Proposal for molecular mechanism of thionins deduced from physico-chemical studies of plant toxins.

We propose a molecular model for phospholipid membrane lysis by the ubiquitous plant toxins called thionins. Membrane lysis constitutes the first major effect exerted by these toxins that initiates a cascade of cytoplasmic events leading to cell death. X-ray crystallography, solution nuclear magnetic resonance (NMR) studies, small angle X-ray scattering and fluorescence spectroscopy provide evidence for the mechanism of membrane lysis. In the crystal structures of two thionins in the family, alpha(1)- and beta-purothionins (MW: approximately 4.8 kDa), a phosphate ion and a glycerol molecule are modeled bound to the protein. (31)P NMR experiments on the desalted toxins confirm phosphate-ion binding in solution. Evidence also comes from phospholipid partition experiments with radiolabeled toxins and with fluorescent phospholipids. This data permit a model of the phospholipid-protein complex to be built. Further, NMR experiments, one-dimensional (1D)- and two-dimensional (2D)-total correlation spectroscopy (TOCSY), carried out on the model compounds glycerol-3-phosphate (G3P) and short chain phospholipids, supported the predicted mode of phospholipid binding. The toxins' high positive charge, which renders them extremely soluble (>300 mg/mL), and the phospholipid-binding specificity suggest the toxin-membrane interaction is mediated by binding to patches of negatively charged phospholipids [phosphatidic acid (PA) or phosphatidyl serine (PS)] and their subsequent withdrawal. The formation of proteolipid complexes causes solubilization of the membrane and its lysis. The model suggests that the oligomerization may play a role in toxin's activation process and provides insight into the structural principles of protein-membrane interactions.

On the nature of a glassy state of matter in a hydrated protein: Relation to protein function.

Diverse biochemical and biophysical experiments indicate that all proteins, regardless of size or origin, undergo a dynamic transition near 200 K. The cause of this shift in dynamic behavior, termed a "glass transition," and its relation to protein function are important open questions. One explanation postulated for the transition is solidification of correlated motions in proteins below the transition. We verified this conjecture by showing that crambin's radius of gyration (Rg) remains constant below approximately 180 K. We show that both atom position and dynamics of protein and solvent are physically coupled, leading to a novel cooperative state. This glassy state is identified by negative slopes of the Debye-Waller (B) factor vs. temperature. It is composed of multisubstate side chains and solvent. Based on generalization of Adam-Gibbs' notion of a cooperatively rearranging region and decrease of the total entropy with temperature, we calculate the slope of the Debye-Waller factor. The results are in accord with experiment.

Modeling and mutational analysis of a putative sodium-binding pocket on the dopamine D2 receptor.

A homology model of the dopamine D2 receptor was constructed based on the crystal structure of rhodopsin. A putative sodium-binding pocket identified in an earlier model (PDB ) was revised. It is now defined by Asn-419 backbone oxygen at the apex of a pyramid and Asp-80, Ser-121, Asn-419, and Ser-420 at each vertex of the planar base. Asn-423 stabilizes this pocket through hydrogen bonds to two of these residues. Highly conserved Asn-52 is positioned near the sodium pocket, where it hydrogen-bonds with Asp-80 and the backbone carbonyl of Ser-420. Mutation of three of these residues, Asn-52 in helix 1, Ser-121 in helix 3, and Ser-420 in helix 7, profoundly altered the properties of the receptor. Mutants in which Asn-52 was replaced with Ala or Leu or Ser-121 was replaced with Leu exhibited no detectable binding of radioligands, although receptor immunoreactivity in the membrane was similar to that in cells expressing the wild-type D2L receptor. A mutant in which Asn-52 was replaced with Gln, preserving hydrogen-bonding capability, was similar to D2L in affinity for ligands and ability to inhibit cAMP accumulation. Mutants in which either Ser-121 or Ser-420 was replaced with Ala or Asn had decreased affinity for agonists (Ser-121), but increased affinity for the antagonists haloperidol and clozapine. Interestingly, the affinity of these Ser-121 and Ser-420 mutants for substituted benzamide antagonists showed little or no dependence on sodium, consistent with our hypothesis that Ser-121 and Ser-420 contribute to the formation of a sodium-binding pocket.

