An enzymatic activation of formaldehyde for nucleotide methylation

Folate enzyme cofactors and their derivatives have the unique ability to provide a single carbon unit at different oxidation levels for the de novo synthesis of amino-acids, purines, or thymidylate, an essential DNA nucleotide. How these cofactors mediate methylene transfer is not fully settled yet, particularly with regard to how the methylene is transferred to the methylene acceptor. Here, we uncovered that the bacterial thymidylate synthase ThyX, which relies on both folate and flavin for activity, can also use a formaldehyde-shunt to directly synthesize thymidylate. Combining biochemical, spectroscopic and anaerobic crystallographic analyses, we showed that formaldehyde reacts with the reduced flavin coenzyme to form a carbinolamine intermediate used by ThyX for dUMP methylation. The crystallographic structure of this intermediate reveals how ThyX activates formaldehyde and uses it, with the assistance of active site residues, to methylate dUMP. Our results reveal that carbinolamine species promote methylene transfer and suggest that the use of a CH2O-shunt may be relevant in several other important folate-dependent reactions.

2. Catalytic mechanism is also important. Because crystal structures were reported in the manuscript, the authors may discuss on the contribution of amino acid resides including Tyr91 and Ser88 to the catalytic mechanism of ThyX. For example, how the H2O molecule is generated during the reaction? Furthermore, it is helpful for general readers to show the conservation of these amino acid residues in sequences as well as structures.
Reviewer #2 (Remarks to the Author): Authors demonstrate that in anaerobic conditions formaldehyde can replace the natural methylene donor CH2THF in ThyX catalyzed methylation of dUMP. These findings are highly interesting and endorsed by crystallographic, NMR and HPLC data. However, comparison of reaction kinetics with the native methylene donor should be included. In the crystal structure a the carbinolamine adduct of reduced FADH-is observed. Authors are convinced this is the methylating agent and not a trapped intermediate that should be activated into a reactive imine since 'The methylene group of the carbinolamine adduct is nicely located only 2.4 Å from C5-dUMP and is ideally poised for an SN2 attack by the activated nucleobase to generate an intermolecular C-C bond.' However, without MD these are not convincing arguments. An MD is the least that should be performed on the obtained model to have a more reliable idea on distances and hydrogen bonding interactions indicated in fig 3d, to endorse the hypothesis that carbinolamine flavin species could be the active species instead of its iminium counterpart, that is so far postulated as the dUMP methylating agent. For comparison: To date, the 5-HOCH2H4folate that was considered as structural evidence for the 5iminium ion intermediate, which is the proposed reactive form of CH2H4folate in ThyA (Biochemistry. 1993 Jul 20;32(28) Fig S4C. The full spectrum clearly shows many signals next to the signal of interest at 11ppm. The legend of Fig S4C mentions 'intense peaks are from phenol', however only the 4 signals above 100ppm arise from phenol. No explanation is given for signals between 50ppm and 100ppm, no full reference spectra are added in conditions without dUMP or ThyX. 'Our structure is perfectly superimposable with that of the ThyX/dUMP/folate ternary complex previously reported (r105)' This statement is hard to judge without having access to the pdb or RMSD values. Fig 3c shows a model of ThyX in complex with the flavin carbinolamine, dUMP and folate that was obtained by superimposing the crystal structure of FADH-308 •CH2O complex with the structure of ThyX in complex with dUMP and folate (pdb, 4gt9). It is impossible to see any steric clashes since only the ribbon diagram of ThyX (determined in this work?) is shown. The legend of fig 3 mentions R174 participating in activation of dUMP, though no reference is given for this statement. Authors argue that 'The structure shows that Tyr91 and Ser88 (Fig 3d), previously shown to be critical for activity by site-directed mutagenesis, could stabilize such a conformation via H-bonding' also here a reference is missing. The legend of fig3d mentions 'This model shows the different orientations adopted by these residues in two different structures of ThyX' though it does not mention structures that are used for this overlay.
