Conceived and designed the experiments: MZ. Performed the experiments: JK AL VV. Analyzed the data: JK. Contributed reagents/materials/analysis tools: RS SB MB. Wrote the paper: OP MZ JK MB.
Current address: Laboratorium für Physikalische Chemie, ETH Zürich, Zürich, Switzerland
The authors have declared that no competing interests exist.
High-resolution solid-state NMR spectroscopy can provide structural information of proteins that cannot be studied by X-ray crystallography or solution NMR spectroscopy. Here we demonstrate that it is possible to determine a protein structure by solid-state NMR to a resolution comparable to that by solution NMR. Using an iterative assignment and structure calculation protocol, a large number of distance restraints was extracted from 1H/1H mixing experiments recorded on a single uniformly labeled sample under magic angle spinning conditions. The calculated structure has a coordinate precision of 0.6 Å and 1.3 Å for the backbone and side chain heavy atoms, respectively, and deviates from the structure observed in solution. The approach is expected to be applicable to larger systems enabling the determination of high-resolution structures of amyloid or membrane proteins.
Structural characterization of membrane proteins and many other biological systems by X-ray crystallography or solution NMR spectroscopy is difficult because of problems with crystallization, solubility or molecular size. Significant advances, however, have been made to construct three-dimensional (3D) molecular structures from solid-state NMR data obtained under Magic Angle Spinning (MAS)
Structure determination from solid-state NMR data typically follows the approach employed by solution-state NMR, namely assignment of backbone and side chain resonances using pulse sequences for sequential correlation of resonances, characterization of torsion angles and detection of tertiary contacts. Unless sample orientation provides a direct route to monitor molecular structure under MAS conditions
Here we combine 13C,13C- and 15N,13C-encoded 1H/1H mixing experiments recorded on a uniformly 13C/15N-labeled sample with a probabilistic assignment algorithm originally developed for the automatic assignment of 1H-1H correlations in Nuclear Overhauser Effect spectra recorded on proteins in solution
Earlier, the backbone fold of the 38-residue potassium channel blocker toxin KTX in the solid phase was deduced from 28 manually assigned interresidue CHHC correlations (ProteinDataBank (PDB) code: 1XSW)
Previously, 15 long-range, 7 medium-range and 6 short-range correlations could be assigned
Signals assigned in the 2D CHHC spectrum of diluted U-[13C, 15N]-KTX recorded with a mixing time of 250 µs are labeled. (B) and (D) show the number of unambiguously assigned distance constraints as a function of residue difference i and j.
Stereo view of the 20 lowest-energy structures are shown.
Shown are residue-based rms deviations of the coordinates of backbone atoms (solid line) and non-hydrogen side chain atoms (dashed line) within the ensemble of 20 lowest energy structures.
Solution | Free | |
Total | 314 | 260 |
Short range | 199 | 165 |
Medium range | 45 | 33 |
Long range | 70 | 62 |
Distance violations (>0.5 Å) | 0 | 2 |
64 | 58 | |
Dihedral violations (>5°) | 0 | 0 |
Total | −1203.3±60.1 | −1365.1±62.3 |
Dihedral | −133.4±5.6 | −119.3±5.1 |
NOE/CHHC | −166.8±34.3 | −167.7±29.3 |
Dihedral | 3.2±1.3 | 4.9±1.0 |
NOE/CHHC | 0.08±0.02 | 0.09±0.01 |
Backbone atoms (Å) | 0.7 | 0.6 |
All heavy atoms (Å) | 1.6 | 1.3 |
Most favored region (%) | 86.7 | 84.0 |
Disallowed region (%) | 3.3 | 3.7 |
Defined as the average rmsd difference between the 20 structures and the mean coordinates.
The two distance restraints G1(Hα1)-I4(Hα) and I4(Hγ11)-C35(Hα) were violated by 0.69 and 0.6 Å, respectively. The two restraints came from weak cross peak in the spectra and were assigned a distance range of 1.8–6.0 Å in the calculations.
Various tests were performed to probe the convergence of the structure calculations and support the accuracy of the high-resolution solid-state structure (see
Recently, a method for automatic assignment of cross peaks in 13C-13C correlation spectra was developed
To enable a direct comparison, we determined the solution structure of KTX employing the identical strategy as used for KTX in the solid phase. 70 long-range, 45 medium-range and 199 sequential NOEs could be assigned unambiguously, closely resembling the amount and distribution of distance restraints obtained from 2D N/CHHC spectra for KTX in the solid phase (
The backbone of the high-resolution solid-state structure of KTX deviates by 1.3 Å from that observed in solution (
(A) Superposition of high-resolution solid-state structure (green) and solution structure (blue; PDB code: 2KTX) of KTX. (B) Comparison of averaged Cα/Cβ chemical shift differences (blue bars; calculated according to 0.256*[ΔδCα2+ΔδCβ2]1/2) with rms deviation between the mean structures (blue line) of KTX in solution and in the solid-phase. Red dots mark residues, for which the backbone dihedral angles predicted by TALOS differ in solution and the solid state. Secondary structure is indicated.
