Enzymatic methylation of the C5 atom of uridyl to form (ribo)thymidyl occurs during the metabolism of DNA and RNA in all organisms. The first enzyme found to catalyze this fundamental reaction was thymidylate synthase ThyA that uses methylene tetrahydrofolate as carbon source and reducing agent to form thymidylate, an essential DNA precursor. Early work also indicated that the S-adenosyl-l-methionine-dependent methyltransferase TrmA catalyzes the site-specific formation of 5-methyluridine (m⁵U) in the so-called T-Psi-loop (position 54) of tRNA. Recently, two novel flavoproteins were discovered that catalyze the formation of thymidyl groups using CH₂-H₄folate as the carbon donor. The two thymidylate synthases (ThyA and ThyX) and ribothymidylate synthases (TrmA and TrmFO) show a mutually exclusive distribution, as almost always only one is present in a given organism. Information obtained from genetic, biochemical, structural and bioinformatics studies allows to conclude that, despite their reaction products are ubiquitous, these (ribo)thymidylate forming enzymes are not evolutionarily related. The discovery of these novel enzymes provides an excellent example of the evolutionary convergence and versatility of the DNA/RNA modification machinery resulting from molecular tinkering.

Introduction: Historical Background

Methylation at the C5 position of the uracil ring and thus formation of thymidine (5-methyluridine) derivatives is essential for DNA biosynthesis and RNA maturation. These modification reactions occur hundreds of thousands or even millions of times per cell division. Early pregenomic studies led to the identification in Bacteria and Eukarya (and much later in Archaea) of two distinct enzymes that can perform this chemical reaction. First, the essential thymidylate synthase, encoded by thyA in most bacteria and some archaea and by the TYMS gene in humans, converts 2′-deoxyridine-5′-monophosphate [dUMP] to the DNA precursor 2′-deoxythymidine-5′-monophosphate [dTMP]. This was established in the 1950s through pioneering studies using thymus gland homogenate, demonstrating that dUMP is the immediate substrate of ThyA.¹ Subsequent biochemical studies of thymidylate synthase ThyA were greatly facilitated by the relative abundance of this enzyme in methotrexate-resistant mutants of the bacterium Lactococcus casei.² A ribothymidylate synthase was discovered independently by two groups in Escherichia coli.^3,4 This enzyme, initially designated as RumT,⁵ is called TrmA in bacteria,⁶ Trm2p in eukarya⁷ and TrmU₅₄ in archaea.⁸ TrmA and its orthologs methylate C5 of a uridine that is invariably found at position 54 of the so-called T-Psi loop of almost all fully matured and functional tRNAs of a large number of organisms and organelles belonging to all three domains of life (Fig. 1). In vitro and in vivo experiments demonstrated that methylation of U54 in E. coli tRNAs stabilizes the 3D-conformation of tRNAs and facilitates decoding of mRNA on the ribosome.

Figure 1

Reactions that transform uridine to thymidine and distribution of the enzymes involved in these reactions in the three domains of life. Deoxyuridylate (dUMP) can be methylated de novo to thymidylate (dTMP), an essential component of DNA. Two distinct (more...)

Despite the identical outcome of the methylation reactions catalyzed by these two distinct uridyl-specific enzymes (note that this chemistry does not involve the sugar), elegant biochemical experiments established already five decades ago that ThyA and TrmA enzymes do not use the same reaction mechanism (Fig. 1). Wahba and Friedkin first demonstrated that ThyA proteins catalyze a reductive methylation reaction where the methylene group of methylenetetrahydrofolate (CH₂-H₄folate) is transferred and then subsequently reduced to form a methyl group.^9,10 This was demonstrated using a direct spectrophotometric test that allowed detection of the oxidation of tetrahydrofolate (H₄folate) to dihydrofolate (H₂folate) by Escherichia coli ThyA. Notably, until today this reaction remains the only known example of a reductive methylation reaction in biological systems. In parallel, labeling studies established that the methyl groups in bacterial tRNA originate from methionine,^3,4 which was later explained by the fact that TrmA directly transfers a methyl group from S-adenosyl-l-methionine (SAM) to C5 of U54 in tRNA.^4,11 These experiments therefore established that synthesis of 5-methyl uracil (thymidylate or dTMP) at the DNA-precursor and RNA polymer level is catalyzed not only by different enzymes, but occurs by radically different mechanisms. This conclusion was subsequently confirmed by detailed biochemical and structural characterizations of ThyA (reviewed in ref. 12) and TrmA proteins (reviewed in ref. 13).

As far as the other SAM-dependent ribothymidine synthases are concerned, two additional enzymes acting posttranscriptionally have been identified recently in E. coli: the first one, designated as RumA (now renamed RlmD), contains a redox active FeS group and catalyzes the site-specific formation of m⁵U at position 1939 in 23s rRNA,^14,15 while the second methyltransferase (RumB, now renamed RlmC), which like TrmA lacks the FeS group, catalyzes the site-specific formation of m⁵U at position 747 in the same 23s rRNA¹⁶ (Fig. 2). These two new SAM-dependent methyltransferases are homologous to E. coli TrmA and belong to the same cluster of orthologous genes COG2265.¹⁷ Comparative genomic and phylogenetic analyses of available orthologs of COG2265, as well as of COG1206 (TrmFO—410 genomes analyzed) can be found in Supplementary Materials in references 8 and 18.

