J. Mol. Biol. (1990) 211, 763401

Analysis of the Structure of a Common Cold Virus, Furnan Rhinovirus 14, Refined at a Resolution of 3-O A Edward Arnold’~’ and Michael G. Rossmann’ Purdue

‘Department of Biological Sciences University, West Lafayette, IN 47907, U.S.A.

?Yenter for Advanced Biotechnology and Medicine, and Department of Chemistry Rutgers University, P.O. Box 759, Piscataway, NJ 08855-0759, U.S.A. (Received 2 May 1989; accepted 21 August 1989) Human rhinovirus 14 has a pseudo T = 3 icosahedral structure in which 60 copies of the three larger capsid proteins VPl, VP2 and VP3 are arranged in an icosahedral surface lattice, reminiscent of T = 3 viruses such as tomato bushy stunt virus and southern bean mosaic virus. The overall secondary and tertiary structures of VPl, VP2 and VP3 are very similar. The structure of human rhinovirus 14, which was refined at a resolution of 3.0 A [R = 0.16 for reflections with F > 30(F)], is here analyzed in detail. Quantitative analysis of the surface areas of contact (proportional to hydrophobic free energy of association) supports the previously assigned arrangement within the promoter, in which interactions between VP1 and VP3 predominate. Major contacts among VPl, VP2 and VP3 are between the b-barrel moieties. VP4 is associated with the capsid interior by a distributed network of contacts with VPl, VP2 and VP3 within a protomer. As the virion assembly proceeds, the solvent-accessible surface area becomes increasingly hydrophilic in character. A mixed parallel and antiparallel seven-stranded sheet is composed of the /3C, /?H, BE and /3F strands of VP3 in one pentamer and /?A, and PA, of VP2 and the VP1 amino terminus in another pentamer. This association plays an essential role in holding pentamers together in the mature virion as this contact region includes more than half of the total short non-bonded contacts between pentamers. Contacts between protomers within pentamers are more extensive than the contacts between pentamers, accounting in part for the stability of pentamers. The previously identified immunogenic regions are correlated with high solvent accessibility, accessibility to large probes and also high thermal parameters. Surface residues in the canyon, the putative cellular receptor recognition site, have lower thermal parameters than other portions of the human rhinovirus 14 surface. Many of the water molecules in the ordered solvent model are located at subunit interfaces. A number of unusual crevices exist in the protein shell of human rhinovirus 14, including the hydrophobic pocket in VP1 which is the locus of binding for the WIN antiviral agents. These may be required for conformational flexibility during assembly and disassembly. The structures of the B-barrels of human rhinovirus 14 VPl, VP2 and VP3 are compared with each other and with the southern bean mosaic virus coat protein.

1. Introduction

and Background

Human rhinovirus 14 (HRV14) is a member of the picornavirus family of animal viruses (Rueckert, 1986). Rhinoviruses account for roughly half of common colds in humans, are characterized by having a large number of serotypes ( > 100) and are unstable at acidic pH. The HRV14 particle is composed of an icosahedrally symmetric protein shell, consisting of 60 copies of each of four distinct polypeptide chains, surrounding a single-stranded 0022~2836/QO/040763-39

$03.00/O

RNA genome. The four picornaviral coat proteins VPl, VP2, VP3 and VP4 have relative molecular masses, M,, of 32,000, 29,000, 26,000 and 7000, respectively. The RNA strand has approximately 7500 bases and contributes roughly 30% of the dry mass of the picornavirion. The structure of HRV14 was determined by a combination of isomorphous and molecular replacement phasing (Rossmann et al., 1985; Arnold et al., 1987a). A number of other picornavirus structures have been solved: human rhinovirus I A (HRVlA;

763

0

1990 Academic

Press Limited

764

E. Arnold and M. G. Rossmann

Kim et al., 1989), poliovirus 1 Mahoney (Hogle et al., 1985) and poliovirus 3 Sabin (Filman et al., 1989) representing the enterovirus class, Mengo virus (Luo et al., 1987) representing the cardiovirus class and a strain of foot-and-mouth disease virus (FMDV; Acharya et al., 1989) representing the aphthovirus class of picornaviruses. For HRV14, the phases beyond 5 A resolution were obtained from molecular replacement phase extension and refinement, relying on the presence of 20 copies of the viral protomeric unit within the crystallographic asymmetric unit. The native data used for structure solution and refinement were obtained at the Cornell High Energy Synchrotron Source (CHESS) and the calculations were performed using the Purdue University CYBER 205 supercomputer. The structure of HRV14 has been refined at a resolution of 3.0 A (1 A = 01 nm) by a restrained least-squares technique in which phases from the molecular replacement procedure were incorporated as observations. The function minimized was: T w,,[(Ao- A,)* + (B,-B,)‘]

+geometric

terms,

where A, and & are the real and imaginary parts of the structure factors whose phases were derived from molecular replacement real-space averaging, and A, and B, are the corresponding terms structural model computed from the current (Arnold & Rossmann, 1988). The refined atomic model includes 6268 protein atoms in 804 amino acid residues in proteins VPl, VP2, VP3 and VP4 within a viral protomeric unit, as well as 272 solvent water molecules and two putative ions. Individual isotropic thermal parameters were varied for all atoms in the model, as were the occupancy factors for solvent water molecules (see Solvent Structure). The quality of the model is indicated by the low R-factor of 0.16 for all data between 6 and 3 A resolution with F > 3a(F). The structures of the coat proteins and their arrangement within the viral capsids for all small icosahedral RNA viruses studied to date are strikingly similar. The capsids of the plant viruses tomato bushy stunt virus (TBSV; Harrison et al., 1978) southern bean mosaic virus (SBMV; AbadZapatero et al., 1980), turnip crinkle virus (TCV; Hogle et al., 1986) and the insect black beetle virus (BBV; Hosur et al., 1987) are composed of 180 covalently identical protein subunits arranged in a T = 3 icosahedral surface lattice (Caspar & Klug, 1962). The icosahedral asymmetric unit contains three polypeptide chains, which, in each case, have a very similar fold and are arranged about a quasi3-fold axis of symmetry. In the plant cowpea mosaic virus (CPMV; Stauffacher et al., 1987) and beanpod mottle virus (BPMV; Chen et al., 1989), two major coat proteins contain three similar folding domains, forming an overall arrangement that is again reminiscent of the T = 3 viruses. The folding motif found in all cases is an eight-stranded antiparallel /?-barrel characterized by two four-stranded sheets (the BIDG and CHEF sheets, see Notation), which

form a sandwich-like structure. The viral proteins VPl, VP2 and VP3 of pieornaviruses exhibit a similar fold (Fig. 1). Sixty copies of each of the three proteins form a pseudo T = 3 arrangement (P = 3) closely similar to the organization of the covalently identical proteins of the true T = 3 viruses such as SBMV. The overall packing of VPl, VP2 and VP3 in the icosahedral HRV14 capsid is illustrated in Figure 2. The three distinct proteins have become differentially ornamented, endowing picornaviruses with complex surface characteristics that allow them to thrive in the face of immune pressure. The extent of the similarity among the three-dimensional structures of all of the small RNA icosahedral virus coat proteins suggests that they are derived from a common ancestor (Rossmann et al., 1985, 1986; Rossmann, 1987). Results from studies of mutants of HRV14 that were able to “escape” neutralization by monoclonal antibodies indicated the presence of four distinct regions to which neutralizing antibodies attach (Sherry & Rueckert, 1985; Sherry et al., 1986; Rossmann et al., 1985). These antibody attachment sites are on protruding portions of the viral surface, and are excluded from a series of depressions on the surface that might function as the cellular receptor recognition site. The “canyon hypothesis” proposes that the common denominator for picornaviruses within a receptor group can reside in this hidden area (inaccessible to presumably blunt antibodies), while the protruding regions containing the antibody-binding sites are hypervariable, permitting serotype diversity. The site of attachment of some antiviral agents, which inhibit uncoating of picornaviruses, has been located in the hydrophobic interior of the VP1 p-barrel (Smith et al., 1986; Badger et aZ., 1988, mutants resistant to these 1989). HRVl4 compounds confirm that binding in this pocket is essential for antiviral activity (Badger et al., 1989). Assembly of picornavirions proceeds from 6 S protomers consisting of VPl, VP3 and VPO, via pentamers of protomers, to mature virions. Assignment of the protomeric unit as that illustrated in Figure 2 was suggested by the intertwining of the amino and carboxyl ends of the proteins in the mature virion, in particular those of VP1 and VP3 (Rossmann et al., 1985). The 6 S protomers are woven into pentamers by the VP3 amino ends, which form a P-cylinder surrounded by the amino ends of VP4. All processing cleavages of the structural proteins are accomplished by virally coded proteases, except for the separation of VP0 into VP2 and VP4, which occurs at the stage of virion maturation. The placement of serine 10 in VP2 adjacent to the carboxyl terminus of VP4, and its among different picornaviruses, conservation suggested a possible autoproteolytic mechanism in which an RNA base could act as the catalyst by converting the serine hydroxyl to the more nucleophilic alkoxide form (Rossmann et aZ., 1985; Arnold et al., 1987b). The relative placement of the amino and carboxyl termini within the protomeric unit,

Rq’ined Structure

sf HRV14

765

VP1

VPI

VP2

VP3

VP3

RNA lnterlor

(b)

Figure 1. Representations of the structures of HRV14 VPl, VP2 and VP3. (a) Drawings (based on a drawing by Jane Richardson for SBMV) that show the overall folding and topology of the proteins. (b) C” tracings of the proteins displayed in the same orientations as in (a).

and the ubiquity of the p-barrel fold, suggests that the o-barrels probably fold during the formation of the protomer, prior to proteolytic processing. The intimate network of inter-barrel contacts presumably forms during the initial folding of the polyprotein prior to cleavage into the component capsid proteins and persists throughout virion morphowith the genesis (Arnold et al., 19876). Comparison structures of comoviruses (Stauffacher et aZ., 1987; Chen el al., 1989), where two of the /?-barrels remain within the same polypeptide chain, supports the hypothesis that the picornavirus polyprotein folds

into a protomer conformation before separate proteins as a way of directing process. 2.

cleavage into the assembly

Notation

Residues in the viral proteins of HRVl4 have been numbered X Y Y Y, where X corresponds to the viral protein number (VPX) and Y Y Y is the residue number within chain X. For example, proline 83 of VP2 is numbered 2083 (and may be referred to as P2083 or Pro2083 in different places in

766

E. Arnold and M. G. Rossmann TBSV, tomato bushy stunt virus; and TCV, turnip crinkle virus. The picornavirus sequence alignments referred to throughout the paper are taken from Palmenberg (1989).

