Quantitative measurement of exchange dynamics in proteins via ¹³C relaxation dispersion of ¹³CHD₂-labeled samples

Abstract

Methyl groups have emerged as powerful probes of protein dynamics with timescales from picoseconds to seconds. Typically, studies involving high molecular weight complexes exploit ¹³CH₃- or ¹³CHD₂-labeling in otherwise highly deuterated proteins. The ¹³CHD₂ label offers the unique advantage of providing ¹³C, ¹H and ²H spin probes, however a disadvantage has been the lack of an experiment to record ¹³C Carr–Purcell–Meiboom–Gill relaxation dispersion that monitors millisecond time-scale dynamics, implicated in a wide range of biological processes. Herein we develop an experiment that eliminates artifacts that would normally result from the scalar coupling between ¹³C and ²H spins that has limited applications in the past. The utility of the approach is established with a number of applications, including measurement of ms dynamics of a disease mutant of a 320 kDa p97 complex.

Introduction

NMR spectroscopy has emerged as a very powerful technique for the study of protein dynamics over timescales that span many orders of magnitude (Mittermaier and Kay 2006). The development of optimal labeling schemes and the design of the appropriate experiments that exploit them have proven critical in extending applications to increasingly larger proteins and protein complexes (Pervushin et al. 1998; Tugarinov et al. 2003). One popular approach for studies of sidechain dynamics makes use of proteins that are highly deuterated, with methyl group probes labeled as either ¹³CH₃ or ¹³CHD₂ (Tugarinov and Kay 2005). Each different methyl labeling approach has unique and often complementary strengths. For example, the ¹³CH₃ label provides the greatest sensitivity for studies of high molecular weight protein systems in general, by exploiting a methyl-TROSY effect (Tugarinov et al. 2003), and also enables studies of both fast time scale dynamics (ps–ns) (Tugarinov et al. 2007) and ms motional processes (Korzhnev et al. 2004) using approaches that exploit the cross-correlated relaxation networks that manifest in this spin system (Werbelow and Grant 1977). The ¹³CHD₂ label presents ¹³C, ¹H and ²H spins for measurement of motion in the ps–ns regime (Ishima et al. 2001) and ¹³C and ¹H spins for quantifying dynamics that are in the μs–ms range via a number of different R_1ρ-, CPMG- and CEST-types of experiments (Baldwin et al. 2010; Brath et al. 2006; Otten et al. 2010; Rennella et al. 2015).

A particularly important goal in protein dynamics studies by NMR is to characterize sparsely populated and transiently formed states that are generated via excursions from the populated, ground conformer, often the only state that is amenable to characterization using more ‘routine’ non-NMR biophysical measurements (Sekhar and Kay 2013). A number of NMR experiments have been developed with this goal in mind, including those exploiting ¹³CH₃ and ¹³CHD₂ methyl probes. Notably, ¹³C CEST-based experiments have been shown to be fivefold to sixfold more sensitive for ¹³CHD₂ methyl groups than for their ¹³CH₃ counterparts in an application involving the half proteasome that has a molecular mass of 360 kDa (Rennella et al. 2015). In addition, a ¹H-CPMG experiment has been developed for the measurement of ms time-scale dynamics in ¹³CHD₂-labeled proteins and applied to studies of proteasome gating (Baldwin et al. 2010). In contrast, corresponding single-quantum (SQ) based ¹H-CPMG experiments using fully protonated methyls are not possible unless magnetization from only $\frac{1}{2}$ manifolds is selected (Tugarinov and Kay 2007), since imperfections in CPMG refocusing pulses lead to artifacts that interfere with the quantification of relaxation dispersion profiles (Korzhnev et al. 2005). Despite these advantages, a significant drawback with using the ¹³CHD₂ label for studies of slow dynamics relates to the fact that it has not been possible to record ¹³C-CPMG dispersion profiles over the complete range of ¹³C pulsing frequencies that are often necessary to fully characterize such exchanging systems (Ishima et al. 1999). This is especially problematic for systems where the exchange rate is relatively slow (on the order of several hundreds/s). At the heart of the problem lies an interference effect resulting from the ‘combined action’ of the one-bond ¹³C–²H scalar couplings and the rapid ²H spin–lattice relaxation in such systems that can severely distort dispersion profiles (see below). Herein we present a ¹³C-SQ CPMG pulse scheme for studies of chemical exchange using ¹³CHD₂ methyls that eliminates this problem. The robustness of the experiment is established and its utility demonstrated with an application to the measurement of exchange dynamics in a 320 kDa construct of p97, a protein that plays a major role in cellular homeostasis (Braun and Zischka 2008).

