Methionine is included in all defined media designed for the growth of trypanosomatids with or without symbionts 20 22 43 , suggesting that these protozoans are incapable of methionine synthesis. However, experimental evidence has shown that homocysteine and/or cystathionine could substitute for methionine in culture media for trypanosomatids 43 45 64 .

Our analyses suggest that regular trypanosomatids and symbiont-harboring trypanosomatids have the necessary genes to produce cystathionine, homocysteine, and methionine from homoserine (Figure 2), whereas the endosymbiont genomes have no gene for the enzymes involved in the synthesis of methionine from homoserine. However, homoserine is produced from aspartate semialdehyde through the mediation of homocysteine methyltransferase (EC:1.1.1.3), which is universally present in the genomes of all the endosymbionts, symbiont-harboring and regular trypanosomatids examined.

With respect to cysteine synthesis, it has been shown that the incubation of cell homogenates of C. fasciculata with ³⁵S-methionine produced radioactive adenosyl-methionine (SAM), adenosyl-homocysteine (SAH), homocysteine, cystathionine, and cysteine 45 . Thus, this trypanosomatid is fully equipped to methylate methionine to produce homocysteine and, thereon, to convert homocysteine into cysteine through the trans-sulfuration pathway. However, with respect to the cystathionine/cysteine interconversion, there is some ambiguity concerning the presence or absence of cystathionine gamma-lyase (EC:4.4.1.1) in regular trypanosomatids. Many sulfhydrolases have a domain composition very similar to that of EC:4.4.1.1, which makes a definitive in silico function assignment to any of them difficult. Specifically, the enzymes cystathionine gamma-synthase (EC:2.5.1.48) and O-acetylhomoserine aminocarboxypropyltransferase (EC:2.5.1.49), and the two versions of cystathionine beta-lyase (EC:4.4.1.8) are possible candidates to mediate the trans-sulfuration step attributed to EC:4.4.1.1, but further research is required to establish which of these enzymes, if any, performs that reaction. We also found that, in addition to the standard pathway for methionine/cysteine synthesis (Figure 2, compounds III-X), all symbiont-harboring and regular trypanosomatids examined had the genes to produce cysteine from serine in a simple two-step reaction, with acetylserine as an intermediate (Figure 2, I-III).

In summary, if regular and symbiont-harboring trypanosomatids are capable of interconverting methionine and cysteine, as shown for C. fasciculata 43 , none of these two amino acids can be considered essential for trypanosomatids as the presence of one renders the other unnecessary. In that case, both can be synthesized by trypanosomatids, without any participation of their symbionts, except in the optional production of aspartate semialdehyde and homoserine. However, the expression of these genes remains to be confirmed.

In trypanosomatids, initial investigations about the nutritional requirements for threonine were controversial. Most results suggested that this amino acid is essential 43 45 48 64 65 66 , but other studies considered the addition of threonine to the growth media of regular trypanosomatids unnecessary 33 . Our genomic analysis favors the latter observations.

Threonine, one of the precursors of isoleucine, can be produced by different biosynthetic pathways. We have examined two of these possible routes, one starting from glycine and the other from aspartate, as presented in Figure 3. The conversion of glycine plus acetoaldehyde into threonine is mediated by threonine aldolase (EC:4.1.2.5). The gene for this enzyme is absent from endosymbionts but present in the genomes of symbiont-harboring trypanosomatids and C. acanthocephali, but not Herpetomonas. It is also absent from the genomes of trypanosomes but present in the genome of Leishmania major (KEGG data).

The pathway from aspartate utilizes the first two enzymes (EC:2.7.2.4 and EC:1.2.1.11) of the DAP pathway from lysine synthesis for the production of aspartate semialdehyde. These genes are present exclusively in the symbiont genomes. Aspartate semialdehyde is then sequentially converted into homoserine, phosphohomoserine, and threonine. The gene encoding homoserine dehydrogenase (EC:1.1.1.3) is universally present in the genomes of the endosymbionts, symbiont-harboring and regular trypanosomatids. It is also present in the genomes of T. cruzi and Leishmania spp. In contrast, the genes for the enzymes leading from homoserine to threonine via phosphohomoserine (EC:2.7.1.39 and EC:4.2.3.1) are present in the genomes of all insect trypanosomatids (including symbiont-harboring ones), of Trypanosoma spp., and Leishmania spp., but totally absent from the endosymbiont genomes.

Thus, the genetic constitution of regular trypanosomatids is consistent with earlier nutritional data showing the insect trypanosomatids, with or without symbionts, to be autotrophic for threonine. This observation suggests that endosymbionts are able to enhance the host cell threonine synthesis by producing the metabolic precursor aspartate semialdehyde that is also involved in other metabolic pathways.

The overall genomic and enzymatic picture is in apparent contradiction with early nutritional findings showing that threonine promoted the growth of trypanosomatids in culture 67 . This contradiction might find its basis in the fact that endogenously produced threonine is required by many metabolic processes, such that supplementation of the culture media could enhance the growth of the trypanosomatids.