3. Primary Structure

Figure 2. A diagram illustrating the overall arrangement of VPl, VP2 and VP3 on the HRV14 surface and the symmetry designations used in the text. The protomerit unit that cousiets of 1 copy each of VPI, VP2, VP3 and VP4 probably corresponding to the 6 S assembly unit is outlined with thickened lines. The reference protamer is cross-hatched and the symmetry relationships of adjacent protomers are indicated (see Notation).

the text). Water molecules are given four-digit names beginning with 5 (e.g. Wat5004). Residues in the reference protomer (see Fig. 2) are not given explicit symmetry designations. The atoms and amino acids in adjacent symmetryrelated protomers are indicated by X Y Y Y ( x P/Q), where P/Q designates the symmetry operation to be applied to the reference protomer. For example, residue X Y Y Y ( x l/5) is in an adjacent protomer in a pentamer, related to the reference protomer by a 72” counter-clockwise rotation about the B-fold axis of icosahedral symmetry. Likewise 215, 315 and 415 indicate protomers related to the reference protomer by rotations of 144”, 216” and 288”, respectively. The adjacent protomer related to the reference protomer by a 2-fold axis of icosahedral symmetry is designated by l/2 (180” rotation) and the 3-foldrelated protomers by l/3 and 213 as shown in Figure 2. For example, a contact between the y-hydroxyl of serine 2054 in the reference protomer and the carbonyl oxygen of proline 2056 in an icosahedrally 2-fold-related protomer is indicated by Ser2054 Oy-Pro2056 0 ( x l/2). The eight-stranded b-barrels in VPI, VP2 and VP3 can also be considered as b-sandwich structures in which two four-stranded antiparallel B-sheets consisting of PC, fiH, BE and /IF, and /lB, /II, /ID and PC: are packed face-to-face. The two fourstranded sheets are abbreviated as CHEF and BIDG, respectively, when discussed in the text. The abbreviations of viruses used in the text are collected here: BBV, black beetle virus; BPMV, beanpod mottle virus; CPMV. cowpea mosaic virus; FMDV, foot-and-mouth disease virus; HRV14, human rhinovirus type 14; HRVlA, human rhinovirus type IA; SBMV, southern bean mosaic virus;

The sequence of the viral RNA of HRV14 was determined independently by two groups (Stanway et al., 1984; Callahan et al., 1985). The sequence of the strain of HRV14 used for crystallographic studies at Purdue was consistent with that reported by Stanway et al. (1984) and was confirmed by sequencing of the capsid protein genes at Wisconsin in conjunction with immunological studies (Sherry & Rueckert, 1985; Sherry et al., 1986). The one place in the experimental electron density map that was inconsistent had been originally obviously sequenced as isoleucine, but appeared to look like leucine in the original structure determination. In difference Fourier maps computed with respect to later data sets, there was a large peak at this site showing that the residue had changed, probably to valine (residue 170 in VP2, Smith et al., 1986). The experimental electron density for position 2170 in the native HRV14 map is shown in Figure 3. It is possible to account for the mutation sequence Be -+ Leu +Val by a single base change in the codon AUA-+CUA+GUA (Smith et al., 1986). This base may be at a sensitive site in the virion RNA structure, permitting change at this apparent “hot spot”. Alternatively, it may be located at a region of the RNA secondary structure that is difficult for the viral RNA replicase to copy accurately, such as a complex secondary or tertiary structural region. The electron density corresponding to the sidechains of a number of amino acids within the coat protein shell was poorly defined giving rise to high temperature factors for these atoms, perhaps due to some disorder. The most mobile residues are located almost exclusively on the interior and exterior surfaces of the virion (Table 1). Most of these residues have extended aliphatic side-chains. A high proportion of the most disordered side-chains belong to lysine residues. There are no disulfide bonds present in the mature HRV14 capsid, although there are nine cysteine residues present in the protomeric unit. Oddly enough, the major binding site (A site) of the KAu(CN)~ reagent used in the initial phasing stages is located in a poorly ordered region of the protein at the icosahedral 3-fold axis of symmetry (Rossmann et al., 1985). The A site is located close to Cys2007. The B and C KAu(CN), sites are also associated with cysteine residues (Table 2). The B site has, in addition to Cys1069, a neighboring Ag1073, possibly serving as a counterion for the halide-like auric dicyanide reagent (Norne et al., 1975). A plausible mode for the detailed binding at site B might involve displacement of Wat5128 with concomitant liganding by Cys1069 Sy and Arg1073 N”

Refined Structure of HRV14

767

N2172

N2172

7

7

Figure 3. The electron density from the native 308 A molecular replacement map (Arnold & Rossmann, 1988) is plotted in the vicinity of residue 2170. This residue was originally sequenced as isoleucine, but the appearance of the density supports the assignment as leucine.

and Arg1073

N”‘.

Cys2061

from Cys2248 Sv and, thus, they form a transient disulfide assembly; both these cysteine in all but HRV2 and HRV89

Sv is only

6.8 A away

it is conceivable that link during folding or residues are conserved of the available

virus, poliovirus and coxsackie virus sequences. The internal position of the Au sites shows that there must be considerable flexibility in the protein structure to permit ion diffusion.

rhino-

Three of the four amino termini

are not visible

Table 1 The most $exible and the most rigid residues in the HRV14 capsid as indicated by extreme values of isotropic thermal parameters (B values) A. The most rigid residues (those with at least 1 atom with B < 9 A ‘) G1040, W1253, V2032, R2103, T2176, T2198, P2239, L3002, Y3117, F3210,

A1041, A1257, V2033, T2107, L2181, 12199, 12240, P3030, T3118, 13211,

11. The most jerible Q1029, R1094 K1236’ G2008’ &2262’ K3061 %i-& -,

T1053, PI258 L2042, V2108, L2182, V2200, V2242, T3031, K3126, A3213

V1077,

K1114,

F1119,

T1120,

Y1190,

F1200,

Y1201,

T2053, H2109, 12183, 12201, T2243, P3032, Y3131,

52054, V2110, F2184, P2202, P2246 13036, L3147,

F2063, L2123, F2185, 52206,

W2080, V?124, F2186, V2207,

K2081, V2125, Q2187, P2208,

L2082, H2130, 12189, 12209,

L2101, L2170, A2197, H2215,

L3043, T3149,

L3044, 13053, H3150, V3152,

Y3103, Y3104, W3167, 13196,

residues (those with at least 1 atom with B > 30 AZ)

K1030 %@i’ G, Y2009 -’ E3063 i%?i:

D1086, K1097 m: 52010,

A1087, D1138, T1275, D2011,

T1088, G1089, 51139, N1145 El276 K1280’ R2012’ i%?& -2 -,

R3075, E3236 Q4043, S4044, -,L4045

54046 -9

11090, Kll61 m’ D2067 -9

D1091 E1162,

N1092, H1093, E1210, E1234,

E2136,

R2152 -1

M4047, L4067,

N4068

E2161,

C. Missing amino-terminal residues (poorly visible or invisible in electwn density maps) VP1 GLGDELEEVI VEKTKQ (100-1016) VP2 SPNVEAC (2001-2007) VP4 GAQVSTQKSG SHENQNILTN GSNQTFTV (400-4028) Underlined residues are associated with the weakest electron density, indicating possible disorder. These residues reside almost exclusively at solvent-accessible surf&es: on interior RNA interface (K1030, E2040, G2008, Y2009, R2012, L4045, 54046); on exterior (D1091, R1094, E1095, K1097, K1161, E1276, K1280, K1283, D2067, R2152, Q2262, K3061, E3063); in the Bfold channel (N1145, K1236).

in

768

E. Arnold and iIf. G. RO88mUnn

Table 2

Table 3

Environment of KAu(CN)~ binding Bites in HRV14 heavy-atom derivative as interpreted from the refined structure

Re8idW?8in HRV14 that have 8U8peCtmuin chain conform4ztbn8 baaed on deviation from structures that have closely matching confwmtions in a data base of well-refined protein structures

Environment Site

Residue

of gold sites

Contacting atom

Judgement

Distance (A) from Au atom Residue

A B

C

Disordered region of VP2 N terminus Cys2007 nearby but disordered SY Cys1069 0 Cys1069 Wat5128 W&t5261 : Arg1073 ( x 4/5) $2 Arg1073 ( x 4/5) Asn3027 06’ 0 Pro3026

25 32 1-Q 4.0 4.5 4.3 4.1 4.3

Cys2248 Cys2248 Pro2039 wat5174 cys2061

2.0 2.8 4% 2.8 58

sy N 0 0 SY

the electron density (Table 1). The amino terminus of VP3 is very clear and appears to have a covalent modification (Arnold & Rossmann, 1988) that has been interpreted as a possible N-methylation. The size and shape of the unassigned density indicates that it probably consists of a single non-hydrogen atom, possibly an oxygen atom (for a nitrosyl or hydroxylamine modification), a carbon atom (for a methyl or methyleneimine modification) or a nitrogen atom (for a hydrazinyl modification). Alternatively, the modification could be an N-formyl group. An N-methyl group has been identified as a post-translational modification in a number of other proteins, primarily in ribosomal proteins in which a-amino groups of methionine (Chen et al., 1977; Brauer & Wittmann-Liebold, 1977) and alanine (Chen et al., 1977) have been found to be monomethylated. Multiple methylations of amino termini have been identified in a ribosomal protein (Lederer et al., 1977) and in a protozoan cytochrome (Pettigrew & Smith, 1977). An unusual arrangement of density was located just below the protein shell around the 5-fold axis of symmetry, near the amino terminus of VP3 and surrounded by the extra residues of the VP4 amino terminus (4025 through 4028 were weakly visible in the electron density map). This density, substantially weaker than that of the ordered protein, consists of a ball of density at the radius of residues 4025 to 4029, below which are two plates of density perpendicular to and centered about the 5-fold axis of icosahedral symmetry. Although the identity of this portion of the virion is not clear, by analogy with poliovirus 1 Mahoney (Chow et al., 1987), it seems possible that this density could be associated with myristylation of the VP4 amino terminus. An analogous feature is present in the Mengo virus