Reviewer #3 (Remarks to the Author): Bou-Nader et al. interrogate the mechanism of an alternative, flavin-dependent thymidylate synthase (FDTS), with biochemical and structural biology methods. FDTS is an attractive antimicrobial target because it isn't found in humans and employs chemistry orthogonal to that of the human enzyme. Using formaldehyde as a proxy for the biological carbon donor, methylenetetrahydrofolate (MTHF), the authors beautifully demonstrate that carbon travels to its final acceptor (dUMP) via a novel N5 flavin carbinolamine intermediate. This intermediate is observed in cristallo and spectroscopically and is shown to be catalytically competent. The work is elegantly performed, with necessary controls, and will be of significant interest not only to the drug-development community, but also to the flavoenzyme community, adding to the growing collection of covalent flavin adducts as reactive species.
One suggestion I have is to make it clearer in the text that formaldehyde is used as a tool here, to avoid potential confusion that authors are claiming formaldehyde as the biological carbon donor in vivo. For instance, the authors might compare Kd values for MTHF (tens of uM) vs formaldehyde (tens of mM) and state that large discrepancy agrees with folate serving as a CH2O carrier/positioner, thanks to its specific interactions with ThyX and its flavin.
Minor comment: Fig. S4: The spectra are very small and difficult to read. Can you please enlarge each spectrum and label peaks of interest, specifically the ~12 ppm peak for dTMP's methyl group in panel c?

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): This manuscript proposed the chemical mechanism of the methylation by ThyX to produce the thymidine nucleotide. Although the tertiary structures of ThyX from several organisms have been determined so far, the reaction mechanism has been unclear. The authors demonstrated the presence of the carbinolamine intermediate by the X-ray crystallography. Also, the authors showed that formaldehyde can be used instead of methylene-THF to produce dTMP from dUMP. Based on these results, a novel chemical mechanism for ThyX is proposed. The elucidation of chemical mechanism is fundamental in biochemistry and this manuscript gave new insight of the enzymatic reaction. However, the following points will be addressed by the authors to justify the proposed mechanism.
We thank the reviewer for his/her positive comments. The reviewer is absolutely right. With formaldehyde, we necessarily have the formation of a flavin-carbinolamine species, which can subsequently evolve to its iminium counterpart. In the case of the reaction with CH2THF, the flavin-iminium is believed to form first via an addition-elimination reaction between the reduced flavin and folate. Likewise, this iminium can subsequently give the carbinolamine.
Of course, the previously proposed flavin-iminium electrophile could be the bona fide methylating agent. However, here we provide further pieces of evidence that we believe are in disfavor of a flavin-iminium. Studies by Kemal & Bruice in 1976 (see in references of the revised article) showed that the solvolysis of a model flavin generated the carbinolamine likely formed by the rapid reaction of the iminium with water, attesting to the extreme instability of the iminium. As a matter of fact, ThyX presents a large active site that is quite exposed to the solvent and which should prevent formation and reaction of the iminium with dUMP. It is therefore expected that formation of the carbinolamine species from the iminium counterpart occurs at a much faster rate than that of the carbon transfer reaction to dUMP. Furthermore the sp3-hybridized carbinolamine presents a favorable distance and geometry for the in-line attack of the nucleophile in carbon transfer and water displacement. This is compared to the geometric requirements a nucleophilic attack on an sp2-hybridized iminium (Fig S14). Efficient attack of nucleophiles on the system of carbonyls or imines occurs along the so-called Bürgi-Dunitz trajectory JACS (1973) 95, 5065-5067 , with the nucleophile attacking the unsaturated carbon at vector ~107° from the C-N or C-O bond in a direction towards the electronegative atom (Light et al 2014). If water were eliminated from the carbinolamine-flavin adduct we observed crystallographically, the methylene carbon of the planar iminium-flavin adduct would be positioned over the C-N double-bond and much too distant to attack the iminium carbon, the C-N vector of the iminium would be almost 180° away from the direction needed for attack.
