Sequence and structure patterns in proteins from an analysis of the shortest helices: implications for helix nucleation

The shortest helices (three-length 3(10) and four-length alpha), most abundant among helices of different lengths, have been analyzed from a database of protein structures. A characteristic feature of three-length 3(10)-helices is the shifted backbone conformation for the C-terminal residue (phi,psi angles: -95 degrees,0 degrees ), compared to the rest of the helix (-62 degrees,-24 degrees ). The deviation can be attributed to the release of electrostatic repulsion between the carbonyl oxygen atoms at the two C-terminal residues and further stabilization (due to a more linear geometry) of an intrahelical hydrogen bond. A consequence of this non-canonical C-terminal backbone conformation can be a potential origin of helix kinks when a 3(10)-helix is sequence-contiguous at the alpha-helix N-terminal. An analysis of hydrogen bonding, as well as hydrophobic interactions in the shortest helices shows that capping interactions, some of them not observed for longer helices, dominate at the N termini. Further, consideration of the distribution of amino acid residues indicates that the shortest helices resemble the N-terminal end of alpha-helices rather than the C terminus, implying that the folding of helices may be initiated at the N-terminal end, which does not get propagated in the case of the shortest helices. Finally, pairwise comparison of beta-turns and the shortest helices, based on correlation matrices of site-specific amino acid composition, and the relative abundance of these short secondary structural elements, leads to a helix nucleation scheme that considers the formation of an isolated beta-turn (and not an alpha-turn) as the helix nucleation step, with shortest 3(10)-helices as intermediates between the shortest alpha-helix and the beta-turn. Our results ascribe an important role played by shortest 3(10)-helices in proteins with important structural and folding implications.     

Differences in the amino acid distributions of 3(10)-helices and alpha-helices.

Local determinants of 3(10)-helix stabilization have been ascertained from the analysis of the crystal structure data base. We have clustered all 5-length substructures from 51 nonhomologous proteins into classes based on the conformational similarity of their backbone dihedral angles. Several clusters, derived from 3(10)-helices and multiple-turn conformations, had strong amino acid sequence patterns not evident among alpha-helices. Aspartate occurred over twice as frequently in the N-cap position of 3(10)-helices as in the N-cap position of alpha-helices. Unlike alpha-helices, 3(10)-helices had few C-termini ending in a left-handed alpha conformation; most 3(10) C-caps adopted an extended conformation. Differences in the distribution of hydrophobic residues among 3(10)- and alpha-helices were also apparent, producing amphipathic 3(10)-helices. Local interactions that stabilize 3(10)-helices can be inferred both from the strong amino acid preferences found for these short helices, as well as from the existence of substructures in which tertiary interactions replace consensus local interactions. Because the folding and unfolding of alpha-helices have been postulated to proceed through reverse-turn and 3(10)-helix intermediates, sequence differences between 3(10)- and alpha-helices can also lend insight into factors influencing alpha-helix initiation and propagation.     

The structure of the lactose permease derived from Raman spectroscopy and prediction methods.

The secondary structure of the lactose permease of Escherichia coli reconstituted in lipid membranes was determined by Raman spectroscopy. The alpha-helix content is approximately 70%, the beta-strand content below 10% and beta-turns contribute 15%. About 1/3 of the residues in alpha-helices and most other residues are exposed to water. Employing a method for structural prediction which accounts for amphipathic helices, 10 membrane-spanning helices are predicted which are either hydrophobic or amphipathic. They are expected to form an outer ring of helices in the membrane. The interior of the ring would be made of residues which are predominantly hydrophilic and, evoking the analogy to sugar-binding proteins, suited to provide the sugar binding site.     

