J. Mol. Biol. (1992) 228, 516-528

Structure ofDthe U2 Strain of Tobacco Mosaic Virus Refined at 3-5 A Resolution Using X-ray Fiber Diffraction Rekha Pattanayek and Gerald Stubbs Vanderbilt

Department University,

(Received 27 March

of Molecular Biology Nashville, TN 37235, U.S.A. 1992; accepted 10 July 1992)

The structure of the U2 strain of tobacco mosaic virus (TMV) has been determined by fiber diffraction methods at 3.5 A resolution, and refined by a combination of restrained leastsquares and molecular dynamics methods to an R-factor of O-096. The structure is extremel) similar to that of the common strain of TMV, with the largest differences being in the protein loop that makes up the inner surface of the virus. and in the C-terminal region on the outer surface. Differences in the inner loop can be correlated wit,h differences in t.hr properties of the two viruses. Keywords.,

virus; tobacco mosaic virus; U2 strain of TMV: fiber diffraction: carboxylate groups

1. Introduction

contained all of the non-hydrogen atoms of t,he coat protein and t,he RNA: and 71 water molecules. The structure of two protein subunits of TMV. together with three nucleot,ides, is shown in Figure 1. Of particular interest were three regions where negative charges were forced by the st,ructure into cxlosr proximit,v: t,wo carboxyl-ctarboxylate pairs across the pro&in subunit interfaces. and a carboxylatc~phosphat’e interaction. It had been known for many years that several groups in TM\’ titrat’ed with anomalous ph’ values near 7, and Caspar (I 963) had suggested that disassembly of TMV was driven by electrostatic inheractions between a~riomalounl~ titrating carboxylate pairs forced into juxtaposition The obsrrvrd c*arhox)-I-by the structure. carboxylaie pairs and the phosphate--c,;trt)c~xylatr pair were identified as these anomalously titrating groups. One of the carboxylate pairs ((:lu95/(:lulOfi) and the phosphate-carboxylate (Asp1 16) pair wercl also believed to be the calcium-binding s&s postI]lated by Gallagher &L Lauffer (1983). Jn t)he work described in this paper. LW hi~vf~ constructed a model for the IT% coat protein based on t.he TMV coat, protein structure. and refinrd this model against fiber diffraction dat,a to 3.5 .%rrsolution using least-squares m&hods restrained (Hendrickson. 1985: Stubbs rf ~tl.. 1986) and molecular dynamics refinement, methods (Briinger pt al., 1987, 1989: modifications to handle fihe~ diffraction data are des~ribetl hy Wang & Stubbs, 1992). Altjhough most, of’ the struc+urc is vrr!’ simi1a.r to t,hat of TM\‘. there are substantial dif’f’erences in the protein conformation at low radius. including changes in the putat,ivr &Gun-binding

The U2 strain of tobacco mosaic virus (TMV?) is closely related to the common strain, having an identical morphology, similar host range, and 72”/” sequence homology (Rentschler, 1967: Altschuh et al.. 1981, 1987) in the coat protein. It does, however, differ from common TLIIIV in its sensitivity to radiation (Siegel et al.. 1956). and in its titration behavior (Butler & Durham. 1972). rnlike other tobamoviruses, which are acetylated, the coat protein of U2 has a free X terminus. A comparison of the radial density distributions (Holmes & Franklin. 1958) showed many years ago that there are significant structural differences between the two viruses near the inner wall of the virion. The t’obamoviruses are rod-shaped viruses 3000 a (1 Lh = 0.1 nm) long and 180 A in diameter, with a central hole of diameter 40 A. Approximately 2130 identical protein subunits of Jfr 17,500 form a righthanded helix of pitch about 23 8, with close to 49 subunits in three turns. In U2: t,here are 49.05 subunits in 3 turns (Franklin & KIug, 1955); in TMV, there are 4902 (Stubbs & Makowski, 1982). A single strand of RNA follows the basic helix between the protein subunits at a radius of 40 8. There are three nucleotides bound to each protein subunit. The structure of the common strain of TMV was determined by fiber diffraction methods at 2.9 x resolution by Namba et a.1. (1989). The final model t Abbreviations used: r.m.s., root-mean-square. 0022%2836/92/220516-13

TMV. tobacco

$08.00/O

mosaic

virus;

516 I(

I992

Acatkwu(~

k’rcsa

I,irrlited

517

Structure of the U2 strain of TM V site, and we have considered the effect that these differences might have on the titration behavior and radiation sensitivity of the virus.

2. Materials and Methods (a) Data collection

and processing

U2 was grown in Nicotiana tabacum var. Samsun and purified by differential centrifugation, following standard procedures designed for TMV. Fiber diffraction specimens were prepared by drawing pellets of the centrifuged virus into quartz X-ray diffraction capillary tubes of nominal diameter 0.7 mm, mixing with small quantities of buffer solution, and moving the column of virus to orient the long particles by shearing forces (Gregory & Holmes, 1965). These specimens were stable for several years. X-ray fiber diffraction data were collected photographically, as described by Namba et al. (1989) and by Namba & Stubbs (1985). The data used in the final refinement of the structure were obtained from a diffraction pattern collected by Dr K. Namba, from a specimen made by Dr Namba in the laboratory of Dr D. L. D. Caspar using virus from the laboratory of Dr K. C. Holmes. A diffraction pattern from this specimen is shown in Fig. 2. The diffracted intensities were determined from the diffraction pattern using a modified version of the angular declonvolution met’hod of Makowski (1978). Intensities were corrected for geometric and other experimental factors as described by Namba & Stubbs (1985). (b) Model

building

(a)

