservation of GTases, the TPases appear to have evolved from different protein ancestors [1]. TPases found in multicellular eukaryotes contain the cysteinyl-phosphatase superfamily motif HCXXXXXR(S/T) named phosphate-binding loop or ‘p-loop’ (TPasePL). This enzyme catalyses a two-step phosphoryl transfer in which the conserved cysteine attacks the γ-phosphorus of the 5′-triphosphate on the nascent RNA to form a covalent protein-cysteinyl-S-phosphate intermediate producing a 5′-diphosphate RNA product [17], [25]. The enzyme-bound phosphate is then hydrolyzed and liberated as inorganic phosphate. In contrast, all investigated microbial eukaryotic TPases to date share structural organization with metal-dependent phosphohydrolases (TPaseMDP) and have different structural configurations and enzymatic characteristics [26]. A current model of the mRNA capping system evolution in eukaryotes suggests that a ‘transitional state organism’ would contain both types of TPases [1]. The fusion of the TPasePL-GTase was followed by a secondary loss or complete divergence of the TPaseMDP, and the last common ancestor (LCA) of plants and metazoans would carry the fused TPasePL-GTase version only. Although this model is supported by the distribution of these enzymes in several microbial eukaryotes [1], and is consistent with eukaryote phylogenies based on few genes supporting a metazoan-plant relationship [23], it is incongruent with phylogenomic and multiple single gene phylogenies [24], [27]. However this mRNA capping centric view of eukaryote phylogeny does not preclude the occurrence of differential gene loss/gain during eukaryotic evolution.

Here, we have characterized the guanosine cap enzymatic apparatus of Trichomonas vaginalis, a divergent microbial eukaryote, that is a member of the Parabasalia and the super-group Excavata [24], [28]. In contrast to microbial eukaryotes, including other members of the Excavata, the diplomonad G. lamblia [19] and kinetoplastids [20], [22] T. vaginalis has a dual TPasePL-GTase capping enzyme (TvCE) resembling the metazoan-plant type, as previously predicted from genome analyses [29]. Moreover, we have demonstrated that T. vaginalis mRNAs contain a complex cap structure with a canonical m⁷G and adjacent modified nucleotides. Phylogenetic analyses of the GTase domain only and TPasePL-GTase alignments are consistent with a common origin of the T. vaginalis and metazoan-plant enzymes, which suggest that the TPasePL-GTase system is likely to be more ancient then previously thought and that complex scenarios of independent gene loss and/or gain events across various eukaryotic lineages may have taken place.

We have demonstrated previously that T. vaginalis mRNAs have a 5′-end protection that can be removed by pyrophosphatase treatment, and that these mRNAs partially precipitate with a monoclonal anti-trimethylguanosine antibody [30]. Therefore, similar to other eukaryotes this divergent protist must harbor an enzymatic capping apparatus. The T. vaginalis genome database (www.trichdb.org) [31] was screened using BLASTp analysis using available homologs for capping enzymes found in protists, yeast, plants and metazoans. As we previously reported based on analyses of the T. vaginalis genome [29] the same putative T. vaginalis capping enzyme gene (locus tag TVAG_187730, RefSeq accession XP_001327945.1, named here TvCE) is identified in searches conducted using the GTase enzyme found in many microbial eukaryotes and those using the fused TPasePL-GTase from plants and metazoans. No genes were identified in searches using TPaseMDP sequences. Both TPasePL-GTase functional domains in TvCE were conserved relative to the human capping enzyme in protein domain analyses as those from other metazoans, plants, green algae and a choanoflagellate, the latter being a member of the Choanomonada: microbial eukaryotes closely related to metazoan [24], [28] (Figure 1 & Figure S1). TvCE shares 30% identity and 47% similarity to the human capping enzyme and also has its TPasePL-domain fused to the N-terminus of a GTase domain (Figure S1). Strikingly, the TvCE cysteinyl-phosphatase superfamily motif HCXXXXXR(S/T) is 100% identical to the human sequence (Figure 1A). Furthermore, TvCE contained the six conserved peptide motifs (I, III, IIIa, IV, V and VI) at the C-terminus that comprise the active site for GTP binding and nucleotidyl transfer of GTase [32] (Figure 1B). Individual BLASTp searches with either the entire TvCE protein, the N-terminal domain encompassing the TPasePL domain (residues 1–258) or the two C-terminal domains (GTase, residues 259–441 and 442–561) identified by protein domain analyses, recovered as top hits animal sequences (see Data S1). These data indicate that T. vaginalis may have a two-component capping system similar to metazoan and plants.

