International Journal of Biological Macromolecules 63 (2014) 119–125
Contents lists available at ScienceDirect
International Journal of Biological Macromolecules journal homepage: www.elsevier.com/locate/ijbiomac
Investigations of Ramachandran disallowed conformations in protein domain families B. Lakshmi a,b , C. Ramakrishnan b , G. Archunan a , R. Sowdhamini c , N. Srinivasan b,∗ a b c
Department of Animal Science, Bharathidasan University, Tiruchirappalli 620024, India Molecular Biophysics Unit, Indian Institute of Science, Bangalore 560012, India National Centre for Biological Sciences, TIFR, GKVK Campus, Bellary Road, Bangalore 560065, India
a r t i c l e
i n f o
Article history: Received 4 May 2013 Received in revised form 13 August 2013 Accepted 24 October 2013 Available online 31 October 2013 Keywords: Disallowed conformations Protein evolution Protein structures Ramachandran map Steric clash
a b s t r a c t In peptide and protein structures, occurrence of (, ) angles in the disallowed region of the Ramachandran map almost always suggests local regions of error or poor accuracy. However, very rarely genuine disallowed conformations occur as noted in the current study in proteins of known structure available at ˚ In the current work, extent of conservation of genuine disallowed conforultra-high resolution (≤1.2 A). mations in homologous proteins of known structures has been analyzed. From a dataset of 124 protein domain families, with structure of at least one constituent member in each family available at a resolution of 1.2 A˚ or better, we have analyzed the conservation of 221 disallowed conformations. It is observed that the disallowed conformation is only moderately conserved in protein domain families. In the gross dataset no particular residue type adopting disallowed conformation elicit high conservation of residue type though there are alignment positions in the dataset with complete conservation of both the residue type and the disallowed conformation. Conserved disallowed conformation in protein domain families play biologically significant role in roughly 50% of the cases. The residues with the disallowed conformation or its flanking residues are often located within or around the functional site of the protein. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The Ramachandran map [1–3] provides the fundamental understanding on the stereochemistry of the polypeptide backbone. Over 50 years ago Ramachandran (, ) map was derived using the model of two-linked peptide units with l-Ala/Gly in the middle and the contact criteria which defines the limits for the distance of shortest approach between two non-bonded atoms. In twolinked peptide units if the inter-atomic distances between all the non-bonded atoms are higher than the shortest possible distance between the corresponding atoms then that pair of , angles is considered “allowed”; however if the distance between any two non-bonded atoms is less than the limit then that , pair is considered “disallowed” [3]. For several decades now Ramachandran map is routinely used as a powerful tool to validate the stereo chemical quality of the protein structures solved using various experimental techniques or modeled using computational techniques [4,5]. Occurrence of , values in disallowed region of the Ramachandran map raises doubt on the stereochemical
∗ Corresponding author. Tel.: +91 80 2293 2837; fax: +91 80 2360 0535. E-mail addresses: [email protected], [email protected] (N. Srinivasan). 0141-8130/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ijbiomac.2013.10.032
quality of the structure [6]. Usually, further refinement of structure is necessary in such locations of potential inaccuracy [7]. However, genuine disallowed conformations do occur very rarely in protein structures and are evident from the ultra-high resolution crystallographic structures [8]. Ramakrishnan et al. [8] have analyzed a set of ultra high resolution crystal structures of proteins and peptides. They have concluded that such genuine disallowed conformations are tolerated with the expected steric clashes relieved by gentle and acceptable deviations of bond lengths and bond angles, from ideal values, in the local region. Gunasekaran et al. [9] showed that the unusual stereo chemical conformations are retained among independently derived protein structures i.e. the same protein solved by independent crystallographic analyses. Further, the genuine disallowed conformations with its strained structures are believed to be involved in possible functional roles [10,11]. If genuine disallowed conformation is essential for the integrity of structure or function then one would expect it to be conserved in the family of homologous proteins. In the present work, using a large dataset of aligned 3-D structures of homologous proteins, we have addressed the question “Are these genuine disallowed conformations conserved in protein domain families?” and if so what would be the extent and the significance of conservation of disallowed conformation among the homologues.
120
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
2. Materials and methods 2.1. Dataset The database of Phylogeny and ALIgnment of homologues protein structures (PALI version 2.8a) [12], which was built using the database of Structural Classification of Proteins (SCOP version 1.75) [13], consists of 2058 protein domain families comprising of over 17,000 protein domains. For the current analysis clearly orphan (families with just one member of known structure) families are unsuitable and therefore are not considered. Further we chose only those multi-member protein families with structure of at least one ˚ In of the members solved at a crystallographic resolution of ≤1.2 A. case of more than one ultra high resolution structure available in a family, the member with better resolution is chosen as the reference. This resulted in 1655 protein domain families. The , values were calculated, initially for ultra high resolution structures in every family. The , values at various residues are automatically categorized using the in-house software “rmapchk” into “Allowed”, “Disallowed” and “Marginally allowed” regions of the Ramachandran map. The “rmapchk” program is based on the geometrical procedure in which the closest point in the boundary representing the Ramachandran allowed region to the point representing (, ) values of interest is determined from the Ramachandran diagram. It is determined if the extension of the line joining these two points, but in the opposite direction, intersects on another point within the same allowed region of the Ramachandran map. If such point exists then the given (, ) value lie in one of the allowed regions of the Ramachandran map [8]. All the marginally allowed conformations are considered as allowed conformations and so the conformations are finally sorted into two groups as “allowed” and “disallowed” [8]. This resulted in 323 , disallowed conformations from ultra-high resolution structures in 138 protein domain families. The ultra-high resolution structure in every family is considered as the representative structure of the family. It was ensured that in each family used in the current analysis at least one structure is present in ultra high resolution of ≤1.2 A˚ with at least one disallowed conformation in it and rest of the members with structures ˚ determined at a resolution of ≤2 A. In addition to using ultra high resolution structures to identify potentially genuine disallowed Ramachandran angles, we have also considered temperature (B) factors at the positions of disallowed conformation. The temperature factor or B-factor gives an indication of uncertainty associated with atomic positions in the crystalline state. The B-factors taken from PDB may be on different scales owing to the application of different refinement procedures [14]. To measure the uncertainty, the B-factors of backbone atoms N, C␣ , C and O where normalized using a previously proposed procedure [15,16]. Finally, from the ultra high resolution structures with at least one disallowed conformation with a normalized Bfactor ≤2 resulted in 221 , values which are considered to be genuinely disallowed Ramachandran angles pertaining to 124 protein domain families. The details of normalization and rigorous statistical analysis of B-factors observed in high resolution protein structures, which forms the basis for the choice of cut-off for the normalized B-factor, are provided in Ref. [15]. The real space correlation coefficient is the measure of the similarity between an electron density map calculated directly from a structure model and one calculated from experimental data. In addition to the B-factor the real space correlation coefficient is also calculated for the possible cases of 83 (, ) values in the disallowed conformation using the Uppsala electron density server (EDS) [17]. Structure factor information for the protein structures housing the rest of the (, ) values are not available in the PDB and therefore the real space correlation coefficient could not be calculated. The real space correlation coefficient calculated for 83 (, ) values is
Fig. 1. The plot of real space correlation coefficient and the frequency of residues in the disallowed conformations.
