SimShift: identifying structural similarities from NMR chemical shifts

MOTIVATION: An important quantity that arises in NMR spectroscopy experiments is the chemical shift. The interpretation of these data is mostly done by human experts; to our knowledge there are no algorithms that predict protein structure from chemical shift sequences alone. One approach to facilitate this process could be to compare two such sequences, where the structure of one protein has already been resolved. Our claim is that similarity of chemical shifts thereby found implies structural similarity of the respective proteins. RESULTS: We present an algorithm to identify structural similarities of proteins by aligning their associated chemical shift sequences. To evaluate the correctness of our predictions, we propose a benchmark set of protein pairs that have high structural similarity, but low sequence similarity (because with high sequence similarity the structural similarities could easily be detected by a sequence alignment algorithm). We compare our results with those of HHsearch and SSEA and show that our method outperforms both in >50% of all cases.     

The revised NMR chemical shift data of carrageenans

A new set of (13)C and (1)H NMR chemical shifts of most common carrageenan types is given relative to DSS as the internal standard according to the IUPAC recommendations. Moreover, the chemical shifts of characteristic signals for pyruvate acetal and floridean starch are reported. Additionally, chemical shifts of common internal standards, such as methanol, DMSO and acetone, were measured at different temperatures and pH values.     

Secondary structural effects on protein NMR chemical shifts

For an amino acid in protein, its chemical shift, (, )_s, is expressed as a function of its backbone torsion angles ( and ) and secondary state (s): (, )_{s=, )_coil+(, )_s}, where (, )_coil represents its chemical shift at coil state (s=coil); (, )_s (s=sheet or helix) is herein defined as secondary structural effect correction factor, which are quantitatively determined from Residue-specific Secondary Structure Shielding Surface (RSS) for ¹³CO, ¹³C, ¹³C,¹H, ¹⁵N, and ¹HN nuclei. The secondary structural effect correction factors defined in this study differ from those in earlier investigations by separating out the backbone conformational effects. As a consequence, their magnitudes are significantly smaller than those earlier reported. The present (, )_sheet and (, )_helix were found varying little with backbone conformation and the 20 amino acids, specifically for ¹³CO, ¹³C, and ¹H nuclei. This study also carries out some useful investigations on other chemical shift prediction approaches – the traditional shielding surfaces, SHIFTS, SHIFTX, PROSHIFT, and identifies some unexpected shortcomings with these methods. It provides some useful insights into understanding protein chemical shifts and suggests a new route to improving chemical shifts prediction. The RSS surfaces were incorporated into the program PRSI [Wang and Jardetzky, J. Biomol. NMR, 28: 327–340 (2004)], which is available for academic users at http://www.pronmr.com or by sending email to the author (yunjunwang@yahoo.com).     

Parameters for the calculation of proton NMR chemical shifts of oligoribonucleotides

A set of empirical parameters which allows the prediction of the proton NMR chemical shifts at 70 C of non-exchangeable heterobase and anomeric protons in oligoribonucleotides has been constructed. The set is based on the highly flexible nature of oligoribonucleotide single strands and the wide range of conformational states which can be populated at relatively high temperatures (70 C or greater). A pairwise subtractive procedure, using 129 ribonucleotide oligomers (all 16 dimers, all 64 trimers, 37 tetramers, and 12 pentamers), shows that significant contributions to the observed chemical shift of protons in a given nucleoside residue are made by first, second, and third neighbors on the 3' and the 5' sides. The majority of the neighbors cause shielding effects with the exception of some first neighbors on the 5' side of a given residue. The magnitude of the shielding effects is greatest for the purine heterobases and follows the order A greater than G greater than C greater than U, with first neighbors on the 3'side showing more pronounced effects than second neighbors and these in turn showing larger effects than third neighbors. Second neighbors on the 5' side showed consistently greater shieldings than first neighbors, a result attributed to the deshielding effects of the first 5' neighbor phosphate group. The parameter Tables are applied to the prediction of proton chemical shifts in one heptamer, four hexamers, and two pentamers and give average absolute differences between predicted and observed shifts less than 0.030 ppm. The parameter approach represents an excellent method of generating initial assignments of proton chemical shifts for any single strand oligoribonucleotide.     

