Fast Magic Angle Spinning NMR with Heteronucleus Detection for Resonance Assignments and Structural Characterization of Fully Protonated Proteins

Abstract

Heteronucleus-detected dipolar based correlation spectroscopy is established for assignments of ¹H, ¹³C, and ¹⁵N resonances and structural analysis in fully protonated proteins. We demonstrate that ¹³C detected 3D experiments are highly efficient and permit assignments of the majority of backbone resonances, as shown in an 89-residue dynein light chain 8, LC8 protein. With these experiments, we have resolved many ambiguities that were persistent in our previous studies using moderate MAS frequencies and lacking the ¹H dimension. The availability of ¹H isotropic chemical shifts measured with the heteronucleus-detected fast-MAS experiments presented here is essential for the accurate determination of the ¹H CSA tensors, which provide very useful structural probe. Finally, our results indicate that ¹³C detection in fast-MAS HETCOR experiments may be advantageous compared with ¹H detection as it yields datasets of significantly higher resolution in the ¹³C dimension than the ¹H detected HETCOR versions.

INTRODUCTION

Magic angle spinning NMR spectroscopy (MAS NMR) is a powerful technique to study, at atomic resolution, three-dimensional structure and molecular motions of large biological molecules, including proteins, nucleic acids and their assemblies (McDermott 2009; Yan et al. 2013). Despite recent impressive advances in the field, sensitivity and resolution still remain a major concern for the MAS NMR studies of large biomolecules. To overcome these limitations, fast MAS conditions (rotation frequencies of 40 kHz and higher) are increasingly employed (Agarwal et al. 2013; Ernst et al. 2001; Webber et al. 2012; Wickramasinghe and Ishii 2006; Wickramasinghe et al. 2009; Yan et al. 2013; Zhou et al. 2007).

Resonance assignment is an essential step in the protocol for structural and dynamics analysis by MAS NMR. By correlating the resonance frequencies of backbone or side-chain atoms in a series of 2D and 3D spectra, site-specific chemical shifts are obtained for a subsequent structure determination using distance restraints and/or residue-specific dynamics characterization. Proton-detected sequences generally produce excellent sensitivity and are the preferred route for structural studies in solution NMR (Clore and Gronenborn 1994). Unlike in solution NMR, the network of ¹H–¹H dipolar coupling in the solid-state is so strong that the proton linewidths are poorly resolved in ¹H-detected SSNMR experiments, at the MAS frequencies below 60 kHz that have been conventionally available to an experimentalist. To overcome this challenge, proton dilution by partial or full deuteration has been demonstrated to be one of the approaches to largely improve ¹H resolution in MAS NMR under moderate MAS frequencies (10–20 kHz) (Akbey et al. 2010; Chevelkov et al. 2006; Linser et al. 2011; Reif et al. 2001). Different deuteration schemes are successfully applied to amyloid fibrils, membrane proteins and protein-RNA complex for precise assignments, structural restraints and interface identification (Asami et al. 2013; Asami and Reif 2013; Linser et al. 2011; Zhou et al. 2012). However, the dilution of proton bath compromises to a great extent the potentially high sensitivity associated with proton detection. Furthermore, the expression protocols required for production of deuterated biomolecules as well as the back-exchange steps needed for reprotonation make this technique not applicable to many proteins and biomolecular assemblies.

In this context, proton detection under fast magic angle spinning conditions is a potential approach that can be widely used to achieve narrow ¹H lindewidths. As demonstrated by Rienstra and coworkers, fast MAS substantially improves the sensitivity and resolution if combined with full protonation (Zhou et al. 2007). In that study, high-resolution spectra have been acquired on fully protonated proteins with proton detection at MAS frequencies ~40 kHz and high magnetic field. Using fast MAS frequency of 60 kHz, Pintacuda and colleagues have demonstrated that the resolved ¹H dimension can be incorporated to help removing assignment ambiguities so that the resonance assignments of backbone ¹H, ¹⁵N, ¹³C nuclei of fully protonated protein can be made in an efficient and precise way (Marchetti et al. 2012). In addition to the ¹H chemical shifts that provide rich information for predicting secondary structure, the ¹H–¹H distance restraints for structural determination can potentially be measured by ¹H detection experiments in fully protonated proteins (Reif et al. 2001).