CoMFA-based prediction of agonist affinities at recombinant wild type versus serine to alanine point mutated D2 dopamine receptors.

Agonist affinity changes dramatically as a result of serine to alanine mutations (S193A, S194A, and S197A) within the fifth transmembrane region of D2 dopamine receptors and other receptors for monoamine neurotransmitters. However, agonist 2D-structure does not predict which drugs will be sensitive to which point mutations. Modeling drug-receptor interactions at the 3D level offers considerably more promise in this regard. In particular, a comparison of the same test set of agonists across receptors differing minimally (point mutations) offers promise to enhance the understanding of the structural bases for drug-receptor interactions. We have previously shown that comparative molecular field analysis (CoMFA) can be applied to comparisons of affinity at recombinant D1 and D2 dopamine receptors for the same set of agonists, a differential QSAR. Here, we predicted agonist K(L) for the same set of agonists at wild type D2 vs S193A, S194A, and S197A receptors using CoMFA. Each model used bromocriptine as the template. ln(1/K(L)) values for the low-affinity agonist binding conformation at recombinant wild type and mutant D2 dopamine receptors stably expressed in C6 glioma cells were used as the target property for the CoMFA of the 16 aligned agonist structures. The resulting CoMFA models yielded cross-validated R(2) (q(2)) values ranging from 0.835 to 0.864 and simple R(2) values ranging from 0.999 to 1.000. Predictions of test compound affinities at WT and each mutant receptor were close to measured affinity values. This finding confirmed the predictive ability of the models and their differences from one another. The results strongly support the idea that CoMFA models of the same training set of compounds applied to WT vs mutant receptors can accurately predict differences in drug affinity at each. Furthermore, in a "proof of principle", two different templates were used to derive the CoMFA model for the WT and S193A mutant receptors. Pergolide was chosen as an alternate template because it showed a significant increase in affinity as a result of the S193A mutation. In this instance both the bromocriptine- and pergolide-based CoMFA models were similar to one another but different from those for the WT receptor using bromocriptine- or pergolide- as templates. The pergolide-based S193A model was more strikingly different from that of the WT receptor than was the bromocriptine-based S193A model. This suggests that a "dual-template" approach to differential CoMFA may have special value in elucidating key differences across related receptor types and in determining important elements of the drug-receptor interaction.

Accurate protein crystallography at ultra-high resolution: valence electron distribution in crambin.

The charge density distribution of a protein has been refined experimentally. Diffraction data for a crambin crystal were measured to ultra-high resolution (0.54 A) at low temperature by using short-wavelength synchrotron radiation. The crystal structure was refined with a model for charged, nonspherical, multipolar atoms to accurately describe the molecular electron density distribution. The refined parameters agree within 25% with our transferable electron density library derived from accurate single crystal diffraction analyses of several amino acids and small peptides. The resulting electron density maps of redistributed valence electrons (deformation maps) compare quantitatively well with a high-level quantum mechanical calculation performed on a monopeptide. This study provides validation for experimentally derived parameters and a window into charge density analysis of biological macromolecules.

Experimental observation of bonding electrons in proteins.

We demonstrate with two examples the success and potential of recent developments in x-ray protein crystallography at ultra high resolution. Our preliminary structural analyses using diffraction data collected for the two proteins crambin and savinase show meaningful deviations from the conventional independent spherical atom approximation. A noise-reduction averaging technique enables bonding details of electron distributions in proteins to be revealed experimentally for the first time. We move one step closer to imaging directly the fine details of the electronic structure on which the biological function of a protein is based.

Crystal structure of C-phycocyanin from Cyanidium caldarium provides a new perspective on phycobilisome assembly.