Figure 2

Localisation of m⁵U in ribosomal and transfer RNAs of E. coli. Panel A) shows a schematic representation of the secondary structure of E. coli 23S rRNA. The conventional numbering of domains is indicated with roman numerals. Panel B) corresponds to an (more...)

For several decades, the canonical enzymes ThyA and TrmA (and its paralogs) were considered the only enzymes capable of formation of DNA precursor deoxyribothymidylate or ribothymidylate within polymer RNAs. However, the presence of alternative (ribo)thymidylate-forming enzymes had been suspected already prior to the genomic era. In the case of the DNA metabolism, genetic studies identified a gene of unknown function that was capable of complementing an unmapped mutation resulting in thymidine auxotrophy in Dictyostelium discoideum. As the complementing gene identified in this genetic screen was not homologous to thyA, it was concluded that it does not encode thymidylate synthase.¹⁹ Indeed, a distinct new type of deoxythymidylate synthase, a flavoprotein dubbed ThyX that uses H₄folate and also FADH₂, was later discovered in bacteria²⁰ and archaea,²¹ with only one eukaryotic representative to date (Fig. 1, box at left, see also below). In the case of the RNA metabolism, early labelling studies established that ribothymidine in the tRNAs of some Gram-positive bacteria (including Enterococcus faecalis and Bacillus subtilis) does not derive from the SAM pool as in E. coli, but rather from the folate pool.^22-24 The first indication for this was the unexpected observation that bulk tRNAs isolated from folate-deprived E. faecalis cells lacked m⁵U54 in their T-Psi loop²⁵ and that in B. subtilis and Micrococcus lysodeikticus, trimethoprim, a specific inhibitor of bacterial dihydrofolate reductase, inhibits formation of m⁵U54 in tRNA in vivo.²⁶ These studies led to the characterization of a new type of tRNA:m⁵U forming enzyme that uses CH₂-H₄folate as donor of a methylene group and, like ThyX, FADH₂ for the reduction of the methylene group²⁷ (or possibly another reducing agent present in cell-free extracts—Figure 1, right part box). Only recently, the gene coding for this enzyme in B. subtilis was unambiguously identified²⁸ and the corresponding recombinant flavoprotein, designated TrmFO (FO for folate), was purified to near homogeneity.¹⁸ From inspection of all genomes sequenced so far, homologs of the B. subtilis trmFO gene seem to be restricted to the bacterial domain of life (Fig. 1).

Below we summarize how bioinformatics, biochemical and structural studies led to the discovery and identification of these two new families of (deoxy)thymidylate synthases (dubbed ThyX) and a tRNA:U54 methyltransferase (dubbed TrmFO).

Folate-Dependent Thymidylate Synthase of the DNA Metabolism

Thymidylate (dTMP) is a key metabolite required for the accurate replication of DNA genomes in all cellular organisms. De novo synthesis of dTMP requires thymidylate synthase that methylates dUMP. The presence of a gene homologous to D. discoideum thy1 in about 30% of microbial genomes that lack the structural gene for ThyA raised the striking possibility that two structurally and mechanistically distinct classes of thymidylate synthases exist.²⁰ This intriguing observation led to the discovery and characterization of a novel, flavin-dependent thymidylate synthase ThyX (dubbed this way to avoid confusion with the Thy1 cell surface antigen).

Thymidylate Synthase ThyX: Structural Information and Key Amino Acid Residues

Intriguingly, around the same time as ThyX was discovered, Kuhn and coworkers at the Joint Center for Structural Genomics in La Jolla determined the X-ray structure of a protein of unknown function from the thermophilic bacterium Thermotoga maritima (PDB code 1kq4). This protein turned out to be ThyX showing a completely different structure and cofactor requirements than the canonical thymidylate synthase ThyA (see Fig. 3). ThyA enzymes are functional homodimers,²⁹ while ThyX proteins studied to date are homotetrameric.^30,31 In the ThyX structure, the four subunits extensively interact with each other, as indicated by the relatively large surface area of approximately 2000 Ų (per monomer) estimated for the ThyX tetrameric structure in Thermotoga maritime.³² The riboflavin moiety of each FAD molecule points toward the surface of the protein and a novel helix-loop-strand FAD binding motif, unique of ThyX enzymes, has been identified.³² Interestingly, three monomers participate in the binding of each of the four FAD molecules bound to the homotetramer, indicating that FAD functions as cofactor and not as substrate, during ThyX catalysis. However, in some cases, it was observed that ThyX activity is substantially increased by the addition of FAD to the reaction mixtures,³³ suggesting that the relatively loose binding of FAD can limit turnover in vitro.

Figure 3

Mechanistic and structural differences between ThyA and ThyX proteins. The reactions catalyzed by ThyA and ThyX proteins are shown. Panel B corresponds to the ribbon representation of the human ThyA dimer (PDB ID: 1hvy; complexed with dUMP and Tomudex, (more...)