Ser4044 Ser2256 Ala3184 Ala4041 Met1058 Phe2OQ6 Ser4046 AsnlO61 Gly1221 Gly2150 Gly2137 His2149 Pro3008 Pro3037 Gln3192 Gly2019 Asp2057 Lys1283 Pro2056 Thr3193 Pro2226 Ser3010 Gly1222

Deviation from the a rg” 333 332 330 323 319 299 297 296 286 286 284 282 2.79 2.75 266 264 2.63 262 259 256 2.54 251 250

Probably wrong

Indeterminate

Probably correct

x X X X X X X X X X X X X X X X X X X x x X X

These positions were reinspected and judged for likelihood of correctness based on the appearance of the electron density map and local stereochemistry. For instance, although Ala3184 has an “unusual” main chain conformation, the carbonyl direction is clearly indicated by the electron density, and a plausible hydrogen bond exists between the carbonyl oxygen of Ala3184 and the hydroxyl oxygen of Tyr3131 at a distance of 25 A.

density map at roughly the same place (Krishnaswamy & Rossmann, 1990; Luo et al., 1987). Although.there is a single-stranded RNA genome with 7500 bases and a small genome-linked protein inside each viral particle, very little of the RNA structure is visible in the electron density of either the molecular replacement or the difference Fourier maps based on phases calculated from the atomic model. Since the 7500 nucleotide RNA strand cannot have icosahedral symmetry, it is unlikely that the RNA will show strong continuous features in an inherently icosahedrally averaged experimental system, such as the HRV14 cubic crystals. However, it is conceivable that there are repeating motifs of RNA secondary structure that can be induced to-have an approximate icosahedral order by interaction with the inner surface of the icosahedrally symmetric protein shell, as is the case for BPMV (Chen et al., 1989) and CPMV (Stauffacher et al., 1987). There is unassigned planar density at a distance of 3.4 A from tryptophan 2038. This density could correspond to an aromatic ring that is stacked with the tryptophan indole ring in some of

Refined Structure of HRV14 the icosahedral asymmetric units. Trials showed that a purine base can be well accommodated by this density. None of the missing amino-terminal residues in VPI, VP2 and VP4 is aromatic (Table l), supporting the possible assignment of this density as a purine base of the RNA. Furthermore, the site of this density corresponds to the ordered RNA structure in BPMV. Other unassigned density on the inner surface appeared as relatively strong peaks with no contacts of less than 4 A with polar groups in the ordered protein shell and, thus, could be due to phosphate groups of the RNA. Alternatively, these peaks might correspond to partially ordered residues at the amino termini of VPl, VP2 and VP4. As a critical test of the quality of the model, a comparison was made to a data base containing 32 well-refined protein structures using a concept and program developed by A. Jones (Jones & Thirup, 1986). This procedure takes sets of five residues of HRV14 at a time and searches for the 20 best fits to the C” atoms. Tt then computes the distance between the carbonyl oxygen atom of the third residue of HRV14 with each of t’he top 20 leastsquares fitted fragment’s from the data base. When the root-mean-square (r.m.s.) value of these distances is greater tha.n 2.5 A or so, then this is a good indication that something may be in error with the HRV14 structure. There were 23 such cases (Table 3) and another 28 cases with an r.m.s. value between 2.5 A and 2.0 A. Inspection of the top 23 cases showed that some of these are at locations where the main-chain density was poor ambiguous, often associated with glycine residues. The results from the data base suggest that the peptide bond in these cases should sometimes be flipped by 180”.

4. Secondary Structure and Main Chain Hydrogen-bonding Patterns The classification of secondary structure in VPl, VP2 and VP3 of HRVI4 is given in Table 4. The assignments have been based largely on the corresponding main chain hydrogen-bonding diagrams (Fig. 4(a) through (d)). Hydrogen bonds were postulated whenever t.he N-O distance was less than 3.5 A and the C-O-N angle was greater than 110”. Each of the three viral proteins contains a /?-sandwich structure in which the two four-stranded antiparallel sheets CHEF and BIDG lay face-to-face to form the familiar eight-stranded /?-barrel (Fig. 1). There is relatively little a-helix in the HRV14 capsid structure; clA is the only helix of greater than two turns that is well conserved among the three proteins. Some vestiges of aB are present in each of VPl, VP2 and VP3, but in no case is aB a very regular a-helix. Other helical stretches include aZ of VPl, VP2 and VP3. All three helical sections aZ, aA and aB are involved in subunit-subunit interactions, with aZ and aA at 2-fold and pseudo-2-fold symmetry sites. Helix aB interacts with an opposing BTDG sheet in a neighboring subunit.

769

Table 4 Assignment of secondary structural units in HRV14 Inclusive Structural

unit

residue numbers in

VP1

VP2

17-65 6G-72 75-84

12-18 21-28 29-55 56-61 64-7 1

fi-Cylinder “,2 Atino-terminal

l-7 arm

;i

100-104 110-119 120-135 146-153

m ji Puff loop 1 Puff loop 2 ;i f2100p PA FMDV loop 5; Carboxy-terminal

VP3

tail

165-170 174-180 182-188 189-196 197-206 201-220 222-229 237-254 255-289

77-82 89-99 lW112 119-127 128-153 158-176 177-184 185-191 195-201 202-209 210-212 218-229 238-253 254-262

8-41 4249 51-53, 69-71 54-68 (knob) 81-85 95-104 105-118 124-132 141-148 149-155 159-165 166-173 174-176 185-196 205-221 222-236

The main chain conformational angles 4 and ij for the four proteins in the HRV14 protomer are plotted in Figure 5(a) and (b). These Ramachandran diagrams have been divided to separate glycyl and non-glycyl residues and the distribution is superposed onto theoretical energy contours for polyglytine and poly-L-alanine, respectively. The main chain conformational angles 4 and @ are good indicators of the quality of the model, since they are not restrained during the refinement process. In Table 5, the conformational angles are given for non-glycyl residues that have positive 4 angles. The total of 21 such cases represents 3y; of the nonglycyl residues. Five of these residues are located in the vicinity of the neutralizing immunogenic sites and, when the 31 glycine residues are included in this list, each of the four immunogenic sites is represented. Glycine 1214 and glycine 1222 are associated with the conformational rearrangements that occur in the FMDV loop (Table 4) upon the binding of the WIN antiviral agents (Smith et al., 1986; Badger et al., 1988, 1989). All but one of the o main chain conformational angles was restrained to be in the trans configuration (w = 180”) during refinement. A cis-proline was detected at position 2083 (the experimental density for this area is shown in Fig. 6). This proline and its immediate neighbors in the sequence (WKLPDAL) are entirely conserved among the polioviruses, rhinoviruses and coxsackie viruses, and thus is probably a cis-proline in each of these viruses. Apparently in Mengo virus, at the analogous position 2085, there is also a cis-proline ((Krishnaswamy & Rossmann, 1990).

770

E. Arnold and M. G. Rossmann

HRV14

(0) Fig. 4.

A large number of /3-turns or “reverse turns” are found in the HRV14 capsid proteins (Table 6). A number of the corners at the wedge-shaped end of the eight-stranded antiparallel p-barrels, however, do not contain a /?-turn, as defined by Crawford et al. (1973). Preferences for certain residues to appear at characteristic positions include: proline at positions 1 and 2; glycine at position 3; and serine at position 4. Positional preferences and other aspects of B-turns have been reviewed (Crawford et al., 1973; Chou 6 Fasman, 1977). The 3,,-helices that appear in the HRV14 capsid protein structures are listed in Table 6. A number of secondary structures involving contri-

butions from neighboring proteins are of particular interest. The unusual “B-cylinder” structure found at the VP3 amino terminus consists of five strands that twist together to form a cylindrically parallel p-sheet (Fig. 7). Each strand makes eight hydrogen bonds, four with each adjacent strand, in this unusual arrangement. Since the adjacent strands are contributed by VP3 molecules in adjacent protomers within a pentamer, the p-cylinder could play a role in pentamer formation during assembly, pentamer stabilization and pentamer breakdown during disassembly. The /lAl/?Az antiparallel sheet in VP2 has associations with two strands in an adjacent pentamer: /?A, forms a small antiparallel

Refined Structure of HRV14

771

(b) Fig. 4.

p-sheet with residues 1026 to 1028 of the VP1 amino terminus, and PA2 forms a parallel /I-sheet with residues 3150 to 3152 of VP3 OF (see Intersubunit Contacts). A stretch of the VP1 amino terminus from residues 1031 to 1038 makes a total of six main chain hydrogen bonds to main chain pieces of VP2 (2188), VP3 (3159 to 3162) and VP4 (4063). These regions of the polyprotein have been translated prior to VPl, and may provide a folding template for this otherwise “disorganized” portion of VPl. Although the HRV14 capsid proteins do not contain any standard P-bulges (Richardson et al., 1978), there are a number of wide /?-bulges that are recurrent in VPl, VP2 and VP3 (Table 7 and

Fig. 4(a) through (c)). In wide P-bulges, the residues at position 2 of strands 1 and 2 and position 3 of strand 3 are not hydrogen-bonded to each other, but an antiparallel sheet that continues from either end is put out of register by one residue. The overall consequence is similar to b-bulges, in that the sidechains for the central residues at positions 2 and 3 in strand 2 are forced to point to the same side of the sheet. The potential importance of these wide /?-bulges is intimated by the apparent conservation from one viral capsid protein to another: the BE-/IH wide /?-bulge occurs at an analogous position in VP2 and VP3, and the /ID-jI1 wide /I-bulge occurs at an analogous position in VP1 and VP3. The JIB-/II wide

n

60T 8-8

D

I(

knob

3-1 (d) Figure 4. A representation of the main chain hydrogen-bonding patterns of the HRV14 capsid proteins. Contacts with adjacent proteins are indicated, where the symmetry designation is given only if the other protein is not within the reference protomer. Residues that define the NIm sites in the escape mutation assays are circled twice. (a) Main chain hydrogen-bonding diagram for HRV14 VPl. (b) Main chain hydrogen-bonding diagram for HRV14 VP2. (c) Main chain hydrogen-bonding diagram for HRV14 VP3. The 5 segments that define the b-cylinder are shown to indicate the nature of this unusual structural unit. (d) Main chain hydrogen-bonding diagram for HRV14 VP4. The plausible hydrogen bond between the hydroxyl group of Ser2010 and the carboxyl group of Asn4068 is indicated.

