MUNIN: A new approach to multi-dimensional NMR spectra interpretation

A new method, MUNIN (Multi-dimensional NMR spectra interpretation), is introduced for the automated interpretation of three-dimensional NMR spectra. It is based on a mathematical concept referred to as three-way decomposition. An NMR spectrum is decomposed into a sum of components, with each component corresponding to one or a group of peaks. Each component is defined as the direct product of three one-dimensional shapes. A consequence is reduction in dimensionality of the spectral data used in further analysis. The decomposition may be applied to frequency-domain or time-domain data, or to a mixture of these. Features of MUNIN include good resolution in crowded regions and the absence of assumptions about line shapes. Uniform sampling of time-domain data, a prerequisite for discrete Fourier transform, is not required. This opens an avenue for the processing of NMR data that do not follow oscillating behaviour, e.g. from relaxation measurements. The application of MUNIN is illustrated for a ¹H-¹⁵N-NOESY-HSQC, where each component is defined as the set of all NOE peaks formed by a given amide group. As a result, the extraction of structural information simply consists of one-dimensional peak picking of the shape along the NOE-axis obtained for each amide group.     

Resonance assignment and structural analysis of acid denatured E. coli [U-15N]-glutaredoxin 3

Virtually complete sequence specific ¹H and ¹⁵N resonance assignments are presented for acid denatured reduced E. coli glutaredoxin 3. The sequential resonance assignments of the backbone rely on the combined use of 3D F₁-decoupled ROESY-¹⁵N-HSQC and 3D ¹⁵N-HSQC-(TOCSY-NOESY)-¹⁵N-HSQC using a single uniformly ¹⁵N labelled protein sample. The sidechain resonances were assigned from a 3D TOCSY-¹⁵N-HSQC and a homonouclear TOCSY spectrum. The presented assignment strategy works in the absence of chemical exchange peaks with signals from the native conformation and without ¹³C/¹⁵N double labelling. Chemical shifts, ³J(H, NH) coupling constants and NOEs indicate extensive conformational averaging of both backbone and side chains in agreement with a random coil conformation. The only secondary structure element persisting at pH 3.5 appears to be a short helical segment comprising residues 37 to 40.Abbreviations HSQC heteronuclear single quantum coherence - NMR nuclear magnetic resonance - NOE nuclear Overhauser effect - NOESY two-dimensional NOE spectroscopy - ROE nuclear Overhauser effect in the rotating frame - ROESY two-dimensional ROE spectroscopy - TOCSY total correlation spectroscopy - TPPI time proportional phase incrementation Correspondence to: G. Otting     

Fully automated sequence-specific resonance assignments of hetero- nuclear protein spectra

Full automation of the analysis of spectra is a prerequisite for high-throughput NMR studies in structural or functional genomics. Sequence-specific assignments often form the major bottleneck. Here, we present a procedure that yields nearly complete backbone and side chain resonance assignments starting from a set of heteronuclear three-dimensional spectra. Neither manual intervention, e.g., to correct lists obtained from peak picking before feeding these to an assignment program, nor protein-specific information, e.g., structures of homologous proteins, were required. By combining two earlier published procedures, AUTOPSY [Koradi et al. (1998) J. Magn. Reson., 135, 288–297] and GARANT [Bartels et al. (1996) J. Biomol. NMR, 7, 207–213], with a new program, PICS, all necessary steps from spectra analyses to sequence-specific assignments were performed fully automatically. Characteristic features of the present approach are a flexible design allowing as input almost any combination of NMR spectra, applicability to side chains, robustness with respect to parameter choices (such as noise levels) and reproducibility. In this study, automated resonance assignments were obtained for the 14 kD blue copper protein azurin from P. aeruginosa using five spectra: HNCACB, HNHA, HCCH-TOCSY, ¹⁵N-NOESY-HSQC and ¹³C-NOESY-HSQC. Peaks from these three-dimensional spectra were filtered and calibrated with the help of two two-dimensional spectra: ¹⁵N-HSQC and ¹³C-HSQC. The rate of incorrect assignments is less than 1.5% for backbone nuclei and about 3.5% when side chain protons are also considered.     

