proteins

STRUCTURE O FUNCTION O BIOINFORMATICS

A novel method to map and compare proteinprotein interactions in spherical viral capsids
Mauricio Carrillo-Tripp, Charles L. Brooks III, and Vijay S. Reddy*
Department of Molecular Biology, The Scripps Research Institute, La Jolla, California 92037

ABSTRACT Viral capsids are composed of multiple copies of one or a few chemically distinct capsid proteins and are mostly stabilized by inter subunit proteinprotein interactions. There have been efforts to identify and analyze these proteinprotein interactions, in terms of their extent and similarity, between the subunit interfaces related by quasiand icosahedral symmetry. Here, we describe a new method to map quaternary interactions in spherical virus capsids onto polar angle space with respect to the icosahedral symmetry axes using azimuthal orthographic diagrams. This approach enables one to map the nonredundant interactions in a spherical virus capsid, irrespective of its size or triangulation number (T), onto the reference icosahedral asymmetric unit space. The resultant diagrams represent characteristic fingerprints of quaternary interactions of the respective capsids. Hence, they can be used as road maps of the proteinprotein interactions to visualize the distribution and the density of the interactions. In addition, unlike the previous studies, the fingerprints of different capsids, when represented in a matrix form, can be compared with one another to quantitatively evaluate the similarity (S-score) in the subunit environments and the associated proteinprotein interactions. The S-score selectively distinguishes the similarity, or lack of it, in the locations of the quaternary interactions as opposed to other well-known structural similarity metrics (e.g., RMSD, TM-score). Application of this method on a subset of T 5 1 and T 5 3 capsids suggests that S-score values range between 1 and 0.6 for capsids that belong to the same virus family/genus; 0.60.3 for capsids from different families with the same T-number and similar subunit fold; and <0.3 for comparisons of the dissimilar capsids that display different quaternary architectures (T-numbers). Finally, the sequence conserved interface residues within a virus family, whose spatial locations were also conserved have been hypothesized as the essential residues for selfassembly of the member virus capsids.
Proteins 2008; 73:644655.
C V 2008 Wiley-Liss, Inc.

INTRODUCTION Virus particles are assembled from multiple copies of one or a few chemically distinct coat protein (CP) subunits packaging viral genome and relevant accessory proteins in vivo. Some of the viral capsid proteins spontaneously form empty capsids when expressed in heterologous expression systems or assembled in vitro.15 Whether the virus particles assemble in vivo or in vitro, the inter subunit proteinprotein interactions are critical for the formation, stability, and integrity of virus particles. The extent of interactions and the type of resulting capsids are encoded into individual capsid protein subunits as they yield mostly uniform capsids of a single kind (T 5 1, T 5 3, etc.). Analysis of proteinprotein interactions in viral capsids within and across members of virus families and capsid types will provide the locations of the hotspots and weak links of the assembled capsids and highlight the extent, distribution, and similarity of these interactions.69 Here, we report a new approach to map proteinprotein interactions onto azimuthal polar orthographic projection diagrams using spherical coordinates. Its only natural to represent the quaternary interactions of spherical viral capsids using a spherical coordinate system. In addition, spherical polar projections provide a normalized way to map and compare interactions as opposed to projections in Cartesian space, where the coordinates have to be scaled up or down with respect to the size of the reference particle. The method described in this work involves mapping all the unique residues in the icosahedral asymmetric unit that interact at the intersubunit interfaces as azimuthal orthographic projections. These diagrams represent the structural fingerprints of quaternary interactions of the respective capsids. Hence, they can be used as roadmaps to visualize the distribution and the extent of proteinprotein interactions with respect to the icosahedral symmetry axes, irrespective of the size and the type (T-number) of spherical
Grant sponsor: National Institutes of Health (NCRR center for Multiscale Modeling Tools for Structural Biology (MMTSB)); Grant number: RR12255. *Correspondence to: Vijay S. Reddy, Department of Molecular Biology, TPC-06, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037. E-mail: reddyv@scripps.edu Received 15 October 2007; Revised 29 February 2008; Accepted 9 March 2008 Published online 19 May 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.22088

Key words: viral capsids; proteinprotein interactions; quaternary interactions; similarity score; structural fingerprints.

644

PROTEINS

C V 2008 WILEY-LISS, INC.

Protein Interactions in Spherical Viral Capsids

virus capsids. Furthermore, these fingerprints can be readily compared and contrasted with one another and can be used to evaluate quantitatively the extent of similarity of proteinprotein interactions in different capsids by representing the projections in a matrix form.

