A new computational approach for real protein folding prediction

An effective and fast minimization approach is proposed forthe prediction of protein folding, in which the ‘relativeentropy’ is used as a minimization function and the off-lattice model is used. In this approach, we only use the informationof distances between the consecutive C atoms along the peptidechain and a generalized form of the contact potential for 20types of amino acids. Tests of the algorithm are performed onthe real proteins. The root mean square deviations of the structuresof eight folded target proteins versus the native structuresare in a reasonable range. In principle, this method is an improvementon the energy minimization approach. Received June 25, 2002; revised June 6, 2003; accepted July 29, 2003.     

SMoS: a database of structural motifs of protein superfamilies

The Structural Motifs of Superfamilies (SMoS) database providesinformation about the structural motifs of aligned protein domainsuperfamilies. Such motifs among structurally aligned multiplemembers of protein superfamilies are recognized by the conservationof amino acid preference and solvent inaccessibility and areexamined for the conservation of other features like secondarystructural content, hydrogen bonding, non-polar interactionand residue packing. These motifs, along with their sequenceand spatial orientation, represent the conserved core structureof each superfamily and also provide the minimal requirementof sequence and structural information to retain each superfamilyfold. Received April 25, 2003; revised September 9, 2003; accepted September 24, 2003.     

Chimeric ribonuclease as a source of human adapter protein for targeted drug delivery

Assembled modular complexes for targeted drug delivery can bebased on strong non-covalent interactions between a cargo modulecontaining an adapter protein and a docking tag fused to a targetingprotein. We have recently constructed a completely humanizedadapter/docking tag system based on interactions between 15amino acid (Hu-tag) and 110 amino acid (HuS) fragments of humanribonuclease I (RNase I). Although recombinant HuS can be expressedand refolded into a functionally active form, the purificationprocedure is cumbersome and expensive, and more importantly,it yields a significant proportion of improperly folded proteins.Here we describe engineering, high-yield expression, and purificationof a chimeric bovine/human RNase (BH-RNase) comprising 1–29N-terminal amino acids of bovine ribonuclease A and 30–127amino acids of human RNase I. Unlike RNase I, the chimeric BH-RNasecan be cleaved by either subtilisin or proteinase K betweenA20 and S21, providing a functionally active HuS. The HuS obtainedfrom chimeric BH-RNase differs from wild-type HuS by an N24Tsubstitution; therefore, we have reverted this substitutionby mutating N24 to T24 in BH-RNase. This BH-RNase mutant canalso be cleaved by subtilisin or proteinase K yielding wild-typeHuS. The affinity of HuS obtained from BH-RNase to Hu-tag isapproximately five times higher than that for recombinant HuS,reflecting a higher percentage of properly folded proteins. Received June 9, 2003; revised August 4, 2003; accepted August 28, 2003.     

An empirical approach for structure-based prediction of carbohydrate-binding sites on proteins

A computer program system was developed to predict carbohydrate-bindingsites on three-dimensional (3D) protein structures. The programssearch for binding sites by referring to the empirical rulesderived from the known 3D structures of carbohydrate–proteincomplexes. A total of 80 non-redundant carbohydrate–proteincomplex structures were selected from the Protein Data Bankfor the empirical rule construction. The performance of theprediction system was tested on 50 known complex structuresto determine whether the system could detect the known bindingsites. The known monosaccharide-binding sites were detectedamong the best three predictions in 59% of the cases, whichcovered 69% of the polysaccharide-binding sites in the targetproteins, when the performance was evaluated by the overlapbetween residue patches of predicted and known binding sites. Received April 24, 2003; revised June 2, 2003; accepted June 10, 2003.     

Role of the tyrosine corner motif in the stability of neocarzinostatin

Although the immunoglobulin-like ß-sandwich fold hasno specifically conserved function, some common structural featureshave been observed, in particular a structural motif, the tyrosinecorner. Such a motif was described in neocarzinostatin (NCS),a bacterial protein the structure of which is very similar tothat of the immunoglobulin domain. Compared with the other ß-sheetproteins, the NCS ‘tyrosine corner’ presents non-standardstructural features. To investigate the role of this motif inthe NCS structure and stability, we studied the properties ofa mutant where the H bond interaction had been eliminated byreplacing the tyrosine with a phenylalanine. This mutation costs4.0 kcal/mol showing that the NCS ‘tyrosine corner’is involved in protein stability as in the other Greek key proteins.This destabilization is accompanied by remote structural effects,including modification of the binding properties, suggestingan increase in the internal flexibility of the protein. Witha view to using this protein for drug targeting, these resultsalong with those obtained previously allow us to define clearlythe limitations of the modifications that can be performed onthis scaffold. Received December 3, 2002; revised June 6, 2003; accepted September 3, 2003.     