Complete 1H, 13C and 15N NMR assignments and secondary structure of the 269-residue serine protease PB92 from Bacillus alcalophilus

Summary The ¹H, ¹³C and ¹⁵N NMR resonances of serine protease PB92 have been assigned using 3D tripleresonance NMR techniques. With a molecular weight of 27 kDa (269 residues) this protein is one of the largest monomeric proteins assigned so far. The side-chain assignments were based mainly on 3D H(C)CH and 3D (H)CCH COSY and TOCSY experiments. The set of assignments encompasses all backbone carbonyl and CH_n carbons, all amide (NH and NH₂) nitrogens and 99.2% of the amide and CH_n protons. The secondary structure and general topology appear to be identical to those found in the crystal structure of serine protease PB92 [Van der Laan et al. (1992) Protein Eng., 5, 405–411], as judged by chemical shift deviations from random coil values, NH exchange data and analysis of NOEs between backbone NH groups.Abbreviations 2D/3D/4D two-/three-/four-dimensional - HSQC heteronuclear single-quantum coherence - HMQC heteronuclear multiple-quantum coherence - COSY correlation spectroscopy - TOCSY total correlation spectroscopy - NOE nuclear Overhauser enhancement (connectivity) - NOESY 2D NOE spectroscopy Experiment nomenclature (H(C)CH, etc.) follows the conventions used elsewhere [e.g. Ikura et al. (1990) Biochemistry, 29, 4659–4667].     

Backbone 1H and 15N resonance assignments of the N-terminal SH3 domain of drk in folded and unfolded states using enhanced-sensitivity pulsed field gradient NMR techniques

Summary The backbone ¹H and ¹⁵N resonances of the N-terminal SH3 domain of the Drosophila signaling adapter protein, drk, have been assigned. This domain is in slow exchange on the NMR timescale between folded and predominantly unfolded states. Data were collected on both states simultaneously, on samples of the SH3 in near physiological buffer exhibiting an approximately 1:1 ratio of the two states. NMR methods which exploit the chemical shift dispersion of the ¹⁵N resonances of unfolded states and pulsed field gradient water suppression approaches for avoiding saturation and dephasing of amide protons which rapidly exchange with solvent were utilized for the assignment.Abbreviations 2D, 3D two-, three-dimensional - drkN SH3 N-terminal SH3 domain of Drosophila drk - HSQC heteronuclear single-quantum spectroscopy - NOE nuclear Overhauser enhancement - SH3 Src homology domain 3 - TOCSY total correlation spectroscopy     

Sensitivity improvement for correlations involving arginine side-chain Nε/Hε resonances in multi-dimensional NMR experiments using broadband 15N 180° pulses

Due to practical limitations in available ¹⁵N rf field strength, imperfections in ¹⁵N 180° pulses arising from off-resonance effects can result in significant sensitivity loss, even if the chemical shift offset is relatively small. Indeed, in multi-dimensional NMR experiments optimized for protein backbone amide groups, cross-peaks arising from the Arg guanidino ¹⁵Nε (~85 ppm) are highly attenuated by the presence of multiple INEPT transfer steps. To improve the sensitivity for correlations involving Arg Nε–Hε groups, we have incorporated ¹⁵N broadband 180° pulses into 3D ¹⁵N-separated NOE-HSQC and HNCACB experiments. Two ¹⁵N-WURST pulses incorporated at the INEPT transfer steps of the 3D ¹⁵N-separated NOE-HSQC pulse sequence resulted in a ~1.5-fold increase in sensitivity for the Arg Nε–Hε signals at 800 MHz. For the 3D HNCACB experiment, five ¹⁵N Abramovich-Vega pulses were incorporated for broadband inversion and refocusing, and the sensitivity of Arg¹Hε-¹⁵Nε-¹³Cγ/¹³Cδ correlation peaks was enhanced by a factor of ~1.7 at 500 MHz. These experiments eliminate the necessity for additional experiments to assign Arg ¹Hε and ¹⁵Nε resonances. In addition, the increased sensitivity afforded for the detection of NOE cross-peaks involving correlations with the ¹⁵Nε/¹Hε of Arg in 3D ¹⁵N-separated NOE experiments should prove to be very useful for structural analysis of interactions involving Arg side-chains.     