MATERIALS AND METHODS

Azimuthal polar orthographic diagrams (APODs)

In this method, spherical coordinates (r,F,w) of all the residues interacting at the intra and inter icosahedral asymmetric unit interfaces were obtained. The amino acids contacting at the subunit interfaces were identified by the residue-pair specific distance cut-off criteria between the centers of mass (COMs) of the side-chain atoms.7 Cartesian coordinates of the COMs of all the interface residues were extracted from VIPERdb (http://viperdb.scripps.edu), where tables of contacting pairs of residues are readily available.10 The COMs were then transformed into unit vectors and projected onto (F,w) space as an azimuthal polar orthographic diagram (APOD), which resembles projection of a sphere (globe) from one of its poles (North Pole). Figure 1(a) shows a point in Cartesian space, P(x,y,z), and its transformation to spherical space representation P(r,F,w), where F is the angle between the projection of the vector ~ r onto the XY plane and the X axis, and w is the angle between the vector ~ and the Z axis. The vector ~ is norr r malized so that the point P will lie in the surface of a unit sphere. Figure 1(b) shows the point P represented on an APOD. The F angle grows counterclockwise going all around from 08 to 3608. The w angle grows radially starting with 08 at the center of the diagram and ending at 908 at the edge of the diagram (outermost circumference; equator of the unit sphere). Each symbol on an APOD represents the location of the center of mass of side chain atoms of an interface residue i or j, where (i,j) is a specific interacting pair, in Fw space. Both residues of the interacting pairs were mapped. Such a representation is unambiguous and better than locating an interaction as the mid point of the interacting pair, particularly when the APODs are being compared to evaluate Sscores (see below). Two types of symbols were used to distinguish intra (circles) and inter (triangles) asymmetric unit interactions. Locations of the icosahedral symmetry axes comprising the central icosahedral asymmetric unit (CIAU) were highlighted on the APODs as fiducials.
Similarity (S) score, a measure of similarity in quaternary interactions

Figure 1
(a) Representation of coordinates of a point P in Cartesian space and its corresponding coordinates in spherical polar space. (b) Projection of the point P onto an azimuthal polar orthographic diagram (APOD). The four relevant icosahedral symmetry axes in the VIPER convention11 are shown in bold. These include two threefold axes (triangles) at [08, 218] and [1808, 218], fivefold axis (pentagon) at [908, 328], and twofold axis (oval) at [08, 08]. The central icosahedral asymmetrical unit, or CIAU, is represented by the black lines connecting the four icosahedral symmetry elements. The corresponding locations of both the X and Y Cartesian axes are also shown. The Z-axis would be perpendicular to the plane of the diagram at the center, growing in positive values above the page towards the reader.

The proteinprotein interactions in the polar orthographic diagrams can also be represented as an M 3 L matrix, N (polar orthographic matrix, or POM), where dimension M represents the range of the F angle (08

3608), and dimension L corresponds to the range of the w angle (08908). M and L are constant for all spherical viral capsids. Each matrix element in N (POM) can be considered as a cell with a small uniform surface area in the Fw space. The differential of the surface area in spherical coordinates is given by dS 5 r2 sin(w) dw dF 5 r2 dcos(w) dF; r2 5 1 for an unit sphere. To obtain cells of constant surface area of the sphere, the range of
PROTEINS

645

M. Carrillo-Tripp et al.

F (083608) was divided into regular intervals as well as the range of cos(w) (01). The F angle was divided into 360 bins, with a dF 5 18 and the cos(w) is divided into 180 bins, which amounts to dcos(w) of 0.0056. Then, for each value of cos(w), the corresponding value of w can be obtained by applying the acos() function. The area of each cell (dS) will be 0.97 3 1024 A2 for a unit sphere. This translates into $1.0 A2 for a sphere of 100 A in ra dius (200 A in diameter). The value of each element (cell) in the matrix N is equal to the number of interactions found in the corresponding cell in Fw space, i.e., an element (or cell) of the matrix will be zero if no interactions were found in that cell or a positive integer equal to the number of interactions present. Since the average resolution of all the X-ray structures found at VIPERdb is $3 A, COM of a residue in Cartesian space was considered to be anywhere inside a cubic box of the size of 3 A (XCOM 1.5 A, YCOM 1.5 A, ZCOM 1.5 A), thus giving rise to a range of possible Fw values dictated by the volume of the box. This smearing of the residue positions was achieved by turning on the eight adjacent elements in the N matrix. Another way to consider the smearing of residue locations would be to increase the area of the cells in the Fw space. The former approach, however, proved to be more sensitive than the latter. P P a b 2 i j Nij Nij i S score  P P h a b Nij 2 Nij 2 i j Similarity (S) score corresponds to a fraction of common locations of quaternary interactions between a pair of spherical virus capsids with respect to the total number of interactions found in the respective capsids. The dot (inner) product of two POMs (N a and N b) provides the number of common contacts between the two capsids a and b. Therefore, the S-score is defined as twice the ratio of common locations of interactions normalized by the total number of points of interaction in both the capsids being compared. The S-score for identical capsids will be 1, and <1 for different capsids. The S-score, with a range between 0 and 1, is a valuable metric to quantitatively assess the similarity in quaternary interactions and subunit environments between capsids.