Analysis and prediction of the different types of beta-turn in proteins

beta-Turns have been extracted from 59 non-identical proteins (resolution 2 A) using the standard criterion that the distance between C alpha (i) and C alpha (i + 3) is less than 7 A (1 A = 0.1 nm). The beta-turns have been classified, using phi, psi angles, into seven conventional turn types (I, I', II, II', IV, VIa, VIb) and a new class of beta-turn, designated type VIII, in which the central residues (i + 1, i + 2) adopt an alpha R beta conformation. Most beta-turn types are found in various topological environments, with the exception of I' and II' beta-turns, where 83% and 50%, respectively, are found in beta-hairpins. Sufficient data have been gathered to enable, for the first time, the separate statistical analysis of type I and II beta-turns. The two turn types have been shown to be strikingly different in their sequence preferences. Type I turns favour Asp, Asn, Ser and Cys at i; Asp, Ser, Thr and Pro at i + 1; Asp, Ser, Asn and Arg at i + 2; Gly, Trp and Met at i + 3, whilst type II turns prefer Pro at i + 1; Gly and Asn at i + 2; Gln and Arg at i + 3. These preferences have been explained by the specific side-chain interactions observed within the X-ray structures. The positional trends for type I and II beta-turns have been incorporated into the simple empirical predictive algorithm originally developed by P.N. Lewis et al. The program has improved the positional prediction of beta-turns, and has enhanced and extended the method by predicting the type of beta-turn. Since the observed preferences reflect local interactions these predictions are applicable not only to proteins, but also to peptides, many of which are thought to contain beta-turns.     

3(10)-helices in proteins are parahelices

The 3(10)-helix is characterized by having at least two consecutive hydrogen bonds between the main-chain carbonyl oxygen of residue i and the main-chain amide hydrogen of residue i + 3. The helical parameters--pitch, residues per turn, radius, and root mean square deviation (rmsd) from the best-fit helix--were determined by using the HELFIT program. All 3(10)-helices were classified as regular or irregular based on rmsd/(N - 1)1/2 where N is the helix length. For both there are systematic, position-specific shifts in the backbone dihedral angles. The average phi, psi shift systematically from approximately -58 degrees, approximately -32 degrees to approximately -90 degrees, approximately -4 degrees for helices 5, 6, and 7 residues long. The same general pattern is seen for helices, N = 8 and 9; however, in N = 9, the trend is repeated with residues 6, 7, and 8 approximately repeating the phi, psi of residues 2, 3, and 4. The residues per turn and radius of regular 3(10)-helices decrease with increasing length of helix, while the helix pitch and rise per residue increase. That is, regular 3(10)-helices become thinner and longer as N increases from 5 to 8. The fraction of regular 3(10)-helices decreases linearly with helix length. All longer helices, N > or = 9 are irregular. Energy minimizations show that regular helices become less stable with increasing helix length. These findings indicate that the definition of 3(10)-helices in terms of average, uniform dihedral angles is not appropriate and that it is inherently unstable for a polypeptide to form an extended, regular 3(10)-helix. The 3(10)-helices observed in proteins are better referred to parahelices.     

Prediction of beta-turns.

An automated computer prediction of the chain reversal regions of globular proteins is described herein using bend frequencies and beta-turn conformational parameters (Pt) determined from 408 beta-turns in 29 proteins calculated from x-ray atomic coordinates. The probability of bend occurrence at residue i is pt = fi X fi+1 X fi+2 X fi+3 with the average bend probability less than Pt greater than = 0.55 X 10(-4). Tetrapeptides with pt greater than 0.75 X 10(-4) ( approximately to 1.5 X less than pt greater than) as well as less than Pt greater than 1.00 and less than Pa greater than less than less than Pt greater than greater than less than P beta greater than are selected by the computer as probable bends. Adjacent probable bends (i.e., 11-14, 12-15, 13-16) are compared pairwise by the computer, and the tetrapeptide with the higher pt value is predicted as a beta-turn. The percentage of bend and nonbend residues predicted correctly for 29 proteins by this computer algorithm is %t+nt = 70%, whereas 78% of the beta-turns were localized correctly within +/- 2 residues. The average beta-turn content in the 29 proteins is 32%, with helical proteins having fewer bends (17%) than beta-sheet proteins (41%). Three proteins having iron-sulfur clusters were found with the highest percentages of beta-turns: Chromatium high potential iron protein (65%), ferredoxin (57%), and rubredoxin (65%). Finally, the bend frequencies at all 12 positions from 457 beta-turns in 29 proteins (Chou and Fasman, 1977) were used to test the effectiveness of predicting bends using 2, 4, 8, and 12 residues as well as different cut-off pt values. The computer analysis showed that 1.25 less than pt greater than to be the best cut-off yielding 70% accuracy in %t+nt for 4 residues and %t+nt = 73% for 12 residues in predicting the bend and nonbend regions of proteins.     

