Dynamic α-helices: Conformations that do not conform

Dynamic a-Helices: Conformations that Do Not Conform

Kuljeet Singh Sandhu and Debasis Dash*

GN Ramachandran Knowledge Center for Genome Informatics, Institute of Genomics and Integrative Biology (IGIB),CSIR, Delhi 110007, India

ABSTRACT Structural transitions are impor-tant for the stability and function of proteins, butthese phenomena are poorly understood. An exten-sive analysis of Protein Data Bank entries reveals103 regions in proteins with a tendency to trans-form from helical to nonhelical conformation andvice versa. We find that these dynamic helices,unlike other helices, are depleted in hydrophobicresidues. Furthermore, the dynamic helices havehigher surface accessibility and conformationalmobility (P-value ¼ 3.35e-07) than the rigid helices.Contact analyses show that these transitions resultfrom protein-ligand, protein-nucleic acid, and crys-tal-contacts. The immediate structural environ-ment differs quantitatively (P-value ¼ 0.003) as wellas qualitatively in the two alternate conformations.Often, dynamic helix experiences more contacts inits helical conformation than in the nonhelicalcounterpart (P-value ¼ 0.001). There is differentialpreference for the type of short contacts observedin two conformational states. We also demonstratethat the regions in protein that can undergo suchlarge conformational transitions can be predictedwith a reasonable accuracy using logistic regres-sion model of supervised learning. Our findingshave implications in understanding the molecularbasis of structural transitions that are coupledwith binding and are important for the functionand stability of the protein. Based on our observa-tions, we propose that several functionally relevantregions on the protein surface can switch overtheir conformation from coil to helix and vice-versa, to regulate the recognition and binding oftheir partner and hence these may work as ‘‘molec-ular switches’’ in the proteins to regulate certainbiological process. Our results supports the ideathat protein structure–function paradigm shouldtransform from static to a highly dynamic one.Proteins 2007;68:109–122. VVC 2007 Wiley-Liss, Inc.

Key words: conformational transition; protein disor-der; a-helix; interaction; crystal contact

INTRODUCTION

The protein structure–function paradigm, which statesthat a protein must fold into rigid three-dimensionalstructure to perform its respective task, is based on theassumption that molecular recognition in the cell requiresprecise geometrical orientation that is achieved by rigid

folds in the protein. The realization that many proteinsremain completely or partially unfolded in their nativefunctional state1–3 has transformed the ‘‘protein struc-ture–function’’ paradigm to ‘‘protein trinity’’ paradigm,which states ‘‘The native proteins can exist in any of thethree thermodynamic states—ordered, molten globule,and random coil (i.e., Ordered$Molten globule $ Ran-dom coil).’’4 According to this theory, not necessarily theordered state but any one of these can be the native func-tional state of the protein. Reversible transitions amongthese states may regulate the cellular processes theseproteins involved in. Over 30% of human proteome is pre-dicted to be unfolded in the native state5 and this subsetincludes various functional classes encompassing func-tions such as cell signaling, gene-regulation, proteinphosphorylation, storage of small molecules, oligomeriza-tion or self-assembly of large multiprotein complexes.3,5

Natively unstructured proteins are discovered morerecently, since the biased biochemical and biophysicalmethods selectively work for globular protein.3 Crystaldiffraction patterns generally do not resolve unstructuredregions, which limit their identification and characteriza-tion. Nevertheless, spectroscopic-based methods likeNMR and CD can characterize these regions. From thedisorder-promoting characteristics of amino acid sequen-ces, disordered regions in the protein can be predicted bycomputational methods.1,2,6–10 Several sequence patternsthat ascribe conformational flexibility in protein struc-tures have been identified and analyzed.11,12

Conformational variability in protein structures hasfunctional significance. For example, these can be in-volved in specific hinge movements to facilitate certaininteractions. Heat shock proteins contain flexible loopsthat play crucial role in recruiting target protein.13

Structural disorder can also ascribe multiple cellular

Abbreviations: a.a., amino acid; DSSP, definition of secondarystructure prediction; HTH, helix turn helix motif; NMR, nuclearmagnetic resonance; PDB, protein data bank.

The Supplementary Material referred to in this article can be foundat http://www.interscience.wiley.com/jpages/0887-3585/suppmat/

Grant sponsor: CSIR; Grant number: CMM0017.

*Correspondence to: Debasis Dash, GN Ramachandran Knowl-edge Center for Genome Informatics, Institute of Genomics andIntegrative Biology (IGIB), CSIR, Mall Road, Delhi 110007, India..E-mail: [email protected]

Received 25 July 2006; Revised 9 October 2006; Accepted 6November 2006

Published online 3 April 2007 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.21328

VVC 2007 WILEY-LISS, INC.

PROTEINS: Structure, Function, and Bioinformatics 68:109–122 (2007)

roles to a protein.14,15 The unstructured regions providean extended number of contact areas for recognition andinteraction with multiple interacting partners. Theseregions can form metastable conformations of high speci-ficity and low affinity and their transition to rigid con-formations due to binding with interacting partnersdecreases the conformational entropy to form a compact,but reversible macromolecular assembly. Specific struc-tural changes also act as conformational switches forbinding or release of ligand.16 Furthermore, the confor-mational transitions may regulate protein turn-over.Conformationally flexible regions of the protein are moresusceptible to protein degradation.15,17 The structuraltransition of flexible regions to more rigid conformationstabilizes the protein against proteolysis.18 Thus, the re-versible transition between different conformationalstates can work as a regulatory switch to control certaincellular process.Although a few attempts have been made using molec-

ular dynamics simulations,19–21 but such studies arelimited due to paucity of computational power and time.Protein Data Bank provides a large number of experi-mentally determined protein structures with a high levelof redundancy, which can be exploited to study the struc-tural polymorphisms. Several groups22–25 have used theredundancy in Protein Data Bank to explore the struc-ture–function relationships. Several interesting struc-tural transitions have been reviewed by Dyson andWright3 and Uversky.4 The analyses of structural transi-tions in proteins require direct experimental evidenceand studies for individual cases, mostly based on CDspectra, exist in literature. This includes increase in ahelical content in Max protein favored by self dimeriz-ation and DNA-binding,26 coil-to-helix transition upont-RNA binding to protamines27 and ligand-controlledrigidity of a-fetoprotein (AFP) and horse apocytochrome-c.28,29 The region 82–94 in apomyoglobin undergoes coil-to-helix transition during the binding of heme moiety.Such a transition from disordered to ordered state pro-tects the holomyoglobin from proteolysis.18 Coil-to-helixtransitions are also observed during phosphorylation ofproteins.30,31 Similarly, pKID domain of the transcriptionfactor CREB protein is intrinsically unstructured butbecomes helical when binds to CREB binding protein(CBP).32 Intrinsically unstructured protein (IUP), p27-KID, exhibits exceptionally high thermodynamic stabilityafter binding to Cdk2/cyclin-A complex.33 Based on thisobservation, the authors proposed the concept of thermo-dynamic tethering, i.e. the macromolecular assembliesformed by IUP binding to existing protein complexesleading to increased thermodynamic stability.33

In an interesting report, Gerstein et al.34 explored sev-eral domain movements in the proteins with a similarapproach. They identified two kinds of domain motionsnamely the ‘‘hinge’’ motion and ‘‘shear’’ motion. Thesemotions were proposed to be of low energy and are im-portant for ligand binding. They have elegantly demon-strated how the dynamics of protein structures can bestudied at large scale without applying computationally

or experimentally intensive methods. Similarly, thestructural changes upon protein–protein association orenzyme–substrate interaction were analyzed by fewgroups.35,36 However, there is a lack of comprehensivestudies aimed to analyze ‘‘structured $ unstructured’’kind of transitions in Protein Data Bank. Here, we iden-tified large conformational transitions in proteins andthen analyzed their inherent features as well as theirimmediate microenvironments. Among all secondarystructure transitions, helix-to-coil transitions are themost common. a-helix-forming small elements in largeunstructured regions were proposed to mediate thecoupled folding and binding events.37 On a similartheme, preformed structural elements, mostly helices, inunfolded proteins are proposed to be the anchoringpoints for interaction and further facilitating theircoupled folding.38 Therefore, we took initiative to com-prehensively analyze ‘‘helix$coil’’ transitions in PDBand showed that these regions have differential prefer-ence for amino acids, surface accessibility, and conforma-tional flexibility. Interestingly, these regions directlyinteract with ligands, nucleic acids or protein moietiesand their immediate structural environments differ intheir alternate conformations.