and refinement

In our initial attempts to solve the U2 structure, we collected data from several heavy-atom derivatives. and used multi-dimensional isomorphous replacement (m.d.i.r.) in an approach similar to that used to solve the structure of TMV (Namba et al.. 1989). The early m.d.i.r. maps were not readily interpretable, however. With data from better-oriented specimens, it became clear that the structures of TMV and C2 were sufficiently similar for us to use molecular replacement. The TMV structure was therefore taken as a starting point for molecular replacement and refinement. A model ronsisting of the atoms common to the 2 structures (about 95”/; of the total number of atoms in the helical repeating unit) was constructed. For most residues. the side-chain was either unchanged or required the omission of only 1 or 2 atoms. This model was used to ralrulate phases, and a map was calculated with coefficients 6!‘,,,, -. 51*:,,,. For a structure of this size and symmetry. 6pobs - 5Fc,,,, maps are the fiber diffraction equivalent of the 2F,,,Fcalc maps used in protein crystallography (Namba & Stubbs. 1987). This map was used t,o make initial estimates of the conformations of side-chains not common to the 2 viruses. Model-building used the program FRODO (Jones. 1982) on an Evans and Sutherland PS340 computer graphics system, and the program CHAIN (distributed by Dr F. A. Quiocho, Baylor College of Medicine, Houston, TX) on a Silicon Graphics IRIS 4D/340 GTX system. The model was initially refined by 171 cycles of restrained least-squares (Hendrickson, 1985) adapted for use with fiber diffraction data (Stubbs et al.. 1986). Difference maps. synthesized using coefficients Fobs- Fcalc and SF,,,- 5FCalC. were periodically calculated, and used to rebuild the structure. “Omit” maps (in which part of the structure is omitted from the calculation of Fealc in order not to bias that region of the electron density map)

vJ

ib)

Figure 1. Tobacco mosaic virus structure. (a) Computer graphics representation of about 1/20th of the TMV particle. Protein subunits are light gray: RNA is dark gray. The RNA is shown extending beyond the end of the protein helix for clarity. Graphics from Namba et al. (1985). (b) A ribbon drawing of 2 subunits of the TMV coat protein. These 2 subunits would fit into the righthand part of (a), as indicated by arrows. The core of the monomer is a bundle of 4 cc-helices, designated (in the terminology of Champness et al.; 1976) left slewed (LS). right slewed (RS), left radial (LR) and right radial (RR). Two shorter helices are labeled N and C. Three nucleotides of RNA (1. 2. 3) represented as GAS, are shown binding between the subunits, and in the enlarged inset.

were also used. In some cases, the strucature without the omitted region was refined for a few cycles to remove any residual bias, but we did not find that this had any effect on t’he maps. In general, the refinement at this stage followed the procedure used for TMV (Samba et aZ., 1989). As the refinement proceeded, it became clear that some parts of the structure. particularly at the inner and outer

R. Pattanayek

518

Figure 2. .A diffraction pattern from an oriented gel of the U2 strain of TMV, taken using flat film and a doublemirror focusing system.

walls of t,he virus, were significantly different from TMV. and that it would not be possible to build these parts unambiguously from difference maps. At this point, the R-factor was @14. We had by then adapted the molecular dynamics refinement program XPLOR (Briinger et al.. 1987, 1989) for fiber diffraction (Wang & Stubbs. 1992). One cycle of simulated annealing refinement was used. followed by several cycles of conventional refinement using XPLOR. interspersed by rebuilding of the model. The simulated annealing cycle included @2 ps at 2000 K followed by @3 ps cooling at 300 K. These times and temperatures were considerably limited by the computing power that was then available. and are. therefore. rather less than ideal (see, for example. Briinger et al.. 1989). Nevertheless. molecular dynamics did ext.end the radius of convergence of the refinement as we had hoped. removing uninterpretable features frotn the difference maps and reducing the R-factor furt,her. Details of the will be published elsewhere (Wang. refinement Pattanayek & Stubbs. unpublished results). A small number of water molecules was added to the model during the late stages of refinement. In view of the relatively low resolution, only water molecules in clear peaks of more than 3 times the standard deviation of electron density were added, and they were only retained if good stereochemistry and high occupancy persisted during refinement.

3. Results (a) The re$ned model The final model contains 1307 atoms, including all of the non-hydrogen atoms in the RNA, the protein. and eight water molecules. The water molecules are all located in or close to protein subunit interfaces. The r.m.s. deviation of bond lengths from the target values is 0019 A; for bond angles the deviation is 4.1”. Other restrained parameters are close to the target values (Table 1). The shortest non-bonded

and G. Stubbs interatomic distance is 2.5 ,&; there are 12 noI]bonded interactions closer than 2.6 8. The mean temperature fact,or for protein atoms is 40, and for nucleic acid atoms 63. There is no significant difference between main-chain and side-chain temperature factors. The starting model of c’% had an R-fact,or of 026. The final R-factor was 0.096. Fiber diffraction R-factors are inherently lower than crystallographic, R-fact,ors, because of the cylindrical averaging of the data. For a structure having the symmetry and dimensions of U2 at 2.9 .A resolution, the R-factor to be expected from a set of atoms randomly distributed within t,he radial limits of the virus would be about 0.32 (Stubbs, 1989; Millane. 1989). The R-factor was close t,o 0.1 throughoutj bhe resolution range, except in the highest range. where it was 0.171 (Table 1). Considerable effort was expended in attempts to reduce the high-resolution range R-factor. The value of 0171 is well below the starting R-fact,or for this range; which was about @25, close to the value t’hat would be expected from a random structure (Stubbs, 1989). It was not. however. possible to reduce it ang further. The high value may reflect poorer quality data near the resolut’ion limit. All of the non-glycine main-chain prot)ein dihedral angles

fall

within

or

very

close

to

t>he allowed

regions of t,hc Ramachandran plot (Fig. 3). Arg92. Ala101 and Qlu106 are the only non-glycine residues in the left-handed r-helical region of the plot. Arg92 is stabilized in this conformation by the binding of t,he RNA. as it is in TM\‘. The conformations of 41alOl and Glu106 are very different in CT2 from their conformations in TMV (see comparison below). Two final difference maps were calculated, one using phases from the complete final model, and one in which the eight water molecules were omitt’ed from the calculation. The complete final difference map

contained

about

20 very

small

peaks

greater

than three times the standard

deviation of the elect,ron density, but none greater than four t,imes the standard deviation. The eight acacepted watcxr

