Achieving pure spin effects by artifact suppression in methyl adiabatic relaxation experiments

Abstract

Recent methyl adiabatic relaxation dispersion experiments provide examination of conformational dynamics across a very wide timescale (10² – 10⁵ sec⁻¹) and, particularly, provide insight into the hydrophobic core of proteins and allosteric effects associated with modulators. The experiments require efficient decoupling of ¹H and ¹³C spin interactions, and some artifacts have been discovered, which are associated with the design of the proton decoupling scheme. The experimental data suggest that the original design is valid; however, pulse sequences with either no proton decoupling or proton decoupling with imperfect pulses can potentially exhibit complications in the experiments. Here, we demonstrate that pulse imperfections in the proton decoupling scheme can be dramatically alleviated by using a single composite π pulse and provide pure single-exponential relaxation data. It allows the opportunity to access high-quality methyl adiabatic relaxation dispersion data by removing the cross-correlation between dipole-dipole interaction and chemical shift anisotropy. The resulting high-quality data is illustrated with the binding of an allosteric modulator (G2BR) to the ubiquitin conjugating enzyme Ube2g2.

Summary:

Recent methyl adiabatic relaxation dispersion experiments provide examination of conformational dynamics across a very wide timescale (10² – 10⁵ sec⁻¹) and, particularly, provide insight into the hydrophobic core of proteins and allosteric effects associated with modulators. The experiments require efficient decoupling of ¹H and ¹³C spin interactions. Here, we demonstrate that pulse imperfections in the proton decoupling scheme can be dramatically alleviated by using a single composite π pulse and provide pure single-exponential relaxation data. The resulting high-quality data is illustrated with the binding of an allosteric modulator (G2BR) to the ubiquitin conjugating enzyme Ube2g2.

Relaxation dispersion experiments in solution-state NMR spectroscopy are one of the most important tools in studying biologically relevant conformational dynamics of macromolecules^1–2. These experiments have provided critical insights in many biological systems^3–7. The well-established experiments^8–11, as well as their analytic solutions^12–14, provide a powerful way to reveal and quantitate minor states that have been demonstrated to play functional roles in many macromolecules. Recently, a new type of relaxation dispersion technique^15–16, adiabatic relaxation dispersion, has been developed, which provides an alternative way to reveal the information for conformational exchange and access a broader timescale^16–17. Studies applied to methyl groups indicated that revamped proton decoupling schemes are needed to improve the experiments¹⁷. Both theoretical and experimental examination of the decoupling design suggested that pulse imperfection could be a factor that impacted some relaxation components in the methyl group due to spin interactions.

Pulse imperfections have long been a culprit for many issues in NMR spectroscopy, such as systematic measurement errors in Car-Purcell-Meiboom-Gill (CPMG) experiments^18–19 and decreased sensitivities in triple resonance experiments^20–21. The imperfections are reflective of finite instrumental limitations and consist of off-resonance effects and radiofrequency field (RF) inhomogeneity. RF inhomogeneity is a function of probe coil design, which is now well optimized, and off-resonance effects reflect the finite power limitations and pulsewidths. The off-resonance effects correlate with the strength of the static magnetic field and will become much more significant as the available magnetic fields increase. In order to compensate the pulse imperfections, several alternatives have been proposed, such as composite pulses^22–24 and shaped pulses^25–27 (e.g. adiabatic pulses^28–29). The selection of compensated pulses for an application often represents a compromise between excitation performance and timing constraints within a specific pulse sequence. In the following demonstrations, we will show the improvements of methyl adiabatic relaxation experiments¹⁷ with composite pulse decoupling and its application to revealing the potential allosteric signaling in ubiquitin ligase proteins that are central to the ubiquitin proteasome system (UPS).

Recently, it has been reported that pulse imperfections can yield artifacts in the methyl spectra when using a simple Heteronuclear Multiple Quantum Coherence (HMQC) sequence with gradients³⁰. These direct spin coupling artifacts are similar to those observed in J-resolved spectra³¹ and can be removed through well-known composite pulses²⁴, 90_x-180_y-90_x or 90_x-240_y-90_x. Similarly, dipole-dipole spin coupling from remote protons can be removed though the use of fully deuterated macromolecules in deuterated buffer. Under these conditions, the ¹H carrier may be placed in the center of the methyl proton region and off-resonance effects are minimized, thus providing the ideal conditions of biological samples for methyl-TROSY spectroscopy. Unfortunately, these ideal conditions may not be satisfied in more sensitive/unstable macromolecules, where perdeuteration may not be possible or may be incomplete and the protein (or conditions) may require the system to be in H₂O. Under these conditions, the pulse sequence is most efficient when the ¹H carrier is on-resonance for H₂O and accommodates water suppression schemes, e.g. Watergate. In order to eliminate the artifacts resulting from the off-resonance effect when the proton carrier frequency is placed on water, we found that an alternative composite pulse²³ of 90_x-270_y-90_x, which had been proposed to minimize the off-resonance effect on the spin inversion, may be used to generate high-quality, artifact-free HMQC spectra (Supporting Fig. 1). Additionally, spin simulations confirm that the 90_x-270_y-90_x composite pulse is better at compensating the off-resonance effects than the 90_x-180_y-90_x composite pulse and the regular π pulse (Supporting Fig. 2).

