Protein structural domain parsing by consensus reasoning over multiple knowledge sources and methods

Domain parsing, or the detection of signals of protein structural domains from sequence data, is a complex and difficult problem. If carried out reliably it would be a powerful interpretive and predictive tool for genomic and proteomic studies. We report on a novel approach to domain parsing using consensus techniques based on Hidden Markov Models (HMMs) and BLAST searches built from a training set of 1471 continuous structural domains from the Dali Domain Dictionary (DDD). Validation on an independent test sample of family-matched structural domain sequences from the Scop database yields a consensus prediction performance rate of 75.5%, well above the 58% obtained by simple agreement of methods.

Efficient translation of mRNAs in influenza A virus-infected cells is independent of the viral 5' untranslated region

We test the hypothesis that the translation machinery in cells infected by influenza A virus efficiently translates only mRNAs that possess the influenza viral 5' untranslated region (5'-UTR) by introducing mRNAs directly into the cytoplasm of infected cells. This strategy avoids effects due to the inhibition of the nuclear export of cellular mRNAs mediated by the viral NS1 protein. In one approach, we transfect in vitro synthesized mRNAs into infected cells and demonstrate that these mRNAs are efficiently translated whether or not they possess the influenza viral 5'-UTR. In the second approach, an mRNA is synthesized endogenously in the cytoplasm of influenza A virus infected cells by a constitutively expressed T7 RNA polymerase. Although this mRNA is uncapped and lacks the influenza viral 5'-UTR sequence, it is efficiently translated in infected cells via an internal ribosome entry site. We conclude that the translation machinery in influenza A virus infected cells is capable of efficiently translating all mRNAs and that the switch from cellular to virus-specific protein synthesis that occurs during infection results from other processes.

Solution NMR structure and folding dynamics of the N terminus of a rat non-muscle alpha-tropomyosin in an engineered chimeric protein

Tropomyosin is an alpha-helical coiled-coil protein that aligns head-to-tail along the length of the actin filament and regulates its function. The solution structure of the functionally important N terminus of a short 247-residue non-muscle tropomyosin was determined in an engineered chimeric protein, GlyTM1bZip, consisting of the first 19 residues of rat short alpha-tropomyosin and the last 18 residues of the GCN4 leucine zipper. A gene encoding GlyTM1bZip was synthesized, cloned and expressed in Escherichia coli. Triple resonance NMR spectra were analyzed with the program AutoAssign to assign its backbone resonances. Multidimensional nuclear Overhauser effect spectra, X-filtered spectra and (3)J(H(N)-H(alpha)) scalar coupling were analyzed using AutoStructure. This is the first application of this new program to determine the three-dimensional structure of a symmetric homodimer and a structure not previously reported. Residues 7-35 in GlyTM1bZip form a coiled coil, but neither end is helical. Heteronuclear (15)N-(1)H nuclear Overhauser effect data showed that the non-helical N-terminal residues are flexible. The (13)C' chemical shifts of the coiled-coil backbone carbonyl groups in GlyTM1bZip showed a previously unreported periodicity, where resonances arising from residues at the coiled-coil interface in a and d positions of the heptad repeat were displaced relatively upfield and those arising from residues in c positions were displaced relatively downfield. Heteronuclear single quantum coherence spectra, collected as a function of temperature, showed that cross-peaks arising from the alpha-helical backbone and side-chains at the coiled-coil interface broadened or shifted with T(M) values approximately 20 degrees C lower than the loss of alpha-helix measured by circular dichroism, suggesting the presence of a folding intermediate. The side-chain of Ile14, a residue essential for binding interactions, exhibited multiple conformations. The conformational flexibility of the N termini of short tropomyosins may be important for their binding specificity.

SPINE: an integrated tracking database and data mining approach for identifying feasible targets in high-throughput structural proteomics

High-throughput structural proteomics is expected to generate considerable amounts of data on the progress of structure determination for many proteins. For each protein this includes information about cloning, expression, purification, biophysical characterization and structure determination via NMR spectroscopy or X-ray crystallography. It will be essential to develop specifications and ontologies for standardizing this information to make it amenable to retrospective analysis. To this end we created the SPINE database and analysis system for the Northeast Structural Genomics Consortium. SPINE, which is available at bioinfo.mbb.yale.edu/nesg or nesg.org, is specifically designed to enable distributed scientific collaboration via the Internet. It was designed not just as an information repository but as an active vehicle to standardize proteomics data in a form that would enable systematic data mining. The system features an intuitive user interface for interactive retrieval and modification of expression construct data, query forms designed to track global project progress and external links to many other resources. Currently the database contains experimental data on 985 constructs, of which 740 are drawn from Methanobacterium thermoautotrophicum, 123 from Saccharomyces cerevisiae, 93 from Caenorhabditis elegans and the remainder from other organisms. We developed a comprehensive set of data mining features for each protein, including several related to experimental progress (e.g. expression level, solubility and crystallization) and 42 based on the underlying protein sequence (e.g. amino acid composition, secondary structure and occurrence of low complexity regions). We demonstrate in detail the application of a particular machine learning approach, decision trees, to the tasks of predicting a protein's solubility and propensity to crystallize based on sequence features. We are able to extract a number of key rules from our trees, in particular that soluble proteins tend to have significantly more acidic residues and fewer hydrophobic stretches than insoluble ones. One of the characteristics of proteomics data sets, currently and in the foreseeable future, is their intermediate size ( approximately 500-5000 data points). This creates a number of issues in relation to error estimation. Initially we estimate the overall error in our trees based on standard cross-validation. However, this leaves out a significant fraction of the data in model construction and does not give error estimates on individual rules. Therefore, we present alternative methods to estimate the error in particular rules.

Protein NMR spectroscopy in structural genomics

Protein NMR spectroscopy provides an important complement to X-ray crystallography for structural genomics, both for determining three-dimensional protein structures and in characterizing their biochemical and biophysical functions.