Table 1 Rejinement r.m.s. deviations

statistics for the jinal

I’% model

from ideal stereochemistry

Bond lengt,hs Uond angles Dihedral itngles Improper rotations Resolution

range (4)

K

Structure

qf the

d

00

. :. 1”

’ 0

.: ~

Figure dihedral marked

3. Ramachandran plot of the main-chain angles in U2 coat protein. Glycine residues are 0.

molecules, which had all persisted during the refinement, reappeared in the omit map with good geometry and with electron densities greater than any of the other peaks. Figure 4 shows omit maps of two regions of the inner loop of the coat protein. The loops from adjacent subunits line the inner surface of the virus. The st,ructures of these residues deviate substantially from the corresponding structures in TMV, and were consequently among the most difficult to refine. Early maps of these residues were particularly difficult to interpret, but the final omit maps are quite clear. Also shown in Figure 4 is an omit map of part of the interface between protein subunits, where Asp50 of one subunit approaches Asp77 from the subunit above. (b) Molecular

structure:

comparison

with

TM F

As expected with such high sequence homology, most of the structure of U2 is very similar to that of TMV. The core of the TMV protein structure consists of a right-handed four-antiparallel-m-helix bundle. similar to the packed helices of hemerythrins and some cytochromes (Richardson, 1981). There are also two very short helices, near the N and C termini. Several of the helices begin or end with one turn of 3,, helix. All of the helices are conserved in U2, and include approximately the same residues. As noted by Namba et al. (1989), the ends of the helices are difficult to define because of irregularities, but in U2 the helices appear to include residues 9 to 14, 21 to 31, 37 to 52, 73 to 87, 111 to 132 and 140 to 148.

l-72 strain

gf TM

V

519

The degree of order of the protein and nucleic acid structures, as judged from distribution of temperature factors, is similar in TMV and U2, but there is somewhat more disorder in U2, particularly around the RNA binding site. In both cases, the temperature factors of the nucleic acid are somewhat higher than those of the protein; the difference is more pronounced in U2. The variation of temperature factor with amino acid sequence is very similar in both viruses (see Fig. 2 of Namba &, Stubbs, 1986). The most striking differences between the two virus structures are in the protein conformations at’ the inner and outer surfaces of the virus: at the inner surface between residues 95 and 106, and at the outer surface in the last four residues of the protein chain, residues 155 to 158 (Fig. 5). All of the differences in a-carbon positions great(er than 3 A are found in these two peptides. The overall r.m.s. difference between a-carbon positions in the two viruses is 1.6 A if the last four residues are not considered. The r.m.s. differences between a-carbon positions in the four core helices is only 0.9 A. The r.m.s. difference between the R1VA atomic positions is 1.7 A. In the inner loop, there are differences in a-carbon positions of more than 6 A. Residues 97 to 100 form the top surface of the loop (Fig. 6). In TMV, these residues are folded into a reverse turn, extended to make a second hydrogen bond between N97 and 0100. In U2, Pro100 disrupts this structure, and a much more open loop is formed, tethered by a possible hydrogen bond between NlOl and 095. In TMV, the 97 to 100 reverse turn is intimately associated with another turn formed by residues 103 to 106 from the subunit above it, with two hydrogen bonds and many van der Waals contacts between the subunits. The 103 to 106 turn packs tightly into the virus structure in TMV. but in U2 the direction of the turn is reversed, so that these residues protrude from the viral surface. Again, this leads to a much more open structure in U2 than the structure found in TMV. The overall result, of these rearrangements (Fig. 6) is to restore the close topto-bottom contacts between the subunits in U2, probably forming two different hydrogen bonds. and certainly packing the loops very closely together. The peptide chain of residues 155 to 158 in U2 wraps around the outer surface of t,he virus, as it does in TMV, but it runs in almost the opposite direction (Fig. 7). In both viruses, this is a relatively disordered part of the protein with very high temperature factors, but omit maps clearly confirm t’he direction of the chain tracing. Tn TMV, the N-terminal residue is serine and t’he terminus is acetylated. U2, unlike other tobamoriruses, has Pro

Figure 4. Stereo views of difference maps of the U2 structure. Light lines: electron density. Heavy lines: parts of the model. The model atoms shown were omitted from the calculations of the maps. (a) Residues 97 to 100. The large density in the top left of the Figure is from the adjacent, symmetrically equivalent subunit. (b) Residues 103 to 106. (c) Asp50 from 1 subunit. together wit’h Asp77 from the subunit above it. and a water molecule (0) between the 2 carboxylate groups.

520

R. Pattanayek

and G. BtubbR

b

b

fi4

ICI Fig. 4.

4

Structure of the U2 strain of TM V

521

158. \

(b)

Figure 5. Comparison

of the a-carbon positions in U2 and TMV, together with the RPU’A of U2. Heavy lines: U2. Light lines: TMV. Selected cc-carbons are marked. The TMV subunit has been rotated and translated slightly to minimize the differences between a-carbon co-ordinates. (a) Two subunits viewed approximately perpendicular to the viral axis. (b) Two subunits viewed approximately parallel to the viral axis.