Domain orientation in β-cyclodextrin-loaded maltose binding protein: Diffusion anisotropy measurements confirm the results of a dipolar coupling study

Maltose binding protein (MBP) is a 370-residue two-domain molecule involved in bacterial chemotaxis and sugar uptake. Rotational diffusion tensors were calculated for a complex between MBP and -cyclodextrin using backbone ¹⁵N T₁ and T₁ relaxation times and steady state ¹H-¹⁵N NOE values. The tensors obtained for each of the two domains in the protein were subsequently used to determine the relative domain orientation in the molecule. The average domain orientation determined using this approach agrees well with results from dipolar coupling data, but differs significantly from the domain orientation deduced from X-ray studies of the complex.     

Backbone assignments and secondary structure of the Escherichia coli enzyme-II mannitol A domain determined by heteronuclear three-dimensional NMR spectroscopy.

This report presents the backbone assignments and the secondary structure determination of the A domain of the Escherichia coli mannitol transport protein, enzyme-IImtl. The backbone resonances were partially assigned using three-dimensional heteronuclear 1H NOE 1H-15N single-quantum coherence (15N NOESY-HSQC) spectroscopy and three-dimensional heteronuclear 1H total correlation 1H-15N single-quantum coherence (15N TOCSY-HSQC) spectroscopy on uniformly 15N enriched protein. Triple-resonance experiments on uniformly 15N/13C enriched protein were necessary to complete the backbone assignments, due to overlapping 1H and 15N frequencies. Data obtained from three-dimensional 1H-15N-13C alpha correlation experiments (HNCA and HN(CO)CA), a three-dimensional 1H-15N-13CO correlation experiment (HNCO), and a three-dimensional 1H alpha-13C alpha-13CO correlation experiment (COCAH) were combined using SNARF software, and yielded the assignments of virtually all observed backbone resonances. Determination of the secondary structure of IIAmtl is based upon NOE information from the 15N NOESY-HSQC and the 1H alpha and 13C alpha secondary chemical shifts. The resulting secondary structure is considerably different from that reported for IIAglc of E. coli and Bacillus subtilis determined by NMR and X-ray.     

Effect of disulfide bridge formation on the NMR spectrum of a protein: Studies on oxidized and reduced Escherichia coli thioredoxin

Summary As a prelude to complete structure calculations of both the oxidized and reduced forms of Escherchia coli thioredoxin (M_r 11 700), we have analyzed the NMR data obtained for the two proteins under identical conditions. The complete aliphatic ¹³C assignments for both oxidized and reduced thioredoxin are reported. Correlations previously noted between ¹³C chemical shifts and secondary structure are confirmed in this work, and significant differences are observed in the C and C shifts between cis- and trans-proline, consistent with previous work that identifies this as a simple and unambiguous method of identifying cis-proline residues in proteins. Reduction of the disulfide bond in the active-site Cys³²-Gly-Pro-Cys³⁵ sequence causes changes in the ¹H, ¹⁵N and ¹³C chemical shifts of residues close to the active site, some of them quite far distant in the amino acid sequence. Coupling constants, both backbone and side chain, show some differences between the two proteins, and the NOE connectivities and chemical shifts are consistent with small changes in the positions of several side chains, including the two tryptophan rings (Trp²⁸ and Trp³¹). These results show that, consistent with the biochemical behavior of thioredoxin, there are minimal differences in backbone configuration between the oxidized and reduced forms of the protein.     

