Previously published results suggested that the structural dynamics of human dUTPase is intimately linked to its biological function: The flexible C-terminal arm is necessary for catalytic activity, as the deletion mutant is inactive, see Fig. 1 and [16]. The closed form has been shown to close the active site, restrict water access by stacking F158 over the substrate and participate in the coordination of the tri-phosphate [13]. Detailed x-ray analysis using the modified substrate, dUpNHpp, capturing the slowed down reaction in multiple conformations [17], together with enzyme kinetic studies, has further suggested a slow but not rate limiting structural isomerization step in the 20 Hz time scale prior to catalysis [12].
Apo form enzyme dynamics
The structural dynamics was first assessed by recording the temperature dependence of the chemical shifts of the apo form of the enzyme. 15N-HSQC spectra were recorded at 3K intervals between 277 and 310 K and superimposed. Examples of the temperature induced chemical shift drifts of the residues of the conserved motifs 2–3 and for the C-terminal arm are shown in Fig. 2.
The temperature dependent behavior of the NH resonances highlights that we have indeed structural dynamics in several time scales for the apo form. For residues E152-R153 the resonances appear to split into two individually detectable populations at lower temperatures, that coalesce close to 300 K. For many residues in the C-terminal arm, for example G154, G156, G157, F158, G159, T160 and T161, the intensity of the resonances decrease as the temperature is increased. In contrast, for residues near motives 2–3, for example L88, A89, A90, K91, H92, I94 and D95, the signal intensities instead decrease as the temperature is decreased. Other residues are unassignable close to these regions, most notably residues 84–87 are not possible to detect or assign in any studied preparation of the enzyme, including different ligands and different temperatures. These observations confirm the presence of substantial slow motions, above-, at-, below- and across the NMR time scale for the apo form of human dUTPase. The weighted cumulative effect of the chemical shift perturbation as a function of temperature is reported in Fig. 3, using the square root of the sum of the chemical shift perturbation square averages, using a scaling factor, α, of 0.14 for the amide nitrogen atoms and 0.35 for the carbonyl carbons on the backbone as estimated from their average standard deviations in the BMRB chemical shift database.
To get more insight into the structural stability of the apo form enzyme and evaluate the 1H-15N amide bond order parameters (S2), the R1, R1ρ and R2 relaxation rates, together with the steady-state 15N-NOE were determined and data presented on Fig. 4. The C-terminal arm of the apo form, housing the strictly conserved Motif 5 (Fig. 1, and [33]), displayed complex dynamics, involving several different time scales.
Most notably, there is a pronounced increase in the R1 and a corresponding decrease in R1ρ/R2 for residues 148–164 compared to the well-folded parts of the protein complex, which reflects an increased flexibility and fast internal motion in the nano- to picosecond scale of the C-terminal arm (Fig. 4a-b). The flexibility can also be observed by the steady-state NOE parameters, that drop from close to 1 for the well folded regions to negative values for the fully flexible residues of the C-terminal arm. The disordered nature of the C-terminal has earlier been qualitatively observed with both NMR [10] and crystallography [34], but herein its disorder is for the first time quantitatively described in solution. The increased flexibility of the C-terminus is reflected as a sharp drop in the fitted generalized order parameters, s2 (Fig. 4d-e).
In the apo form, negative 15N-NOE were confirmed for residues G154, G156, G157, G159, G162, K163 and N164, meaning that these residues are highly flexible, having effective correlation times reminiscent of a small molecule in solution rather than inheriting the slow tumbling of the protein complex with a rotational correlation time of approximately 28 ns (Fig. 4c). It should be noted that the interspersed residues in the C-terminal stretch from L148 to N164 produced weak resonances that were difficult to integrate accurately. Together with the temperature dependent observations above, we conclude that the C-terminal arm is highly flexible in terms of internal rotations. Importantly, the spectral properties of the C-terminal are also greatly affected by the arms participation in a slow, more concerted motion in the millisecond regime, plausibly an opening and closing of the C-terminal arm relative to the core of the enzyme.
