Erratum: 29(2) p 408-412 Molecular surface directionality of the DNA-binding protein surface on the earth map

Protein-DNA interactions play a pivotal role in both the transcriptional control and the maintenance of genome integrity, and these are two properties that are closely linked to the development of an organism, differentiation, physiology and to the progression of diseases. Chemical and geometric properties are typically two of the key components in any analysis that aims to understand the precise origin of specificity and elucidate the atomic features of a protein-DNA interface. In this study, we have developed a unique representation of the directionality of the molecular surface of a DNA-binding protein. The stereo-orientation of the normal vector that signifies the geometric properties of a protein surface was projected onto a two-dimensional surface (referred to here as an earth map). We identified considerably diverse patterns of the vector distribution of the protein surface, and besides this, the DNA-contact surface, a subset of an entire protein surface, has also been found to contain diverse patterns. At the same time, the direction of the DNA-contact surface was also tracked onto the earth map on a base-pair basis and distinct intertwining properties particular to the specific family of that DNA-binding protein are revealed.

One of the ultimate goals of functional genomics is to identify and record the entire transcription map of any given species among all living organisms.The interactions between DNA-binding proteins and DNA are, after all, responsible for the initiation, elongation and the termination of transcription processes.Aside from this, the maintenance of genome integrity, the control of epigenetics and the recombination of DNA are all molecular processes that involve protein-DNA interactions.Both the manner in which specificity is born and in which proteins search and anchor onto DNA during protein-DNA interactions have been the focus of a great deal of intensive research for much of the past decades (Pabo and Sauer, 1984;Choo and Klug, 1997;Garvie and Wolberger, 2001;Jayaram and Jain, 2004).
While static images of protein-DNA complexes can be obtained by means of X-ray crystallography and nuclear magnetic resonance, the actual realization of a dynamic portrait of a protein-DNA recognition process has yet to be fully achieved.In search of specific atomic details, protein-DNA interfaces have been well examined, but thus far, only two properties have been identified to explain the complementary properties on the interfaces: chemical and geometric complementarity.As concerns the former, it has been established that most protein-DNA interfaces are more polar and that they contain more hydrogen bonds than do protein-protein interfaces (Jones et al., 1999;Nadassy et al., 1999) despite the comparable interfacial gap volume between the two (Jones et al., 1999).And certainly not to be ignored, given the importance of specific DNA sequences and the flexibility of DNA to protein-DNA interactions, several other studies have been centered on changes in DNA conformation (Suzuki and Yagi, 1994;Meierhans et al., 1997;Dickerson, 1998;Mandel-Gutfreund and Margalit, 1998;Segal and Barbas, 2000;Maris et al., 2002;Ahmad et al., 2004;Havranek et al., 2004;Paillard et al., 2004).
Turning to geometric complementarity, what is striking is the way protein shape "follows" DNA conformation, or conversely, the way DNA "adjusts" its conformation such that it closely parallels the shape of a protein surface.To quantitate exactly how the shape of a protein surface stays in line with that of a DNA molecule, we have previously investigated the relationship between the "direction" of a protein surface (i.e., the "normal vector" of a protein surface) and that of a DNA molecule (i.e., the "axis" of a base-pair plane) (Yeh et al., 2003).In the analysis of a set of non-redundant protein-DNA complexes with known three-dimensional structures, strong evidence has substantiated that a significant correlation exists between the direction of a protein surface and the conformation of a DNA molecule.Thus, in that research (Yeh et al., 2003), a new geometric property of protein-DNA interfaces was determined, showing that the 'shape complementarity' of protein-DNA recognition unambiguously bears the property of 'directionality'.
Equally important, our goal in this study was to investigate the distribution of the orientation that best depicts the direction of the molecular surface of a DNA-binding protein.To achieve this we, first of all, moved the root of each normal vector to the same origin so that we could compare all of the vectors by only seeing the arrow heads of the vector from a shared origin.To facilitate the examination of the vector distribution, we employed the Hammer-Aitoff projection method to map a projection of the vector tips from the three-dimensional unit sphere to a two-dimensional plane.The transformation formulas are as follows (Snyder, 1993): where φ is the latitude of each vector tip (+ if north and -if south); λ is the longitude (+ if eastward and -if westward); R is the radius of the sphere (R = 1); and s and t are the positions on the projected two-dimensional plane (i.e., the earth map).The results from the two-dimensional perspective of the normal vectors of the protein surface were enlightening.Take the human YY1 zinc finger protein-DNA complex structure as an example (Figure 1A).It is readily observed that the normal vectors are unevenly distributed on the earth map, with some strips and zones showing a clustering of vectors (Figure 2A).The pattern of this map is specific to the YY1 zinc finger protein surface.That is, it has its own special pattern as opposed to, say, being P53 specific with its own special pattern (Figure 1B, 2B).We also explored the normal vectors of the DNAbinding proteins that are involved in DNA contact.Since the molecular surface of the protein involved in each DNA contact is but a subset of the whole surface of all the proteins, only some of the vectors in Figure 2A are shown in Figure 2C.Characteristic of the vector of the YY1 protein surface in contact with the DNA are three strips; however, there is an isolated patch of vectors which features the normal vector of the P53 protein surface in contact with the DNA (Figure 2D).