shown in Fig. 1 where it is clearly shown that the residues with disallowed conformation show excellent correlation of greater than 0.8. All 124 protein domain families were considered for pairwise structural comparison by involving the ultra high resolution structure in each pair. A subset of 56 protein domain families with at least 7 members in each family with at least one of the members in the family with structure determined at ultra-high resolution and the rest of the structures in the family are available at a resolution of 2 A˚ or better are considered for multiple structure comparison. The structural alignments are performed using DALI [18] for pairwise comparison and MUSTANG [19] for multiple comparison. Classification of data into various categories using pairwise comparisons is shown in Fig. 2 which would be explained in the later sections. 3. Results and discussion 3.1. General distribution of disallowed conformations in ultra-high resolution structures The 221 disallowed , values at non-Gly residues in the ultra-high resolution structures in our dataset are plotted in the Ramachandran map (Fig. 3). The details of 221 disallowed , values are given in Supplementary Table S1. From Fig. 2 it can be noticed that disallowed conformations are not sparsely distributed all along the disallowed regions. Interestingly clusters of disallowed conformations could be noticed especially in the right bottom quadrant of the , plane. Reasons for such a clustering have been discussed by Gunasekaran et al. [20]. The homologues proteins domain structures are superimposed (both pairwise and multiple) and the conservation of disallowed conformation is analyzed with respect to the representative (ultra-high resolution) structure in every protein family in our dataset. Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j. ijbiomac.2013.10.032. 3.2. Conservation of residue types and conservation of disallowed conformations In the present work, the conservation of disallowed conformations in the homologues of protein domain families as well as conservation of residue types that adopt disallowed conformation has been analyzed.
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
121
Fig. 2. Distribution of disallowed conformations in the dataset from the pairwise structural alignments.
Fig. 3. Ramachandran plot for 221 non-Gly residues in disallowed conformations from protein domains of known structures determined at ultra-high resolution ˚ (≤1.2 A).
For pairwise structural comparison, in each pair at least one ultra high resolution structure with at least one disallowed conformation for a non-Gly residue is considered in 124 protein domain families. From 221 disallowed phi, psi values in 124 families there are 1687 pairwise alignment positions considered for further analysis (Fig. 2). A subset of 56 families with at least 7 members with at least one of them corresponding to an ultra high resolution structure was considered for multiple structural comparisons. From the structural alignments, the alignment position of the residue with the disallowed conformation in ultra-high resolution structure is compared with the homologous domains. At each of these alignment positions the conservation of residues with disallowed conformation and its disallowed nature is analyzed. Conservation of residue types is classified as, (i) RES–RES (RES refers to a non-Gly residue with disallowed conformation): the residue with disallowed conformation is conserved and is topologically equivalent with its homologues
in the corresponding alignment position. RES–RES also indicates conservation of residue type. (ii) RES–X (X is any non-Gly residue other than RES): the residue with disallowed conformation is topologically equivalent with the residue X in its homologues in the alignment position but the residue types are not conserved. (iii) RES–G (G – Glycine): the residue with the disallowed conformation is topologically equivalent to Glycine in the alignment position with its homologues. (iv) RES–: the residue with disallowed conformation is deleted in the homologue according to the alignment. For 56 families the multiple structural alignments were generated and in each family the conservation of the residue type is analyzed at the position with disallowed conformation in the ultra-high resolution structure. As an example in Fig. 4a the multiple structural alignment of a family with 15 members in the structure with ultra high resolution is shown in red color and the position considered for conservation is shown in cyan. The phi, psi values of the homologues were also calculated and conformational multiple structural alignments were generated for each of the family as shown in Fig. 4b. In this alignment, each residue is represented based on its conformational state if the residue is allowed (A) or disallowed (D) according to the Ramachandran map or not-applicable (N) (mainly for Gly and terminal residues where it is not possible to calculate one of phi or psi). The conservation of residue type (Fig. 4a) and the disallowed nature (Fig. 4b) are shown in cyan color. There are 83 alignment positions from 56 families considered and at each of these positions the percentage conservation of residue types, disallowed nature and simultaneous conservation of residue type and disallowed nature were calculated and the results are shown in Fig. 5. In Fig. 5 it is clear that in more than 50% of alignments positions the residue types are not conserved and only in a few positions the residue with the disallowed conformation is conserved. But the conservation of disallowed nature at these alignment positions shows moderately better conservation compared to the conservation of residue types which is represented as red dots. Simultaneous conservation of both the residue type and the disallowed conformation, which is represented as green dots is generally poor though there are few alignment positions which show very high conservation. The significance of the conservation of these residues is explained in later sections. In the cases with residue type conserved in at least 50% of the members, the disallowed nature is usually well conserved. In pairwise comparisons, the homologues members were structurally aligned by considering two homologues at a time with one
122
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
Fig. 4. An example of multiple structural alignment along with annotation of allowed or disallowed conformation. (a) Multiple structural alignment for a protein domain family with the representative structure highlighted in red. The residue position marked in cyan is the residue in the disallowed conformation and the residues in the equivalent position in its homologues are shown. (b) Conformational alignment for same protein domain family. Here A indicates allowed conformation, D indicates disallowed conformation according to the Ramachandran map and N refers to “not applicable” which corresponds to Gly and terminal residues. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