Lipari-Szabo mapping: A graphical approach to Lipari-Szabo analysis of NMR relaxation data using reduced spectral density mapping

In this paper, we explore connections between the Lipari-Szabo formalism and reduced spectral density mapping, and show how spectral density estimates can be associated with Lipari-Szabo parameters via a simple geometric construction which we call Lipari-Szabo mapping. This relationship can be used to estimate Lipari-Szabo parameters from spectral density estimates without the need for nonlinear optimization, and to perform 'model selection' in a graphical manner. The Lipari-Szabo map also provides insight into the Lipari-Szabo model, and allows us to determine when a given set of experimental spectral densities are inconsistent with the Lipari-Szabo formalism. Practical applications of Lipari-Szabo mapping in conjunction with more traditional analysis methods are discussed.

A Bayesian statistical method for the detection and quantification of rotational diffusion anisotropy from NMR relaxation data

It has recently become more widely appreciated that the presence of rotational diffusional anisotropy in proteins and other macromolecules can have a significant affect on the interpretation of NMR relaxation data in terms of molecular motion. In this paper, we show how commonly used NMR relaxation data (R(1), R(2), and NOE) obtained at two spectrometer frequencies can be analyzed using a Bayesian statistical approach to reliably detect and quantify the degree of rotational diffusion anisotropy. Our approach differs from previous methods in that it does not make assumptions concerning the internal motions experienced by the residues which are used to quantify the diffusion anisotropy, but rather averages the results over all internal motions consistent with the data. We demonstrate our method using synthetic data corresponding to isotropic, axially symmetric anisotropic, and fully asymmetric anisotropic rotational diffusion, as well as experimental NMR data. We compare the Bayesian statistical approach with a widely used method for extracting tumbling parameters using both synthetic and experimental data. While it can be difficult to separate the effects of chemical exchange from rotational anisotropy using this "standard" method, these effects are readily separated using Bayesian statistics. In addition, we find that the Bayesian statistical approach requires considerably less CPU time than an equivalent standard analysis.

Partial NMR assignments for uniformly (13C, 15N)-enriched BPTI in the solid state

We demonstrate that high-resolution multidimensional solid state NMR methods can be used to correlate many backbone and side chain chemical shifts for hydrated micro-crystalline U-13C,15N Basic Pancreatic Trypsin Inhibitor (BPTI), using a field strength of 800 MHz for protons, magic angle sample spinning rates of 20 kHz and proton decoupling field strengths of 140 kHz. Results from two homonuclear transfer methods, radio frequency driven dipolar recoupling and spin diffusion, were compared. Typical 13C peak line widths are 0.5 ppm, resulting in Calpha-Cbeta and Calpha-CO regions that exhibit many resolved peaks. Two-dimensional carbon-carbon correlation spectra of BPTI have sufficient resolution to identify and correlate many of the spin systems associated with the amino acids. As a result, we have been able to assign a large number of the spin systems in this protein. The agreement between shifts measured in the solid state and those in solution is typically very good, although some shifts near the ion binding sites differ by at least 1.5 ppm. These studies were conducted with approximately 0.2 to 0.4 micromol of enriched material; the sensitivity of this method is apparently adequate for other biological systems as well.

Solution NMR evidence for a cis Tyr-Ala peptide group in the structure of [Pro93Ala] bovine pancreatic ribonuclease A

Proline peptide group isomerization can result in kinetic barriers in protein folding. In particular, the cis proline peptide conformation at Tyr92-Pro93 of bovine pancreatic ribonuclease A (RNase A) has been proposed to be crucial for chain folding initiation. Mutation of this proline-93 to alanine results in an RNase A molecule, P93A, that exhibits unfolding/refolding kinetics consistent with a cis Tyr92-Ala93 peptide group conformation in the folded structure (Dodge RW, Scheraga HA, 1996, Biochemistry 35:1548-1559). Here, we describe the analysis of backbone proton resonance assignments for P93A together with nuclear Overhauser effect data that provide spectroscopic evidence for a type VI beta-bend conformation with a cis Tyr92-Ala93 peptide group in the folded structure. This is in contrast to the reported X-ray crystal structure of [Pro93Gly]-RNase A (Schultz LW, Hargraves SR, Klink TA, Raines RT, 1998, Protein Sci 7:1620-1625), in which Tyr92-Gly93 forms a type-II beta-bend with a trans peptide group conformation. While a glycine residue at position 93 accommodates a type-II bend (with a positive value of phi93), RNase A molecules with either proline or alanine residues at this position appear to require a cis peptide group with a type-VI beta-bend for proper folding. These results support the view that a cis Pro93 conformation is crucial for proper folding of wild-type RNase A.

Comparison of local and global stability of an analogue of a disulfide-folding intermediate with those of the wild-type protein in bovine pancreatic ribonuclease A: identification of specific regions of stable structure along the oxidative folding pathway

We have identified specific regions of the polypeptide chain of bovine pancreatic ribonuclease A (RNase A) that are critical for stabilizing the oxidative folding intermediate des-[40-95] (with three native disulfide bonds but lacking the fourth native Cys40-Cys95 disulfide bond) in an ensemble of largely disordered three-disulfide precursors (3S if des-[40-95]). A stable analogue of des-[40-95], viz., [C40A, C95A] RNase A, which contains three out of four native disulfide pairings, was previously found to have a three-dimensional structure very similar to that of the wild-type protein. However, it is determined here from GdnHCl denaturation experiments to have significantly reduced global stability, i.e., = 4.5 kcal /mol at 20 degrees C and pH 4.6. The local stability of [C40A, C95A] RNase A was also examined using site-specific amide (2)H/(1)H exchange measurements at pD 5.0 to determine the individual unfolding free energy of specific residues under both strongly native (12 degrees C) and more destabilizing (20 degrees C) conditions. Comparison of the relative stabilities at specific amide sites of [C40A, C95A] RNase A at both temperatures with the corresponding values for the wild-type protein at 35 degrees C corroborates previous experimental evidence that unidentified intramolecular contacts in the vicinity of the preferentially formed native one-disulfide (C65-C72) loop are crucial for stabilizing early folding intermediates, leading to des-[40-95]. Moreover, values of for residues at or near the third alpha-helix, and in part of the second beta-sheet of [C40A, C95A] RNase A, indicate that these two regions of regular backbone structure contribute to stabilizing the global chain fold of the des-[40-95] disulfide-folding intermediate in the wild-type protein. More significantly, we have identified numerous specific residues in the first alpha-helix and the first beta-sheet of the protein that are stabilized in the final step of the major oxidative regeneration pathway of RNase A (des-[40-95] --> N).