Both ThyA and ThyX enzymes catalyze the methylation of deoxyuridine 5′-monophosphate (dUMP) to deoxythymidine 5′-monophosphate (dTMP). However, the amino acid residues essential for catalysis in ThyA are absent in ThyX proteins and no sequence or structural similarity exists between the two thymidylate synthases. Moreover, differently from ThyX (see below), in the ThyA-catalyzed reaction methylenetetrahydrofolate (CH₂-H₄folate) provides a methylene-group and functions as genuine reductant (Fig. 3). Also, the nature and role in catalysis of the key conserved amino acid residues in ThyA are different (reviewed in refs. 12, 34). In brief, it appears that only a few amino acids are catalytically essential in ThyA, including a conserved cysteine residue (C146 in E. coli numbering) in the active site located at the interface of the ThyA homodimer. Interestingly, a single change of a conserved Asn at position 177 to Asp in E. coli ThyA or the equivalent conserved Asn 229 in Lactobacillus casei, changes the specificity of these ThyA proteins from dUMP to dCMP.^35,36 When this mutation was coupled with a mutation His199Ala, also in the active site of L. casei ThyA, the enzyme has lost almost all its capability to methylate the C5 atom of dUMP and selectively methylated dCMP instead.³⁷ Finally, the enzymatic step corresponding to the reduction of the methylene intermediate by hydride transfer from H₄folate, is determining the rate of the reaction.³⁸ In the novel thymidylate synthase ThyX, two strictly conserved residues (histidine and serine) are essential for catalysis, the latter participating also in dUMP binding.^20,30 Surprisingly, although the substrate binding sites of ThyX proteins may be formed by two subunits,³² the active site of tetrameric ThyX proteins is located at the interface of three monomers.³⁹ Obviously, this unexpected and rare situation, together with the observed structural plasticity of ThyX proteins, implies reconfiguration of the active sites during catalysis.^30,40

Enzymatic Mechanism of ThyX

A major difference between two thymidylate synthases is that the ThyX proteins rely on an FAD/NADPH couple as reductant to form a methyl group (Fig. 3). The role of FAD in mediating hydride transfer from NADPH during ThyX catalysis has been directly observed using steady-state and single turnover kinetics,^21,33,41 explaining why ThyX proteins produce tetrahydrofolate as reaction product and not dihydrofolate as ThyA.⁴² This key difference in ThyA and ThyX catalysis has important implications for the bacterial folate metabolism that have been discussed elsewhere.^43,44

Steady-state kinetic analyses of ThyA have revealed that this enzyme uses a sequential reaction mechanism where the two substrates, dUMP and CH₂-H₄folate, are bound at the same time to the active site of the enzyme. This intermediate has been trapped using F-dUMP that forms a covalent reaction intermediate between the enzyme, the nucleotide analog F-dUMP and CH₂-H₄folate.⁴⁵ The nucleophile in ThyA proteins is the thiol group of a conserved cysteine residue (C146 in E. coli numbering) that activates the uridine group of dUMP by cross-linking to the C6 position of the uridine ring. This results in the formation of an enolate ion at the C5 atom that is consequently capable of attacking CH₂-H₄folate.¹²

Details of the reaction catalyzed by ThyX are still under debate, in part because a covalent intermediate with F-dUMP has not been observed directly. This presumably reflects the fact that, different from ThyA proteins, ThyX enzymes use the serine hydroxyl group as nucleophile, which in turn certainly influences the reaction mechanism. Indeed, in ThyA the catalytic cysteine thiol (or thiolate anion) alone is quite nucleophilic and likely to attack dUMP, while the less nucleophilic serine hydroxyl of ThyX most likely requires activation by other residues. In the case of serine proteases, the activation of a catalytic serine is achieved by histidine and aspartate residues.⁴⁶ It is worth mentioning that chemical modification and mass spectroscopy experiments allowed identification of an essential histidine residue of PBCV-1 ThyX (His53), located in the vicinity of the redox active N5 atom of the FAD coenzyme ring system.³⁰ Steady-state kinetic studies support a reaction mechanism where ThyX catalysis proceeds via the formation of distinct ternary complexes between the enzyme, NADPH or CH₂-H₄folate and dUMP without formation of a methyl-enzyme intermediate (Fig. 4). In this proposed reaction scheme, dUMP plays a key role in regulating the substrate traffic in a single active site that binds all three ThyX substrates. This notion is also supported by the fact that the structures of noncovalent dUMP-ThyX complexes have been resolved for T. maritima and Mycobacterium tuberculosis.^32,47 Nevertheless the formation of a methyl-enzyme intermediate has been proposed for Chlamydia spp. ThyX proteins.⁴²

Figure 4

Cleland plot for the proposed catalytic mechanism of ThyX proteins. Note that the order in which dUMP and NADPH bind and products are released from the active site has not yet been investigated in detail. The figure has been reproduced with permission (more...)

An additional discrepancy concerning the ThyX reaction mechanism has been raised by the observation that in contrast to practically all ThyX proteins studied where the K_{m, NADPH} is relatively low, in T. maritima it is very high, thus raising the possibility that NADPH is not necessarily the physiological reductant of all ThyX proteins and that other reductants may be used for reduction of methylene within the ThyX protein family.⁴⁸ Whereas the reduction of the methylene intermediate by hydride transfer from H₄folate is considered the rate limiting step of ThyA catalysis,³⁸ in the case of ThyX catalysis either NADPH oxidation and/or FAD reduction or associated conformational changes determine the final rate of the reaction.³³ The observation that ThyX-bound FAD exchanges protons with the solvent prior to the reduction of an intermediate methylene also suggests that redox chemistry may limit ThyX turnover.⁴⁸ Obviously, the two thymidylate synthases ThyA and ThyX use distinct strategies to overcome the rate-limiting step of the reduction reaction leading to the transfer of a methylene group to dUMP.