The crystal structure of the light-harvesting protein phycocyanin from the cyanobacterium Cyanidium caldarium with novel crystal packing has been solved at 1.65-A resolution. The structure has been refined to an R value of 18.3% with excellent backbone and side-chain stereochemical parameters. In crystals of phycocyanin used in this study, the hexamers are offset rather than aligned as in other phycocyanins that have been crystallized to date. Analysis of this crystal's unique packing leads to a proposal for phycobilisome assembly in vivo and for a more prominent role for chromophore beta-155. This new role assigned to chromophore beta-155 in phycocyanin sheds light on the numerical relationships among and function of external chromophores found in phycoerythrins and phycoerythrocyanins.

Crystal structure of Ser-22/Ile-25 form crambin confirms solvent, side chain substate correlations.

It is not agreed that correlated positions of disordered protein side chains (substate correlations) can be deduced from diffraction data. The pure Ser-22/Ile-25 (SI form) crambin crystal structure confirms correlations deduced for the natural, mixed sequence form of crambin crystals. Physical separation of the mixed form into pure SI form and Pro-22/Leu-25 (PL form) crambin and the PL form crystal structure determination (Yamano, A., and Teeter, M. M. (1994) J. Biol. Chem. 269, 13956-13965) support the proposed (Teeter, M. M., Roe, S. M., and Heo, N. H. (1993) J. Mol. Biol. 230, 292-311) correlation model. Electron density of mixed form crambin crystals shows four possible pairs of side chain conformations for heterogeneous residue 22 and nearby Tyr-29 (2(2) = 4, two conformations for each of two side chains). One combination can be eliminated because of short van der Waals' contacts. However, only two alternates have been postulated to exist in mixed form crambin: Pro-22/Tyr-29A and Ser-22/Tyr-29B. In crystals of the PL form, Pro-22 and Tyr-29A are found to be in direct van der Waals' contact (Yamano, A., and Teeter, M. M. (1994) J. Biol. Chem. 269, 13956-13965). Comparison of the SI form structure with the mixed form electron density confirms that the fourth combination of side chains does not occur and that side chain correlations are mediated by water networks.

Structure of human plasminogen kringle 4 at 1.68 a and 277 K. A possible structural role of disordered residues.

Despite considerable effort to elucidate the functional role of the kringle domains, relatively little is known about interactions with other protein domains. Most of the crystal structures describe the interactions at the kringle active site. This study suggests a novel way to interpret structural results such as disorder located away from an active site. The crystal structure of human plasminogen kringle 4 (PGK4) has been refined against 10-1.68 A resolution X-ray data (R(merge) = 3.7%) to the standard crystallographic R = 14.7% using the program X-PLOR. The crystals of PGK4 showed significant instability in cell dimensions (changes more than 1.5 A) even at 277 K. The refinement revealed structural details not observed before [Mulichak, Tulinsky & Ravichandran (1991). Biochemistry, 30, 10576-10588], such as clear density for additional side chains and more extensive disorder. Discrete disorder was detected for residues S73, S78, T80, S89, S91, S92, Ml12, S132, C138 and K142. Most of the disordered residues form two patches on the surface of the protein. This localized disorder suggests that these residues may play a role in quaternary interactions and possibly form an interface with the other domains of proteins that contain kringles, such as plasminogen. Although, an additional residue D65 was refined at the beginning of the sequence, still more residues near the peptide cleavage site must be disordered in the crystal.

Expression, purification and characterization of recombinant crambin.