MATERIALS AND METHODSDataset

The Protein Data Bank (http://www.rcsb.org/pdb/)entries are clustered using PDB-REPRDB server (http://mbs.cbrc.jp/pdbreprdb-cgi/reprdb_menu.pl) as on October12, 2005 with the following criteria; resolution � 3 A, R-factor < 0.3, length > 40, and sequence identity < 90%.This is the parent dataset of 4685 clusters.

Secondary structure was assigned using Definition ofSecondary Structure Prediction program (DSSP) (http://bioweb.pasteur.fr/seqanal/interfaces/dssp.html). The clus-ters with a sequence stretch of length � 8 found in bothhelical and nonhelical conformations were analyzed. Thedata set composed of 103 clusters wherein each clusterincluded sequences that had 100% identity and elimi-nated all other close homologs. The representative pairsfor helical and corresponding nonhelical conformationswere then extracted out based on sequence length(where the structure coordinates are available) and reso-lution. Protein sequences with comparable lengthswithin any cluster and with lowest resolution were pre-ferred. The final dataset was composed of 103 pairs ofsequence segments (of length �8) having helical andcomplementary nonhelical conformations (see, the flowchart, supplementary Fig. 1).

Amino Acid Composition

The frequency of each amino acid was calculated forboth dynamic and rigid helices. To plot, we subtractedthe a.a. frequency in rigid helices from that of dynamichelices. Thus, the positive values show the relativeabundance of particular amino acid in the dynamic heli-ces and the vice-versa.

110 K.S. SANDHU AND D. DASH

PROTEINS: Structure, Function, and Bioinformatics DOI 10.1002/prot

Mean Hydrophobic Moment

We calculated the mean hydrophobic moments usinghmoment program available in EMBOSS package (http://emboss.sourceforge.net/).

Surface Accessibility and CrystallographicB-Factors

Information on the accessible surface area (ASA) wastaken from DSSP profile. A surface residue is that withASA at least 25% of its nominal maximum surface area.Crystallographic B-factors (temperature factors orDebye–Waller factors) for each Ca atom were taken fromPDB file and normalized using the formula:

B0 ¼ B� B

r

where B is the actual B-factor of Ca atom of the residueB is the average B-factor, and r is standard deviationfor all Ca atoms in each PDB.

Machine Learning

Weka-3.4 machine learning java libraries (http://www.cs.waikato.ac.nz/ml/weka/) were used for the pre-diction of dynamic and rigid helices. The detailedstrategy adapted for the prediction is given in the sup-plementary Figure 2.

Dihedral Angle Difference

Dihedral angle difference (DAD) between helical andnonhelical conformations in a pair was calculated as

DADAi ¼ juA

i � uBi j þ jwA

i � wBi j

where A and B are two structures to be compared and iis the residue. This value is averaged over the entiresegment.

Contact Analysis

Crystal environment (for crystal structures) was com-puted using Cryco server (http://ligin.weizmann.ac.il/�lpgerzon/cryco5.0/cryco/cryst1.cgi). All close contactsaround the helical and nonhelical structures were listedout at atom to atom distance cut-off of <4 A.

Images

Structural images were generated by Visual MolecularDynamics (http://www.ks.uiuc.edu/Research/vmd/) andRasMol (http://www.umass.edu/microbio/rasmol/index2.htm) softwares. The heatmap was produced using anexcel macroscript.

Statistics

Suitable tests of significance were applied. Whereverthe data does not follow the normal distribution (as con-firmed by Shapiro–Wilk normality test), we applied non-parametric test of significance namely the Wilcoxon-testbesides the traditional t-test.

RESULTS AND DISCUSSIONDefinitions

‘‘Dynamic helices’’ are those helices that have alter-nate nonhelical conformation in the representative pair,while the ‘‘rigid helices’’ are those that are structurallyconserved in the representative pair. There were 103dynamic and 398 rigid helices derived from the samePDB dataset (refer to ‘‘Materials and Methods’’ section).‘‘Structural environment’’ is the number and type ofshort contacts experienced in the immediate vicinity(atom to atom distance < 4 A) by the protein segmentconsidered. Water molecules were not included in thisstructural environment, instead, these were analyzedseparately.

Comparison of Dynamic and Rigid Helices

We compared the length, amino acid composition, sur-face accessibility, and conformational flexibility ofdynamic helices with that of rigid helices derived fromthe same protein.

Dynamic helices are relatively short anddepleted in hydrophobic amino acids

The length distribution (Fig. 1) of dynamic and rigidhelices shows that the former are relatively short (8–12residues, i.e. 2–3 complete turns of helix) in length ascompared to the later ones. Structural transition in lon-ger protein segments would involve higher energetic costand hence these are infrequent in our dataset. Thismight also relate to the fact that the long disordered seg-ments of protein structure are generally not resolved inX-ray diffraction pattern and hence are not available inPDB.

Certain combinations of amino acid residues relatedirectly to the local mobility of the protein backbone. In

Fig. 1. Length distribution of dynamic and rigid helices.

111DYNAMIC �-HELICES


a preliminary analysis, we compare the a.a. compositionof dynamic and rigid helices, which differs in the twodatasets. Dynamic helices are deficient in hydrophobica.a. and slightly enriched in hydrophilic and chargeda.a. In dynamic helices, Asn, Lys, Arg are the mostabundant and Val, Leu, Ala the least [Fig. 2(A)]. The v2

tests give P-values that are significant (P ¼ 0.024 and0.022) for under-representation of Val and Cys, andnear-significant (P ¼ 0.067) for over-representation ofLys (supplementary Table I). In particular, under-repre-sentation of Cys may relate to loose packing of dynamichelices, since Cys can contribute to rigidity by makingdisulfide bonds. Similarly, over-representation of Glymay contribute to the dynamic nature of these helices,since Gly is the most flexible residue in terms of tor-sional rotation of the backbone.

Dynamic helices are more surface-accessible

As the surface residues of a monomer may get buriedafter interaction with other monomers/interacting part-ners, we calculated ASA in the monomeric form of theprotein to nullify the effect of interaction(s) on surface

accessibility. We found more surface-exposed residues indynamic helices than in the rigid helices [Fig. 2(B)]. Thetwo-tailed t-test for the difference in proportion ofsurface-exposed residues in dynamic and rigid helices ishighly significant (P-value ¼ 2.50577e-08). Thus, thestructural transitions are more common on the surfacethan the buried regions of a protein. This agrees withthe fact that the protein core is generally rigid andstructurally conserved.