In the previous implementation of adiabatic relaxation dispersion experiments on methyl groups (methyl-geoHARD)¹⁷, we proposed to use deuterated and ILV-labeled proteins; however it was found that the proton decoupling scheme is important and very sensitive to pulse imperfections. To remove the cross-correlation between the dipole-dipole interaction and chemical shift anisotropy, it is necessary to incorporate a proton decoupling scheme that avoids perturbing the proton-proton spin-flip mechanism and yields clean/mono-exponential intrinsic relaxation decays (and adiabatic relaxation dispersion profiles). It can be shown that decoupling is achieved by uniformly inverting all proton spins within the protein (all methyl protons, residual protonation, and exchangeable protons when the solvent is H₂O) at the center of the adiabatic spin-lock blocks (Fig. 1). Spin simulations found that uniform spin inversion causes little or no perturbation on the adiabatic relaxation dispersion profiles, and complete spin inversion effectively removes the adverse effects of cross-correlation between the ¹H-¹³C dipole-dipole interaction and ¹³C chemical shift anisotropy (Fig. 2). The cross-correlation relaxation effects can become substantial, since, although the methyl ¹³C CSA is small, the experiment measures the reduced R₂ rate from the inner two components of the manifolds, where the R₂ rate is reduced by the cross-correlation between ¹H-¹³C dipolar interactions. The cross-correlation effects can become more severe in large protein complexes with anisotropic tumbling. Experimentally, non-ideal inversion with a simple square 180_x pulse leads to considerable artifacts (non-exponential decays of longitudinal magnetizations)¹⁷, which complicates data analyses. Hence, it is important to remove such effects and improve the quality of relaxation dispersion data. The artifacts will worsen at higher magnetic fields, where off-resonance effects increase, unless the pulse width of the 180_x pulse decreases with the field. Therefore, it is crucial to use an efficient composite pulse that can uniformly invert all proton spins within the protein. The observations are first: the effects of the cross-correlation between dipole-dipole interaction and chemical shift anisotropy can be significantly suppressed by efficient composite pulse decoupling, which completely removes the potential bi-exponential decay of transverse magnetization for the methyl groups (Supporting Fig. 3). The deleterious effects of bi-exponential decay of transverse magnetization are exacerbated when the biomolecule has an asymmetric globular shape and experiences anisotropic tumbling. Secondly, our experimental data shows that a single 90_x-270_y-90_x composite pulse has little or no effect on the relaxation component due to the proton-proton spin-flip mechanism between methyl protons and external protons, as desired (Supporting Fig. 4). In this work and our previous report on methyl geoHARD¹⁷, we examined four schemes relating to proton decoupling (Supporting Fig. 5). The use of repeated 180_x pulses leads to considerable non mono-exponential behavior (Supporting Fig. 5c and Fig. S6¹⁷). The use of a single π-pulse during the relaxation period (Supporting Figs. 5d,e, 6) significantly reduces undesired relaxation behavior yielding mono-exponential decays, and the best performance is obtained with the single 90_x-270_y-90_x composite pulse. The final pulse sequence is shown in Fig. 1.

Figure 1.