Nucleic acid helix structure determination from NMR proton chemical shifts

We present a method for de novo derivation of the three-dimensional helix structure of nucleic acids using non-exchangeable proton chemical shifts as sole source of experimental restraints. The method is called chemical shift de novo structure derivation protocol employing singular value decomposition (CHEOPS) and uses iterative singular value decomposition to optimize the structure in helix parameter space. The correct performance of CHEOPS and its range of application are established via an extensive set of structure derivations using either simulated or experimental chemical shifts as input. The simulated input data are used to assess in a defined manner the effect of errors or limitations in the input data on the derived structures. We find that the RNA helix parameters can be determined with high accuracy. We finally demonstrate via three deposited RNA structures that experimental proton chemical shifts suffice to derive RNA helix structures with high precision and accuracy. CHEOPS provides, subject to further development, new directions for high-resolution NMR structure determination of nucleic acids.     

13C NMR chemical shifts can predict disulfide bond formation

The presence of disulfide bonds can be detected unambiguously only by X-ray crystallography, and otherwise must be inferred by chemical methods. In this study we demonstrate that ¹³C NMR chemical shifts are diagnostic of disulfide bond formation, and can discriminate between cysteine in the reduced (free) and oxidized (disulfide bonded) state. A database of cysteine ¹³C C and C chemical shifts was constructed from the BMRB and Sheffield databases, and published journals. Statistical analysis indicated that the C shift is extremely sensitive to the redox state, and can predict the disulfide-bonded state. Further, chemical shifts in both states occupy distinct clusters as a function of secondary structure in the C/C chemical shift map. On the basis of these results, we provide simple ground rules for predicting the redox state of cysteines; these rules could be used effectively in NMR structure determination, predicting new folds, and in protein folding studies.     

Protein structural class identification directly from NMR spectra using averaged chemical shifts

Knowledge of the three-dimensional structure of proteins is integral to understanding their functions, and a necessity in the era of proteomics. A wide range of computational methods is employed to estimate the secondary, tertiary, and quaternary structures of proteins. Comprehensive experimental methods, on the other hand, are limited to nuclear magnetic resonance (NMR) and X-ray crystallography. The full characterization of individual structures, using either of these techniques, is extremely time intensive. The demands of high throughput proteomics necessitate the development of new, faster experimental methods for providing structural information. As a first step toward such a method, we explore the possibility of determining the structural classes of proteins directly from their NMR spectra, prior to resonance assignment, using averaged chemical shifts. This is achieved by correlating NMR-based information with empirical structure-based information available in widely used electronic databases. The results are analyzed statistically for their significance. The robustness of the method as a structure predictor is probed by applying it to a set of proteins of unknown structure. Our results show that this NMR-based method can be used as a low-resolution tool for protein structural class identification.     

Measurement of carbonyl chemical shifts of excited protein states by relaxation dispersion NMR spectroscopy: comparison between uniformly and selectively 13C labeled samples

Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion nuclear magnetic resonance (NMR) spectroscopy has emerged as a powerful method for quantifying chemical shifts of excited protein states. For many applications of the technique that involve the measurement of relaxation rates of carbon magnetization it is necessary to prepare samples with isolated (13)C spins so that experiments do not suffer from magnetization transfer between coupled carbon spins that would otherwise occur during the CPMG pulse train. In the case of (13)CO experiments however the large separation between (13)CO and (13)C(alpha) chemical shifts offers hope that robust (13)CO dispersion profiles can be recorded on uniformly (13)C labeled samples, leading to the extraction of accurate (13)CO chemical shifts of the invisible, excited state. Here we compare such chemical shifts recorded on samples that are selectively labeled, prepared using [1-(13)C]-pyruvate and NaH(13)CO(3,) or uniformly labeled, generated from (13)C-glucose. Very similar (13)CO chemical shifts are obtained from analysis of CPMG experiments recorded on both samples, and comparison with chemical shifts measured using a second approach establishes that the shifts measured from relaxation dispersion are very accurate.     

Resolving nitrogen-15 and proton chemical shifts for mobile segments of elastin with two-dimensional NMR spectroscopy

In this study, one- and two-dimensional NMR experiments are applied to uniformly (15)N-enriched synthetic elastin, a recombinant human tropoelastin that has been cross-linked to form an elastic hydrogel. Hydrated elastin is characterized by large segments that undergo "liquid-like" motions that limit the efficiency of cross-polarization. The refocused insensitive nuclei enhanced by polarization transfer experiment is used to target these extensive, mobile regions of this protein. Numerous peaks are detected in the backbone amide region of the protein, and their chemical shifts indicate the completely unstructured, "random coil" model for elastin is unlikely. Instead, more evidence is gathered that supports a characteristic ensemble of conformations in this rubber-like protein.     

