An Alternative Flavin-Dependent Mechanism for Thymidylate Synthesis

See allHide authors and affiliations

Science  05 Jul 2002:
Vol. 297, Issue 5578, pp. 105-107
DOI: 10.1126/science.1072113


Although deoxythymidylate cannot be provided directly by ribonucleotide reductase, the gene encoding thymidylate synthase ThyA is absent from the genomes of a large number of nonsymbiotic microbes. We show that ThyX (Thy1) proteins of previously unknown function form a large and distinct class of thymidylate synthases.ThyX has a wide but sporadic phylogenetic distribution, almost exclusively limited to microbial genomes lackingthyA. ThyX and ThyA use different reductive mechanisms, because ThyX activity is dependent on reduced flavin nucleotides. Our findings reveal complexity in the evolution of thymidine in present-day DNA. Because ThyX proteins are found in many pathogenic microbes, they present a previously uncharacterized target for antimicrobial compounds.

All deoxythymidine 5′-monophosphate (dTMP) in bacteria and eukarya is thought to be formed either de novo by thymidylate synthase (ThyA)–dependent methylation of deoxyuridine 5′-monophosphate (dUMP) or by thymidine kinase (Tdk)–dependent salvage of thymidine compounds from the growth medium (1). ThyA uses tetrahydrofolate (H4folate) as a reductant and forms dihydrofolate (H2folate) as a product of the methylation reaction (2). Because reduced folate derivatives are essential for a variety of biological processes, H2folate formed by ThyA is rapidly reduced to H4folate by dihydrofolate reductase (DHFR). This functional, and often physical, coupling of ThyA and DHFR proteins was thought to be essential for de novo thymidylate synthesis in virtually all actively dividing cells.

The hyperthermophilic anaerobic archaeon Pyrococcus abyssi,which lacks thymidine kinase, incorporates label from extracellular uracil, but not from thymidine, into its DNA (3). This implies that P. abyssi must synthesize thymidylate de novo. However, iterative similarity searches (4) of the three completed Pyrococcusgenomes did not reveal any candidate genes for thyA or for their distantly related putative archaeal homologs (5). When analyzing additional genomes, we detected many other examples of archaea and bacteria that lacked known pathways for formation of dTMP (see Table 1 for a nonexhaustive listing). Many organisms missing thyA also lacked recognizable dihydrofolate reductase genes (Table 1). Together, these in silico observations suggested that, in many microbes, alternative enzymes might function in thymidylate metabolism.

Table 1

Nonexhaustive list of genomes apparently lackingthyA and tdk genes. Clinically relevant bacteria are indicated in bold type. N.A., not analyzed.

View this table:

A gene of unknown function (6) had previously been shown to complement a thymidine-requiring mutant of Dictyostelium discoideum to thymidine prototrophy. This gene was namedthy1, but, to avoid confusion with the Thy-1 cell surface antigen, we will refer to this gene and its homologs asthyX. ThyA and ThyX (Thy1) lack any sequence similarity, and there was no functional evidence that ThyX is a thymidylate synthase. Using thyX from D. discoideum (GenBank accession number g135829) as bait in manual similarity searches, we have found many additional thyXgenes with a wide phylogenetic distribution in up to 30% of microbial genomes (Fig. 1) (7). Although the average pairwise sequence identity of the ThyX homologs is only 28.4% (8), sequence alignments revealed a specific sequence motif RHRX7S (“ThyX motif”) common to this family of proteins (fig. S1) (other positively charged amino acid residues can substitute for the arginine residues) (9). With the exception of Corynebacterineagenomes, thyX and thyA genes have mutually exclusive phylogenetic patterns (10), on the basis of which we predict that ThyX has substituted for the unrelated ThyA protein.

Figure 1

An unrooted phylogenetic tree based on ThyX homologs retrieved by iterative similarity searches was obtained by maximum likelihood (ML) analyses with the program PROTML [quick-add search with 500 replicates, JJT-F (16) amino acid substitution model (17)]. The tree shown is the best ML tree obtained. Bootstrap values were computed using the resampling of estimated likelihood (RELL) method (values higher than 90% are indicated by closed circles). Similar results were also obtained using the neighbor joining algorithm. The scale bar represents the number of substitutions per 100 sites. Additional, more distant, candidatethyX genes were detected in Chlamydia andThermoplasma species but were excluded from the analysis shown here. A, archaea; E, eukarya; + and –, Gram-positive and Gram-negative bacteria, respectively. Viral sequences are underlined.