(a) Pulse sequences for the improved version of methyl adiabatic relaxation dispersion experiments. As described previously, adiabatic hyperbolic secant pulses (HS_n) are used for the ¹³C spin-lock periods. R_1ρ (or R₁) relaxation delays (b) and R_2ρ (or R₂) relaxation delays (c) are followed after the filter element with ε = 0.6666 ms. All 90° (180°) pulses are indicated by narrow (wide) rectangles and are applied along the x axis, unless indicated otherwise. The shape pulses on the proton channel are for water flip-back and Watergate, and the gray wide rectangles are composite pulses, 90_x-270_y-90_x. Four adiabatic pulses (b and c) are concatenated in a MLEV-4 fashion (y, y, −y, −y), and each spin-lock unit is represented by a composite adiabatic pulse (16 ms) in parentheses for both R_1ρ and R_2ρ experiments. Δ is set to 1.8 ms. Phases are ϕ₁ = x, −x; ϕ₂ = x, x, −x, −x; ϕ₃ = x; and ϕ_rec = x, −x, −x, x. Quadrature detection in the t₁ dimension is achieved by shifting the phase of ϕ₃ by 90° in the states-TPPI manner. Gradient magnitudes for G1–G6 in units of (ms, G/cm) are (1, 10), (0.5, 6), (1, 50), (0.2, 10), (1, −20), and (0.5, 25), respectively.

Figure 2.

Simulated improvement due to the implementation of the proton decoupling scheme using a π pulse with perfect spin inversion. Data of adiabatic R_1ρ, adiabatic R_2ρ, and R₁ at two (14.1 and 18.8 T) magnetic fields are first simulated and then analyzed using a geometric approximation to output the dynamic parameters. Adiabatic relaxation dispersion data sets (400) are all simulated using the 12 × 12 density matrix with random input dynamic parameters as described previously. No errors are introduced in this data analysis. J coupling values (J) and transverse cross-correlation rates (G_xy) are varied as indicated above panels (a, d), (b, e), and (c, f). In (a–c), the relaxation dispersion data were simulated without a proton spin inversion during the spin lock periods; in (d–f), the relaxation dispersion data were simulated with a proton spin inversion during the spin lock periods. The output results are plotted against the input values, and those with large standard deviations (SD of k_ex > 10^0.2) after Monte Carlo sampling are omitted.

In order to demonstrate the broader application of this new sequence, Ube2g2, an ubiquitin-conjugating enzyme E2 with known anisotropic tumbling, will be used to test the new experiments. Ube2g2 functions with the ubiquitin ligase E3, gp78, as part of the endoplasmic reticulum associated degradation (ERAD) pathway^32–33, in a concerted mechanism to conjugate ubiquitin to substrates. In particular, different cytosolic domains of gp78 bind to Ube2g2 to induce allosteric effects. Through crystallography and solution-state NMR, structures and interfaces of Ube2g2 with various domains of gp78 bound to Ube2g2 reveal very subtle structural differences. Current studies of conformational dynamics have been focused on the backbone dynamics of Ube2g2 and suggest a role for dynamics driving the allostery³⁴. However, what is unclear is how the allostery is transmitted through Ube2g2 upon gp78 binding. Due to the limited exchange rate window of the conventional CPMG experiments and the lower sensitivity of the backbone amide experiments, it is beneficial to probe the conformational dynamics of Ube2g2 in the presence of other binding partners using side-chain methyl-geoHARD.

All NMR experiments for methyl-geoHARD were performed on {U-[¹⁵N, ²H]; Ileδ1-[¹³CH₃]; Leu, Val-[¹³CH₃, ¹²CD₃]}-labeled samples of Ube2g2 (~300 μM) at 15 °C in 50 mM Tris buffer (pH 7.4) containing 0.5 mM TCEP and 10% D₂O. Data were acquired using Bruker Avance III spectrometers equipped with a helium-temperature TCI cryoprobe at 800 MHz and a nitrogen-temperature TCI Prodigy probe at 600 MHz. As described previously¹⁷, the adiabatic R_1ρ and adiabatic R_2ρ experiments monitor the decay of magnetization for each composite shape (HS_nc) beginning at 0 ms and extended by multiples of a 16 ms block of composite adiabatic pulses to spin lock the magnetization (0, 16, 32, 48 ms… for N = 0, 1, 2, 3… in Fig. 1b,c). The total experiments consisted of adiabatic R_1ρ with 0~128 ms of the spin lock, adiabatic R_2ρ with 0~80 ms of the spin lock, R₁ with 0~400 ms of the delay, and R₂ with 0~128 ms of the spin echo, which took 3 days of acquisition in each static magnetic field. All relaxation experiments were carried out with a 2 s recycle delay, 16 scans, 128 complex points in the ¹³C dimension (19 ppm), and 2048 complex points in the ¹H dimension (14 ppm). Data are processed with linear prediction for another 128 complex points in the ¹³C dimension with NMRPipe³⁵ and subsequently analyzed with Sparky³⁶.