HYPER: a hierarchical algorithm for automatic determination of protein dihedral-angle constraints and stereospecific C beta H2 resonance assignments from NMR data

A new computer program, HYPER, has been developed for automated analysis of protein dihedral angle values and C beta H2 stereospecific assignments from NMR data. HYPER uses a hierarchical grid-search algorithm to determine allowed values of phi, psi, and chi 1 dihedral angles and C beta H2 stereospecific assignments based on a set of NMR-derived distance and/or scalar-coupling constraints. Dihedral-angle constraints are valuable for restricting conformational space and improving convergence in three-dimensional structure calculations. HYPER computes the set of phi, psi, and chi 1 dihedral angles and C beta H2 stereospecific assignments that are consistent with up to nine intraresidue and sequential distance bounds, two pairs of relative distance bounds, thirteen homo- and heteronuclear scalar coupling bounds, and two pairs of relative scalar coupling constant bounds. The program is designed to be very flexible, and provides for simple user modification of Karplus equations and standard polypeptide geometries, allowing it to accommodate recent and future improved calibrations of Karplus curves. The C code has been optimized to execute rapidly (0.3-1.5 CPU-sec residue-1 using a 5 degrees grid) on Silicon Graphics R8000, R10000 and Intel Pentium CPUs, making it useful for interactive evaluation of inconsistent experimental constraints. The HYPER program has been tested for internal consistency and reliability using both simulated and real protein NMR data sets.

Automated analysis of NMR assignments and structures for proteins

Recent developments in protein NMR technology have provided spectral data that are highly amenable to analysis by advanced computer software systems. Specific data collection strategies, coupled with these computer programs, allow automated analysis of extensive backbone and sidechain resonance assignments and three-dimensional structures for proteins of 50 to 200 amino acids.

Estimation of dynamic parameters from NMR relaxation data using the Lipari-Szabo model-free approach and Bayesian statistical methods

In order to analyze NMR relaxation data in terms of parameters which describe internal motion, one must first obtain a description of the overall tumbling of the macromolecule in solution. Methods currently used to estimate these global parameters may not always provide reliable estimates of their values and uncertainties. In this paper, we present a general data analysis formalism based on products of Bayesian marginal probability densities which can be used to efficiently combine the information content from multiple experiments, such as R(1), R(2), and NOE data collected at multiple magnetic field strengths, or data from cross-correlation or rotating frame relaxation dispersion experiments. Our approach allows the estimation of global tumbling and internal dynamical parameters and their uncertainties without some of the assumptions which are made in the commonly-used methods for model-selection and global parameter estimation. Compared to an equivalent classical statistical approach, the Bayesian method not only is more computationally efficient, but also provides greater insight into the information content of the data. We demonstrate that this approach can be used to estimate both the isotropic rotational correlation time in the context of the original and "extended" Lipari-Szabo formalisms [Lipari & Szabo, J. Am. Chem. Soc. 1982, 104, 4546; Clore et al., J. Am. Chem. Soc. 1990, 112, 4989], as well as the rotational diffusion coefficients for axially symmetric anisotropic tumbling.

Sensitivity-enhanced sim-CT HMQC PFG-HBHA(CO)NH and PFG-CBCA(CO)NH triple-resonance experiments

Short transverse relaxation times of Calpha and Cbeta single-quantum coherences reduce the sensitivity of triple-resonance experiments involving transfers of Calpha/Cbeta or Halpha/Hbeta coherences. Multiple-quantum line-narrowing techniques improve the relaxation properties of 13C coherences, thereby increasing the sensitivity of the experiment. In the present work, we describe PFG-CBCA(CO)NH and PFG-HBHA(CO)NH experiments that utilize heteronuclear multiple-quantum coherences in a simultaneous constant-time period to obtain completely decoupled spectra with improved sensitivity. Results indicate that approximately 30% of cross peaks show an average enhancement of approximately 15% in the CBCA(CO)NH experiment. In the related HBHA(CO)NH experiment, approximately 97% of the cross peaks show an average enhancement of approximately 40%.

RNA binding by the novel helical domain of the influenza virus NS1 protein requires its dimer structure and a small number of specific basic amino acids

The RNA-binding/dimerization domain of the NS1 protein of influenza A virus (73 amino acids in length) exhibits a novel dimeric six-helical fold. It is not known how this domain binds to its specific RNA targets, one of which is double-stranded RNA. To elucidate the mode of RNA binding, we introduced single alanine replacements into the NS1 RNA-binding domain at specific positions in the three-dimensional structure. Our results indicate that the dimer structure is essential for RNA binding, because any alanine replacement that causes disruption of the dimer also leads to the loss of RNA-binding activity. Surprisingly, the arginine side chain at position 38, which is in the second helix of each monomer, is the only amino-acid side chain that is absolutely required only for RNA binding and not for dimerization, indicating that this side chain probably interacts directly with the RNA target. This interaction is primarily electrostatic, because replacement of this arginine with lysine had no effect on RNA binding. A second basic amino acid, the lysine at position 41, which is also in helix 2, makes a strong contribution to the affinity of binding. We conclude that helix 2 and helix 2', which are antiparallel and next to each other in the dimer conformation, constitute the interaction face between the NS1 RNA-binding domain and its RNA targets, and that the arginine side chain at position 38 and possibly the lysine side chain at position 41 in each of these antiparallel helices contact the phosphate backbone of the RNA target.

