Host Range and Variability of Calcium Binding by Surface ... - CiteSeerX

doi:10.1006/jmbi.2000.3868 available online at http://www.idealibrary.com on

J. Mol. Biol. (2000) 300, 597±610

Host Range and Variability of Calcium Binding by Surface Loops in the Capsids of Canine and Feline Parvoviruses Alan A. Simpson1, Veda Chandrasekar1, BenoõÃt HeÂbert1, Gail M. Sullivan2, Michael G. Rossmann1* and Colin R. Parrish2 1

Department of Biological Sciences, Purdue University West Lafayette IN 47907-1392, USA 2

James A. Baker Institute of Animal Health, College of Veterinary Medicine, Cornell University, Ithaca NY 14853, USA

Canine parvovirus (CPV) emerged in 1978 as a host range variant of feline panleukopenia virus (FPV). This change of host was mediated by the mutation of ®ve residues on the surface of the capsid. CPV and FPV enter cells by endocytosis and can be taken up by many non-permissive cell lines, showing that their host range and tissue speci®city are largely determined by events occurring after cell entry. We have determined the structures of a variety of strains of CPV and FPV at various pH values and in the presence or absence of Ca2‡. The largest structural difference was found to occur in a ¯exible surface loop, consisting of residues 359 to 375 of the capsid protein. This loop binds a divalent calcium ion in FPV and is adjacent to a double Ca2‡-binding site, both in CPV and FPV. Residues within the loop and those associated with the double Ca2‡-binding site were found to be essential for virus infectivity. The residues involved in the double Ca2‡-binding site are conserved only in FPV and CPV. Our results show that the loop conformation and the associated Ca2‡binding are in¯uenced by the Ca2‡ concentration, as well as pH. These changes are correlated with the ability of the virus to hemagglutinate erythrocytes. The co-localization of hemagglutinating activity and host range determinants on the virus surface implies that these properties may be functionally linked. We speculate that the ¯exible loop and surrounding regions are involved in binding an as yet unidenti®ed host molecule and that this interaction in¯uences host range. # 2000 Academic Press

*Corresponding author

Keywords: parvovirus; canine and feline host range; calcium-ion binding; pH dependence; crystal structures

Introduction Viral host range and tissue tropism can be determined at many different stages of a viral life-cycle, including attachment, entry pathway, uncoating, transcription, translation, and assembly. Many viruses enter cells by endocytosis and subsequently undergo pH-dependent structural changes necessary for productive infection (Kim et al., 1990; Giranda et al., 1992). These structural changes are Abbreviations used: CPV, canine parvovirus; FPV, feline panleukopenia virus; ssDNA, single-stranded DNA; HA, hemagglutination; MVM, minute virus of mice; PPV, porcine parvovirus; PDB, Protein Data Bank; NCS, non-crystallographic symmetry. E-mail address of the corresponding author: [email protected] 0022-2836/00/030597±14 $35.00/0

often mediated by pH-dependent Ca2‡ binding (Durham et al., 1977; Robinson & Harrison, 1982; Haynes et al., 1993). Parvoviruses enter cells by endocytosis, after which the virus is transported through the cytoplasm to the nucleus (Parker & Parrish, 2000; Vihinen-Ranta et al., 2000). For some autonomous mammalian parvoviruses, such as canine parvovirus (CPV) and feline panleukopenia virus (FPV), there appears to be an early critical event, after cell entry and before genome replication in the nucleus, that controls tissue speci®city and host range (Parrish et al., 1991; Horiuchi et al., 1992). In the case of CPV, it has been shown that this event is mediated by the capsid protein (Chang et al., 1992; Spitzer et al., 1996), possibly through the binding of a host factor to the capsid. Although the identity of the putative host factor is # 2000 Academic Press