We tested directly whether P. abyssi or Helicobacter pylori thyX (GenBank accession numbers g14916853 and g7464006, respectively) can functionally complement growth defects of anEscherichia coli strain specifically impaired in thymidylate synthase activity. Using an arabinose inducible promoter, these proteins were produced in E. coli strain χ2913 carrying a well-defined deletion in the thyA gene (Fig. 2A). In the absence of thymidine,E. coli χ2913 transformants, carrying the H. pylori thyX gene on plasmid, formed yellowish colonies on solid minimal media only in the presence of arabinose (Fig. 2B). The level of growth for individual colonies in our complementation tests was approximately 50% of that observed in the presence of thymidine. The growth was inhibited by trimethoprim, a specific inhibitor of dihydrofolate reductase, but only in the presence of a relatively high dose of this drug [100 μg/ml (table S1)]. Changing Ser107 of H. pylori ThyX [the last residue in the conserved ThyX motif (fig. S1)] into Ala and ochre (stop) codons abolished complementation (table S1). These genetic data show directly that H. pylori thyX can functionally replaceE. coli thyA in dTMP synthesis. In similar experiments,P. abyssi thyX did not complement E. coli χ2913 to thymidine prototrophy, presumably reflecting either the incapability of this ”hyperthermophilic“ protein to function under ”mesophilic“ conditions or the presence of chemically modified folates in Pyrococcus sp. (11).

Figure 2

(A) Expression analyses of P. abyssi (strain Orsay) and H. pylori (strain 26695) ThyX proteins. Western immunoblot analyses using whole-cell lysates of noninduced (–arabinose) and induced (+0.2% L-arabinose) cell cultures of E. coli χ2913 (ΔthyA) with monoclonal antibodies to Gly-Lys-Pro-Ile-Pro-Asn-Pro-Leu-Leu-Gly-Leu-Asp-Ser-Thr (V5) epitope (V5) are shown. (B) The ability of P. abyssi and H. pylori thyX to permit thymidine-independent growth of E. coli thymidine auxotroph χ2913 (ΔthyA) was scored after 3 to 4 days in the presence or absence of 0.2% L-arabinose on minimal M9 agar lacking thymidine (18). Complementation of χ2913 (ΔthyA) to thymidine prototrophy was brought about by arabinose induction of H. pylori ThyX. X2913, E. coli χ2913.

We purified H. pylori ThyX carrying a histidine-tag at its carboxy-terminus from cell-free extracts of E. colistrain χ2913. The obtained protein preparations (purity >95%) typically contained 1 to 2 mg/ml of protein, with an apparent molecular mass of ≈31 kD on SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gels (the expected molecular mass of H. pyloriThyX is 31.5 kD) (Fig. 3A), and were bright yellow in color. Superdex 200 (Phamacia, Sweden) size-exclusion chromatography revealed a native molecular mass of 111 kD for this protein, which suggests that its active form could correspond to a homotetramer (12). Spectroscopic analyses of the isolated (oxidized) protein revealed absorbance characteristics typical of a flavoprotein (Fig. 3B), with broad peaks at 447.5 and 375 nm, which were absent in the dithionite-reduced enzyme. Similar absorption characteristics were found for the cofactor after its release from the protein by heat denaturation at 80°C for 5 min.

Figure 3

Biochemical analyses of H. pyloriThyX. (A) 12% SDS-PAGE and immunoblot analyses of isolatedH. pylori ThyX protein. 1.5 (lane 1) and 0.3 (lane 2) μg of pure protein was detected using Coomassie Brilliant Blue staining or monoclonal antibodies to V5 epitope (Invitrogen) with enhanced chemiluminescence detection, respectively. The expected molecular weight of H. pylori ThyX is 31.5 kD (B) Spectroscopic analyses of H. pylori ThyX indicate that it is a flavoprotein. Absolute spectra of 10 μM isolated enzyme (black line) and the cofactor after its release from protein (gray line). The inset shows spectra for the oxidized enzyme (Ox) and the dithionite reduced protein (Red). (C) dTMP-forming activity of H. pylori ThyX protein. Positions of radioactive peaks were determined by collecting 1-mL fractions, followed by the determination of radioactivity using scintillation counting (3). The complete reaction is indicated by a smooth line; the control reaction omitting ThyX protein is indicated by a dashed line.