773

Refined Structure of HRV14

180

I .----------:---------I

_-_-_--_-_ .

x

,---------:----------:-----------

I I

, I I I 4

I I

I 0 I I !

I

la I I I I

I I I

-180

180 (a)

180

---L-_--~hyy-g-_-$ ______-_/_-_ c-q; ;c): 3

.____ ----,----I

x

I

_ _

---..---:----------

---..----;--- iqy?q _ - ;y-

i :

i “1

.

. .

x: :

N

I80

-180 (b) Figure 5. Plots (Ramachandran diagrams) HRV14 protomer. (a) Distribution of main contours of the calculated energies for a (b) Distribution of main chain conformational energies for a polypeptide containing glycine

of the main chain conformational angles 4 and rj for the 4 proteins in the chain conformational angles for non-glycyl residues superimposed on polypeptide containing L-alanine residues (Brant & Schimmel, 1967). angles for glycyl residues superimposed on contours of the calculated residues (Brant & Schimmel, 1967).

774

E. Arnold and M. G. Rossmann

Table 5 Main chin conformational angles for non-glycyl residues in HRVl4 capsid proteins that have positive 4 vaiues Residue

Q, (“)

((I (7

OJ0

His1059 Lys1097 Ser1139 His1249 Asn1274

57 55 25 71 43

30 37 102 52 78

-176 175 175 -172 174

Asn2030 Tyr2035 Ala2036 Asp2057 Asn2172 Met2173 Cys2248 Lys2257 Gln2262

58 58 58 58 53 64 53 104 65

-158 24 30 -127 21 19 43 101

-179 178 175 -169 - 179 177 -179 -174

Asn3027 Asn3074 Asn3077 Ser3194 Glu3200 Leu3221

73 48 39 173 68 62

10 23 -14 158 13 83

-178 170 -168 180 -177 -177

Leu4045 Met4047

47 60

36 95

176 177

Comments

NImIA NImIB

VP2 C-terminal

residue

NImIII NImIII 3-Fold axis

Figure 7. Model of the /?-cylinder

/?-bulge in VP2, at the edge of the BIDG sheet, occurs at the same location as the VP3 knob, suggesting the possibility that wide p-bulges can serve as convenient sites for insertions. There is a large loop present in all three major capsid proteins of HRV14, located in the middle of the /?G strand. An analogous feature was found in SBMV. This loop wraps the carboxyl end of the fiG sheet around the blunt end of the b-barrels, and is located close to the pseudo-3-fold axis that is at an interprotomer

interface

in picornaviruses.

Compari-

son of HRV14, Mengo virus, SBMV and TBSV coat protein amino acid sequence alignments (Luo et al., 1987) shows that seven of the eight proteins have a

Figure 6. Stereodiagram phases.

of cis-proline

of HRV14 that /3-strands contributed by 5 VP3 molecules in a pentamer. The hydrogen bonding distances and the main chain conform&,ional angles are designated. consists

of 5 parallel

P-X sequence at the beginning of the loop, with X being an aromatic residue (Y, W or I?). Indeed, every picornaviral VPl, VP2 and VP3 sequence that is available (see Notation) satisfies this requirement. The aromatic residue at position X fills in the gap in the barrel created by the divergence of flD and /?G at the site of the loop. Interestingly, in each of the three /3G loops in the HRV14 capsid proteins,

2083 in electron density computed from the 3.08 A molecular

replacement

Refined Structure

of HRV14

775

Table 6 P-Turns R2

VP2

VP3

VP4 13. “Nrar” VP1 VP3

L46 A87 v217 D285 c34 D57 T72 D84 TI 15 c: I ,50 Dl68 Y171 Nl90 P23 I LT G9 I,14 E63 F86 F91 P120 TlXO c214 D34 P49

VP2

capsid proteins

R4

42 (7

$2 (“1

43 (“)

*3 (“1

D49 190 H220 S288 E37 V60 575 K87 H118 G153 Y171 N174 T193 A234 SlO Ql2 D17 S66 D89 T95 S123 S183 F217 537 F52

-48 -52 -56 -66 58 -63 -60 -41 -76 -65 -60 53 -44 -58 -54 -54 -60 -60 -59 -56 -57 -54 -73 -74 -62

-39 124 -14 -33 24 -15 -20 -32 12 -24 -47 21 -37 128 139 147 -12 -4 -16 -29 -34 - 45 -6 -20 -28

-90 69 -109 -71 58 -84 -96 - 100 -109 -64 -57 64 -70 60 -105 68 -107 -98 -96 -102 -85 -81 -66 -74 -63

20 25 -18 -14 30 -22 4 6 -18 -34 -9 I9 - 14 26 -l4 -2 3 -7 12 -10 12 21 -36 -10 -28

Type

d LQt

Location

P47 T88 L21X 1286 Y35 T58 T73 A85 Kll6 El51 Vl69 Nl72 L19I T232 P8 SIO T15 V64 I87 K93 K121 YlXl P215 A35 SRO

548 G89 N219 K287 A36 559 G74 L86 F117 R152 1170 Ml73 R192 GP33 G9 Gil T16 N65 G88 T94 L122 T182 D216 A36 K51

I II I III III’ 111 I I I Ill I 111’ III II II II I I I I I I III 111 III

3.1 2.7 28 32 30 31 2.8 2.7 31 3.0 3.3 29 2.5 2.9 31 2.9 32 30 3.1 3.0 33 2.9 33 3.1 2.9

FMDV loop Carboxyl terminus

Puff Puff Puff Corner Corner End of j-cylinder Knob

P-turns with prolin~ at position i + 1 (angular deviations outside the accepted range or N-O distance > 3.3 A) v3 1 I’ I .54 L25 P198

(‘. d,,-Helicrs VI’1

in the HRV14

in the HRV14 caps&i

A. B-Twns VP1

R3

and 3,0-helices

P32 P155 P26 PI99

133 G156 N27 E200

L34 Al57 Y28 T201

-69 -53 -64 -52

- 13 119 133 157

T39 G40 E95 A96 D46 A47 T147 H148 N180 L181

-64 -68 -58 -61 -64 -63 -46 -63 -47 -46

-9 -16 -40 -43 -36 -26 -44 -26 -49 -31

-94 118 73 68

-3 -14 10 13

I II II II

3.5 32 3.5 3.4

-16 -24 -43 -23 -26 -9 -26 I6 -31 12

111 III III Ill III I 111 I III I

32 32 2.8 30 31 31 2.8 31 2.9 3.1

Bottom of canyon Corner

in thr HR VI4 capsid

A36 N3i x92 H93 P43 I)44 1’144 T145 I JI77 I.178

N37 E38 H93 R94 D44 v45 T145 F146 L178 G179

E38 T39 R94 E95 V45 D46 F146 T147 G179 Nl80

-68 -76 -61 -69 -63 -96 -63 -104 -46 -98

NlmIA NlmlA

Residues R, through R, are the residues rounding the bends, and the 4 and IJ angles are given for the central residues in the turns. Classification of the turn ‘type” is according to Crawford et al. (1973). The differences between observed and accepted torsional values were tolerated if less than 200”, and a single deviation of up to 300” was allowed. Distances (t) from the carbonyl oxygen of residue i to the amide nitrogen of residue i+3 were accepted at 3.3 A or less. The canonical definition for each type is: Type I II II’ Ill III’

42 (“) -60 -60 60 -60 60

** (7 - 30 I20 -120 -30 30

43 (7 -90 80 -80 -60 60

*a (7 0 0 0 -30 30

either the proline (in VP2) or the next residue (119OY in VP1 and 3167W in VP3) is among 78 residues with B < 9 A2, implying a considerable rigidity of this portion of the structure. The main chain hydrogen-bonding diagram for VP4 (Fig. 4(d)) shows a distinct lack of secondary structural features, save a few turns, a short u-helix and some intraprotomer hydrogen bonds with VPl, VP2 and VP3. Although the HRV14 electron

density map is not clear in the N-terminal region of VP4, by analogy with poliovirus and the presence of uninterpreted density on the 5-fold axis (see Primary Structure), it is likely that HRV14 VP4 is myristylated and that the five strands in a pentamer form a bundle that surrounds the VP3 p-cylinder. Thus, the carboxyl terminus of VP4 is suspended from this canopy as spokes radiating from the center of the pentamer. A significant amount of

776

E. Arnold and M. G. Rossmann

Table 7 Wide b-bulges in th HRV14 capsid proteins (Richardson et al., 1978) A representation of a wide b-bulge Strand

I 0

N

0

N

0

N

i I Strand

2

‘&

Residues in strand 1

Residues in strand 2

At position

1

2

3

1

2

3

4

VP1 VP2 VP2 VP3 VP2 VP3 VP1 VP2

D125 G105 Cl21 K126 L123 1128 H245 V243

F124 s104 G120 Al25 L122 L127 Y244 T242

R123 R103 5119 5124 Cl21 K126 V243 1241

K248 E247 A226 T193 1223 w196 A75 L66

H249 C248 P227 5194 P224 Y191 C76 D67

v250 T249 L228 L195 1225 Q192 v77 S68

E251 E250 T229 1196 A226 T193 H78 K69

surface of contact is observed for VP4 with the rest of the pentamer, and thus it may be attached relatively non-specifically. Its precise position may, therefore, vary from one picornavirus to another and is, indeed, rather different in Mengo virus (Luo et al., 1987).