of the two corresponding to the ultra-high resolution structure. Pairwise comparisons have been made for 124 families. Totally 984 pairwise structural alignments were generated involving 1687 alignment positions (Fig. 2) involving 221 residues in disallowed conformation observed in ultra-high resolution structures. In these 1687 alignment positions the residue with the disallowed conformation is compared with the equivalent alignment position in its homologue and classified into one of the residue type conservation as mentioned earlier as RES–RES, RES–X etc. The conservation of disallowed conformation and its disallowed nature for each of the 20 amino acid types are given in detail in Table 1. It is clear from the table that either the residue types or the disallowed conformation is not very well conserved in almost all the cases although the residue in the alignment position corresponding to the residue with disallowed conformation is topologically equivalent. i.e., though the residues are topologically equivalent the disallowed nature is also not very well conserved. Ala, Lys, Asp show moderate
conservation in residue type and the disallowed nature whereas in case of Cys both the residue and disallowed nature conservation is high though the data size for Cys is low. None of the Cys in disallowed conformation is involved in disulphide bond formation. For Met, Tyr and Trp the conservation of residue type is not very high but when these residue types are conserved the disallowed nature is also highly conserved. In case of Asn, Ser and Ala the residue in the disallowed conformation is often replaced with Gly in its homologues but in case of Ser and Ala at many positions the residues are topologically equivalent and its disallowed nature is also moderately conserved. The percentage conservation of disallowed conformation of each of the 20 amino acid types is shown in Fig. 6. The disallowed conformation is highly conserved in Cys, Lys and Met, but in the rest of the cases it is only moderately conserved. So, out of 1687 pairwise alignment positions considered, only in 298 positions the disallowed nature is conserved while in slightly more than 50% of positions the residue type is conserved
Table 1 Extent of conservation of residue types and disallowed conformations. Amino acids
Number of residues with disallowed conformation
Number of alignments positions
A C D E F H I K L M N P Q R S T V W Y
29 14 28 9 14 6 3 21 15 6 39 3 12 16 37 21 10 9 12
202 57 147 41 79 30 17 83 52 77 255 31 54 63 164 123 72 68 72
Conservation of disallowed conformation
Conservation of residue types
RES–RES
RES/res–
RES–X (X#RES/G)
RES–G
RES–RES
RES–X (X#RES)
26 11 45 3 6 0 7 24 7 1 27 2 10 3 27 17 4 18 17
26 16 26 12 31 3 0 18 16 16 59 5 24 10 24 22 11 10 14
95 4 11 4 8 12 4 8 3 27 42 2 5 11 57 22 13 20 27
25 9 4 1 0 0 0 10 2 5 36 0 1 6 24 8 7 8 0
18 10 20 3 3 0 2 15 6 0 20 0 4 3 19 11 3 14 13
24 3 3 1 2 1 0 6 0 23 11 0 2 3 12 11 1 4 0
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
123
Fig. 7. Frequencies of conservation of disallowed conformation nature and residue types from pairwise alignments.
Fig. 5. Percentage conservation of disallowed conformation and residue types in multiple structural alignments. The red dots correspond to percentage conservation of residue type and the disallowed conformation whereas the green dots correspond to percentage conservation of both residue type and the disallowed conformation.
(Fig. 2). It is also evident from Fig. 7 that in 164 alignment positions both the residue type and the disallowed nature are conserved. However in 107 positions only the disallowed nature is considered, but in many positions neither of them is conserved. Overall conservation of residue positions with disallowed conformation in ultra high resolution structures, either as residue types or disallowed conformation is not high among the homologues. 3.3. Conservation of both residue type and disallowed conformations We addressed the question “If the type of the residue with disallowed conformation is completely conserved in the homologues of known 3-D structure, is the disallowed conformational nature also conserved? Table 2 lists the families from our data set in which the type of the residue with disallowed conformation
in ultra-high resolution structure is completely conserved. Fig. 8 shows the Ramachandran plot of these conserved residues from these domain families. In Family 1 of glycosyl hydrolase family (c.1.8.4) with 8 homologues members (Fig. 8a), Trp in disallowed conformation is conserved in all its homologues along with its disallowed conformation. Lys from Triosephosphate isomerase family (c.1.1.1) (Fig. 8b) is conserved in all its 9 homologues both in terms of residue type and disallowed conformation. Asp from Subtilases (c.41.1.1) family (Fig. 8c) is conserved in all the 12 homologues. However the disallowed nature of the conformation is conserved only in 11 out of 12 homologues, It can be seen in Fig. 8c that in one case (phi,psi) values at Asp is located just inside the allowed regions of the Ramachandran map. Overall, it is evident that if the residue type with disallowed conformation is conserved, the disallowed nature is also almost completely conserved.
3.4. Functional significance of disallowed conformations in protein domain families It is generally believed that genuine disallowed conformations have some important role to play in protein function. It is also evident from the earlier studies that Ramachandran disallowed conformations are generally believed to have structural and also possible functional roles [10]. Gunasekaran et al. [9] suggested that the strained backbone conformations may have a special significance in the active sties of the proteins. Osnat Herzberg et al. [11], by analyzing the rare significant steric strains in protein structures, noted that those conformations occur in the functional regions of proteins.
Fig. 6. Percentage conservation of disallowed conformations in 20 residue types in pairwise structural alignments.
124
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
Table 2 List of protein domain families in which the type of the residue with disallowed conformation in ultra-high resolution structure is completely conserved. S.no
Protein domain family name
1
Family 1 of glycosyl hydrolase (c.1.8.4) Triosephosphate isomerase (c.1.1.1) Subtilases (c.41.1.1)
2 3
Number of members in the protein domain family
SCOP domain code
Residue in the disallowed conformation
Number of members in which disallowed conformation is conserved
8
1e4mm
W140
8
9
1n55a
K12
9
12
1ea7a
D34
11
Fig. 8. Ramachandran plot for the non-Gly residues in disallowed conformation in the members of protein domain families with both residue type and the disallowed conformation conserved. (a) Family 1 of glycosyl hydrolase family (c.1.8.4). (b) Triosephosphate isomerase family (c.1.1.1). (c) Subtilases family (c.41.1.1).