Crambin, a small hydrophobic protein (4.7 kDa and 46 residues), has been successfully expressed in Escherichia coli from an artificial, synthetic gene. Several expression systems were investigated. Ultimately, crambin was successfully expressed as a fusion protein with the maltose binding protein, which was purified by affinity chromatography. Crambin expressed as a C-terminal domain was then cleaved from the fusion protein with Factor Xa protease and purified. Circular dichroism spectroscopy and amino acid analysis suggested that the purified material was identical to crambin isolated from seed. For positive identification the protein was crystallized from an ethanol-water solution, by a novel method involving the inclusion of phospholipids in the crystallization buffer, and then subjected to crystallographic analysis. Diffraction data were collected at the Brookhaven synchrotron (beamline-X12C) to a resolution of 1.32 A at 150 K. The structure, refined to an R value of 9.6%, confirmed that the cloned protein was crambin. The availability of cloned crambin will allow site-specific mutagenesis studies to be performed on the protein known to the highest resolution.

Refinement of purothionins reveals solute particles important for lattice formation and toxicity. Part 1: alpha1-purothionin revisited.

The three-dimensional structure of alpha(1)-purothionin (alpha(1)-PT), a wheat-germ protein and a basic lytic toxin, was previously solved by molecular-replacement methods using an energy-minimized predicted model and refined to an R-factor of 21.6% [Teeter, Ma, Rao & Whitlow (1990). Proteins Struct. Funct. Genet. 8, 118-1321. Some deficiencies of the model motivated us to revisit the structure and to continue the refinement. Here we report a significantly improved structure refined to an R-factor of 15.5% with excellent geometry. The refinement of this relatively low resolution structure ( approximately 2.8 A) is well suited to test the limitations of classical methods of refinement and to address the problem of overfitting, The final structure contains 434 atoms including 330 protein atoms, 70 waters, three acetates, two glycerols, one sec-butanol and one phosphate. The key solute molecules (acetate ion and phosphate ion) play a crucial role in the lattice formation. Phosphate and glycerol found in the structure may be important for biological activity of the toxins.

Refinement of purothionins reveals solute particles important for lattice formation and toxicity. Part 2: structure of beta-purothionin at 1.7 A resolution.

The crystal structure of beta-purothionin (beta-PT) has been determined at 1.7 A resolution. beta-PT and previously solved alpha(l)-PT belong to a family of membrane-active plant toxins homologous to crambin. (beta-PT crystallizes in the same space group as alpha(l)-PT (1422) but with the c axis 3 A longer than (alpha(l)-PT. The unit-cell dimensions of beta-PT crystals are a = b = 53.94 and c = 72.75 A. Two data sets were collected on a multiwire area detector, each with R(sym) around 6.0%, and were merged to get a single data set at 1.7 A, (R(merge) = 9.6%). The X-ray structure of alpha(l)-PT was used to build a starting model for beta-PT. The beta-PT model was refined using the program PROLSQ from 10 to 1.7 A resolution to an R-factor of 19.8% with very good geometry. The final structure contains 439 atoms including 337 protein atoms, 77 waters, two acetates, two glycerols and one phosphate. The high-resolution structure of the beta-PT agreed well with that of the lower resolution alpha(l)-PT structure only after the latter was extensively rerefined. Both refinements revealed phosphate and glycerol molecules which are important in lattice formation. The binding of phosphate and glycerol molecules to purothionins (PT) was confirmed by NMR and was implicated in the biological activity of toxins. Modeling of phospholipid binding to PT based on glycerol and phosphate-binding site could shed light on the lytic toxicity of this protein-toxin family. Although the structures of (alpha(l)-PT and beta-PT preserve the overall fold of crambin, they exhibit key differences that are directly relevant to the toxicity of thionins.

Full-matrix refinement of the protein crambin at 0.83 A and 130 K.