Here a stereo view showing ThyX active site with the dUMP and the flavin-carbinolamine and in which the OH of the carbinolamine adduct is rotated to its reactive position (either for displacement or elimination). It's obvious from the figure that elimination to form the iminium puts the CH2 in a place where it simply cannot react; it would take more than just a little dynamical changes to achieve a reactive conformation. In obvious contrast, the carbinolamine is poised to react. Therefore, these stereoelectronic considerations lead us to favor the carbinolamine as the actual carbon-transfer agent. We have discussed these issues in the revised manuscript.
2. Catalytic mechanism is also important. Because crystal structures were reported in the manuscript, the authors may discuss on the contribution of amino acid resides including Tyr91 and Ser88 to the catalytic mechanism of ThyX. For example, how the H2O molecule is generated during the reaction?
Although Pka of these residues are not available in ThyX, serine and tyrosine could act as acid by protonating the carbinolamine. Alternatively, these residues could polarize a water molecule into the active site of ThyX, which would serve as a proton donor. However the fact that phenolic oxygen of Tyr 91 is within hydrogen-bonding distance to a guanidinium nitrogen of Arg 90, which is also conserved among ThyX homologues, leads us to favor the first scenario.
We now further elaborate on the catalytic mechanism by stating: "The carbinolamine geometry favors an SN2-like attack by dUMP, which is activated as a nucleophile by deprotonation of N3 through electrostatic interaction with R174, enhancing the enaminecharacter of the pyrimidine moiety. Nucleophilic attack leads to the formation of a covalent FAD-CH2-dUMP adduct, previously proposed based upon the fragments obtained from rapid-quenching by base 27 , and the displacement of a water molecule. The two conserved residues, Ser88 and Tyr91, at hydrogen bonding distance with the carbinolamine hydroxyl group may assist the SN2 reaction by acting as acids and promoting the displacement of the -hydroxyl leaving group (Fig 3d). It is worth noting that the phenolic oxygen of Tyr 91 is within hydrogen-bonding distance to a guanidinium nitrogen of Arg 90 (also conserved); the proximity of such a (presumably) positive charge could enhance the acidity of Tyr 91, which might be the general acid catalyst that assists the displacement of the leaving water." Furthermore, it is helpful for general readers to show the conservation of these amino acid residues in sequences as well as structures.

Reviewer #2 (Remarks to the Author):
Authors demonstrate that in anaerobic conditions formaldehyde can replace the natural methylene donor CH2THF in ThyX catalyzed methylation of dUMP. These findings are highly interesting and endorsed by crystallographic, NMR and HPLC data. However, comparison of reaction kinetics with the native methylene donor should be included.
We thank the reviewer for his/her positive feedback. A kinetic trace with the CH2THF has been added in the new figure 2A.