Even though ¹H detection offers large sensitivity enhancements in various applications discussed above, heteronucleus (¹³C) detection is advantageous in the cases where the resolution is inadequate for ¹H-detected experiments. For instance, in paramagnetic proteins the ¹H linewidths are often paramagnetically broadened beyond the detectable limit (Bermel et al. 2003; Bermel et al. 2006a; Bertini et al. 2005). Another example is natively unfolded proteins, which have very small chemical shift dispersion in the proton dimension, making ¹H detection impractical for resonance assignments of many unfolded proteins (Bermel et al. 2006b). Finally, the ¹³C chemical shifts in diamagnetic proteins span ca. 200 ppm, and covering the entire chemical shift range with adequate digital resolution is time consuming if ¹³C is sampled in an indirect dimension when full spectral widths need to be covered. The use of nonuniform sampling may alleviate this challenge to some extent; however, in the case of ¹³C experiments, the number of points that need to be sampled to retain the intensity information is still rather large (Suiter et al. 2014). Therefore, ¹³C detected experiments may be preferred for the acquisition of high-resolution spectra in the above situations. Indeed, ¹³C-detected scalar coupling based experiments have been reported to have high efficiency and resolution under fast MAS conditions and enable nearly complete backbone resonance assignments of fully protonated dimeric 153-residue protein in its diamagnetic and paramagnetic states (Barbet-Massin et al. 2013). However, heteronucleus detection in the context of dipolar based fast-MAS 3D experiments have not been yet demonstrated.

In this work, we present heteronucleus-detected dipolar-based 3D correlation spectroscopy for assignments of ¹H, ¹³C, and ¹⁵N resonances and structural analysis in fully protonated proteins. We establish this approach on an 89-residue microtubule-associated protein, dynein light chain 8 (LC8), whose primary sequence and secondary structure are shown in Figure 1. LC8 is an integral subunit of cytoplasmic dynein and plays important roles in both dynein-dependent and dynein-independent cellular functions. It has been implicated in viral infections and cancer progression (Barbar 2008; Vadlamudi et al. 2004). Previously, we have employed paramagnetically doped LC8 for the development of fast MAS NMR methods (40 kHz) for sensitivity and resolution enhancement needed for structural analysis of large protein assemblies (Sun et al. 2012). Here, we demonstrate that high-resolution ¹³C-detected 3D HNCA, HNCO, and HNCOCX spectra can be acquired in fully protonated U-¹³C,¹⁵N-LC8, in a time-efficient way, at the magnetic field of 19.9 T and MAS frequency of 62 kHz. The resolution in the ¹H dimension is excellent and allows for unambiguous assignments of most backbone ¹H resonances of LC8. With this set of 3D experiments, we have accomplished backbone resonance assignments for the majority of the residues in LC8 and resolved many ambiguities that were persistent in our previous studies using moderate MAS frequencies and experiments lacking the ¹H dimension (Sun et al. 2011). The secondary structure predicted based on the ¹H, ¹³C, and ¹⁵N chemical shifts agrees very well with the X-ray crystal structure 3DVT (PDB ID) (Lightcap et al. 2008). Finally, we compare heteronucleus and ¹H detection under fast MAS conditions using 2D HETCOR experiments and demonstrate that heteronucleus detection is advantageous from the resolution and time-saving standpoints.

Figure 1.

Amino acid sequence, secondary structure (top) and 3D X-ray structure (bottom) of Drosophila dynein light chain LC8. The structure is generated from PDB entry file 3DVT (Lightcap et al. 2008). LC8 is shown as homodimer. The α-helices are shown in green and β-sheets are in purple.