Hydrogen-bonded turns in proteins: the case for a recount

Beta-turns are sites at which proteins change their overall chain direction, and they occur with high frequency in globular proteins. The Protein Data Bank has many instances of conformations that resemble beta-turns but lack the characteristic N-H(i) --> O=C(i - 3) hydrogen bond of an authentic beta-turn. Here, we identify potential hydrogen-bonded beta-turns in the coil library, a Web-accessible database utility comprised of all residues not in repetitive secondary structure, neither alpha-helix nor beta-sheet (http://www.roselab.jhu.edu/coil). In particular, candidate turns were identified as four-residue segments satisfying highly relaxed geometric criteria but lacking a strictly defined hydrogen bond. Such candidates were then subjected to a minimization protocol to determine whether slight changes in torsion angles are sufficient to shift the conformation into reference-quality geometry without deviating significantly from the original structure. This approach of applying constrained minimization to known structures reveals a substantial population of previously unidentified, stringently defined, hydrogen-bonded beta-turns. In particular, 33% of coil library residues were classified as beta-turns prior to minimization. After minimization, 45% of such residues could be classified as beta-turns, with another 8% in 3(10) helixes (which closely resemble type III beta-turns). Of the remaining coil library residues, 37% have backbone dihedral angles in left-handed polyproline II structure.     

A neural network method for prediction of beta-turn types in proteins using evolutionary information

MOTIVATION: The prediction of beta-turns is an important element of protein secondary structure prediction. Recently, a highly accurate neural network based method Betatpred2 has been developed for predicting beta-turns in proteins using position-specific scoring matrices (PSSM) generated by PSI-BLAST and secondary structure information predicted by PSIPRED. However, the major limitation of Betatpred2 is that it predicts only beta-turn and non-beta-turn residues and does not provide any information of different beta-turn types. Thus, there is a need to predict beta-turn types using an approach based on multiple sequence alignment, which will be useful in overall tertiary structure prediction. RESULTS: In the present work, a method has been developed for the prediction of beta-turn types I, II, IV and VIII. For each turn type, two consecutive feed-forward back-propagation networks with a single hidden layer have been used where the first sequence-to-structure network has been trained on single sequences as well as on PSI-BLAST PSSM. The output from the first network along with PSIPRED predicted secondary structure has been used as input for the second-level structure-to-structure network. The networks have been trained and tested on a non-homologous dataset of 426 proteins chains by 7-fold cross-validation. It has been observed that the prediction performance for each turn type is improved significantly by using multiple sequence alignment. The performance has been further improved by using a second level structure-to-structure network and PSIPRED predicted secondary structure information. It has been observed that Type I and II beta-turns have better prediction performance than Type IV and VIII beta-turns. The final network yields an overall accuracy of 74.5, 93.5, 67.9 and 96.5% with MCC values of 0.29, 0.29, 0.23 and 0.02 for Type I, II, IV and VIII beta-turns, respectively, and is better than random prediction. AVAILABILITY: A web server for prediction of beta-turn types I, II, IV and VIII based on above approach is available at http://www.imtech.res.in/raghava/betaturns/ and http://bioinformatics.uams.edu/mirror/betaturns/ (mirror site).     

