Report

An unprecedented mechanism of nucleotide methylation in organisms containing thyX

See allHide authors and affiliations

Science  29 Jan 2016:
Vol. 351, Issue 6272, pp. 507-510
DOI: 10.1126/science.aad0300

A special way to make T

The genomes of all cell-based life consist of DNA. Blocking DNA synthesis is thus lethal, and if targeted selectively, its inhibition can provide cancer and antibiotic treatments. For example, the drug methotrexate interferes with the synthesis of thymidine, the base T in DNA. Mishanina et al. found that the enzyme that carries out the last step of thymidine synthesis in several human pathogens, which cause tuberculosis, anthrax, and typhus, uses a previously undescribed mechanism. Knowing the mechanism may allow the development of specific inhibitors for this enzyme.

Science, this issue p. 507

Abstract

In several human pathogens, thyX-encoded flavin-dependent thymidylate synthase (FDTS) catalyzes the last step in the biosynthesis of thymidylate, one of the four DNA nucleotides. ThyX is absent in humans, rendering FDTS an attractive antibiotic target; however, the lack of mechanistic understanding prohibits mechanism-based drug design. Here, we report trapping and characterization of two consecutive intermediates, which together with previous crystal structures indicate that the enzyme’s reduced flavin relays a methylene from the folate carrier to the nucleotide acceptor. Furthermore, these results corroborate an unprecedented activation of the nucleotide that involves no covalent modification but only electrostatic polarization by the enzyme’s active site. These findings indicate a mechanism that is very different from thymidylate biosynthesis in humans, underscoring the promise of FDTS as an antibiotic target.

Enzymes involved in DNA biosynthesis are primary targets of chemotherapeutic and antibiotic agents. One such enzyme is thymidylate synthase, or TSase [Enzyme Commission (EC) number 2.1.1.45]. Encoded by the thyA gene (TYMS in humans), TSase produces thymidylate (dTMP), a precursor of one of the DNA bases, thymine. TSase catalyzes the reductive methylation of the nucleotide deoxyuridine monophosphate (dUMP) with the folate derivative CH2H4fol (Fig. 1). The reaction also produces dihydrofolate (H2fol), which is restored to tetrahydrofolate (H4fol) for reuse in catalysis by dihydrofolate reductase (DHFR). TSase is successfully targeted by drugs such as 5-fluorouracil and raltitrexed, and DHFR is the target of methotrexate and trimethoprim.

Fig. 1 Reactions catalyzed by TSase, DHFR, and FDTS.

Highlighted are the reducing hydrogen in the TSase reaction (red), methylene (purple), and nucleotide under study (blue). R, 2′-deoxyribose-5′-phosphate; R′, (p-aminobenzoyl)glutamate; NADP+, nicotinamide adenine dinucleotide phosphate; NADPH, reduced form of NADP+.

However, in several human pathogens thymidylate formation is catalyzed by thyX-encoded flavin-dependent thymidylate synthase, or FDTS (EC 2.1.1.148), which carries out the functions of both TSase and DHFR (Fig. 1) (1, 2). Many pathogens depend solely on thyX for thymine, including all Rickettsia (causing typhus, spotted fever, and other diseases). Pathogens containing both thyX and thyA, such as Mycobacterium tuberculosis, can synthesize thymidylate through either pathway and often develop multidrug resistance. As multi- and extreme-drug resistance in these pathogens becomes more common, the addition of an FDTS inhibitor to the cocktail could prove essential for treatment. Further information regarding the prevalence of thyX gene in human pathogens, and its potential as a drug target, is provided in the supplementary materials. Hitherto, no drugs are known to selectively inhibit FDTS, and the mechanistic intricacies necessary for the rational design of mechanism-based inhibitors are yet to be resolved. FDTS is genetically and structurally dissimilar not only from the canonical TSase and DHFR but from other flavoenzymes (3, 4). The current report provides a road map to better understanding of FDTS catalysis and to rational design of inhibitory drugs.