Structural diversity in the small heat shock protein superfamily: control of aggregation by the N-terminal region

The small heat shock protein superfamily, extending over allkingdoms, is characterized by a common core domain with variableN- and C-terminal extensions. The relatively hydrophobic N-terminusplays a critical role in promoting and controlling high-orderaggregation, accounting for the high degree of structural variabilitywithin the superfamily. The effects of N-terminal volume onaggregation were studied using chimeric and truncated proteins.Proteins lacking the N-terminal region did not aggregate abovethe tetramers, whereas larger N-termini resulted in large aggregates,consistent with the N-termini packing inside the aggregates.Variation in an extended internal loop differentiates typicalprokaryotic and plant superfamily members from their animalcounterparts; this implies different geometry in the dimericbuilding block of high-order aggregates. Received March 15, 2003; revised May 20, 2003; accepted September 4, 2003.     

On the rotational operators in protein structure simulations

The reduction of the computational complexity of the algorithmsdealing with protein structure analysis and conformation predictionsis of prime importance. One common element in most of thesealgorithms is the process of transforming geometrical informationbetween dihedral angles and Cartesian coordinates of the atomsin the protein using rotational operators. In the literature,the operators used in protein structures are rotation matrices,quaternions in vector and matrix forms and the Rodrigues–Gibbsformula. In the protein structure-related literature, the mostwidely promoted rotational operator is the quaternions operator.In this work, we studied the computational efficiency of themathematical operations of the above rotational operators appliedto protein structures. A similar study applied to protein structureshas not been reported previously. We concluded that the computationalefficiency of these rotational operators applied to proteinchains is different from those reported for other applications(such as mechanical machinery) and the conclusions are not analogous.Rotation matrices are the most efficient mathematical operatorsin the protein chains. We examined our findings in two proteinmolecules: Ab1 tyrosine kinase and heparin-binding growth factor2. We found that the rotation matrix operator has between 2and 187% fewer mathematical operations than the other rotationaloperators. Received June 6, 2003; revised July 17, 2003; accepted August 20, 2003.     

Structure-based substitutions for increased solubility of a designed protein

Manipulation of protein solubility is important for many aspectsof protein design and engineering. Previously, we designed aseries of consensus ankyrin repeat proteins containing one,two, three and four identical repeats (1ANK, 2ANK, 3ANK and4ANK). These proteins, particularly 4ANK, are intended for useas a universal scaffold on which specific binding sites canbe constructed. Despite being well folded and extremely stable,4ANK is soluble only under acidic conditions. Designing interactionswith naturally occurring proteins requires the designed proteinto be soluble at physiological pH. Substitution of six leucineswith arginine on exposed hydrophobic patches on the surfaceof 4ANK resulted in increased solubility over a large pH range.Study of the pH dependence of stability demonstrated that 4ANKis one of the most stable ankyrin repeat proteins known. Inaddition, analogous leucine to arginine substitutions on thesurface of 2ANK allowed the partially folded protein to assumea fully folded conformation. Our studies indicate that replacementof surface-exposed hydrophobic residues with positively chargedresidues can significantly improve protein solubility at physiologicalpH. Received June 23, 2003; revised August 22, 2003; accepted August 28, 2003.     

Differential effects of zinc on functionally distinct human growth hormone mutations

Human growth hormone (hGH) binds and activates lactogenic receptorsby a sequential receptor dimerization mechanism. The affinityfor the first lactogenic receptor is increased due to one zincmolecule linking hGH residues H18 and E174, located in helices1 and 4, respectively, with two adjacent residues in the lactogenicreceptor (D187 and H188). Two functionally unique groups ofmutant hGHs have been identified. Addition of 25 µM zincto lactogenic bioassays differentially affects mutant activitiesbased on which group they belong to. One mutation (G120R) islocated within the binding surface of hGH that interacts withthe second lactogenic receptor. In the presence of endogenouszinc, G120R reduces the maximal activity of hGH without alteringeither the agonist or antagonist phases of the bell-shaped dose–responsecurve. Addition of zinc to this assay further reduces the activityof this protein. In contrast, mutations within a hydrophobicmotif in hGH that functionally couples the two lactogenic receptorbinding sites decrease the sensitivity (right-shift) of theagonist phase of the dose–response curve without similarlyaffecting the antagonist phase. The addition of zinc to theselactogenic assays increases the sensitivity (left-shifts) ofthe dose–response curves, largely negating the effectof these mutations. The effects of zinc differentiate betweenmutations within these two distinct functional motifs by limitingthe pool of potential conformations that are available for bindingwithin either of the receptor binding sites of this ligand. Received April 10, 2003; revised June 7, 2003; accepted June 8, 2003.     

