Novel ATP-cone-driven allosteric regulation of ribonucleotide reductase via the radical-generating subunit

Essential revisions:

1) In the present work, dADP is shown for the first time to act as an allosteric inhibitor. dADP appears to have half the affinity for the activity site compared to the canonical allosteric inhibitor, dATP. Does this finding imply that dADP has a physiological role in regulation?

Actually, dADP has an affinity which is only a fifteenth of the dATP affinity (Figure 8F) and because of that we believe that dATP is the main physiological allosteric effector. However, the relative affinity for dADP compared to dATP is still higher than in most other RNRs and we cannot exclude the possibility that local concentrations around the enzyme might be high enough to actually have a physiological effect. Nevertheless, we have modified the text to make it clear that we believe that dATP is the main physiological effector of the two.

What is the effect of dADP on oligomerization compared to dATP? The k_on for dADP should be compared between L. blandensis RNR and the triphosphate-forming kinases that convert dADP to dATP to see whether such a role is feasible.

The main message of our current study is the transfer of an ATP-cone domain from the catalytic subunit of an RNR to its corresponding radical-containing subunit, and that the ATP-cone confers similar effects of dATP and ATP to the oligomeric status of the NrdB subunit. We would therefore prefer not to focus the current study on the binding of dADP to the ATP-cone, as we have shown that the binding of dADP is only a fifteenth compared to binding of dATP. Both dADP and dATP induce the formation of NrdB tetramers but the effect of dADP is weaker. We have replaced the dADP experiment in Figure 4A with a more representative one. The main conclusion, that both nucleotides induce the formation of NrdB tetramers, is the same. The isotherms for calculating k_on values for dADP association with L. blandensis NrdB have a subtle shoulder relating to the fact that the two sites are probably not totally equivalent and the rough estimates are 3x10³ s^-1M^-1s^-1, compared to human nucleoside diphosphate kinase with k_on values for dADP in the order of 4.5x10⁶ s^-1M^-1s^-1 (Schaertl & al, J Biol Chem 1998, 273: 5662). Kinetic parameters derived from the ITC isotherm were obtained using the AFFINImeter software assuming a 1:1 interaction model, according to the method described in J. Am. Chem. Soc., 2012, 134(1), 559-565 (see eq. 4 in Suppl. Information) and Dumas et al. (2016), Methods in Enzymology, Vol. 567, Chap. 7, A.Feig Ed.

In light of this unusual finding concerning dADP regulation, it is important to carry out a comparison of the binding sites between the L. blandensis structure and the ligand-bound ATP cones from E. coli and/or human RNR. Such an analysis may reveal why dADP is not an allosteric inhibitor in E. coli or human RNR when it is in L. blandensis.

The L. blandensis ATP cone belongs to the recently-identified family of ATP cones that bind two molecules of dATP rather than only one, as in the human and E. coli enzymes. A detailed comparison of the two types of ATP cone was already made for its occurrence in the NrdA protein from Pseudomonas aeruginosa in Johansson et al., 2016. The ability to bind a second dATP molecule is due to the presence of a unique sequence motif in the C-terminal helix of the ATP cone domain that is not present in human or E. coli RNR. The interactions are essentially identical in the L. blandensis ATP cone, thus such a comparison would be superfluous in this paper. Nevertheless, one can speculate that, since the two dATP molecules bind with their triphosphate tails coordinating a Mg²⁺ ion, the loss of the γ phosphate group in dADP would have a relatively minor effect on affinity, as this group makes only one polar interaction with the protein in one of the two dATP molecules. On the other hand, this appears to be the case for the E. coli and human ATP cones as well. We hypothesize that the effect of loss of this single interaction is less when two dATP molecules are bound than when there is only one.

A comparative ITC experiment would be useful to shed light on dADP binding between species, while an anomalous scattering experiment would help solidify the identity of the metal ions observed in the L. blandensis RNR metal cluster and activity site.

Anomalous diffraction experiments have been done at three wavelengths (1.72 Å, 1.87 Å and 1.92 Å) that should help discriminate between Fe, Mn and Ca. The results suggest that the metal centre is filled with Ca²⁺ ions from the crystallization medium. This data was added to the main text. ITC studies on dADP binding to other species are beyond the scope of our current study, as we have determined dADP binding to be more than an order of magnitude worse than dATP binding and in vivo concentrations of dNDP in general are known to be low compared to dNTP concentrations.

2) Two ATP molecules are found to be bound in the activity site of L. blandensis RNR. What is the role of the second one? Based on early work with the eukaryotic RNRs, it was shown that oligomerization is induced by ATP or dATP binding – what is the driving force for oligomerization here? Please address whether the oligomerization phenomena be validated in this system using site-directed mutants.

For the new type of ATP cone observed in this enzyme, this question was already addressed quite extensively from a structural point of view in Johansson et al., 2016, where the structure of such a domain was seen for the first time in NrdA from P. aeruginosa. The domain was characterized biochemically in a previous paper from the Hofer lab (Jonna et al., 2015). Site-directed mutants were made in the P. aeruginosa NrdA interaction surface that abolished tetramerization. Since the interface residues are conserved in L. blandensis NrdB and the interactions between ATP cones are almost identical, we do not think it necessary to re-make these mutations.