Isoleucine, valine, and leucine are considered essential nutrients for the growth of all trypanosomatids, except symbiont-harboring ones. The canonic pathway for the synthesis of isoleucine is depicted in Figure 4. Oxobutanoate (alpha-ketoglutaric acid) is the starting point of the pathway, and can be produced in two ways: from threonine (Figure 4, compounds II-III) or from pyruvate (Figure 4, compounds I, IX). The conversion of threonine into oxobutanoate is mediated by threonine deaminase (EC:4.3.1.19). The specific activity of this enzyme was higher in symbiont-enriched subcellular fractions of symbiont-harboring trypanosomatid homogenates than in any other cell fraction or in the cytosol, suggesting that this enzyme was located in the symbiont 33 . However, genes for EC:4.3.1.19 are present in the genomes of endosymbionts, as well as those of symbiont-harboring and regular trypanosomatids (except Leishmania and Trypanosoma), contrasting with enzymatic determinations showing the absence of enzyme activity in regular trypanosomatids 33 . Since the presence of the gene does not guarantee the functionality of the enzyme for that specific reaction, the issue remains to be experimentally verified. The next enzymatic step, the transference of the acetaldehyde from pyruvate to oxobutanoate, is mediated by the enzyme acetolactate synthase (EC:2.2.1.6), which is present exclusively in the genomes of endosymbionts. Also present only in symbionts are the genes for the next four enzymes of the pathway, which are common for valine and isoleucine synthesis. However, the gene for a branched-chain amino acid transaminase (EC:2.6.1.42), mediating the last step in the synthesis of isoleucine, valine, and leucine, is present in the genomes of symbiont-harboring and regular trypanosomatids, but not endosymbionts.

The first step of the valine pathway is the conversion of pyruvate into hydroxymethyl ThPP, mediated by an enzyme of the pyruvate dehydrogenase complex (EC:1.2.4.1) whose gene is present in the genomes of endosymbionts and symbiont-harboring and regular trypanosomatids. The next reaction, leading to acetolactate, is mediated by acetolactate synthase (EC:2.2.1.6), whose gene is present exclusively in the genomes of the endosymbionts. The reactions that follow from acetoacetate into valine involve the same endosymbiont genes from isoleucine synthesis.

Synthesis of leucine uses oxoisovalerate, an intermediate metabolite of the valine pathway that is converted into isopropylmalate by 2-isopropylmalate synthase (EC:2.3.3.13), encoded by a gene present only in the endosymbionts – as are the genes for the enzymes catalyzing the next three steps for leucine biosynthesis. The presence of the gene for this branched-chain amino acid transaminase (EC:2.6.1.42) in the genomes of regular trypanosomatids explains the earlier finding that oxopentanoate and oxoisovalerate, the immediate precursors of isoleucine, valine, and leucine could substitute for these amino acids when added to regular trypanosomatid synthetic culture media 43 . Interestingly, this gene is present in all symbiont-harboring and regular trypanosomatid genomes examined, but absent from endosymbiont genomes (Figure 4). It is also present in the genomes of T. brucei and the leishmaniae available from KEGG. In addition to isoleucine, valine, and leucine biosynthesis, this enzyme also participates in the degradation of these amino acids for their use in other metabolic processes in the cell, which might explain the presence of this enzyme as the only representative of the pathway in all regular trypanosomatids examined.

A coupled biosynthetic pathway of the branched-chain amino acids was also described for the symbiotic bacterium Buchnera and its aphid host, where the symbiont has the capability to synthesize the carbon skeleton of these amino acids but lacks the genes for the terminal transaminase reactions 68 69 . The aphid possesses genes hypothesized to accomplish these missing steps, even if orthologs of those are found in other insects and carry out different functions 70 . The branched-chain amino acid transaminase (EC:2.6.1.42) encoded by an aphid gene was shown to be up-regulated in the bacteriocytes, supporting the cooperation of Buchnera and its host in the synthesis of essential amino acids 71 . Since this transamination involves the incorporation of amino-N and the aphid diet is low in nitrogen, the host mediation of this step would be a way of maintaining a balanced profile of amino acids through transamination between those that are over abundant and those that are rare 71 72 .

In summary, the presence in endosymbionts of most genes involved in isoleucine, valine and leucine synthesis explains why symbiont-harboring trypanosomatids, but not regular ones, are autotroph for these essential amino acids. However, it is worth noting that the presence of the branched-chain amino acid transaminase in trypanosomatids indicates that the host might control amino acid production according to their necessity and the nutrient availability in the medium.

There are no enzymatic data concerning the synthesis of phenylalanine, tryptophan, and tyrosine in trypanosomatids. However, it is well known that these amino acids are essential in defined culture media designed for regular trypanosomatids, but not for symbiont-harboring ones 20 22 43 44 . The biosynthetic routes for these three amino acids use chorismate, which is produced from phosphoenolpyruvate (PEP) via the shikimate pathway, as a common substrate. The genomes of all endosymbionts contain the genes for this route, while the genomes of symbiont-harboring and regular trypanosomatids do not (Figure 5).