We then examined the molecular surface contact between the protein and the DNA in detail by profiling the normal vectors of the protein surface averaged on a basepair unit.We calculated the vector sum of the normal vectors of the protein surface in contact with each DNA basepair and also illustrated the vector sum on the earth map based upon the base-pair unit.As shown in the example of the YY1 zinc finger protein, the normal vectors of the protein surface intertwining through the major groove of the DNA (Figure 1A and 1C) traverses longitudinally through a circle on the map (almost 360°) (Figure 2E).This particular "intertwining" feature is dramatically different from that shown for P53, which zigzags in a latitudinal fashion (Figure 2F).
We also employed the clustering of the normal vectors of the DNA-contact protein surface averaged on a base-pair unit on the earth map using the dynamic programming method.We found that the directional features of the molecular surface of DNA-binding proteins, including helix-loop-helix, zinc finger, β hairpin/ribbon and helixturn-helix families, formed distinct groups in the dendrogram (Figure 3), demonstrating that DNA-binding proteins of the same family have similar directionality patterns.
We have also constructed a website (named "ShapeCom" at http://140.121.200.163/shapecom.htm) to present the details of different protein-DNA interactions.On the website, the geometric properties of the interface from various protein-DNA complexes are visually represented, and valuable links to other web-sites are provided for the examination of the coordinates and the chemical properties of protein-DNA interfaces.The web contents for each protein-DNA complex include: Data & links 1) links to the Protein Data Bank (PDB, http://www.rcsb.org/pdb/)(Deshpande et al., 2005); 2) the coordinates of the protein part, the DNA part and the protein-DNA complex (Deshpande et al., 2005); 3) links to the web-site of the Biomolecular Structure and Modelling (BSM) group, University College, London (http://www.biochem.ucl.ac.uk/bsm/prot_dna/prot_dna_c over.html); and 4) links to PDBsum (Laskowski et al., 2005).http:// swissmodel.expasy.org/spdbv/(Guex and Peitsch, 1997)); 6) stereo images of the normal vectors of a protein surface and the axes of a DNA base-pair plane (Yeh et al., 2003); 7) the angles between the normal vectors of a protein surface and the axes of a DNA base-pair plane (Yeh et al., 2003); 8) a two-dimensional projection of the normal vectors of a protein surface (this study, Figures 2A, 2B, 2C and 2D); and 9) a two-dimensional projection of the normal vectors of a protein surface in contact with DNA, averaged on a base-pair unit (this study, Figures 2E and 2F).
On the weight of the evidence from our previous investigation, we concluded that the normal vectors of a DNA-contacting protein surface distinctly prefer certain angles.This enables them to align with certain axes that characterize the conformation of the DNA.We have now extended our 'shape complementarity' studies of protein-DNA recognition to encompass the topographic properties of the 'directionality' of the protein surface.By employing the two-dimensional projection techniques presented here, we found that the distribution of the normal vectors on the earth map is uneven and, at the same time, that it varies among different DNA-binding proteins.Beyond this, most of the vectors in contact with DNA also have their own distinct pattern depending on the protein under investigation.Our two-dimensional representation of the normal vectors of a protein surface can be regarded as a natural extension of the current trends in protein research.The protein surface has long been used to categorize and predict the functions of proteins and their interaction with other biological molecules (Lichtarge et al., 1996;Jones and Thornton, 1997).To simplify the representation and comparison, the protein surface was approximated using the spherical harmonic function (Duncan and Olson, 1993).More specifically, a spherical approximation of the protein surface was used to analyze the surface features within homologous families and to predict the conservation and divergence of protein functions and protein-protein interactions (Pawlowski and Godzik, 2001).Worth bearing in mind is that the normal vectors of a protein surface were previously used in the protein-protein and protein-ligand docking problem (Norel et al., 1995;Norel et al., 1999).To the best of our knowledge, however, never before have the normal vectors been represented on a two-dimensional projection (earth map); nor have the distinct patterns for the "directionality" of proteins and their DNA-contact patches been visualized on an earth map.As further detailed in our web-site (ShapeCom), DNA-binding proteins in different families utilize the family-specific protein surface in DNA contacts.The specific directionality of each family of DNA-binding proteins may very well play an important role when it comes to under-standing the recognition process of protein-DNA and protein-protein interactions.
Erratum: 29(2) p 408-412 Molecular surface directionality of the DNA-binding protein surface on the earth map This article has received corrections in agreement with the ERRATUM published in Volume 29 Number 4.