To delineate the mechanism of FDTS catalysis, we report on the trapped reaction intermediates, that is, intermediate species that have been chemically modified by a reaction quencher (here, acid or base). Characterization of these trapped compounds suggested a unique nucleotide methylation path. Our previous report on rapid-quenching experiments with Thermatoga maritima FDTS (TmFDTS) revealed a substantial accumulation of an acid-modified intermediate, identified as 5-hydroxymethyl-dUMP (5). In the current study, we used a basic quencher to stop enzymatic turnover and trap different derivatives of the intermediates. Indeed, with radiolabeled nucleotide, [2-14C]-dUMP, we observed a previously unknown radioactive species, distinct from the acid-modified 5-hydroxymethyl-dUMP. The base-modified species was also observed with methylene-labeled folate, [11-14C]-CH2H4fol, indicating that the nucleotide intermediate acquired the methylene before being trapped by the base (fig. S1).

The top panel in Fig. 2 shows the accumulation and decay of the base-trapped intermediate (blue) with [2-14C]-dUMP as a substrate. Upon base addition, dTMP product (black) is formed even though the flavin is in a reduced state before the quencher’s addition (green stopped-flow trace). A comparison of the time course of the acid- and base-modified intermediate derivatives (Fig. 2, bottom) suggests that at least two different reaction intermediates (I1 and I2) are trapped in acid as 5-hydroxymethyl-dUMP. In base, I1 is trapped in a different chemical form than in acid, and I2 is converted to dTMP (see further discussion of this phenomenon in the supplementary materials).

Fig. 2 Oxidative half-reaction kinetics of FDTS.

(Top) Base-quenched data (dots) globally fitted to a single-intermediate model (lines), overlaid with stopped-flow traces at 420 nm (flavin absorbance, A420, in green) (6). Each time point was obtained from a radiogram like that shown in fig. S1A, with dUMP in red, dTMP in black, and intermediate in blue. (Bottom) Base-trapped intermediate kinetics (blue) overlaid with acid-trapped intermediate data (red), globally fitted to a two-intermediate model (red curves). The sum of the two intermediates is shown as a dotted red curve.

What is the base-modified intermediate? Its mass was determined to be [M-H] 705.1212 daltons (fig. S2A). Tandem mass spectrometry (MS/MS) of the compound displayed a loss of 320 daltons, corresponding to CH2-dUMP (fig. S2B), which accords with the 14C found in this intermediate when using 14C labeled dUMP or 14CH2H4fol. Hence the [M-H] 385 ion, corresponding to the mass difference between the trapped species and CH2-dUMP moiety (fig. S2B), must have belonged to an adduct covalently linked to the trapped nucleotide.

All FDTS mechanisms proposed heretofore (69) (e.g., fig. S3) postulate that the methylene is transferred directly from CH2H4fol to the nucleotide, as in the TSase reaction (10), and the absorbance spectrum of the trapped intermediate (fig. S4) accords with folate absorbance. Consequently, the most logical candidate for the adduct was the pterin moiety of CH2H4fol. However, when using CH2H4fol radiolabeled at pterin or benzoyl moieties, we found no radioactivity in the base-modified intermediate, ruling out folate as a component of the base-trapped intermediate. Furthermore, the MS/MS ion of 385 daltons in the base-trapped intermediate did not match any buffer or protein constituents.

The only remaining part of the quenched FDTS reaction that could have provided the mysterious adduct was a flavin adenine dinucleotide (FAD) derivative. To investigate, we turned to FDTS reconstituted with isotopically labeled FAD (11). FDTS reconstituted with FAD labeled at the adenine moiety yielded no labeled intermediate, but the reaction of [7a,8a-3H]-FAD-FDTS produced a tritiated trapped intermediate (fig. S5B), indicating that the labeled dimethylbenzene portion of isoalloxazine (fig. S5A) was a part of the trapped species. Additionally, when we used FDTS containing FAD uniformly labeled with 13C and 15N at its dioxopyrimidine (ring “C” in fig. S5), that is, 6 daltons heavier than unlabeled FAD, the mass of the isolated trapped intermediate was heavier by only 2 daltons than that produced by the unlabeled enzyme (fig. S2C). This finding suggests that only two labeled atoms of the dioxopyrimidine ring of the flavin are retained in the base-modified intermediate.