In the crystal structure a the carbinolamine adduct of reduced FADH-is observed. Authors are convinced this is the methylating agent and not a trapped intermediate that should be activated into a reactive imine since 'The methylene group of the carbinolamine adduct is nicely located only 2.4 Å from C5-dUMP and is ideally poised for an SN2 attack by the activated nucleobase to generate an intermolecular C-C bond.' However, without MD these are not convincing arguments. An MD is the least that should be performed on the obtained model to have a more reliable idea on distances and hydrogen bonding interactions indicated in fig 3d, to endorse the hypothesis that carbinolamine flavin species could be the active species instead of its iminium counterpart, that is so far postulated as the dUMP methylating agent. For comparison: To date, the 5-HOCH2H4folate that was considered as structural evidence for the 5-iminium ion intermediate, which is the proposed reactive form of CH2H4folate in ThyA (Biochemistry. 1993 Jul 20;32 (28) We thank the reviewer for his proposal regarding the molecular dynamics. However, the complexity of the system, a tetrameric protein with 4 active sites, each of them containing a flavin-carbinolamine, dUMP and folate, makes parametrization very complicated. We fully agree that this may be a direction to explore in the future and it is a very demanding study of investigations in itself. In the current state, the structure seems to us to be fairly conclusive on the fact of the presence of a reaction intermediate, which acts as a methylating agent for dUMP. However, we here provide further pieces of evidence that we believe are in disfavor of a flavin-iminium. Studies by Kemal & Bruice in 1976 (see in references of the revised article) showed that the solvolysis of a model flavin generated the carbinolamine likely formed by the rapid reaction of the iminium with water, attesting to the extreme instability of the iminium. As a matter of fact, ThyX presents a large active site that is quite exposed to the solvent and which should prevent formation and reaction of the iminium with dUMP. It is therefore expected that formation of the carbinolamine species from the iminium counterpart occurs at a much faster rate than that of the carbon transfer reaction to dUMP. Furthermore the sp3-hybridized carbinolamine presents a favorable distance and geometry for the in-line attack of the nucleophile in carbon transfer and water displacement. This is compared to the geometric requirements a nucleophilic attack on an sp2-hybridized iminium (Fig S14). Efficient attack of nucleophiles on the system of carbonyls or imines occurs along the so-called Bürgi-Dunitz trajectory JACS (1973) 95, 5065-5067 , with the nucleophile attacking the unsaturated carbon at vector ~107° from the C-N or C-O bond in a direction towards the electronegative atom (Light et al 2014). If water were eliminated from the carbinolamine-flavin adduct we observed crystallographically, the methylene carbon of the planar iminium-flavin adduct would be positioned over the C-N double-bond and much too distant to attack the iminium carbon, the C-N vector of the iminium would be almost 180° away from the direction needed for attack.
Here a stereo view showing ThyX active site with the dUMP and the flavin-carbinolamine and in which the OH of the carbinolamine adduct is rotated to its reactive position (either for displacement or elimination). It's obvious from the figure that elimination to form the iminium puts the CH2 in a place where it simply cannot react; it would take more than just a little dynamical changes to achieve a reactive conformation. In obvious contrast, the carbinolamine is poised to react. Therefore, these stereoelectronic considerations lead us to favor the carbinolamine as the actual carbon-transfer agent. We have discussed these issues in the revised manuscript.
The top spectrum in fig 2A is an expansion of the full NMR spectrum (not RMN!) in Fig S4C. The full spectrum clearly shows many signals next to the signal of interest at 11ppm. The legend of Fig S4C mentions 'intense peaks are from phenol', however only the 4 signals above 100ppm arise from phenol. No explanation is given for signals between 50ppm and 100ppm, no full reference spectra are added in conditions without dUMP or ThyX.
We corrected the term NMR. We also corrected our typo error and replaced the term dUMP by dTMP in the legend of figure S4A. We also added figure S4D (reference spectra without ThyX), which does not have the signal of interest at 11 ppm. As the reviewer mentioned, the peaks above 100 ppm (at 115, 121, 129 and 155 ppm) arise from phenol. As for the other peaks between 50 and 100 ppm, the peak at 81 ppm arises from formaldehyde as shown in the NMR spectrum of formaldehyde ( Figure S4B), the peak at 62 ppm probably arises from Tris-Cl buffer and is present in activity tests without ThyX ( Figure S4D). The other minor peaks at 59, 72, 87 and 88 ppm could not be attributed but are also present in activity tests without ThyX (Figure S4D;) indicating that they are not arising from an activity of ThyX. These peaks could arise from impurities in phenol used for products extraction.
'Our structure is perfectly superimposable with that of the ThyX/dUMP/folate ternary complex previously reported (r105)' This statement is hard to judge without having access to the pdb or RMSD values.
We now state in the result section "The structures show that the homotetrameric enzyme does not undergo significant conformational changes with overall root-mean-square deviation (RMSD) of 0.2 Å over 167 residues".