values between 0 and 1. The magnitude of its value is not dependent on protein size. Because TM-score weights a close match stronger than a distant match, TM-score is more sensitive than the RMSD. A TM-score 0.17 indicates that the structures are unrelated while a TM-score !0.45 suggests the structures display similar topology.12,13

RESULTS AND DISCUSSION Earlier studies on comparing proteinprotein interactions in spherical virus capsids were focused on evaluating the conservation of residue pairs (interactions) as a fraction of total number of contacts (Q-scores) across quasi-equivalent interfaces within individual capsids.7 The Q-scores were calculated based on the contact pairs, whose identities were defined by their residue numbers, which are specific to the respective CP subunits. The equivalences of the contacting pairs between different capsids cannot be made easily without the prior identification of the corresponding residue contacts based on sequence alignments. Hence, this method could not be extended for comparing the interactions across different capsids. A more recent work by Shepherd and Reddy8 and Bahadur et al.9 calculated the extent of proteinprotein interactions as a fraction of buried surface area of individual subunits upon capsid formation relative to isolated subunits in spherical virus capsids of various architectures. They suggested that the extent of buried surface area provides a characteristic estimate of quaternary associations and discussed their implications on capsid formation and assembly pathways. Quite sometime ago, Rossmann et al.14 analyzed the similarity in tertiary and quaternary structures by structural superposition of three spherical plant viruses: satellite tobacco necrosis virus, STNV (2buk), southern bean mosaic virus, SBMV (4sbv) and tomato bushy stunt virus, TBSV (2tbv). They showed that despite considerable differences in the properties of these viruses and low sequence similarities between the capsid protein subunits (CPs), the topology and the tertiary b-barrel fold of the CPs are remarkably similar. In addition, they observed a marginal similarity in the locations of contacts (e.g., salt bridges) at subunit interfaces by explicitly superimposing subunits on one another. In this study, we highlight a new method of mapping the locations of the residues involved in subunitsubunit interactions on APODs. These normalized diagrams are useful as roadmaps for the visualization of the density, distribution and characteristics of the residues at the subunit interfaces. Furthermore, we discuss the utility of representing the APODs in a matrix form to quantify the overall similarity as well as identify the common fingerprints of quaternary interactions shared by groups of capsids from within and across families. In addition,

Structural similarity metrics used in the current study

The root mean square deviation (RMSD) measures the extent of structural similarity between two proteins by aligning the structurally analogous regions followed by least squares optimization of the distances between the aligned CA atoms. Low RMSD value (<2 A) with at least 30% of the residues aligned indicate significant structural similarity. TM-score12 estimates the similarity in the topologies of two protein structures with a range of

646

PROTEINS

Protein Interactions in Spherical Viral Capsids

locations of the sequence-conserved interface residues present in the spatially conserved fingerprint of the four members of the Nodaviridae family were identified. We suggest that these residues could be of importance for the self-assembly of Nodavirus capsids. APODs for the capsids of STNV,15 SBMV,16 and TBSV17 are shown in Figure 2. The diagrams for SBMV and TBSV capsids look very similar. However, there are differences near the threefold axis at [1808, 218] arising due to the differences in the extent of the ordered peptide arm of the C-subunit forming the b-annulus. In addition, the diagram for TBSV shows an extra set of interactions near the middle of each of the three sides of the triangular central icosahedral asymmetrical unit (CIAU), in comparison to that of SBMV. These contacts (residues) protruding out of the CIAU correspond to the interactions resulting from the P domains of the TBSV subunits, which are absent in SBMV. Locations of the interactions at the intraunit (inside the CIAU), interunit interfaces and their relative extents at these interfaces are readily discernible from these diagrams. For example, the P domains of the TBSV capsid participate only in the interunit interactions while the S domains are involved in the contacts at intraunit as well as interunit interfaces. The P domain of the TBSV CP contributes on average 33 more interactions per subunit compared with that of the SBMV subunit.
Comparison of interactions in STNV, TBSV, and SBMV capsids