Catalytic Differences between ThyA and ThyX Influence Genome Evolution

Despite their importance in DNA metabolism, both ThyA and ThyX are relatively inefficient catalysts. Studies from several different laboratories with various ThyA and ThyX proteins have indicated that ThyA enzymes are at least 10 times more efficient than the most catalytically active ThyX enzymes examined so far.⁴⁹ A priori, this could be explained by assay conditions for ThyX proteins that are perhaps not yet optimal. However, using a combination of genetic and biochemical studies, we have recently investigated the possibility that the differences in the catalytic efficiencies of both thymidylate synthases are physiologically relevant and possibly have influenced prokaryotic genome evolution. In support of this idea, we have demonstrated that the DNA replication speed in bacteria and archaea that contain the low-activity ThyX enzyme is up to 10-fold decreased compared with species that contain the catalytically more efficient ThyA.⁴⁹ Moreover, our statistical studies of more than 400 genomes indicated that ThyA proteins are preferred for the replication of large genomes, thus providing further evidence that the thymidylate metabolism is limiting expansion of prokaryotic genomes. Since both ThyX and ThyA enzymes participate in frequent reciprocal gene replacement events,²⁰ these observations also indicate that the bacterial metabolism continues to modulate the size and composition of prokaryotic genomes. Thus, the increased kinetic efficiency of thymidylate synthesis may have contributed to the extension of the prokaryotic evolutionary potential (Fig. 5A,B).

Figure 5

Size distribution and nucleotide counts of thyA- and thyX-containing genomes. A) Distributions of genome size, GC% and growth rate are shown. In each case, thyX and thyA containing organisms were analyzed individually. B) A scatter plot demonstrating (more...)

Folate-Dependent Ribothymidylate Synthase of the RNA Metabolism

Ribothymidylate (5-methyluridine monophosphate, rT) is a ubiquitous modified nucleoside that is invariably found at position 54 (m⁵U54) of the so-called T-Psi loop of most tRNAs. As explained above, early labeling and inhibitory studies had indicated that in addition to the canonical SAM-dependent tRNA:m⁵U methyltransferase (TrmA), an alternative tRNA ribothymidylate synthase exists in some bacteria that use CH₂-H₄folate as one-carbon donor and reduced flavin adenine dinucleotide (FADH₂).²⁷ The observation that both flavoproteins, ThyX of the DNA metabolism and TrmFO of the RNA metabolism, use CH₂-H₄folate as carbon donor and depend on an FAD coenzyme raised the possibility that these two enzymes are derived from the common protein precursor. To detect this possible relationship, sequence and structural information about ThyX and TrmFO was needed.

tRNA Substrate Recognition Problem and Catalysis

Whereas thymidylate synthases methylate the relatively small free nucleotide dUMP with a molecular mass of 352 Da (in the case of the disodium salt), the rather complex tRNA substrate molecule for TrmFO is 70 to 95 nucleotides long, folded in a cloverleaf secondary structure, which in turn is folded into a compact (but still flexible) L-shaped 3D architecture. This is schematized in Figure 6A for a Class I tRNA (without a long extra-arm). All tRNAs have a molecular mass of 25,000 (+/−3,000) Da and are likely the comparable relative size to the TrmFO enzymes constituted by two identical subunits (2 × 58,000 Da for B. subtilis enzyme). It should be emphasized that the cellular tRNAs likely exist in several alternative conformations during tRNA maturation (see chapter by Hayrapetyan et al in this volume), thus making the substrate recognition difficult. The uracil methylation by TrmFO is even further complicated by the fact that in fully mature and functional tRNAs, ribothymine-54 is implicated in internal base pairing with A58 through a reverse Hoogsteen base pair that stacks on the conserved Watson-Crick G53-C62 pairs. This induces the two pyrimidines at positions 59 and 60 to loop out, thus leaving these tRNAs with a mini-loop of only 3 bases instead of 7 as in the unstructured tRNA molecule. Also, conserved bases rT55 and C56 are interacting with two conserved guanosine bases at positions 18 and 19 m thus forming the characteristic ubiquitous 3D-core of the tRNA molecules (Fig. 6A). Consequently, in the fully mature tRNA, ribothymine-54 (or unmodified U54) is fairly rigid and inaccessible.

Figure 6

Schematic representation of tRNA indicating localization of uridines that are modified by TrmFO or its homolog GidA. Panel A) corresponds to schematic representations of the L-shaped 3-D structure of a tRNA molecule of class I (lacking the long extra (more...)

The recent structure of E. coli monomeric TrmA in a covalent complex with a 19-mer oligonucleotide corresponding to the T-Psi arm solved at 2.43 Å resolution⁵² reveals that the protein-RNA interface includes only six nucleotides from the T-Psi-loop and two base pairs of the T-stem. When comparing the T-Psi-loop conformation as bound to the crystal structure of the TrmA enzyme with the same T-Psi-loop within the tRNA crystal structure (Fig. 6B), it appears that methylation of C5 in U54 by TrmA requires refolding of the T-Psi-loop into a conformation different from the free state. Such refolding of the U54 target-containing loop into this new base-stacked conformation displays a triplet of nucleotides (U54-U55-C56) to the active site cavity of the enzyme. If this model applies to the binding of the TrmFO enzyme to the same U54 within an ‘exposed’ T-Psi-loop (not within a T-Psi-loop engaged in interaction with the D-loop as in the canonical L-shaped architecture, Fig. 6A), then the binding of the ribothymidine synthase-tRNA complex should imply at least two steps: one for a rapid association of the enzyme with the T-Psi-loop within the tRNA molecule and a second one for a probably slow ‘accommodation/looping out’ of a triplet of nucleotides of the T-Psi-loop (encompassing the target U54) within the active site cavity of the enzyme. The possibility also exists that efficient and appropriate binding process depends on the dynamic properties of the enzyme itself (examples for the other enzyme/tRNA couples can be found in the chapter by Byrne et al in this volume). To date, nothing is known about how the dimeric form of TrmFO enzymes accesses and recognizes the U54 target within the tRNA. The importance of dynamic properties of the enzyme and of the cofactors (FAD/H₄folate, see below) on the binding process needs to be established.