Our study demonstrates that high-resolution 3D structures of globular proteins can be obtained from solid-state NMR data. The combination of 15N,13C-encoded 1H/1H mixing experiments with a probabilistic cross peak assignment algorithm is particularly powerful, as short distances between protons provide the principal source of long-range structural information. Depending on the molecule under investigation, the presented approach can be combined with other solid-state NMR spectroscopic methods. Applications to larger proteins may benefit from the use of block
Unlabelled and uniformly [13C,15N]-labeled KTX was prepared recombinantly as previously described
Solution-state NMR experiments were carried out at 298 K on a Bruker AVANCE 600 spectrometer. NMR samples contained 2 mM unlabelled KTX in 95% H2O/5% D2O, pH 7.5. 2D TOCSY (mixing time of 60 ms using MLEV17) and NOESY experiments (mixing time of 200 ms) were used to facilitate backbone assignment
An anisotropic medium for measurement of residual dipolar couplings was prepared by addition of Pf1 filamentous phages (Asla, Riga, Latvia) to a concentration of 12 mg/ml
Solid-state NMR data comprised three CHHC spectra (100, 175 and 250 µs (1H,1H) mixing) and one NHHC spectrum (100 µs (1H,1H) mixing)
Two-dimensional CHHC, NHHC and NOESY spectra were automatically peak picked using Sparky 3 (T. D. Goddard and D. G. Kneller, University of California, San Francisco). Diagonal peaks were manually removed. Peak intensities were classified into four ranges and converted into distance ranges of 1.8–2.7, 1.8–3.3, 1.8–5.0, and 1.8–6.0 Å. Lists of cross peaks were subjected to the automated cross-peak assignment and structure calculation algorithm PASD implemented in Xplor-NIH
PASD was applied largely following published procedures
PASD structures do not represent fully-refined NMR structures
We performed several tests to probe the reliability of the solid-state 3D structure of KTX: (i) use of CHHC spectra recorded with longer mixing times and using undiluted 13C/15N-labeld KTX; (ii) use of different conformations for calculating likelihood estimates in the PASD analysis; (iii) influence of chemical shift tolerances; (iv) influence of distance ranges; (v) influence of disulphide bond restraints; (vi) combination of solid-state distance restraints with solution-state dihedral angles (and vice versa).
To i): In contrast to the measurements performed in this study, CHHC spectra were previously recorded on 13C/15N-labeld KTX that was not diluted by unlabeled protein
To ii): At the end of pass 1 and 2 the PASD algorithm calculates likelihood estimates that each particular assignment associated with a cross-peak is correct. The likelihoods are calculated using the ensemble of structures present at the end of the corresponding pass. Thus, they are a metric of how consistent a given assignment is with the ensemble of structures at the end of each calculation pass
To iii): For the calculations reported in the main part of the manuscript, tolerances for matching chemical shifts to cross-peaks were set to 0.38 ppm and 0.60 ppm in the acquisition and indirect dimension, respectively. We repeated the structure calculations with chemical shift tolerances of 0.38 ppm and 0.4 ppm in the acquisition and indirect dimension, respectively. The resulting structure deviated by less than 0.7 Å (rms value for all N, Cα, CO backbone atoms) from the structure shown in
To iv): Peak intensities obtained from the 2D CHHC and NHHC spectra were classified into four ranges and converted into distance ranges of 1.8–2.7, 1.8–3.3, 1.8–5.0 and 1.8–6.0 Å, respectively. The classification was done independently for the four proton-proton correlation spectra (see main manuscript). To test the sensitivity of the solid-state structure to the used distance ranges, we repeated the structure calculations assigning to all N/CHHC correlations a distance range of 2.4–6.0 Å. The resulting structure deviated by less than 0.3 Å (rms value for all N, Cα, CO backbone atoms) from the structure shown in
To v): For both KTX(solution) and KTX(solid), structure calculations were performed without and with restraints for the three disulphide bonds. The resulting structures did not differ (backbone rms deviation below 0.5 Å) and only the results of calculations, in which the disulphide bonds were not enforced, were reported.
To vi): Are the structural differences due to an uncertainty in the analysis of N/CHHC spectra? To address this question, we recalculated the structure (using XPLOR-NIH and starting from an extended strand) using the same solid-state N/CHHC distance restraints, but supplementing them with the dihedral angles obtained by TALOS from the solution-state chemical shifts (instead of those obtained from the solid-state chemical shifts). The backbone of the resulting structure deviated by 0.5 Å from the high-resolution solid-state structure. The coordinate precision for backbone and all heavy atoms was 0.7 Å and 1.7 Å, respectively. However, two dihedral angle violations were introduced (for residues 2 and 24) and residue 24 moved into the disallowed region of the Ramachandran plot. In addition, the total energy increased from −1307±54 kcal/mol to −1032±48 kcal/mol, the dihedral angle energy from −110±6 kcal/mol to −16±36 kcal/mol and the distance restraint energy from −157±28 kcal/mol to −18±67 kcal/mol (when compared to the pure solid-state structure calculation). Similarly, when the solution-state distance restraints were combined with the solid-state dihedral angles, one dihedral angle violation (for S9) was introduced, the total energy was increased from 1203±60 kcal/mol to −1154±65 kcal/mol, the dihedral angle energy from −133±6 kcal/mol to −55±32 kcal/mol and the distance restraint energy from −167±34 kcal/mol to −31±53 kcal/mol (when compared to the pure solution-state structure calculation). The backbone of the resulting structure deviated by 0.6 Å from the high-resolution solution-state structure. The coordinate precision for backbone and all heavy atoms was 0.8 Å and 1.9 Å, respectively. These data demonstrate that the solid-state distance restraints are only in agreement with the solid-state backbone chemical shifts, and the solution-state distance restraints are only in agreement with the solution-state backbone chemical shifts.
(0.50 MB DOC)
(0.21 MB DOC)
(0.10 MB DOC)
We thank John Kuszewski and Christian Griesinger for discussions.