The APODs of different capsids can be compared with one another to assess the degree of similarity in the locations of their proteinprotein interactions using S-scores as well as to obtain a common characteristic structural fingerprint of quaternary proteinprotein interactions among a group of capsids. Similarity in quaternary interactions (S-scores) were calculated between the three viruses studied by Rossmann et al.14 An S-score of 0.60 between TBSV and SBMV capsids suggests that there is moderate similarity between the locations of interface residues in both viruses, even though the capsids have only 26% overall sequence identity and belong to different virus families. As noted earlier, STNV forms T 5 1 capsids, which are composed of 60 subunits, while TBSV and SBMV form T 5 3 capsids containing 180 subunits. Not surprisingly, comparison of the APOD of STNV with those of TBSV and SBMV resulted in lower S-scores; 0.16 and 0.17, respectively. To test whether or not the poor scores are simply due to comparing the interactions of the T 5 1 capsid of STNV capsid having 1 subunit in the CIAU with those of the T 5 3 capsids of TBSV and SBMV with three subunits occupying the CIAU, we analyzed the environments of the individual subunits. The interface residues corresponding to the individual subunits (A/B/C) of TBSV

and SBMV capsids were identified and transformed according to the structural superposition of the respective (T 5 3) subunits on the STNV subunit. The resultant transformed APODs of the individual subunits of SBMV and TBSV capsids were compared with the APOD of the STNV subunit (see Fig. 3). Interestingly, the Sscores calculated for each individual subunits of SBMV and TBSV with respect to STNV were 0.21, 0.24, and 0.22 for the SBMVA, SBMVB and SBMVC respectively, and 0.20, 0.21, and 0.17 for the TBSVA, TBSVB and TBSVC, respectively. The subunit wise S-scores are only slightly better than the S-scores (0.16 and 0.17) for comparing the conventional APODs of the CIAUs of the TBSV and SBMV capsids. Taken together, this suggests that the individual subunit environments of the SBMV and TBSV capsids are indeed different compared with that of the STNV subunit. Additionally, environments of the individual A, B, and C subunits of the TBSV and SBMV capsids were compared with one another within each capsid and between the capsids (Table I). The S-scores for all inter-subunit pairs within the individual CIAUs of TBSV and SBMV are listed in Table I, respectively. In both cases, the C subunits display distinct environments (S-score: h0.7i) in comparison to that of the A or B subunits, while the environments of the A and B subunits compare well (Sscore: h0.85i) with each other. This distinction in the environments of the C-subunits is likely due to selective ordering of a segment of the N-terminal peptide arm responsible for forming the b-annulus structure. Furthermore, S-scores for the intersubunit pairs between the SBMV and TBSV capsids showed a similar pattern (Table I). This provides a quantitative assessment of the extent of variation in the environments of the C-subunits compared to the A and B-subunits both in intra- or intercapsid analysis of the two T 5 3 capsids.
Analysis of parvo, sobemo, and nodavirus capsids

In order to gain further insights into the relationship between the S-scores and the structural similarity within and across virus families, the analysis was extended to a few selected groups of virus capsids. The intrafamily similarity scores for capsid pairs in three different virus families, Parvoviridae (T 5 1), Sobemoviridae (T 5 3), and Nodaviridae (T 5 3), are shown in Table II. Starting with the Parvoviridae family, there is a clear difference between the densovirus capsid (PDB-ID:1dnv, genus: densovirus; S-score 5 h0.33i)18 and the other members of the family, which primarily belong to the parvovirus genus (S-score 5 h0.88i). This difference was clearly highlighted in the orthographic diagrams (not shown), which have a lower density of interactions around the threefold axis at [08, 218] of capsid 1dnv compared to the other member capsids in the family. These differences
PROTEINS

647

M. Carrillo-Tripp et al.

Figure 2
Schematic illustrations and APODs for STNV (a), SBMV (b), and TBSV (c) capsids. Illustrated on the left are schematic representation of the respective capsids with the subunits that occupy the CIAU are shown in color and the symmetry related copies are shown in gray. Icosahedral lattices are shown as black lines surrounding the individual capsids. Shown on the right are the corresponding APODs. Two types of symbols were used to distinguish intra (circles) and inter (triangles) asymmetric unit interactions. The locations of the interacting residues are color coded according to the corresponding subunits: A-subunit in blue, B-subunit in red, and C-subunit in green. Each subunit is surrounded by a border approximately delimiting the spatial extent of the subunits. The CIAU is depicted as a triangle.

648

PROTEINS

Protein Interactions in Spherical Viral Capsids

Figure 3
Comparison of APODs of STNV (T 5 1) (a) with those of the transformed subunits of SBMV: A-subunit (b), B-subunit (c), and C-subunit (d), which were oriented with respect to the STNV subunit.

seen in the APODs of parvo and densovirus capsids highlight the characteristic differences in their quaternary interactions even though they belong to the same virus family. This suggests that there could be significant dif-

ferences in the extent of quaternary interactions even among members of the same family. This finding is consistent with the differences in the estimated intersubunit
Table II

Pairwise Intra-Family S-Scores for Three Families of Virus Capsids

Table I
TBSV17

Pairwise Inter-Subunit S-Scores for TBSV and SBMV Capsids A0 1.00 0.86 0.75 A 1.00 0.85 0.76 A 0.54 0.54 0.50 B0 0.86 1.00 0.68 B 0.85 1.00 0.65 B 0.49 0.51 0.48 C0 0.75 0.68 1.00 C 0.76 0.65 1.00 C 0.37 0.48 0.57