Figure 1 shows the pulse sequence that has been developed for measurement of ¹³C SQ CPMG relaxation dispersion profiles using ¹³CHD₂ spin systems, where T _relax is the CPMG interval for recording effective ¹³C transverse relaxation rates, R _2,eff , as a function of the frequency of application of ¹³C chemical shift refocusing pulses, ν _CPMG . During this period a ¹H continuous wave (CW) decoupling field is applied at high strength (≥15 kHz) (Jiang et al. 2015), ensuring that ¹³C magnetization remains in-phase with respect to the coupled ¹H spin in the ¹³CHD₂ methyl. This eliminates the effects of imbalance in relaxation rates of in-phase and anti-phase ¹³C magnetization (with respect to ¹H) that can give rise to spurious dispersion profiles. The ¹³C–²H scalar coupled evolution of ¹³C magnetization during the CPMG interval can also lead to a ν _CPMG dependent contribution to the dispersion profile that is independent of chemical exchange, complicating extraction of robust exchange parameters. This is illustrated by an application to a sample of the B1 domain from immunoglobulin binding protein G (GB1). GB1 does not show ms time-scale dynamics in other experiments (Korzhnev et al. 2005), yet pathologic dispersion profiles are observed for all residues using the scheme of Fig. 1 when ²H decoupling is not applied during the CPMG interval, Fig. 1b (left). These ‘erroneous’ profiles, showing large R _2,eff rates at low CPMG pulsing frequencies, result from ¹³C–²H scalar coupled evolution of ¹³C magnetization in concert with ²H longitudinal relaxation that effectively interconverts the ¹³C multiplet components as they evolve (Abragam 1961; Grzesiek et al. 1993). For ν _CPMG  ≥ 250 Hz the attached deuterons are effectively decoupled and this region of the resulting dispersion profile is thus flat in the absence of exchange. Application of a uniform 1 kHz CW ²H decoupling field improves performance at the low CPMG frequency end but artifacts are predicted when ν _CW  = (2k − 1)^.ν _CPMG , where ν _CW is the strength of the ²H CW field in Hz and k is a natural number. As can be seen in Fig. 1b (middle) large spikes in R _2,eff values are observed at 350 Hz (k = 2) and 1 kHz (k = 1), as predicted, as well as smaller spikes for k > 2 (inset). These arise when the centers of the ¹³C and ²H pulses coincide, leading to evolution of ¹³C-²H scalar coupling during T _relax . In contrast, when a ²H decoupling scheme is used such that ν _CW  = 2k ^. ν _CPMG for ν _CPMG  ≤ 250 Hz (Vallurupalli et al. 2007) and ‘no field’ for ν _CPMG  > 250 Hz, flat dispersion profiles are obtained over the complete ν _CPMG range, Fig. 1b (right). Here we have kept the ²H CW field close to 1 kHz, with changes ~± 10 % for different ν _CPMG rates to ensure that ν _CW  = 2k ^. ν _CPMG , Fig. 2. The required power levels for the ²H CW decoupling field as a function of ν _CPMG can be calculated using a C-based program that is provided in Supporting Information and by the authors upon request.