If the disallowed conformation is conserved in a family one might expect the residue to be involved in the function. Therefore, in the current study, residues in genuine disallowed conformation in homologous protein structures which shows conservation of its disallowed nature are further considered for the analysis on its functional significance. For this purpose 298 alignment positions with disallowed conformation nature conserved from the dataset of pairwise alignments were considered. The three flanking residues on either side of the residue with disallowed conformation have been considered for their potential participation in function. All these seven residues were further analyzed to investigate if any of these seven residues is/are involved in some functional role. In order to identify if these residues are involved in function we used pdbSUM [21]. It is noted that in 149 out of 298 alignment positions (50%) with disallowed nature conserved one or more residues in the seven residue stretch are involved in function of the protein. For example in Human Rab5a GTPase domain (PDB code: 1R2Q)
(Fig. 9b) the Ala30 is in disallowed conformation and its flanking residues are interacting with the ligand Phosphoaminophosphonic acid-guanylate ester. In 3,4-dihydroxy-2-butanone 4-phosphate synthase (PDB code: 1K4I) (Fig. 9a) Arg 36 is in disallowed conformation and the residues flanking interacting with magnesium [22]. In many cases these sterically strained regions are involved in metal or ligand interaction through hydrophobic or hydrogen bond interactions. Also it is evident from the flowchart (Fig. 2) that in many cases the disallowed conformation is conserved and also having functional significance. There are also substantial number of cases with conservation of disallowed conformation but do not show any functional significance. Reason for conservation of disallowed conformation in these cases is currently obscure. Out of 221 disallowed phi, psi values considered for analysis 116 phi, psi values are involved in at least one pairwise alignment with disallowed conformation conserved. Further when
Fig. 9. Examples of disallowed conformation showing functional significance. The residues with disallowed conformation and its flanking residues are shown in red and cyan color respectively. The ligand is shown in pink and orange color stick models, metals as a small spheres and catalytic residues are marked. (a) 3,4-Dihydroxy-2-Butanone 4-Phosphate Synthase (PDB code: 1K4I) the flanking residues interact with the metal. (b) Human Rab5a GTPase Domain (PDB code: 1R2Q) the residue in the disallowed conformation is Ala30 and its flanking residues interact with the ligand phosphoaminophosphonic acid–guanylate ester (GNP).
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
these cases were subjected to the analysis of functional significance in roughly half the number of cases the disallowed region is in the functional environment. Almost in all of these cases the residues in disallowed conformation are mentioned in “PDBSITE” record according to Protein Data Bank entry which means that these residues are in the functional environment in the protein. The details of the functional significance are provided in the supplementary data (Supplementary Table S2). In many of the examples analyzed the residue in disallowed conformation and/or its flanking residues are involved in function. In many cases the residues involved are found to be in beta hairpins. Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijbiomac. 2013.10.032. 4. Conclusions The present analysis illustrated that the Ramachandran disallowed conformations do occur in ultra-high resolution protein structures in locations with reasonably low temperature factor. Such genuine disallowed conformations are only moderately conserved among the homologues in protein domain families. The extent of conservation of disallowed nature is higher when the disallowed residue type is also conserved compared to when residue type is not conserved. It is also clear that none of the residue types are highly preferred for the conservation of disallowed conformation. The analysis on the functional significance of the disallowed conformations suggests that in roughly half the number of cases, with disallowed conformation conserved, the residue in the disallowed conformation or its flanking residues are associated with function of the protein.
125
Acknowledgements This work is supported by Council of Scientific and Industrial Research (CSIR), DBT-AIST (Indo-Japan collaborative project) and by DBT, Government of India. References [1] G.N. Ramachandran, C. Ramakrishnan, V. Sasisekharan, J. Mol. Biol. 7 (1963) 95–99. [2] C. Ramakrishnan, G.N. Ramachandran, Biophys. J. 5 (1965) 909–933. [3] G.N. Ramachandran, V. Sasisekharan, Adv. Protein Chem. 23 (1968) 283–438. [4] T.A. Jones, J. Appl. Crystallogr. 11 (1978) 268–272. [5] A.L. Morris, M.W. MacArthor, E.G. Hutchinson, J.M. Thornton, Proteins: Struct Funct Genet 12 (1992) 345–364. [6] G.J. Kleywegt, Acta Crystallogr. D Biol. Crystallogr. 56 (2000) 249–265. [7] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thornton, J. Appl. Cryst. 26 (1993) 283–291. [8] C. Ramakrishnan, B. Lakshmi, A. Kurien, D. Devipriya, N. Srinivasan, Protein Pept. Lett. 14 (2007) 672–682. [9] K. Gunasekaran, C. Ramakrishnan, P. Balaram, J. Mol. Biol. 264 (1996) 191–198. [10] Z. Jia, M. Vandonselaar, J.W. Quail, L.T. Delbaere, Nature 361 (1993) 94–97. [11] O. Herzberg, J. Moult, Proteins 11 (1991) 223–229. [12] S. Balaji, S. Sujatha, S.S. Kumar, N. Srinivasan, Nucleic Acids Res. 29 (2001) 61–65. [13] A.G. Murzin, S.E. Brenner, T. Hubbard, C. Chothia, J. Mol. Biol. 247 (1995) 536–540. [14] D.E. Tronrud, J. Appl. Crystallogr. 29 (1996) 100–104. [15] S. Parthasarathy, M.R. Murthy, Protein Sci. 6 (1997) 2561–2567. [16] Z. Yuan, J. Zhao, Z.X. Wang, Protein Eng. 16 (2003) 109–114. [17] G.J. Kleywegt, M.R. Harris, J.Y. Zou, T.C. Taylor, A. Wahlby, T.A. Jones, Acta Crystallogr. D Biol. Crystallogr. 60 (2004) 2240–2249. [18] L. Holm, P. Rosenstrom, Nucleic Acids Res 38 (2010) W545–W549. [19] A.S. Konagurthu, J.C. Whisstock, P.J. Stuckey, A.M. Lesk, Proteins 64 (2006) 559–574. [20] K. Gunasekaran, R. Nussinov, J. Mol. Biol. 365 (2007) 257–273. [21] R.A. Laskowski, E.G. Hutchinson, A.D. Michie, A.C. Wallace, M.L. Jones, J.M. Thornton, Trends Biochem. Sci. 22 (1997) 488–490. [22] D.I. Liao, Y.J. Zheng, P.V. Viitanen, D.B. Jordan, Biochemistry 41 (2002) 1795–1806.