Parvoviridae 1dnv18 1dnv 1.00 1c8d 0.33 1c8h 0.33 1k3v 0.34 Sobemoviridae 1smv21 1smv 1.00 1x35 0.92 1f2n 0.69 4sbv 0.88 1ng0 0.69 Nodaviridae 1nov25 1nov 2bbv 1f8v fhv 1.00 0.86 0.81 0.86

A0 B0 C0 SBMV16 A B C SBMV/TBSVa A0 B0 C0
a 0

1c8d19 0.33 1.00 0.92 0.83 1x3522 0.92 1.00 0.69 0.87 0.69 2bbv26 0.86 1.00 0.80 0.90

1c8h19 0.33 0.92 1.00 0.83 1f2n23 0.69 0.69 1.00 0.73 0.76 1f8v27 0.81 0.80 1.00 0.78

1k3v20 0.34 0.83 0.83 1.00 4sbv16 0.88 0.87 0.73 1.00 0.71 fhv28 0.86 0.90 0.78 1.00 1ng024 0.69 0.69 0.76 0.71 1.00

A , B0 C0 indicate A, B and C subunits in the TBSV capsid.

PROTEINS

649

M. Carrillo-Tripp et al.

association energies for these capsids, which are available at the VIPERdb website.10 The orthographic diagrams of Sobemoviridae show that they all have very similar densities of interacting residues except for the differences occurring due to selective ordering of the peptide arm of the C subunit; the comparison of S-scores of the members of the Sobemoviridae family (Table II) show that they fall into two groups: sesbania mosaic virus21 (1smv, 1x35) and southern bean mosaic virus16 (4sbv) correspond well with each other (S-score 5 h0.90i) as they share a common configuration of the peptide arm of the C-subunit that makes a sharp U turn near the twofold axis and extending out towards the threefold axis at [1808, 218]. Whereas the rice yellow mottle virus23 (1f2n) and cocksfoot mottle virus24 (1ng0) capsids (S-score 5 0.76) are distinct from the first group with the peptide arm adopting a different configuration from those of the SBMV, SMV, and TBSV capsids by extending towards to the opposite threefold axis at [08, 218] without making the U turn. In a similar fashion, members of the Nodaviridae family share the same subunit topology and organization with respect to the icosahedral symmetry axes and show higher S-score values (h0.82i) for the comparisons with one another. Differences arising mainly due to the N-terminal arms of the A subunits, which were considered earlier to be part of the C-subunits. Pariacotovirus27 (1f8v), however, shows a higher density of interactions in the intraunit interfaces and has 30% more interactions than the other three members mainly due to ordering of the nearly complete peptide arm of the A-subunit. As mentioned earlier, Sobemoviridae and Nodaviridae families share the same triangulation number (T 5 3), while Parvoviridae form T 5 1 capsids. As expected, the differences in quaternary architectures are reflected in the low values of the S-scores (h0.22i) between capsids of Parvoviridae and Sobemoviridae or Nodaviridae (Table III). On the other hand, comparison of the members of the T 5 3 capsids display higher S-scores (h0.60i) as shown in Table III. On the basis of the range (01) of S-scores, spherical capsids can be clustered into three groups reflecting the similarity or lack of it in the quaternary interactions. The first group has S-scores in the range of (1.000.6) and is composed of capsids that display the same quaternary architecture (T-number) and similar subunit topology with minor variations at the termini and in the loops (Table II). These capsids are mostly structural homologs and belong to the same family and/or genus. The second group consists of capsids that exhibit the same quaternary architecture and contain CPs with minor differences in their topology and/or with large insertions relative to one another (e.g., TBSV vs. SBMV). In this group of capsids, the subunit interfaces remain at similar locations with respect to the CIAU. In addition, individual capsids in this group belong to different virus families and will

Table III
P vs. S P1 P2 P3 P4 P vs. N P1 P2 P3 P4 N vs. S N1 N2 N3 N4

Pairwise Inter-Family S-Scores for Three Families of Virus Capsidsa S1 0.18 0.23 0.23 0.23 N1 0.19 0.21 0.22 0.21 S1 0.64 0.59 0.61 0.62 S2 0.17 0.23 0.23 0.23 N2 0.20 0.23 0.22 0.22 S2 0.64 0.60 0.62 0.62 S3 0.17 0.22 0.22 0.23 N3 0.22 0.24 0.24 0.24 S3 0.60 0.59 0.60 0.61 S4 0.17 0.23 0.24 0.23 N4 0.20 0.24 0.24 0.24 S4 0.66 0.62 0.64 0.63 S5 0.55 0.55 0.55 0.56 S5 0.16 0.23 0.23 0.23

a Parvoviridae (P): 1dnv (P1), 1c8d (P2), 1c8h (P3) and 1k3v (P4), Sobemoviridae (S): 1smv (S1), 1x35 (S2), 1f2n (S3), 4sbv (S4) and 1ng0 (S5), Nodaviridae (N): 1nov (N1), 2bbv (N2), 1f8v (N3) and fhv (N4).