2D and 3D TROSY-enhanced NOESY of 15N labeled proteins

Recently, several TROSY-based experiments have been designed for backbone chemical shift assignment and measurement of the NOEs of ²H, ¹³C and ¹⁵N labeled proteins. Here, we present TROSY-enhanced NOESY experiments, namely the 2D S3E-NOESY-S3E, 3D TROSY-NOESY-S3E and S3E-NOESY-TROSY experiments. These experiments use the spin-state selective excitation method (S3E), and have the TROSY effect in all the indirectly and directly detected dimensions, and so provide optimal resolution for amide protons. The first two experiments provide an additional useful feature in that the diagonal peaks of the amide proton region are cancelled or greatly reduced, allowing clear identification of NOE cross peaks that are close to diagonal peaks.     

Direct methods and residue type specific isotope labeling in NMR structure determination and model-driven sequential assignment

Direct methods in NMR based structure determination start from an unassigned ensemble of unconnected gaseous hydrogen atoms. Under favorable conditions they can produce low resolution structures of proteins. Usually a prohibitively large number of NOEs is required, to solve a protein structure ab-initio, but even with a much smaller set of distance restraints low resolution models can be obtained which resemble a protein fold. One problem is that at such low resolution and in the absence of a force field it is impossible to distinguish the correct protein fold from its mirror image. In a hybrid approach these ambiguous models have the potential to aid in the process of sequential backbone chemical shift assignment when ¹³C^β and ¹³C′ shifts are not available for sensitivity reasons. Regardless of the overall fold they enhance the information content of the NOE spectra. These, combined with residue specific labeling and minimal triple-resonance data using ¹³C^α connectivity can provide almost complete sequential assignment. Strategies for residue type specific labeling with customized isotope labeling patterns are of great advantage in this context. Furthermore, this approach is to some extent error-tolerant with respect to data incompleteness, limited precision of the peak picking, and structural errors caused by misassignment of NOEs. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Identification of HN-methyl NOEs in large proteins using simultaneous amide-methyl TROSY-based detection

A pair of HN-methyl NOESY experiments that are based on simultaneous TROSY-type detection of amide and methyl groups is described. The preservation of cross-peak symmetry in the simultaneous ¹H–¹⁵N/¹³CH₃ NOE spectra enables straightforward assignments of HN-methyl NOE cross-peaks in large and complex protein structures. The pulse schemes are designed to preserve the slowly decaying components of both ¹H–¹⁵N and methyl ¹³CH₃ spin-systems in the course of indirect evolution (t ₂) and acquisition period (t ₃) of 3D NOESY experiments. The methodology has been tested on {U-[¹⁵N,²H]; Ileδ1-[¹³CH₃]; Leu,Val-[¹³CH₃,¹²CD₃]}-labeled 82-kDa enzyme Malate Synthase G (MSG). A straightforward procedure that utilizes the symmetry of NOE cross-peaks in the time-shared 3D NOE data sets allows unambiguous assignments of more than 300 HN-methyl interactions in MSG from a single 3D data set providing important structural restraints for derivation of the backbone global fold.     

3D 13C/1H NMR-based assignments for side-chain resonances of Lactobacillus casei dihydrofolate reductase. Evidence for similarities between the solution and crystal structures of the enzyme

Summary ¹³C-based three-dimensional ¹H–¹H correlation experiments have been used to determine essentially complete ¹³C and ¹H resonance assignments for the amino acid side chains of uniformly ¹³C/¹⁵N labelled L. casei dihydrofolate reductase in a complex with the drug methotrexate. Excellent agreement is observed between these assignments and an earlier set of partial assignments made on the basis of correlating nuclear Overhauser effect and crystal structure data, indicating that the tertiary structure of the enzyme is similar in solution and in the crystal state.To whom correspondence should be addressed.     