Homology modeling of an RNP domain from a human RNA-binding protein: Homology-constrained energy optimization provides a criterion for distinguishing potential sequence alignments

We have recently described an automated approach for homology modeling using restrained molecular dynamics and simulated annealing procedures (Li et al, Protein Sci., 6:956-970,1997). We have employed this approach for constructing a homology model of the putative RNA-binding domain of the human RNA-binding protein with multiple splice sites (RBP-MS). The regions of RBP-MS which are homologous to the template protein snRNP U1A were constrained by "homology distance constraints," while the conformation of the non-homologous regions were defined only by a potential energy function. A full energy function without explicit solvent was employed to ensure that the calculated structures have good conformational energies and are physically reasonable. The effects of mis-alignment of the unknown and the template sequences were also explored in order to determine the feasibility of this homology modeling method for distinguishing possible sequence alignments based on considerations of the resulting conformational energies of modeled structures. Differences in the alignments of the unknown and the template sequences result in significant differences in the conformational energies of the calculated homology models. These results suggest that conformational energies and residual constraint violations in these homology-constrained simulated annealing calculations can be used as criteria to distinguish between correct and incorrect sequence alignments and chain folds.

Propagation of experimental uncertainties using the Lipari-Szabo model-free analysis of protein dynamics

In this paper we make use of the graphical procedure previously described [Jin, D. et al. (1997) J. Am. Chem. Soc., 119, 6923-6924] to analyze NMR relaxation data using the Lipari-Szabo model-free formalism. The graphical approach is advantageous in that it allows the direct visualization of the experimental uncertainties in the motional parameter space. Some general 'rules' describing the relationship between the precision of the relaxation measurements and the precision of the model-free parameters and how this relationship changes with the overall tumbling time (tau m) are summarized. The effect of the precision in the relaxation measurements on the detection of internal motions not close to the extreme narrowing limit is analyzed. We also show that multiple timescale internal motions may be obscured by experimental uncertainty, and that the collection of relaxation data at very high field strength can improve the ability to detect such deviations from the simple Lipari-Szabo model.

Solution NMR structure and backbone dynamics of the major cold-shock protein (CspA) from Escherichia coli: evidence for conformational dynamics in the single-stranded RNA-binding site

The major cold-shock protein (CspA) from Escherichia coli is a single-stranded nucleic acid-binding protein that is produced in response to cold stress. We have previously reported its overall chain fold as determined by NMR spectroscopy [Newkirk, K., Feng, W., Jiang, W., Tejero, R., Emerson, S. D., Inouye, M., and Montelione, G. T. (1994) Proc. Natl. Acad. Sci. U.S.A. 91, 5114-5118]. Here we describe the complete analysis of 1H, 13C, and 15N resonance assignments for CspA, together with a refined solution NMR structure based on 699 conformational constraints and an analysis of backbone dynamics based on 15N relaxation rate measurements. An extensive set of triple-resonance NMR experiments for obtaining the backbone and side chain resonance assignments were carried out on uniformly 13C- and 15N-enriched CspA. Using a subset of these triple-resonance experiments, the computer program AUTOASSIGN provided automatic analysis of sequence-specific backbone N, Calpha, C', HN, Halpha, and side chain Cbeta resonance assignments. The remaining 1H, 13C, and 15N resonance assignments for CspA were then obtained by manual analysis of additional NMR spectra. Dihedral angle constraints and stereospecific methylene Hbeta resonance assignments were determined using a new conformational grid search program, HYPER, and used together with longer-range constraints as input for three-dimensional structure calculations. The resulting solution NMR structure of CspA is a well-defined five-stranded beta-barrel with surface-exposed aromatic groups that form a single-stranded nucleic acid-binding site. Backbone dynamics of CspA have also been characterized by 15N T1, T2, and heteronuclear 15N-1H NOE measurements and analyzed using the extended Lipari-Szabo formalism. These dynamic measurements indicate a molecular rotational correlation time taum of 4.88 +/- 0.04 ns and provide evidence for fast time scale (taue < 500 ps) dynamics in surface loops and motions on the microsecond to millisecond time scale within the proposed nucleic acid-binding epitope.

Modification of a receptor-binding surface of epidermal growth factor (EGF): analogs with enhanced receptor affinity at low pH or at neutrality

Six mutants of human epidermal growth factor (EGF), which carry single point substitutions within a surface patch proposed to juxtapose the bound receptor, were prepared and characterized for receptor affinity and mitogenicity. Receptor affinities relative to EGF are G12Q > H16D > Y13W > Q43A approximately = H16A approximately = EGF >> L15A. Notably, the reduced receptor affinity of mutant L15A indicates that Leu15 probably contributes substantially to receptor binding whereas unaltered receptor affinities observed for analogs H16A and Q43A indicate that neither His16 nor Gln43 contributes significantly to this interaction. On the other hand, the observed enhanced receptor affinities of analogs G12Q, Y13W and H16D highlight surface loci where additional productive receptor-binding contacts can be introduced. Interestingly, at acidic pH analog H16A reveals substantially greater receptor affinity than that of EGF, a property which may offer enhanced therapeutic utility in acidic environments in vivo.

The structure of the N-terminus of striated muscle alpha-tropomyosin in a chimeric peptide: nuclear magnetic resonance structure and circular dichroism studies

Tropomyosins (TMs) are highly conserved, coiled-coil, actin binding regulatory proteins found in most eukaryotic cells. The amino-terminal domain of 284-residue TMs is among the most conserved and functionally important regions. The first nine residues are proposed to bind to the carboxyl-terminal nine residues to form the "overlap" region between successive TMs, which bind along the actin filament. Here, the structure of the N-terminus of muscle alpha-TM, in a chimeric peptide, TMZip, has been solved using circular dichroism (CD) and two-dimensional proton nuclear magnetic resonance (2D 1H NMR) spectroscopy. Residues 1-14 of TMZip are the first 14 N-terminal residues of rabbit striated alpha-TM, and residues 15-32 of TMZip are the last 18 C-terminal residues of the yeast GCN4 transcription factor. CD measurements show that TMZip forms a two-stranded coiled-coil alpha-helix with an enthalpy of folding of -34 +/- 2 kcal/mol. In 2D1H NMR studies at 15 degrees C, pH 6.4, the peptide exhibits 123 sequential and medium range intrachain NOE cross peaks per chain, characteristic of alpha-helices extending from residue 1 to residue 29, together with 85 long-range NOE cross peaks arising from interchain interactions. The three-dimensional structure of TMZip has been determined using these data plus an additional 509 intrachain constraints per chain. The coiled-coil domain extends to the N-terminus. Amide hydrogen exchange studies, however, suggest that the TM region is less stable than the GCN4 region. The work reported here is the first atomic-resolution structure of any region of TM and it allows insight into the mechanism of the function of the highly conserved N-terminal domain.