This paper describes the first successful full-matrix least-squares (FMLS) refinement of a protein structure. The example used is crambin which is a small hydrophobic protein (4.7 kDa, 46 residues). It proves the feasibility of refining such large molecules by this classic method, routinely applied to small molecules. The final structure with 381 non-H protein atoms (54 protein atoms in alternative positions), 367 H atoms, 162 water molecules (combined occupancy 93) and one disordered ethanol molecule converged to a standard unweighted crystallographic R-factor of R = 9.0% when refined against F with reflections stronger than F > 2sigma(F) and R = 9.5% when refined against F(2). The programs RFINE [Finger & Prince (1975). Natl Bur. Stand. (US) Tech. Note 854. A System of Fortran IV Computer Programs for Crystal Structure Computations] and SHELXL93 [Sheldrick (1993). SHELXL93. Program for Crystal Structure Refinement, Univ. of Göttingen, Germany] were used for FMLS refinement with the high-resolution low-temperature (0.83 A, 130 K) data set of a mixed-sequence form of crambin. A detailed analysis of the models obtained in FMLS and PROLSQ [restrained least squares or RLS; Teeter, Roe & Heo (1993). J. Mol. Biol. 230, 292-311] refinements with the same data set is presented. The differences between the models obtained by both FMLS and RLS refinements are systematic but negligible and advantages and shortcomings of both methods are discussed. The final structure has very good geometry, fully comparable to the geometry of other structures in this resolution range. Ideal values used in PROLSQ and those by Engh & Huber [Engh & Huber (1991). Acta Cryst. A47, 392-400] differ significantly from this refinement and we recommend a new standard. FMLS refinement constitutes a sensitive tool to detect and model disorder in highly refined protein structures. We describe the modeling of temperature factors by the TLS method [Schomaker & Trueblood (1968). Acta Cryst. B24, 63-76]. Rigid body-TLS refinements led to a better understanding of different modes of vibrations of the molecule. Refinements using F(2) or F protocols converged and reached slightly different minima. Despite theoretical support for F(2)-based refinement, we recommend refinement on structure factors.

Crambin: a direct solution for a 400-atom structure.

The crystal structure of crambin, a 46-residue protein containing the equivalent of approximately 400 fully occupied non-H-atom positions, was originally solved at 1.5 A by exploiting the anomalous scattering of its six S atoms at a single wavelength far removed from the absorption edge of sulfur. The crambin structure has now been resolved without the use of any anomalous-dispersion measurements. The technique employed was an ab initio 'shake-and-bake' method, consisting of a phase-refinement procedure based on the minimal function alternated with Fourier refinement. This method has successfully yielded solutions for a smaller molecule (28 atoms) using 1.2 A data, and a crambin solution was obtained at 1.1 A.

Homology modeling of the dopamine D2 receptor and its testing by docking of agonists and tricyclic antagonists.

We present the first model of dopamine D2 receptor transmembrane helices constructed directly from the bacteriorhodopsin (bR) coordinates derived from two-dimensional electron diffraction experiments. We have tested this model by its ability to accommodate rigid agonist and semirigid antagonist molecules which were docked into the putative binding pocket with stabilizing interactions. The model is consistent with structure-activity relationships of agonists and antagonists that interact with the receptor. It also illuminates data on a Na+ site for regulation of receptor function. The plausibility of the model is increased by its consistency with many mutagenesis studies on G protein-coupled receptors. Further, this model provides a basis to suggest testable molecular mechanisms for changes in the D2 conformational states for high- and low-affinity binding and signal transduction. Changes in the conformational state of the receptor are hypothesized to be due partly to movement of helix 7. In contrast to the model presented here, other published models were built using ideal helical structures or following the sense of the bacteriorhodopsin structure rather than the actual available coordinates. The presented model for the dopamine G protein-coupled receptor can be reconciled with the recent rhodopsin projection structure (Schertler, G. F. X.; Villa, C.; Henderson, R. Projection Structure of Rhodopsin.

Correlated disorder of the pure Pro22/Leu25 form of crambin at 150 K refined to 1.05-A resolution.