Contents lists available at ScienceDirect
International Journal of Biological Macromolecules journal homepage: www.elsevier.com/locate/ijbiomac
Investigations of Ramachandran disallowed conformations in protein domain families B. Lakshmi a,b , C. Ramakrishnan b , G. Archunan a , R. Sowdhamini c , N. Srinivasan b,∗ a b c
Department of Animal Science, Bharathidasan University, Tiruchirappalli 620024, India Molecular Biophysics Unit, Indian Institute of Science, Bangalore 560012, India National Centre for Biological Sciences, TIFR, GKVK Campus, Bellary Road, Bangalore 560065, India
a r t i c l e
i n f o
Article history: Received 4 May 2013 Received in revised form 13 August 2013 Accepted 24 October 2013 Available online 31 October 2013 Keywords: Disallowed conformations Protein evolution Protein structures Ramachandran map Steric clash
a b s t r a c t In peptide and protein structures, occurrence of (, ) angles in the disallowed region of the Ramachandran map almost always suggests local regions of error or poor accuracy. However, very rarely genuine disallowed conformations occur as noted in the current study in proteins of known structure available at ˚ In the current work, extent of conservation of genuine disallowed conforultra-high resolution (≤1.2 A). mations in homologous proteins of known structures has been analyzed. From a dataset of 124 protein domain families, with structure of at least one constituent member in each family available at a resolution of 1.2 A˚ or better, we have analyzed the conservation of 221 disallowed conformations. It is observed that the disallowed conformation is only moderately conserved in protein domain families. In the gross dataset no particular residue type adopting disallowed conformation elicit high conservation of residue type though there are alignment positions in the dataset with complete conservation of both the residue type and the disallowed conformation. Conserved disallowed conformation in protein domain families play biologically significant role in roughly 50% of the cases. The residues with the disallowed conformation or its flanking residues are often located within or around the functional site of the protein. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The Ramachandran map [1–3] provides the fundamental understanding on the stereochemistry of the polypeptide backbone. Over 50 years ago Ramachandran (, ) map was derived using the model of two-linked peptide units with l-Ala/Gly in the middle and the contact criteria which defines the limits for the distance of shortest approach between two non-bonded atoms. In twolinked peptide units if the inter-atomic distances between all the non-bonded atoms are higher than the shortest possible distance between the corresponding atoms then that pair of , angles is considered “allowed”; however if the distance between any two non-bonded atoms is less than the limit then that , pair is considered “disallowed” [3]. For several decades now Ramachandran map is routinely used as a powerful tool to validate the stereo chemical quality of the protein structures solved using various experimental techniques or modeled using computational techniques [4,5]. Occurrence of , values in disallowed region of the Ramachandran map raises doubt on the stereochemical
∗ Corresponding author. Tel.: +91 80 2293 2837; fax: +91 80 2360 0535. E-mail addresses: [email protected], [email protected] (N. Srinivasan). 0141-8130/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ijbiomac.2013.10.032
quality of the structure [6]. Usually, further refinement of structure is necessary in such locations of potential inaccuracy [7]. However, genuine disallowed conformations do occur very rarely in protein structures and are evident from the ultra-high resolution crystallographic structures [8]. Ramakrishnan et al. [8] have analyzed a set of ultra high resolution crystal structures of proteins and peptides. They have concluded that such genuine disallowed conformations are tolerated with the expected steric clashes relieved by gentle and acceptable deviations of bond lengths and bond angles, from ideal values, in the local region. Gunasekaran et al. [9] showed that the unusual stereo chemical conformations are retained among independently derived protein structures i.e. the same protein solved by independent crystallographic analyses. Further, the genuine disallowed conformations with its strained structures are believed to be involved in possible functional roles [10,11]. If genuine disallowed conformation is essential for the integrity of structure or function then one would expect it to be conserved in the family of homologous proteins. In the present work, using a large dataset of aligned 3-D structures of homologous proteins, we have addressed the question “Are these genuine disallowed conformations conserved in protein domain families?” and if so what would be the extent and the significance of conservation of disallowed conformation among the homologues.
120
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
2. Materials and methods 2.1. Dataset The database of Phylogeny and ALIgnment of homologues protein structures (PALI version 2.8a) [12], which was built using the database of Structural Classification of Proteins (SCOP version 1.75) [13], consists of 2058 protein domain families comprising of over 17,000 protein domains. For the current analysis clearly orphan (families with just one member of known structure) families are unsuitable and therefore are not considered. Further we chose only those multi-member protein families with structure of at least one ˚ In of the members solved at a crystallographic resolution of ≤1.2 A. case of more than one ultra high resolution structure available in a family, the member with better resolution is chosen as the reference. This resulted in 1655 protein domain families. The , values were calculated, initially for ultra high resolution structures in every family. The , values at various residues are automatically categorized using the in-house software “rmapchk” into “Allowed”, “Disallowed” and “Marginally allowed” regions of the Ramachandran map. The “rmapchk” program is based on the geometrical procedure in which the closest point in the boundary representing the Ramachandran allowed region to the point representing (, ) values of interest is determined from the Ramachandran diagram. It is determined if the extension of the line joining these two points, but in the opposite direction, intersects on another point within the same allowed region of the Ramachandran map. If such point exists then the given (, ) value lie in one of the allowed regions of the Ramachandran map [8]. All the marginally allowed conformations are considered as allowed conformations and so the conformations are finally sorted into two groups as “allowed” and “disallowed” [8]. This resulted in 323 , disallowed conformations from ultra-high resolution structures in 138 protein domain families. The ultra-high resolution structure in every family is considered as the representative structure of the family. It was ensured that in each family used in the current analysis at least one structure is present in ultra high resolution of ≤1.2 A˚ with at least one disallowed conformation in it and rest of the members with structures ˚ determined at a resolution of ≤2 A. In addition to using ultra high resolution structures to identify potentially genuine disallowed Ramachandran angles, we have also considered temperature (B) factors at the positions of disallowed conformation. The temperature factor or B-factor gives an indication of uncertainty associated with atomic positions in the crystalline state. The B-factors taken from PDB may be on different scales owing to the application of different refinement procedures [14]. To measure the uncertainty, the B-factors of backbone atoms N, C␣ , C and O where normalized using a previously proposed procedure [15,16]. Finally, from the ultra high resolution structures with at least one disallowed conformation with a normalized Bfactor ≤2 resulted in 221 , values which are considered to be genuinely disallowed Ramachandran angles pertaining to 124 protein domain families. The details of normalization and rigorous statistical analysis of B-factors observed in high resolution protein structures, which forms the basis for the choice of cut-off for the normalized B-factor, are provided in Ref. [15]. The real space correlation coefficient is the measure of the similarity between an electron density map calculated directly from a structure model and one calculated from experimental data. In addition to the B-factor the real space correlation coefficient is also calculated for the possible cases of 83 (, ) values in the disallowed conformation using the Uppsala electron density server (EDS) [17]. Structure factor information for the protein structures housing the rest of the (, ) values are not available in the PDB and therefore the real space correlation coefficient could not be calculated. The real space correlation coefficient calculated for 83 (, ) values is
Fig. 1. The plot of real space correlation coefficient and the frequency of residues in the disallowed conformations.