MATERIALS AND METHODS

Sample Preparation

The expression and purification of U–¹³C,¹⁵N-LC8 was reported by us previously (Lightcap et al. 2008). ¹⁵N–NH₄Cl and U–¹³C₆–glucose were used for isotopic enrichment. Purified U–¹³C,¹⁵N LC8 was dialyzed against 10 mM MES buffer (10 mM MgCl₂, pH = 6.0) and then concentrated to 30 mg/ml. To make paramagnetically doped sample, Cu(II)-EDTA solution was added to both concentrated LC8 solution and PEG-3350 solution (32%, w/v) to the final concentration of 5 mM. The PEG-3350 solution was gradually added into the LC8 solution following the controlled precipitation protocol (Marulanda et al. 2004) to get protein precipitates. The detailed preparation procedures of Cu(II)-EDTA doped LC8 have been described previously (Sun et al. 2012). Finally 3.1 mg of U–¹³C,¹⁵N LC8 solid sample containing 5 mM Cu(II)-EDTA was centrifuged into 1.3 mm Bruker MAS rotor for subsequent solid-state NMR experiments.

MAS NMR Spectroscopy

The MAS NMR experiments were carried out on Bruker AVIII 850 MHz spectrometer at the magnetic field of 19.9 T using a 1.3 mm ¹H/¹³C/¹⁵N triple-resonance probe. All 3D spectra were recorded at the MAS frequency of 62 kHz with apparent temperature controlled at −33±1 °C (sample temperature was 2 °C). The ¹H, ¹³C and ¹⁵N chemical shifts were referenced with respect to DSS, admantane and ammonium chloride used as external referencing standards. The pulse sequences for three-dimensional HNCA, HNCO and HN(CO)CX experiments in U–¹³C,¹⁵N LC8 doped with Cu(II)-EDTA are shown in Figure 2 A–B. For HNCA and HNCO experiments, ¹H–¹⁵N Hartmann–Hahn cross-polarization (CP) was set after ¹H evolution and followed by ¹⁵N–¹³C specific cross-polarization. For HN(CO)CX experiment, to establish ¹³C–¹³C correlations, RFDR was used with a mixing time of 2.6 ms after the double cross-polarization (DCP) (Schaefer et al. 1979). Low-power TPPM decoupling was used during the acquisition and indirect-dimension evolution periods (Bennett et al. 1995). All the 3D experiments were acquired with 2064 complex points in direct ¹³C dimension (t₃), 48 and 24 complex points in indirect ¹⁵N (t₂) and ¹H dimension (t₁) respectively. The experimental time were 3.5 days for HNCA, 1.5 days for HNCO and 7 days for HNCOCX. It has been demonstrated that with the paramagnetic dopant the ¹H longitudinal relaxation time T₁ of LC8 is significantly shortened and the ¹⁵N transverse relaxation time T₂ is slightly affected and close to that of neat samples. However, most chemical shifts of LC8 resonances are unperturbed and only a few resonances exhibiting small chemical shift perturbations in the 5mM Cu(II)-EDTA doped sample (Sun et al. 2012).

Figure 2.