Fig 3c shows a model of ThyX in complex with the flavin carbinolamine, dUMP and folate that was obtained by superimposing the crystal structure of FADH-308 •CH2O complex with the structure of ThyX in complex with dUMP and folate (pdb, 4gt9). It is impossible to see any steric clashes since only the ribbon diagram of ThyX (determined in this work?) is shown. Fig S8 shows the structural superposition of our structures with the previously reported structure of ThyX bound to CH2THF and dUMP (PDB 4gt9). We have expanded this by plotting the RMSD for each residue of each subunit. This further confirms that minor rearrangements of some side chains occur while the overall structures are identical. Fig S12) showing a stereo view of all the side chains in the active site at the vicinity of the flavin-carbinolamine in our structures in the presence of dUMP and CH2THF (pdb 4gt9). There is only the CH2OH moiety on the N5 of flavin that clashes with dUMP.

Additionally, we have added a figure in the main manuscript (new
The legend of fig 3 mentions R174 participating in activation of dUMP, though no reference is given for this statement.
The reference DOI 10.1021/bi500648n has been added.
Authors argue that 'The structure shows that Tyr91 and Ser88 (Fig 3d), previously shown to be critical for activity by site-directed mutagenesis, could stabilize such a conformation via Hbonding' also here a reference is missing.
The reference DOI 10.1021/bi500648n has been added.
The legend of fig3d mentions 'This model shows the different orientations adopted by these residues in two different structures of ThyX' though it does not mention structures that are used for this overlay.

The structures used for this overlay are: (i) ThyX-FADHsoaked with 20 mM Formaldehyde vs (ii) pdb 4gt9
Bou-Nader et al. interrogate the mechanism of an alternative, flavin-dependent thymidylate synthase (FDTS), with biochemical and structural biology methods. FDTS is an attractive antimicrobial target because it isn't found in humans and employs chemistry orthogonal to that of the human enzyme. Using formaldehyde as a proxy for the biological carbon donor, methylenetetrahydrofolate (MTHF), the authors beautifully demonstrate that carbon travels to its final acceptor (dUMP) via a novel N5 flavin carbinolamine intermediate. This intermediate is observed in cristallo and spectroscopically and is shown to be catalytically competent. The work is elegantly performed, with necessary controls, and will be of significant interest not only to the drug-development community, but also to the flavoenzyme community, adding to the growing collection of covalent flavin adducts as reactive species.
We thank the reviewer for his/her positive feedback.
One suggestion I have is to make it clearer in the text that formaldehyde is used as a tool here, to avoid potential confusion that authors are claiming formaldehyde as the biological carbon donor in vivo. For instance, the authors might compare Kd values for MTHF (tens of uM) vs formaldehyde (tens of mM) and state that large discrepancy agrees with folate serving as a CH2O carrier/positioner, thanks to its specific interactions with ThyX and its flavin.
To make it clearer that formaldehyde is used as a tool in our study we have added the following statements: In the introduction we now say, "we show that a CH2O-shunt can replace the natural methylene donor for ThyX-dependent dUMP methylation". In the result section we emphasized this by stating "Taken together, these results confirmed that ThyX uses CH2O as a direct methylene donor for dTMP synthesis and can therefore substitute for CH2THF by forming a flavin intermediate that could be a carbinolamine adduct." In the discussion we have added: "The tighter binding of ThyX for CH2THF (KD~ 4M 27 ) compared to CH2O (KD~ 20 mM) confirms that methylene tetrahydrofolate acts as the biological carbon donor for ThyX, serving as a CH2O carrier/transporter and thus avoiding genotoxic effects 28-30 .
Minor comment: Fig. S4: The spectra are very small and difficult to read. Can you please enlarge each spectrum and label peaks of interest, specifically the ~12 ppm peak for dTMP's methyl group in panel c?
We enlarged all NMR spectra and added a new figure (S4D), which is a reference spectrum without ThyX.