5. Thermal Parameter Variation A plot of the isotropic thermal parameters?, averaged over all atoms in each residue, is shown in Figure 8 for VPl, VP2, VP3 and VP4, and the positions of secondary structural elements are indicated. The mean value of the temperature factor for all protein atoms is 14 A’, with values ranging from a minimum around 5 A2 to a maximum of nearly 50 A’. Given the relatively low resolution of the data used for refinement (3 A limit), the absolute values of the thermal parameters may be distorted. However, the pattern of variation is sensible: (1) there is a similar pattern of temperature factor variation for each of the three p-barrels (see Fig. 8); (2) a large proportion of residues located on the inner and outer surface have large thermal parameters, particularly those involved in viral recognition by antibodies (see Correlation of Physical with Immunogenicity); conversely, Properties (3) atoms in residues situated in the cores of the t Crystallographic temperature (thermal) parameters represent the combination of spatial differences in different unit cells with thermal motion. The latter is usually negligible compared to the former in biological macromolecules (Frauenfelder et al., 1979; Artymiuk et al., 1979). Here, temperature factors are also averaged over the 20 non-crystallographically related units of a virus particle.

P-barrels have the smallest thermal parameters, where there might be a restricted degree of conformational mobility (Fig. 9). Some of these residues are involved in large conformational changes on binding certain antiviral compounds that inhibit viral uncoating (Smith et al., 1986; Badger et al., 1988, 1989), and even these have low thermal parameters (Table 8). The interfaces between b-barrels in the protomer are also relatively rigid. Regions of the intertwining network of the amino termini, such as the VP3 region between residues 3010 and 3040, are also seen to have relatively low conformational mobility, even though the secondary structures present in this stretch are induced by intersubunit interactions. An error in the definition of non-crystallographic symmetry could be compensated in part by increasing thermal parameters as a fun&ion of radius. The presence of regions of atoms with high thermal parameters on the internal virion surface thus provides a good control for the significance of the high thermal parameters of atoms on the external virion surface. “Hot” regions on the interior surface include a region of VP4, between residues 4041 and 4047, which makes limited contact with other proteins, the carboxy terminus of VP4 and the nearby ordered portion of the VP2 amino terminus. This latter region may have a functional requirement for conformational mobility, as the apparently autoproteolytic cleavage of VP0 takes place in a maturing virion (Rossmann et al., 1985; Arnold et al., 1987b). Alternatively, there may be some spatial disorder in this region, due to the few remaining uncleaved VP0 chains in mature virions (Rueckert , 1986). Thermal parameters and surface accessibility esti-

Rejhxl

Structure

of HRV14

777

Table 8 Thermal paramterg and solvent-amem&% aurface area eetirnatea for reskdues involved in the bdn4h.g of the WIN antiviral agenab to HRV14

&Id.. acccxkbility (A2)

(B) a22) Main Residue chain A. Rekdwa

that

Side chain wwkgo

All atoms the

Main chein

Side Chlh

Iarge& umformahd

All atoms

ckungeadw lo

binding of riceWIN ixmpmda

(i) FMDV loop in VP1 cubes a*bo HHHl--k+H

h

-Pk!aF H-It+ HH

PI H

Yl213 Gl214 11216 Tl216 V1217 L1218 N1219 Hl220 Ml221 G1222 51223 Ml224

IS 18 19 19 18 20 21 19 17 19 19 18

‘2 21 19 17 18 23 21 ‘7 20 21

0

5 35 5 27 4 50 25 6 21 4

5 5 55 19 83 4 58 36 8 1 21 4

0 2 1 0 9 5 5 3 11

3 9 3 3 20 0 1 65

3 11 4 3 30 5 5 4 76

0 9 6 8 5 9 9 0 2 7 0

18 27 14 80 0 11 9 27 8 14 2

18 36 20 88 5 20 18 27 10 21 2

13 18 20 19 18 19 22 20 17 19 19 19

5 20 14 6 0 8 11 2 1 0 0

12 11 10 10 10 11 14 17 23 20 19 14 20 17 19 15 I6 16 13 12

(ii) Carboql end of/M c0rn.w Ml151 Yl152 v1153 PI154 Pl155 Gl156 All57 Pll58 Nl159

12 11 10 10 11 11 14 8:

12 11 10 10 10 13 16 25

(iii) BC and folbwi~ re8idua NllOO DllOl w1102 Kl103 11104 Nl105 L1106 51107 Sl108 LllO9 VlllO

19 17 15 17 17 17 16 15 15 14 13

21 20 14 22 ii 14 17 17 13 12

B. Residueawithin 3.6 A

Figure 8. Plots of the isotropic thermal parameters averaged for all atoms in each residue for the HRV14 crtpsid proteins. (a) VPl; (b)VPB; (c) VP3; and (d) VP4. The locakions of the secondary structural units are designated at the top of each plot.

K1103 11104 L1106 51107 Llll6 Yl128 Y1152 P1174 Vl176 Fll86 Vll88 v1191 Y1197 Cl199 Tl216 NE19 H1220 Ml221 Ml224 A3024

17 17 16 15 12 12 11 13 12 12 10 11 11 11 19 21 19 17 18 11

22 17 14 17 11 12 11 13 11 12 11 10 10 12 19 23 21 17 21 11

of tL boundWIN 20 17 15 16 12 12 11 13 11 12 10 11 10 11 19 22 20 17 19 11

8 5 9 0 0 0 2 1 0 0 0 1 0 20 14 8 11 2 0 0

c43q~7~nd8

80 0 9 27 14 4 9 3 2 7 5. 8 7 8 5 50 25 6 3 2

88 5 18 27 14 4 11 4 2 7 5 9 7 27 19 58 36 8 3 2

778

E. Arnold and M. G. Rossmann

Table 9 Thermal parameters and solvent-accessible surface area estimates for residues in the canyon

GO (A*) Residue

Figure 9. Diagram of the protomeric unit of the HRV14 capsid proteins color-coded by the atomic thermal parameters. Atoms with B factors 40 are yellow (B values are in A*). The display was generated using the program HYDRA (R. Hubbard, unpublished results).

mates for residues in the canyon (Rossmann & Palmenberg, 1988) are given in Table 9. When compared with the average for all surface residues (Table 10 gives thermal parameter estimates as a function of amino acid type for all HRV14 residues, outer surface residues, inner surface residues and canyon residues; given also is the corresponding information for solvent-accessible surface area as a function of amino acid type), the thermal parameters for these residues are somewhat lower, indicating a possible role for conformational stability of the viral recognition of a cellular receptor. However, in Mengo virus there exists disorder in many of the residues lining the “pit”, the putative receptor attachment site (Luo et al., 1987; Krishnaswamy & Rossmann, 1990). Alternatively, since these residues reside in a depression as opposed to a protrusion, they are in an inherently more densely packed region of the surface and have somewhat less room to roam. This topology constrains the ability of the virus to accept mutations in a cleft and, hence, regulates the rate of genetic variability in the putative receptor-recognition site. Similar constraints should prevail for the variability of surface residues in other proteins. The overall sample size of HRV14 canyon

residues

is

limited,

and

further

viral

receptor attachment surfaces would need to be characterized to develop a stronger correlation. The thermal parameters for residues involved with the binding site of the WIN compounds (Smith et al., 1986) are given in Table 8. Residues in the

H1078 v1079 T1080 DllOl w1102 K1103 P1155 G1156 N1159 D1164 Y1166 V1217 H1220 51223 El231 V1278 F3086 K3093 D3177 P3178 D3179 T3180 53183 Q3226 T3227 s3229 A3233

Surface accessibility

(.!I’)

Main chain

Side chain

All atoms

Main chain

Side chain

All atoms

12 16 19 17 16 17 11 11 21 21 16 18 19 19 21 20 14 12 15 16 16 17 15 20 18 18 16

13 14 19 20 14 22 10

12 15 19 19 15 20 10 11 21 23 17 18 21 20 24 20 13 13 18 15 17 17 15 24 18 18 16

0 25 7 9 6 8 9 5 11 23 9 6 11 0 11 26 4 2 6 19 8 5 4 21 17 10 21

31 35 9 27 14 80 20 65 42 30 27 25 21 51 20 42 37 46 36 22 46 35 65 13 52 35

31 59 16 36 20 88 30 5 76 65 39 33 36 21 62 46 47 39 53 -5 i;O 51 39 86 30 62 56

21 25 17 17 21 21 27 19 12 14 20 14 19 17 15 27 18 19 15

hydrophobic pocket inside VPl, including 1152 and methionine 1221, which both significant displacement upon binding of viral agents, do not have unusually high Residues at the mouth of the pocket, asparagine 1219, show considerably higher

tyrosine undergo the antimobility. such as mobility.

6. Solvent Structure A total

of 272 water

molecules

were included

in

the final model for the HRV14 protomer (Arnold & Rossmann, 1988). The locations of these relative to the C” backbones of the viral proteins in the protomer are shown in Figure 10. The density corresponding to a typical region of the difference Fourier used for locating the water molecules is shown in Figure 11. As indicated by the appearance of this difference Fourier density, the bound water is, in general, well ordered in the HRV14 cubic crystals. Some of the features of the solvent structure were not fully resolved. For example, some pairs of water molecules within hydrogen-bonding distance showed connecting density. Since the water positions were not refined, emphasis in this discussion will be placed on water demography and the general nature of the contacts. The distances of water molecules to the nearest protein and solvent atoms were examined for the number of short water

ReJined Structure of HRV14

Table 10 Average B and surface accessibility for HRV14 capsid amino acid residues as a function subdivided by location Inner surface residues

All capsid residues Residue type Ala Ax ASn Asp cys Gln Glu Gly His He Leu LYS Met Phe Pro Ser Thr Trp Tyr Val

5% fractional accessibility rewu.3 a Gly-X-Gly reference (see the text). Inner surface is that protein surface facing the internal RNA. Outer surface is that protein surface facing the virus exterior, including the canyon residues. Canyon residues are those defined by Rossmann & Palmenberg (1988). (surfj, average surface accessibility.

contacts with polar (non-carbon) atoms in the virion coat, other water molecules, hydrophobic (carbon) atoms, charged atoms, main chain polar atoms, side-chain polar atoms and the number of distinct proteins that have a polar atom contact with each water (Table 11). Greater than 60% of the solvent-accessible surface area of the isolated proteins becomes buried in subunit interfaces in the final virion structure (Table 12). Thus, there is plenty of opportunity for

the formation of intersubunit cavities. A significant proportion of the water molecules in the final model make short (less than 35 A) polar contacts with more than one protein subunit and are, therefore, involved with the formation of subunit interfaces (Table 11). These water molecules can be thought of as being post-translational modifications of the protein structure. A quality factor was devised for ranking the approximate scattering power of water molecules

Figure 10. Location of water molecules relative to the capsid proteins of HRV14. The positions of the water oxygen atoms in the refined VP4 of the reference

HRV14 model (indicated by circles) are shown alongside the C’ backbones of VPl, VP2, VP3 and protomer.