The genes for the enzymes converting chorismate into prephenate and for transforming this compound into phenylalanine and tyrosine are present in all endosymbiont genomes. Symbiont-harboring and regular trypanosomatid genomes also have the genes for the last step in the synthesis of phenylalanine and tyrosine, but it is not known whether all of these enzymes are functional. The gene for phenylalanine-4-hydroxylase (EC:1.14.16.1), which converts phenylalanine into tyrosine, is present in symbiont-harboring and regular trypanosomatids, including the leishmaniae, but not in endosymbionts. Similarly, this enzyme is present only in the aphid concerning the metabolic partnership between Buchnera and its insect host. Furthermore, the gene encoding this enzyme is up-regulated in bacteriocytes, thus enhancing the production and interconversion of such amino acids 71 . On the other hand, endosymbionts have an additional route for the synthesis of phenylalanine from prephenate, involving enzymes aromatic-amino-acid aminotransferase (EC:2.6.1.57) and prephenate dehydratase (EC:4.2.1.51), whose genes are absent in symbiont-harboring and regular trypanosomatid genomes.

The case of the last enzyme of the tryptophan pathway is rather interesting. Tryptophan synthase (EC:4.2.1.20) possesses two subunits. This bi-enzyme complex (a tetramer of two alpha and two beta subunits) channels the product of the alpha subunit (indole) to the beta subunit, which condenses indole and serine into tryptophan 73 . Both subunits are present in the endosymbionts, whereas the genomes of symbiont-harboring trypanosomatids and H. muscarum have only the beta subunit. None of the other trypanosomatid genomes examined presented either subunit of tryptophan synthase.

In summary, the endosymbionts have all the genes for the different routes leading from chorismate to tryptophan, tyrosine, and phenylalanine, which are absent from symbiont-harboring and regular trypanosomatid genomes. This obviously prevents regular trypanosomatids from synthesizing any of these three amino acids and growing without supplementation. It is worth observing that the presence of phenylalanine hydroxylase, which converts phenylalanine into tyrosine, in trypanosomatids but not in endosymbionts indicates that the host might control tyrosine production.

Histidine is derived from three precursors: the ATP purine ring furnishes a nitrogen and a carbon, the glutamine contributes with the second ring nitrogen, while PRPP donates five carbons. Histidine is a truly essential amino acid for most trypanosomatids, as corroborated by its obligatory presence in every synthetic media so far devised for regular trypanosomatid growth 22 43 44 . Accordingly, symbiont-harboring and regular trypanosomatid genomes do not seem to carry a single gene for histidine synthesis (Figure 6). All genes for the enzymes that participate in its biosynthesis, except the gene for histidinol-phosphate phosphatase (HPP, EC:3.1.3.15), which converts histidinol phosphate into histidinol, are present in the endosymbiont genomes. Since symbiont-harboring trypanosomatids do not require histidine, it is presumed that the absent EC:3.1.3.15 is replaced by an equivalent enzyme yet to be characterized (see Other observation on amino acid pathway peculiarities).

Organisms autotrophic for ornithine use the glutamate pathway 74 for its synthesis via acetylated compounds as represented in Figure 7 (I-VI). All genes for this pathway are present in the genomes of endosymbionts. The last step in the synthesis of ornithine can also be performed by the enzymes aminoacylase (EC:3.5.1.14) or acetylornithine deacetylase (EC:3.5.1.16), which convert acetylornithine into ornithine and are present in the genomes of symbiont-harboring and regular trypanosomatids, but not endosymbionts.

As represented in Figure 7, organisms lacking the glutamate pathway for the synthesis of ornithine can nevertheless produce it by different routes utilizing either citrulline or arginine 37 39 54 . Ornithine can be produced from the hydrolysis of citrulline mediated by citrulline hydrolase (EC:3.5.1.20). This activity is present in cell homogenates of all trypanosomatids, except the leishmaniae and trypanosomes, but the corresponding gene has not yet been identified to date in any organism, making it impossible to perform similarity searches. Ornithine can also be produced from arginine by means of arginase (EC:3.5.3.1), which splits arginine into ornithine and urea. The gene for arginase is present in the genomes of symbiont-harboring trypanosomatids and some regular trypanosomatids (Leishmania and C. acanthocephali), but not in the genomes of endosymbionts or H. muscarum – although a fragment was found in the later (see HGT and arginine and ornithine biosynthesis).

Arginine can be synthesized from ornithine through a recognized universal enzymatic pathway 74 , the first step of which is the conversion of ornithine and carbamoyl phosphate into citrulline mediated by OCT (ornithine carbamoyl transferase, EC:2.1.3.3). The gene for OCT was found in the genomes of all endosymbionts and also in Herpetomonas, but was absent from other regular, as well as symbiont-harboring, trypanosomatid genomes examined. These findings confirm earlier immunocytochemical ultrastructural experiments showing the presence of OCT in the symbiont of Angomonas deanei 36 . The absence of the OCT gene renders most trypanosomatids unable to make citrulline from ornithine 75 . However, the genes for the remaining enzymes leading from citrulline into arginine are all present in the genomes of all regular and symbiont-harboring trypanosomatids, but absent from the endosymbiont genomes. These data are in full accordance with earlier enzymatic determinations for argininosuccinate synthase (EC:6.3.4.5), argininosuccinate lyase (EC:4.3.2.1), and arginase (EC:3.5.3.1) in cell homogenates of trypanosomatids 38 39 56 .