Schematic representations of pulse sequences for 3D ¹H-based heteronucleus-detected experiments and 2D HETCOR conducted at fast MAS conditions. Solid and open bars represent π/2 and π pulse. A) The pulse sequence of HNCA and HNCO 3D experiments. The ¹³C carrier frequency was set to 55.0 ppm for HNCA and 170.0 ppm for HNCO. B) The pulse scheme of HN(CO)CX 3D experiment. A rotor synchronized RFDR mixing sequence was used to establish the ¹³C–¹³C correlations, with a mixing time of 2.6 ms. No proton decoupling was applied during the RFDR period at fast MAS frequency. C) HC HETCOR with carbon detection. D) CH HETCOR with proton detection. Phase cycling was applied to suppress water signal during the ¹³C–¹³C mixing period. The phases of the RF pulses are as follows: φ₁ = x, x, –x, –x; φ₂ = x, x, y, y, –x, –x, –y, –y; φ₃ = x, –x, –y, y, –x, x, y, –y; φ₄ = y, –y, –y, y, x, –x, –x, x, –y, y, y, –y, –x, x, x, –x; φ₅ = x, x, –x, –x, y, y, –y, –y; φ₆ = x, –x, –x, x, y, –y, –y, y; φ₇ = x, x, –x,– x; φ₈ = x, x, x, x, –x, –x, –x, –x, y, y, y, y, –y, –y, –y, –y; φ₉ = x, –x, –x, x, –x, x, x, –x, y, –y, –y, y, –y, y, y, –y.

To examine the resolution benefits of carbon-detected spectra, we conducted 2D heteronuclear (HETCOR) experiments with both ¹³C detection and ¹H detection at the MAS frequency of 60 kHz. The 2D HETCOR experiments exploited coherence transfers through dipolar couplings; the pulse schemes are displayed in Figure 2 C–D (Maudsley and Ernst 1977). For ¹H–¹³C carbon-detected HETCOR, swept-frequency TPPM decoupling (swfTPPM) was employed during the ¹³C FID acquisition (Thakur et al. 2006) following the Hartmann–Hahn polarization transfer. In the ¹³C–¹H proton-detected HETCOR, the WALTZ-16 broadband decoupling was executed during the FID acquisition, after the magnetization was transferred back to ¹H. Phase cycling was applied to suppress the water signal during the ¹³C–¹³C spin diffusion. In both spectra, the total experiment time was 4.5 hours. The HC HETCOR was collected with 128 scans for each t₁ point, while 32 scans were required for the CH HETCOR. The spectra were processed in NMRpipe (Delaglio et al. 1995) and analyzed with CCPNMR analysis (Stevens et al. 2011). Detailed information on the acquisition and processing parameters is given in Table S1.

RESULTS AND DISCUSSION

¹³C-Detected Fast MAS NMR Spectroscopy: Resonance Assignments of Dynein Light Chain 8

For resonance assignments of LC8, a set of ¹³C-detected dipolar-based 3D HNCA, HNCO and HN(CO)CX spectra was acquired. As expected, these spectra are of high quality, and the resolution is excellent with most of the cross peaks well separated. From these spectra, the resonance assignments could be readily derived as discussed below.

Several 2D H–C planes extracted from the 3D HNCA and HN(CO)CX spectra at different ¹⁵N chemical shifts are presented in Figure 3. The correlations between the amide proton and Cα of the same residue (i) or Cα of the previous residue (i-1) clearly reveal the connectivity of adjacent amino acids (shown as dashed line in Figure 3 A–C). From all slices it is easily seen that most cross peaks are resolved and distinguishable. Even though the resolution of the carbonyl region of HN(CO)CX spectrum is generally slightly worse than that of the aliphatic region of the same spectrum, most the peaks in the carbonyl region can be assigned unambiguously when the data set is combined with the HNCO spectrum. Notably, the experimental times for these high-quality 3D heteronuclear-detected data sets are relatively short: 3.5 days for HNCA, 1.5 days for HNCO and 7 days for HNCOCX. Therefore, ¹³C detection under fast MAS is a time efficient method yielding high sensitivity and resolution and enabling facile resonance assignments in fully protonated proteins.

Figure 3.

