Efficient analysis of protein 2D NMR spectra using the software packageEASY

Summary The programEASY supports the spectral analysis of biomacromolecular two-dimensional (2D) nuclear magnetic resonance (NMR) data. It provides a user-friendly, window-based environment in which to view spectra for interactive interpretation. In addition, it includes a number of automated routines for peakpicking, spin-system identification, sequential resonance assignment in polypeptide chains, and cross peak integration. In this uniform environment, all resulting parameter lists can be recorded on disk, so that the paper plots and handwritten notes which normally accompany manual assignment of spectra can be largely eliminated. For example, in a protein structure determination by 2D¹H NMR,EASY accepts the frequency domain datasets as input, and after combined use of the automated and interactive routines it can yield a listing of conformational constraints in the format required as input for the calculation of the 3D structure. The program was extensively tested with current protein structure determinations in our laboratory. In this paper, its main features are illustrated with data on the protein basic pancreatic trypsin inhibitor.     

A general Monte Carlo/simulated annealing algorithm for resonance assignment in NMR of uniformly labeled biopolymers

We describe a general computational approach to site-specific resonance assignments in multidimensional NMR studies of uniformly ¹⁵N,¹³C-labeled biopolymers, based on a simple Monte Carlo/simulated annealing (MCSA) algorithm contained in the program MCASSIGN2. Input to MCASSIGN2 includes lists of multidimensional signals in the NMR spectra with their possible residue-type assignments (which need not be unique), the biopolymer sequence, and a table that describes the connections that relate one signal list to another. As output, MCASSIGN2 produces a high-scoring sequential assignment of the multidimensional signals, using a score function that rewards good connections (i.e., agreement between relevant sets of chemical shifts in different signal lists) and penalizes bad connections, unassigned signals, and assignment gaps. Examination of a set of high-scoring assignments from a large number of independent runs allows one to determine whether a unique assignment exists for the entire sequence or parts thereof. We demonstrate the MCSA algorithm using two-dimensional (2D) and three-dimensional (3D) solid state NMR spectra of several model protein samples (α-spectrin SH3 domain and protein G/B1 microcrystals, HET-s_218–289 fibrils), obtained with magic-angle spinning and standard polarization transfer techniques. The MCSA algorithm and MCASSIGN2 program can accommodate arbitrary combinations of NMR spectra with arbitrary dimensionality, and can therefore be applied in many areas of solid state and solution NMR.     

Automated NMR structure determination of stereo-array isotope labeled ubiquitin from minimal sets of spectra using the SAIL-FLYA system

Stereo-array isotope labeling (SAIL) has been combined with the fully automated NMR structure determination algorithm FLYA to determine the three-dimensional structure of the protein ubiquitin from different sets of input NMR spectra. SAIL provides a complete stereo- and regio-specific pattern of stable isotopes that results in sharper resonance lines and reduced signal overlap, without information loss. Here we show that as a result of the superior quality of the SAIL NMR spectra, reliable, fully automated analyses of the NMR spectra and structure calculations are possible using fewer input spectra than with conventional uniformly ¹³C/¹⁵N-labeled proteins. FLYA calculations with SAIL ubiquitin, using a single three-dimensional “through-bond” spectrum (and 2D HSQC spectra) in addition to the ¹³C-edited and ¹⁵N-edited NOESY spectra for conformational restraints, yielded structures with an accuracy of 0.83–1.15 Å for the backbone RMSD to the conventionally determined solution structure of SAIL ubiquitin. NMR structures can thus be determined almost exclusively from the NOESY spectra that yield the conformational restraints, without the need to record many spectra only for determining intermediate, auxiliary data of the chemical shift assignments. The FLYA calculations for this report resulted in 252 ubiquitin structure bundles, obtained with different input data but identical structure calculation and refinement methods. These structures cover the entire range from highly accurate structures to seriously, but not trivially, wrong structures, and thus constitute a valuable database for the substantiation of structure validation methods.     

Iterative algorithm of discrete Fourier transform for processing randomly sampled NMR data sets

Spectra obtained by application of multidimensional Fourier Transformation (MFT) to sparsely sampled nD NMR signals are usually corrupted due to missing data. In the present paper this phenomenon is investigated on simulations and experiments. An effective iterative algorithm for artifact suppression for sparse on-grid NMR data sets is discussed in detail. It includes automated peak recognition based on statistical methods. The results enable one to study NMR spectra of high dynamic range of peak intensities preserving benefits of random sampling, namely the superior resolution in indirectly measured dimensions. Experimental examples include 3D ¹⁵N- and ¹³C-edited NOESY-HSQC spectra of human ubiquitin.     