There is a more subtle but noteworthy increase in both R1 and R1ρ/R2 for the stretch of residues between 78 and 104, who are involved in substrate recognition (including the strictly conserved Motif 2 and half Motif 3). There is also a moderate increase in R1 for residues 40–46 (including the strictly conserved Motif 1). The increase in R1 generally indicates increased motions in the nanosecond regime affecting the order parameters while there is a simultaneous increase in R1ρ/R2 (instead of a decrease which would have been expected for a highly flexible region). This leads to the conclusion that there are additional motions that contribute to the spectral density functions in the nanosecond regime with additional slow motions closer to the NMR time scale (in the millisecond regime), that contribute to line broadening (and R2). This is supported by the absence of response in the 15N-NOE parameter for these regions [35]. A heatmap highlighting the most disordered parts of the enzyme is presented in Fig. 5. Interestingly, this disorder is not restricted to motifs directly involved in substrate binding.
The complex dynamics of the enzyme was expected to manifest in relaxation dispersions, an experiment that is sensitive to exchange processes in the millisecond regime. Relaxation dispersion data acquired at 600 and 850 MHz did indeed display slow dynamics for the backbone of a large number of residues. Analysis in Nessy, using both datasets simultaneously, achieved the best fit to a slow dynamics model (model 7 in the Nessy documentation) over the model devoid of slow dynamics (model 1 in the Nessy documentation) for 46 residues evenly distributed over the sequence, out of the 122 residues that could be reliably integrated. A slightly more conservative report from analyzing only the 600 MHz dataset, which was the highest quality dataset, in PINT is reported in Fig. 4f.
In conclusion, the dynamic picture of the apo form enzyme reveals an enzyme that is relatively active in pulsating and “breathing” motions, adopting the conformation of the open active site for substrate binding.
Substrate binding
The addition of 1.2 molar equivalents of the modified “uncleavable” substrate, dUpNHpp, results in dramatic changes to the spectral quality as well as the chemical shifts of the protein resonances (Fig. 6a-d). The heterogenic change is not concentration dependent, and addition of up to 4 molar equivalents does not further affect the spectrum (Figure S1 in the SI). The assignments of the apo form could be successfully transferred to the dUpNHpp holo form, and when doing so the assignment was annotated to the major resonance for the heterogenic peak.
The substrate binding caused a slowing down of the slow NMR time scale dynamics observed for the apo form, into the milliseconds to seconds regime where multiple sub-conformations became individually observable. This was reflected by the Rex contributions to the relaxation dispersions observed for the apo form were abolished upon substrate binding (Fig. 7e), except for the single residue, Q146.
The binding of the modified substrate had substantial effects on the relaxation parameters, as shown in the difference plots in Fig. 7. The clearest change is a stabilization of the C-terminal arm, reducing the degree of negative NOE for residues G156-N164. These residues interact with the triphosphate of the substrate, partially stabilizing the closed conformation of the C-terminal arm. There was also a systematic reduction of R2 upon substrate binding, likely attributed to the reduction of Rex contributions from slow conformational exchange, as the enzyme backbone is stabilized. The differences in the order parameters (Fig. 7A-D) had a few hotspots along the sequences that the generated difference plots didn’t necessarily captures because of some residues being undetectable after substrate binding, thereby giving a zero contribution to the plot. For example, the R1 of the C-terminal arm is substantially changed for residues L148, D149, T151, G154, G156, G159 and K163, but in addition to that, residues S155, F158, S160 and T161 become undetectable, and does therefore not contribute to the difference plot. Two residues, G154 and G156, go against the trend and are slower for the apo form, but these resonances are also of very poor quality in the substrate bound form with big uncertainties in the integration. For example, the R1 of G154 is 1.22 ± 0.06 for apo and 1.45 ± 0.23 for holo. The R2 rates follows the same trend for the C-terminal, and together with the NOEs consistently shows that the C-terminal arm behaves more like the well-folded core of the protein in the substrate bound form than in the apo form, reflecting that the closed form of the C-terminal arm is stabilized by substrate binding.
One more patch shows changes in both R1 and R2, comprising residues A98, V100, E103, with G99, I101 and D102 becoming undetectable upon binding. In this case however, both the R1 and R2 rates become slower upon substrate binding, suggesting multiple time scales being involved. The R2 heatmap largely corresponds to the relaxation dispersion heatmap, in that the Rex contributions from the slow ms-µs motions are lost when the substrate stabilize the enzyme. This patch is directly involved in the substrate-protein interactions and is also part of conserved motif 3.
The raw relaxation parameters are supplied in the Supporting information, Figure S2.