780

E. Arnold and M. a. Rossmann

Figure 11. Solvent omit difference Fourier map computed using molecular replacement phases. Observe the wellresolved water molecules in chemically reasonable positions. The following plausible hydrogen bonds have been omitted for clarity:

Wat5020

0

Wat5028

0

WA5028

0

Wat5028

0

0 Oy 0 N

Ala3145 Ser3188 Ala3130 Ser3188

for which both the occupancy, Q, and the individual isotropic temperature factors, B, were varied. The criterion used by James k Sielecki (1983), Qua1 = Q/B’, was found to be overly sensitive to errors in the value of B. In this work, the quality factor for water molecules, Quulwat, is defined as: Qwhat

308 w

2.86 A 315 A 3.15 A

where Res is the maximum resolution of the data used for refinement. This quality factor for water molecules represents the relative scattering factors of solvent atoms at the limit of resolution of the data used. The actual values of Qualwat indicate the percentage of actual scattering verBu8 a hypothetical “strongest” water molecule with Q = 1-O and B = 0 AZ. The water molecules included in the final

= 100 Q exp (- B/4Res2),

Table 11 Summary of water contacts in the re&ed HRV14 wwdel

%.x(4 A. To To To To To To To To To To

31

Number of CrmtaGtebetween water mokcu~ee and other atom8 separated by leaa than R,,t polar (non-carbon) atoms in protein shell 430 hydrophobic (carbon) atoms in protein shell 47 charged atoms in protein ehell$ 78 polar main chain atoms 245 polar side-chain atoms 185 other water molecules 68 polar atoms in VP1 155 polar atoms in VP2 134 polar atoms in VP3 128 polar atoms in VP4 15

Number of contacts of each type B. Bvyddown of typea of conhct8 between w&f Polar contacts Water contacts Different proteins contacted Hydrophobic contacts Charged contacts Main chain polar contacts Side-chain polar contacts

0 &ma

1

2

and other atom8 when R, 16 212 16 183 IQ4 68 102

66 46 169 66 61 117 108

101 14 82 18 16 62 51

3

33

35

570 119 96 326 244 74 210 179 158 23

722 278 121 415 307 76 257 230 204 31

4

5

6

Total contacts

26 0 0 2 0 6 1

3 0 0 0 0 2 0

1 0 0 0 0 0 0

570 74 119 96 326 244

= 33 A# 69 0 5 3 1 17 10

t Since a water molecule can make multiple contacts, a given water molecule may contribute more than once to a given item as well &8 to multiple items. 3 Charged atoms include ionizable atoms of aepartic acid, glutimic acid, lysine and arginine side-chains. § The entries give the number of water molecules that fall into & given category. For example, the number of water molecules meking contacti less then 33 A with 2 different protein molecules is 82.

781

Rejined Structure of HRV14

Table 12 Solvent-accessible surface area that becomes concealed during the folding and assembly process of HR V14 Surface area exposed by Isolated proteins

-

Protomer

--+

Pentamer

--+

Virion

-

Hydrated virion

VP1 Surface Change

18336 -t 6705

11631 3692

7939 -* 705

7234 + 1709

5525

Hydrophobic Charged Main

10480/57 2581114 394612 1

6278154 1700/15 2699123

4195153 1278/16 1851/23

3757152 1255117 1695123

2923153 1104/20 1164/21

VP2 Surface Change Hydrophobic Charged Main

12734 4 3932

8802 4 1122

7680 2515

5165 1410

3755

7135/56 1506/12 2903123

4762154 1116/13 2034123

4192155 941112 1776123

2843155 815/16 1297125

2158157 572115 916124

VP3 Surface Change

15844 6111

9733 4 3786

5947 1301

4646 --* 1316

3330

Hydrophobic Charged Main

9457160 1606/10 3581/23

5534157 1162112 2239123

3195/54 836/14 1250121

2447153 613/13 1087/23

1782153 499115 651/20

VP4 Surface Change

4638 2189

2449 242

2207 122

2085 + 206

1879

2829161 351/s 1123124

1443159 191/s 641126

1298/59 168/8 597127

1199/57 150/7 589/28

1101/59 131/7 523128

51552 18937

32615 8842

19130 4641

14489

29901/58 6044fl2 11553/22

18017/55 4169/13 7613/23

23773 4643 12880154 3223114 5474123

10246154 2833/15 4668124

7964155 2306/16 3254122

Hydrophobic Charged Main Total Surface Change Hydrophobic Charged Main

Each entry for surface and change is given in A’. Entries indicating the breakdown by chemical type (hydrophobic, charged or main chain) show exposed surface area in AZ/% of the protein in the given intermediate that has the characteristic (e.g. the percentage of the exposed surface of VP2 in a pentamer that is attributable to main chain atoms is 23%).

refined model for HRV14 at 3-O A had Qualwat values ranging from as high as 82.9 (Q = 1.0, B = 68 A’) to as low as 12.2 (Q = @20, B = 17.9 8’). Only seven of 272 water molecules have a Qualwat value below 20. A large number of weaker peaks were observed in a final difference Fourier at stereochemically reasonable positions for water molecules. A correlation between the number of protein atoms within 35 A of a water molecule and Qualwat values is shown in Figure 12. The solvent,-accessible surface area for the HRV14 virion that remains exposed after addition of the 272 water molecules is given in the final column of Table 12. The percentage of exposed surface that is hydrophobic or polar remains relatively constant in comparison to the dry virion, thus there is no preferential covering of the surface by the more ordered water molecules in HRV14. Surprisingly, the fraction of exposed surface that is charged does not decrease in the hydrated model. This latter trend may be due to an inability to

detect water molecules bound to the side-chains of charged residues such aa lysine and glutamic acid, which appear to be considerably more mobile than the surface residues taken as a whole (Table 10). A number of networks exist, in which three or more water molecules are linked by a chain of possible hydrogen bonds. Networks that include more than three water molecules connected by potential hydrogen bonds are: (5016, 5389 ( x 4/5), 5845 (x 3/5), 5511, 5076), (5062, 5310, 5688, 5651) and (5017, 5856, 5690, 5227) (water network listed in parentheses, adjacent molecules in the list are potentially hydrogen bonded to each other).

7. Solvent-accessible Surface Area of the Protein Shell of HRVl4 Measurements of the solvent-accessible surface area of the HRV14 protein shell have been made using the algorithm and program of Lee & Richards

782

E. Arnold

and M. G. Rossmunn

x

0

I

2 Number of protein

3 atoma

1 0/0 versus Gly-X-Gly reference) is the most selective (S = O-29) on the basis of the fraction of correct residues reiative to the total number of surface residues identified by the criterion. Scheme 5, mixing information from both surface accessibility to a medium-sized probe with thermal parameter information, is only marginally less selective (S = 623). Not all of the residues that were identified as sites of escape mutations are necessarily directly recognized by the neutralizing antibodies. The only requirement for these escape mutations to be identified is that the mutant virus survives in the presence of the neutralizing monoclonal antibody. Asn3072 in NImIII was not identified in any of the predictive schemes (Table 14) and may, therefore, be a residue not directly involved in antibody binding. This Asn might, for instance, be involved in a conformational change that occurs when the virus is neutralized, which may no longer be possible when this residue is altered. At least one site, located at the top of the VP3 knob, has all of the correct characteristics to be predicted as immunogenic in all of the methods used, but no escape mutants have been found at this site in HRVl4. The analogous portion of VP3 was found to be immunogenic in poliovirus, and in poliovirus 1 Mahoney is the dominant immunogen (Diamond et al., 1985; Page et al., 1988). Possibly, this region is not very immunogenic in mice, due to coincidental similarity with some mouse protein, and that antibodies would be generated in another

Figure 13. Plots of the fractional solvent accessibility relative to a Gly-X-Gly reference value (see the text) for each residue in the HRV14 capsid proteins calculated in the context of the entire capsid. (a) VPl; (b) VPZ; (c) VP3; and (d) VP4. The locations of the secondary structural units are designated at the top of each plot.