522

R. Pattanayek and G. Stubbs

as residue 1: and is not’ acetylated (Wittmann, 1965). The charged N terminus of U2 does not make any ion-pair interactions with other residues; the nearest negative charge is the C t,erminus, 9 L%away. The rings of Pro1 and Pro156 in 112 form a santlwith-like structure, however, which probably stabilizes the direction of the 155 t,o 158 chain. Tn TMV, five carboxylate groups have been implicated in the mechanism of disassembly of the virus (Namba & Stubbs, 1986; Namba et al.. 1989). Glu50 and Asp77 from separate subunits are close together in the top-to-bottom subunit interfaces, while Glu95 and GlulO6 approach each other in the side-to-side interfaces. Asp116 is very close to one of the RNA phosphate groups. It has been suggested (Caspar, 1963: Namba, et al., 1989) that under cellular conditions the electrostatic repulsion between these charged groups contributes to the disassembly of the virus. Four of the five carboxylate groups are conserved in U2, and Glu50 in TMV is replaced by Asp in C2. Residues 50 and 77 are about 4 L%apart in TMV, with no density t.o indicate the presence of a water molecule or a calcium ion. Tn U2, the corresponding carboxylates are about 5.5 14 apart, and the electron densit.y indicates that they are bridged by a water molecule (Fig. 4(c)). Asp116 is about 3.5 Lh from one of the R,NA phosphate oxygens in TMV, with electron densit,? t,o indicat.c the presence of a calcium ion between t’he groups. 117 the U2 model, one of the side-chain oxygen atoms from Asp1 16 is a little over 3 ,% from the RSA phosphat,e oxygen. but there is no indication in the map of the presence of a calcium ion or bridging water molecule. Glu95 and QlulO6 are bot)h in t’hr inner loop of the protein chain, and appear to form a calcium-binding site in TMV (Pattanayek rl al.. 1992). but, the conformation of this loop differs so much between the two viruses (Fig. 6(c) and (d)) that in C2 the t,wo groups’are more than 8 A4apart. The protein structure is stabilized by a large hvdrophobic region near the outer surface of the virus. forming a continuous hydrophobic ribbon that follows the viral helix. Most of the aromatic residues of the protein are in t.his ribbon, interacting extensively within and between subunits, and making edge-to-center contacts as in many other proteins (Burley & Petsko, 198.5). There is only one change involving an aromatic residue between TMV and 112: Lys68 from TMV becomes Tyr in U2. The tyrosine ring is incorpora.ted as part; of the hydrophobic ribbon. interacting with Phe62 and Tyr17. The conformation of t,he RNA in c’2 is essentially the same as that, of TMV. The conformations of the three sugar rings are all approximately 3’.rndo. although ribose 2 appears to be in a somewhat closer t’o altered form of this conformation. (‘-1’.rndo (Saenger, 1984). Bases 1 and 2 are in the common nnti conformat,ion, with the base pointing away from the ribose; base 3 is in the unusual syn conformation, as it is in TMV (Stubbs & Stauffacher, 1981). Most of the other torsional angles are in the same energetic minima that the.

occupy in T,MV. A t&v excrpt,ions (*an all be at,tributed t’o the fact that base 2 in the model of vi” is moved about I x to higher radius in t)he virus. relative to its posit,ion in TMV. The R?iA structure in tobamoviruaes is rather crowded (Na,mba it (I/.. 1989), and this small shift’ has t)he effect of relieving several slightly unfavorable torsion angles. In par& cular. the very unusual value of 8 in TM1’ nucleotide 2, 73” (samba it al., 1989), is changed in IT2 to a much more usual 162”. We cannot exclude the possibility t,hat’ even these small differenws are dur t,o the enhanced abilit? of simulated annealing refinement to find prevmusly inaccessible energ!. minima: since TMV was not refined bv t.his method, the unusual value of E found for that virus should be treat,ed with caution until simulated annealing refinement of TMV. now in progress: is c*omplet~e. Tntera,ctions bet)ween the protein and t ht> nucleic acid are also extremely similar in TMV and (‘2. Four conserved arginine residues, 41. 90. 92 and 113. neutralize the charged phosphate groups. =Zs noted above: the unusually close approach in TMV between the highly conserved Asp1 16 and phosphate 2 is also seen in I 12. \Ve have suggested t,hat in TMV this int,eract,ion is probably a c*alciurnbinding site (Namba Pt nl.. 1989). In the I’2 map. there is no electron density t,hat would indicate thr presence of a calcium ion. hut one or two oxygen atoms each from both the phosphate and the aspartate; the ring oxygen from ribose 3. and the ribost hydroxyl from nucleot,idc I, are still calustrr+d in a way that, suggests a binding sit,r for a metal. par1 i cularly c*alcium (Einspahr & Rugg. 1984). A feature of the KNA-binding sitmein TM\’ is that one of t,he three base- hinding sites appears to have a specific affinity for guanine, although it (*an also bind other bases. This affinity has been implicated in assembly and disassembly of the virus (Namba rt al., 1989). Guanine is important in the recognition of TMV R,NA by the coat prot,ein dtlring viral assembly: assembly of TMV begins with the binding of a loop of RNA which includes the squen(ae (XXG), (Zimmern, 1977). rt may also br import,ant. in the disassembly of the virus. The first 69 nucltbotides of the T,MV genome include no guanint~ bases. and it, is known that the protein subunits binding t’o t.his part of the RNA must. be removed in 1he tharl>, stages of disassembly, so that ribosotncbs can bind to the first start codon. and displacae thcl rest of thcl coat protein during translation (Wilson. 1984: Sha\% it al., 1986). The specific guaninr-binding site is also seen iti CJ2(Fig. 8). The base lies between two intrr-subunit salt-bridges. Argl22-Asp88 and Aq~ll5--Argll3. Hydrogen bonds can be formed by the guaninr O-6 and N-7 with Arg122. and by N-2 with the carbong oxygen of Arg90.