Conclusions and Perspectives

The methylation enzymes we have described in this chapter catalyze chemical reactions with an identical and relatively simple product. Nevertheless, the reaction mechanisms of these methyltransferases per se can be quite complex and vary greatly from one enzyme to another (Fig. 1). The available data suggest that the use of a redox couple, such as CH₂-H₄folate and FAD/NADPH, to form a methyl group (as exemplified by ThyX and TrmFO) is by far less efficient than methylation mechanisms used by ThyA or TrmA proteins. We discussed above that both, the use of a cysteine thiol as nucleophile and the use of H₄folate as reductant, which is not released from the active site, significantly contribute to the increased catalytic efficiency of the ThyA proteins in comparison to the ThyX enzymes. Notably, TrmA homologs catalyze the methylation of U54 using S-adenosyl-l-methionine (S-AdoMet or SAM) as methyl donor. SAM is created by adenylation of methionine resulting in the activation of a methyl group via the positive charge on the adjacent sulfur atom. Consequently, the methyl transfer potential of SAM is much higher than that of CH₂-H₄folate. This may explain why SAM is the preferred methyl donor in the majority of biosynthetic methylation reactions, such as DNA and RNA methylation, but also in methylation of proteins and many cellular metabolites.^57,58 At present, the identity of the nucleophilic residue of TrmFO proteins is not known, raising the possibility that TrmFO proteins use yet a different variant of the methylation reaction than of ThyX. More experimental work is required to address this point as well as to understand whether TrmFO and TrmA proteins use different strategies to access U54 that is buried in the 3D-core of the tRNA molecules.

The discoveries of ThyX and TrmFO are fine examples of how comparative genomics allowed the identification of previously uncharacterized DNA and RNA modification enzymes. These discoveries also allowed the inhibitor screening projects on ThyX, which due to its essential role in many pathogenic microrganisms and its sequence and structural differences to ThyA can be used as new type of antimicrobial target and to identify new compounds with anti-microbial potential (see for instance ref. 59). Non-orthologous gene replacement events are perhaps more common than previously thought and may have played a key role in the evolution of metabolic pathways.⁶⁰ Apparently the sporadic phylogenetic distribution of the mutually exclusive ThyA and ThyX proteins as well as of TrmFO and TrmA raised the possibility that mechanistic constraints of the four proteins might have influenced their evolutionary histories and the genome composition in current day organisms.⁴⁹ One difference between the mutually exclusive enzyme couples, ThyA/ThyX and TrmA/TrmFO, is that the phylogenetic distribution of the latter follows more closely the universal tree of life than the former which suggests that TrmFO and TrmA have participated less frequently in lateral gene transfer events than ThyX and ThyA proteins.

For TrmFO, improved biochemical assays are clearly required to investigate details of the methylation reaction, especially the nature of the real in vivo reductant molecules. This will eventually allow a comparison with the reactions catalyzed by ThyX (see above) and TrmA. Also, ribothymidine formation within a polymeric RNA obviously presents additional challenging problems compared to ribothymidine formation in the free deoxynucleotide dUMP. More detailed mechanistic and structural studies in the future will substantially increase our general understanding of the extraordinary versatility of DNA and RNA modification systems. It is of note that the evolution of DNA/RNA methyltransferases is very complex, as the all three enzymes (ThyA, ThyX and TrmFO) we have described in this chapter are almost certainly evolutionary unrelated (i.e., they did not emerge from the same ancestral precursor protein), while, in contrast, TrmFO and GidA proteins are clearly homologous and thus probably emerged from a common ancestor.

In conclusion, the four enzyme families described above represent at least three (possibly four) different chemical strategies of how evolution has used chemical building blocks available in the cell to achieve the methylation (more complex modification in the case of GidA) reaction at the C5 atom of uracil ring. As the cells have clearly done this repeatedly, not like an engineer designing from scratch, but rather without knowing beforehand what could have been the most efficient outcome, this is a good example of evolutionary tinkering (bricolage moléculaire)⁶¹ with genes and their products.

Acknowledgements

Work on ThyX proteins has been financed by Agence Nationale de la Recherche, CNRS and Ministère de la Recherche (to H.M. and U. L.). H.M. thanks the INSERM AVENIR program and the Fondation Bettencourt for financial support. H.G. (emeritus scientist) thanks Pr. Jean-Pierre Rousset from the Université d'Orsay for providing facilities to continue enjoying to work in a laboratory. Particular thanks to all persons who have been involved in the work on ThyX and TrmFO proteins. We thank Steve Douthwaite (University Odense, Denmark) and O. Namy for helping us in elaborating Figure 2 and 6, respectively. We also thank A.D. Hanson, J. Urbonavicius and G.R. Bjork for helpful comments on manuscript.

References

1.