Precise and efficient cleavage of recombinant fusion proteins using mammalian aspartic proteases

Expression of recombinant proteins as translational fusionsis commonly employed to enhance stability, increase solubilityand facilitate purification of the desired protein. In general,such fusion proteins must be cleaved to release the mature proteinin its native form. The usefulness of the procedure dependson the efficiency and precision of cleavage and its cost perunit activity. We report here the development of a general procedurefor precise and highly efficient cleavage of recombinant fusionproteins using the protease chymosin. DNA encoding a modifiedpro-peptide from bovine chymosin was fused upstream of hirudin,carp growth hormone, thioredoxin and cystatin coding sequencesand expressed in a bacterial Escherichia coli host. Each ofthe resulting fusion proteins was efficiently cleaved at thejunction between the pro-peptide and the desired protein bythe addition of chymosin, as determined by activity, N-terminalsequencing and mass spectrometry of the recovered protein. Thesystem was tested further by cleavage of two fusion proteins,cystatin and thioredoxin, sequestered on oilbody particles obtainedfrom transgenic Arabidopsis seeds. Even when the fusion proteinwas sequestered and immobilized on oilbodies, precise and efficientcleavage was obtained. The precision, efficiency and low costof this procedure suggest that it could be used in larger scalemanufacturing of recombinant proteins which benefit from expressionas fusions in their host organism. Received June 5, 2003; revised August 1, 2003; accepted August 20, 2003.     

Vicinal disulfide turns

The formation of a disulfide bond between adjacent cysteineresidues is accompanied by the formation of a tight turn ofthe protein backbone. In nearly 90% of the structures analyzeda type VIII turn was found. The peptide bond between the twocysteines is in a distorted trans conformation, the omega torsionangle ranges from 159 to –133°, with an average valueof 171°. The constrained nature of the vicinal disulfideturn and the pronounced difference observed between the oxidizedand reduced states, suggests that vicinal disulfides may beemployed as a ‘redox-activated’ conformational switch. Received December 16, 2002; revised June 30, 2003; accepted July 30, 2003.     

Prediction of protein secondary structure with a reliability score estimated by local sequence clustering

Most algorithms for protein secondary structure prediction arebased on machine learning techniques, e.g. neural networks.Good architectures and learning methods have improved the performancecontinuously. The introduction of profile methods, e.g. PSI-BLAST,has been a major breakthrough in increasing the prediction accuracyto close to 80%. In this paper, a brute-force algorithm is proposedand the reliability of each prediction is estimated by a z-scorebased on local sequence clustering. This algorithm is intendedto perform well for those secondary structures in a proteinwhose formation is mainly dominated by the neighboring sequencesand short-range interactions. A reliability z-score has beendefined to estimate the goodness of a putative cluster foundfor a query sequence in a database. The database for predictionwas constructed by experimentally determined, non-redundantprotein structures with <25% sequence homology, a list maintainedby PDBSELECT. Our test results have shown that this new algorithm,belonging to what is known as nearest neighbor methods, performedvery well within the expectation of previous methods and thatthe reliability z-score as defined was correlated with the reliabilityof prediction. This led to the possibility of making very accuratepredictions for a few selected residues in a protein with anaccuracy measure of Q₃ > 80%. The further development ofthis algorithm, and a nucleation mechanism for protein foldingare suggested. Received March 27, 2003; revised June 30, 2003; accepted August 22, 2003.     

A long insertion reverts the functional effect of a substitution in acetylcholinesterase

Proteins are thought to undertake single substitutions, deletionsand insertions to explore the fitness landscape. Nevertheless,the ways in which these different kind of mutations act togetherto alter a protein phenotype remain poorly described. We introducedincrementally the single substitution W290A and a 26 amino acidlong insertion at the 297 location in the Nippostrongylus brasiliensisacetylcholinesterase B sequence and analysed in vitro the inducedchanges in the hydrolysis rate of three hemi-substrates: pirimicarb,paraoxon methyl and omethoate. The substitution decreased thehydrolysis rate of the three hemi-substrates. The insertiondid not influence this kinetic alteration induced by the substitutionfor the former hemi-substrate, but reverted it for the two others.These results show that two different kinds of mutations caninteract together to influence the direction of a protein’sadaptative walk on the fitness landscape. Received January 28, 2003; revised April 25, 2003; accepted June 6, 2003.     