To examine the enzymatic activity of recombinant TvCE we attempted to express either the full-length protein or each domain separately. We found that the single GTase domain was insoluble and that the TPase, although soluble, was inactive. Therefore, the full-length recombinant TvCE was used to analyse enzyme activity. The capping enzyme TPase specifically hydrolyses the γ-phosphate from the 5′-terminus of RNAs allowing its activity to be monitored by either the production of inorganic phosphate or ATP-ADP conversion. The TPase activity of recombinant TvCE was tested in the absence of metals at varying pH and the protein was found to be active in pH ranging from 4.5 to 6.5 (Figure 2A). Addition of 1–5 mM MgCl₂ inhibited TvCE TPase activity and addition of EDTA could reverse this inhibition (Figure 2B). Activity in an acidic pH range and inhibition by MgCl₂ are typical of a cysteinyl-phosphatase TPase and are not observed in classic microbial eukaryotic TPases [1]. In mammalian cysteinyl-phosphatase TPase, a transient phosphocysteine-enzyme intermediate can be trapped using a short incubation time at low temperature and acidic pH [17]. We asked whether TvCE also exhibits this property and found that the phospho-TvCE intermediate was detected in the presence of [γ³²P]ATP, but not in the presence of [α³²P]ATP as predicted, and within a restricted acidic pH (3.5–4.0) (Figure 2C). Phospholabeling of TvCE was lost after treatment with iodine but not hydroxylamine, supporting the presence of a thiophosphate linkage predicted to be at the C126 within the ‘P-loop’ (Figure 2D, left panel). To determine whether C126 is involved in this linkage, TvCE was subjected to specific amino acid mutations and tested for the ability to form a thiophosphate linkage. As shown in Figure 2D (right panel), serine substitution of either C126 or H125 abrogates formation of phospho-TvCE. When enzymatic affinity was compared across substrate concentration, we observed that P-loop mutants (H125S and C126S) and the double mutant (C126S and R526A) were inert to release γ-phosphate from ATP (Figure 3A), which is in agreement with the inability of these mutants to form a phosphocysteine-enzyme intermediate (Figure 2D). However, a mutation in the GTase domain alone (R526A) did not significantly affect TPase activity (Figure 3A). The utilization of ATP and GTP by TvCE as a function of nucleotide concentration is similar (Figure 3B). A relatively low conversion of the substrate is observed which may indicate that TvCE has either a low turnover rate or is partially inactivated during the 2-step purification. Together, these data indicate that C126 and H125 coordinate the cleavage and release of the γ-phosphate and demonstrate that TvCE has characteristics of a typical cysteinyl-phosphatase TPase [1].

The capping enzyme GTase uses a ‘ping-pong’ reaction mechanism for nucleotidyl transfer through a covalent enzyme-(lysyl-N)-GMP intermediate [33]. This allows GTase activity to be detected by ³²P transfer from [α-³²P]GTP to the enzyme [14], [16], [18], [34]. To determine whether TvCE also uses this reaction mechanism, we incubated recombinant TvCE with [α-³²P]GTP in the presence or absence of divalent cations. The formation of the SDS-stable ³²P-labeled enzyme was then evaluated by SDS-PAGE (Figure 4). TvCE GTase activity was detected by [α-³²P]GTP labeling in a broad pH window (data not shown) and required the presence of either MnCl₂ or MgCl₂. Calcium did not support activity (Figure 4A). Metal dependence, specifically Mn²⁺ and Mg²⁺, is a typical characteristic of an RNA capping GTase.