Figure 1 -
Figure 1 -Stereo image of the structure (A and B) and molecular surface (C and D) of the YY1 zinc finger-DNA complex (A and C, PDB code 1ubd) and the P53-DNA complex (B and D, PDB code 1tsr).The molecular surface of the protein is marked in grey, and the surface in contact with the DNA is in yellow.For clarity, the DNA is shown as a ribbon.(For the views of different orientations, please see http://140.121.200.163/molecule.asp?pathn=1ubd and http://140.121.200.163/molecule.asp?pathn=1tsr).
Figure 2 -Orientation distribution of the normal vectors of a protein surface shown on the earth map (A, C and E for the YY1 zinc finger-DNA complex; B, D and F for the P53-DNA complex).The normal vectors for the whole protein are shown in A and B, while the normal vectors involved in the DNA contact are shown in C and D. The normal vectors of the protein surface involved in DNA contact are averaged on a base-pair unit and are shown on the earth map (E, F).Each diamond-shaped symbol represents a base-pair, and they are linked with a dashed line to trace the overall orientation of the average of the normal vectors.Also shown is the nucleotide sequence of the YY1 zinc finger protein-bound DNA (nucleotide sequence T5-C6-T7-C8-C9-A10-T11-T12-T13-T14-G15-A17, PDB ID: 1ubd) (E) and the P53-bound DNA (A9-C10-T12-G13-C14-C16-A17, PDB ID: 1tsr) (F).

Figure 3 -
Figure 3 -A dendrogram for the clustering of the normal vectors of the DNA-contact protein surface averaged on a base-pair unit on the earth map.For each leaflet, PDB ID, protein name and the DNA-binding protein family name are shown.

Figure 2 -
Figure 2 -Orientation distribution of the normal vectors of a protein surface shown on the earth map (A, C and E for the YY1 zinc finger-DNA complex; B, D and F for the P53-DNA complex).The normal vectors for the whole protein are shown in A and B, while the normal vectors involved in the DNA contact are shown in C and D. The normal vectors of the protein surface involved in DNA contact are averaged on a base-pair unit and are shown on the earth map (E, F).Each diamond-shaped symbol represents a base-pair, and they are linked with a dashed line to trace the overall orientation of the average of the normal vectors.Also shown is the nucleotide sequence of the YY1 zinc finger protein-bound DNA (nucleotide sequence T5-C6-T7-C8-C9-A10-T11-T12-T13-T14-G15-A17, PDB ID: 1ubd) (E) and the P53-bound DNA (A9-C10-T12-G13-C14-C16-A17, PDB ID: 1tsr) (F).