Humphreys GK, Greenberg DM. Studies on the conversion of deoxyuridylic acid to thymidylic acid by a soluble extract from rat thymus. Arch Biochem Biophys. 1958;78(2):275–287. [PubMed: 13618009]

2.

Dunlap RB, Harding NG, Huennekens FM. Thymidylate synthetase from amethopterin-resistant Lactobacillus casei. Biochemistry. 1971;10(1):88–97. [PubMed: 5538615]

3.

Fleissner E, Borek E. A new enzyme of RNA synthesis: RNA methylase. Proc Natl Acad Sci USA. 1962;48:1199–1203. [PMC free article: PMC220932] [PubMed: 13893516]

4.

Svensson I, Boman HG, Eriksson KG. et al. Studies on microbial rna. I. Transfer of methyl groups from methionine to soluble rna from Escherichia coli. J Mol Biol. 1963;7:254–271. [PubMed: 14065310]

5.

Santi DV, Hardy LW. Catalytic mechanism and inhibition of tRNA (uracil-5-)methyltransferase: evidence for covalent catalysis. Biochemistry. 1987;26(26):8599–8606. [PubMed: 3327525]

6.

Bjork GR. Transductional mapping of gene trmA responsible for the production of 5-methyluridine in transfer ribonucleic acid of Escherichia coli. J Bacteriol. 1975;124(1):92–98. [PMC free article: PMC235869] [PubMed: 1100617]

7.

Nordlund ME, Johansson JO, von Pawel-Rammingen U. et al. Identification of the TRM2 gene encoding the tRNA(m5U54)methyltransferase of Saccharomyces cerevisiae. RNA. 2000;6(6):844–860. [PMC free article: PMC1369962] [PubMed: 10864043]

8.

Urbonavicius J, Auxilien S, Walbott H. et al. Acquisition of a bacterial RumA-type tRNA(uracil-54, C5)-methyltransferase by Archaea through an ancient horizontal gene transfer. Mol Microbiol. 2008;67(2):323–335. [PubMed: 18069966]

9.

Wahba AJ, Friedkin M. Direct spectrophotometric evidence for the oxidation of tetrahydrofolate during the enzymatic synthesis of thymidylate. J Biol Chem. 1961;236:PC11–12. [PubMed: 13782537]

10.

Wahba AJ, Friedkin M. The enzymatic synthesis of thymidylate. I. Early steps in the purification of thymidylate synthetase of Escherichia coli. J Biol Chem. 1962;237:3794–3801. [PubMed: 13998281]

11.

Mandel LR, Borek E. The biosynthesis of methylated bases in ribonucleic acid. Biochemistry. 1963;2:555–560. [PubMed: 14069547]

12.

Carreras CW, Santi DV. The catalytic mechanism and structure of thymidylate synthase. Annu Rev Biochem. 1995;64:721–762. [PubMed: 7574499]

13.

Garcia GA, Goodenough-Lashua DM. 1998. Mechanisms of RNA-modifying and editing enzymes. In: Grosjean H, Benne R, eds. Modification and Editing of RNA. Washington DC: ASM Press; pp. 135–168.

14.

Agarwalla S, Kealey JT, Santi DV. et al. Characterization of the 23 S ribosomal RNA m5U1939 methyltransferase from Escherichia coli. J Biol Chem. 2002;277(11):8835–8840. [PubMed: 11779873]

15.

Agarwalla S, Stroud RM, Gaffney BJ. Redox reactions of the iron-sulfur cluster in a ribosomal RNA methyltransferase, RumA: optical and EPR studies. J Biol Chem. 2004;279(33):34123–34129. [PMC free article: PMC1237038] [PubMed: 15181002]

16.

Madsen CT, Mengel-Jorgensen J, Kirpekar F. et al. Identifying the methyltransferases for m(5)U747 and m(5)U1939 in 23S rRNA using MALDI mass spectrometry. Nucleic Acids Res. 2003;31(16):4738–4746. [PMC free article: PMC169892] [PubMed: 12907714]

17.

Tatusov RL, Fedorova ND, Jackson JD. et al. The COG database: an updated version includes eukaryotes. BMC Bioinformatics. 2003;4:41. [PMC free article: PMC222959] [PubMed: 12969510]

18.

Urbonavicius J, Brochier-Armanet C, Skouloubris S. et al. In vitro detection of the enzymatic activity of folate-dependent tRNA (Uracil-54,-C5)-methyltransferase: evolutionary implications. Methods Enzymol. 2007;425:103–119. [PubMed: 17673080]

19.

Dynes JL, Firtel RA. Molecular complementation of a genetic marker in dictyostelium using a genomic DNA library. Proc Natl Acad Sci USA. 1989;86(20):7966–7970. [PMC free article: PMC298193] [PubMed: 2813371]

20.

Myllykallio H, Lipowski G, Leduc D. et al. An alternative flavin-dependent mechanism for thymidylate synthesis. Science. 2002;297(5578):105–107. [PubMed: 12029065]

21.

Giladi M, Bitan-Banin G, Mevarech M. et al. Genetic evidence for a novel thymidylate synthase in the halophilic archaeon Halobacterium salinarum and in Campylobacter jejuni. FEMS Microbiol Lett. 2002;216(1):105–109. [PubMed: 12423760]

22.

Delk AS, Rabinowitz JC. Biosynthesis of ribosylthymine in the transfer RNA of Streptococcus faecalis: a folate-dependent methylation not involving S-adenosylmethionine. Proc Natl Acad Sci USA. 1975;72(2):528–530. [PMC free article: PMC432345] [PubMed: 804695]

23.