Taking all these data together, we can conclude that regular trypanosomatids require exogenous sources of arginine or citrulline in their culture medium to produce ornithine. This is related to the fact that regular trypanosomatids lack the glutamate pathway for ornithine synthesis. Furthermore, ornithine cannot substitute for arginine or citrulline because most regular trypanosomatids lack OCT. Conversely, symbiont-harboring trypanosomatids are autotrophic for ornithine. This is due to the fact that, although the symbiont lacks most genes for ornithine production, it contains sequences for key enzymes such as those for the glutamate route and OCT, which converts ornithine into citrulline thus completing the urea cycle.

As shown in Figure 7, putrescine, a polyamine associated with cell proliferation, can be produced from ornithine in a one-step reaction mediated by ODC (ornithine decarboxylase, EC:4.1.1.17), whose gene is present in the genomes from the genus Angomonas and regular trypanosomatids, but not in endosymbionts or Strigomonas. Interestingly, it was proposed that the symbiont can enhance the ODC activity of A. deanei by producing protein factors that increase the production of polyamines in the host trypanosomatid 76 . Such high ODC activity may be directly connected to the lowest generation time described for trypanosomatids that is equivalent to 6 hours 13 . Putrescine could also be produced from agmatine since the genomes of regular and symbiont-harboring trypanosomatids have the gene for agmatinase (EC:3.5.3.11), converting agmatine into putrescine. However, the gene for the enzyme arginine decarboxylase (EC:4.1.1.19), which synthesizes agmatine, is present solely in the genomes of endosymbionts, thus completing the biosynthetic route for this polyamine, via agmatinase, in symbiont-harboring trypanosomatids. Putrescine is then converted to spermidine and spermine by enzymes S-adenosylmethionine decarboxylase (EC:4.1.1.50) and spermidine synthase (EC:2.5.1.16). The genes for these enzymes are present in the regular and symbiont-harboring trypanosomatids, but not in endosymbionts (Figure 7). Enzyme EC:2.5.1.16, converting S-adenosylmethioninamine and putrescine into S-methyl-5’-thioadenosine and spermidine, also participates in a reaction from the methionine salvage pathway. This pathway is present, complete in all symbiont-harboring and regular trypanosomatids examined (Additional file 3), although there are questions regarding the step catalyzed by acireductone synthase (EC:3.1.3.77, see HGT and methionine and cysteine biosynthesis).

Our data on the phylogeny of the genes for essential amino acid biosynthesis have clearly shown that the genes present in the symbionts are of betaproteobacterial origin (for an illustrative example, see Figure 8), as shown before for the genes of heme synthesis 58 and many others across the endosymbiont genomes 25 . The symbiont-harboring and regular trypanosomatid genomes, on the other hand, present a rather different situation. Thus, 18 of the 39 genes required for the biosynthesis of essential amino acids exhibited at least some phylogenetic evidence of having been horizontally transferred from a bacterial group to a trypanosomatid group, with three other genes presenting undetermined affiliation (see Additional file 2 for a summary of the phylogenetic analyses results). As detailed below, horizontal gene transfer (HGT) events seem to have originated from a few different bacterial taxa, although in some cases the exact relationship was not completely clear. Also, while some transfers are common to all trypanosomatid groups examined, others were found to be specific to certain subgroups. This could be due to multiple HGT events from associated bacteria at different points of the family’s evolutionary history or, alternatively, to HGT events that occurred in the common ancestor of all trypanosomatids, whose corresponding genes were later differentially lost in certain taxa. Given the low number of genomes currently known in the family, it is difficult to assign greater probability to either scenario.

Regarding the taxonomic affiliation of the putative origin of these HGT events, it is possible to notice a preponderance of bacteria from a few phyla with three or more genes transferred, i.e. Firmicutes, Bacteroidetes, and Gammaproteobacteria, plus a few other phyla with two or less genes represented, like Actinobacteria, Betaproteobacteria, Acidobacteria, and Alphaproteobacteria. In a few other cases, the trypanosomatid genes grouped inside diverse bacterial phyla, in which case the assignment of a definite originating phylum was not possible. However, given the sometimes high rate of HGT in prokaryotic groups, it is difficult to assess with confidence the correct number of putative HGT events from Bacteria to Trypanosomatidae. It is possible that some of the genes that seem to have originated from different phyla could actually have come from one bacterial line that was itself the recipient of one or more previous HGT events from other bacteria.

Analysis of all generated phylogenetic inferences has uncovered a clear pattern for the HGT events, which were shown to be concentrated preferentially in pathways or enzymatic steps that are usually reported to be absent in eukaryotes, particularly animals and fungi. Thus, the HGT events identified in this study involve pathways for the synthesis of lysine, cysteine, methionine, threonine, tryptophan, ornithine, and arginine (Figures 1, 2, 3, 5, and 7) and also the synthesis of a few non-essential amino acids such as glycine, serine, and proline. The detailed analysis of these events in different genes and pathways follows.

Some enzymes are common to a number of pathways involving key precursors to many compounds. Homoserine dehydrogenase (EC:1.1.1.3), for example, participates in the aspartate semialdehyde pathway for the synthesis of lysine, cysteine, methionine, and threonine (Figures 1, 2, and 3). The gene for EC:1.1.1.3 present in symbiont-harboring and regular trypanosomatid genomes seems to have been transferred from a member of the Firmicutes, clustering most closely with Solibacillus silvestris, Lysinibacillus fusiformis, and L. sphaericus with bootstrap support value (BSV) of 100 (Figure 8). On the other hand, the endosymbiont ortholog groups deep within the Betaproteobacteria, more specifically in the Alcaligenaceae family, as expected in the case of no HGT of this gene into the endosymbiont genomes.