Immobilized oxidoreductase as an additive for refolding inclusion bodies: application to antibody fragments

Three foldases—the apical domain of GroEL (mini-chaperone)and two oxidoreductases (DsbA and DsbC) from Escherichia coli—werestudied in refolding a protein with immunoglobulin fold (immunoglobulin-foldedprotein) that had been produced as inclusion bodies in E.coli.The foldases promoted the refolding of single-chain antibodyfragments from denaturant-solubilized and reduced inclusionbodies in vitro, and also effectively functioned as alternativesfor labilizing agent and oxidizing reagent in the stepwise dialysissystem. Immobilization of the oxidoreductases enhanced refoldingand recovery of functional single-chain antibody in the dialysissystem, suggesting that immobilized oxidoreductases can be usedas an effective additive for refolding immunoglobulin-foldedproteins in vitro. Received April 7, 2003; revised June 5, 2003; accepted June 6, 2003.     

PROSPECT II: protein structure prediction program for genome-scale applications

A new method for fold recognition is developed and added tothe general protein structure prediction package PROSPECT (http://compbio.ornl.gov/PROSPECT/).The new method (PROSPECT II) has four key features. (i) We havedeveloped an efficient way to utilize the evolutionary informationfor evaluating the threading potentials including singletonand pairwise energies. (ii) We have developed a two-stage threadingstrategy: (a) threading using dynamic programming without consideringthe pairwise energy and (b) fold recognition considering allthe energy terms, including the pairwise energy calculated fromthe dynamic programming threading alignments. (iii) We havedeveloped a combined z-score scheme for fold recognition, whichtakes into consideration the z-scores of each energy term. (iv)Based on the z-scores, we have developed a confidence index,which measures the reliability of a prediction and a possiblestructure–function relationship based on a statisticalanalysis of a large data set consisting of threadings of 600query proteins against the entire FSSP templates. Tests on severalbenchmark sets indicate that the evolutionary information andother new features of PROSPECT II greatly improve the alignmentaccuracy. We also demonstrate that the performance of PROSPECTII on fold recognition is significantly better than any othermethod available at all levels of similarity. Improvement inthe sensitivity of the fold recognition, especially at the superfamilyand fold levels, makes PROSPECT II a reliable and fully automatedprotein structure and function prediction program for genome-scaleapplications. Received March 20, 2003; revised June 28, 2003; accepted July 8, 2003.     

Tandem repeats of Sushi3 peptide with enhanced LPS-binding and -neutralizing activities

Endotoxin, also known as lipopolysaccharide (LPS), is the majormediator of septic shock due to Gram-negative bacterial infection.Chemically synthesized S3 peptide, derived from Sushi3 domainof Factor C, which is the endotoxin-sensitive serine proteaseof the limulus coagulation cascade, was previously shown tobind and neutralize LPS activity. For large-scale productionof this peptide and to mimick other pathogen-recognizing molecules,tandem multimers of the S3 gene were constructed and expressedin Escherichia coli. The recombinant tetramer of S3 (rS3-4mer)was purified by anion exchange and digested into monomers (rS3-1mer).Both the rS3-4mer and rS3-1mer were functionally analyzed fortheir ability to bind LPS by an ELISA-based method and surfaceplasmon resonance. The LAL inhibition and TNF-release test showedthat rS3-1 mer can neutralize the LPS activity as effectivelyas the synthetic S3 peptide, while rS3-4mer displays an enhancedinhibitory effect on LPS-induced activities. Both recombinantpeptides exhibited low cytotoxicity and no haemolytic activityon human cells. This evidence suggests that the recombinantsushi peptides have potential use for the detection, removalof endotoxin and/or anti-endotoxin strategies. Received September 5, 2002; revised June 18, 2003; accepted June 20, 2003.     