To determine the specificity of TvCE GTase, competition reactions were performed using increasing concentrations of cold NTPs. The recombinant TvCE GTase displayed an absolute specificity for a GTP substrate since ATP, CTP or UTP were not found to inhibit this enzyme (Figure 4B). Guanylation of TvCE is dependent on nucleotide concentration (Figure 4C). In accordance with previous reported structure/function studies [18], the single arginine substitution at GTase motif VI (R526A) abolished TvCE activity (Figure 4C). Interestingly, we also found that mutations in the TPase domain (H125S and C126S) significantly affected GTase activity, suggesting that these residues may exert a cis-structural effect on the GTase domain (Figure 4C). Using size exclusion chromatography, we found both TPase and GTase activities in a single discrete peak corresponding to ∼71 kDa indicating that TvCE is monomeric (Figure S2).

In addition to using nucleotide substrates to assess the activity of TvCE we have also evaluated its ability to transfer GMP from GTP to the 5′ triphosphate end of an in vitro transcribed RNA, using the full-length enzyme with both active domains. As shown in Figure 5, TvCE is capable of labeling an RNA substrate provided this molecule harbors a 5′ triphosphate (lane 2). When the substrate is dephosphorylated by alkaline phosphatase digestion prior to incubation with TvCE, no transfer of [α-³²P]GTP to the substrate is achieved (lane 3). In order for the GTase to transfer [α-³²P]GMP from [α-³²P]GTP to the 5′-end of the RNA substrate, the γ-phosphate must be first removed by the TPase (lane 4). When TPase or GTase activity is dependent on a TvCE mutant that is inactive in the corresponding domain, no labeling of the substrate is observed (lanes 5 & 8). On the other hand, when TPase or GTase activity is dependent on a TvCE mutant that inactivates the opposite domain, labeling of the substrate is achieved (lanes 6 & 7). These data are consistent with those characterizing individual domains using nucleotide substrates and TvCE mutants (Figure 3A and 4C) and demonstrate that capping of an RNA substrate strictly depends on the activity of both domains (Figure 5).

The observed similarities between metazoa and T. vaginalis RNA capping enzymatic activities prompted us to investigate the mRNA cap structure in this organism. The nucleotides adjacent to the m⁷G-cap structure are methylated to different extents in eukaryotes. Metazoans can have 2′-O-ribose methylations of the first and second transcribed nucleotides forming the cap structure m⁷GpppNmpNmpNp, where the first transcribed nucleotide is an adenosine. However, yeast is found to have no modifications beyond the 5′ m⁷G cap [4], [6]. We developed a multi-step protocol to purify mRNA suitable for structural analysis of nucleotides (see Materials and Methods). None of the steps alone, including consecutive passages over poly-dT chromatography columns, was sufficient to remove abundant RNA species such as tRNA and rRNA (Figure S3). As these contaminants contain hypermodified nucleotides in relatively high abundance that interfered with our analysis, we found it necessary to also immunoprecipitate RNAs with anti-TMG (which cross-reacts with the 7-methyl guanosine cap of mRNAs) to remove uncapped RNA contaminants (Figure S3). As a result, in vivo ³²P-labeled mRNA obtained by this purification protocol was shown to be free of contaminating rRNA and tRNA by gel electrophoresis (data not shown) and structural analysis of nucleotides after P1 digestion (Figure 6A and Figure S3). The abundance of adenosine and uridine, relative to guanosine and cytosine, in this heterogeneous mRNA population (Figure 6A) is consistent with the reported ∼65% AT content of T. vaginalis genes [29]. Digestion of the purified mRNA sample with Tobacco Acid Pyrophosphatase (TAP), which specifically hydrolyzes the phosphoric acid anhydride bonds in the triphosphate bridge of a 5′-end cap structure, released m⁷GMP (Figure 6A). The identity of this modified nucleotide was confirmed by demonstrating that it can be converted to m^2,2,7GMP using Schizosaccharomyces pombe trimethylguanosine synthase or SpTgs (Figure 6A) [35], [36].