Selective suppression of excipient signals in 2D <Superscript>1</Superscript>H–<Superscript>13</Superscript>C methyl spectra of biopharmaceutical products

While the use of ¹H–¹³C methyl correlated NMR spectroscopy at natural isotopic abundance has been demonstrated as feasible on protein therapeutics as large as monoclonal antibodies, spectral interference from aliphatic excipients remains a significant obstacle to its widespread application. These signals can cause large baseline artifacts, obscure protein resonances, and cause dynamic range suppression of weak peaks in non-uniform sampling applications, thus hampering both traditional peak-based spectral analyses as well as emerging chemometric methods of analysis. Here we detail modifications to the 2D ¹H–¹³C gradient-selected HSQC experiment that make use of selective pulsing techniques for targeted removal of interfering excipient signals in spectra of the NISTmAb prepared in several different formulations. This approach is demonstrated to selectively reduce interfering excipient signals while still yielding 2D spectra with only modest losses in protein signal. Furthermore, it is shown that spectral modeling based on the SMILE algorithm can be used to simulate and subtract any residual excipient signals and their attendant artifacts from the resulting 2D NMR spectra.     

AUREMOL-RFAC-3D,combination of R-factors and their use for automated quality assessment of protein solution structures

We present here the computer program AUREMOL-RFAC-3D that is a generalization of the previously published program RFAC for the fully automated estimation of residual indices (R-factors) from 2D NOESY spectra. It is part of the larger AUREMOL software package (www.auremol.de). RFAC-3D calculates R-factors directly from two-dimensional homonuclear NOESY spectra as well as from three-dimensional ¹⁵N or ¹³C edited NOESY-HSQC spectra and thus extends the application range to larger proteins. The fully automated method includes automated peak picking and integration, a Bayesian noise and artifact recognition and the use of the complete relaxation matrix formalism. To enhance the reliability of the calculated R-factors the method is also generalized to calculate combined R-factors from a set of 2D and 3D-spectra. For an optimal combination of the information derived from different sources a plausible formalism had to be derived. In addition, we present a novel direct R-factors based measure that correlates an R-factors as defined in this paper to the root mean square deviation of the actual structure from the optimal structure. The new program has been successfully tested on the histidine containing phosphocarrier protein (HPr) from Staphylococcus carnosus and on the Ras-binding domain (RBD) of the Ral guanine-nucleotide dissociation stimulation factor (RalGDS).     

Proton NMR studies of myohemerythrin from Themiste zostericola

 One- and two-dimensional NMR experiments have been carried out on different forms of myohemerythrin (MHr), a monomeric 13.9-kDa oxygen carrier, focusing on paramagnetically shifted proton resonances. Compared to the corresponding forms of octameric hemerythrin (Hr), all of the MHr forms exhibit spectra with better resolution and signal-to-noise ratios. The metMHr spectra allow the differentiation of the signals from the N_δ-H protons of the five N_ε-coordinated His ligands and those from the bridging Asp and Glu ligands. The 1D spectra of deoxyMHr exhibit a number of relatively sharp features including three solvent-exchangeable peaks that account for five protons. One of these His N-H protons exchanges more slowly with solvent than the other four and is assigned to His 54, which, by analogy to the crystal structure of deoxyHr, is the only His ligand that is hydrogen-bonded to an amino acid residue, Glu24 in this case. One-dimensional NOE results on the non-exchangeable signals clearly show the connectivities among the α and β protons of the bridging Asp111, and the α, β, and γ protons of the bridging Glu58 ligands. One-dimensional NOE experiments performed on the N-H proton signals of the coordinated His ligands, together with the COSY results, help to identify the geminal β protons of the His ligands. Upon the binding of N₃ ^– to one of the Fe(II) sites in deoxyMHr, the overlapping His N_δ-H proton signals observed in the deoxyMHr spectrum are resolved into individual signals; these have been correlated to the corresponding signals in deoxyMHr by saturation transfer experiments. Similarly, all five His N-H protons are resolved in the ¹H NMR spectrum of the deoxy form of the single point mutant L103N MHr. However, all five N-H protons readily exchange with solvent, indicating that the mutation affects the hydrogen-bonding interaction between His54 and Glu24. Received: 20 May 1996 / Accepted: 24 October 1996     

Peak fitting in 2D <Superscript>1</Superscript>H–<Superscript>13</Superscript>C HSQC NMR spectra for metabolomic studies