The two genes of the lysine pathway (Figure 1) that were found in trypanosomatid genomes presented evidence of HGT. H. muscarum was the only trypanosomatid analyzed containing the next to last gene, for diaminopimelate epimerase (EC:5.1.1.7), which clusters strongly with the phylum Bacteroidetes, with BSV of 99 (Additional file 4). The last gene, for diaminopimelate decarboxylase (EC:4.1.1.20), was present in the symbiont-harboring and regular trypanosomatids. In the phylogeny, this particular gene has Actinobacteria as sister group (BSV of 79), although also grouping with a few other eukaryotic genera, most closely Dictyostelium, Polysphondylium, and Capsaspora, with BSV of 65 (Additional file 5). There are, overall, very few Eukaryota in the tree for 4.1.1.20, making it hard to reach a definite conclusion on the direction of transfer for this gene, since other eukaryotes are also present basally to this substantially large group of Actinobacteria plus Trypanosomatidae, with the high BV of 98.

Using the C. acanthocephali gene for EC:4.1.1.20 in a manual search against the L. major genome has shown a small fragment with significant similarity (57% identity and 67% similarity, from amino acid 177 to 227), but containing stop codons. Search against predicted L. major proteins yielded no results. These sequence remains suggest that Leishmania could have lost DAP-decarboxylase in a relatively recent past.

The pathways for cysteine and methionine synthesis (Figure 2) present the highest number of HGT events identified among the pathways studied here. The gene for the enzyme EC:2.3.1.30, necessary for the conversion of serine to cysteine, seems to have been transferred from Bacteria to the genomes of host trypanosomatids. EC:2.3.1.30 of symbiont-harboring and regular trypanosomatids grouped inside a large cluster of diverse Bacteria (predominantly Bacteroidetes and Betaproteobacteria), with BSV of 80 (Additional file 6). An even deeper branch, which separates the subtree containing the trypanosomatids from the rest of the tree, has BSV of 97. The evolutionary history of the other enzyme with the same functionality, EC:2.5.1.47, is unclear and can not be considered a case of HGT given current results. Its gene is present in symbiont-harboring and regular trypanosomatids (including one sequence from T. cruzi CL Brener) and clusters as a sister group of Actinobacteria, although with low BSV (Additional file 7). Although there are many other eukaryotes in the tree, they are not particularly close to the subtree containing the Trypanosomatidae. Interestingly, one Entamoeba dispar sequence is a sister group to the Trypanosomatidae, although with low BSV, raising the possibility of eukaryote-to-eukaryote HGT, as previously reported (reviewed in 77 ).

The gene for EC:2.3.1.46, the first in the pathway converting homoserine to cystathionine, is present in all symbiont-harboring trypanosomatids and Herpetomonas, but in no other regular trypanosomatid examined. This trypanosomatid gene groups within Bacteroidetes, with BSV of 53 and, in a deeper branch, BSV of 89, still clustering with Bacteroidetes only (Additional file 8).

The gene for EC:2.1.1.37, responsible for the first step in the conversion of S-adenosylmethionine into homocysteine, is present in all symbiont-harboring and regular trypanosomatids, although the sequence is still partial in the genome sequences of the Angomonas species. Almost all organisms in the tree are Bacteria of several different phyla (Additional file 9), with the few Eukaryota present forming a weakly supported clade. KEGG shows that many Eukaryota do possess a gene for enzyme EC:2.1.1.37, but their sequences are very different from that present in the trypanosomatids (and other eukaryotes) studied here. This therefore suggests a bacterial origin for the EC:2.1.1.37 from the Eukaryota in our tree, although the specific donor group cannot be currently determined with confidence. It is interesting to note that, besides the Trypanosomatidae, the clade of eukaryotes is composed of Stramenopiles and green algae (both groups that have, or once had, plastids), with a Cyanobacteria close to the base of the group. Although the BSV of 54 does not allow strong conclusions regarding this group, it is interesting to speculate about the possibility of eukaryote-to-eukaryote gene transfer, as previously seen (reviewed in 77 ), after the acquisition of this gene from a so-far unidentified bacterium.

The genes for EC:2.5.1.48, EC:2.5.1.49, and EC:4.4.1.8 (two versions) are quite similar in sequence and domain composition. Therefore, similarity searches with any one of these genes also retrieves the other three. In spite of the similarities, these genes are found in rather different phyletic and phylogenetic patterns on the trypanosomatids (Additional file 10). EC:2.5.1.48 is present in all symbiont-harboring and regular trypanosomatids examined, plus Trypanosoma sp. and a few other Eukaryota (mostly Apicomplexa and Stramenopiles), all within a group of Acidobacteria (BSV of 94). The gene for EC:2.5.1.49 is present in the symbiont-harboring trypanosomatids and Herpetomonas, but in none of the other regular trypanosomatids examined. This trypanosomatid gene also clusters with diverse groups of Bacteria, although low BSV makes it hard to confidently identify its most likely nearest neighbor, and it is not possible to conclude with reasonable certainty that this gene is derived from HGT. The gene for EC:4.4.1.8 occurs, in symbiont-harboring and regular trypanosomatids, as two orthologs presenting very different evolutionary histories. One of the orthologs clusters with eukaryotes, with BSV of 95, while the other seems to be of bacterial descent, grouping mostly with Alphaproteobacteria of the Rhizobiales order, with BSV of 99.