Crystal structure of the unique RNA-binding domain of the influenza virus NS1 protein

The nonstructural protein (NS1 protein) of the influenza A virus binds to several types of RNAs. X-ray crystallographic analysis of the RNA-binding domain reveals a unique topology for the monomer as well as a novel six-helix structure for the dimer.

A novel RNA-binding motif in influenza A virus non-structural protein 1

The solution NMR structure of the RNA-binding domain from influenza virus non-structural protein 1 exhibits a novel dimeric six-helical protein fold. Distributions of basic residues and conserved salt bridges of dimeric NS1(1-73) suggest that the face containing antiparallel helices 2 and 2' forms a novel arginine-rich nucleic acid binding motif.

Structural characterization of an analog of the major rate-determining disulfide folding intermediate of bovine pancreatic ribonuclease A

The major rate-determining step in the oxidative regeneration of bovine pancreatic ribonuclease A (RNase A) proceeds through des-[40-95] RNase A, a three-disulfide intermediate lacking the Cys40-Cys95 disulfide bond. An analog of this intermediate, [C40A, C95A] RNase A, has been characterized in terms of regular backbone structure and thermodynamic stability at pH 4.6. Nearly complete backbone 1H, 15N, and 13C resonances, and most 13Cbeta side-chain resonances have been assigned for the mutant RNase A using triple-resonance NMR data and a computer program, AUTOASSIGN, for automated analysis of resonance assignments. Comparisons of chemical shift data, 3J(1HN-1Halpha) coupling constants, and NOE data for the mutant and wild-type proteins reveal that the overall chain folds of the two proteins are very similar, with localized structural perturbations in the regions spatially adjacent to the mutation sites in [C40A, C95A] RNase A. More significantly, 1H/2H amide exchange and thermodynamic data reveal a global destabilization of the mutant protein characterized by a significant difference in the midpoint of the thermal transition curves (DeltaTm of 21.8 degrees C) and a significant increase in the slowest exchanging backbone amide 1H/2H exchange rates (10(2)-10(6)-fold faster in the hydrophobic core of [C40A, C95 A] RNase A). Comparisons of the entropy DeltaS degrees (T) and enthalpy DeltaH degrees (T) of unfolding between wild-type and [C40A, C95A] RNase A reveal that some of the global destabilization of the mutant protein arises from entropic and enthalpic changes in the folded state. Implications of these observations for understanding the role of des-[40-95] in the folding pathway of RNase A are discussed.

High-resolution solution NMR structure of the Z domain of staphylococcal protein A

Staphylococcal protein A (SpA) is a cell-wall-bound pathogenicity factor from the bacterium Staphylococcus aureus. Because of their small size and immunoglobulin (IgG)-binding activities, domains of protein A are targets for protein engineering efforts and for the development of computational approaches for de novo protein folding. The NMR solution structure of an engineered IgG-binding domain of SpA, the Z domain (an analog of the B domain of SpA), has been determined by simulated annealing with restrained molecular dynamics on the basis of 671 conformational constraints. The Z domain contains three well-defined alpha-helices corresponding to polypeptide segments Lys7 to Leu17 (helix 1), Glu24 to Asp36 (helix 2), and Ser41 to Ala54 (helix 3). A family of ten conformers representing the solution structure of the Z domain was computed by simulated annealing of restrained molecular dynamics using the program CONGEN. The average of the root-mean-square deviations (r.m. s.d.) of the individual NMR conformers, relative to the mean coordinates, for the backbone atoms N, Calpha and C' of residues Phe5 through Ala56 is 0.69 A; the corresponding backbone r.m.s.d. for the three-helical core is 0.44 A. Helices 1, 2 and 3 are antiparallel in orientation (Omega12=-170(+/-4) degrees , Omega13=+16(+/-3) degrees , Omega23=+173(+/-7) degrees ). A comparison of backbone amide hydrogen/deuterium exchange rates in free and IgG-bound Z domains demonstrates that the amide protons of helices 1, 2 and 3 are protected from rapid exchange in both states, indicating that all three helices are also intact in the IgG-bound state. These solution NMR results differ from the previously determined X-ray structure of the similar SpA B domain in complex with the Fc fragment of a human IgG antibody, where helix 3 is not observed in the electron density map and from the solution NMR structure of the B domain, where helix 3 is observed but helix 1 is tilted by approximately 30 degrees with respect to helices 2 and 3. Hydrogen-bonded N-cap and C-cap formation is observed for all three helices of the Z domain; these capping interactions appear to be highly conserved in the five homologous domains of SpA.

Automated analysis of protein NMR assignments using methods from artificial intelligence

An expert system for determining resonance assignments from NMR spectra of proteins is described. Given the amino acid sequence, a two-dimensional 15N-1H heteronuclear correlation spectrum and seven to eight three-dimensional triple-resonance NMR spectra for seven proteins, AUTOASSIGN obtained an average of 98% of sequence-specific spin-system assignments with an error rate of less than 0.5%. Execution times on a Sparc 10 workstation varied from 16 seconds for smaller proteins with simple spectra to one to nine minutes for medium size proteins exhibiting numerous extra spin systems attributed to conformational isomerization. AUTOASSIGN combines symbolic constraint satisfaction methods with a domain-specific knowledge base to exploit the logical structure of the sequential assignment problem, the specific features of the various NMR experiments, and the expected chemical shift frequencies of different amino acids. The current implementation specializes in the analysis of data derived from the most sensitive of the currently available triple-resonance experiments. Potential extensions of the system for analysis of additional types of protein NMR data are also discussed.

