Dynamic allostery in substrate binding by human thymidylate synthase

Curation statements for this article:
  • Curated by eLife

    eLife logo

    Evaluation Summary:

    The authors analyze the mechanisms of entropically driven cooperativity in the human thymidylate synthase (hTS), an enzyme essential for DNA replication and a promising target for anticancer drugs. The authors conclude that the cooperative binding of dUMP ligands to its two identical sites arises from a disproportionate reduction in the enzyme's conformational entropy upon binding the first ligand. The results provide rare insights into the mechanisms of ligand binding for an essential human protein and should be of great interest to readers interested in enzyme structure/dynamics/function relationships, cooperativity and allostery, and possible drug targeting of thymidylate synthase.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)”

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Human thymidylate synthase (hTS) is essential for DNA replication and therefore a therapeutic target for cancer. Effective targeting requires knowledge of the mechanism(s) of regulation of this 72 kDa homodimeric enzyme. Here, we investigate the mechanism of binding cooperativity of the nucleotide substrate. We have employed exquisitely sensitive methyl-based CPMG and CEST NMR experiments enabling us to identify residues undergoing bifurcated linear 3-state exchange, including concerted switching between active and inactive conformations in the apo enzyme. The inactive state is populated to only ~1.3%, indicating that conformational selection contributes negligibly to the cooperativity. Instead, methyl rotation axis order parameters, determined by 2 H transverse relaxation rates, suggest that rigidification of the enzyme upon substrate binding is responsible for the entropically-driven cooperativity. Lack of the rigidification in product binding and substrate binding to an N-terminally truncated enzyme, both non-cooperative, support this idea. In addition, the lack of this rigidification in the N-terminal truncation indicates that interactions between the flexible N-terminus and the rest of the protein, which are perturbed by substrate binding, play a significant role in the cooperativity—a novel mechanism of dynamic allostery. Together, these findings yield a rare depth of insight into the substrate binding cooperativity of an essential enzyme.

Article activity feed

  1. Author Response:

    Evaluation Summary:

    The authors analyze the mechanisms of entropically driven cooperativity in the human thymidylate synthase (hTS), an enzyme essential for DNA replication and a promising target for anticancer drugs. The authors conclude that the cooperative binding of dUMP ligands to its two identical sites arises from a disproportionate reduction in the enzyme's conformational entropy upon binding the first ligand. The results provide rare insights into the mechanisms of ligand binding for an essential human protein and should be of great interest to readers interested in enzyme structure/dynamics/function relationships, cooperativity and allostery, and possible drug targeting of thymidylate synthase.

    We would like to add that the disproportionate reduction in conformational entropy is entirely dependent on the presence of the flexible N-terminus, even though the N-terminus itself undergoes no detectable change in conformational entropy.

    Reviewer #1 (Public Review):

    Human thymidylate synthase (hTS) is relatively large for NMR standards (~72 kDa dimer) and so the authors use a battery of advanced, TROSY-based NMR experiments to investigate the structure and conformational dynamics of the enzyme in multiple binding states. In particular, they have acquired multiple and single quantum methyl CPMG and CEST data to probe us-ms dynamics. These experiments showed that hTS undergoes exchange between active and inactive conformations. Analysis of residual dipolar couplings and chemical shift perturbation experiments indicated that the major conformational state revealed by CPMG and CEST corresponds to the active hTS conformation. This finding suggests that conformational selection is not the primary mechanism mediating cooperativity in hTS.

    To investigate if binding cooperativity in hTS is due to modulation of conformational entropy upon ligand binding, the authors have investigated ps-ns dynamics in hTS by means of 2H relaxation measurements. These measurements suggest that rigidification of the protein upon the first binding event is the primary origin of cooperativity in the hTS dimer. Indeed, acquisition of control experiments on systems that do not show binding cooperativity (i.e., the complex formed by dUMP with N-terminal truncated hTS and the complex formed by TMP with full-length hTS) do not show the same modulation of conformational entropy observed upon formation of the dUMP-hTS complex. Overall, I found this manuscript interesting and well-written. I found particularly fascinating the observation that cooperativity is driven by modulation of conformational disorder in the unstructured N-terminal tail, which is not directly involved in ligand binding. The experimental approach and analysis protocols are sound and the conclusions are well supported by the experimental data.

    We appreciate the positive comments. We note that the last statement about cooperativity is slightly misleading, as it is the presence of the N-terminal tail that enables modulation by conformational entropy, even though the entropy “at play” appears not to be in the tail itself.