Fig. 1

a Pulse sequence for measurement of ¹³C SQ CPMG relaxation dispersion profiles using ¹³CHD₂ spin systems. 90° (180°) pulses applied at maximum power are denoted by narrow (wide) black bars (with the exception of ¹³C pulses in the shaded element where a 17 kHz field is used), with pulse phases assumed to be x unless otherwise indicated. Between points A and B the ¹H carrier is placed in the center of the methyl region (1 ppm), while at point B the carrier is jumped to the water line (4.7 ppm); ¹³C and ²H carriers are positioned at 20 ppm and 1 ppm, respectively. ¹H, ¹³C and ²H WALTZ-16 decoupling elements (Shaka et al. 1983) are applied with fields of 6.2, 2 kHz and 710 Hz, respectively. Continuous wave (CW) elements are applied during the constant-time CPMG element (Mulder et al. 2001) of duration T _relax for N ≠ 0. A ¹H field of approximately 17 kHz is employed (Jiang et al. 2015), while for ²H a field of approximately 1 kHz is used for ν _CPMG  ≤ 250 Hz. The exact strength of the ²H CW element, ν _CW (Hz), is adjusted to match the condition ν _CW  = 2k ^. ν _CPMG (Vallurupalli et al. 2007) where k is a natural number. On our system heating is not observed with ²H decoupling and thus no compensation is used for ν _CPMG  > 250 Hz. In contrast to ²H, no adjustment is required for the ¹H CW field because a strong field is applied and the matching condition can be neglected (Jiang et al. 2015). For the case where N = 0 (reference plane) the ¹H CW element is applied immediately following data acquisition to ensure uniform heating, independent of the value of N. The two synchronous DIPSI-2 (Shaka et al. 1988) elements are employed for heteronuclear cross-polarization (8 ms) using a field of 8 kHz. Striped boxes indicate purge elements (3.6 kHz) of durations 2 (x-axis) and 3.4 ms (y-axis) for water suppression (Messerlie et al. 1989). The delays used are: τ_a = 1.8 ms, τ_b = 2 ms, τ_cp = T _relax /(4·N), t ^MAX₁  = maximum t₁ duration, δ = 460 μs. The delay t_eq before the CPMG element (partially) restores equilibrium magnetization values, depending on the value of k _ex . The phase cycle is as follows: φ ₁ = x, −x; φ ₂ = 4{x}, 4{−x}; φ ₃ = 2{y}, 2{−y}; φ ₄ = y; φ _R  = x, −x, −x, x. Quadrature detection in the indirect dimension is obtained using the gradient enhanced sensitivity approach (Kay et al. 1992; Schleucher et al. 1993) whereby separate data sets are recorded for each t₁ point with (φ ₄, g6) and (−φ ₄, −g6). Gradients are applied with the following durations (ms) and strengths (G/cm): g₀: (1, 21.4), g₁:(0.5, 16), g₂: (0.8, 26.8), g₃: (1, −26.8), g₄: (0.9, −16), g₅: (0.3, −21.4), g₆: (1.02, −48.2), g₇: (0.26, 48.2), g₈: (0.22, −16), g₉: (0.45, −16). The g₅ gradient pair is inverted with g₆. b Experimental CPMG profiles measured on GB1. (left) ²H CW decoupling is not employed, (center) a constant 1 kHz ²H CW field is used, (right) using a ²H CW field with ν _CW  = 2k ^. ν _CPMG (Vallurupalli et al. 2007) for ν _CPMG  ≤ 250 Hz, ν _CW  = 0 for ν _CPMG  > 250 Hz, as in Fig. 2. A 16.7 kHz ¹H CW decoupling field is applied in all cases (Jiang et al. 2015)

Fig. 2

Modulation of the ²H CW decoupling field (ν _CW ) as a function of ν _CPMG to ensure that the condition ν _CW  = 2k ^. ν _CPMG is satisfied for ν _CPMG  ≤ 250 Hz (see Fig. 1b, right). An approximate 1 kHz ²H decoupling field is chosen. See Supporting Information for C-code used to generate power levels for ²H CW decoupling