Dynamic helices are less amphiphilic thanrigid helices

Mean hydrophobic moment was calculated to estimatethe amphiphilicity of dynamic and rigid helices. Wefound dynamic helices less amphiphilic than the rigidhelices [Fig. 2(C), P-value of t test ¼ 0.077]. The amphi-philicity of helices play crucial role in proper proteinfolding by exposing its hydrophilic face toward exteriorand hydrophobic toward protein-interior. Since thedynamic helices are depleted in hydrophobic residues,they are less amphiphilic than the other helices. Thismight contribute to their loose packing on the proteinsurfaces and eventually to their dynamic nature.

Fig. 2. Comparison of dynamic and rigid helices. (A) Difference in the amino acid frequencies in dynamic and rigid helices. (B) Distribution ofsurface accessible residues in dynamic and rigid helices. (C) Distribution of hydrophobic moment in dynamic and rigid helices. (D) Distribution ofnormalized B-factor for surface accessible residues in dynamic and rigid helices.



Dynamic helices are inherently flexible

The crystallographic B-factors represent the extent ofdispersal of the electron density due to atomic oscilla-tions around the equilibrium position. Severalauthors15,39–43 have used normalized B-factors to esti-mate the degree of local flexibility in the protein back-bone or side chains.Here, we compared the distribution of normalized B-

factors between dynamic and rigid helices. Since most ofthe residues in dynamic helices are surface-exposed, weanalyzed only surface residues and found these to differ[Fig. 2(D)]. The two-tailed t-test exhibits significant P-value for the higher mean B-factors in dynamic helices(P-value ¼ 3.35118e-07). Thus, the surface residues inthe dynamic helices are more flexible than those in therigid helices. This confirms the inherent dynamic natureof these helices.

Dynamic and rigid helices can be predictedusing ‘‘logistic regression’’ model

In the above analysis, we show several discriminatingproperties of dynamic and rigid helices. These discrimi-nating characteristics can be exploited as attributes topredict the dynamic and rigid helices in proteins. Wenow show that we, indeed, can discriminate the dynamicand rigid helices in proteins using the present dataset.We used the logistic regression model to train the classi-fier on our dataset. Standard 10-fold crossvalidationstrategy is used to test the reliability of the prediction.The overall methodology adopted for prediction is givenas a flow chart in supplementary Figure 2. We predict83.83% of total 501 instances correctly with a meanabsolute error of 0.211. The TP rate, FP rate, precision,recall, and F-measure (product of precision and recall)shown in Table I represent reasonable accuracy of pre-diction. The precision can be improved by training onmore number of dynamic helices (by including closehomologs in the starting dataset). These observationshave important implications in identifying regions proneto conformational transitions in proteins. This can alsobe useful in designing synthetic ‘‘conformationalswitches’’ in proteins.In a Raman optical activity (ROA)-based study, McColl

et al.44 proposed two types of a-helices in proteins/poly-peptides with respect to hydration patterns observed.These were (1) canonical a-helices or ac-helices thatdominate primarily in the hydrophobic environment andappears to be unhydrated. The main chain dihedralangles (uiþ1 and ci) group in the range of �66 (�5)8 and

�41 (�6)8 and (2) Open conformation a-helices or ao-hel-ices, that dominate in the hydrophilic environment andare hydrated. The uiþ1 and Ci dihedral angles group inthe range between �59 (�6)8 and �44 (�6)8. The ac-heli-ces are stabilized by the H-bonding with the water mole-cules or the hydrophilic side chains with the peptidylcarbonyl. Based on discriminatory features of dynamichelices, one can probably relate the dynamic helices toao category. Assuming that the surface-exposed regionsexperience more hydrophilic interactions than the in pro-tein interior, we can safely conclude that the dynamichelices, like ao-helices, prefer to be in the hydrophilicenvironment. However, the comparison of uiþ1 and Ci di-hedral angles in dynamic and rigid helices did not revealany significant grouping in the characteristic range of ac

or ao helices. Both the helices showed almost equalgrouping in two characteristic ranges (data not shown).The comparison of hydration pattern would not be justi-fied here, since there is a significant difference in thesolvent accessibility of dynamic and rigid helices. There-fore, we cannot categorize the dynamic helices solely toao category.

Analysis of Dynamic Helix and ComplementaryNon-Helical ConformationConformational perturbation in dynamic helices

The right-handed a-helical conformation falls in therange �1208 to �108 and �1208 to 208, of u–c angles,respectively, in the Ramachandran plot. It is expectedthat, upon transition from helix to nonhelix, the u–cangles of the right-handed a-helical conformation will beperturbed. The Ramachandran plots in Figure 3 indeedshow that after transition, the u–c values spread outfrom the allowed region of right-handed helical confor-mation. The average dihedral angle difference (DAD)varies from 11.278 to 208.648, suggesting a wide range ofperturbations in the dynamic helices. In 89 out of 103cases, the average DAD values were above 408 (supple-mentary Fig. 3). Thus, in most cases, the conformationwas noticeably different in the two conformationalstates. The exact conformation (as given by DSSP) of thenonhelical counterparts of the dynamic helices wasmostly turn (48%), bend (15%), and 310 helix (20%). TheDSSP does not calculate the PPII helical conformation.Using an in-house computer program that identifies PPIIconformation using backbone torsion angles (centeredaround �758 and 1458 values of u, c angles respectivelywith deviation of �208), we found only 2% residues fall-ing in the u–c range of PPII conformation. The datasetcompletely lacks a complete turn of PPII helix (i.e., fourcontinuous residues in PPII conformation). To nullify theeffect of crystal contacts, we analyzed a subset, whereinthe PDB structures of nonhelical counterparts of thedynamic helices were determined using NMR method.Again, we do not see any preference for the PPII confor-mation. Out of total 205 points, only two points fall in thePPII range of u, c angles. This is inconsistent to a reportof Blanch et al.45 on human lysozyme, where PPII is pro-

TABLE I. Performance Measures for Prediction ofDynamic and Rigid Helices

Class TP rate FP rate Precision Recall F-measure

Dynamic helix 0.73 0.13 0.59 0.73 0.65Rigid helix 0.87 0.27 0.93 0.87 0.90



posed to be preferred conformation upon helix-unfolding.However, unlike our study, their study involved fullyhydrated and temperature-driven partially denaturedstate of the protein. Also, they proposed that PPII is anintermediate conformation during helix-to-sheet transi-tion pathway, since the main chain torsion angles forPPII and b-sheet conformations are in close vicinity inthe Ramachandran plot. This probably justifies the dis-crepancy observed in the two studies.

Structural environment differs in twoconformational states

Since the multiple structures of the same protein pres-ent in the PDB may differ in their structural environ-ment, we analyzed the structural environment for therepresentative pairs of helical and nonhelical conforma-tions.