Representative 2D H–C planes of the ¹³C-detected 3D HNCOCX (purple) and HNCA (green) spectrum of U–¹³C,¹⁵N-LC8 at (A–C) δ_15N = 128.3 ppm; (D–F) δ_15N = 120.8 ppm; (G–I) δ_15N = 118.1 ppm. The correlations between amide proton and Cα of the same residue (i) or Cα of the previous residue (i-1) clearly reveal the connectivity of adjacent amino acids (shown as dashed line in Figure 3. A–C). All the spectra were acquired at the magnetic field of 19.9 T (850 MHz ¹H frequency) and MAS frequency of 62 kHz. The experimental times are 3.5 days for HNCA, 1.5 days for HNCO and 7 days for HNCOCX.

Using the above approach, we have successfully extracted assignments for 82 out of 89 residues of LC8. The representative sequential backbone walk for the A21–A28 segment is shown in Figure 4. Correlations in different spectra are found by matching the same amide ¹H and amide ¹⁵N chemical shifts. The incorporation of ¹H dimension helps remove many ambiguities that have previously impeded our resonance assignments of LC8 from experiments executed at the MAS frequency of 14 kHz (Sun et al. 2011). The residues that could not be assigned because the corresponding cross peaks were missing in the spectra include M1, S2, D3, R4, K5 (at the N-terminus), H68 (in a loop region) and G89 (at the C-terminus). As these residues are located in dynamic regions they may be either undergoing motions on intermediate timescales or are conformationally heterogeneous, either of which would lead to broader or undetectable peaks. For several other residues only backbone resonances are partially assigned. The amide ¹H and ¹⁵N resonances are not obtained for P52 and E69. P52 does not have an amide ¹H. E69 is in a loop region and may be undergoing conformational exchange on the experimental timescales. The backbone carbonyl ¹³C resonances are missing for A6, E16, L29, E30, K31, Y51, T67, T70, F86 and K87. Most of these residues are adjacent to mobile residues, and therefore we speculate that dynamics interferes with polarization transfer. The solid-state chemical shifts of LC8 obtained from 3D HNCA, HNCO, and HN(CO)CX spectra are listed in Table 1. Compared to our previous resonance assignments, under the fast MAS conditions we have detected and assigned 31 new resonances and also assigned the overwhelming majority of the backbone ¹H chemical shifts. In addition to the sequential assignments, we have detected and assigned a number of additional side chain resonances in HN(CO)CX spectra, which were not present in the NCOCX and NCACX experiments performed by us previously at the MAS frequencies of 10–20 kHz. On the basis of the HN(CO)CX spectra alone, we have assigned Cβ resonances for 29 residues (Table 1) and Cγ resonances for 5 residues (see Table S2). Furthermore, we have detected the aromatic side chain resonances of F86 as well (Table S2).

Figure 4.

Sequential backbone walk for residues A21–A28 of LC8 using 3D fast MAS experiments: HNCA (green), HNCO (black) and HNCOCX (purple). The solid-state NMR experiments were carried out on Bruker Ascend 850 MHz spectrometer at the magnetic field of 19.9 T. All 3D spectra were recorded at the MAS frequency of 62 kHz with apparent temperature controlled at −33±1 °C (sample temperature was 2 °C).

Table 1.

Chemical shifts of U-¹³C,¹⁵N-LC8 assigned based on 3D MAS NMR HNCA, HNCO and HNCOCX spectra.