Having established a robust experimental scheme for recording ¹³C dispersion profiles in ¹³CHD₂ labeled proteins we next recorded experiments on a sample of an Ile, Leu, Val ¹³CHD₂-labeled G48A Fyn SH3 domain that interconverts between a highly populated native, folded conformation and a sparsely populated ensemble that corresponds to the unfolded state. The insets in Fig. 3 show ¹³C dispersion profiles for Leu 29^δ2 and Val 55^γ2 recorded with the pulse scheme of Fig. 1. Profiles from all residues have been fit simultaneously to a two state model of chemical exchange, $G {\mathop{\rightleftarrows}\limits_{ {k_{{EG}} } }^{ {k_{{GE}} } }} E$, where G and E denote ground and excited states, respectively. Extracted exchange parameters, k _ex  = 104 ± 2 s⁻¹, p _E  = 10.0 ± 0.2 % (25 °C), where p _E is the fractional population of the rare state, are in excellent agreement with those obtained previously from ¹³C CEST, k _ex  = 105 ± 2 s⁻¹, p _E  = 9.5 ± 0.1 %. Further verification of the methodology can be obtained by comparing the extracted chemical shifts, |Δϖ| (ppm), between the exchanging states obtained from the present study with those from ¹³C-CEST, Fig. 3, and the agreement is excellent.

Fig. 3

Linear correlation plot of extracted chemical shifts, |Δϖ| (ppm), obtained from analysis of ¹³C SQ dispersion profiles recorded on a ¹³CHD₂-sample of the G48A Fyn SH3 domain, 25 °C (y-axis) with those from analysis of a ¹³C-CEST dataset obtained under identical conditions (Rennella et al. 2015). Insets show a pair of methyl dispersion profiles, 600 MHz, recorded with the scheme of Fig. 1

We were interested in comparing the sensitivity of the present scheme with previously published methods for recording ¹³C CPMG profiles of ¹³CH₃-labeled proteins, focusing in particular on high molecular weight complexes. To this end we have compared spectra recorded on the half proteasome, α₇α₇, from T. Acidophilum (Sprangers and Kay 2007), labeled as [U-²H; Ileδ1-¹³CH₃; Leu,Val-¹³CH₃/¹²CD₃; Met-¹³CH₃] or [U-²H; Ileδ1-¹³CHD₂; Leu,Val-¹³CHD₂/¹³CHD₂; Met-¹³CH₃]. The α₇α₇ complex has an aggregate molecular mass of 360 kDa and an estimated overall tumbling time of 125 ns at 50 °C (Sprangers and Kay 2007). Figure 4 plots signal-to-noise (S/N) values as a function of methyl group for CPMG spectra recorded with T _relax  = 25 ms, 50 °C, using a methyl-TROSY scheme (Korzhnev et al. 2004) (black squares, ¹³CH₃-labeled protein), the present pulse sequence (dark grey circles, ¹³CHD₂-labeling) and a previously published ¹³C SQ-based experiment that was intended for studies of small to medium sized proteins (Lundstrom et al. 2007) (light grey diamonds, ¹³CH₃). As expected the methyl-TROSY scheme provides the best sensitivity (by roughly a factor of 2-3, on average, over the SQ ¹³CHD₂ experiment), with the SQ ¹³CHD₂ version significantly better than its SQ ¹³CH₃ counterpart. It is clear that from the point of view of sensitivity, often limiting in applications involving high molecular weight proteins, that the methyl-TROSY scheme is advantageous.