Delk AS, Romeo JM, Nagle DP Jr. et al. Biosynthesis of ribothymidine in the transfer RNA of Streptococcus faecalis and Bacillus subtilis. A methylation of RNA involving 5,10-methylenetetrahydrofolate. J Biol Chem. 1976;251(23):7649–7656. [PubMed: 826533]

24.

Romeo JM, Delk AS, Rabinowitz JC. The occurrence of a transmethylation reaction not involving S-adenosylmethionine in the formation of ribothymidine in Bacillus subtilis transfer-RNA. Biochem Biophys Res Commun. 1974;61(4):1256–1261. [PubMed: 4218103]

25.

Samuel CE, Rabinowitz JC. Initiation of protein synthesis by folate-sufficient and folate-deficient Streptococcus faecalis R: partial purification and properties of methionyl-transfer ribonucleic acid synthetase and methionyl-transfer ribonucleic acid formyltransferase. J Bacteriol. 1974;118(1):21–31. [PMC free article: PMC246635] [PubMed: 4206871]

26.

Arnold HH, Kersten H. Inhibition of the tetrahydrofolate-dependent biosynthesis of ribothymidine in tRNAs of B. subtilis and M. lysodeikticus by trimethoprim. FEBS Lett. 1975;53(2):258–261. [PubMed: 806472]

27.

Delk AS, Nagle DP Jr, Rabinowitz JC. Methylenetetrahydrofolate-dependent biosynthesis of ribothymidine in transfer RNA of Streptococcus faecalis. Evidence for reduction of the 1-carbon unit by FADH2. J Biol Chem. 1980;255(10):4387–4390. [PubMed: 6768721]

28.

Urbonavicius J, Skouloubris S, Myllykallio H. et al. Identification of a novel gene encoding a flavin-dependent tRNA:m5U methyltransferase in bacteria—evolutionary implications. Nucleic Acids Res. 2005;33(13):3955–3964. [PMC free article: PMC1178002] [PubMed: 16027442]

29.

Hardy LW, Finer-Moore JS, Montfort WR. et al. Atomic structure of thymidylate synthase: target for rational drug design. Science. 1987;235(4787):448–455. [PubMed: 3099389]

30.

Graziani S, Bernauer J, Skouloubris S. et al. Catalytic mechanism and structure of viral flavin-dependent thymidylate synthase ThyX. J Biol Chem. 2006;281(33):24048–24057. [PubMed: 16707489]

31.

Kuhn P, Lesley SA, Mathews II. et al. Crystal structure of thy1, a thymidylate synthase complementing protein from Thermotoga maritima at 2.25 A resolution. Proteins. 2002;49(1):142–145. [PubMed: 12211025]

32.

Mathews II, Deacon AM, Canaves JM. et al. Functional analysis of substrate and cofactor complex structures of a thymidylate synthase-complementing protein. Structure. 2003;11(6):677–690. [PubMed: 12791256]

33.

Graziani S, Xia Y, Gurnon JR. et al. Functional analysis of FAD-dependent thymidylate synthase ThyX from Paramecium bursaria Chlorella virus-1. J Biol Chem. 2004;279(52):54340–54347. [PubMed: 15471872]

34.

Finer-Moore JS, Santi DV, Stroud RM. Lessons and conclusions from dissecting the mechanism of a bisubstrate enzyme: thymidylate synthase mutagenesis, function and structure. Biochemistry. 2003;42(2):248–256. [PubMed: 12525151]

35.

Hardy LW, Nalivaika E. Asn177 in Escherichia coli thymidylate synthase is a major determinant of pyrimidine specificity. Proc Natl Acad Sci USA. 1992;89(20):9725–9729. [PMC free article: PMC50205] [PubMed: 1409689]

36.

Liu L, Santi DV. Mutation of asparagine 229 to aspartate in thymidylate synthase converts the enzyme to a deoxycytidylate methylase. Biochemistry. 1992;31(22):5100–5104. [PubMed: 1606134]

37.

Agarwalla S, LaPorte S, Liu L. et al. A novel dCMP methylase by engineering thymidylate synthase. Biochemistry. 1997;36(50):15909–15917. [PubMed: 9398324]

38.

Kanaan N, Marti S, Moliner V. et al. A quantum mechanics/molecular mechanics study of the catalytic mechanism of the thymidylate synthase. Biochemistry. 2007;46(12):3704–3713. [PubMed: 17328531]

39.

Leduc D, Graziani S, Lipowski G. et al. Functional evidence for active site location of tetrameric thymidylate synthase X at the interphase of three monomers. Proc Natl Acad Sci USA. 2004;101(19):7252–7257. [PMC free article: PMC409905] [PubMed: 15123820]

40.

Ulmer JE, Boum Y, Thouvenel CD. et al. Functional analysis of the Mycobacterium tuberculosis FAD-dependent thymidylate synthase, ThyX reveals new amino acid residues contributing to an extended ThyX motif. J Bacteriol. 2008;190(6):2056–2064. [PMC free article: PMC2258874] [PubMed: 18192395]

41.

Gattis SG, Palfey BA. Direct observation of the participation of flavin in product formation by thyX-encoded thymidylate synthase. J Am Chem Soc. 2005;127(3):832–833. [PubMed: 15656610]

42.