NMR structural analysis of an analog of an intermediate formed in the rate-determining step of one pathway in the oxidative folding of bovine pancreatic ribonuclease A: automated analysis of 1H, 13C, and 15N resonance assignments for wild-type and [C65S, C72S] mutant forms

A three-disulfide intermediate, des-[65-72] RNase A, lacking the disulfide bond between Cys65 and Cys72, is formed in one of the rate-determining steps of the oxidative regeneration pathways of bovine pancreatic ribonuclease A (RNase A). An analog of this intermediate, [C65S, C72S] RNase A, has been characterized in terms of structure and thermodynamic stability. Triple-resonance NMR data were analyzed using an automated assignment program, AUTOASSIGN. Nearly all backbone 1H, 13C, and 15N resonances and most side-chain 13C(beta) resonances of both wild-type (wt) and [C65S, C72S] RNase A were assigned unambiguously. Analysis of NOE, 13C(alpha) chemical shift, and 3J(H(N)-H(alpha)) scalar coupling data indicates that the regular backbone structure of the major form of [C65S, C72S] RNase A is very similar to that of the major form of wt RNase A, although small structural differences are indicated in the mutation site and in spatially adjacent beta-sheet structures comprising the hydrophobic core. Thermodynamic analysis demonstrates that [C65S, C72S] RNase A (Tm of 38.5 degrees C) is significantly less stable than wt RNase A (Tm of 55.5 degrees C) at pH 4.6. Although the structural comparison of wt RNase A and this analog of an oxidative folding intermediate indicates only localized effects around the Cys65 and Cys72 sites, these thermodynamic measurements indicate that formation of the fourth disulfide bond, Cys65-Cys72, on this oxidative folding pathway results in global stabilization of the native chain fold. This conclusion is supported by comparisons of amide 1H/2H exchange rates which are significantly faster throughout the entire structure of [C65S, C72S] RNase A than in wt RNase A. More generally, our study indicates that the C65-C72 disulfide bond of RNase A contributes significantly in stabilizing the structure of the hydrophobic core of the native protein. Formation of this disulfide bond in the final step of this oxidative folding pathway provides significant stabilization of the native-like structure that is present in the corresponding three-disulfide folding intermediate.

Homology modeling using simulated annealing of restrained molecular dynamics and conformational search calculations with CONGEN: application in predicting the three-dimensional structure of murine homeodomain Msx-1

We have developed an automatic approach for homology modeling using restrained molecular dynamics and simulated annealing procedures, together with conformational search algorithms available in the molecular mechanics program CONGEN (Bruccoleri RE, Karplus M, 1987, Biopolymers 26:137-168). The accuracy of the method is validated by "predicting" structures of two homeodomain proteins with known three-dimensional structures, and then applied to predict the three-dimensional structure of the homeodomain of the murine Msx-1 transcription factor. Regions of the unknown protein structure that are highly homologous to the known template structure are constrained by "homology distance constraints," whereas the conformations of nonhomologous regions of the unknown protein are defined only by the potential energy function. A full energy function (excluding explicit solvent) is employed to ensure that the calculated structures have good conformational energies and are physically reasonable. As in NMR structure determinations, information on the consistency of the structure prediction is obtained by superposition of the resulting family of protein structures. In this paper, our homology modeling algorithm is described and compared with related homology modeling methods using spatial constraints derived from the structures of homologous proteins. The software is then used to predict the DNA-bound structures of three homeodomain proteins from the X-ray crystal structure of the engrailed homeodomain protein (Kissinger CR et al., 1990, Cell 63:579-590). The resulting backbone and side-chain conformations of the modeled yeast Mat alpha 2 and D. melanogaster Antennapedia homeodomains are excellent matches to the corresponding published X-ray crystal (Wolberger C et al., 1991, Cell 67:517-528) and NMR (Billeter M et al., 1993, J Mol Biol 234:1084-1097) structures, respectively. Examination of these structures of Msx-1 reveals a network of highly conserved surface salt bridges that are proposed to play a role in regulating protein-protein interactions of homeodomains in transcription complexes.

Application of multiple-quantum line narrowing with simultaneous 1H and 13C constant-time scalar-coupling evolution in PFG-HACANH and PFG-HACA(CO)NH triple-resonance experiments

Many triple-resonance experiments make use of one-bond heteronuclear scalar couplings toestablish connectivities among backbone and/or side-chain nuclei. In medium-sized(15-30 kDa) proteins, short transverse relaxation times of Calpha single-quantum stateslimit signal-to-noise (S/N) ratios. These relaxation properties can be improved usingheteronuclear multiple-quantum coherences (HMQCs) instead of heteronuclear single-quantumcoherences (HSQCs) in the pulse sequence design. In slowly tumbling macromolecules, theseHMQCs can exhibit significantly better transverse relaxation properties than HSQCs.However, HMQC-type experiments also exhibit resonance splittings due to multiple two- andthree-bond homo- and heteronuclear scalar couplings. We describe here a family of pulsed-field gradient (PFG) HMQC-type triple-resonance experiments using simultaneous 1H and13C constant-time (CT) periods to eliminate the t1 dependence of these scalar couplingeffects. These simultaneous CT PFG-(HA)CANH and PFG-(HA)CA(CO)NH HMQC-typeexperiments exhibit sharper resonance line widths and often have better S/N ratios than thecorresponding HSQC-type experiments. Results on proteins ranging in size from 6 to 30 kDashow average methine CalphaH HMQC:HSQC enhancement factors of 1.10 +/- 0.15, withabout 40% of the cross peaks exhibiting better S/N ratios in the simultaneous CT-HMQCversions compared with the HSQC versions.