Conserved motifs
It is well established that dynamics play a role in ligand recognition [36]. Generally, if certain types of dynamics are important for function, one would expect these to coincide with conserved parts within a protein family. One finds examples of such conservation in e.g. the NAGK protein family, where normal mode analysis indicates that slow movements of key residues are conserved [37]. Recent relaxation dispersion experiments have also successfully informed which residues are important for both function and dysfunction where dynamics are key [38, 39]. One interesting observation is that the conserved Motifs 1–5 are all represented in the relaxation dispersion heatmap, reflecting that these conserved patches often display structural dynamics in the ms-µs timescale (Fig. 8). The secondary structure content of these motif includes mostly beta-strands (Motif 1 and 4), helical (Motif 2) and mostly loops (Motif 3 and 5). More generally, the unstructured C-terminal contains two conserved motifs, Motif 4 and Motif 5, the latter which comprise residues participating in binding interactions with the triphosphate of the substrate. Notably, there is a sequence segment, VVKTDI in the human sequence that resides between Motif 1 and Motif 2 which displays the characteristic ms-µs dynamics, and also has a quite high conservation. Conversely, many structured parts of the sequence are not conserved, and does not take part in slow dynamics. Taken together, this strongly suggests that essential features of dUTPase is correlated to its structural dynamics, both in the binding and release of substrate, and in the structural rearrangements involved in catalysis.
Asymmetry between subunits and homotropic allosterism
It has previously been suggested that dUTPase displaying non-linear spectral response to ligand titration is the result of homotropic allosterism [10], where a binding event in one subunit cause a change in ligand affinity in another subunit of the homotrimer. This has later been convincingly refuted by the construction of chimera trimers [16]. A possible explanation for the observed unexpectedly efficient quenching of the NMR signal intensity upon ligand binding is that slow exchange between bound and free ligands, or slow exchange between conformations in the bound state, contribute to the relaxation by a Rex term, making the system deviate from a 2-state model and make the binding sites appear not being independent of each other. The observed heterogeneity upon dUpNHpp binding displayed several species in complex ratios (Fig. 6a-c) but does not imply allosterism. The observed heterogeneity is not concentration dependent, which means that the observed heterogeneity represents binding saturation at equilibrium. The most plausible explanation for this observation is that the protein gets slowed down while also becoming trapped by the uncleavable substrate in the conformational isomerization step, previously suggested in the kinetic model by Toth et al. [12]. This would make multiple species individually observable on the NMR time scale, in populations that results from the k1 and k-1 of the conformational equilibrium prior to the catalytic step, together with the possible combinations of states in the three subunits. The kinetic model suggests 21.2 Hz for the forward process and 3.7 Hz for the backward process. If we approximate that to a 5:1 ratio for simplicity and combine it with the possible eight states of the three subunits, that reduces to 1:3:3:1, we get a combination of 1:3:3:1 with 5:1, which would translate to 1:12:48:64 probabilities of each population. In other words, we would expect to find two major peaks, one slightly more intense than the other, and one, possibly two, minor peaks, one being very weak, per residue if the spectra corresponded to the expected populations based on the kinetic model where the chemical step has been blocked. Several residues in Fig. 6a-c indeed fits this pattern, for example residues G80, A83, V109 and G126. This lends a different kind of support to the suggested kinetic model, while also providing a plausible explanation for the observed heterogeneity upon dUpNHpp binding.
Hydration
The stabilization of the C-terminal arm as well as several core residues can also be observed as a reduction in amide proton exchange with the bulk water, detected by the CLEANEX-PM experiment (Fig. 9). This is the result of these residues being less exposed to the bulk solvent, again reflecting a stabilization of the closed form by the substrate binding. The same observation can be made for residues A83, A89, A90, Y105, R106 and V112, in the active site, belonging to the conserved motifs 2 and 3, as well as D127, R128 and E135 belonging to the conserved motif 4.
Changes in the dipole-dipole ROESY peaks to the bulk water, with opposite sign to the bulk water exchange peaks, were also observed (Fig. 10). In the presence of the modified substrate, dipole-dipole ROEs could be detected for residues G97, A98 and V100, which were either absent or weaker for the apo form. These residues’ amide protons are close in space to the catalytic water that is coordinated by V100 and the sidechain of the strictly conserved D102. This suggests that not only the enzyme backbone dynamics is stabilized by substrate binding, but also some of the bound water molecules increase their residence time.