shown in Fig. 1 where it is clearly shown that the residues with disallowed conformation show excellent correlation of greater than 0.8. All 124 protein domain families were considered for pairwise structural comparison by involving the ultra high resolution structure in each pair. A subset of 56 protein domain families with at least 7 members in each family with at least one of the members in the family with structure determined at ultra-high resolution and the rest of the structures in the family are available at a resolution of 2 A˚ or better are considered for multiple structure comparison. The structural alignments are performed using DALI [18] for pairwise comparison and MUSTANG [19] for multiple comparison. Classification of data into various categories using pairwise comparisons is shown in Fig. 2 which would be explained in the later sections. 3. Results and discussion 3.1. General distribution of disallowed conformations in ultra-high resolution structures The 221 disallowed , values at non-Gly residues in the ultra-high resolution structures in our dataset are plotted in the Ramachandran map (Fig. 3). The details of 221 disallowed , values are given in Supplementary Table S1. From Fig. 2 it can be noticed that disallowed conformations are not sparsely distributed all along the disallowed regions. Interestingly clusters of disallowed conformations could be noticed especially in the right bottom quadrant of the , plane. Reasons for such a clustering have been discussed by Gunasekaran et al. [20]. The homologues proteins domain structures are superimposed (both pairwise and multiple) and the conservation of disallowed conformation is analyzed with respect to the representative (ultra-high resolution) structure in every protein family in our dataset. Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j. ijbiomac.2013.10.032. 3.2. Conservation of residue types and conservation of disallowed conformations In the present work, the conservation of disallowed conformations in the homologues of protein domain families as well as conservation of residue types that adopt disallowed conformation has been analyzed.
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
121
Fig. 2. Distribution of disallowed conformations in the dataset from the pairwise structural alignments.
Fig. 3. Ramachandran plot for 221 non-Gly residues in disallowed conformations from protein domains of known structures determined at ultra-high resolution ˚ (≤1.2 A).
For pairwise structural comparison, in each pair at least one ultra high resolution structure with at least one disallowed conformation for a non-Gly residue is considered in 124 protein domain families. From 221 disallowed phi, psi values in 124 families there are 1687 pairwise alignment positions considered for further analysis (Fig. 2). A subset of 56 families with at least 7 members with at least one of them corresponding to an ultra high resolution structure was considered for multiple structural comparisons. From the structural alignments, the alignment position of the residue with the disallowed conformation in ultra-high resolution structure is compared with the homologous domains. At each of these alignment positions the conservation of residues with disallowed conformation and its disallowed nature is analyzed. Conservation of residue types is classified as, (i) RES–RES (RES refers to a non-Gly residue with disallowed conformation): the residue with disallowed conformation is conserved and is topologically equivalent with its homologues
in the corresponding alignment position. RES–RES also indicates conservation of residue type. (ii) RES–X (X is any non-Gly residue other than RES): the residue with disallowed conformation is topologically equivalent with the residue X in its homologues in the alignment position but the residue types are not conserved. (iii) RES–G (G – Glycine): the residue with the disallowed conformation is topologically equivalent to Glycine in the alignment position with its homologues. (iv) RES–: the residue with disallowed conformation is deleted in the homologue according to the alignment. For 56 families the multiple structural alignments were generated and in each family the conservation of the residue type is analyzed at the position with disallowed conformation in the ultra-high resolution structure. As an example in Fig. 4a the multiple structural alignment of a family with 15 members in the structure with ultra high resolution is shown in red color and the position considered for conservation is shown in cyan. The phi, psi values of the homologues were also calculated and conformational multiple structural alignments were generated for each of the family as shown in Fig. 4b. In this alignment, each residue is represented based on its conformational state if the residue is allowed (A) or disallowed (D) according to the Ramachandran map or not-applicable (N) (mainly for Gly and terminal residues where it is not possible to calculate one of phi or psi). The conservation of residue type (Fig. 4a) and the disallowed nature (Fig. 4b) are shown in cyan color. There are 83 alignment positions from 56 families considered and at each of these positions the percentage conservation of residue types, disallowed nature and simultaneous conservation of residue type and disallowed nature were calculated and the results are shown in Fig. 5. In Fig. 5 it is clear that in more than 50% of alignments positions the residue types are not conserved and only in a few positions the residue with the disallowed conformation is conserved. But the conservation of disallowed nature at these alignment positions shows moderately better conservation compared to the conservation of residue types which is represented as red dots. Simultaneous conservation of both the residue type and the disallowed conformation, which is represented as green dots is generally poor though there are few alignment positions which show very high conservation. The significance of the conservation of these residues is explained in later sections. In the cases with residue type conserved in at least 50% of the members, the disallowed nature is usually well conserved. In pairwise comparisons, the homologues members were structurally aligned by considering two homologues at a time with one
122
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
Fig. 4. An example of multiple structural alignment along with annotation of allowed or disallowed conformation. (a) Multiple structural alignment for a protein domain family with the representative structure highlighted in red. The residue position marked in cyan is the residue in the disallowed conformation and the residues in the equivalent position in its homologues are shown. (b) Conformational alignment for same protein domain family. Here A indicates allowed conformation, D indicates disallowed conformation according to the Ramachandran map and N refers to “not applicable” which corresponds to Gly and terminal residues. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