have S-scores in the range of (0.600.3) (Table III). Finally, the third group will have S-scores below 0.30. This group contains capsids that display different triangulation numbers and belong to different virus families with similar or dissimilar subunit topologies (e.g., Parvoviridae vs. Sobemoridae/Nodaviridae, Table III).
S-score versus other structural similarity metrics

Here, we describe the correlation of the S-score with other metrics of structural similarity, namely, root mean square deviation (RMSD), TM-score12,13 and sequence identity. The pair wise TM-score analysis for intra-family structures suggest that capsids with S-scores in the range of 1.00 to 0.6 display the same topology in the core regions of the CP subunit [Fig. 4(b)]. TM-scores greater than 0.8 were observed for most of the intra-family capsid pairs except for the comparisons with the Densovirus (1dnv) of Parvoviridae. Not surprisingly, the intra-family comparisons have RMSD values 2 A [Fig. 4(a)]. There is, however, a clear dependence of the S-score on the sequence identity as well as on RMSD, shown by the intra-family analysis [Fig. 4(a,c)]. Even though there is a strong correlation between the S-score and the other metrics mentioned above, only the S-score estimates the similarity in the quaternary (structure) interactions, while the other metrics primarily suggest the similarity in the CPs primary and tertiary structure. Inter-family pairwise analysis of TM-score and RMSD versus S-score clearly illustrates the lack of similarity in the proteinprotein interactions (S-score), even though there is a good deal of structural similarity in the tertiary structure of the CP subunits [Fig. 4(d,e)]. How-

650

PROTEINS

Protein Interactions in Spherical Viral Capsids

Figure 4
Comparison of S-scores with other structural similarity metrics. S-score versus RMSD (left column), versus TM-score (middle column), and versus sequence identity percentage (right column) between pairs of spherical virus capsids from three families. Results from the intra family analysis are shown in the first row: Parvoviridae (diamonds), Sobemoviridae (squares), and Nodaviridae (asterisk). The results of the inter-family analysis are shown in the second row: Sobemoviridae vs Parvoviridae (triangles), Nodaviridae vs Parvoviridae (Xs), and Sobemoviridae vs Nodaviridae (circles). A-subunits of the T 5 3 capsids were used in the structural superpositions and sequence comparisons with the reference subunit of the T 5 1 capsids. Structural alignments were done using TM-align.12

ever, CPs forming the same T number capsids display greater S-scores, in agreement with the similarity in the quaternary organization as opposed to capsids with different T numbers [Fig. 4(f)]. Taken together this suggests that greater structural similarity in the tertiary structure of CPs is necessary but not sufficient to form similar quaternary structures.

Identification and role of sequence and space conserved residues

As an immediate application of the APODs, we identified the conserved interface residues in the Fw space among the capsids within a single family. This was Q k S achieved by calculating the matrix product Nij n Nij , k where Nij is a polar orthographic matrix (POM) representing the APOD of capsid k in a family (set) of n capsids. The resulting matrix, NS, represents spatially conserved locations of the quaternary interactions in all the capsids being considered. Only the nonzero matrix elements ij common to all n POMs will also be nonzero in

NS. Another way of looking at this is as if all the chosen APODs were stacked in register as layers and only the (Fw) locations common to all the layers within a certain cutoff interval were selected as the quaternary fingerprint. The above procedure was employed on structurally characterized members of the Nodaviridae family: 2BBV, 1NOV, 1F8V, and FHV. Figure 5(a-d) shows the APODs of the individual members in the stack. Figure 5(e) shows the nonzero cells (locations) that are common to all the APODs in the stack representing spatially conserved interface residues, a total of 58, among the four Nodaviruses. A subset (18) of these residues is also sequence conserved and remarkably, the majority of these sequence and space conserved residues (SSCs) surround the icosahedral symmetry axes [Fig. 5(f)]. Figure 6 shows an alignment of the amino acid sequence of the four CPs highlighting the locations of the 58 space conserved and the 18 sequence space conserved residues (SSCs). Only six of the 18 SSCs are present in all the A, B and C subunits (Fig. 6; stars with gray background). The remaining (12) residues are present
PROTEINS

651

M. Carrillo-Tripp et al.

Figure 5
Illustration of the APODs of the four Nodaviruses and the locations of the sequence and spatially conserved interface residues. 2BBV (a), 1NOV (b), 1F8V (c), and FHV (d) capsids of the Nodaviridae family. (e) Spatially conserved locations of the interface residues, which are common to all the four members. (f) A subset of residue locations in the panel (e), which are also sequence conserved among the four Nodaviruses.