Several nuclear magnetic resonance (NMR) experiments were performed to determine the structure of the base-trapped intermediate (Fig. 3, figs. S6 and S7, and table S2). Critically, the methylene (U7) originating in the CH2H4fol had bonded to C5 of the dUMP moiety and coupled to FC5a and FC4a of the degraded flavin in the NMR heteronuclear multiple-bond correlation (HMBC) spectrum (Fig. 3). These results and others presented in the supplementary materials show that CH2H4fol-derived methylene bridges dUMP and degraded flavin, connecting C5 of dUMP and N5 of the decomposed isoalloxazine moiety. Other structural features of the base-trapped intermediate derivative are described in the supplementary materials.

Fig. 3 Structure of the base-modified FDTS reaction intermediate.

(Top) Structural components originating in dUMP (red) and CH2H4fol (black) are referred to as “U subunit” and the structural parts derived from FAD (blue) as “F subunit.” Atomic numbering in F subunit follows convention adapted from the original FAD (fig. S5A). (Bottom) Overlay of 1H/13C heteronuclear multiple-quantum coherence (black; red marks folded cross-peaks) and HMBC (green) spectra. The cross-peaks are labeled with the first letter representing the subunit (U or F). The cross-peaks of the HMBC spectrum circled in red unambiguously establish that the CH2H4fol-drived methylene (U7) bridges dUMP and flavin derivative. ppm, parts per million.

The presence of a covalent methylene bridge between dUMP and flavin indicates a previously unconsidered mechanism substantially different from earlier mechanisms (6, 7, 9) (fig. S3). This mechanism, given in Fig. 4 and fig. S8, postulates that N5 of the reduced flavin accepts the activated methylene from the CH2H4fol Schiff base (step 1), and then passes it to C5 of the enzyme-polarized dUMP (steps 2 and 3). This mechanism agrees well with crystal structures of FDTS complexes with FAD, dUMP, and folates [Protein Data Bank identifications (PDB IDs): 4GT9, 4GTA, and 4GTB], in which the isoalloxazine moiety of FAD is sandwiched between the folate ring and uracil (8) (Fig. 4). Thus, whereas protein-bound flavins typically carry electrons from one side of the flavin ring to the other, the FAD of FDTS carries a methylene.

Fig. 4 Proposed chemical mechanism for FDTS.

(Left) Methylene (CH2, purple) is transferred from the N5 of H4fol (green) via N5 of FAD cofactor (gold) to C5 of dUMP substrate (cyan). Oxygens are in red, nitrogens in blue, and the hydride transferred in step 4 in brown. The proposed steps of flavin oxidative half-reaction are numbered. I1 (boxed) is the reaction intermediate that in base is converted to the compound presented in Fig. 3 and in acid undergoes water addition (fig. S10 and supplementary text). I2 (boxed) refers to the intermediate that in base forms dTMP, and in acid reacts with water (fig. S10 and supplementary text). The “420 nm absorbance” versus “colorless” labels of flavins relate the 420-nm time trace (green line in Fig. 2) to the mechanistic model. A rigorous electron-pushing description of this mechanism is presented in fig. S8. R, 2′-deoxyribose-5′-phosphate; R′, (p-aminobenzoyl)glutamate; and R″, adenosine-5′-pyrophosphate–ribityl. (Right) Active site of FDTS [PDB ID 4GTA (8)] in complex with FAD, dUMP, and folinic acid, a stable analog of the activated CH2H4fol. For clarity, the carbonyl oxygen of folinic acid is not shown, and the structure is inversed to match the orientation of ligands in the left-hand structures. The formyl carbon (magenta) on folinic acid represents the methylene to be transferred. Black arrows mark the methylene-transfer trajectory from donor (N5 of folate) to proposed relay atom (N5 of flavin) and final acceptor (C5 of dUMP).

The proposed intermediates I1 and I2 are modified in base to the compound identified in Fig. 3 and product, respectively. However, both are converted to 5-hydroxymethyl-dUMP in acid. The mechanisms of their degradation are discussed in the supplementary materials and presented in fig. S10.