Among 55 isotope-labeled methyl groups from Ile, Leu, and Val in Ube2g2 (~18 kDa; PDB: 2CYX), 47 methyl resonances are selected for data analysis, based on resolution and resonance overlap at the two magnetic field strengths. All the selected methyl resonances have mono-exponential decay of transverse magnetizations using the new pulse sequences at the two magnetic fields. For those methyl groups with relaxation dispersion, analyses assumed a two-site exchange model. The chemical exchange rates (k_ex) of 13 methyl groups can be reported for Ube2g2 (Fig. 3a and Supporting Fig. 7). (Only the k_ex rates with small standard deviations are reported.) Comparatively, the carbon single-quantum CPMG experiment³⁷ only shows 6 methyl groups with measurable dispersions (Supporting Fig. 8). Our experimental data reveal two clusters of methyl groups in Ube2g2 (Supporting Fig. 7b) with well-defined exchange processes. One group (L9_CD2, L16_CD1, V116_CG1, V116_CG2), with a wider range of chemical exchange rates from 2000 s⁻¹ to 9000 s⁻¹, is located at the junction between helix α1 and helix α2. The second group (I24_CD1, I41_CD1, V124_CG2), with chemical exchange rates around 4000 s⁻¹, is located at the junction between helix α2 and the β turns, which form the binding site for the G2BR peptide³⁸. In previous reports, the allostery crucial for the biological function is observed when G2BR peptide binds to Ube2g2 and greatly enhances (approximately 50-fold) the RING binding affinity to Ube2g2³³. Interestingly, the binding site for RING domain is located at the N-terminal end of helix α1. In order to reveal the role of conformational dynamics in communicating the allostery as compared to a simple shift in a single conformation, we carried out a second set of experiments when Ube2g2 is saturated with the non-deuterated G2BR peptide (~5 kDa) (PDB:3H8K). Our second set of experimental data reveals detectable k_ex rates of only 7 methyl groups throughout Ube2g2 (Supporting Fig. 9), and no cluster of methyl groups as observed in free Ube2g2 (Fig. 3b). The decreased number of methyl groups with detectable k_ex values could result from a significantly decreased population of the minor states, which supports the previous hypothesis that G2BR can lock Ube2g2 into a closed state priming RING binding^33–34. The CPMG experiment³⁷ for the Ube2g2:G2BR complex shows all methyl groups with little or no relaxation dispersion (Supporting Fig. 10), similar to previous NH CPMG data³⁴. However, the new version of methyl geoHARD detects conformational exchange for 7 methyl groups in Ube2g2:G2BR. The trimmed average of the observed k_ex rates for these methyl groups decreases from ~6000 s⁻¹ to ~3000 s⁻¹ upon the binding of G2BR. The reduction implies that the energy barrier between the two exchanging conformations becomes higher when G2BR binds to Ube2g2. This is in agreement with our previous interpretation of exchange between the closed (bound) state of Ube2g2 and the other state (most probably the open state), based on MD trajectories³⁴.

Figure 3.

Conformational dynamics of methyl groups in Ube2g2 without (a) or with (b) the binding of G2BR, which are characterized by the new version of methyl-geoHARD. A total of 55 methyl groups are available as reporters throughout Ube2g2. Those with large standard deviations (SD of k ex > 10^0.2) after Monte Carlo sampling are not reported. (a) The k_ex rates of 13 methyl groups are reported in Ube2g2 without the binding of G2BR. (PDB: 2CYX) (b) The k_ex rates of 7 methyl groups are reported for Ube2g2 bound to G2BR. (PDB: 3H8K)

Finally, we conclude that the improved version of methyl-geoHARD experiments with composite pulse decoupling can provide high-quality data to reveal complex conformational dynamics. One may envision other, more complex composite pulses³⁹ to be used for proton decoupling schemes to minimize the effects of pulse imperfections; however, the total length of the proton pulse should be less than ~160 μs (<1% of the spinlock period), to avoid undesired mixing of trajectories and creation of multiple quantum terms during overlap of the proton pulse and the adiabatic pulses. Such pulses must be validated to leave the proton-proton spin-flip mechanism unperturbed, which is especially important for proteins bound with non-deuterated ligands, such as synthetic peptides, nucleotides, or small molecule drugs. The present use of 90_x-270_y-90_x performs very well and will be effective for all accessible magnetic field strengths, including the forthcoming 1.2 GHz systems. Utilizing these new tools, we reveal a potential allosteric path in Ube2g2 that is essential for its biological functions. Subsequent applications in studying conformational dynamics of multi-component complexes, as well as the hydrophobic cores of proteins, will improve our understanding in allosteric effects due to interactions among different components and allosteric signaling transmitted within the cores of proteins.