784

E. Arnold and M. G. Roeamann

Table 13 Thermal parameters, solvent accessibility and contact surface to medium and large-sized probes for residues in the vicinity of the immunogenic sites in HRV14 R= 1.4

Probe radius (A)

NIm site

Residue

NImIA

Ile1090 AsplOQl* Asn1092 His1093 ArglO94 Glu1095* Ala1096

36 42 34 28 32 35 27

Ile1082 GlnlO83* Asn1084 Lys1085* Asp1086 Pro1 137 Asp11385 Ser1139 Asp1 140

NImIB

NImII

NImIII

W

WL2)

R=5

Solvent-accessible surfaoe/frtlctional accessibility relative to Gly-X-Gly (A’/%)

R= 10

Contact surface/fractional contact surface relative to Gly-X-Gly (AZ/%)

(AZ/%)

59/33 148190 62/42 22111 101144 140175 15/14

4112 17/68 2110 o/o 4112 14154 010

6134 O/l o/o o/o 5132 o/o

20 23 22 26 30 28 34 30 26

o/o 2ljll 11/7 60/29 66140 32122 90164 82157

o/o o/o o/o

Ala1209 Glu1210* Thrl211 His2135 Glu2136* Gly2137 Ser2157 Ser2158* Ale2159* Asn2160 Glu2161* Va12162* Gly2163*

22 28 22 22 26 19 21 23 24 26 28 23 21

43136 120170 21115 35/17 75143

Ile1286 Lys1287* Ser1288 Leu307 1 Asn3072* Ala3073 Asn3074 Arg3075* Gln3076 Asn3077 Glu3078* Gln3079 Thr3202 Gly3203* Gln3204

21 23 23 13 16 17 21 27 23 25 23 15 22 23 24

18/Q 78/38 30123 o/o 36/23 32128 75146 102/42 56130 52129 60133 27115 99/66 44152 79141

w3

112

64150 46/38 79177 87152 78144 132179 lO/lO

217

3114 o/o a/44 8/33, o/o

o/o o/o o/o 010 o/o o/o l/8 2112 o/o

219 13/63 l/3 010 4115 010 5/18

l/5 7138 o/o o/o l/7 o/o l/4

21s

218 1l/66 3112 3/14 14149 010 o/o 319 l/4 o/o o/o o/o

l/6

Q/28 219

218

4118 010 10142 3122 2110

012

6149 o/o

l/6

Q/40 o/o o/o

o/2

o/o o/o o/o o/o o/o 4/16 l/3 l/4

l/6

o/o

3/19 o/o 010

An asterisk (*) indicates escape mutations. Residues are listed if they are within 1 residue of an escape mutant, or if they are located in a contiguous stretch of polypeptide containing 2 nearby mutations (as between D1091 and El095 in NImIA). The solvent-accessible surface was computed by the algorithm of Lee & Richards (1971), using parameters described in section 7. This is distinct from the contact surface (Lee & Richards, 1971), whioh corresponds to patches of the ven der Weals’ surface of the protein that come into contact with the probe. The solvent-accessible surface area becomes very large with increasing probe size.

animal. The presence of lysine 61 near the top of the VP3 knob suggests the additional possibility that this loop becomes cleaved by some trypsin-like protease before the virion comes in contact with the immune system. This situation is probably geared toward detecting antigenicity (not immunogenicity) and so sites such as the VP3 knob are probably antigenic, although they do not necessarily lead to a neutralizing immunogenic response.

9. Analysis of HRV14 Assembly: Burled Surfaces Due to the Ixmtcrsctions between Protomers and between Pentamers An approximate accounting of the energetics of the HRV14 capsid assembly (see Introduction and Background) is attempted here by assessing the magnitude of the solvent-accessible surface area of the viral protein subunits that becomes buried at

R&aed Structure of HRV14

Table 14 Prediction of immunogenic residues: comparison of selection schemes based on thermal parameters and residue accessibility to solvent and larger probes Prediction

scheme

ESlXp?

NIm site

mutant

1

2

3

4

5

6

NImIA

D1091 El095

+ +

+. +

+ +

NImIB

Q1083 K1085 D1138 81139 El210 E2136 S2158 A2159 E2161 V2162

+ + + + + + -

+ + + + + + + + + +

+ + + + + + + + +

K1287 N3072 R3075 E3078 G3203

+ -

+ + + +

+ + + +

+ + + + + + + + + + + + + + + +

+ + + + + + + + + + + + + + +

+ + + + + + + + + + + + + -

16

15

16

15

13

89

80

NImII

NImIII

Number correctly

predicted 9

Additional

residues incorrectly 37

Selectivity coefficient number identified)

93

(S = number 0.20

Criteria for immunogenic Scheme Criteria 1 2 3 4 5 6

identified)

0.15

prediction

correctly 0.14

0.17

50 32 predicted/total 023

0.29

schemes

Mean residue B factor >25 A2 Mean residue B factor >20 8’ Fractional accessibility > 25% for 1.4 A probe Fractional accessibility > 10% for 1.4 A probe; mean residue B factor >20 A2 Contact surface for 5 A radius probe is >5% relative to Gly-X-Gly reference; mean residue B factor >20 A2 Contact surface for 10 A radius probe is > 1 y0 relative to Gly-X-Gly reference

each successive assembly stage. This buried surface analysis depends on the assumption that the respective assembly intermediates have a similar conformation to that found in the crystallized mature virion. While this assumption seems reasonable for the protomeric units (Arnold et al., 19873), it seems less reasonable that the non-barrel components of the individual protein subunits remain unaltered. Many small and some large movements are expected to have taken place at the loci of subunit interaction during virion assembly, and it has been observed that some small perturbations of structure do occur when complementary protein surfaces associate (Colman et al., 1987). Thus, it is expected that the energetic analysis presented here is at least qualitatively correct. It has been suggested that the predominant driving force involved in the folding and association

785

of proteins is the hydrophobic free energy gained when amino acid residues are transferred from a bulk solvent phase to the substantially less polar protein interior (Kauzmann, 1959). In a series of papers, Chothia (Chothia, 1974, 1975; Chothia & Janin, 1975) showed that the hydrophobic free energy changes accompanying different stages of protein folding and association could be estimated by measuring the amount of solvent-accessible molecular surface area that becomes buried as a result of folding or subunit-subunit interactions. This analysis depends on the recognition that hydrophobic surface burial is favorable as a result of solvent exclusion and the consequent increased entropy of solvent water molecules, and that hydrophilic or polar surface burial is allowable as a result of the compensating polar interactions (hydrogen bonds, salt bridges) that are seen to exist at complementary protein interfaces (Chothia, 1975). Chothia found that burial of 40 A2 of surface area corresponds to a hydrophobic free energy decrease of roughly 1 kcal/mol (Chothia, 1974, 1975). The amount of solvent-accessible protein surface that becomes buried during the assembly process is given in Table 12. The results are given for the individual viral proteins as well as the presumed assembly intermediates and are broken down into the fraction due to main chain atoms, hydrophobic atoms and charged atoms. Thus, applying Chothia’s relationship, the decrease in hydrophobic free energy due to association of chains within a protomer is approximately 475 kcal/mol, of a protomer within a pentamer 220 kcal/mol, and of a protomer in one pentamer interacting with another pentamer approximately 115 kcal/mol. Thus, the biggest saving of energy occurs in the initial association of /?-barrels within the Pl polyprotein, while the least saving of energy occurs for the assembly of the 12 S pentameric units. This is qualitatively supported by the available data on picornaviral morphogenesis (Rueckert, 1986). The approximate hydrophobic energy decrease due to the internal folding interactions of the protein domains relative to an extended chain can also be estimated. From the formula (Chothia, 1975) A, = 1.44 M, where A, is the theoretical solventaccessible surface area for an all-trans extended polypeptide chain that has mass M, the values for the relevant portion of each chain before any folding occurs are: VPl, 43,500 A2 (273 residues); VP2, 40,500 A2 (255 residues); VP3, 37,500 A2 (236 residues); VP4, 6000 A2 (40 residues); and all VPs, 127,500 AZ. The reduction in solvent-accessible surface of 75,000 A2 from the fully extended chains to the folded proteins (before considering the association within protomers) leads to a hydrophobic energy decrease of roughly 1900 kcal/mol. The change in hydrophobic free energy upon encapsidation of the viral RNA cannot be computed in an analogous fashion, since the details of the RNA structure were not illuminated by the crystallographic work. However, at least a rough estimate can be made. If we assume that 50% of the inner

786

E. Arnold and M. G. Rosswmnn

surface is in contact with the RNA, and that there is roughly 5000 8’ (at a radius of 110 A, a sphere has surface area of 152,000 L%‘;divide by 60 and assume a roughness that doubles this ideal value) of exposed interior surface area per protomer, then perhaps 2500 A2 of protein solvent-accessible surface area becomes buried by the interaction with the RNA. The hydrophobic free energy decrease would then correspond to roughly 60 kcal/mol. This may be an overestimate, however, as it is likely that the heterogeneous RNA strand will not have interactions with the protein shell that are as neatly complementary as the protein-protein interactions within the capsid, leading to fewer compensating interactions for buried polar groups. Cations such as Mg2+ and polyamines are known to neutralize the negatively charged phosphates of the RNA in rhinoviruses and polioviruses, respectively (Rueckert,l986), and may play a role in liganding some of the buried polar groups of both the inner protein surface and the RNA.

10. Buried Surfaces for all Protein-Protein Subunit Interactions in the HRV14 Capsid The proteins for the context surface pairs of

surface accessibility of the viral capsid in the protomeric unit was computed both isolated proteins and the proteins in the of the neighboring proteins. The buried estimates thus obtained for all interacting proteins in the capsid are given in Table 15.