A modified Lorentzian distribution function is used to model peaks in two-dimensional (2D) ¹H–¹³C heteronuclear single quantum coherence (HSQC) nuclear magnetic resonance (NMR) spectra. The model fit is used to determine accurate chemical shifts from genuine signals in complex metabolite mixtures such as blood. The algorithm can be used to extract features from a set of spectra from different samples for exploratory metabolomics. First a reference spectrum is created in which the peak intensities are given by the median value over all samples at each point in the 2D spectra so that ¹H–¹³C correlations in any spectra are accounted for. The mathematical model provides a footprint for each peak in the reference spectrum, which can be used to bin the ¹H–¹³C correlations in each HSQC spectrum. The binned intensities are then used as variables in multivariate analyses and those found to be discriminatory are rapidly identified by cross referencing the chemical shifts of the bins with a database of ¹³C and ¹H chemical shift correlations from known metabolites.     

Computerized Interpretation of the Electron Ionization Mass Spectra of Nucleoside TMS Derivatives

Abstract

A computer program has been developed for the automated interpretation of mass spectra of TMS derivatives of nucleosides found in human urine. The m/z values in the unknown spectrum are compared to m/z values of 3 different ion series commonly observed in the mass spectra of nucleoside TMS derivatives.¹ If a correlation exists, the unknown spectra are marked with color according to the scheme: 1) blue—molecular ion series, 2) red—base ion series and 3) yellow—sugar ion series. The program suggests a structural assignment for each of the marked ions and calculates a series related ion current. The calculated ion current is used to assign the of sugar contained in the unknown nucleoside.     

Direct evidence for a phenylalanine site in the regulatory domain of phenylalanine hydroxylase

The hydroxylation of phenylalanine to tyrosine by the liver enzyme phenylalanine hydroxylase is regulated by the level of phenylalanine. Whether there is a distinct allosteric binding site for phenylalanine outside of the active site has been unclear. The enzyme contains an N-terminal regulatory domain that extends through Thr117. The regulatory domain of rat phenylalanine hydroxylase was expressed in Escherichia coli. The purified protein behaves as a dimer on a gel filtration column. In the presence of phenylalanine, the protein elutes earlier from the column, consistent with a conformational change in the presence of the amino acid. No change in elution is seen in the presence of the non-activating amino acid proline. ¹H–¹⁵N HSQC NMR spectra were obtained of the ¹⁵N-labeled protein alone and in the presence of phenylalanine or proline. A subset of the peaks in the spectrum exhibits chemical shift perturbation in the presence of phenylalanine, consistent with binding of phenylalanine at a specific site. No change in the NMR spectrum is seen in the presence of proline. These results establish that the regulatory domain of phenylalanine hydroxylase can bind phenylalanine, consistent with the presence of an allosteric site for the amino acid.     

The measurement of transmembrane helices by the deconvolution of CD spectra of membrane proteins: A review

The interpretation of the CD spectra of proteins to date requires additional secondary structural information of the proteins to be analyzed, such as x-ray or nmr data. Therefore, these methods are inappropriate for a CD data base whose secondary structures are unknown, as in the case of the membrane proteins. The Convex Constraint Analysis algorithm [A. Perczel, M. Hollósi, G. Tusnády, and G. D. Fasman (1991) Protein Engineering, Vol. 4, 669–679], on the other hand, operates only on a collection of spectral data to extract the common spectral components with their spectral weights. The linear combinations of these derived “pure” CD curves can reconstruct the original data set with great accuracy. For a membrane protein data set, the five-component spectra so obtained from the deconvolution consisted of two different types of α-helices (the α-helix in the soluble domain and the α_T-helix, for the transmembrane α-helix), a β-pleated sheet, a class C-like spectrum related to β-turns, and a spectrum correlated with the unordered conformation. The deconvoluted CD spectrum for the α_T-helix was characterized by a positive red-shifted band in the range 195–200 nm (+95,000 deg cm² dmol^?l), with the intensity of the negative band at 208 nm being slightly less negative than that of the 222 nm band (?50,000 and ?60,000 deg cm² dmol^?1, respectively) in comparison with the regular α-helix, with a positive band at 190 nm and two negative bands at 208 and 222 nm with magnitudes of + 70,000, ?30,000, and ?30,000 deg cm² dmol^?1, respectively. © 1994 John Wiley & Sons, Inc.     