Helical conformation experiences more numberof contacts. The supplementary Table II shows thenumber of crystal and noncrystal contacts (consideringcenter to center distance <4 A). The structural environ-ment or the number of short contacts experienced differssignificantly between the two conformational states (P-valuewilcoxon-test, paired, two-sided ¼ 0.003474, P-valuet-test,paired, two-sided ¼ 0.023). The dynamic helices experiencemore contacts in their helical conformation than in theirnonhelical counterparts (P-valuewilcoxon-test, paired, one-sided

¼ 0.001737, P-valuet-test, paired, one-sided ¼ 0.01166). Innumbers, there are 81 cases in the dataset of 103, wherethe number of contacts differs in two states and this sub-set encompasses 53 cases where dynamic helices experi-ence more contacts in the helical state (Fig. 4). However,using the present methodology to derive the representa-

tive pairs for comparison, we cannot guarantee that allpossible pairs of helical and nonhelical conformations dif-fer in their structural environment. The dynamic helicesmay experience the same environment in both the confor-mational states in certain conditions. Conformational dif-ferences in such cases may arise as a consequence of longrange interactions or inherent nature of the structure orother physical changes such as temperature and pH. Inany case, it is not easy to discern the real cause for theconformational transitions by computational methods.Naturally, any technical artifacts during the determina-tion of the protein structure may also affect the results.With these issues in the forefront, we cannot statewhether the conformational changes occur due solely toaberrations in the environment or are inherent to the pro-tein structure.

The observation that helical state of dynamic helicesexperiences more contacts than the nonhelical state sug-gests that these transitions represent the events ofcoupled folding and binding. The few cases, where theopen conformation experiences more contacts than thecorresponding helical conformation may arise due to dis-crepancy from the routine stabilizing interactions withthe immediate surroundings (discussed later). It can alsobe a specific conformational switch for ligand-binding orrelease [as in the case of Fig. 8(I)].

Further, we compared the total number of contactsmade in each kind of interaction (oligomer, ligand, crys-tal, etc.) in both the conformations. We find that theoligomer-interactions, in contrast to crystal- and ligand-interactions have larger proportion of contacts in thenonhelical conformation than in the helical conformation(P-value < 2.2e-16). This suggests that these are mostlycrystal and ligand-contacts that lead to ordered struc-tures upon interaction than the oligomer-contacts. Thisis true since the appropriate crystal environment is bi-ased toward ordered compact protein structure, similarlyligand interaction requires precise and rigid geometricalfold for specificity, but the same is not true for the pro-tein–protein interaction.

Fig. 5. Amino acid frequencies in oligomer-, crystal-, and ligand-con-tacts (small molecules, nucleic acids, metal ions). The two bars are forhelical and non-helical conformation of dynamic helix respectively. Leftpanel shows hydrophobic amino acids and right panel shows hydrophilicamino acids.

Fig. 4. Heatmap representation of short contacts experienced bytwo alternate conformations of dynamic helices. NH and NX representthe number of short contacts experienced in helical and nonhelical con-formation, respectively. The color bars represent the relative numbersof short contacts.

Fig. 3. Ramachandran plots for dynamic helices in their (A) a-helicalconformation (B) non-helical conformation.



Aminoacid preference for short-contacts. To verifywhether amino acids that make short contacts indynamic helices are represented differentially in the dif-ferent kind of contacts, we plotted amino acid frequenciesin oligomer-, crystal-, and ligand-contacts (small mole-cule, nucleic acid, metal ion contacts) (Fig. 5). The prefer-ence of hydrophilic amino acid in crystal contacts (longerblack colored bars in hydrophilic panel in Fig. 5) is in co-herence with the earlier results by Dasgupta et al.46 Thisshows the strong selection against hydrophobic aminoacids in crystal lattice.47,48 By lattice model simulations,it has been shown that more hydrophobic contacts lead todisordered aggregates, not the organized crystal lattices.Although, the burial of hydrophobic surface results indecrease in the entropy and hence stabilizing the interac-tion, only optimum representation of hydrophobic surfaceburial determines this phenomenon.47,48 This optimalitydiscriminates the organized protein crystals from the dis-ordered aggregates. This is one of the reasons for the dif-ficulties faced in crystallizing the membrane proteinsthat are enriched in hydrophobic amino acids. Discrep-ancy in the a.a. frequencies for short-contacts is observedwhen helical and nonhelical states of dynamic helices arecompared. A deeper insight on such discrepancies ininteraction is given in the next subsection.In addition, we compared the amino acid frequencies in

ligand-contacts with that represented in binding pocket

(BPK) and LIG databases.49 In the case of helical confor-mation, we find significant correlations (P-values ¼ 0.0037,0.0036) between our dataset and BPK/LIG databases (Fig.6). The similar analysis with nonhelical conformationreveals insignificant correlations due to insufficient andinappropriate representation of routine ligand-interac-tions. Thus, the routine ligand-interactions actually preferthe ordered conformation. This is in agreement with thefact that ligand-binding requires precise geometrical orien-tation for binding, in contrast to other interactions thatcan be achieved by linear unstructured regions.

Type of short contacts differs in helical and non-helical conformations. We analyzed the type of con-tacts observed in the three classes (Fig. 7). All contactsare classified into carbon–carbon (C–C), carbon–heteroa-tom (C–X), and heteroatom–heteroatom (X–X) interac-tions. C–C contacts reflect the hydrophobic interactions,in contrast to X–X contacts that estimate salt bridgesand hydrogen bonds. In the case of helical conformation,no significant difference was observed between oligomer-and crystal-contacts. This is also consistent with theearlier observation by Dasgupta et al.46 where onlymoderate differences were observed. The ligand-contactsshow more X–X type of contacts and less C–C contacts.This is consistent with the fact that ligand-binding (smallmolecule, nucleic acid, etc.) prefer salt bridges or hydrogenbond interactions.49,50 Again, the observations deviate in

Fig. 6. Pearson’s correlations between our ‘‘ligand-contacts’’ dataset and BPK/LIG datasets: (A) for con-tact-residues in helical conformation; (B) for contact-residues in nonhelical conformation.



the case of nonhelical conformation. When compared tohelical conformation, nonhelical conformation showedmore C–C contacts in crystal- and ligand-interactionsthan in the oligomer-interactions (P-valuev-squared test ¼0.055, Fig. 7). In contrast, more X–X contacts were seen inoligomer-interactions than in the crystal- and ligand-interactions when compared to helical conformation(P-valuesv-squared test ¼ 2.32e-09 and 8.33e-53). Morehydrophobic interactions in the crystal and ligand con-tacts may lead to disorderliness in the local conformationof the protein. This is in coherence with the earlier studiesthat show that more hydrophobic crystal-lattice interac-tions lead to less ordered crystal structures.47,48 On theother hand, the oligomer-contacts enriched in hydrophilicinteractions generally lead to flexible interfaces.51 Thissuggests that there are interactions that favor the orderli-ness of the protein structure, in contrast to other ‘‘few’’that disfavor. In particular, it’s the relative proportion ofhydrophobic and hydrophilic interactions that differs.However, our observation is limited only to dynamic heli-

ces. To consolidate this observation, one needs to analyzeall the interactions on the protein surface with respect toorderliness and disorderliness of the structure.

Thus, we showed that there are quantitative as wellas qualitative differences in the immediate structuralenvironments that are observed around the stabilizedhelical conformation and the unstable open conforma-tion. The aberrations in the immediate structural envi-ronment assisted with the inherent tendency of struc-tural deformability result to local structural transitionsin proteins. Again, the relative contribution of each ofthe factors involved cannot be estimated using computa-tional method.

Based on hydration pattern of a-helices in X-raycrystal structures, Sundaralingam and Sekharudu52 pro-posed that water molecules stabilize intermediate confor-mations (like turns) in the helix-unfolding pathway. Ouranalysis on crystal structures (78 representative pairs)did not reveal any preference for close water contacts(X–X type of contacts with a distance cutoff of < 3A8) inthe nonhelical conformations of the dynamic helices. TheP-values of significance tests were insignificant for thedifference in the number of water-contacts in the twostates (Pt-test, two-sided ¼ 0.6037, Pwilcoxin-test, two-sided ¼0.2470). A major proportion (47.4%) of this dataset didnot show any close water contact in either of the confor-mations. Based on this observation, we believe thatwater molecules does not play a significant role here andtherefore the structural changes in our dataset appear tobe mainly due to aberrations in the structural environ-ment other than water.