The presence of two genes identified as EC:4.4.1.8 raises the possibility of them performing different enzymatic reactions. Given the overall domain composition similarities of several of the genes of the methionine and cysteine synthesis pathways, it is possible that one of the enzymes identified as EC:4.4.1.8 is actually the enzyme EC:4.4.1.1, for which no gene has been found in our searches of the Trypanosomatidae genomes, as detailed above (Methionine and cysteine).

Genes for two of the enzymes for the last step in the methionine synthesis, EC:2.1.1.10 and EC:2.1.1.14 (Additional files 11 and 12), are present in all regular and symbiont-harboring trypanosomatids (except for Herpetomonas, which lacks the latter). EC:2.1.1.14 appears to be of bacterial origin, grouping within the Gammaproteobacteria with moderate (74) bootstrap support. While EC:2.1.1.10 also groups near Gammaproteobacteria, BSV is low and this gene cannot be considered a case of HGT given current data.

As seen above, most genes in the de novo methionine synthesis pathway seem to have originated in one or more HGT events. Enzymes from the methionine salvage pathway (Additional file 3), on the other hand, are notably different. Of these, only S-methyl-5-thioribose kinase (EC:2.7.1.100), found in C. acanthocephali and Herpetomonas but not in the endosymbionts and symbiont-harboring trypanosomatids, seems to have originated in a bacterial group (Additional file 13). These two organisms’ enzymes group deep within the Gammaproteobacteria, with BSV of 97.

Enzyme acireductone synthase (EC:3.1.3.77) presents an intriguing case, being the only methionine salvage pathway enzyme absent from the symbiont-harboring trypanosomatid genomes. This enzyme is of eukaryotic origin (not shown), and present in both H. muscarum and C. acanthocephali, but was not found in any other of the regular trypanosomatids available from KEGG. Interestingly, KEGG data for Trypanosoma brucei also shows the two enzymes preceding EC:3.1.3.77 as missing, which raises the question of whether this important pathway is in the process of being lost in trypanosomatids. If that is not the case, and given that all other enzymes from the pathway are present, the Trypanosomatidae must have a different enzyme (or enzymes) to perform the required reactions.

The gene for the enzyme that interconverts glycine and threonine (Figure 3), EC:4.1.2.5, was identified in all symbiont-harboring and regular trypanosomatids (except Herpetomonas), but the evolutionary histories of symbiont-harboring and regular trypanosomatid genes are very different (Additional file 14). The gene found in the regular trypanosomatids Leishmania sp. and C. acanthocephali groups deep within the Firmicutes, most closely Clostridium, with BSV of 63. The symbiont-harboring trypanosomatid genes, on the other hand, cluster as the most basal clade of one of the two large assemblages of eukaryotes present in this phylogeny, although all BSV are low and there is a large group of Bacteria from diverse phyla between the symbiont-harboring trypanosomatids plus a few other eukaryotic groups and the other eukaryotes in this part of the tree. It is therefore difficult to conclude whether the symbiont-harboring trypanosomatid gene is of bacterial or eukaryotic origin.

Tryptophan synthase beta subunit (EC:4.2.1.20), present in the symbiont-harboring trypanosomatids and Herpetomonas, is the last enzyme of the tryptophan biosynthesis pathway, and the only one present in trypanosomatids for this pathway. Its gene groups robustly (BSV of 97) with the Bacteroidetes phylum (Additional file 15). It is also highly similar (around 80% identity and 90% similarity) to the corresponding genes of this phylum, suggesting either a very recent transfer or high sequence conservation. Given that the protein alignment of the orthologs (not shown) presents a maximum patristic distance value of 84.04% and a median of 47.22%, it is therefore likely that the transfer of EC:4.2.1.20 to the Trypanosomatidae is relatively recent.

The arginine and ornithine synthesis pathway has been influenced by HGT events in a few key steps. As discussed above, one of the entry points for the urea cycle is through ornithine synthesized from glutamate. The last step, converting N-acetylornithine to ornithine, can be performed by either EC:3.5.1.14 or EC:3.5.1.16 (Figure 7). We have found that the genes for both enzymes, present in all symbiont-harboring and regular trypanosomatid genomes, originated from HGT events. All gene copies for EC:3.5.1.14 group as one clade with a gammaproteobacterium (BSV of 98), and with Bacteria of different phyla (predominantly Firmicutes) as nearest sister group, although with low BSV (Additional file 16). The few other eukaryotic groups present in the tree are very distant from the trypanosomatid group. The multiple copies of the gene for EC:3.5.1.16 in symbiont-harboring and regular trypanosomatids group together in a monophyletic clade (Additional file 17), which clusters within a large group of mostly Betaproteobacteria with BSV of 80, including the Alcaligenaceae, the family to which the endosymbionts belong. However, it seems highly unlikely that this sequence has been transferred from the endosymbiont genomes to their hosts genomes because the nuclear sequences are firmly removed from the Alcaligenaceae, and many regular trypanosomatids (including Trypanosoma spp.) also present this gene in the same part of the tree.