    Reviewer #2 (Public Review):

    The principal objective of this work is to detail the basis for the enzyme's observed cooperative binding to dUMP, which was reported by the authors in a previous publication (Bonin et al. 2019 Biophys J). That paper showed (via ITC) that the binding of dUMP ligands to the protein's two identical sites cannot be explained by a simple thermodynamic model with a single affinity, but rather requires a cooperative model in which the second binding event is more favorable by 1.3 kcal/mol (~2RT), due in part to a much more favorable entropy change -TΔS. In this paper, the authors set out to test two possible cooperativity models consistent with that observation: (1) that binding of the first ligand results in stabilization of a binding-competent conformation (conformational selection), or (2) that a broad reduction in protein dynamics (conformational entropy ΔS_conf) upon binding the first ligand results in a smaller ΔS_conf penalty for binding the second ligand, and therefore a more favorable ΔG.

    The authors perform an extensive series of sophisticated NMR experiments using a range of samples with specialized labeling patterns, particularly ILV methyl-13C, and ILV-methyl-13C-HD_2. These labeling patterns allow the investigators to record high-quality methyl NMR spectra on the large ~70 kDa hTS dimer, its Δ25 N-terminal deletion, alone and in complex with dUMP and dTMP. Insights into exchange dynamics come from 13C methyl CPMG and CEST relaxation measurements, which are sensitive to motions on timescales spanning µs to ms. Insights into ps-time scale dynamics and conformational entropy come from methyl-2H_1H_2 relaxation measurements, and are extrapolated using an empirical "entropy meter". Structural insights are obtained from measured and predicted amide 1H-15N residual dipolar couplings, solvent PRE measurements, and chemical shift perturbations.

    The structural context is largely framed by prior reports of hTS crystallizing in two distinct conformations, termed 'active' and 'inactive' (Chen et al. 2017). These states are described (page 3) as differing by the conformations of an active site loop. The authors posit that if the enzyme is exchanging between these two states, with the 'inactive' state being dominant, binding of a first dUMP to the enzyme will shift the population towards the 'active' state, therefore favoring an additional dUMP binding event. The structural differences between the 'active' and 'inactive' states are not well described, however, and since the enzyme must bind both the substrate dUMP and its co-substrate MTHF, it's not entirely clear why this is a reasonable premise. RCSB coordinates 5X5A and 1YPV are used as the reference structures for the 'active' and 'inactive' states, respectively. Computed RDC data (Fig. 4) indicate that they are quite different, but it would be helpful to have a description of the differences in the structures, why it is reasonable to hypothesize that one or more of them might have different affinities for dUMP, and how the sampling of the other state might be manifest in the subsequent NMR data.

    We agree that the manuscript would be improved if there is a more detailed description of the existing structural states for hTS. These structures correspond to apo and dUMP-bound hTS. In TS, in general the dUMP-bound conformations are generally highly similar to those with nucleotide and cofactor (or cofactor analog) both bound in the active site.

    The authors indeed observe strong dispersions in methyl CPMG relaxation data for ligand-free hTS (Fig. 2), and more than one dip in CEST profiles (Fig. 1). However, these data (esp. CPMG) are not well described by a global exchange process with a single set of rate constants and populations, indicating a more complex exchange between three or more states (Fig S1, S2). (This point could be better described - the authors conclude that the data do not fit a 2-state model, but it would be helpful to describe in the main text the analysis that brought them to that conclusion.) Since the data are not well described by a two-state model, the authors fit the data to a three-state "BAC" model, in which the major state A exchanges with two other states, B and C; the A-B exchange is referred to as "slow" (~240/s) and the A-C exchange as "fast" (>2000/s)(Fig 2). It could be clearer why that model is preferred over alternative three-state models.

    We appreciate the Reviewer’s accurate summary of the CPMG and CEST NMR relaxation on hTS. The point is well-taken that the steps taken during the fitting trials were not described in detail, and we will expand on the descriptions to make the process that led to the BAC model more clear.