On singularities of molecular relaxation in water solutions

In this paper, we analyzed spectra of liquid water and water solutions in a frequency domain, characteristic of the collective dynamics of water molecules (from 0 to 200 cm^?1). Particular attention is paid to the relaxation processes, one of which is observed in the terahertz region of the spectrum (～5–50 cm^?1). The physical essence of this process at the molecular level is still unclear. Based on data obtained this process is strongly suggested to interpret as a monomolecular relaxation of unbound water molecules.     

A hierarchical MS2/MS3 database search algorithm for automated analysis of phosphopeptide tandem mass spectra

A novel hierarchical MS²/MS³ database search algorithm has been developed to analyze MS²/MS³ phosphopeptides proteomic data. The algorithm is incorporated in an automated database search program, MassMatrix. The algorithm matches experimental MS² spectra against a supplied protein database to determine candidate peptide matches. It then matches the corresponding experimental MS³ spectra against those candidate peptide matches. The MS² and MS³ spectra are used in concert to arrive at peptide matches with overall higher confidence rather than combining MS² and MS³ data searched separately. Receiver operating characteristic analysis showed that hierarchical MS²/MS³ database searches with MassMatrix had better sensitivity and specificity than the two‐stage MS²/MS³ database searches obtained with MassMatrix, MASCOT, and X!Tandem. A greater number of true peptide matches at a given false rate were identified by use of this new algorithm for data collected on both LCQ and LTQ‐FTICR mass spectrometers. The additional MS³ spectral data also improved the overall reliability and the number of true positives (TPs) due to the fact that the TPs of the MS²/MS³ search results had higher scores than those of the MS².     

Escherichia coli B 5-methyltetrahydrofolate-homocysteine cobalamin methyltransferase: circular dichroism of the methylated and propylated holoenzymes

Solutions (35 μm) of [¹⁴C] methylcobalamin (MeB₁₂) enzyme and propylcobalamin (PrB₁₂) enzyme were prepared in 0.1 m phosphate buffer, pH 7.4. Their circular dichroism (CD) spectra were taken over the wavelength region from 310–650 nm and compared to the CD spectra given by the corresponding alkylcobalamins and alkylcobinamides. Although an exact match-up with one of the free alkylcorrinoids was not found, the CD spectrum of the [¹⁴C]MeB₁₂ enzyme most closely resembled that of free, base-on MeB₁₂ at pH 7.4. No evidence for a base-off MeB₁₂-histidine complex on the enzyme was obtained. In contrast, the CD spectrum of the PrB₁₂ enzyme was strongly indicative of PrB₁₂ with its base predominantly off. Upon testing various solvents, it was observed that in benzyl alcohol-ethanol (9:1) the CD spectra of both free MeB₁₂ and PrB₁₂ were altered to mimic more closely those of the [¹⁴C]MeB₁₂ and PrB₁₂ enzymes, respectively. In this solvent mixture, free PrB₁₂ also displayed a base-off type of absorption spectrum.     

Raman scattering of chymotrypsinogen A, ribonuclease, and ovalbumin in the aqueous solution and solid state

The Raman spectra of three globular proteins, beef pancreas chymotrypsinogen A, beef pancreas ribonuclease, and hen egg white ovalbumin have been obtained in the solid state and aqueous solution. X-ray diffraction and circular dichroism evidence have indicated that these proteins have a low α-helical content and a large fraction of the residues in the unordered and β-sheet conformation. The frequencies and intensities of the amide I and amide III lines are consistent with assignments based on the Raman spectra of polypeptides. The intense amide III lines observed in all the spectra would be expected for proteins with a low fraction of the residues in the α-helical conformation. Several spectra changes occur upon dissolution of the proteins in water and may be associated with further hydration of the proteins. The spectrum of thermally denatured chymotrypsinogen is presented. A 3 cm^–1 decrease in the frequency of the amide I line of the protein dissolved in D₂O upon heating was observed. This observation is consistent with a denaturation mechanism allowing only slight changes in the secondary structure but an increase in solvent penetration upon going from the native to the reversibly denatured state.     