The high resolution crystal structure of crambin has been based on the crystals containing two sequence forms (the mixed form). Here, we report the crystal structure of the sequence isomer having Pro and Leu at residues 22 and 25 (the PL form). This elimination of the sequence heterogeneity resulted in a simpler structure which permits a more accurate modeling of protein disorder. In the observed disorder, the PL form structure and the mixed form structure have significant differences: 1) the disorder caused by the sequence heterogeneity (Pro2/Ser22, Leu/Ile25, Tyr29) is absent in the PL form; 2) Phe13 and Glu23 disordered in the mixed form have only one conformation in the PL form; and 3) Asn12 has multiple conformations in the PL form. During the study of disorder in the PL form structure, we found that conformational correlation can be inferred from a structure determined with Bragg's reflections by introducing fundamental stereochemical information, van der Waals contact (Gursky, O., Badger, J., Li, Y., and Caspar, D. L. D. (1992) Biophys. J. 63, 1210-1220), although an x-ray structure is an image averaged over a large number of copies and the period of data collection and does not carry direct evidence about correlations. The correlations among Thr2,Arg10, and Ile34 present the clearest example. The dimension of this correlation is comparable with the short-range (4-8 A) correlations in the atomic displacements concluded from the x-ray diffuse scattering experiments (Caspar, D. L. D., Clarage, J. B., Salunke, D. M., and Clarage, M. S. (1988) Nature 322, 659-662; Clarage, J. B., Clarage, M. S., Phillips, W. C., Sweet, R. M., and Caspar, D. L. D. (1992) Proteins 12, 145-157).

Improvement of turn structure prediction by molecular dynamics: a case study of alpha 1-purothionin.

Because of the problems in predicting a correct conformation for loop regions in homology-based prediction, disagreements are often found between the predicted models and the refined X-ray structures of the same protein in loop regions. Such a situation has been encountered for alpha 1-purothionin (alpha 1-PT). Hence, attempts have been made to improve the predicted model of alpha 1-PT by limited molecular dynamics using both AMBER and XPLOR. With molecular dynamics, the previously predicted incorrect turn region reverts to the correct conformation as seen in the X-ray refined structure. In contrast to the model which is not subjected to molecular dynamics, the improved model refines with the X-ray data of alpha 1-PT in fewer cycles, without any manual rebuilding and with comparable or better refinement statistics. Also, the improved model serves as a better starting model in the determination of the structure with the molecular replacement methods.

Atomic resolution (0.83 A) crystal structure of the hydrophobic protein crambin at 130 K.

To enhance the already high quality of diffraction data for crystals of the hydrophobic protein crambin, X-ray data were collected at 130 K by the method of H. Hope to 0.83 A resolution. Refinement with PROLSQ yields a model with an R value of 10.5%. The final model had three parameter anisotropic vibration factors for all atoms, which included 367 protein heavy atoms, 372 hydrogen atoms and 144 solvent atoms with one ethanol molecule. Dihedral angles and hydrogen-bonding distances generally agree with earlier studies of high-resolution protein structures, but some new patterns are noted. Solvent-related helix distortions are reminiscent of those described by others. Helix and beta-sheet regions show distinct patterns in their side-chain conformations. Despite crambin's hydrophobic nature, its accessible surface area in the crystal is surprisingly close to that of water-soluble proteins like myoglobin and carboxypeptidase A. More of crambin's hydrophobic surface is buried in the crystal, perhaps accounting for its high order of diffraction. A total of 24% of the 46 residues show discrete disorder at 130 K. This includes five side-chains at both 300 and 130 K, and six more side-chains and an ethanol molecule at 130 K. Disorder is associated with the sequence microheterogeneity at Pro/Ser22 and Leu/Ile25, with space filling or with solvent disorder. Correlated conformations extend over three to five residues. The patterns of disorder in this structure reveal important principles of protein structure and its dynamics. Finding disordered groups correlated over 5 to 8 A suggests that co-ordinated motion extends in groups rather than simply as uncorrelated movement around an atom center. Thermal diffuse scattering experiments on insulin and lysozyme are consistent with this interpretation. Nearly all of the protein-bound solvent has been located. Less than 1% of protein accessible surface area remains uncovered by solvent or crystal contacts. Preliminary analysis of the solvent network reveals two main networks in each of four solvent regions.