Next, a sample of unlabeled mRNA was used for analysis of the first transcribed nucleotide (position +1). RNAs were subjected to TAP or a TAP-mock treatment, alkaline phosphatase digestion and then 5′-end labeled using T4 polynucleotide kinase. These were then digested completely by nuclease P1, and nucleotides were resolved on 2D-TLC plates. Appearance of distinct spots presence only in the TAP treated sample should reveal whether the first transcribed mRNA nucleotide (position +1), which is protected by an m⁷G cap, is modified (Figure 6B). Our result demonstrated that T. vaginalis mRNAs have a typical metazoan 2′-O-ribose methylated cap 1 nucleotide. The cap 1 nucleotide is either an adenosine (80%) or a cytosine (20%) based on a comparison of the Am/Cm ratio (Figure 6B). This is in agreement with ∼75% of T. vaginalis protein-coding genes being preceded by the conserved initiator element (Inr) TCAT/_CT/_A that dictates transcription initiation at the underlined adenosine [29], [37]. The four unmodified nucleotides (A, C, G and U) observed in both samples likely result from partially degraded RNAs that would not require TAP treatment to be 5′-end labeled or the presence of intact RNA contaminants that lack a cap.

There is a strong bias for a uridine at positions +2 and +3 in most T. vaginalis mRNAs [37], [38]. To demonstrate whether the uridine +2 is 2′-O-ribose methylated forming a cap 2 structure, in vivo labeled mRNA was digested with RNase T2 prior to anti-TMG precipitation, the last step of the mRNA purification protocol (Figure 7A). 2′-O-ribose methylation of this nucleotide would render the adjacent 3′-5′ phosphodiester linkage resistant to RNase T2 as this enzyme cleaves RNA via 2′-3′phosphate cyclization. These samples were then immunoprecipitated with anti-TMG agarose beads and 3′-5′ phosphodiester linkages were cleaved by on-bead RNase P1 treatment, to restrict analysis to the cap 2 modified nucleotide, if present, and its adjacent cap 1 nucleotide. Released nucleotides were then analyzed by 2D-TLC. If nucleotide +2 is not 2′-O-ribose methylated only the cap 1 unmodified nucleotide, mostly uridine, would be released, compared to the release of both 2′-O-ribose methylated nucleotides (position +2) and the unmodified nucleotide (position +3) if a cap 2 modified nucleotide is present (Figure 7A). As a result, no modified nucleotides were detected indicating the absence of a cap 2 structure in T. vaginalis mRNAs (Figure 7B). As predicted the most abundant unmodified nucleotide detected was uridine. Although these data cannot exclude the presence of a cap 2 structure in a small subset of mRNAs, if present these are highly underrepresented.

The unexpected finding of a metazoan/plant-like capping apparatus and a cap 1 modified nucleotide on T. vaginalis mRNAs led us to investigate the phylogeny of TvCE. A global taxa sampling of animal, fungal, plant, microbial and iridovirus GTases as well as a subset dataset including exclusively the TPasePL-GTase structural organization were aligned and subjected to protein maximum likelihood phylogenetic analyses (Figure 8 & Figure 9). Although generally poorly resolved, the global GTase phylogeny recovered the T. vaginalis sequence with modest support value (64% bootstrap proportion, BP but increasing to 70% and 77% when one or two of the most divergent sequences were removed) in a clan of exclusively TPasePL-GTase configured sequences, consistent with a common origin of their structural organization (Figure 8). The TPasePL-GTase maximum likelihood tree revealed that TvCE does not cluster with proteins from animals (Figure 9) and hence, at face value, does not support lateral gene transfer (LGT) between T. vaginalis or a Parabasalid ancestor and their animal hosts. These data are consistent with an ‘Unikonts’/‘Bikonts’ split [39] with on one hand plants and green algae forming a clan with T. vaginalis supported with 58% BP, raising to 77% in the absence of the iridovirus sequence (Figure 9), and the metazoan and choanoflagellates forming the other clan in line with phylogenomic data [24]. As the position of the eukaryotic root is currently unknown [39] it is not clear how the iridovirus sequence relates to the eukaryotic sequences; its divergent sequence makes inference of its position in both analyses tentative. Consistent with potential long-branch attraction (LBA) issues for this sequence, its removal in the TPasePL-GTase alignment improved the BP of several nodes including the one for the T. vaginalis-plant/green algae clan and the metazoan clan (Figure 9). Recoding the 20 amino acids into four categories, to allow optimization of the rate matrix and reduce composition heterogeneity between sequences and mitigate potential LBA artefacts, recovered a similar tree with reduced BP for the split between the TvCE-plant/algae clan and the metazoan clan in maximum likelihood analyses. Together our data favor a tree topology consistent with phylogenomics data [24] and cluster TvCE with plant and green algae sequences; however, LGT from a metazoan donor to a parabasalid cannot be strictly rejected due to lack of strong signal.