Fig. 4

S/N as a function of methyl group for CPMG spectra recorded on ¹³CHD₂- and ¹³CH₃-labeled α₇α₇ samples, with T _relax  = 25 ms, 50 °C. S/N ratios obtained using a methyl-TROSY scheme (Korzhnev et al. 2004) (black squares ¹³CH₃-labeled protein), the present pulse sequence (dark grey circles ¹³CHD₂-labeling) and a previously published ¹³C SQ-based experiment (Lundstrom et al. 2007) (light grey diamonds ¹³CH₃) are plotted

Despite the sensitivity loss in comparison to the methyl-TROSY experiment there is an advantage with the ¹³CHD₂-scheme. As discussed previously, methyl-TROSY CPMG profiles are sensitive to both ¹³C and ¹H chemical shift differences between spins in the exchanging states (Korzhnev et al. 2004). This offers the appealing possibility of obtaining both ¹³C and ¹H methyl chemical shifts of the rare conformer from a single experiment. However, in practice it can be difficult to simultaneously obtain accurate Δϖ values for both ¹³C and ¹H, unless S/N is high, and experiments that probe one Δϖ value at a time, as is possible for ¹³CHD₂- but not for ¹³CH₃-labeled samples using the methyl-TROSY approach (Korzhnev et al. 2005), can thus be desirable. By means of illustration, we have carried out simulations showing that while similar errors in ¹³C Δϖ values are obtained from fits of SQ and methyl-TROSY data when ¹H Δϖ = 0, fitted shift parameters from methyl-TROSY dispersions can become less accurate as ¹H shift differences increase because dispersion size correlates inversely with the magnitude of ¹H Δϖ (Fig. S1). For example, ¹³C SQ CPMG dispersion profiles that are on the order of 5–10 s⁻¹ for ¹³CH₃-labeled G48A Fyn SH3 decrease significantly (to ~2–3 s⁻¹) in methyl-TROSY CPMG data sets, making it challenging to extract accurate ¹³C Δϖ values from fits. This emphasizes the tradeoff between high sensitivity (methyl-TROSY) and robustness of data fitting that is higher in the SQ experiment where dispersion profiles are insensitive to ¹H Δϖ.

Figure 5 shows selected ¹³C and ¹H SQ dispersion profiles recorded on a [U-²H; Ileδ1-¹³CHD₂; Leu,Val-¹³CHD₂/¹³CHD₂; Met-¹³CHD₂]-labeled sample of an R95G mutant of p97. Notably, both ¹³C and ¹H profiles can be recorded on a single sample. The p97 protein is a ubiquitous ATPase, comprising ~1 % of the total protein in the cell (Song et al. 2003), with the R95G mutation responsible for a neurodegenerative disease that leads to dementia (Kimonis et al. 2008). To address the role of dynamics in p97 function/misfunction we have recorded ¹³C/¹H CPMG dispersion profiles of this protein. An extensive conformational exchange process has been discovered involving residues at domain interfaces that can be fit globally to a two-site exchange model with k _ex  = 2300 ± 400 s⁻¹.

Fig. 5

Experimental ¹³C (black) and ¹H (grey) CPMG profiles (circles) measured on a 0.6 mM sample of ¹³CHD₂-methyl labeled R95G p97, 50 °C, 800 MHz, together with fits of dispersion profiles to a two-site exchange model using the Bloch–McConnell equations (McConnell 1958) (solid lines). Details are provided in Supporting Information

In summary, we have presented an experiment for recording ¹³C SQ CPMG dispersion profiles of ¹³CHD₂-labeled proteins. Artifacts that normally result from the presence of coupled deuterons are eliminated using a ²H decoupling field that varies as a function of ν _CPMG . The utility of the approach has been established through tests on small proteins, showing the expected flat dispersion profiles where there is no exchange, while correct exchange parameters are fit for systems that interconvert. Although the ¹³CHD₂-based experiment is significantly less sensitive than the methyl-TROSY CPMG scheme, it can offer advantages with regards to robustness of extracted chemical shift parameters, and the ¹³CHD₂-labeling does offer the possibility of performing a wide range of different (¹H, ¹³C, ²H) relaxation measurements. Notably, high quality data can be obtained on a 320 kDa fragment of p97. The ¹³CHD₂-CPMG experiment, in concert with a previously published ¹H-CPMG scheme for ¹³CHD₂-labeled proteins (Baldwin et al. 2010), enables accurate measurement of both ¹H and ¹³C chemical shifts of rare protein conformers that will provide an important starting point for characterizing how these states are involved in biological function.