Illustrative examples

We now examine few illustrative examples of confor-mational transitions.

Nucleic acid interaction. Several unstructuredregions in DNA/RNA binding proteins have beenreported to interact with DNA or RNA molecules.53

These unstructured regions may become structured uponinteraction with nucleic acid.26,54 The solution structureof ribosomal protein S15 from Thermus thermophilusshows several molten-globule regions [Fig. 8(A)]55 thatbecome highly structured upon interaction with RNAmolecule to form 30S ribosomal subunit [Fig. 8(A)].56,57

Contact analysis revealed that several hydrophilic andcharged residues of the protein (Lys5, Lys8, Gln9, Ile12,His50, His51, Ser52, Arg54, Gly55, Leu57, Gln62, Arg64,Arg65, Arg68, Tyr69, and Arg72) interact with the basesand phosphate backbone of the RNA molecule (supple-mentary Table III). Average DAD was 95.698 in thiscase. The S15 protein along with S8 and S17 bind tothree way junction formed by H20, H21, and H22 (RNAhelix nomenclature) of the 30S subunit. Such inter-actions coupled with transitions make compact macro-molecular assembly of the central domain and help theRNA to acquire proper tertiary fold. The other exampleis Trp repressor protein.58,59 The solution structures ofboth free and DNA-bound forms show the conformationalchange in the DNA interacting regions of the protein

Fig. 7. Type of contacts in oligomer-, crystal, and ligand-interactions.C–C stand for carbon–carbon contacts, C–X for carbon-heteroatomcontacts, and X–X for heteroatom-heteroatom contacts: (A) for contactsmade by helical conformation; (B) for contacts made by nonhelical con-formation.



[Fig. 8(B)]. A helix, present in the DNA–protein interfacein the DNA-bound form of the protein, acquires flexibleconformation in the unbound free-state [Fig. 8(B)]. Thesetransitions may decrease the conformational entropythat is compensated by favorable enthalpic interactions

and entropy of solvent exclusion. Such transitions areexpected to be reversible, since entropic cost associatedwith the transition shifts the equilibrium constant moretoward the value 1. The reversibility is important in theregulation of process these proteins involved in. Similar

Fig. 8. Examples of conformational transitions in the a-helices. The dynamic helices are colored in red. (A) Ribosomal protein S15 from Thermusthermophilus. The uncomplexed form (1AB3) of the protein has several unstructured regions (red-colored) that adopts helical conformation uponbinding to ribosomal complex (1FJ3). RNA molecule is shown as space-filled model. (B) Trp repressor from Eescherichia coli. 1WRT is the NMRstructure of free form and 1RCS is the NMR structure of DNA (space-filled model)-bound form of the protein. (C) Ribosomal protein L25 fromEescherichia coli. 1B75 is the solution structure of the uncomplexed form and 1DFU is the RNA (space-filled model) complexed form of the protein.(D) NMR structures of cellular retinol-binding protein II from Rattus norvegicus. 1B4M is the apo-form and 1EII is the holo-form of the enzyme (ligandis shown in iceblue color and CPK model). (E) 1BLR is the solution structure and 1CBS is the ligand-bound crystal structure of cellular retinoic acidbinding protein II (CRABP II). Ligand is shown in blue and neighboring crystal units are shown in grey wireframe model. (F) Apo- and holo-forms ofMouse Ferro-Cytochrome. 1IET is the solution structure of the free-form and 1B5A is the solution structure of heme-bound form. Ligand is shown iniceblue color and CPK model. (G) 1APC is the NMR structure of free-form and 256B is the crystal structure bound-form of Cytochrome B562. (H)Crystal structures of Homoserine Kinase C. 1H73 is complex with HSE þ ANPPNP and 1H72 is complex with Thr þ ANPPNP. The Ser133 is shownas VDW model. The difference in the ser133 and ligand interaction can be seen. In 1H72, the Oxygen atom of Ser133 is in close contacts with ligand(blue colore, CPK model). The same is not true with 1H73 where the region shows the non-helical conformation. (I) Crystal structures of Bombyxmori Pheromone-Binding Protein (BmorPBP). 1DQE is the holo- and 1GM0 is the apo-form of the protein. Two helices showing transitions are col-ored as red and iceblue. Ligand is shown in magenta color and CPK model. Transition in the red helix is a conformational switch to release the ligandfrom the binding pocket at lower pH. (J) Crystal structures of Annexin V protein. 1ANX shows formation of helix upon binding to a Ca2þ ion (cyancolor) and a phospholipids molecule (VDW model). Ca2þ ion is coordinated with carbonyls of D226 and T229. (K) Crystal structures of l-serine dehy-dratase from rat liver. PDB 1PWE is the apo form and 1PWH is the holo form. The dynamic helices are shown as cartoon models and neighboringcrystal units as grey wire frames. (L) Zoomed image for the chain A of 1PWE and 1PWH. Ligand is shown in blue and potassium ion (Kþ) is shownin cyan color. (M) Trp repressor from Eescherichia coli. 1RCS is the DNA (wire-frame) bound solution structure and 1P6Z is the unbound crystalstructure of the protein (neighboring crystals are shown as wire diagram). (N) Crystal structures of human hemoglobin. 1IRD is the carbonmonoxy-hemoglobin and 1FN3 is nickel reconstituted hemoglobin, crystal contacts are shown in gray wireframe and ligand as cyan CPK model. (O) and (P)are the examples of transition due to crystal contacts. Close crystal contacts are shown in blue threads in 1JOB-N and as light-grey threads in1CAV. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]



changes are observed in the case of Ribosomal L25 pro-tein from Escherichia coli.60,61 The unbound form con-tains an unstructured region that acquires helical con-formation when interacts with RNA molecule [Fig. 8(C)].

Small molecule interaction. Ligand-induced con-formational stability has been reviewed in the litera-ture.62 We have several such examples in our dataset.Residues 28–36 form a helix in the holo form of cellular

Figure. 8. (Continued)



retinol-binding protein II (CRBP II) [Fig. 8(D)]63 and reti-nol molecule is completely buried in the protein. The he-lix, which is a part of helix-turn-helix (HTH) motif, ismissing in the apo form of the protein. Thr30, Arg31,Ala34, residues of this helix were in the close contacts

with the retinol molecule. There were several other con-formational changes in the structure upon binding to reti-nol, for example, several b-strands are induced uponligand binding. These changes were also discussed by theauthor who solved the ligand-bound solution structure of

Figure. 8. (Continued)



the protein.64 It is known that retinol dissociates fromCRBP II more rapidly than in the case of CRBP. It hasbeen proposed that the conformational perturbations inthe HTH motif facilitate the retinol release from CRBP II.Similar changes were observed in case of cellular retinoicacid binding protein II (CRABP II).65,66 The solution struc-ture of the protein shows dynamic disorder in the bindingpocket and ligand-binding induces significant conforma-tional changes [Fig. 8(E)]. The solution structure of Ferro-Cytochrome is a molten globule structure [Fig. 8(E)]67 andan unstructured region present toward the C terminal ofthe protein adopts helical conformation upon interactionwith the heme moiety [Fig. 8(F)].68 The same is true foranother heme-binding protein Cytochrome B562 [PDB ids:1APC and 256B, Fig. 8(G)].69,70 This is consistent with anearlier observation on apo- and holo-myoglobin.18