Why wasn't such a mechanism considered when the structure of FDTS complex with its substrates was solved in 2012 (8)? The reason lies in reports of activity of FDTS reconstituted with 5-deaza-5-carba-FAD (henceforth 5-deaza-FAD) (7, 9). In 5-deaza-FAD, N5 of the flavin has been replaced with a carbon, which prohibits several chemical conversions proposed in Fig. 4. The identification of the base-modified intermediate prompted us to revisit this finding. We synthesized 5-deaza-FAD and incorporated it into apo-FDTS. In doing so, we found that the procedure used in the past for removing FAD from holo-TmFDTS (salting out) (4, 9) leaves behind variable amounts of FAD, producing the same level of (residual) activity in the salted-out “apoenzyme” as in 5-deaza-FAD–reconstituted FDTS. By denaturing the apoenzyme with 8 M urea after salting out, completely removing FAD, then refolding and reconstituting with either FAD or 5-deaza-FAD, we found no activity for the apoenzyme, fully recovered activity for FAD-reconstituted enzyme, and no activity for the enzyme reconstituted with 5-deaza-FAD (fig. S9). This indicates that 5-deaza-FAD-FDTS is actually not active at all (<10−6 at the signal/noise limit, see supplementary materials), removing the conceptual barrier for the mechanism proposed in Fig. 4.

Flavin’s N5-mediated methylene transfer might be rare, but it has been proposed for TrmFO tRNA methyltransferase (12). However, in TrmFO the uracil is activated via Michael addition of active-site cysteine to C6 of the pyrimidine ring, and a second cysteine activates and delivers the methylene to the flavin. In FDTSs, no such enzymatic nucleophile is available (7), and the cause of the uracil’s initial activation has been a matter of uncertainty. One hypothesis posits that the N5 hydride of reduced FAD nucleophilically activates dUMP (7) (fig. S3A), whereas another suggests that the enzyme polarizes the pyrimidine ring of dUMP for an attack on the CH2H4fol methylene (6, 9) without a Michael nucleophile (fig. S3B). The latter chemistry is unprecedented in nucleotide methylation, and the need for pyrimidine activation via Michael addition to its C6 has been emphasized in enzymatic and nonenzymatic systems (6). The only supporting pieces of evidence so far were the absence of deuterated trapped intermediates (6) and ultraviolet-visible features consistent with those of reduced flavin [Fig. 2, green trace, and (5, 9)], but these observations can be rationalized by more conventional mechanisms (6). Such nucleotide activation was thus met with healthy skepticism.

The current findings provide the strongest supporting evidence so far for pyrimidine activation via polarization in the reduced enzyme’s active center. Because the methylene-carrier role for flavin’s N5 proposed here (Fig. 4) requires FAD to be reduced during carbon transfer to dUMP, flavin oxidation during substrate activation (fig. S3A) can be definitively ruled out. The mechanism proposed in Fig. 4 also agrees with the observation (7) that, when the TmFDTS reaction is conducted in D2O at near-physiological temperatures, a deuterium is incorporated at C7 of the dTMP product while some deuterium is found at C6 of the product at low temperature (fig. S8).

The data presented above dictate a thymidylate biosynthesis mechanism (Fig. 4) in thyX-dependent human pathogens that is fundamentally different from that found in humans, lending hope for development of mechanism-based FDTS inhibitors as a new class of nontoxic antibiotics (see supplementary materials).

Supplementary Materials

www.sciencemag.org/content/351/6272/507/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S10

Tables S1 and S2

References (1338)

References and Notes

Acknowledgments: We thank D. Roston for assistance with kinetic data fitting, and M. S. Hossain and F. W. Foss Jr. for assistance with preparation of 5-deaza-riboflavin. This work was funded by NIH R01 GM110775 to A.K. and a fellowship from the Iowa Center of Biocatalysis and Bioprocessing (NIH T32 GM008365) to T.V.M. and K.K.
View Abstract

Stay Connected to Science

Subjects

Navigate This Article