of the two corresponding to the ultra-high resolution structure. Pairwise comparisons have been made for 124 families. Totally 984 pairwise structural alignments were generated involving 1687 alignment positions (Fig. 2) involving 221 residues in disallowed conformation observed in ultra-high resolution structures. In these 1687 alignment positions the residue with the disallowed conformation is compared with the equivalent alignment position in its homologue and classified into one of the residue type conservation as mentioned earlier as RES–RES, RES–X etc. The conservation of disallowed conformation and its disallowed nature for each of the 20 amino acid types are given in detail in Table 1. It is clear from the table that either the residue types or the disallowed conformation is not very well conserved in almost all the cases although the residue in the alignment position corresponding to the residue with disallowed conformation is topologically equivalent. i.e., though the residues are topologically equivalent the disallowed nature is also not very well conserved. Ala, Lys, Asp show moderate
conservation in residue type and the disallowed nature whereas in case of Cys both the residue and disallowed nature conservation is high though the data size for Cys is low. None of the Cys in disallowed conformation is involved in disulphide bond formation. For Met, Tyr and Trp the conservation of residue type is not very high but when these residue types are conserved the disallowed nature is also highly conserved. In case of Asn, Ser and Ala the residue in the disallowed conformation is often replaced with Gly in its homologues but in case of Ser and Ala at many positions the residues are topologically equivalent and its disallowed nature is also moderately conserved. The percentage conservation of disallowed conformation of each of the 20 amino acid types is shown in Fig. 6. The disallowed conformation is highly conserved in Cys, Lys and Met, but in the rest of the cases it is only moderately conserved. So, out of 1687 pairwise alignment positions considered, only in 298 positions the disallowed nature is conserved while in slightly more than 50% of positions the residue type is conserved
Table 1 Extent of conservation of residue types and disallowed conformations. Amino acids
Number of residues with disallowed conformation
Number of alignments positions
A C D E F H I K L M N P Q R S T V W Y
29 14 28 9 14 6 3 21 15 6 39 3 12 16 37 21 10 9 12
202 57 147 41 79 30 17 83 52 77 255 31 54 63 164 123 72 68 72
Conservation of disallowed conformation
Conservation of residue types
RES–RES
RES/res–
RES–X (X#RES/G)
RES–G
RES–RES
RES–X (X#RES)
26 11 45 3 6 0 7 24 7 1 27 2 10 3 27 17 4 18 17
26 16 26 12 31 3 0 18 16 16 59 5 24 10 24 22 11 10 14
95 4 11 4 8 12 4 8 3 27 42 2 5 11 57 22 13 20 27
25 9 4 1 0 0 0 10 2 5 36 0 1 6 24 8 7 8 0
18 10 20 3 3 0 2 15 6 0 20 0 4 3 19 11 3 14 13
24 3 3 1 2 1 0 6 0 23 11 0 2 3 12 11 1 4 0
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
123
Fig. 7. Frequencies of conservation of disallowed conformation nature and residue types from pairwise alignments.
Fig. 5. Percentage conservation of disallowed conformation and residue types in multiple structural alignments. The red dots correspond to percentage conservation of residue type and the disallowed conformation whereas the green dots correspond to percentage conservation of both residue type and the disallowed conformation.
(Fig. 2). It is also evident from Fig. 7 that in 164 alignment positions both the residue type and the disallowed nature are conserved. However in 107 positions only the disallowed nature is considered, but in many positions neither of them is conserved. Overall conservation of residue positions with disallowed conformation in ultra high resolution structures, either as residue types or disallowed conformation is not high among the homologues. 3.3. Conservation of both residue type and disallowed conformations We addressed the question “If the type of the residue with disallowed conformation is completely conserved in the homologues of known 3-D structure, is the disallowed conformational nature also conserved? Table 2 lists the families from our data set in which the type of the residue with disallowed conformation
in ultra-high resolution structure is completely conserved. Fig. 8 shows the Ramachandran plot of these conserved residues from these domain families. In Family 1 of glycosyl hydrolase family (c.1.8.4) with 8 homologues members (Fig. 8a), Trp in disallowed conformation is conserved in all its homologues along with its disallowed conformation. Lys from Triosephosphate isomerase family (c.1.1.1) (Fig. 8b) is conserved in all its 9 homologues both in terms of residue type and disallowed conformation. Asp from Subtilases (c.41.1.1) family (Fig. 8c) is conserved in all the 12 homologues. However the disallowed nature of the conformation is conserved only in 11 out of 12 homologues, It can be seen in Fig. 8c that in one case (phi,psi) values at Asp is located just inside the allowed regions of the Ramachandran map. Overall, it is evident that if the residue type with disallowed conformation is conserved, the disallowed nature is also almost completely conserved.
3.4. Functional significance of disallowed conformations in protein domain families It is generally believed that genuine disallowed conformations have some important role to play in protein function. It is also evident from the earlier studies that Ramachandran disallowed conformations are generally believed to have structural and also possible functional roles [10]. Gunasekaran et al. [9] suggested that the strained backbone conformations may have a special significance in the active sties of the proteins. Osnat Herzberg et al. [11], by analyzing the rare significant steric strains in protein structures, noted that those conformations occur in the functional regions of proteins.
Fig. 6. Percentage conservation of disallowed conformations in 20 residue types in pairwise structural alignments.
124
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
Table 2 List of protein domain families in which the type of the residue with disallowed conformation in ultra-high resolution structure is completely conserved. S.no
Protein domain family name
1
Family 1 of glycosyl hydrolase (c.1.8.4) Triosephosphate isomerase (c.1.1.1) Subtilases (c.41.1.1)
2 3
Number of members in the protein domain family
SCOP domain code
Residue in the disallowed conformation
Number of members in which disallowed conformation is conserved
8
1e4mm
W140
8
9
1n55a
K12
9
12
1ea7a
D34
11
Fig. 8. Ramachandran plot for the non-Gly residues in disallowed conformation in the members of protein domain families with both residue type and the disallowed conformation conserved. (a) Family 1 of glycosyl hydrolase family (c.1.8.4). (b) Triosephosphate isomerase family (c.1.1.1). (c) Subtilases family (c.41.1.1).