Regarding the role of the second dATP molecule, this was also covered in Johansson et al., 2016. Inhibition is driven by dATP binding to both kinds of ATP cones (binding one or two dATPs). This induces conformational changes in the domain that leads them to associate, either with each other (as in yeast, human, P. aeruginosa and L. blandensis) or with NrdB (as in E. coli). The ATP-induced oligomerization is unique to eukaryotic RNRs and has not been characterized structurally. The L. blandensis type of ATP cone domain uses a different surface to oligomerize (building tetramers) than what eukaryotes or E. coli do, and interestingly this is the surface closest to the second dATP molecule.

3) Some additional explanation is required with the GEMMA analysis. In Figure 4B, the NrdBΔ99 mutant appears to form monomers, dimers, and trimers without any effector present. Please explain how this is possible. Is it due to error in the dimer and trimer molecular weights, or is there a problem with the approach? Figure 4C shows NrdA is able to form monomers and dimers in the absence of an effector. Please explain why this is – is dimerization due to mass action? In the second trace, where 50 μM of dATP is added, monomers and dimers both appear; however, the monomer is slightly more populated than the dimer. Please explain. Is there any reason, as with other RNR's (such as eukaryotic or E. coli), why the dimer population is not dominant?

The NrdBΔ99 mutant is mainly a monomer but with a slight tendency to aggregate non-specifically in the ammonium acetate buffer. A typical pattern of aggregation is a ladder of dimers, trimers, tetramers etc, which you also see in the figure. Another possibility could also be false interactions, but by comparing the analysis at many different flow-rates (different pressures), we found that this is less likely. Generally, the role of tiny peaks (now shown in brackets) should not be overemphasized in GEMMA (the trimers, tetramers etc can be regarded minor in this sense whereas the monomers and dimers are much higher). It is not surprising that the NrdA protein is in a monomer-dimer equilibrium already without effector, since there is a great variation between species. For example, the mammalian protein is a monomer and the E. coli protein is in a monomer-dimer equilibrium. The relative amounts of monomers and dimers are dependent on the protein concentration. As the reviewer has pointed out, it is a bit surprising that nucleotide effectors do not have as strong effect on dimerization as in most other species studied. All these points are now explained in the text.

4) Several assumptions have been made based on the structure in interpreting the ITC data. For example, the number of dATP binding sites is assumed to be two because that is what is seen crystallographically. This stoichiometry should be dealt in an unbiased way by fitting the data to different numbers of sites and using the fit with least error to select the correct model.

N values in ITC fits are strongly affected by the content of active protein material as opposed to total protein concentration. In all the nucleotide titrations the fitted apparent N value was significantly above 1 when total protein concentration was used. However, for the N value to be meaningful in terms of stoichiometry it has to be an integer. Using N=1 for dATP binding the active protein concentration becomes more than 100% of the total concentration, whereas using N=2 gives an active protein concentration of 60%, in our experience a typical fraction of active RNR (the error of the fit will not change whether the N value or the active protein concentration is fixed). The proportion of inactive protein can be explained by e.g. aggregation as observed by SEC-MALS and DLS. Moreover, our X-ray crystallography shows a stoichiometry of 2 bound dATP molecules. For meaningful results, we therefore used the active protein concentration for calculation of ATP and dADP binding stoichiometries.

5) Comparison of the ΔH values between the different ligands gives a rather surprising value for dADP, which is at least 24 kJ/mol higher than the dATP. This difference implies that several additional hydrogen bonds, ion-pairs, and/or van der Waals interactions are made by dADP compared to dATP. Yet the structure indicates that dATP makes only a single an additional hydrogen bond compared to the dADP. This discordance between the ITC and structural data should be explained.

The overall enthalpy of the reaction, measured by calorimetry, is affected not only by direct protein interaction with the ligand, but also by water molecules in the binding pocket (Klebe, Nature Rev Drug Discovery 2015, 14: 95). Since calorimetry measures the overall effect of all changes in the binding pocket, it is hard to correlate and compare enthalpies without structural data for the dADP bound complex.

6) The significance of the work should be better clarified for a general (non-RNR) audience. Given the wealth of important enzyme systems in biology, why are the current findings important for advancing our understanding of broad issues, such as allosteric regulations, the appropriate control of nucleotide levels to avoid disease states, etc. Along these lines, the discussion of the potential applications of the findings is brief and unsatisfying. It is claimed that the new observations offer a possible route to regulate engineered enzymes by dATP inhibition. Please elaborate on this concept and how it might work.

We have added a sentence to the Abstract, a paragraph to the Discussion, plus a few references in the Introduction, describing the degree of novelty of our finding. That allosteric regulation controlled by modular domains is evolutionarily dynamic has been shown before (by us and others), but we are not aware of other enzyme complexes in which a domain has transferred to a non-homologous component. We would like to argue that this, in addition to the general importance of ribonucleotide reduction and its control in all organisms, makes our finding interesting to a broad audience of biologists interested in evolution of proteins, nucleotide metabolism and enzyme structure and function. In addition, we have added a few sentences at the end of the fourth paragraph in the Introduction describing briefly that RNRs are potential targets both for antibacterial and antitumor therapies. We have also added a sentence to the Introduction describing in some more detail how we foresee that ATP-cones could be used to confer regulation of engineered enzymes by control of nucleotide concentrations, at least for enzymes that are sensitive to oligomerization states. Since this is not within our direct area of expertise, we fear that further detail would become too speculative.