The loss of tritium from [5-3H]dUMP during the formation of dTMP allows the quantification of thymidylate synthase activity after removal of the radioactive nucleotides from the reaction mixtures (13). We found that ThyX catalyzes in vitro the release of tritium from [5-3H]dUMP in a protein concentration– and CH2H4folate-dependent manner (12), revealing the first biochemical activity for ThyX proteins. However, the activity detected under conditions described for ThyA proteins was too low to explain the results of our complementation tests. In further experiments, we found that including oxidized flavin mononucleotide (FMN) or flavin adenine dinucleotide (FAD) and reduced pyridine nucleotides in the assays substantially increased the tritium release activity of H. pylori ThyX (Table 2). The simplest explanation for this observation is that an electron flow from reduced pyridine nucleotides via flavin nucleotides is necessary for catalytic activity of ThyX proteins. To exclude the possibility that H. pylori ThyX is a uridine 5′ monophosphate (UMP) using methyl transferase, we have shown that UMP cannot compete with dUMP in the reaction (Table 2). In addition, tritium-release activity of H. pylori ThyX is inhibited by micromolar concentrations of dTMP (12). Moreover, with the use of [6-3H]dUMP and high-pressure liquid chromatography (Fig. 3C), we have analytically shown that H. pylori ThyX possesses a dTMP-forming activity (≈75 nmol/mg protein in a 60-min incubation). These results have established that the physiologically relevant activity of H. pylori ThyX is that of a dUMP using thymidylate synthase.

Table 2

Optimized assay conditions for H. pyloriThyX protein.

View this table:

Analogous to CH2H4folate- and FADH2-dependent ribothymidyl synthase ofStreptococcus faecalis (14), our data suggest that CH2H4folate functions in the ThyX catalyzed reaction only as a one-carbon donor, whereas electrons required for the formation of the methyl moiety originate from reduced-flavin nucleotides. Similar to S. faecalisribothymidyl synthase, ThyX catalysis should then result in the formation of H4folate as a reaction product. Because of the intrinsic instability of CH2H4folate and reduced flavin nucleotides in the presence of molecular oxygen, we have not yet directly tested this proposal. However, the proposed reaction mechanism for ThyX is supported by the absence of dihydrofolate reductase genes in many ThyX-containing organisms (Table 1), whereas all bacteria and eukarya with thyA contain DHFR (12). The existence of different reaction mechanisms for the two classes of thymidylate synthases is also supported by the observation that ThyX proteins lack sequence motifs that are thought to be essential for catalysis and substrate binding in ThyA proteins (2).

ThyA and/or thyX genes are present in all completed genome sequences, which suggests that these two proteins are the only thymidylate synthases. The lack of any sequence similarity between ThyA and ThyX proteins, together with the differences in their enzymatic mechanisms, suggests an independent origin for the two distinct thymidylate synthases. This suggestion was confirmed by the very recently determined structure of a eubacterial ThyX homotetramer (see note added in proof). The peculiar sporadic distribution of thyX in the three domains of life (Fig. 1) likely results from several independent lateral gene transfer or non-orthologous gene replacement events (e.g., between α-proteobacteria and eukarya or from bacteriophages to mycobacteria). These multiple gene–transfer events prevent a definitive conclusion regarding the evolutionary origin of the two pathways for thymidylate synthesis, but indicate the hitherto unnoticed complexity in the evolution of thymidine-containing genetic material (15). Notably, thyX is present in a number of human pathogenic bacteria, but absent in the human genome, making this class of thymidylate synthases an attractive target for specifically inhibiting microbial growth.

Note added in proof: The structure of the Thermotoga maritima ThyX homotetramer (1KQ4) in complex with a flavin adenine dinucleotide has been solved very recently by the Joint Center for Structural Genomics ( It reveals that ThyA and ThyX proteins are not structurally related, thus confirming the independent origin for the two distinct classes of thymidylate synthases.

  • * To whom correspondence should be addressed. E-mail: hannu.myllykallio{at} (H.M.) and ursula.liebl{at} (U.L.)


View Abstract

Navigate This Article