4. Discussion (a) Propertirs

qf the cirus

The structural differences between U2 and TMV may explain some of the differences in their proper-

Structure of the UZ strain of TM V

523

(a 1 P102

TMV

-6 El06

Figure 6. Comparison of the inner loop structures of U2 and TMV. (a) U2: stereo views of residues 95 to 102, together with residues 101 to 106 of the subunit above (+ 16 subunits in the viral helix). (b) TMV: as for (a). (c) U2: residues 95 to 108, together with the same residues in the laterally adjacent subunit (+ 1 subunit in the viral helix). (d) TMV: as for (c). The reverse turn 103 to 106 in TMV has the opposite orientation in U2; the turn 97 to 100 is present in TMV, but not in U2. Heavy lines mark the protein backbones. Views are from inside the viruses, but rotated slightly for clarity. Views in (a) and (b) are identical, but different from the view used in (c) and (d); the dots at the bottom of each figure are arbitrary points, marked to allow the superimposition of (a) and (b), and of (c) and (d).

R. Pattanayek and G. fhbbs

524

0

0

TMV

0 l

(d) Fig. 6.

ties, particularly the radial density distributions (Holmes & Franklin, 1958) and the sensitivity of U2 to radiation (Siegel et aZ., 1956). Some insight is also available into the different titration behavior of the two viruses (Butler & Durham, 1972), but it is not) possible to explain completely a phenomenon as complex as titration behavior on the basis of a model derived at only 3.5 Lk resolution. Most of the structural differences between the two viruses are found in the inner protein loops, particularly residues 95 to 106, and on the outer surface of the virus, particularly in the C-terminal residues. A major cause of the large conformation differences in the inner loop is probably the fact that Ala100 in TMV is replaced by Pro in U2. The proline sidechain disrupts the reverse turn formed by residues 97 to 100: in TMV, the dihedral angle 4 of residue 100 is about - 170”, which is much too low to accommodate the ring structure of a proline residue. The rearrangement of the 95 to 101 loop requires a rearrangement of the 103 to 106 loop in the subunit,

thr close contacts above, in order to maintain between the subunits. Other changes may also he important, however; for example. the conformation may be influenced by the fact that the charge on Clu95 in TMV is stabilized by t’he proximity of Argl12. In U2, Argl12 becomes Gin. and Glu95 is not stabilized by the close positive charge. Because of its own conformational change and the generaIll. more open structure of U2, Glu95 has more access to the free solvent in ITS. The inner loop may be under less evolutionarv restraint than other part)s of the protein, since it ‘Is disordered in the free coat protein (Bloomer et rrl.. 1978; Jardetzky rt al.. 1978). The changes in the inner loop account for a number of the observed differences between the properties of TMV and U2. Calculations from the atomic co-ordinates show that the more open structure, extending to lower radii between residues 97 and 106 in L’Z, explains the altered radial density distribution observed by Holmes & Franklin (1958). Tt also appea,rs that the increased accessibility of the

Structure

qf

the U2 strain

of TM V

525

(a)

P156

yi” (b) Figure 7. Comparison of the with residues 1 and 2. (b) TMV: different path of the chain from the superimposition of (a) and

C-terminal regions of U2 and TMV. (a) U2: stereo views of residues 152 to 158, together as for (a). The interaction between Pro1 and Pro152 in U2 may account for the markedly 155 to 158 in the 2 viruses. The dots in each figure are arbitrary points, marked to allow (b).

RPU’A caused by this open structure could explain the great sensitivity of U2 to radiation damage (by diffusion of free radicals), compared to TMV. The sensitivity of U2 t’o ultraviolet radiation is in fact not significantly different from the sensitivity of free U2 RNA. In sharp contrast, intact TMV has less than one fifth of the sensitivity of free TMV RNA (Siegel et al.. 1956). The tobacco mosaic virus structure includes three sites of electrostatic repulsion, believed to be important in viral disassembly. While all of the carboxylate groups involved in these sites are conserved in U2, only two of the three sites are found in the U2 structure. The low-radius site in TMV has been

identified as a calcium-binding site (IVamba & Stubbs, 1986; Pattanayek et al., 1992). The calcium ligands include Glu95 and Glu106. These residues (and all of the other ligands) are part of the inner loop of the protein, and we do not see any likely metal-binding site in this region of U2. In support of the apparent loss of this site in U2, we note that in TMV the site is known to bind lead (Caspar, 1956: Stubbs et al., 1977; Pattanayek et al., 1992). Lead binds to a number of tobamoviruses near the inner viral surface; in addition to TMV, it binds to cucumber green mottle mosaic virus, watermelon strain (CGMMV-W) (Lobert et aZ., 1987), and to ribgrass mosaic virus (RMV) (D. Allen, R.

526

i

R. Pattanayek: and G’. S’tubbs

Asp

A88

Figure 8. Part of the binding site for base 1 in U2. The residues shown are from laterally adjacent subunits: residues 115 and 122 are from one subunit, while residues AM, A90 and All3 are from the other. The base lies in the intersubunit interface, interacting with base 3 (out of the plane of the Figure) and with 2 inter-subunit saltbridges. Potential hydrogen bonds between O-6 and Arg122 (dashed lines) favor the binding of guanine in this site, as does a potential hydrogen bond between K-2 and the carbonyl oxygen of residue 90. A hydrogen bond between Thr89 and N-l (not shown) could probably form

with K-3 in a pyrimidine, thus accommodating any base. This figure may be compared with Fig. 11 of Namba et a,l. (1989). but the directions of view are not identical.