Crystal structures of homoserine kinase C are solved incomplex with HSE [Homoserine] þ AMPPNP [50-adenyly-limidotriphosphate] (PDB id: 1H72) and Threonine þAMPPNP (PDB id: 1H73).71 We found that the region127–134 forms a helix in PDB 1H72. The C-b and Oatoms of Ser133 make close contacts with AMPPNP mole-cule [Fig. 8(H)]. These contacts were absent in the PDB1H73, probably because of different orientation andfitting of ligand and, in turn, the region does not formhelical conformation. We propose that this might be a con-sequence of differential ligand-binding in these two cases.There were a few cases where the scenario is vice-

versa, i.e., the ligand bound form of the protein containsa disordered region (region 131–141) that acquires heli-cal structure in the ligand free form. For example, theC-terminal of Bombyx mori pheromone-binding protein(BmorPBP) is unstructured in pheromone bound state16

and adopts a helical structure in pheromone-free state[Fig. 8(I)]. It has been proposed that local drop in pHleads to protonation of acidic residues of the C-terminalthat further acquires helical conformation and movesinto the binding pocket to displace the ligand.16 Thus,the coil to helix transition here is a conformationalswitch to release the ligand from the binding pocket.The N-terminal (region 2–11) of the same protein is dis-ordered in the ligand-free form and adopts helical confor-mation upon ligand interaction, similar to other casesdiscussed above.Ion binding proteins generally show structural poly-

morphism in apo and holo form.72–75 Annexin is thefamily of proteins that bind to membrane in calcium-de-pendent manner. The conformations of different calcium-binding loops change upon calcium and lipid binding.76–79

In our dataset, we observe that region 221–228 of theAnnexin V binds to calcium ion and phospholipid mole-cule and forms a helical structure [Fig. 8(J)]. The calciumion is coordinated with the carbonyls of D226 and T229.

Protein–protein interaction and crystal contacts.The disordered regions may adopt rigid secondary

structure upon interaction with other proteins.32,33 Wehave few such cases in our dataset. For example, PDB1PWE and 1PWH both belongs to l-serine dehydratase,

the former is apo-form and latter is holo-form [Fig.8(K,L)].80 The authors, who solved the structures ofthese two forms, reported conformational change in theregion 193–234. The loss of helical segment 200–207 (asreflected in our dataset too) is proposed to be the conse-quence of release of ligand from the holo form, however,the ligand itself does not interact with this helical seg-ment. We hereby, propose another possible factor thatcan contribute to such conformational perturbation. Ourcontact analysis shows that the structural environmentof helical and the nonhelical structures differs (supple-mentary Table III). For each monomeric unit of thecomplex, it is found that helical conformation in theholo-form experiences more short contacts than the non-helical conformation in the apo-form. These contactseither arise from neighboring monomers or the crystalpacking [Fig. 8(K,L), supplementary Table III].

There are several examples in our dataset where con-formational transitions due to crystal packing areobserved [Fig. 8(M–P)]. These were not discussed in thecorresponding articles at all. The difference in the con-formational flexibility (normalized B-factor) upon crys-tal packing has previously been observed by Eyalet al.,81 however, our observation is the first report forsuch large conformational changes induced by closecrystal contacts. Thus, the transitions studied here arenot always the consequence of physical interaction of bi-ological significance, but may also be an artifact of theexperimental method. This observation has implicationsin homology-based protein structure prediction. Suchcases should be treated with caution while training anycomputational model for protein structure prediction.Our library of structural transitions upon interactioncan also be important in designing predictive dockingalgorithms.

Our observations show the intrinsic tendency of con-formational transitions in certain regions on the proteinsurface in response to their immediate structural envi-ronment. Since the observed number of cases for morecontacts in helical conformation is significantly higherthan that for the nonhelical conformation, we proposethat these regions undergo folding and binding that arecoupled together. The transitions coupled with bindingmay regulate the biological process they are involved in.In the presence of specific and favorable interaction, theopen extended region may adopt the helical conformationthat further facilitates appropriate binding with thepartner. Thus, the dynamic helices may represent the‘‘molecular switches’’ that regulate the recognition andbinding of the partner and hence regulating the biologi-cal process downstream. The molecular functions ofthese proteins are mostly nucleic acid and ligand-binding(supplementary Table II) that are crucial for severalimportant biological processes in the cell. Since thedynamic helices are surface-exposed, can adopt the openconformation, and are enriched in certain hydrophilicamino acids that are preferred in proteolysis/ubiquitina-tion (like Lys, Asn, Glu, Ser etc), these may also regulatethe protein turn over in the cell.



CONCLUSIONS

The database level study of protein structures revealsseveral short segments prone to ‘‘helix$coil’’ transitions.Our analysis confirms their higher surface accessibilityand conformational mobility. We demonstrate usingsequence and structure features that we can predict theregions in protein that can potentially undergo such con-formational transitions. Further, these regions are sensi-tive to their structural environment. In particular, thenative helical conformation and the alternate nonhelicalconformation differ significantly in their structural envi-ronment. Several of the cases identified are in agreementwith the literature available. We propose that thesedynamic helices may function as ‘‘molecular switches’’ inthe proteins for the regulation of certain biological process.

ACKNOWLEDGMENTS

We thank Dr. Eran Eyal for help in generating thecrystal environments for our PDB dataset. Authorsthank Dr. Shantanu Chaudhary, Dr. S. P. Modak, Dr.Souvik Maiti, Mythily Ganapathi, and Gajinder PalSingh for their useful suggestions in writing manuscript.

REFERENCES

1. Uversky VN. What does it mean to be natively unfolded? Eur JBiochem 2002;269:2–12.

2. Uversky VN, Gillespie JR, Fink AL. Why are ‘‘natively unfolded’’proteins unstructured under physiologic conditions? Proteins2000;41:415–427.

3. Dyson HJ, Wright PE. Intrinsically unstructured proteins andtheir functions. Nat Rev Mol Cell Biol 2005;6:197–208.

4. Uversky VN. Natively unfolded proteins: a point where biologywaits for physics. Protein Sci 2002;11:739–756.

5. Ward JJ, Sodhi JS, McGuffin LJ, Buxton BF, Jones DT. Predic-tion and functional analysis of native disorder in proteins fromthe three kingdoms of life. J Mol Biol 2004;337:635–645.

6. Vucetic S, Brown CJ, Dunker AK, Obradovic Z. Flavors of pro-tein disorder. Proteins 2003;52:573–584.

7. Ward JJ, McGuffin LJ, Bryson K, Buxton BF, Jones DT. TheDISOPRED server for the prediction of protein disorder. Bioin-formatics 2004;20:2138–2139.

8. Weathers EA, Paulaitis ME, Woolf TB, Hoh JH. Reduced aminoacid alphabet is sufficient to accurately recognize intrinsicallydisordered protein. FEBS Lett 2004;576:348–352.

9. Linding R, Jensen LJ, Diella F, Bork P, Gibson TJ, Russell RB.Protein disorder prediction: implications for structural proteo-mics. Structure 2003;11:1453–1459.

10. Dosztanyi Z, Csizmok V, Tompa P, Simon I. The pairwise energycontent estimated from amino acid composition discriminatesbetween folded and intrinsically unstructured proteins. J MolBiol 2005;347:827–839.