Molecular docking of substrates and inhibitors in the catalytic site of CYP6B1, an insect cytochrome P450 monooxygenase

Furanocoumarins represent plant toxins that are used in thetreatment of a variety of skin diseases and are metabolizedby cytochrome P450 monooxygenases (P450s) existing in insectssuch as Papilio polyxenes (the black swallowtail). To elucidatethe active site in the CYP6B1 protein that is the principalP450 existing in this species, we have constructed a homologymodel of it based on sequence and structure alignments withthe bacterial CYP102 protein whose crystal structure has beendefined and with the insect CYP6B4 protein that also metabolizesfuranocoumarins. In the derived CYP6B1 model, Phe116 and His117in SRS1, Phe371 in SRS5 and Phe484 in SRS6 contribute to theformation of a resonant network that stabilizes the P450’scatalytic site and allows for interactions with its furanocoumarinsubstrates. The first two of these residues are absolutely conservedin all members of the insect CYP6B subfamily and the last twoare variable in different members of the CYP6B subfamily. Acombination of theoretical and experimental docking analysesof two substrates (xanthotoxin and bergapten) and two inhibitors(coumarin and pilocarpine) of this P450 provide significantinformation on the positioning of furanocoumarins within thiscatalytic pocket. Molecular replacement models based on theresults of variations at two of these critical amino acids providesupport for our furanocoumarin-docked model and begin to rationalizethe altered substrate reactivities observed in experimentalanalyses. Received December 4, 2002; revised June 8, 2003; accepted June 23, 2003.     

Reduction of protein sequence complexity by residue grouping

It is well known that there are some similarities among variousnaturally occurring amino acids. Thus, the complexity in proteinsystems could be reduced by sorting these amino acids with similaritiesinto groups and then protein sequences can be simplified byreduced alphabets. This paper discusses how to group similaramino acids and whether there is a minimal amino acid alphabetby which proteins can be folded. Various reduced alphabets areobtained by reserving the maximal information for the simplifiedprotein sequence compared with the parent sequence using globalsequence alignment. With these reduced alphabets and simplifiedsimilarity matrices, we achieve recognition of the protein foldbased on the similarity score of the sequence alignment. Thecoverage in dataset SCOP40 for various levels of reduction onthe amino acid types is obtained, which is the number of homologouspairs detected by program BLAST to the number marked by SCOP40.For the reduced alphabets containing 10 types of amino acids,the ability to detect distantly related folds remains almostat the same level as that by the alphabet of 20 types of aminoacids, which implies that 10 types of amino acids may be thedegree of freedom for characterizing the complexity in proteins. Received November 20, 2002; revised March 10, 2003; accepted April 4, 2003.     

Optimizing the search algorithm for protein engineering by directed evolution

An in silico protein model based on the Kauffman NK-landscape,where N is the number of variable positions in a protein andK is the degree of coupling between variable positions, wasused to compare alternative search strategies for directed evolution.A simple genetic algorithm (GA) was used to model the performanceof a standard DNA shuffling protocol. The search effectivenessof the GA was compared to that of a statistical approach calledthe protein sequence activity relationship (ProSAR) algorithm,which consists of two steps: model building and library design.A number of parameters were investigated and found to be importantfor the comparison, including the value of K, the screeningsize, the system noise and the number of replicates. The statisticalmodel was found to accurately predict the measured activitiesfor small values of the coupling between amino acids, K 1.The ProSAR strategy was found to perform well for small to moderatevalues of coupling, 0 K 3, and was found to be robust to noise.Some implications for protein engineering are discussed. Received January 2, 2003; revised May 13, 2003; accepted June 19, 2003.     

Have we seen all structures corresponding to short protein fragments in the Protein Data Bank? An update

Assembling short fragments from known structures has been awidely used approach to construct novel protein structures.To what extent there exist structurally similar fragments inthe database of known structures for short fragments of a novelprotein is a question that is fundamental to this approach.This work addresses that question for seven-, nine- and 15-residuefragments. For each fragment size, two databases, a query databaseand a template database of fragments from high-quality proteinstructures in SCOP20 and SCOP90, respectively, were constructed.For each fragment in the query database, the template databasewas scanned to find the lowest r.m.s.d. fragment among non-homologousstructures. For seven-residue fragments, there is a 99% probabilitythat there exists such a fragment within 0.7 Å r.m.s.d.for each loop fragment. For nine-residue fragments there isa 96% probability of a fragment within 1 Å r.m.s.d., whilefor 15-residue fragments there is a 91% probability of a fragmentwithin 2 Å r.m.s.d.. These results, which update previousstudies, show that there exists sufficient coverage to modeleven a novel fold using fragments from the Protein Data Bank,as the current database of known structures has increased enormouslyin the last few years. We have also explored the use of a gridsearch method for loop homology modeling and make some observationsabout the use of a grid search compared with a database searchfor the loop modeling problem. Received October 23, 2002; revised March 3, 2003; accepted March 30, 2003.