3D structure of bovine pancreatic ribonuclease A in aqueous solution: An approach to tertiary structure determination from a small basis of1H NMR NOE correlations

Summary A method is proposed to generate initial structures in cases where the distance geometry method may fail, such as when the set of¹H NMR NOE-based distance constraints is small in relation to the size of the protein. The method introduces an initial correlation between the and backbone angles (based on empirical observations) which is relaxed in later stages of the calculation. The obtained initial structures are refined by well-established methods of energy minimization and restrained molecular dynamics. The method is applied to determine the solution structure of Ribonuclease A (124 residues) from a NOE basis consisting of 467 NOE cross-correlations (97 intra-residue, 206 sequential, 23 medium-range and 141 long-range) obtained at 360 MHz. The global shape and backbone overall fold of the eight final refined structures are close to those shown by the crystal structure. A meaningful difference in the positioning of the catalytically important His¹¹⁹ side chain in the solution and crystal structures has been detected.     

A suite of Mathematica notebooks for the analysis of protein main chain 15N NMR relaxation data

A suite of Mathematica notebooks has been designed to ease the analysis of protein main chain ¹⁵N NMR relaxation data collected at a single magnetic field strength. Individual notebooks were developed to perform the following tasks: nonlinear fitting of ¹⁵N-T ₁ and -T ₂ relaxation decays to a two parameter exponential decay, calculation of the principal components of the inertia tensor from protein structural coordinates, nonlinear optimization of the principal components and orientation of the axially symmetric rotational diffusion tensor, model-free analysis of ¹⁵N-T ₁, -T ₂, and {¹H}–¹⁵N NOE data, and reduced spectral density analysis of the relaxation data. The principle features of the notebooks include use of a minimal number of input files, integrated notebook data management, ease of use, cross-platform compatibility, automatic visualization of results and generation of high-quality graphics, and output of analyses in text format.L. Spyracopoulos is an AHFMR Medical Research Senior Scholar     

Protein NMR structure determination with automated NOE-identification in the NOESY spectra using the new software ATNOS

Novel algorithms are presented for automated NOESY peak picking and NOE signal identification in homonuclear 2D and heteronuclear-resolved 3D [¹H,¹H]-NOESY spectra during de novoprotein structure determination by NMR, which have been implemented in the new software ATNOS (automated NOESY peak picking). The input for ATNOS consists of the amino acid sequence of the protein, chemical shift lists from the sequence-specific resonance assignment, and one or several 2D or 3D NOESY spectra. In the present implementation, ATNOS performs multiple cycles of NOE peak identification in concert with automated NOE assignment with the software CANDID and protein structure calculation with the program DYANA. In the second and subsequent cycles, the intermediate protein structures are used as an additional guide for the interpretation of the NOESY spectra. By incorporating the analysis of the raw NMR data into the process of automated de novoprotein NMR structure determination, ATNOS enables direct feedback between the protein structure, the NOE assignments and the experimental NOESY spectra. The main elements of the algorithms for NOESY spectral analysis are techniques for local baseline correction and evaluation of local noise level amplitudes, automated determination of spectrum-specific threshold parameters, the use of symmetry relations, and the inclusion of the chemical shift information and the intermediate protein structures in the process of distinguishing between NOE peaks and artifacts. The ATNOS procedure has been validated with experimental NMR data sets of three proteins, for which high-quality NMR structures had previously been obtained by interactive interpretation of the NOESY spectra. The ATNOS-based structures coincide closely with those obtained with interactive peak picking. Overall, we present the algorithms used in this paper as a further important step towards objective and efficient de novoprotein structure determination by NMR.     

HNCAN pulse sequences for sequential backbone resonance assignment across proline residues in perdeuterated proteins

A TROSY-based triple-resonance pulse scheme is described which correlates backbone ¹H and ¹⁵N chemical shifts of an amino acid residue with the ¹⁵N chemical shifts of both the sequentially preceding and following residues. The sequence employs ¹ J _NC and ² J _NC couplings in two sequential magnetization transfer steps in an `out-and-back' manner. As a result, N,N connectivities are obtained irrespective of whether the neighbouring amide nitrogens are protonated or not, which makes the experiment suitable for the assignment of proline resonances. Two different three-dimensional variants of the pulse sequence are presented which differ in sensitivity and resolution to be achieved in one of the nitrogen dimensions. The new method is demonstrated with two uniformly ²H/¹³C/¹⁵N-labelled proteins in the 30-kDa range.     