11. Lise S, Jones DT. Sequence patterns associated with disorderedregions in proteins. Proteins 2005;58:144–150.

12. Singh GP, Ganapathi M, Sandhu KS, Dash D. Intrinsic unstruc-turedness and abundance of PEST motifs in eukaryotic pro-teomes. Proteins 2006;62:309–315.

13. Quigley PM, Korotkov K, Baneyx F, Hol WG. A new nativeEcHsp31 structure suggests a key role of structural flexibilityfor chaperone function. Protein Sci 2004;13:269–277.

14. Tompa P, Szasz C, Buday L. Structural disorder throws newlight on moonlighting. Trends Biochem Sci 2005;30:484–489.

15. Sandhu KS, Dash D. Conformational flexibility may explainmultiple cellular roles of PEST motifs. Proteins 2006;21:21.

16. Lautenschlager C, Leal WS, Clardy J. Coil-to-helix transitionand ligand release of Bombyx mori pheromone-binding protein.Biochem Biophys Res Commun 2005;335:1044–1050.

17. Fontana A, Fassina G, Vita C, Dalzoppo D, Zamai M, Zambo-nin M. Correlation between sites of limited proteolysis andsegmental mobility in thermolysin. Biochemistry 1986;25:1847–1851.

18. Picotti P, Marabotti A, Negro A, Musi V, Spolaore B, ZamboninM, Fontana A. Modulation of the structural integrity of helix Fin apomyoglobin by single amino acid replacements. Protein Sci2004;13:1572–1585.

19. Best RB, Chen YG, Hummer G. Slow protein conformational dy-namics from multiple experimental structures: the helix/sheettransition of arc repressor. Structure 2005;13:1755–1763.

20. Verkhivker GM, Bouzida D, Gehlhaar DK, Rejto PA, Freer ST,Rose PW. Simulating disorder-order transitions in molecular rec-ognition of unstructured proteins: where folding meets binding.Proc Natl Acad Sci USA 2003;100:5148–5153.

21. Shamsir MS, Dalby AR. One gene, two diseases and three con-formations: molecular dynamics simulations of mutants ofhuman prion protein at room temperature and elevated temper-atures. Proteins 2005;59:275–290.

22. Flores TP, Orengo CA, Moss DS, Thornton JM. Comparison ofconformational characteristics in structurally similar proteinpairs. Protein Sci 1993;2:1811–1826.

23. Zhao S, Goodsell DS, Olson AJ. Analysis of a data set of paireduncomplexed protein structures: new metrics for side-chain flex-ibility and model evaluation. Proteins 2001;43:271–279.

24. Babor M, Greenblatt HM, Edelman M, Sobolev V. Flexibility ofmetal binding sites in proteins on a database scale. Proteins2005;59:221–230.

25. Eyal E, Najmanovich R, Edelman M, Sobolev V. Protein side-chain rearrangement in regions of point mutations. Proteins2003;50:272–282.

26. Horiuchi M, Kurihara Y, Katahira M, Maeda T, Saito T, UesugiS. Dimerization and DNA binding facilitate a-helix formation ofMax in solution. J Biochem (Tokyo) 1997;122:711–716.

27. Warrant RW, Kim SH. a-Helix-double helix interaction shown inthe structure of a protamine-transfer RNA complex and a nucle-oprotamine model. Nature 1978;271:130–135.

28. Uversky VN, Narizhneva NV. Effect of natural ligands on thestructural properties and conformational stability of proteins.Biochemistry (Mosc) 1998;63:420–433.

29. Stellwagen E, Rysavy R, Babul G. The conformation of horseheart apocytochrome c. J Biol Chem 1972;247:8074–8077.

30. Breitenlechner C, Engh RA, Huber R, Kinzel V, Bossemeyer D,Gassel M. The typically disordered N-terminus of PKA can foldas a helix and project the myristoylation site into solution. Bio-chemistry 2004;43:7743–7749.

31. Vetter SW, Leclerc E. Phosphorylation of serine residues affectsthe conformation of the calmodulin binding domain of humanprotein 4.1. Eur J Biochem 2001;268:4292–4299.

32. Radhakrishnan I, Perez-Alvarado GC, Parker D, Dyson HJ, Mont-miny MR, Wright PE. Solution structure of the KIX domain of CBPbound to the transactivation domain of CREB: a model for activa-tor:coactivator interactions. Cell 1997;91:741–752.

33. Bowman P, Galea CA, Lacy E, Kriwacki RW. Thermodynamiccharacterization of interactions between p27(Kip1) and activatedand non-activated Cdk2: intrinsically unstructured proteins asthermodynamic tethers. Biochim Biophys Acta 2006;1764:182–189.

34. Gerstein M, Lesk AM, Chothia C. Structural mechanisms fordomain movements in proteins. Biochemistry 1994;33:6739–6749.

35. Betts MJ, Sternberg MJ. An analysis of conformational changeson protein-protein association: implications for predictive dock-ing. Protein Eng 1999;12:271–283.

36. Gutteridge A, Thornton J. Conformational changes observed inenzyme crystal structures upon substrate binding. J Mol Biol2005;346:21–28.

37. Oldfield CJ, Cheng Y, Cortese MS, Romero P, Uversky VN,Dunker AK. Coupled folding and binding with a-helix-formingmolecular recognition elements. Biochemistry 2005;44:12454–12470.

38. Fuxreiter M, Simon I, Friedrich P, Tompa P. Preformed struc-tural elements feature in partner recognition by intrinsicallyunstructured proteins. J Mol Biol 2004;338:1015–1026.

39. Ringe D, Petsko GA. Study of protein dynamics by X-ray diffrac-tion. Methods Enzymol 1986;131:389–433.



40. Parthasarathy S, Murthy MR. Analysis of temperature factordistribution in high-resolution protein structures. Protein Sci1997;6:2561–2567.

41. Yuan Z, Zhao J,Wang ZX. Flexibility analysis of enzyme active sites bycrystallographic temperature factors. ProteinEng 2003;16:109–114.

42. Radivojac P, Obradovic Z, Smith DK, Zhu G, Vucetic S, BrownCJ, Lawson JD, Dunker AK. Protein flexibility and intrinsic dis-order. Protein Sci 2004;13:71–80.

43. Bhalla J, Storchan GB, MacCarthy CM, Uversky VN, Tcherkas-skaya O. Local flexibility in molecular function paradigm. MolCell Proteomics 2006;5:1212–1223.

44. McColl IH, Blanch EW, Hecht L, Barron LD. A study of a-helixhydration in polypeptides, proteins, and viruses using vibrationalraman optical activity. J Am Chem Soc 2004;126:8181–8188.

45. Blanch EW, Morozova-Roche LA, Cochran DA, Doig AJ, HechtL, Barron LD. Is polyproline II helix the killer conformation? ARaman optical activity study of the amyloidogenic prefibrillarintermediate of human lysozyme. J Mol Biol 2000;301:553–563.

46. Dasgupta S, Iyer GH, Bryant SH, Lawrence CE, Bell JA. Extentand nature of contacts between protein molecules in crystal lat-tices and between subunits of protein oligomers. Proteins1997;28:494–514.

47. Patro SY, Przybycien TM. Simulations of reversible protein ag-gregate and crystal structure. Biophys J 1996;70:2888–2902.

48. Patro SY, Przybycien TM. Simulations of kinetically irreversibleprotein aggregate structure. Biophys J 1994;66:1274–1289.