Phase labeling of C-H and C-C spin-system topologies: application in constant-time PFG-CBCA(CO)NH experiments for discriminating amino acid spin-system types

Triple-resonance experiments facilitate the determination of sequence-specific resonance assignments of medium-sized 13C,15N-enriched proteins. Some triple-resonance experiments can also be used to obtain information about amino acid spin-system topologies by proper delay tuning. The constant-time PFG-CBCA(CO)NH experiment allows discrimination between five different groups of amino acids by tuning (phase labeling) independently the delays for proton-carbon refocusing and carbon-carbon constant-time frequency labeling. The proton-carbon refocusing delay allows discrimination of spin-system topologies based on the number of protons attached to C alpha and C beta atoms (i.e. C-H phase labeling). In addition, tuning of the carbon-carbon constant-time frequency-labeling delay discriminates topologies based on the number of carbons directly coupled to C alpha and C beta atoms (i.e. C-C phase labeling). Classifying the spin systems into these five groups facilitates identification of amino acid types, making both manual and automated analysis of assignments easier. The use of this pair of optimally tuned PFG-CBCA(CO)NH experiments for distinguishing five spin-system topologies is demonstrated for the 124-residue bovine pancreatic ribonuclease A protein.

Phase labeling of C-H and C-C spin-system topologies: application in PFG-HACANH and PFG-HACA(CO)NH triple-resonance experiments for determining backbone resonance assignments in proteins

Triple-resonance experiments can be designed to provide useful information on spin-system topologies. In this paper we demonstrate optimized proton and carbon versions of PFG-CT-HACANH and PFG-CT-HACA(CO)NH 'straight-through' triple-resonance experiments that allow rapid and almost complete assignments of backbone H(alpha), 13C(alpha), 15N and H(N) resonances in small proteins. This work provides a practical guide to using these experiments for determining resonance assignments in proteins, and for identifying both intraresidue and sequential connections involving glycine residues. Two types of delay tunings within these pulse sequences provide phase discrimination of backbone Gly C(alpha) and H(alpha) resonances: (i) C-H phase discrimination by tuning of the refocusing period tau(a_f); (ii) C-C phase discrimination by tuning of the 13C constant-time evolution period 2T(c). For small proteins, C-C phase tuning provides better S/N ratios in PFG-CT-HACANH experiments while C-H phase tuning provides better S/N ratios in PFG-CT-HACA(CO)NH. These same principles can also be applied to triple-resonance experiments utilizing 13C-13C COSY and TOCSY transfer from peripheral side-chain atoms with detection of backbone amide protons for classification of side-chain spin-system topologies. Such data are valuable in algorithms for automated analysis of resonance assignments in proteins.

Simulated annealing with restrained molecular dynamics using a flexible restraint potential: theory and evaluation with simulated NMR constraints

A new functional representation of NMR-derived distance constraints, the flexible restraint potential, has been implemented in the program CONGEN (Bruccoleri RE, Karplus M, 1987, Biopolymers 26:137-168) for molecular structure generation. In addition, flat-bottomed restraint potentials for representing dihedral angle and vicinal scalar coupling constraints have been introduced into CONGEN. An effective simulated annealing (SA) protocol that combines both weight annealing and temperature annealing is described. Calculations have been performed using ideal simulated NMR constraints, in order to evaluate the use of restrained molecular dynamics (MD) with these target functions as implemented in CONGEN. In this benchmark study, internuclear distance, dihedral angle, and vicinal coupling constant constraints were calculated from the energy-minimized X-ray crystal structure of the 46-amino acid polypeptide crambin (ICRN). Three-dimensional structures of crambin that satisfy these simulated NMR constraints were generated using restrained MD and SA. Polypeptide structures with extended backbone and side-chain conformations were used as starting conformations. Dynamical annealing calculations using extended starting conformations and assignments of initial velocities taken randomly from a Maxwellian distribution were found to adequately sample the conformational space consistent with the constraints. These calculations also show that loosened internuclear constraints can allow molecules to overcome local minima in the search for a global minimum with respect to both the NMR-derived constraints and conformational energy. This protocol and the modified version of the CONGEN program described here are shown to be reliable and robust, and are applicable generally for protein structure determination by dynamical simulated annealing using NMR data.

Simulated annealing with restrained molecular dynamics using CONGEN: energy refinement of the NMR solution structures of epidermal and type-alpha transforming growth factors

The new functionality of the program CONGEN (Bruccoleri RE, Karplus M, 1987, Biopolymers 26:137-168; Bassolino-Klimas D et al., 1996, Protein Sci 5:593-603) has been applied for energy refinement of two previously determined solution NMR structures, murine epidermal growth factor (mEGF) and human type-alpha transforming growth factor (hTGF alpha). A summary of considerations used in converting experimental NMR data into distance constraints for CONGEN is presented. A general protocol for simulated annealing with restrained molecular dynamics is applied to generate NMR solution structures using CONGEN together with real experimental NMR data. A total of 730 NMR-derived constraints for mEGF and 424 NMR-derived constraints for hTGF alpha were used in these energy-refinement calculations. Different weighting schemes and starting conformations were studied to check and/or improve the sampling of the low-energy conformational space that is consistent with all constraints. The results demonstrate that loosened (i.e., "relaxed") sets of the EGF and hTGF alpha internuclear distance constraints allow molecules to overcome local minima in the search for a global minimum with respect to both distance restraints and conformational energy. The resulting energy-refined structures of mEGF and hTGF alpha are compared with structures determined previously and with structures of homologous proteins determined by NMR and X-ray crystallography.

High-level production of uniformly ¹⁵N- and ¹³C-enriched fusion proteins in Escherichia coli

An approach to produce 13C- and 15N-enriched proteins is described. The concept is based on intracellular production of the recombinant proteins in Escherichia coli as fusions to an IgG-binding domain, Z, derived from staphylococcal protein A. The production method provides yields of 40-200 mg/l of isotope-enriched fusion proteins in defined minimal media. In addition, the Z fusion partner facilitates the first purification step by IgG affinity chromatography. The production system is applied to isotope enrichment of human insulin-like growth factor II (IGF-II), bovine pancreatic trypsin inhibitor (BPTI), and Z itself. High levels of protein production are achieved in shaker flasks using totally defined minimal medium supplemented with 13C(6)-glucose and (15NH4)2SO4 as the only carbon and nitrogen sources. Growth conditions were optimized to obtain high protein production levels and high levels of isotope incorporation, while minimizing 13C(6)-glucose usage. Incorporation levels of 13C and/or 15N isotopes in purifies IGF-II, BPTI, and Z were confirmed using mass spectrometry and NMR spectroscopy. More than 99% of total isotope enrichment was obtained using a defined isotope-enriched minimal medium. The optimized systems provide reliable, high-level production of isotope-enriched fusion proteins. They can be used to produce 20-40 mg/l of properly folded Z and BPTI proteins. The production system of recombinant BPTI is state-of-the-art and provides the highest known yield of native refolded BPTI.