The results are subdivided into intraprotomer interactions, interprotomer interactions within a pentamer and interpentamer interactions. These results lend quantitative support to the previous assignment (Fig. 2) of the protomeric unit. The interaction between VP3 is more extensive with each of VPl, VP2 and VP4 in the proposed protomer than for any other grouping. Note, for example, that the sum of buried surface area between VP3 and VPl, VP2 and VP4 in the reference protomeric unit is roughly 6500 AZ, as opposed to only 2200 8’ for the interactions between VP3 ( x l/5) in the neighboring protomer and VPl, VP2 and VP4 of the reference protomer. This corresponds to a decrease of at least a factor of 3 in the hydrophobic free energy for the proposed protomer arrangement versus any alternative arrangement. The ordered portion of VP4, spanning residues 4029 to 4068, makes roughly equal contact with VP1 and VP3 within a protomer, and somewhat less with VP2. The contacts that this portion of VP4 makes with proteins in other protomers and pentamers are not as extensive. The missing residues at the amino terminus of VP4 may make extensive interactions with other VP4s within a pentamer, as intimated by the location of the lessordered extension corresponding to residues 4024 to 4028, which forms a flat annulus perpendicular to and encircling the 5-fold axis. As previously noted, the most extensive interactions within the protomeric unit occur between VP1 and VP3. These interactions include not only

Table 15 Solvent-accessible surface area (A’) Within

a protomer

(underlined

that is buried by interactions HR V14 capsid

of neighboring

protein

subunits in the

diagonal entries refer to the accessible surface area for the isolated protein subunit)

VP1 VP2 VP4 VP3 VP1 __ 18836 2144 897 4057 VP2 1987 __ 12734 556 1699 1621 VP3 3986 883 15844 VP4 994 601 p638 833 The paired results below and above the diagonal are not exactly the same due to the different surface being buried for each protein. For example, the amount of VP3 surface buried by interactions with VP1 (3986 A’) 1s not the same as the VP1 surface concealed by VP3 (4057 AZ) because VP3 has a different surface being concealed than does VP1 in the interacting region Within

a pentamer VP2 (x1/5)

VP1 (x1/5) VP1 VP2 VP3 VP4

1043 452 327 0

0 0 0 0

VP3 (x1/5)

VP4 (x1/5)

1385 765 788 73

0 0 9 29

VP1 (x4/5)

VP2 (x4/5)

VP3 (x4/5)

VP4 (x4/5)

VP3 (x2/5)

VP4 (x2/5)

VP3 (x3/5)

VP4 (x3/5)

450 0 740 0

269 0 820 9

4 0 75 26

0 0 153 43

0 0 115 0

0 0 131 120

0 0 41 0

1034 0 1387 0

Between pentamers

VP1 VP2 VP3 VP4

VP2 ( x 1/a

VP1 (x l/3)

VP2 (X l/3)

VP3 (x l/3)

VP4 (X l/3)

VP2 ( x 213)

VP3 ( x 213)

VP4 ( x 2/3)

0 691 0 0

0 703 0 115

0 29 3 0

0 1228 92 0

0 77 0 0

734 30 1290 74

0 4 86 0

100 0 0 0

Refined Structure of HRV14

Table 16 Reduction of accessible surface area due to interactions of pairs of structural elements in the HRVl4 A. Intraprotomer

capsid

interaction& 11111111112222222222333333333344

12345678901234567890123456789012345678901 ;;A2

;:‘L

11114

4

3 * 7

1

451257 *5 Nterm C B - 3 AB-422332 4* c 2 1 8 174 5 :: CB -3 3 7 ;t CCBCB-3 3*2* BCC c-794*2 1 2 2 VP2 BE c BBB C-416 264 3 11 puff CABC CBB-521 1 1 3 12 lo PF CBB CCBCBC-13 !Fi CB CBCCC-4 3 1 1 3 CBBCCBC cccPI - 4 B Nterm B cc B1 51 6 c 15 az C C -6211 14 knob ccc c BB-2332 1 3* PB CC32 261 PC B ccc -3 2 1722 CtA ABCCB-3 B B C 3*3*27 20 pD cc VP3 PE c-4*5*1 g A

PH

PI Cterm Nterm B B

BCC A

C B

C

PC g

2 1

3 2

1

B

c

C

C

B C

VP1 PE UB F'd B FMDV

B B

B

fi!

B

C

cterm

c

c

1 1

141

17

5 4

2

B AA

cc BB

VP3

VP3

($3)

BA,A, arm ps aA j?D puff /x PI

1

2 I

* 2

1 * 2 5 3

4

cc

C B CCBCBC

C

B

interactions 22 0 17 0 64 0 62 1

(~$3)

VP3 ($3)

VP3 (!;3)

3 1

1

-4

3

4* c-492

7 6 5 * 5 2 3 9 2*2 *3 3

B-2 CCA-5 22 B-5113 BBC B-2 3 CB CBBCBB-4 BBCCBC CA c - 2 BB CB B-

3-Fold related

B. Interpentamer

2*

1

5

CCCB-2 C A

C

VP2 VP2 VP2 VP2 VP2 VP2 VP2 VP2

3*

2

CBB-522 c-3 4 3 CCBC c-2219 CCB CBBCC-2 BBBCCBC ccc-2* BB A A-23 8 AB CB BA-4 CB BC C BA-1 51 B A-5212 B1 cc

35

40

6 I.

C

C 25

2

2

* 4

2-Fold related VP3 ($3)

VP1 (x&12;3)

VP2 (“x”;;‘;,

VP2

VP2

VP2

VP2

(~“$2)

(x8:12)

(:72)

($2)

0 8 0 0

8 265 0 41

0 0 0 0

0 45 0 235

37 82 7 30

36

95

7

36

0

(A’) 43 64 195 33 0 41 0 26

283 68 28 0 38 0 35 24

36 0 0 0 0 0 0 0

0 0 64 0 76 0 0 159

528 96 0 0 14 0 45 0

t Entries correspond to interactions between the secondary structural elements listed in the 1st column and another element that is referenced by a corresponding number. Values above the diagonal indicate the number of kcal/mol (derived from buried surface divided by 40) of hydrophobic free energy of association: blank for no interaction, 0 to 9 for 1 to 400 A’ and * for >400 A’. Below the diagonal is a single letter indicating the degree of hydrophobicity of the contact (A, 0 to 33%; B, 33 to 67%; C, 67 to 100%) measured by percentage of the buried surface belonging to carbon atoms. For instance, the association between VP4 (no. 1) and the VP2 BA,A, (no. 2) strands leads to a reduction of hydrophobic free energy of 3 kcal/mol with less than l/3 of the buried area being due to carbon atoms.

Structural

unit

VPI N-terminal arm

VP2 WI (x ‘lid

VP2

VP3 PF

BJQ

VP3 BE

VP3 ,ljH

VP3 BC

(X’l8)

(b) Figm 14. Representations of the ‘I-strand interpentamer sheet. (a) The VP2 /?A, and PA2 strands are from one pentamer, whereas the “CHEF” sheet of VP3 and the portion of the VP1 amino terminus are contributed by an adjacent 3-fold-related pentemer. The horizontal and slanted bars between residues in adjacent strands indicate hydrogen bonds. Of the 7 adjacent strand pairs, only VP2 PA, ( x l/3) and VP3 /SF have a parallel relationship. The overall appearance of the sheet is relatively flat, since the 2 antiparallel sheets are joined by the parallel pairing in the center. (b) Below is shown a stereo pair of the atomic model of this interpentamer sheet. Representative labels give reference positions in each strand. The sheet orientation is parallel to the roughly flat interpentamer interfaces, with the VP1 amino terminus residing on the internal surface of the virion and the /lC strand of VP3 on the outer capsid surface.

789

Rejined Structure of HRV14 those between barrels at the pseudo-2-fold axis, but also the intricate net of interactions between the amino termini and a smaller, but substantial, interaction between the carboxy termini of VP1 and VP3. The structure suggests that the unprocessed protomeric unit may have an appearance similar to that found in the mature virion (Arnold et al., 19876). The amino terminus of VPl, translated after the VP3 barrel may have folded, is nestled against the inner surface of the VP3 barrel. Its lack of regular secondary structure may be due to the availability of the VP3 barrel to act as a template for folding. Table 15 provides a convenient means by which to assess the overall degree of contact between proteins in separate assembly units. For example, it is clear that VP1 and VP3 make the most extensive contacts with other proteins within a pentamer. Interpentamer contacts are largely due to VP2 (in the reference protomer) interactions with VP1 ( x 2/3) and VP3 ( x 2/3) in an adjacent pentamer and with a symmetry-related VP2 ( x l/2). A complete tabulation of the surface areas of contact between all pairs of secondary structural units in the HRV14 capsid is presented in Table 16. The order of secondary structural units in the Table corresponds to the order of translation in the polyprotein. The pattern of contact surface area for VP1 , VP2 and VP3 secondary structure interactions is very similar and appears to represent a “fingerprint” for this type of domain. The b-sandwich description is appropriate for this fold, as consecutive interactions between adjacent strands in the CHEF and BIDG sheets are the most extensive, followed by less extensive interactions between the faces of the sheet involving /?BbC, flH/?I, BE/ID and flF/lG in pairwise strand interactions. The strands

/zlI and /ID have the most extensive contacts with helices aA in each barrel, whereas strands BH, PE and BF have the most extensive contacts with uB. Much of the contacts between viral proteins within a protomer involve the portions of the chains that are not part of the P-barrel, for example the VP2 puff interaction with the VP1 FMDV loop and with the VP1 carboxyl terminus, and interactions between the amino termini on the viral interior surface. Interactions between the paired helices crA and ctZ of VP1 and VP3 contribute to interactions across pseudo-2-fold axes. In an analogous interpentamer interaction, 2-fold-related aA helices of adjacent VP2 molecules are in contact as well as a region corresponding to the aZ position in VP1 and VP3. By far the most extensive surface of contact between elements in adjacent pentamers involves residues in the seven-stranded intersubunit /?-sheet and nearby structural elements (Fig. 14(a) and (b)).

11. Non-covalent Interactions: Amino Acid Residue Contacts within the HRV14 Capsid After the preceding examination of the overall surface of contact between assembly intermediates, this section follows with a breakdown by the chemical nature of the contacts. In poliovirus, pentamers are stable building blocks and require 2 M-urea for disruption, whereas poliovirions can be dissociated into pentamers in small concentrations of EDTA (Koch & Koch, 1985). Overall, interactions between protomers include more hydrophobic contacts than between pentamers, as reflected both by the buried surface calculations and the relative proportion of short non-bonded contacts at each type of interface (Table 17). The huge number of short non-bonded

Table 17 Distribution

of types of short non-bonded contacts in HRV14 for interactions within protomers, between protomers in a pentamer and between pentamers Hydrophobic

Mixed

Polar

860 69 313 478

1448 475 524 449

43 22 10 11

2602 579 909 1121

85 23 28 34

4 3 0 1

205 34 73 99

5 3 2 0

139 35 35 69

Attractive

charged

Total

A. Within protom,er All contacts (‘ontarts < 3.0 a

30A

Analysis of the structure of a common cold virus, human rhinovirus 14, refined at a resolution of 3.0 A.

Human rhinovirus 14 has a pseudo T = 3 icosahedral structure in which 60 copies of the three larger capsid proteins VP1, VP2 and VP3 are arranged in a...
4MB Sizes 0 Downloads 0 Views