    The authors compare backbone amide RDCs measured from the major state of the enzyme and its complex with dUMP with RDCs computed from the crystallographic "active" and "inactive" structures (Fig. 4). On the basis of its better agreement, they conclude that the major state is the "active" conformation. This may be a reasonable conclusion but merits additional discussion. Why are the predicted RDCs so different if the conformations only differ in a loop (as described on page 3)? Were the same alignment tensors obtained from the structural analysis? Provided the major state in solution is the "active" conformation, they conclude that they can rule out the conformational exchange mechanism of allostery. Again, this might be a reasonable interpretation, but it would be strengthened by describing the evidence that the crystallographically observed "active" and "inactive" conformations will have different affinities for dUMP.

    While the active site loop is one region of the protein that is dramatically different in the active and inactive structures, there are differences between the two beyond this loop. We will expand upon the description of these differences to clarify this point. We will also add details supporting the idea that the two conformations have different affinities for dUMP.

    The authors further examine differences in the methyl spectra between full-length hTS, which exhibits cooperative dUMP binding by ITC, and the Δ25 mutant, which does not (Figure 7). Since the methyl spectra are nearly superimposable, they conclude that the N-terminal region does not perturb the structure, though it is responsible for the observed cooperative behavior. Again, this might be a reasonable interpretation, but it is tempered by the inherent limitations of the observables, as the spectra only reflect the structure experienced by the labeled methyl groups, so the data are silent about other areas of the protein that might reflect structural changes.

    This is true that the NMR probes are not used for every atom of the protein, and that methyl groups are in discrete locations of the structure. Nevertheless, the amide NH spectra contain information from nearly all residues, and we believe that methyl density is sufficient to draw basic structural inferences. We note that while there are fewer probes at the dimer interface (low density of methyls and lack of amides from extremely slow back exchange), significant structural changes there should be sufficient to detect chemical shift changes at nearby observable probes.

    Having ruled out structural changes and conformational exchange as responsible for the cooperative behavior, the authors quantify the intrinsic conformational entropy of the enzyme. They use the 2H relaxation rates of suitably labeled methyl groups to compute the magnitude of the order parameter S^2 of each labeled methyl axis. They compute the change in conformational entropy ΔS_conf using the change in S^2 for each methyl as a proxy, and an empirically-derived "entropy meter" (Fig. 5). From this analysis they find a larger 'unfavorable' entropy change upon binding dUMP than to TMP, meaning that a larger reduction in conformational entropy is associated with cooperative binding. The reason is that if more than half of this entropic penalty is paid upon binding the first ligand, the second binding event can occur with a smaller entropy penalty and thus a more favorable affinity. These are not unreasonable conclusions; however, there are significant uncertainties in the data and the underlying assumptions. At a minimum, these uncertainties should be considered and discussed.

    We agree that the entropy meter is a method for estimation of entropy, with uncertainties. However, the method has been shown to be useful for a large set of proteins (and widely adopted) and gives an overall sense of conformational entropic effects. We have been rigorous about measurements of error in the 2H relaxation data, and in fact switched to 2H relaxation in CHD2 groups after determining from our own data that methyl 1H relaxation in CH3 groups appears to be less reliable. In the end, the trends presented from use of the entropy meter are also easily observed from changes in the raw methyl axis order parameters.

    The ΔS_conf conclusion at which the authors arrive is unfortunately mechanistically uninformative. In a statistical mechanical sense, a reduction in entropy arises from a reduction in accessible conformational states. Might one quantify the states that are excluded upon ligand binding, and one might gain an understanding of the link between structure (ensembles) and thermodynamics. The "entropy meter" approach is not informative about 'which' states are lost, only that a reduction in disorder, extrapolated over the full protein, is associated with a bulk change in entropy.

    It would be nice to have the ability to identify such specific microscopic and transient states, but the current state of NMR cannot provide such a high level of this kind of conformational detail. We would like to reiterate the point that through the experimental strategy, we were able to identify the flexible N-terminus as a key element (i.e. mechanism) in the entropy effect underlying cooperativity in hTS dUMP binding.

  2. Evaluation Summary:

    The authors analyze the mechanisms of entropically driven cooperativity in the human thymidylate synthase (hTS), an enzyme essential for DNA replication and a promising target for anticancer drugs. The authors conclude that the cooperative binding of dUMP ligands to its two identical sites arises from a disproportionate reduction in the enzyme's conformational entropy upon binding the first ligand. The results provide rare insights into the mechanisms of ligand binding for an essential human protein and should be of great interest to readers interested in enzyme structure/dynamics/function relationships, cooperativity and allostery, and possible drug targeting of thymidylate synthase.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)”

  3. Reviewer #1 (Public Review):

    Human thymidylate synthase (hTS) is relatively large for NMR standards (~72 kDa dimer) and so the authors use a battery of advanced, TROSY-based NMR experiments to investigate the structure and conformational dynamics of the enzyme in multiple binding states. In particular, they have acquired multiple and single quantum methyl CPMG and CEST data to probe us-ms dynamics. These experiments showed that hTS undergoes exchange between active and inactive conformations. Analysis of residual dipolar couplings and chemical shift perturbation experiments indicated that the major conformational state revealed by CPMG and CEST corresponds to the active hTS conformation. This finding suggests that conformational selection is not the primary mechanism mediating cooperativity in hTS.