Peptide design using alpha,beta-dehydro amino acids: from beta-turns to helical hairpins

Incorporation of alpha,beta-dehydrophenylalanine (DeltaPhe) residue in peptides induces folded conformations: beta-turns in short peptides and 3(10)-helices in larger ones. A few exceptions-namely, alpha-helix or flat beta-bend ribbon structures-have also been reported in a few cases. The most favorable conformation of DeltaPhe residues are (phi,psi) approximately (-60 degrees, -30 degrees ), (-60 degrees, 150 degrees ), (80 degrees, 0 degrees ) or their enantiomers. DeltaPhe is an achiral and planar residue. These features have been exploited in designing DeltaPhe zippers and helix-turn-helix motifs. DeltaPhe can be incorporated in both right and left-handed helices. In fact, consecutive occurrence of three or more DeltaPhe amino acids induce left-handed screw sense in peptides containing L-amino acids. Weak interactions involving the DeltaPhe residue play an important role in molecular association. The C--H.O==C hydrogen bond between the DeltaPhe side-chain and backbone carboxyl moiety, pi-pi stacking interactions between DeltaPhe side chains belonging to enantiomeric helices have shown to stabilize folding. The unusual capability of a DeltaPhe ring to form the hub of multicentered interactions namely, a donor in aromatic C--H.pi and C--H.O==C and an acceptor in a CH(3).pi interaction suggests its exploitation in introducing long-range interactions in the folding of supersecondary structures.     

Combinations of turns in proteins.

We observed that beta- and gamma-turns in protein structure may be associated as peptides representing combinations of turns that span between nine and 26 amino acid residues along the polypeptide backbone chain and often correspond to loops in the protein structure. Around 475 peptides resulted from the analysis of a non-redundant data set corresponding to 248 protein crystal structures selected from the Protein Data Bank. Nearly 40% protein chains are associated with two or more peptides and the peptides with nine and 10 amino acid residues are more frequent. A maximum of four distinct peptides varying in number of amino acid residues were observed in at least 10 proteins along the same protein chain. Nearly 80% peptides comprise type IV beta-turns that are associated with irregular dihedral angle values suggesting this may be important for the conformational diversity associated with the loops in proteins. In general, predominant interactions that possibly stabilize these peptides involve main-chain and side-chain interactions with solvent, in addition to hydrogen bond, salt-bridge and non-bonded interactions. Majority of the peptides were observed in hydrolase, oxidoreductase, transferase, serine proteinase/inhibitor complex, electron transport/electron transfer and lyase proteins.     

The location and characterisation of the O-linked glycans of the human insulin receptor

O-linked glycosylation is a post-translational and post-folding event involving exposed S/T residues at beta-turns or in regions with extended conformation. O-linked sites are difficult to predict from sequence analyses compared to N-linked sites. Here we compare the results of chemical analyses of isolated glycopeptides with the prediction using the neural network prediction method NetOGlyc3.1, a procedure that has been reported to correctly predict 76% of O-glycosylated residues in proteins. Using the heavily glycosylated human insulin receptor as the test protein six sites of mucin-type O-glycosylation were found at residues T744, T749, S757, S758, T759, and T763 compared to the three sites (T759 and T763- correctly, T756- incorrectly) predicted by the neural network method. These six sites occur in a 20 residue segment that begins nine residues downstream from the start of the insulin receptor beta-chain. This region which also includes N-linked glycosylation sites at N742 and N755, is predicted to lack secondary structure and is followed by residues 765-770, the known linear epitope for the monoclonal antibody 18-44.     

Prediction of beta-turns in proteins using neural networks

The use of neural networks to improve empirical secondary structure prediction is explored with regard to the identification of the position and conformational class of beta-turns, a four-residue chain reversal. Recently an algorithm was developed for beta-turn predictions based on the empirical approach of Chou and Fasman using different parameters for three classes (I, II and non-specific) of beta-turns. In this paper, using the same data, an alternative approach to derive an empirical prediction method is used based on neural networks which is a general learning algorithm extensively used in artificial intelligence. Thus the results of the two approaches can be compared. The most severe test of prediction accuracy is the percentage of turn predictions that are correct and the neural network gives an overall improvement from 20.6% to 26.0%. The proportion of correctly predicted residues is 71%, compared to a chance level of about 58%. Thus neural networks provide a method of obtaining more accurate predictions from empirical data than a simpler method of deriving propensities.     