598

Calcium Binding in Canine and Feline Parvoviruses

or near the capsid surface. The major structural difference between CPV and FPV is in a surface loop between VP2 residues 359 and 375 (Figures 1 and 2). This loop is more ¯exible in CPV (see Table 2A, ID1) than in FPV (Table 2A, ID3), and its Ê in equivalent position differs by as much as 7.3 A a C positions (Agbandje et al., 1993). The change of conformation is probably determined by residue 375, which is Asn in CPV and Asp in FPV. It is shown here that this loop can adopt a variety of conformations or can be completely disordered. Therefore, this loop will be called the ``¯exible'' loop, although in previous publications it was referred to as the ``hemagglutinating'' loop, because residues involved in hemagglutination (HA) are adjacent to or within the loop and the conformation of the loop appeared to be correlated with HA activity (Barbis et al., 1992; Chang et al., 1992; Agbandje et al., 1993). Mutations at residues 93, 103, 299, 300, 301, and 323 have been shown to affect CPV host range (Parrish, 1991; Chang et al., 1992; Llamas-Saiz et al., 1996; Parker & Parrish, 1997), and all but residue 103 are on the viral surface (Figure 1). Residues 93, 103, and 323 differ consistently between CPV and FPV, whereas residue 300 varies in recent isolates (Parrish, 1991). Residues equivalent to 316 and 320 in CPV have been implicated in MVM tissue tropism (Ball-Goodrich & Tattersall, 1992), and residues equivalent to 386, 440, and 570 have been similarly implicated for porcine parvovirus (PPV) (Bergeron et al., 1996). These residues in MVM and PPV are close to both the ¯exible loop and residues that are involved in HA (Figure 1). During infection, the virus encounters a change of pH from about 7.5 outside the cell to about 5.5 inside the endosomes. Infection of CPV and FPV can be blocked by compounds that inhibit endosomal acidi®cation, indicating that low pH is essential to infection (Basak & Turner, 1992; Parker & Parrish, 2000). CPV is able to hemagglutinate Rhesus monkey erythrocytes at both pH 6.2 and 7.5, whereas FPV can do so only at pH 6.2 (Chang

not known, the differences in the surface structures of CPV and FPV suggest which residues are responsible for binding such a factor and, hence, allow us to propose a mechanism for tissue and host cell speci®city. Parvoviruses encapsidate a linear singlestranded DNA (ssDNA) genome in an icosahedral Ê in diameter protein shell approximately 260 A (Arella et al., 1990). The capsids contain 60 copies of a combination of the viral structural proteins VP1, VP2, and VP3, of which about 54 are VP2 or VP3. VP3 is formed by cleavage of 19 amino acid residues from the amino termini of some VP2 subunits after assembly of DNA-containing capsids (Weichert et al., 1998). VP1 and VP2 are formed by alternative splicing, such that the complete sequence of VP2 is present in VP1 (Rhode, 1985). In CPV and FPV, the additional amino-terminal domain in VP1 is 143 amino acid residues in length. The three-dimensional structures of CPV (Tsao et al., 1991; Wu & Rossmann, 1993), FPV (Agbandje et al., 1993), minute virus of mice (MVM, a parvovirus of mice) (Agbandje-McKenna et al., 1998), and Galleria mellonella densovirus (an insect parvovirus) (Simpson et al., 1998) have been determined to near-atomic resolution, and the structure of the Ê human parvovirus B19 has been determined to 8 A resolution (Agbandje et al., 1994). Each subunit has an eight-stranded, antiparallel b-barrel, similar to those found in many other viruses (Liljas, 1986; Rossmann & Johnson, 1989). The b-strands are connected by long loops that form most of the viral surface. The amino termini of some of the VP2 subunits, but none of the VP1s (Weichert et al., 1998), are externalized such that a conserved glycine-rich sequence is located in a channel that runs along each ®vefold vertex. There are only seven amino acid residues (80, 93, 103, 323, 375, 564, and 568) that differ consistently between CPV and FPV in the capsid protein, although other residues vary between different isolates (Table 1A). Five of the seven residues are on

Table 1. Differences between CPV and FPV A. Amino acid differencesa Residue (VP2) 80b CPV-Alabama R CPV-d CPV-d, A300 ! D FPV K Surface No