Assignment Strategies for Large Proteins by Magic-Angle Spinning NMR: The 21-kDa Disulfide-Bond-Forming Enzyme DsbA

We present strategies for chemical shift assignments of large proteins by magic-angle spinning solid-state NMR, using the 21-kDa disulfide-bond-forming enzyme DsbA as prototype. Previous studies have demonstrated that complete de novo assignments are possible for proteins up to  ∼ 17 kDa, and partial assignments have been performed for several larger proteins. Here we show that combinations of isotopic labeling strategies, high field correlation spectroscopy, and three-dimensional (3D) and four-dimensional (4D) backbone correlation experiments yield highly confident assignments for more than 90% of backbone resonances in DsbA. Samples were prepared as nanocrystalline precipitates by a dialysis procedure, resulting in heterogeneous linewidths below 0.2 ppm. Thus, high magnetic fields, selective decoupling pulse sequences, and sparse isotopic labeling all improved spectral resolution. Assignments by amino acid type were facilitated by particular combinations of pulse sequences and isotopic labeling; for example, transferred echo double resonance experiments enhanced sensitivity for Pro and Gly residues; [2-¹³C]glycerol labeling clarified Val, Ile, and Leu assignments; in-phase anti-phase correlation spectra enabled interpretation of otherwise crowded Glx/Asx side-chain regions; and 3D NCACX experiments on [2-¹³C]glycerol samples provided unique sets of aromatic (Phe, Tyr, and Trp) correlations. Together with high-sensitivity CANCOCA 4D experiments and CANCOCX 3D experiments, unambiguous backbone walks could be performed throughout the majority of the sequence. At 189 residues, DsbA represents the largest monomeric unit for which essentially complete solid-state NMR assignments have so far been achieved. These results will facilitate studies of nanocrystalline DsbA structure and dynamics and will enable analysis of its 41-kDa covalent complex with the membrane protein DsbB, for which we demonstrate a high-resolution two-dimensional ¹³C-¹³C spectrum.     

Cysteine ligand vibrations are responsible for the complex resonance Raman spectrum of azurin

 In the redox center of azurin, the Cu(II) is strongly coordinated to one thiolate S from Cys 112 and two imidazole Ns from His 46 and 117. This site yields a complex resonance Raman (RR) spectrum with >20 vibrational modes between 200 and 1500 cm^–1. We have investigated the effects of ligand-selective isotope replacements on the RR spectrum of Pseudomonas aeruginosa azurin to determine the relative spectral contribution from each of the copper ligands. Growth on ³⁴S-sulfate labels the cysteine ligand and allows the identification of a cluster of bands with Cu–S(Cys) stretching character between 370 and 430 cm^–1 whose frequencies are consistent with the trigonal or distorted tetrahedral coordination in type 1 sites. In type 2 copper-cysteinate sites, the lower ν (Cu–S) frequencies between 260 and 320 cm^–1 are consistent with square-planar coordination. Addition of exogenous ¹⁵N-labeled imidazole or histidine to the His117Gly mutant generates type 1 or type 2 sites, respectively. Because neither the above nor the His46Gly mutant reconstituted with ¹⁵N-imidazole exhibits significant isotope dependence, the histidine ligands can be ruled out as important contributors to the RR spectrum. Instead, a variety of evidence, including extensive isotope shifts upon global substitution with ¹⁵N, suggests that the multiple RR modes of azurin are due principally to vibrations of the cysteine ligand. These are resonance-enhanced through kinematic coupling with the Cu–S stretch in the ground state or through an excited-state A-term mechanism involving a Cu-cysteinate chromophore that extends into the peptide backbone. Received: 29 July 1996 / Accepted: 9 November 1996