49. Najmanovich R, Kuttner J, Sobolev V, EdelmanM. Side-chain flex-ibility in proteins upon ligand binding. Proteins 2000;39: 261–268.

50. Ahmad S, Gromiha MM, Sarai A. Analysis and prediction ofDNA-binding proteins and their binding residues based on com-position, sequence and structural information. Bioinformatics2004;20:477–486.

51. Korn AP, Burnett RM. Distribution and complementarity of hy-dropathy in multisubunit proteins. Proteins 1991;9:37–55.

52. Sundaralingam M, Sekharudu YC. Water-inserted a-helical seg-ments implicate reverse turns as folding intermediates. Science1989;244:1333–1337.

53. Minezaki Y, Homma K, Kinjo AR, Nishikawa K. Human tran-scription factors contain a high fraction of intrinsically disor-dered regions essential for transcriptional regulation. J Mol Biol2006;359:1137–1149.

54. Hyre DE, Klevit RE. A disorder-to-order transition coupled toDNA binding in the essential zinc-finger DNA-binding domainof yeast ADR1. J Mol Biol 1998;279:929–943.

55. Berglund H, Rak A, Serganov A, Garber M, Hard T. Solutionstructure of the ribosomal RNA binding protein S15 from Ther-mus thermophilus. Nat Struct Biol 1997;4:20–23.

56. Carter AP, Clemons WM, Brodersen DE, Morgan-Warren RJ,Wimberly BT, Ramakrishnan V. Functional insights from thestructure of the 30S ribosomal subunit and its interactions withantibiotics. Nature 2000;407:340–348.

57. Wimberly BT, Brodersen DE, Clemons WM Jr, Morgan-WarrenRJ, Carter AP, Vonrhein C, Hartsch T, Ramakrishnan V. Struc-ture of the 30S ribosomal subunit. Nature 2000;407:327–339.

58. Zhang H, Zhao D, Revington M, Lee W, Jia X, Arrowsmith C,Jardetzky O. The solution structures of the trp repressor-opera-tor DNA complex. J Mol Biol 1994;238:592–614.

59. Zhao D, Arrowsmith CH, Jia X, Jardetzky O. Refined solutionstructures of the Escherichia coli trp holo- and aporepressor. JMol Biol 1993;229:735–746.

60. Stoldt M, Wohnert J, Gorlach M, Brown LR. The NMR structureof Escherichia coli ribosomal protein L25 shows homology togeneral stress proteins and glutaminyl-tRNA synthetases.EMBO J 1998;17:6377–6384.

61. Lu M, Steitz TA. Structure of Escherichia coli ribosomal proteinL25 complexed with a 5S rRNA fragment at 1.8-A resolution.Proc Natl Acad Sci USA 2000;97:2023–2028.

62. Uversky VN, Narizhneva NV. Effect of natural ligands on thestructural properties and conformational stability of proteins.Biochemistry (Mosc.) 1998;63:420–433.

63. Lu J, Lin CL, Tang C, Ponder JW, Kao JL, Cistola DP, Li E. Thestructure and dynamics of rat apo-cellular retinol-binding pro-tein II in solution: comparison with the X-ray structure. J MolBiol 1999;286:1179–1195.

64. Lu J, Lin CL, Tang C, Ponder JW, Kao JL, Cistola DP, Li E.Binding of retinol induces changes in rat cellular retinol-bindingprotein II conformation and backbone dynamics. J Mol Biol2000;300:619–632.

65. Wang L, Li Y, Abildgaard F, Markley JL, Yan H. NMR solutionstructure of type II human cellular retinoic acid binding protein:implications for ligand binding. Biochemistry 1998;37:12727–12736.

66. Kleywegt GJ, Bergfors T, Senn H, Le Motte P, Gsell B, Shudo K,Jones TA. Crystal structures of cellular retinoic acid bindingproteins I and II in complex with all-trans-retinoic acid and asynthetic retinoid. Structure 1994;2:1241–1258.

67. Dangi B, Sarma S, Yan C, Banville DL, Guiles RD. The origin ofdifferences in the physical properties of the equilibrium forms ofcytochrome b5 revealed through high-resolution NMR structuresand backbone dynamic analyses. Biochemistry 1998;37:8289–8302.

68. Falzone CJ, Mayer MR, Whiteman EL, Moore CD, Lecomte JT.Design challenges for hemoproteins: the solution structure ofapocytochrome b5. Biochemistry 1996;35:6519–6526.

69. Feng Y, Sligar SG, Wand AJ. Solution structure of apocyto-chrome b562. Nat Struct Biol 1994;1:30–35.

70. Hamada K, Bethge PH, Mathews FS. Refined structure of cyto-chrome b562 from Escherichia coli at 1.4 A8 resolution. J MolBiol 1995;247:947–962.

71. Krishna SS, Zhou T, Daugherty M, Osterman A, Zhang H.Structural basis for the catalysis and substrate specificity ofhomoserine kinase. Biochemistry 2001;40:10810–10818.

72. Gatewood JM, Schroth GP, Schmid CW, Bradbury EM. Zinc-induced secondary structure transitions in human sperm prot-amines. J Biol Chem 1990;265:20667–20672.

73. Uversky VN, Gillespie JR, Millett IS, Khodyakova AV, VasilenkoRN, Vasiliev AM, Rodionov IL, Kozlovskaya GD, Dolgikh DA,Fink AL, Doniach S, Permyakov EA, Abramov VM. Zn(2þ)-mediated structure formation and compaction of the ‘‘nativelyunfolded’’ human prothymosin alpha. Biochem Biophys ResCommun 2000;267:663–668.

74. Engel J, Taylor W, Paulsson M, Sage H, Hogan B. Calcium bind-ing domains and calcium-induced conformational transition ofSPARC/BM-40/osteonectin, an extracellular glycoprotein exp-ressed in mineralized and nonmineralized tissues. Biochemistry1987;26:6958–6965.

75. Yoo SH, Albanesi JP. Ca2(þ)-induced conformational change andaggregation of chromogranin A. J Biol Chem 1990;265:14414–14421.

76. Neumann JM, Sanson A, Lewit-Bentley A. Calcium-induced changes in annexin V behaviour in solution as seenby proton NMR spectroscopy. Eur J Biochem 1994;225:819–825.

77. Lewit-Bentley A, Bentley GA, Favier B, L’Hermite G, RenouardM. The interaction of metal ions with annexin V: a crystallo-graphic study. FEBS Lett 1994;345:38–42.

78. Sopkova J, Gallay J, Vincent M, Pancoska P, Lewit-Bentley A.The dynamic behavior of annexin V as a function of calcium ionbinding: a circular dichroism, UV absorption, and steady-stateand time-resolved fluorescence study. Biochemistry 1994;33:4490–4499.

79. Sopkova J, Renouard M, Lewit-Bentley A. The crystal structureof a new high-calcium form of annexin V. J Mol Biol 1993;234:816–825.

80. Yamada T, Komoto J, Takata Y, Ogawa H, Pitot HC, Takusa-gawa F. Crystal structure of serine dehydratase from rat liver.Biochemistry 2003;42:12854–12865.

81. Eyal E, Gerzon S, Potapov V, Edelman M, Sobolev V. The limitof accuracy of protein modeling: influence of crystal packing onprotein structure. J Mol Biol 2005;351:431–442.



Dynamic α-helices: Conformations that do not conform

Documents

Transcript of Dynamic α-helices: Conformations that do not conform