The only trypanosomatid analyzed which presented ornithine carbamoyl transferase (OCT, EC:2.1.3.3) was Herpetomonas muscarum. Our phylogenetic analysis of this gene indicates that it is of eukaryotic origin (not shown). The symbiont-harboring trypanosomatids utilize the OCT provided by their endosymbionts, and their OCT genes group firmly inside the Alcaligenaceae family, next to Taylorella and Advenella, as expected.

The genes for EC:6.3.4.5 and EC:4.3.2.1 present similar evolutionary patterns: both are absent from endosymbiont genomes and present in all symbiont-harboring and regular trypanosomatid genomes – the only exception being the lack of the latter in Leishmania spp. The Trypanosomatid genes form monophyletic groups in their respective trees, grouping within Firmicutes in both cases (Additional files 18 and 19). BSV is higher (82) in the tree of EC:4.3.2.1 than in that of EC:6.3.4.5 (69). In both cases, support falls for deeper branches in the trees. Although the host genomic sequences are still incomplete and in varying degrees of contiguity, it is interesting to note that the genes for EC:6.3.4.5 and EC:4.3.2.1 are present in tandem in one contig in all symbiont-harboring trypanosomatids (Additional file 1). The flanking genes are eukaryotic: terbinafine resistance locus protein and a multidrug resistance ABC transporter. As seen in the genome browser at TriTrypDB (http://tritrypdb.org), Leishmania spp. have most of these same genes, although in a slightly different order (EC:6.3.4.5 occurring after the two eukaryotic genes instead of between them) and lacking EC:4.3.2.1. L. braziliensis seems to be in the process of additionally losing EC:6.3.4.5, which is annotated as a pseudogene. These phylogenetic and genomic data strongly suggest that EC:4.3.2.1 and EC:6.3.4.5 have been transferred together from a Firmicutes bacterium to the common ancestor of the symbiont-harboring and regular trypanosomatids studied, and that these transferred genes have been or are being lost from Leishmania at least.

The final enzyme in the urea cycle, arginase (EC:3.5.3.1), is present in all symbiont-harboring and regular trypanosomatids examined here. However, the sequence from Herpetomonas presents a partial arginase domain; while the protein sequence length is as expected, the domain match starts only after 70 amino acids. We speculate that this divergence could be responsible for the lack of arginase activity previously seen in Herpetomonas. Differently from most other enzymes in this work, there are different evolutionary histories for the arginase genes: all trypanosomatid genes but that from Herpetomonas cluster together with very high bootstrap support of 98, within Eukaryota (Additional file 20). The sequence from Herpetomonas on the other hand is the sister group (BSV of 79) of a large assemblage of Bacteria from several different phyla, but predominantly Deltaproteobacteria, Firmicutes, Actinobacteria, and Cyanobacteria. It is therefore clear that Herpetomonas must have acquired a different arginase than that present in the other trypanosomatids studied, which possess eukaryotic genes. Furthermore, this gene seems to be undergoing a process of decay, given its lack of significant similarity to the known arginase domain in a significant portion of the protein.

Ornithine cyclodeaminase (EC:4.3.1.12) converts ornithine directly into proline, a non-essential amino acid. In our analyses, we have found that the gene for EC:4.3.1.12 of symbiont-harboring trypanosomatid genomes is very similar to those from Betaproteobacteria of the Alcaligenaceae family, to which the endosymbionts belong. Regular trypanosomatid and endosymbiont genomes do not contain the gene for this enzyme. Accordingly, the phylogeny shows the symbiont-harboring trypanosomatid gene grouping close to several Alcaligenaceae, although the clade is not monophyletic and presents BSV of 63 (Additional file 21). This grouping, together with the gene presence in symbiont-harboring trypanosomatid genomes only, poses the possibility that EC:4.3.1.12 has been transferred from the ancestral endosymbiont to the corresponding host, before the radiation of symbiont-harboring trypanosomatids into the two genera and five species analyzed here.

Some interesting peculiarities of specific genes from a few pathways deserve to be discussed. Interestingly, the gene for branched-chain-amino-acid transaminase (EC:2.6.1.42), the last step in the synthesis of isoleucine, valine, and leucine (Figure 4), was identified in all bacteria of the Alcaligenaceae family present in KEGG, except for the endosymbionts’ closest relatives, Taylorella spp. (parasitic) and Advenella kashmirensis (free-living), which also lack the gene. The question is raised then of whether the common ancestor of Taylorella and the endosymbionts, which are sister groups 25 , had already lost the gene. Another possibility is that independent loses occurred in endosymbionts, Taylorella, and Advenella. Considering that the rest of the pathway is present in these organisms and that the free-living Advenella would need the last gene to complete synthesis of these amino acids, it is reasonable to speculate that their EC:2.6.1.42 is novel or at least very different and thus could not be identified by similarity searches.