NMR chemical shifts and structure refinement in proteins

Summary Computation of the ¹³C chemical shifts (or shieldings) of glycine, alanine and valine residues in bovine and Drosophila calmodulins and Staphylococcal nuclease, and comparison with experimental values, is reported using a gauge-including atomic orbital quantum-chemical approach. The full 24 ppm shielding range is reproduced (overall r.m.s.d.=1.4 ppm) using optimized protein structures, corrected for bond-length/bond-angle errors, and rovibrational effects.To whom correspondence should be addressed.     

Temperature dependence of NMR chemical shifts: Tracking and statistical analysis

Isotropic chemical shifts measured by solution nuclear magnetic resonance (NMR) spectroscopy offer extensive insights into protein structure and dynamics. Temperature dependences add a valuable dimension; notably, the temperature dependences of amide proton chemical shifts are valuable probes of hydrogen bonding, temperature‐dependent loss of structure, and exchange between distinct protein conformations. Accordingly, their uses include structural analysis of both folded and disordered proteins, and determination of the effects of mutations, binding, or solution conditions on protein energetics. Fundamentally, these temperature dependences result from changes in the local magnetic environments of nuclei, but correlations with global thermodynamic parameters measured via calorimetric methods have been observed. Although the temperature dependences of amide proton and nitrogen chemical shifts are often well approximated by a linear model, deviations from linearity are also observed and may be interpreted as evidence of fast exchange between distinct conformational states. Here, we describe computational methods, accessible via the Shift‐T web server, including an automated tracking algorithm that propagates initial (single temperature) ¹H? ¹⁵N cross peak assignments to spectra collected over a range of temperatures. Amide proton and nitrogen temperature coefficients (slopes determined by fitting chemical shift vs. temperature data to a linear model) are subsequently calculated. Also included are methods for the detection of systematic, statistically significant deviation from linearity (curvature) in the temperature dependences of amide proton chemical shifts. The use and utility of these methods are illustrated by example, and the Shift‐T web server is freely available at http://meieringlab.uwaterloo.ca/shiftt .     

The prediction of protein structural class using averaged chemical shifts

Knowledge of protein structural class can provide important information about its folding patterns. Many approaches have been developed for the prediction of protein structural classes. However, the information used by these approaches is primarily based on amino acid sequences. In this study, a novel method is presented to predict protein structural classes by use of chemical shift (CS) information derived from nuclear magnetic resonance spectra. Firstly, 399 non-homologue (about 15% identity) proteins were constructed to investigate the distribution of averaged CS values of six nuclei ((13)CO, (13)Cα, (13)Cβ, (1)HN, (1)Hα and (15)N) in three protein structural classes. Subsequently, support vector machine was proposed to predict three protein structural classes by using averaged CS information of six nuclei. Overall accuracy of jackknife cross-validation achieves 87.0%. Finally, the feature selection technique is applied to exclude redundant information and find out an optimized feature set. Results show that the overall accuracy increased to 88.0% by using the averaged CSs of (13)CO, (1)Hα and (15)N. The proposed approach outperformed other state-of-the-art methods in terms of predictive accuracy in particular for low-similarity protein data. We expect that our proposed approach will be an excellent alternative to traditional methods for protein structural class prediction.     

Rapid and accurate calculation of protein 1H, 13C and 15N chemical shifts

A computer program (SHIFTX) is described which rapidly and accurately calculates the diamagnetic ¹H, ¹³C and ¹⁵N chemical shifts of both backbone and sidechain atoms in proteins. The program uses a hybrid predictive approach that employs pre-calculated, empirically derived chemical shift hypersurfaces in combination with classical or semi-classical equations (for ring current, electric field, hydrogen bond and solvent effects) to calculate ¹H, ¹³C and ¹⁵N chemical shifts from atomic coordinates. The chemical shift hypersurfaces capture dihedral angle, sidechain orientation, secondary structure and nearest neighbor effects that cannot easily be translated to analytical formulae or predicted via classical means. The chemical shift hypersurfaces were generated using a database of IUPAC-referenced protein chemical shifts – RefDB (Zhang et al., 2003), and a corresponding set of high resolution (<2.1 Å) X-ray structures. Data mining techniques were used to extract the largest pairwise contributors (from a list of 20 derived geometric, sequential and structural parameters) to generate the necessary hypersurfaces. SHIFTX is rapid (< 1 CPU second for a complete shift calculation of 100 residues) and accurate. Overall, the program was able to attain a correlation coefficient (r) between observed and calculated shifts of 0.911 (¹H), 0.980 (¹³C), 0.996 (¹³C), 0.863 (¹³CO), 0.909 (¹⁵N), 0.741 (¹HN), and 0.907 (sidechain ¹H) with RMS errors of 0.23, 0.98, 1.10, 1.16, 2.43, 0.49, and 0.30 ppm, respectively on test data sets. We further show that the agreement between observed and SHIFTX calculated chemical shifts can be an extremely sensitive measure of the quality of protein structures. Our results suggest that if NMR-derived structures could be refined using heteronuclear chemical shifts calculated by SHIFTX, their precision could approach that of the highest resolution X-ray structures. SHIFTX is freely available as a web server at http://redpoll.pharmacy.ualberta.ca.     