in either two out of the three subunits or only in a single subunit. We hypothesize that the six SSCs common to all the three subunits might be of importance for self-assembly of Nodavirus capsids. Currently, there is no experimental data available on the role of these residues in the assembly of Nodaviruses. Figure 7 shows the locations of all (34) SSCs (11 in A-subunit, 13 in B-subunit and 10

in C-subunit) found in the CIAU of the Nodaviruses mapped on to the structure of the Black beetle virus. It is interesting to note that all the SSCs found in the icosahedral asymmetric unit lie nearly on a single plane and the majority of them are clustered near the icosahedral symmetry axes, particularly at the icosahedral fivefold and threefold axes [Fig. 7(a)].

652

PROTEINS

Protein Interactions in Spherical Viral Capsids

Figure 6
Sequence alignment of the four Nodaviruses, highlighting the space conserved residues (gray and black stripes) and sequence and space conserved residues (SSCs) (black stripes). The asterisks, semicolons and dots correspond to identical, similar and homologous residues in the alignment. The six SSCs that are common in all the three distinct (A/B/C) subunits are identified by the asterisk with the gray background.

PROTEINS

653

M. Carrillo-Tripp et al.

Figure 7
Illustrations highlighting the locations of sequence and space conserved interface residues (SSCs) in the Nodaviridae family mapped on to the 3Dstructure of black beetle virus (2BBV). (a) A tube representation of the icosahedral asymmetric unit of black beetle virus (PDB-ID: 2BBV), a view down the quasi threefold axis, showing the traces of the A(blue), B(red), and C(green) subunits and C-alpha locations of the SSCs (spheres of all colors) in the respective subunits. Residue (SSC) locations in the A, B, and C subunits are shown as blue, red and green spheres respectively while the orange spheres correspond to the six SSCs common to all the three distinct subunits. Perpendicular views showing the locations of the SSCs at the (b) AB, (c) BC, and (d) CA interfaces.

CONCLUSIONS APODs of viral capsids provide a normalized fingerprint of protein-protein interactions irrespective of the particle size or the triangulation (T) number. Hence, they can be used as road maps of quaternary interactions in spherical viral capsids to visualize the distribution and the relative extent of interactions with respect to the icosahedral symmetry axis. Furthermore, these fingerprints when represented in a matrix form facilitate the correlation of the fingerprints with one another. The similitude in these fingerprints represented as the S-score provides a measure of overall similarity in the inter subunit (quaternary) interactions between pairs or a group of capsids. The S-score accurately identifies the quaternary structure similarity or lack of it, which the other metrics (RMSD, TM-score, %Sequence Identity) fail to discern. A limited analysis of viral capsids in terms of the APODs and Sscores suggests that the intra family/genus comparisons

will have S-scores mostly in the range of 1.0 to 0.6. Inter-family capsid pairs, which exhibit similar subunitfold and the same capsid architecture (T-number) will have S-scores in the range of 0.6 to 0.3 and the capsid pairs with different T numbers and/or subunit tertiary structures display S-scores below 0.3. Sequence-conserved residues with the spatially conserved locations within a family of virus capsids may play an important role in the self-assembly of the member capsids. Exhaustive interand intra-family wide analysis of the other structurally known capsids is underway. The resultant APODs and their correlations will soon be available at the VIPERdb website (http://viperdb.scripps.edu). ACKNOWLEDGMENTS We thank Ian Borelli for help in devising some of the algorithms used in this work and maintaining VIPERdb up to date.