The mechanism of binding staphylococcal protein A to immunoglobin G does not involve helix unwinding

Structural changes in staphylococcal protein A (SpA) upon its binding to the constant region (Fc) of immunoglobulin G (IgG) have been studied by nuclear magnetic resonance and circular dichroism (CD) spectroscopy. The NMR solution structure of the engineered IgG-binding domain of SpA, the Z domain (an analogue of the B domain of SpA), has been determined by simulated annealing with molecular dynamics, using 599 distance and dihedral angle constraints. Domain Z contains three alpha-helices in the polypeptide segments Lys7 to His18 (helix 1), Glu25 to Asp36 (helix 2), and Ser41 to Ala54 (helix 3). The overall chain fold is an antiparallel three-helical bundle. This is in contrast to the previously determined X-ray structure of the similar SpA domain B in complex with Fc, where helix 3 is not observed in the electron density map [Deisenhofer, J. (1981) Biochemistry 20, 2361-2370], but similar to the solution NMR structure of domain B, which is also a three-helical bundle structure [Gouda, H., et al. (1992) Biochemistry 31, 9665-9672]. In order to characterize possible secondary structural changes associated with IgG binding, far-UV CD spectra were collected for the Z domain, an engineered repeat of this molecule (ZZ), recombinant Fc from IgG subclass 1 (Fc1), recombinant Fc from IgG subclass 3 (Fc3), and mixtures of Z/Fc1, Z/Fc3, ZZ/Fc1, and ZZ/Fc3. Fc3 was included as a control for possible changes of the CD spectrum in the mixture of noncomplexed molecules, since SpA is known not to bind Fc3. From these CD spectra, it was concluded that the third alpha-helix in Z is not disrupted in its complexes with Fc1. Similar results were obtained for the ZZ molecule. However, in both Z and ZZ there are some perturbations in CD spectra at high energy wavelengths (i.e., lambda < 215 nm) accompanying complex formation. On the basis of the combined CD and NMR results, as well as previously described binding studies of Z mutant proteins to Fc1, we conclude that the Z domain maintains its three-helical bundle structure in the Z-Fc complex, though there may be a small structural change involved in the binding mechanism.

An amino-terminal polypeptide fragment of the influenza virus NS1 protein possesses specific RNA-binding activity and largely helical backbone structure

The NS1 protein of influenza A virus has the unique property of binding to three apparently different RNAs: poly A; a stem-bulge in U6 small nuclear RNA; and double-stranded RNA. One of our major goals is to determine how the NS1 protein recognizes and binds to its several RNA targets. As the first step for conducting structural studies, we have succeeded in identifying a fragment of the NS1 protein that possesses all the RNA-binding activities of the full-length protein. The RNA-binding fragment consists of the 73 amino-terminal amino acids of the protein. We have developed procedures for obtaining large amounts of the polypeptide in pure form. This has enabled us to establish the RNA-binding properties of this polypeptide and to demonstrate that it retains the ability to dimerize exhibited by the full-length protein. In addition, far-UV CD spectroscopy indicates that this RNA-binding polypeptide is largely (approximately 80%) helical, suggesting that the mode of dimerization of the NS1 protein and of its interaction with RNA is mediated, at least in part, by helices.

Automated analysis of nuclear magnetic resonance assignments for proteins

Recent developments in protein NMR technology provide spectral data that are highly amendable to analysis by computer software systems. Automated methods of analysis use constraint satisfaction, pseudoenergy minimization, directed search, neural net, simulated annealing, and/or genetic algorithms to establish sequential links and sequence-specific assignments. The most advanced systems provide automated analysis of complete backbone and extensive side-chain resonance assignments for proteins of 50-150 amino acids.

Classification of amino acid spin systems using PFG HCC(CO)NH-TOCSY with constant-time aliphatic 13C frequency labeling

We have developed a useful strategy for identifying amino acid spin systems and side-chain carbon resonance assignments in small 15N-, 13C-enriched proteins. Multidimensional constant-time pulsed field gradient (PFG) HCC(CO)NH-TOCSY experiments provide side-chain resonance frequency information and establish connectivities between sequential amino acid spin systems. In PFG HCC(CO)NH-TOCSY experiments recorded with a properly tuned constant-time period for frequency labeling of aliphatic 13C resonances, phases of cross peaks provide information that is useful for identifying spin system types. When combined with 13C chemical shift information, these patterns allow identification of the following spin system types: Gly, Ala, Thr, Val, Leu, Ile, Lys, Arg, Pro, long-type (i.e., Gln, Glu and Met), Ser, and AMX-type (i.e., Asp, Asn, Cys, His, Phe, Trp and Tyr).

Crankshaft motions of the polypeptide backbone in molecular dynamics simulations of human type-alpha transforming growth factor

Order parameters for the backbone N-H and C alpha-H bond vectors have been calculated from a 150 ps molecular dynamics (MD) simulation of human type-alpha transforming growth factor in H2O solvent. Two kinds of 'crankshaft motions' of the polypeptide backbone are observed in this MD trajectory. The first involves small-amplitude rocking of the rigid peptide bond due to correlated changes in the backbone dihedral angles psi i-1 and phi i. These high-frequency 'librational crankshaft' motions are correlated with systematically smaller values of motional order parameters for backbone N-H bond vectors compared to C alpha-H bond vectors. In addition, infrequent 'crankshaft flips' of the peptide bond from one local minimum to another are observed for several amino acid residues. These MD simulations demonstrate that comparisons of N-H and C alpha-H order parameters provide a useful approach for identifying crankshaft librational motions in proteins.