A resonance Raman study on the interaction of rifampicin with Escherichia coli RNA polymerase

The technique of resonance Raman spectroscopy has been used to investigate the interaction of the antibiotic rifampicin with Escherichia coli RNA polymerase. Spectra were analyzed by generating the first derivative of each recorded spectrum using the Savitsky-Golay algorithm. The only band that shifted significantly in the resonance Raman spectrum of rifampicin upon the formation of the drug-core polymerase complex was the amide III band. It underwent an 8 cm^?1 shift from 1306 cm^?1 in aqueous solution to 1314 cm^?1. A comparable shift was observed for the rifampicin-holoenzyme complex. Thus, the interaction of the sigma subunit with the core polymerase does not significantly alter the manner in which rifampicin interacts with RNA polymerase. The nature of this shift has been analyzed further by recording the resonance Raman spectrum of rifampicin in a variety of solvents with different hydrogen-bonding ability. In non-hydrogen-bonding solvents (benzene and carbon disulfide) the amide III band was observed at approximately 1220 cm^?1; in dimethyl sulfoxide, a weak hydrogen-bond acceptor, 1274 cm^?1; in water, a strong hydrogen-bonding solvent, 1306 cm^?1; and finally, in triethylamine, a stronger hydrogen-bonding solvent than water, it was observed at 1314 cm^?1. Thus, as the hydrogen-bonding ability of the solvent increased, the amide III band shifted to higher frequency. Based on these results, the rifampicin binding site in RNA polymerase provides a stronger hydrogen-bonding environment for the amidic proton of rifampicin than is encountered when rifampicin is free in aqueous solution.     

Structural differences between apolipoprotein E3 and E4 as measured by 19F NMR

The apolipoprotein E family contains three major isoforms (ApoE4, E3, and E2) that are directly involved with lipoprotein metabolism and cholesterol transport. ApoE3 and apoE4 differ in only a single amino acid with an arginine in apoE4 changed to a cysteine at position 112 in apoE3. Yet only apoE4 is recognized as a risk factor for Alzheimer''s disease. Here we used ¹⁹F NMR to examine structural differences between apoE4 and apoE3 and the effect of the C-terminal domain on the N-terminal domain. After incorporation of 5-¹⁹F-tryptophan the 1D ¹⁹F NMR spectra were compared for the N-terminal domain and for the full length proteins. The NMR spectra of the N-terminal region (residues 1–191) are reasonably well resolved while those of the full length wild-type proteins are broad and ill-defined suggesting considerable conformational heterogeneity. At least four of the seven tryptophan residues in the wild type protein appear to be solvent exposed. NMR spectra of the wild-type proteins were compared to apoE containing four mutations in the C-terminal region that gives rise to a monomeric form either of apoE3 under native conditions (Zhang et al., Biochemistry 2007; 46: 10722–10732) or apoE4 in the presence of 1 M urea. For either wild-type or mutant proteins the differences in tryptophan resonances in the N-terminal region of the protein suggest structural differences between apoE3 and apoE4. We conclude that these differences occur both as a consequence of the Arg158Cys mutation and as a consequence of the interaction with the C-terminal domain.     

Automated backbone assignment of labeled proteins using the threshold accepting algorithm

The sequential assignment of backbone resonances is the first step in the structure determination of proteins by heteronuclear NMR. For larger proteins, an assignment strategy based on proton side-chain information is no longer suitable for the use in an automated procedure. Our program PASTA (Protein ASsignment by Threshold Accepting) is therefore designed to partially or fully automate the sequential assignment of proteins, based on the analysis of NMR backbone resonances plus C information. In order to overcome the problems caused by peak overlap and missing signals in an automated assignment process, PASTA uses threshold accepting, a combinatorial optimization strategy, which is superior to simulated annealing due to generally faster convergence and better solutions. The reliability of this algorithm is shown by reproducing the complete sequential backbone assignment of several proteins from published NMR data. The robustness of the algorithm against misassigned signals, noise, spectral overlap and missing peaks is shown by repeating the assignment with reduced sequential information and increased chemical shift tolerances. The performance of the program on real data is finally demonstrated with automatically picked peak lists of human nonpancreatic synovial phospholipase A₂, a protein with 124 residues.     

Water–Protein Hydrogen Exchange in the Micro-Crystalline Protein Crh as Observed by Solid State NMR Spectroscopy

We report site-resolved observation of hydrogen exchange in the micro-crystalline protein Crh. Our approach is based on the use of proton T₂^′ -selective ¹H–¹³C–¹³C correlation spectra for site-specific assignments of carbons nearby labile protein protons. We compare the proton T₂^′ selective scheme to frequency selective water observation in deuterated proteins, and discuss the impacts of deuteration on ¹³C linewidths in Crh. We observe that in micro-crystalline proteins, solvent accessible hydroxyl and amino protons show comparable exchange rates with water protons as for proteins in solution, and that structural constraints, such as hydrogen bonding or solvent accessibility, more significantly reduce exchange rates.