Residue	H (ppm)	N (ppm)	C′ (ppm)	Cα (ppm)	Cβ (ppm)
K5			176.1
A6	9.3	134.9		51.6
V7	9.1	123.4	175.4	61.7	34.2
I8	9.1	129.0	176.0	61.4
K9	8.3	129.0	176.9	56.2
N10	8.2	114.4	172.5	53.6
A11	8.8	124.7	175.2	52.0	22.4
D12	8.9	122.5	173.7	53.5
M13	8.1	120.7	174.9	55.9
S14	9.4	123.9	175.1	58.5	64.2
E15	9.3	123.6	179.1	60.5
E16	9.1	118.7		60.0
M17	7.5	120.8	177.5	59.6
Q18	8.5	119.0	178.1	59.4	28.9
Q19	7.7	117.0	178.3	58.2
D20	7.8	119.4	178.7	57.4
A21	8.7	125.1	178.7	56.0	17.6
V22	7.8	118.7	179.8	66.8	31.7
D23	8.9	124.0	178.6	58.0
C24	8.9	120.7	177.1	63.3	27.7
A25	8.6	120.1	178.3	55.0	19.8
T26	8.6	115.5	176.6	68.0
Q27	7.7	120.1	178.7	58.7	28.4
A28	8.3	122.1	175.9	55.1	20.9
L29	8.2	116.6		56.9	42.2
E30	7.3	117.5		57.8
K31	7.1	116.9		58.3	35.5
Y32	8.6	118.5	173.5	56.5	33.6
N33	6.9	111.2	175.8	53.2	41.8
I34	9.6	124.2	178.1	61.5
E35	10.2	130.1	178.6	62.6	28.4
K36	8.5	116.7	177.1	59.9
D37	6.8	118.9	178.5	56.7
I38	7.7	122.2	177.0	65.4	38.4
A39	8.4	119.1	178.6	54.8	17.9
A40	7.9	119.3	178.3	55.0	18.6
Y41	7.7	117.5	177.5	61.9	38.3
I42	7.8	117.6	177.4	65.0	38.4
K43	8.4	118.8	178.7	61.1	35.4
K44	8.4	117.4	180.5	59.8
E45	8.0	119.4	178.9	59.0
F46	8.5	123.1	177.9	63.5
D47	8.6	119.9	178.7	57.3	40.2
K48	7.3	117.2	177.7	59.0
K49	7.9	117.6	177.9	59.2
Y50	8.6	114.3	176.8	58.1
N51	7.0	113.1		55.2
P52			178.3	60.4
T53	7.9	121.6	172.0	64.8
W54	9.7	128.3	174.3	55.9
H55	8.5	118.5	174.8	55.1	34.5
C56	8.3	120.9	171.4	56.7	30.9
I57	9.4	132.4	173.6	59.8
V58	8.4	125.0	175.8	59.6	36.3
G59	9.7	112.6	172.7	46.8
R60	8.6	118.5	178.2	56.7
N61	7.7	117.3	172.6	54.9
F62	8.0	119.4	173.4	57.5
G63	9.6	107.8	172.0	43.8
S64	8.2	111.8	172.4	57.2
Y65	9.0	122.6	173.2	61.5
V66	8.0	119.9	177.3	57.2
T67	9.0	122.6		61.6
H68
E69			179.5	60.7
T70	8.1	123.4		64.9
R71	9.0	123.3	176.6	58.7
H72	8.6	116.2	173.3	56.7
F73	8.1	120.8	172.6	58.3
I74	8.6	129.0	169.1	61.8
Y75	8.5	127.3	174.8	54.8	41.1
F76	9.4	125.3	170.1	54.8
Y77	9.3	118.2	176.5	55.6
L78	8.9	123.1	176.6	54.1
G79	8.9	114.3	175.1	47.0
Q80	8.9	124.4	174.5	56.3
V81	8.8	121.2	172.5	62.7
A82	8.5	128.8	176.2	50.3	20.7
I83	8.8	120.2	173.3	60.5	39.8
L84	9.3	128.0	173.9	53.1	46.2
L85	8.9	130.2	174.1	53.9
F86	9.0	121.2		56.1
K87	9.1	116.3		54.6
S88	7.4	117.1
G89

We also note that the availability of ¹H isotropic chemical shifts measured directly under MAS conditions is essential for the determination of the ¹H CSA tensors. We have recently developed an experimental approach for deriving ¹H CSA tensors in proteins and other biological molecules, in a residue-specific way (Hou et al. 2014; Hou et al. 2013), and corroborated that the principal components are strongly correlated with hydrogen bonding environments, providing very useful structural probe. At the same time, we could not record ¹H isotropic chemical shifts under moderate MAS conditions and had to rely on solution values, an approach, which is likely to be error-prone due to the different environments of the solid-like and solution states. The heteronucleus-detected fast-MAS experiments presented here overcome this limitation.