Prediction of turn types in protein structure by machine-learning classifiers

We present machine learning approaches for turn prediction from the amino acid sequence. Different turn classes and types were considered based on a novel turn classification scheme. We trained an unsupervised (self-organizing map) and two kernel-based classifiers, namely the support vector machine and a probabilistic neural network. Turn versus non-turn classification was carried out for turn families containing intramolecular hydrogen bonds and three to six residues. Support vector machine classifiers yielded a Matthews correlation coefficient (mcc) of approximately 0.6 and a prediction accuracy of 80%. Probabilistic neural networks were developed for beta-turn type prediction. The method was able to distinguish between five types of beta-turns yielding mcc > 0.5 and at least 80% overall accuracy. We conclude that the proposed new turn classification is distinct and well-defined, and machine learning classifiers are suited for sequence-based turn prediction. Their potential for sequence-based prediction of turn structures is discussed.     

Novel protein structural motifs containing two-turn and longer 3(10)-helices.

The 3(10)-helix constitutes a small but significant fraction of secondary structural elements in proteins. Protein data base surveys have shown these helices to be present as alpha-helical extensions, in loops and as connectors between beta-strands. The present work focuses on two-turn and longer 3(10)-helices where we establish that two-turn and longer 3(10) helices, unlike the more abundant single-turn 3(10)-helices, frequently occur independent of any other contiguous secondary structural elements. More importantly, a large fraction of these independent two-turn and longer 3(10)-helices, along with alpha-helices and beta-strands, are found to form novel super-secondary structural motifs in several proteins with possible implications for protein folding, local conformational relaxation and biological functions.     

Local sequence information in cellular retinoic acid-binding protein I: specific residue roles in beta-turns

We have recently shown that two of the beta-turns (III and IV) in the ten-stranded, beta-clam protein, cellular retinoic acid-binding protein I (CRABP I), are favored in short peptide fragments, arguing that they are encoded by local interactions (K. S. Rotondi and L. M. Gierasch, Biochemistry, 2003, Vol. 42, pp. 7976-7985). In this paper we examine these turns in greater detail to dissect the specific local interactions responsible for their observed native conformational biases. Conformations of peptides corresponding to the turn III and IV fragments were examined under conditions designed to selectively disrupt stabilizing interactions, using pH variation, chaotrope addition, or mutagenesis to probe specific side-chain influences. We find that steric constraints imposed by excluded volume effects between near neighbor residues (i,i+2), favorable polar (i,i+2) interactions, and steric permissiveness of glycines are the principal factors accounting for the observed native bias in these turns. Longer-range stabilizing interactions across the beta-turns do not appear to play a significant role in turn stability in these short peptides, in contrast to their importance in hairpins. Additionally, our data add to a growing number of examples of the 3:5 type I turn with a beta-bulge as a class of turns with high propensity to form locally defined structure. Current work is directed at the interplay between the local sequence information in the turns and more long-range influences in the mechanism of folding of this predominantly beta-sheet protein.     

The elusive π-helix

Central to protein architecture is the local arrangement or secondary structure of the polypeptide backbone. Thirty to forty percent of protein domains are α-helices with 3.6 residues per turn. π-Helices, in which the peptide chain is more loosely coiled (4.4 residues per turn), have also been proposed. However, such structures necessitate an energetically unfavorable ～1 Å central helical hole. We show that rather than being composed of idealized π-helices, helical regions formed from putative π-helices actually consist of a series of concatenated wide turns with unique elliptical configurations. These structures have a larger helical radius akin to that of a π-helix, but without the loss of favorable cross-core van der Waals interactions. This not only obviates the helical void, but also endows proteins with important functionalities, including metal ion coordination, enhanced flexibility and specific enzyme-substrate binding interactions.     