Pattanayek & G. Stubbs, unpublished observations). It does not, however bind to U2 (our unpublished observations). The high-radius site of electrostatic repulsion, in the TMV subunit interface between Glu50 of one subunit and Asp77 from the subunit above it, is retained in U2, although residue 50 in U2 is Asp. Tn TMV, this site was not’ identified as a calciumbinding site. No electron density was seen that might correspond to a calcium ion, and more significantly, no other metal ligands were available. In U2, Asnl30 (Val in TMV) could be such a ligand; in that case, there would be potentially five oxygen atoms available to bind calcium, and a calcium ion could occupy the site in which we have located a water moleclule. We are not, however, aware of any titration or other data that might support’ this speculation. The site of the electrostatic repulsion between Asp1 16 and a phosphate appears to be intact in U2. but on the basis of the electron density in the map. there does not appear to be an ion bound at this site. Density in the TMV map was interpreted as a

bound calcium ion (Samba rt nl.. 1989). It seems likely, however. that this differencacarsists sitnpll because the concent,ration of c*alcium in the L-h specimen was lower. The TMi: spec4mrn used for data collection had originally been prepared in 1960 (Namba rf c&1., 1989): at, that, time laborat,or) distilled wat,er supplies generally contained higher levels of calcium ions than they do now. Butler & Durham (1972) titrated several strains of TMV including U2, a,nd found that the tit,ration curve for U2 was shifted to higher pH, suggesting that the pK of one of the abnormalI? titrating groups in I:2 might be higher than that of the corresponding group in TMV. The titration behavior of proteins is the sum of contribut~ions from many groups, and it is not, possiblr to explain completely the differences between TMV and 11%on the basis of t,hr struct,ura.l models. .It is calear that despite t,he sequence similarities between TMV a,nd 1:2, the structures in the vicinity of some of thr anomalously titmt.ing carboxylatr groups arc’ significantly different. lt. is not clear, however. that the most, obvious changes would lead to higher ph’ values. Raised carboxylate pK values are causetl 1)~ t,he pr0ximit.y of negat.ive charges, and four of the five carboxylate groups suggested as anomalously titrating groups in TMV (50, 77. 95 and 106) are actually furt.her away from negative charges in I-2. It may be that, the unusually high phi in 1’2 does not come from an inter-subunit carboxylate pair at all. The intra-subunit. pair AsplB/Glu22 is about 1 LA apart in TMV. and 3.6 ,& apart in the (‘2 model. In TMV, the pair is stabilized by lysines 53 and 6X. In C2, residue 68 is tyrosinr. and the losti of this stabilizing positive charge may be the most signiticant factor in raising one pK of I he clarboxylate pair. Tn addition, Lys53 in tr2 has movrd awn> from Glu22. t.oward Asp%). Finally. the proximit>y of Asp66 to the Asp19/Glu22 pair in both viruses may also affect the titration behavior: agait). in TMV the charge of Asp66 is partially nrut,ralizrd 1, the proximity of Lys68. The RSA and its immediately surrounding prot,ein are rat,her more disordered in 112 than they arcs in TMV, if we are t,o judge by temperature fac+ors. The mean t,emperature fact)or of the RNA atoms in T,MV is 45 (Samba pf al.. 1989): for I’2 it is 63. The\ overall mean t,etnperat)ure factors of t hfl prot’eill atoms are much c4oser together: 3.5 for TMY. and 10 for G’2. Although the limited resolution of t.he strut: ture determinations tneans that. t.hese numbers must be treated with caution, the differences are suf% ciently marked t.o suggest that disorder may eontribute t#o the greater susceptibilit,y of I:2 Tao degradation (Siegel et nl., 1956). The four C-terminal residues have high trtnperkLture factors in both viruses. Their very different. conformations

may be due bo the large dikerences

in

the nearby N termini of the two proteins: in TMV. residue 1 is acet,ylSer; in 1‘2 this residue is Pro. and is not acet,ylated. As discussed above, int)eractions between Pro1 and Pro156 may affect the dire&ion of the C-terminal peptide chain.

Structure

of the U2 strain

(h) Tobamovirus structure The determination of the U2 structure represents an important step toward the goal of establishing which structural features are specific to tobacco mosaic virus, and which are general, at least within the tobamovirus group. Sequence analysis is of limited value, since only 25 residues are invariant in seven members of the group (Altschuh et al., 1987). It is clear that the a-helical core of the coat protein structure is structurally conserved; at low resolution, this is also true of GGMMV-W (Lobert & Stubbs, 1990), and is likely to be true of the tobamovirus group as a whole. Peripheral parts of the structure are not conserved, however. The inner loop of the protein is functionally important: it undergoes a transition between disorder and order during viral assembly, and intersubunit interactions between carboxylate groups located in this loop are believed to help drive disassembly. Even so, it is the least structurally conserved part of the protein. The protein--protein interactions in other parts of the structure are better conserved, but even the carboxyl-carboxylate interaction between Asp50 and Asp77 in U2> deep inside the virus structure, is somewhat different from the corresponding Glu50Asp77 interaction in TMV. This interaction is missing altogether in CGMMV-W, and it appears that interactions of this type migrate freely within subunit interfaces in viruses (Lobert et aE., 1987; Namba et aZ., 1989). Protein-nucleic acid interactions, in contrast’ to protein-protein interactions, are well conserved. This is not’ unexpected: firstly, because among the 25 residues that are conserved in the tobamovirus sequences. 11 are directly involved in RNA binding et al., 1989); and (Altschuh et al., 1987; Ramba secondly, because evolutionary pressures would resist viable mutations in the nucleic acid binding site. whereas mutations in other parts of the structure can be compensated for by complement,ary mutations in spatially nearby residues. The conservation of the specific arginine-guanine interaction reflects the finding that closely related tobamoviruses generally encapsidate each other’s coat prot,eins efficiently (Atabekov et aZ., 1970; Okada et al., 1970). The interaction between arginine and guanine has been seen in other proteinnucleic acid specific interactions, for example in a zinc finger-DKA complex (Pavletich & Pabo, 1991), and in a complex between the catabolite gene activator protein (CAP) and DNA (Schultz et al.. 1991): such interactions were predicted by Weber & Steitz (1984) and by Seeman et al. (1976). Since the (XXG), sequence is found in the origin of assembly sequences of tobamoviruses as distantly related as TMV (Zimmern, 1977) and CGMMV-W (Meshi et al., 1983), it seems likely that the arginine-guanine interaction will be retained in all members of the group. Turner et al. (1988) found that the identity of t’he non-guaninr residues in the sequence was not import,ant,