This is the first report of the functional characterization of a microbial eukaryote harboring a typical metazoan/plant-like capping apparatus containing a fused dual-functional TPasePL-GTase (TvCE). In addition to this unusual capping apparatus in T. vaginalis, mRNAs from this unicellular eukaryote were found to contain a cap 1 structure, another feature of metazoan mRNAs. The GTase domain of TvCE selectively binds GMP. The transfer of GMP to the 5′ triphosphate end of the RNA is strictly dependent on the prior removal of the γ-phosphate by the TPase activity of TvCE. The TPase domain of TvCE has typical features of a metazoan cysteinyl-phosphatase enzyme; it is active in the absence of metals as a bifunctional monomeric enzyme, prefers an acidic pH, and forms a phospho-enzyme through a cysteine-thiophosphate linkage within a conserved ‘P-loop’ active site which is necessary for phosphatase activity. An apparent cis-structural effect in TvCE was observed between the two active domains as a single amino acid change in the GTase domain results in a detectable reduction of TPase activity. We also observed that purification of the individual TvCE TPase domain resulted in an inactive enzyme. Together, these data suggest that the GTase domain can affect the activity of its neighboring TPase domain.

Analyses of the cap structure of mRNAs in this divergent microbial eukaryote revealed the presence of a canonical m⁷G cap 0 nucleotide, consistent with the presence of two conserved RNA (guanine-7) MTase genes in the T. vaginalis genome (http://trichdb.org) that could convert the guanosine at the 5′ end of the mRNA by TvCE to an m⁷G. We previously showed that T. vaginalis has an atypical trimethylguanosine synthase (TgS) that produces m^2,7G from m⁷G RNA substrates [36]. These observations led us to speculate that T. vaginalis mRNAs might contain m^2,7G caps. Contrary to this prediction, the analyses presented here demonstrate the presence of a canonical m⁷G cap nucleotide on T. vaginalis mRNAs.

A general comparison of cap structures among different eukaryotes suggested that complexity increases following a phylogenetic trend across the evolution of eukaryotes [5]. This is illustrated by the presence of a complex cap 2 structure on mRNAs of metazoans but only cap 0 on budding yeast mRNAs. Prior to the studies reported here, the unusual cap 4 in trypanosomatids [7]–[11] appeared to be the exception to this rule. T. vaginalis mRNAs which contain a conserved cap 1 structure composed primarily of m⁷GpppAmpUp or m⁷GpppCmpUp are now also exceptions. No cap 2 structure was detected in T. vaginalis mRNA and a strong bias to uridines at position +2 was observed, making this cap structure less complex than metazoans. It should also be noted however, that the use of a heterogeneous population of mRNA may have obscured the detection of a cap 2 nucleotide in a smaller subpopulation of mRNAs. Similarly, although we consider it unlikely, we cannot strictly preclude that immunoprecipitation of mRNAs using the anti-TMG antibody led to the exclusion of a subset of mRNAs with a cap 2 nucleotide or an alternative cap structure that are not efficiently bound by this antibody.

The observation that ∼80% of cap 1 nucleotides in T. vaginalis mRNAs are adenosines (A) while ∼20% are cytosines (C) indicates that transcription initiation by RNA polymerase II in this organism can occur at cytosine, as well as adenosine. In turn this suggests the presence of either unknown variants of the initiator (Inr) motif that surrounds the start site of transcription of T. vaginalis mRNAs [37], [38] or unrelated motifs that can direct transcription to initiate at a cytosine. This is consistent with our previous prediction that only ∼75% of T. vaginalis genes appear to use a classic Inr leading to transcription initiation at an A [29].