Constituents of Papaver bracteatum: O-methyl-α-thebaol and 10-n-nonacosanol. Lanthanide-induced chemical shifts in 1H NMR and 13C NMR

Two non-alkaloidal constituents were identified in Papaver bracteatum: O-methyl-α-thebaol and 10-n-nonacosanol. O-Methyl-α-thebaol is a new natural compound. The presence of isothebaine is confirmed. Lanthanide-induced chemical shifts can be used for the assignments of the ₁₃C NMR chemical shifts of isothebaine and phenanthrenes. The use of lanthanide-induced chemical shifts in the identification of methoxyl resonances in ₁H NMR is discussed.     

QSAR study on (13)C NMR chemical shifts on carbinol carbon atoms

QSAR calculations of (13)C NMR chemical shifts (ppm, TMS=0) on carbinol carbon atoms have been attempted using a large set of distance based topological indices: Wiener (W)-, Szeged (Sz)-, PI (Padmakar-Ivan) and Connectivity ((m)chi, (m)chi(v)) indices. The regression analysis has shown that excellent results are obtained in multiparametric regression. The predictive power of the proposed models are discussed using cross-validated parameters.     

Changes in C NMR chemical shifts of DNA as a tool for monitoring drug interactions

The antibiotic drug, netropsin, was complexed with the DNA oligonucleotide duplex [d(GGTATACC)]₂ to explore the effects of ligand binding on the ¹³C NMR chemical shifts of the DNA base and sugar carbons. The binding mode of netrospin to TA-rich tracts of DNA has been well documented and served as an attractive model system. For the base carbons, four large changes in resonance chemical shifts were observed upon complex formation: −0.64 ppm for carbon 4 of either Ado4 or Ado6, 1.36 ppm for carbon 2 of Thd5, 1.33 ppm for carbon 5 of Thd5 and 0.94 for carbon 6 of Thd5. AdoC4 is covalently bonded to a heteroatom that is hydrogen bonded to netropsin; this relatively large deshielding is consistent with the known hydrogen bond formed at AdoN3. The three large shielding increases are consistent with hydrogen bonds to water in the minor groove being disrupted upon netropsin binding. For the DNA sugar resonances, large changes in chemical shifts were observed upon netropsin complexation. The 2′, 3′ and 5′ ¹³C resonances of Thd3 and Thd5 were shielded whereas those of Ado4 and Ado6 were deshielded; the ¹³C resonances of 1′ and 4′ could not be assigned. These changes are consistent with alteration of the dynamic pseudorotational states occupied by the DNA sugars. A significant alteration in the pseudorotational states of Ado4 or Ado6 must occur as suggested by the large change in chemical shift of −1.65 ppm of the C3′ carbon. In conclusion, ¹³C NMR may serve as a practical tool for analyzing structural changes in DNA-ligand complexes.     

The projection analysis of NMR chemical shifts reveals extended EPAC autoinhibition determinants

EPAC is a cAMP-dependent guanine nucleotide exchange factor that serves as a prototypical molecular switch for the regulation of essential cellular processes. Although EPAC activation by cAMP has been extensively investigated, the mechanism of EPAC autoinhibition is still not fully understood. The steric clash between the side chains of two conserved residues, L273 and F300 in EPAC1, has been previously shown to oppose the inactive-to-active conformational transition in the absence of cAMP. However, it has also been hypothesized that autoinhibition is assisted by entropic losses caused by quenching of dynamics that occurs if the inactive-to-active transition takes place in the absence of cAMP. Here, we test this hypothesis through the comparative NMR analysis of several EPAC1 mutants that target different allosteric sites of the cAMP-binding domain (CBD). Using what to our knowledge is a novel projection analysis of NMR chemical shifts to probe the effect of the mutations on the autoinhibition equilibrium of the CBD, we find that whenever the apo/active state is stabilized relative to the apo/inactive state, dynamics are consistently quenched in a conserved loop (β2-β3) and helix (α5) of the CBD. Overall, our results point to the presence of conserved and nondegenerate determinants of CBD autoinhibition that extends beyond the originally proposed L273/F300 residue pair, suggesting that complete activation necessitates the simultaneous suppression of multiple autoinhibitory mechanisms, which in turn confers added specificity for the cAMP allosteric effector.