A deep dense inception network for protein beta-turn prediction

Beta-turn prediction is useful in protein function studies and experimental design. Although recent approaches using machine-learning techniques such as support vector machine (SVM), neural networks, and K nearest neighbor have achieved good results for beta-turn prediction, there is still significant room for improvement. As previous predictors utilized features in a sliding window of 4-20 residues to capture interactions among sequentially neighboring residues, such feature engineering may result in incomplete or biased features and neglect interactions among long-range residues. Deep neural networks provide a new opportunity to address these issues. Here, we proposed a deep dense inception network (DeepDIN) for beta-turn prediction, which takes advantage of the state-of-the-art deep neural network design of dense networks and inception networks. A test on a recent BT6376 benchmark data set shows that DeepDIN outperformed the previous best tool BetaTPred3 significantly in both the overall prediction accuracy and the nine-type beta-turn classification accuracy. A tool, called MUFold-BetaTurn, was developed, which is the first beta-turn prediction tool utilizing deep neural networks. The tool can be downloaded at http://dslsrv8.cs.missouri.edu/~cf797/MUFoldBetaTurn/download.html .     

Prediction of protein folding class from amino acid composition

An empirical relation between the amino acid composition and three-dimensional folding pattern of several classes of proteins has been determined. Computer simulated neural networks have been used to assign proteins to one of the following classes based on their amino acid composition and size: (1) 4α-helical bundles, (2) parallel (α/β)₈ barrels, (3) nucleotide binding fold, (4) immunoglobulin fold, or (5) none of these. Networks trained on the known crystal structures as well as sequences of closely related proteins are shown to correctly predict folding classes of proteins not represented in the training set with an average accuracy of 87%. Other folding motifs can easily be added to the prediction scheme once larger databases become available. Analysis of the neural network weights reveals that amino acids favoring prediction of a folding class are usually over represented in that class and amino acids with unfavorable weights are underrepresented in composition. The neural networks utilize combinations of these multiple small variations in amino acid composition in order to make a prediction. The favorably weighted amino acids in a given class also form the most intramolecular interactions with other residues in proteins of that class. A detailed examination of the contacts of these amino acids reveals some general patterns that may help stabilize each folding class. © 1993 Wiley-Liss, Inc.     

Addition of side-chain interactions to 3(10)-helix/coil and alpha-helix/3(10)-helix/coil theory.

An increasing number of experimental and theoretical studies have demonstrated the importance of the 3(10)-helix/ alpha-helix/coil equilibrium for the structure and folding of peptides and proteins. One way to perturb this equilibrium is to introduce side-chain interactions that stabilize or destabilize one helix. For example, an attractive i, i + 4 interaction, present only in the alpha-helix, will favor the alpha-helix over 3(10), while an i, i + 4 repulsion will favor the 3(10)-helix over alpha. To quantify the 3(10)/alpha/coil equilibrium, it is essential to use a helix/coil theory that considers the stability of every possible conformation of a peptide. We have previously developed models for the 3(10)-helix/coil and 3(10)-helix/alpha-helix/ coil equilibria. Here we extend this work by adding i, i + 3 and i, i + 4 side-chain interaction energies to the models. The theory is based on classifying residues into alpha-helical, 3(10)-helical, or nonhelical (coil) conformations. Statistical weights are assigned to residues in a helical conformation with an associated helical hydrogen bond, a helical conformation with no hydrogen bond, an N-cap position, a C-cap position, or the reference coil conformation plus i, i + 3 and i, i + 4 side-chain interactions. This work may provide a framework for quantitatively rationalizing experimental work on isolated 3(10)-helices and mixed 3(10)-/alpha-helices and for predicting the locations and stabilities of these structures in peptides and proteins. We conclude that strong i, i + 4 side-chain interactions favor alpha-helix formation, while the 3(10)-helix population is maximized when weaker i, i + 4 side-chain interactions are present.