654

PROTEINS

Protein Interactions in Spherical Viral Capsids

REFERENCES
1. Choi YG, Dreher TW, Rao AL. tRNA elements mediate the assembly of an icosahedral RNA virus. Proc Natl Acad Sci USA 2002;99:655660. 2. Fox JM, Wang G, Speir JA, Olson NH, Johnson JE, Baker TS, Young MJ. Comparison of the native CCMV virion with in vitro assembled CCMV virions by cryoelectron microscopy and image reconstruction. Virology 1998;244:212218. 3. Hsu C, Singh P, Ochoa W, Manayani DJ, Manchester M, Schneemann A, Reddy VS. Characterization of polymorphism displayed by the coat protein mutants of tomato bushy stunt virus. Virology 2006;349:222229. 4. Sorger PK, Stockley PG, Harrison SC. Structure and assembly of turnip crinkle virus. II. Mechanism of reassembly in vitro. J Mol Biol 1986;191:639658. 5. Zhao X, Fox JM, Olson NH, Baker TS, Young MJ. In vitro assembly of cowpea chlorotic mottle virus from coat protein expressed in Escherichia coli and in vitro-transcribed viral cDNA. Virology 1995;207:486494. 6. Reddy VS, Giesing HA, Morton RT, Kumar A, Post CB, Brooks CL, III, Johnson JE. Energetics of quasiequivalence: computational analysis of protein-protein interactions in icosahedral viruses. Biophys J 1998;74:546558. 7. Damodaran KV, Reddy VS, Johnson JE, Brooks CL, III. A general method to quantify quasi-equivalence in icosahedral viruses. J Mol Biol 2002;324:723737. 8. Shepherd CM, Reddy VS. Extent of protein-protein interactions and quasi-equivalence in viral capsids. Proteins 2005;58:472477. 9. Bahadur RP, Rodier F, Janin J. A dissection of the protein-protein interfaces in icosahedral virus capsids. J Mol Biol 2007;367:574590. 10. Shepherd CM, Borelli IA, Lander G, Natarajan P, Siddavanahalli V, Bajaj C, Johnson JE, Brooks CL, III, Reddy VS. VIPERdb: a relational database for structural virology. Nucleic Acids Res 2006;34(Database issue):D386D389. 11. Reddy VS, Natarajan P, Okerberg B, Li K, Damodaran KV, Morton RT, Brooks CL, III, Johnson JE. Virus particle explorer (VIPER), a website for virus capsid structures and their computational analyses. J Virol 2001;75:1194311947. 12. Zhang Y, Skolnick J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Res 2005;33:23022309. 13. Zhang Y, Skolnick J. Scoring function for automated assessment of protein structure template quality. Proteins 2004;57:702710. 14. Rossmann MG, Abad-Zapatero C, Murthy MR, Liljas L, Jones TA, Strandberg B. Structural comparisons of some small spherical plant viruses. J Mol Biol 1983;165:711736.

15. Liljas L, Unge T, Jones TA, Fridborg K, Lovgren S, Skoglund U, Strandberg B. Structure of satellite tobacco necrosis virus at 3.0 A resolution. J Mol Biol 1982;159:93108. 16. Abad-Zapatero C, Abdel-Meguid SS, Johnson JE, Leslie AGW, Rayment I, Rossmann MG, Suck D, Tsukihara T. Structure of southern bean mosaic virus at 2.8 A resolution. Nature (London) 1980;286:3339. 17. Harrison SC, Olson AJ, Schutt CE, Winkler FK. Tomato bushy stunt virus at 2.9 A resolution. Nature 1978;276:368373. 18. Simpson AA, Chipman PR, Baker TS, Tijssen P, Rossmann MG. The structure of an insect parvovirus (Galleria mellonella densovirus) at 3.7 A resolution. Structure 1998;6:13551367. 19. Simpson AA, Chandrasekar V, Hebert B, Sullivan GM, Rossmann MG, Parrish CR. Host range and variability of calcium binding by surface loops in the capsids of canine and feline parvoviruses. J Mol Biol 2000;300:597610. 20. Simpson AA, Hebert B, Sullivan GM, Parrish CR, Zadori Z, Tijssen P, Rossmann MG. The structure of porcine parvovirus: comparison with related viruses. J Mol Biol 2002;315:11891198. 21. Bhuvaneshwari M, Subramanya HS, Gopinath K, Savithri HS, Nayudu MV, Murthy MR. Structure of sesbania mosaic virus at 3 A resolution. Structure 1995;3:10211030. 22. Sangita V, Lokesh GL, Satheshkumar PS, Saravanan V, Vijay CS, Savithri HS, Murthy MR. Structural studies on recombinant T 5 3 capsids of sesbania mosaic virus coat protein mutants. Acta Crystallogr D Biol Crystallogr 2005;61:14021405. 23. Qu C, Liljas L, Opalka N, Brugidou C, Yeager M, Beachy RN, Fauquet CM, Johnson JE, Lin T. 3D domain swapping modulates the stability of members of an icosahedral virus group. Structure 2000;8:10951103. 24. Tars K, Zeltins A, Liljas L. The three-dimensional structure of cocksfoot mottle virus at 2.7 A resolution. Virology 2003;310:287 297. 25. Zlotnick A, Natarajan P, Munshi S, Johnson JE. Resolution of space-group ambiguity and structure determination of nodamura virus to 3.3 A resolution from pseudo-R32 (monoclinic) crystals. Acta Crystallogr D Biol Crystallogr 1997;53:738746. 26. Wery JP, Reddy VS, Hosur MV, Johnson JE. The refined three dimensional structure of an insect virus at 2.8 A resolution. J Mol Biol 1994;235:565586. 27. Tang L, Johnson KN, Ball LA, Lin T, Yeager M, Johnson JE. The structure of pariacoto virus reveals a dodecahedral cage of duplex RNA. Nat Struct Biol 2001;8:7783. 28. Fisher AJ, Johnson JE. Ordered duplex RNA controls capsid architecture in an icosahedral animal virus. Nature 1993;361:176179.

PROTEINS

655