As mentioned above, the histidine pathway biosynthesis is performed by the endosymbionts and all enzymes, with the exception of histidinol-phosphate phosphatase (HPP, EC:3.1.3.15), have been identified. This is also the only enzyme of this pathway missing in other Betaproteobacteria available in KEGG. Recently, it was reported that such a gap in the histidine biosynthesis pathway in other organisms was completed by novel HPP families 78 79 . Our searches for the novel C. glutamicum HPP (cg0910, an inositol monophosphatase-like gene) have identified two possible candidate genes in the endosymbionts (BCUE_0333 and BCUE_0385, in C. K. blastocrithidii). As in Corynebacterium, neither of these genes is in the same operon as the known histidine synthesis genes. Given the absence of any other inositol phosphate metabolism genes in the endosymbiont genomes, except for these two IMPases, it is reasonable to hypothesize that at least one of the two aforementioned candidates could be the HPP.

The symbiont-harboring trypanosomatid species genomes sequenced here were: Strigomonas oncopelti TCC290E, S. culicis TCC012E, S. galati TCC219, Angomonas deanei TCC036E, and A. desouzai TCC079E. These symbiont-harboring trypanosomatids harbor, respectively, the symbionts: Candidatus Kinetoplastibacterium oncopeltii, Ca. K. blastocrithidii, Ca. K. galatii, Ca. K. crithidii and Ca. K. desouzaii 24 , and were previously sequenced 25 . In addition, we have also sequenced the genomes of two regular trypanosomatids, i.e. Herpetomonas muscarum TCC001E and Crithidia acanthocephali TCC037E. These organisms are cryopreserved at the Trypanosomatid Culture Collection of the University of São Paulo, TCC-USP. Symbiont-harboring trypanosomatids were grown in Graces’ medium (Gibco). Regular trypanosomatids were grown in LIT media 82 .

Total genomic DNA was extracted by the phenol-chloroform method 83 . We applied kDNA depletion methods to minimize the presence of this type of molecule, as previously described 58 , which result in less than about 5% of remaining kDNA in the sample. After kDNA depletion, about 5 μg of DNA were submitted to each Roche 454 shotgun sequencing run, according to the manufacturer’s protocols. Different genomes have so far been sequenced to different levels of draft quality, with estimated coverages of 15X to 23X (considering a genome of ~30 Mbp). Sequences were assembled using the Newbler assembler version 2.3, provided by Roche. Resulting assemblies are available from GenBank, as detailed in “Availability of Supporting Data” below. Endosymbiont genomes were finished to a closed circle as previously described 25 .

Endosymbiont genes were used as previously published 25 . In an initial scan of the genome, trypanosomatid genes were discovered and mapped to metabolic pathways using ASGARD 84 , employing as reference the UniRef100 85 and the Kyoto Encyclopedia of Genes and Genomes, KEGG 86 databases. The identified segments of DNA were then extracted from the genome and manually curated for completion and proper location of start and stop codons by using the GBrowse genome browser 87 . Putative sequence functions were confirmed by domain searches against NCBI’s Conserved Domain Database 88 . Genes and annotations from other trypanosomatids were used when needed and as available at KEGG. All trypanosomatid genes characterized in this study have been submitted to NCBI’s GenBank and accession numbers are available from Additional file 22. All endosymbiont genes analyzed here have been previously sequenced 25 ; gene identifiers are available from Additional file 23.

Due to the incomplete nature of our trypanosomatid assemblies, a set of criteria were used to avoid including contaminant sequences in our analyses. A gene was accepted as legitimate only when satisfying at least two of the following: genomic context compatible with a trypanosomatid gene (i.e. long stretches of genes in the same orientation in the contig, most neighboring genes similar to genes from other, previously sequenced trypanosomatids); sequencing coverage in the gene similar to, or higher than, that of the gene and genome averages (since contaminants that are difficult to detect will almost always be in small contigs of low coverage); GC percent content consistent with that of the neighboring genes, and of the overall genome; and phylogenetic congruence (i.e. whether genes from more than one trypanosomatid formed monophyletic assemblages). Genomic context and GC content graphs were drawn by GBrowse 87 and graphically edited for better use of space.

For phylogenetic analysis of each enzyme characterized in this work, corresponding putative orthologous genes from all domains of life were collected from the public databases by BLAST search (E-value cutoff of 1e-10, maximum of 10,000 matches accepted) against the full NCBI NR protein database, collecting sequences from as widespread taxonomic groups as possible and keeping one from each species (except for alignments with more than ~1,500 sequences, in which case one organism per genus was kept). Only sequences that were complete and aligned along at least 75% of the length of the query were selected. All analyses were performed at the protein sequence level. Sequences were aligned by Muscle v. 3.8.31 89 . Phylogenetic inferences were performed by the maximum likelihood method, using RAxML v. 7.2.8 90 and employing the WAG amino acid substitution model 91 , with four gamma-distributed substitution rate heterogeneity categories and empirically determined residue frequencies (model PROTGAMMAWAGF). Each alignment was submitted to bootstrap analysis with 100 pseudo-replicates. Trees were initially drawn and formatted using TreeGraph2 92 and Dendroscope 93 , with subsequent cosmetic adjustments performed with the Inkscape vector image editor (http://inkscape.org). Phylogenetic conclusions have been displayed as strong in the summary table for phylogenetic results (Additional file 2) if BSV was 80 or greater, and moderate if BSV was between 50 and 80 – with one exception, EC:2.1.1.37, described in the results. Resulting phylogenetic trees are available from TreeBase (accession number S14564), as detailed in “Availability of Supporting Data” below.