The structural organization of the mRNA caps enzyme machinery in eukaryotes has been considered a marker for inference of eukaryote phylogeny [1]. Fungi and most sampled microbial eukaryotes have separate TPaseMDP and GPase capping enzymes with the TPaseMDP structurally and mechanistically distinct from the TPasePL that is fused to a GTase in metazoan and plants. Thus, the acquisition of the metazoan-type dual-function enzyme after the divergence of unicellular and multicellular eukaryotes has been proposed [1]. However recent genome samplings have broadened the taxonomic diversity of TPasePL-GTase encoding taxa, which now includes green algae, a choanoflagellate and T. vaginalis. These additional data and our structural, functional and phylogenetic analyses of TvCE complicates the earlier simple dichotomy observed between metazoan-plants and microbial eukaryotes. Indeed, the proponent of this dichotomy acknowledged that “The scheme is certainly oversimplified…” due to poor taxa sampling [1].

Differential gene losses could explain the unusual presence of a metazoan/plant-like capping apparatus in T. vaginalis, green algae and choanoflagellates. The LCA of microbial eukaryotes, plants and animals may have contained the fused TPasePL-GTase and independent gene losses subsequently occurred in most currently sampled microbial eukaryotes, leaving T. vaginalis, green algae and choanoflagellates as rare microbial eukaryotes carrying this prototype. Notably the green algae and choanoflagellate TPasePL-GTase sequences were recovered in the expected clans as defined by phylogenomics [24], pushing the acquisition of this configuration among a likely microbial ancestor deeper in eukaryotic evolution. Alternatively, the LCA of plants/green algae and animals/choanoflagellates had the split system seen today in many microbial eukaryotes and subsequently lost this and acquired a new fused version. Similar but independent events would than be invoked for the acquisition of TvCE, leading to an enzyme with little sequence relatedness to its counterpart in plants and animals. However the global GTase phylogeny suggests that the GTase from the TPasePL-GTase fusion was shared by the LCA between T. vaginalis, plants/algae and choanoflagellate/animals; as expected if the gene was present in the LCA of all eukaryotes. Alternatively, the GTase for the taxa with TPasePL-GTase configurations was independently acquired by different lineages from similar sources with the same structural organization and/or T. vaginalis acquired TvCE by LGT. Existing phylogenies do not provide positive evidence for these scenarios. As Parabasalia and Diplomonads appear to be closely related within the excavates [24] an independent origin scenario of the TPasePL-GTase fusion could have been suggested if the TvCE GTase formed a clan with the G. lamblia GTase, however this was not found in our maximum likelihood phylogenetic analyses (Figure 8, Figure S4). These data clearly reinforce the importance of further genome sampling among various microbial eukaryotes and viruses before evolutionary hypotheses based on the mRNA capping system can be appropriately assessed and contrasted to the current hypothesis of eukaryote phylogeny [39]. Phylogenomic data obtained so far consistently provide evidence for at least six major eukaryotic lineages [24], [27] that do not match those suggested by the distribution of the mRNA capping enzymes [1]. Our phylogenetic analyses do not support LGT as the source of TvCE in T. vaginalis reducing the conflict between the distribution of the mRNA capping machinery and the six major lineages recovered by phylogenomics [39]. In contrast, relationships among microbial eukaryote GTases functioning with TPaseMDP seem more complex and in conflict with phylogenomic data in terms of major group relationships. For instance non-monophyly is observed for microsporidial GTases, which do not cluster with the Fungi, and the red algae cluster with the Amoebozoa. Could these be explained by LGT events as described for the acquisition of a trifunctional capping enzyme in a mimivirus, thought to be derived from its amoeba host [40]? Likewise the iridovirus TPasePL-GTase may have been acquired from an animal host. The evolution of capping enzymes in eukaryotes appears to have proceeded via multiple events that led to the independent loss and/or gain of genes in different lineages. Polarization of such events will require denser sampling of mRNA capping genes and additional robust independent phylogenetic analyses.