Griffin J, Roshick C, Iliffe-Lee E. et al. Catalytic mechanism of Chlamydia trachomatis flavin-dependent thymidylate synthase. J Biol Chem. 2005;280(7):5456–5467. [PubMed: 15591067]

43.

de Crecy-Lagard V, El Yacoubi B, de la Garza RD. et al. Comparative genomics of bacterial and plant folate synthesis and salvage: predictions and validations. BMC Genomics. 2007;8:245. [PMC free article: PMC1971073] [PubMed: 17645794]

44.

Leduc D, Escartin F, Nijhout HF. et al. Flavin-dependent thymidylate synthase ThyX activity: implications for the folate cycle in bacteria. J Bacteriol. 2007;189(23):8537–8545. [PMC free article: PMC2168944] [PubMed: 17890305]

45.

Santi DV, McHenry CS. 5-Fluoro-2′-deoxyuridylate: covalent complex with thymidylate synthetase. Proc Natl Acad Sci USA. 1972;69(7):1855–1857. [PMC free article: PMC426818] [PubMed: 4505665]

46.

Dodson G, Wlodawer A. Catalytic triads and their relatives. Trends Biochem Sci. 1998;23(9):347–352. [PubMed: 9787641]

47.

Sampathkumar P, Turley S, Ulmer JE. et al. Structure of the Mycobacterium tuberculosis flavin dependent thymidylate synthase (MtbThyX) at 2.0 A Resolution. Journal of Molecular Biology. 2005;352(5):1091. [PubMed: 16139296]

48.

Agrawal N, Lesley SA, Kuhn P. et al. Mechanistic studies of a flavin-dependent thymidylate synthase. Biochemistry. 2004;43(32):10295–10301. [PubMed: 15301527]

49.

Escartin F, Skouloubris S, Liebl U. et al. Flavin-dependent thymidylate synthase X limits chromosomal DNA replication. Proc Natl Acad Sci USA. 2008;105(29):9948–9952. [PMC free article: PMC2481370] [PubMed: 18621705]

50.

Dym O, Eisenberg D. Sequence-structure analysis of FAD-containing proteins. Protein Sci. 2001;10(9):1712–1728. [PMC free article: PMC2253189] [PubMed: 11514662]

51.

Delk AS, Nagle DP Jr, Rabinowitz JC. 1979. Purification of methylenetetrahydrofolate-dependent methyltransferase catalysizing biosynthesis of ribothymidine in transfer RNA of Streptococcus faecalis. In: Kisliuk RL, Brown GM, eds. Chemistry and Biology of Pteridines. New York: Elsevier/North Holland Publishing; pp. 389–394.

52.

Alian A, Lee TT, Griner SL. et al. Structure of a TrmA-RNA complex: A consensus RNA fold contributes to substrate selectivity and catalysis in m5U methyltransferases. Proc Natl Acad Sci USA. 2008;105(19):6876–6881. [PMC free article: PMC2383949] [PubMed: 18451029]

53.

Cicmil N. Crystallization and preliminary X-ray crystallographic characterization of TrmFO a folate-dependent tRNA methyltransferase from Thermotoga maritima. Acta Crystallogr Sect F Struct Biol Cryst Commun. 2008;64(Pt 3):193–195. [PMC free article: PMC2374164] [PubMed: 18323606]

54.

Bregeon D, Colot V, Radman M. et al. Translational misreading: a tRNA modification counteracts a +2 ribosomal frameshift. Genes Dev. 2001;15(17):2295–2306. [PMC free article: PMC312767] [PubMed: 11544186]

55.

Meyer S, Scrima A, Versees W. et al. Crystal structures of the conserved tRNA-modifying enzyme GidA: implications for its interaction with MnmE and substrate. J Mol Biol. 2008;380(3):532–547. [PubMed: 18565343]

56.

Scrima A, Vetter IR, Armengod ME. et al. The structure of the TrmE GTP-binding protein and its implications for tRNA modification. EMBO J. 2005;24(1):23–33. [PMC free article: PMC544919] [PubMed: 15616586]

57.

Schubert HL, Blumenthal RM, Cheng X. Many paths to methyltransfer: a chronicle of convergence. Trends Biochem Sci. 2003;28(6):329–335. [PMC free article: PMC2758044] [PubMed: 12826405]

58.

Fontecave M, Atta M, Mulliez E. S-adenosylmethionine: nothing goes to waste. Trends Biochem Sci. 2004;29(5):243–249. [PubMed: 15130560]

59.

Esra Onen F, Boum Y, Jacquement C. et al. Design, synthesis and evaluation of potent thymidylate synthase X inhibitors. Bioorg Med Chem Lett. 2008;18(12):3628–3631. [PubMed: 18513963]

60.

Galperin MY, Koonin EV. Functional genomics and enzyme evolution. Homologous and analogous enzymes encoded in microbial genomes. Genetica. 1999;106(1-2):159–170. [PubMed: 10710722]

61.

Jacob F. Evolution and tinkering. Science. 1977;196(4295):1161–1166. [PubMed: 860134]

62.

Kealey JT, Gu X, Santi DV. Enzymatic mechanism of tRNA (m5U54)methyltransferase. Biochimie. 1994;76(12):1133–1142. [PubMed: 7748948]

63.

Almog R, Waddling CA, Maley F. et al. Crystal structure of a deletion mutant of human thymidylate synthase {{Delta}} (7-29) and its ternary complex with tomudex and dUMP. Protein Science. 2001;10(5):988. [PMC free article: PMC2374201] [PubMed: 11316879]