If the disallowed conformation is conserved in a family one might expect the residue to be involved in the function. Therefore, in the current study, residues in genuine disallowed conformation in homologous protein structures which shows conservation of its disallowed nature are further considered for the analysis on its functional significance. For this purpose 298 alignment positions with disallowed conformation nature conserved from the dataset of pairwise alignments were considered. The three flanking residues on either side of the residue with disallowed conformation have been considered for their potential participation in function. All these seven residues were further analyzed to investigate if any of these seven residues is/are involved in some functional role. In order to identify if these residues are involved in function we used pdbSUM [21]. It is noted that in 149 out of 298 alignment positions (50%) with disallowed nature conserved one or more residues in the seven residue stretch are involved in function of the protein. For example in Human Rab5a GTPase domain (PDB code: 1R2Q)
(Fig. 9b) the Ala30 is in disallowed conformation and its flanking residues are interacting with the ligand Phosphoaminophosphonic acid-guanylate ester. In 3,4-dihydroxy-2-butanone 4-phosphate synthase (PDB code: 1K4I) (Fig. 9a) Arg 36 is in disallowed conformation and the residues flanking interacting with magnesium [22]. In many cases these sterically strained regions are involved in metal or ligand interaction through hydrophobic or hydrogen bond interactions. Also it is evident from the flowchart (Fig. 2) that in many cases the disallowed conformation is conserved and also having functional significance. There are also substantial number of cases with conservation of disallowed conformation but do not show any functional significance. Reason for conservation of disallowed conformation in these cases is currently obscure. Out of 221 disallowed phi, psi values considered for analysis 116 phi, psi values are involved in at least one pairwise alignment with disallowed conformation conserved. Further when
Fig. 9. Examples of disallowed conformation showing functional significance. The residues with disallowed conformation and its flanking residues are shown in red and cyan color respectively. The ligand is shown in pink and orange color stick models, metals as a small spheres and catalytic residues are marked. (a) 3,4-Dihydroxy-2-Butanone 4-Phosphate Synthase (PDB code: 1K4I) the flanking residues interact with the metal. (b) Human Rab5a GTPase Domain (PDB code: 1R2Q) the residue in the disallowed conformation is Ala30 and its flanking residues interact with the ligand phosphoaminophosphonic acid–guanylate ester (GNP).
B. Lakshmi et al. / International Journal of Biological Macromolecules 63 (2014) 119–125
these cases were subjected to the analysis of functional significance in roughly half the number of cases the disallowed region is in the functional environment. Almost in all of these cases the residues in disallowed conformation are mentioned in “PDBSITE” record according to Protein Data Bank entry which means that these residues are in the functional environment in the protein. The details of the functional significance are provided in the supplementary data (Supplementary Table S2). In many of the examples analyzed the residue in disallowed conformation and/or its flanking residues are involved in function. In many cases the residues involved are found to be in beta hairpins. Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ijbiomac. 2013.10.032. 4. Conclusions The present analysis illustrated that the Ramachandran disallowed conformations do occur in ultra-high resolution protein structures in locations with reasonably low temperature factor. Such genuine disallowed conformations are only moderately conserved among the homologues in protein domain families. The extent of conservation of disallowed nature is higher when the disallowed residue type is also conserved compared to when residue type is not conserved. It is also clear that none of the residue types are highly preferred for the conservation of disallowed conformation. The analysis on the functional significance of the disallowed conformations suggests that in roughly half the number of cases, with disallowed conformation conserved, the residue in the disallowed conformation or its flanking residues are associated with function of the protein.
125
Acknowledgements This work is supported by Council of Scientific and Industrial Research (CSIR), DBT-AIST (Indo-Japan collaborative project) and by DBT, Government of India. References [1] G.N. Ramachandran, C. Ramakrishnan, V. Sasisekharan, J. Mol. Biol. 7 (1963) 95–99. [2] C. Ramakrishnan, G.N. Ramachandran, Biophys. J. 5 (1965) 909–933. [3] G.N. Ramachandran, V. Sasisekharan, Adv. Protein Chem. 23 (1968) 283–438. [4] T.A. Jones, J. Appl. Crystallogr. 11 (1978) 268–272. [5] A.L. Morris, M.W. MacArthor, E.G. Hutchinson, J.M. Thornton, Proteins: Struct Funct Genet 12 (1992) 345–364. [6] G.J. Kleywegt, Acta Crystallogr. D Biol. Crystallogr. 56 (2000) 249–265. [7] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thornton, J. Appl. Cryst. 26 (1993) 283–291. [8] C. Ramakrishnan, B. Lakshmi, A. Kurien, D. Devipriya, N. Srinivasan, Protein Pept. Lett. 14 (2007) 672–682. [9] K. Gunasekaran, C. Ramakrishnan, P. Balaram, J. Mol. Biol. 264 (1996) 191–198. [10] Z. Jia, M. Vandonselaar, J.W. Quail, L.T. Delbaere, Nature 361 (1993) 94–97. [11] O. Herzberg, J. Moult, Proteins 11 (1991) 223–229. [12] S. Balaji, S. Sujatha, S.S. Kumar, N. Srinivasan, Nucleic Acids Res. 29 (2001) 61–65. [13] A.G. Murzin, S.E. Brenner, T. Hubbard, C. Chothia, J. Mol. Biol. 247 (1995) 536–540. [14] D.E. Tronrud, J. Appl. Crystallogr. 29 (1996) 100–104. [15] S. Parthasarathy, M.R. Murthy, Protein Sci. 6 (1997) 2561–2567. [16] Z. Yuan, J. Zhao, Z.X. Wang, Protein Eng. 16 (2003) 109–114. [17] G.J. Kleywegt, M.R. Harris, J.Y. Zou, T.C. Taylor, A. Wahlby, T.A. Jones, Acta Crystallogr. D Biol. Crystallogr. 60 (2004) 2240–2249. [18] L. Holm, P. Rosenstrom, Nucleic Acids Res 38 (2010) W545–W549. [19] A.S. Konagurthu, J.C. Whisstock, P.J. Stuckey, A.M. Lesk, Proteins 64 (2006) 559–574. [20] K. Gunasekaran, R. Nussinov, J. Mol. Biol. 365 (2007) 257–273. [21] R.A. Laskowski, E.G. Hutchinson, A.D. Michie, A.C. Wallace, M.L. Jones, J.M. Thornton, Trends Biochem. Sci. 22 (1997) 488–490. [22] D.I. Liao, Y.J. Zheng, P.V. Viitanen, D.B. Jordan, Biochemistry 41 (2002) 1795–1806.