    To investigate if binding cooperativity in hTS is due to modulation of conformational entropy upon ligand binding, the authors have investigated ps-ns dynamics in hTS by means of 2H relaxation measurements. These measurements suggest that rigidification of the protein upon the first binding event is the primary origin of cooperativity in the hTS dimer. Indeed, acquisition of control experiments on systems that do not show binding cooperativity (i.e., the complex formed by dUMP with N-terminal truncated hTS and the complex formed by TMP with full-length hTS) do not show the same modulation of conformational entropy observed upon formation of the dUMP-hTS complex. Overall, I found this manuscript interesting and well-written. I found particularly fascinating the observation that cooperativity is driven by modulation of conformational disorder in the unstructured N-terminal tail, which is not directly involved in ligand binding. The experimental approach and analysis protocols are sound and the conclusions are well supported by the experimental data.

  4. Reviewer #2 (Public Review):

    The principal objective of this work is to detail the basis for the enzyme's observed cooperative binding to dUMP, which was reported by the authors in a previous publication (Bonin et al. 2019 *Biophys J*). That paper showed (via ITC) that the binding of dUMP ligands to the protein's two identical sites cannot be explained by a simple thermodynamic model with a single affinity, but rather requires a cooperative model in which the second binding event is more favorable by 1.3 kcal/mol (~2RT), due in part to a much more favorable entropy change -TΔS. In this paper, the authors set out to test two possible cooperativity models consistent with that observation: (1) that binding of the first ligand results in stabilization of a binding-competent conformation (conformational selection), or (2) that a broad reduction in protein dynamics (conformational entropy ΔS_conf) upon binding the first ligand results in a smaller ΔS_conf penalty for binding the second ligand, and therefore a more favorable ΔG.

    The authors perform an extensive series of sophisticated NMR experiments using a range of samples with specialized labeling patterns, particularly ILV methyl-13C, and ILV-methyl-13C-HD_2. These labeling patterns allow the investigators to record high-quality methyl NMR spectra on the large ~70 kDa hTS dimer, its Δ25 N-terminal deletion, alone and in complex with dUMP and dTMP. Insights into exchange dynamics come from 13C methyl CPMG and CEST relaxation measurements, which are sensitive to motions on timescales spanning µs to ms. Insights into ps-time scale dynamics and conformational entropy come from methyl-2H_1H_2 relaxation measurements, and are extrapolated using an empirical "entropy meter". Structural insights are obtained from measured and predicted amide 1H-15N residual dipolar couplings, solvent PRE measurements, and chemical shift perturbations.

    The structural context is largely framed by prior reports of hTS crystallizing in two distinct conformations, termed 'active' and 'inactive' (Chen et al. 2017). These states are described (page 3) as differing by the conformations of an active site loop. The authors posit that if the enzyme is exchanging between these two states, with the 'inactive' state being dominant, binding of a first dUMP to the enzyme will shift the population towards the 'active' state, therefore favoring an additional dUMP binding event. The structural differences between the 'active' and 'inactive' states are not well described, however, and since the enzyme must bind both the substrate dUMP and its co-substrate MTHF, it's not entirely clear why this is a reasonable premise. RCSB coordinates 5X5A and 1YPV are used as the reference structures for the 'active' and 'inactive' states, respectively. Computed RDC data (Fig. 4) indicate that they are quite different, but it would be helpful to have a description of the differences in the structures, why it is reasonable to hypothesize that one or more of them might have different affinities for dUMP, and how the sampling of the other state might be manifest in the subsequent NMR data.