Secondary Structure Prediction in LC8

To predict the secondary structure from the resonance assignments of LC8, we used TALOS+ (Shen et al. 2009) to derive the backbone torsion angles ϕ and ψ of each residue. Overall, the secondary structure predicted from the fast-MAS NMR chemical shifts is generally consistent with the X-ray structure of LC8 (PDB file 3DVT) and with our previous results (Sun et al. 2011).

The plots of derived torsion angles and cartoon of predicted secondary structure are shown in Figure 5. The α-helices are shown in green and β-sheets are in purple. Since the chemical shifts of residues M1–R4 are unavailable, the predicted results are presented from K5. Our prediction is identical to the secondary structure derived from the X-ray data in the two α-helices segments and the β-sheet structure at the end (V81–S88). In other segments slight discrepancies exist. For instance, G59–N61 and T70–H72 are predicted to have no secondary structure and belong to loop regions, instead of being at the termini of two β-sheets in the X-ray structure 3DVT. In our case, this is not surprising since the resonances of these two segments are missing in most of our previous heteronuclear spectra (Sun et al. 2011; Sun et al. 2012). Even though we were able to detect the resonances corresponding to these residues in the 3D experiments performed with fast MAS at 850 MHz reported here, the peak intensities of these residues are extremely weak compared to those of other peaks. These results suggest that the G59–N61 and T70–H72 segments are mobile on the experimental timescales, and therefore whether these residues belong to the beginning and end of a loop connecting the two β sheets or to the corresponding β-sheet termini regions is likely dependent on the sample conditions. To this end, we note that the X-ray structure was determined under cryogenic conditions at temperature of 100 K and the termini of secondary structure elements appear rigid, unlike the MAS NMR experiments conducted under the temperatures close to physiological conditions, T = 2 °C.

Figure 5.

Backbone torsion angles Φ (top) and φ (bottom) derived by TALOS+ (Shen et al. 2009) based on the isotropic chemical shifts of backbone ¹H, ¹³C and ¹⁵N atoms in LC8. The chemical shifts (see Table 1) were recorded by 3D fast MAS experiments. The amino acid sequence and predicted secondary structure of LC8 by TALOS+ are shown on top. The α-helices are shown in green and β-sheets are in purple.

Comparison of Heteronucleus Detection and ¹H Detection

As discussed above, heteronucleus-detected 3D fast-MAS experiments are very efficient for resonance assignments. In certain instances, ¹³C-detection is anticipated to be advantageous vis-à-vis ¹H detection at fast MAS in terms of spectral resolution. One such example is H-C HETCOR experiments.

The HC (carbon-detected) and CH (proton-detected) HETCOR spectra acquired with the same total experiment time are shown in Figure 6 A and D, and the magnified representations of ¹H_N-¹³Cα regions are given in B and C. The number of scans for CH HETCOR was set to be four times as small as the number for HC HETCOR, in order to keep the total time of the CH experiment reasonable. We note that this does not affect the resolution comparison between the two experiments. 1D slices of ¹H and ¹³C dimensions extracted from 2D HETCOR are shown in Figure 6. The ¹H and ¹³C linewidths of the cross peaks in the ¹H_N-¹³Cα region are listed in Table S3. For the majority of the cross peaks, the linewidths of the ¹³C dimensions in the HC HETCOR are smaller than those in the CH HETCOR, indicating distinct resolution improvement in the ¹³C dimension in the carbon detection experiment. The resolution in ¹H dimension is very similar in both experiments: half of the cross peaks have somewhat smaller ¹H linewidths in the ¹³C-detected experiment while for the other half of the peaks, the linewidths in ¹H dimension are somewhat larger. One such example is shown in Figure 6 E, where the ¹H linewidth of the cross peak at 5.9 ppm is 269 Hz for carbon-detected HETCOR (red curve) and 322 Hz for proton-detected HETCOR (black curve). The ¹H-detected HETCOR thus does not exhibit obvious advantages in terms of the ¹H resolution either. Our results generally suggest that ¹H detection may not offer resolution benefits in the cases where indirect-dimension spectral width is large, such as ¹³C experiments. In these cases, experiment times of ¹H-detected measurements are excessively long to attain sufficient resolution for unambiguous assignments, if full spectral width in the ¹³C dimension needs to be covered.