Nevertheless.

provided

that

since distantly

X

was

related

not

cytosine.

members

of the

qf TM

V

527

group cross-assemble with only limited efficiency (Holoubeck, 1962), there may well be differences in the other base-binding sites of some tobamoviruses. The guanine-binding sites of TMV and U2, however, are very similar. In both viruses, there are hydrogen bonds between guanine and Arg122. Tn TMV, there appeared to be a hydrogen bond between guanine and Aspl15; in U2, this hydrogen bond is apparently replaced by one between guanine and the carbonyl oxygen of residue 90. it is not clear, however, whether this observation reflects a real difference. a consequence of the different methods of refinement, or a consequence of the different resolutions of the two analyses. Molecular dynamics refinement in TMV and studies of CGMMV-W and RMV now in progress are expected to provide more information on this subject. We thank Keiichi Namba and Hong Wang for many valuable discussions. This work was supported by NH grant GM33265 Computers were purchased with funds from NIH grant l-SlO-RR02506 and SSF grants BBS-8607624 and DIR-9011014. The refined atomic coordinates have been deposited with the Brookhaven Protein Data Bank. The identification code is IVTM.

References Altschuh. I>., Reinbolt. ,J. & Van Regenmortel, M. H. V. (1981). Sequence and antigenic activity of the region 93 to 113 of the coat protein of strain IT2 of tobacco

mosaic virus. J. Gen. viral. 52, 363-366. Altschuh. D., Lesk. A. Al., Bloomer. A. C. bz Klug, A. (1987). Correlations of co-ordinatrd amino acid substitutions with function in viruses related to tobacco mosaic virus. J. Mol. Biol. 193. 693-707. Atabekov, ?J. G.. Novikov, V. K., Vishnichenko, V. K. & Kaftanova. A. 8. (1970). Some properties of hybrid viruses reassembled in vitro. Virology, 41. 519-532. Bloomer, A. C., Champness, cJ. N.. Bricogne. G., Staden, R. & Klug, A. (1978). Protein disk of tobacco mosaic virus at 2% A resolution showing the interactions within and between subunits. Naturr (London), 276. 362-368. Burley, S. K. & Pet’sko, G. A. (1985). Aromatic-aromatic interaction: a mechanism of prot’ein structure stabilizat,ion. Science. 229, 23-28. Butler, P. ,J. G. & Durham, A. C. H. (1972). Structures

and roles of the polymorphic forms of tobacco mosaic virus. J. .+IoC.Biol. 72, 19-24. Briinger. A. T., Kuriyan. ,J. & Karplus. M. (1987). Crystallographic R-factor refinement by molecular dynamics. Science, 235, 45X-460. Briinger. A. T.. Karplus. M. & Petsko, G. A. (1989). Crystallographic refinement by simulated annealing: application to crambin. Acta Crystallogr. 45, 50-61. Caspar. I). L. D. (1956). Structure of tobacco mosaic virus. Xature (London), 177. 928. Caspar, D. L. D. (1963). Assembly and stability of the tobacco mosaic virus particle. Advnn. Protein Chem. 18, 37-121. Champness. ,I. ru’.. Bloomer, A. C.. Bricogne, G.: Butler, P. J. G. & Klug, A. (1976). The structure of the protein disk of tobacco mosaic virus to 5 A resolution. Nature (London), 259, 20-24. Einspahr. H. & Bugg, C. E. (1984). (Trystal structure studies of calcium complexes and implications for

528 biological

systems.

In

Metal

Ions

R.

Pattanayek

in

Biological

and its Role in Biology (&gel, H., ed.), pp. 51-97, Dekker, New York. Franklin. R. E. & Klug, A. (1955). The splitting of laye;el lines in X-ray fiber diffraction diagrams of helical structures: application to tobacco mosaic virus. Acta Systems.

Crystallogr.

vol.

II:

Calcium

8. 777-780.

Gallagher, W. H. & Lauffer, M. A. (1983). Calcium ion binding by tobacco mosaic virus. J. Mol. Biol. 170. 905-919. Gregory, J. & Holmes, K. C. (1965). Methods of preparing orientated tobacco mosaic virus sols for X-ray diffraction. J. Mol. Biol. 13, 796-801. Hendrickson, W. A. (1985). Stereochemically restrained refinement of macromolecular structures. Methods Enzymol.

115, 252-270.

Holmes, K. C. & Franklin, R. E. (1958). The radial density dist,ribut,ion in some strains of tobacco mosaic virus. Virology, 6. 328-336. Holoubeck. V. (1962). Mixed reconstitution between protein from common tobacco mosaic virus and ribo18, 401 nucleic> arid from other strains. Virology, 404. Jardet,zky, O., Akasaka, K., Vogel. D.. AMorris, S. & Holmes, K. C1. (1978). Unusual segmental flexibilit) in a region of tobacco mosaic virus coat protein. IVature (London), 273. 564-566. Jones. T. A. (1982). A graphics fitting program for macromolecules. In Computational Crystallography (Sayre. I).. rd.), pp. 303-317, Oxford University Press. Oxford. Lobert, S. & Stubbs, G. (1990). Fiber diffraction analysis of cucumber green mott,le mosaic virus using limited numbers of heavy-atom derivatives. Acta Crystallogr. sect. A, 46,

11, 273-283.

Meshi, T., Kiyama, It., Ohno, T. & Okada. Y. (1983). Nucleotide sequence of the coat protein cistron and 3’ noncoding region of cucumber green mottle mosaic virus (watermelon strain) RNA. ViroZogy, 127, 54-64. Millane, R. P. (1989). R factors in X-ray fiber diffraction. II. Largest likely R, factors for N overlapping terms. Acta Crystallogr. sect. A, 45, 573-576. Namba. K. & Stubbs, G. (1985). Solving the phase problem in fiber diffraction. Application to tobacco mosaic virus at 3.6 A resolution. Acta Crystallogr. Sect. A, 41. 252-262. Namba, K. & Stubbs, 0. (1986). Structure of tobacco mosaic virus at 36 a resolut,ion: implications for assembly. Science, 231, 140-1406. Namba. K. & Stubbs, G. (1987). Difference Fourier (%ystallogs. syntheses in fiber diffraction. Acta sect. A, 43,

533-539.