    The authors indeed observe strong dispersions in methyl CPMG relaxation data for ligand-free hTS (Fig. 2), and more than one dip in CEST profiles (Fig. 1). However, these data (esp. CPMG) are not well described by a global exchange process with a single set of rate constants and populations, indicating a more complex exchange between three or more states (Fig S1, S2). (This point could be better described - the authors conclude that the data do not fit a 2-state model, but it would be helpful to describe in the main text the analysis that brought them to that conclusion.) Since the data are not well described by a two-state model, the authors fit the data to a three-state "BAC" model, in which the major state A exchanges with two other states, B and C; the A-B exchange is referred to as "slow" (~240/s) and the A-C exchange as "fast" (>2000/s)(Fig 2). It could be clearer why that model is preferred over alternative three-state models.

    The authors compare backbone amide RDCs measured from the major state of the enzyme and its complex with dUMP with RDCs computed from the crystallographic "active" and "inactive" structures (Fig. 4). On the basis of its better agreement, they conclude that the major state is the "active" conformation. This may be a reasonable conclusion but merits additional discussion. Why are the predicted RDCs so different if the conformations only differ in a loop (as described on page 3)? Were the same alignment tensors obtained from the structural analysis? Provided the major state in solution is the "active" conformation, they conclude that they can rule out the conformational exchange mechanism of allostery. Again, this might be a reasonable interpretation, but it would be strengthened by describing the evidence that the crystallographically observed "active" and "inactive" conformations will have different affinities for dUMP.

    The authors further examine differences in the methyl spectra between full-length hTS, which exhibits cooperative dUMP binding by ITC, and the Δ25 mutant, which does not (Figure 7). Since the methyl spectra are nearly superimposable, they conclude that the N-terminal region does not perturb the structure, though it is responsible for the observed cooperative behavior. Again, this might be a reasonable interpretation, but it is tempered by the inherent limitations of the observables, as the spectra only reflect the structure experienced by the labeled methyl groups, so the data are silent about other areas of the protein that might reflect structural changes.

    Having ruled out structural changes and conformational exchange as responsible for the cooperative behavior, the authors quantify the intrinsic conformational entropy of the enzyme. They use the 2H relaxation rates of suitably labeled methyl groups to compute the magnitude of the order parameter S^2 of each labeled methyl axis. They compute the change in conformational entropy ΔS_conf using the change in S^2 for each methyl as a proxy, and an empirically-derived "entropy meter" (Fig. 5). From this analysis they find a larger 'unfavorable' entropy change upon binding dUMP than to TMP, meaning that a larger reduction in conformational entropy is associated with cooperative binding. The reason is that if more than half of this entropic penalty is paid upon binding the first ligand, the second binding event can occur with a smaller entropy penalty and thus a more favorable affinity. These are not unreasonable conclusions; however, there are significant uncertainties in the data and the underlying assumptions. At a minimum, these uncertainties should be considered and discussed.

    The ΔS_conf conclusion at which the authors arrive is unfortunately mechanistically uninformative. In a statistical mechanical sense, a reduction in entropy arises from a reduction in accessible conformational states. Might one quantify the states that are excluded upon ligand binding, and one might gain an understanding of the link between structure (ensembles) and thermodynamics. The "entropy meter" approach is not informative about 'which' states are lost, only that a reduction in disorder, extrapolated over the full protein, is associated with a bulk change in entropy.

  5. Reviewer #3 (Public Review):

    The manuscript by Bonin et al describes the use of sensitive methyl-based nuclear magnetic resonance (NMR) spectroscopy methods to characterize the mechanism of substrate binding in human thymidylate synthase. Using NMR experiments that probe protein motions on the micro-to-millisecond timescale, they first show that the activation loop and substrate binding regions undergo conformational exchange between active and inactive states and in some cases a previously undescribed third state. Using a variety of NMR methods, the authors show that the active state is the dominant (>98%) conformer in the solution. With a small population (<2%), the inactive state cannot contribute significantly to cooperative substrate binding. Then using side-chain methyl-group relaxation methods, which are sensitive to protein motions on the nano-to-picosecond, they determine that thymidylate synthase becomes more rigid upon substrate binding; a result that is consistent with an entropic, dynamic version of cooperativity. Through the use of mutants and substrate-bound forms of the enzyme, the authors are able to trace the origin of this dynamic allostery to the unstructured N-terminus of the protein, which is lacking in bacterial versions of the enzyme and does not show the same cooperativity. In short, this work highlights the power of solution state NMR for the use of studying protein motions over 9 orders of magnitude in time and how these motions contribute to the thermodynamics and regulation of enzymatic activity.