Figure 6.

2D H–C ¹³C-detected HETCOR spectrum (A) and 2D C–H ¹H-detected HETCOR spectrum (D) of U–¹³C,¹⁵N-LC8 acquired at 60 kHz MAS and 850 MHz. Both spectra were processed with 30-degree sine bell apodization. The lowest contour level is set at 10 × σ RMSD with multiplier of 1.2. The zoom-in representations of the ¹H_N-¹³Cα region of each HETCOR spectrum are displayed in B and C. The experimental time was the same in both cases (4.5 hours) while different numbers of scans were used: 128 scans for HC HETCOR and 32 scans for CH HETCOR. Note that several cross peaks that appear in the ¹H-detected HETCOR are not visible in ¹³C-detected experiment, because of the lower sensitivity of the latter. The peaks in the ¹³C-detected HETCOR that appear around δ_13C = 20–45 ppm/δ_1H = 5.2 ppm (¹H) are most likely artifacts. 1D slices were extracted from 2D HETCOR spectra processed with 60-degree sine bell at the position of crosshairs shown in B and C: (E) ¹H slices of HC HETCOR (red) and CH HETCOR (black) extracted at δ_13C = 61.5 ppm. The ¹H linewidth of the peak at δ_1H = 5.9 ppm is 269 Hz for ¹³C-detected HETCOR and 322 Hz for ¹H-detected HETCOR. F) ¹³C slices of HC HETCOR (red) and CH HETCOR (black) extracted at δ_1H = 4.25 ppm.

While better resolution is an advantage of ¹³C-detected HETCOR, sensitivity is considerably higher in the ¹H-detected version, consistent with prior reports by other investigators (Marchetti et al. 2012; Zhou et al. 2007). In the spectra reported here, the average signal-to-noise ratio (SNR) for the cross peaks is 69 and 208 for the ¹³C- and ¹H-detected HETCOR, respectively. In both cases, the experiment time was 4.5 hours. The HC HETCOR was collected with 128 scans, and the SNR is ca. 6 per scan for the ¹³C-detection. The CH HETCOR was acquired with 32 scans, and the SNR is 37 per scan for the ¹H-detection. Curiously, several cross peaks that appear in the ¹H-detected experiment vanished in the ¹³C-detected spectrum, e. g., the peaks in the region of δ_13C = 20–35 ppm/δ_1H = 4–5 ppm, δ_13C = 59 ppm/δ_1H = 2.8 ppm, and δ_13C = 69 ppm/δ_1H = 2.9 ppm. This is likely due to the lower sensitivity of the ¹³C detection.

CONCLUSIONS

¹³C detected 3D heteronuclear correlation spectroscopy under fast MAS conditions is highly efficient for resonance assignments and structural analysis of fully protonated proteins. The sensitivity and the resolution of 3D HNCA, HNCO, and HNCOCX experiments are excellent, and with these three datasets we have assigned the majority of the backbone and a number of sidechain ¹H/¹³C/¹⁵N resonances in dynein’s LC8. With these experiments, we have resolved many ambiguities that were persistent in our previous studies using moderate MAS frequencies and lacking the ¹H dimension. Finally, we demonstrate that ¹³C detection in the fast-MAS HETCOR experiments may be advantageous vis-à-vis ¹H detection as it yields datasets of significantly higher ¹³C resolution.