Namba. Ii.. Caspar, 1). L. D. 8r Stubbs, U. (1985). C.:omputer graphics representation of levels of organization in tobacco mosaic virus structure. Sciancr.

227. 773-776. Namba, K., Pattanayek, R. Visualization of protein-nucleic virus. Refined structure of virus at 2.9 A resolution by J. Mol. Biol. 208, 307-325. Okada. Y., Ohashi, Y., Ohno,

Sequential reconstitution of’ tobacco mosaic. virus. Virology. 42. 243-245. Pattanayek. R., Elrod, M. & Stubbs, (;. (1992). Characterization of a putative calcium-binding sit,e in tobacco mosaic virus. Proteins, 12, 128-132. Pavletich. N. P. & Pabo. (1. 0. (1991). Zinc finger-DNA recognition: crystal structure of a Zif268-DNA complex at 2.1 A. Scien,ce, 252, 8094317. Rentschler. L. (1967). Aminosiiuresequenzrn uncl pines physikorhemisches Verhalten dea Hiillproteins Wildstammes des Tabakmosaikvirus. ;Mol. CPV. Oerwt. 100. 84-95. Richardson. CJ. S. (1981). The anatomy and taxonomy of Chem. 34, 167 339. protein structure. Advan. Protein Saenger. W. (1984). Principlrs of L%Lcleic dcid r\trwturr. Springer-Verlag, New York. Schultz, S. C”.. Shields, G. (‘. & Steitz, T. A. (1991). (‘rystal structure of a CAP-D8A romplex: t,hr I),NA is bent by 90”. ScirrLce, 253. 1001- 1007. Seeman. M. (1.. Rosenberg, .J. M. & Rich. A. (1976). Sequence-specific recognition of double hrlicxal nucleic< acids by proteins. Proc. Nat. Acad. Nci.. I:.S.A 73. X04&808. Shaw, J. G., Plaskitt,. Ki. A. & Wilson, ‘I’. ICI. A. (1!#6). Evidence that t’obacco mosaic virus particles disassemble cLotranslationally in vivo. I’irology, 48. 326

336. Siegel. A.. Wildman. S. C. & (:inoza. \V. (19.56). Sensitivit? to ult,ra-violet light of infectious tobaccao mosaic virus nucleic acid. ,$‘aturv (London). 178. lll7~1118. Stubbs. (:. (1989). The probability distribution of X-ra) intensities in fiber diffrac%ion: largest likely values for fiber diffraction R facbtors. Actn Urystallmqr. ,swt. A,

45, 254-~258.

993-997.

Lobert. S.. Heil. I’., Namba. K. b Stubbs, C:. (1987). Preliminary X-ray fiber diffraction studies of cucumber green mottle mosaic virus. watermelon sbrain. J. Mol. Biol. 196. 935-938. Makowski. L. (1978). Processing of X-ray diffraction dat.a from partially oriented specimens. J. Appl. Crystallogr.

and G. Stubbs

&Z Stubbs. G. (1989). acid interactions in a intact tobacco mosaic X-ray fiber diffraction. T.

&

Ivozu.

L’.

St’ubbs. (:. & Makowxki. I,. (198%). Coordinated use ot isomorphous replacement and layer-line splitting in the phasing of fiber diffraction dat)a. Actor (‘r~@allo~qr. suet. A, 38. 417 -425. Stubbs. (:. & St,auffacher. (‘. (1981). Structure of thr RNA in t,obacco mosaic virus. .J. Mol. Biol. 152. 387- 396. Stubbs, G.. Warren, S. & Holmes, K. (1977). Structure of RNA and RNA binding site in tobacco mosaic virus from a 4 A map calculated from X-ray fibre diagrams. Nature (London), 267, 216--221. Stubbs. C.. Namba, K. &, Makowski. 1.. (1986). Application of restrained least-squares refinement t,o fiber diffraction from macromolecular assemblies. Biophys. J. 49. 58-60. Turner, I). R.. ?Joyce. I,. E. & Butler. I’. J. C. (1988). The tobacco mosaic virus origin RNA. assembly Functional czharacteristics defined by directed mut,agenesis. .J. Mol. Biol. 203. 531-547. dynamics in Wang, H. & Rtubbs. G. (1992). Molecular refinement against fiber diffraction data. ,4&u (‘rystallogr. sect. A. in the press. Weber. 1. 7’. $ Steitz. T. A. (1984). Modrl of spe(*ifica cbomples between catabolite gene activator prot,ein and R-DNA suggested by elrctrostatic~ cornplrmrntarity. t’soc. ,Vat. ilr&. Sci.. I’.S..-l. 81. 3973~-3977. Wilson. T. M. .\. (I 984). (‘otranslational disassembly of t’oba,cco mosaic virus. l’irology. 137. 255 -265. Wittmann. H. (G. (1965). Dir primlre I’roteinstrukt~~~t~ van StGmmrn des Tabakmosaikvirus. %. Snturforsch.

B20. 1213 1223. Zimmern.

I). (1977). The nucleotidr sequence at the origin on tobac.cao mosaic virus R&B. (‘VII, 11. 463.-482. for

(1970).

Edited

by I).

LIeRosier

;tssernbly

Structure of the U2 strain of tobacco mosaic virus refined at 3.5 A resolution using X-ray fiber diffraction.

The structure of the U2 strain of tobacco mosaic virus (TMV) has been determined by fiber diffraction methods at 3.5 A resolution, and refined by a co...
4MB Sizes 0 Downloads 0 Views