93b N

103b A

232c I

300b A

323b N

375 N

K Yes

V No

V Yes

D A Yes

D Yes

D Near

396 E E Yes 2

397 R R Yes 2

B. Amino acid residues implicated in hemagglutination Residue (VP2) 323 375 377 CPV N N R FPV D D R Surface Yes Near Yes Reference 1 1 2 a

Blanks indicate identity with CPV-Alabama. These residues contribute to the host range determination. c These residues are variable in different strains. d References: 1, Chang et al. (1992); 2, Tresnan et al. (1995). b

386c K Q Q Q Yes

484c V

564b S

568b G

I No

N Near

A Near

Calcium Binding in Canine and Feline Parvoviruses

599

Figure 1. Roadmap of CPV. Surface amino acid residues that differ Ê in position by more than 2.5 A between CPV and FPV are shown in magenta. Shown also are residues that have been identi®ed in determining host range (turquoise), HA properties (purple), and Ca2‡ binding (blue). Symmetry-related residues are not labeled. The arrow shows the approximate viewing direction of the ¯exible loop in Figures 2 and 5. The region containing a cluster of residues that might be implicated in binding a host factor is outlined. This Figure was prepared using the program ROADMAP (Chapman, 1993).

et al., 1992). Changes in the charge distribution and structure caused by pH are likely, therefore, to affect both cell infectivity and HA. Residues that are known to alter the HA properties of CPV and FPV (Table 1B) map close to the ¯exible loop, implying that structural changes in this region may affect both HA and host range. We have undertaken to study the changes in the structures of CPV and FPV between pH 5.5 and 7.5. Whereas the earlier published structures of parvoviruses were derived from X-ray diffraction data collected at 277 K and pH 7.5, we report here structures of CPV and FPV determined from frozen crystals maintained at 100 K over the pH range 5.5-7.5, in the presence and absence of Ca2‡. Cryo-crystallography allows data of adequate completeness to be collected from only one to three crystals per data set, thus improving the quality of the data. In order to obtain reliable comparisons, both the old and new structures have now been re®ned using standard crystallographic techniques. We ®nd that three Ca2‡-binding sites affect the conformation of the ¯exible loop in a pH-dependent manner and that changes in Ca2‡ binding are likely to be responsible for differentiating some of the properties of CPV and FPV.

Results and Discussion A considerable body of structural information is now available for mammalian parvoviruses, with near-atomic resolution structures available for canine, feline, and murine parvoviruses (Table 2). We present here the re®ned structures of CPV-d, A300 ! D mutant (Table 2A, ID4) and of FPV

(ID3) (both data sets were measured at 277 K and at pH 7.5). Also, we present the newly determined structures of CPV-d at pH 6.2 (ID8) and pH 5.5 (ID9), FPV at pH 7.5 (ID6) and pH 6.2 (ID7), FPV at pH 6.2 in the presence of 5 mM EDTA (ID10) and CPV-d at pH 6.2 in the presence of 22 mM GdCl3 (ID11), all determined at 100 K. By far the largest structural differences among the CPV and FPV structures occur in the ¯exible loop (Figure 2). The mobility of the loop varies between the different viral strains and different solvent conditions, as is apparent in the plot of temperature factors (Figure 3). The loop can adopt two distinctly different conformations, neither of which is as well ordered as most of the rest of the capsid protein; it can also be entirely disordered (Table 2A). The ability of the loop to alter its conformation is aided by the occurrence of glycine residues at positions 360, 362, 363, and 374 at the ends of the loop. Conformation A (Figure 2(b)) is de®ned as that in CPV at 277 K and pH 7.5 (ID1), and conformation B (Figure 2(c)) is de®ned as that in FPV at 277 K and pH 7.5 (ID3). The structure of FPV at 277 K and pH 7.5 showed three electron density features whose protein environment suggested that they corresponded to bound cations. Indeed, 8 mM CaCl2 had been used in the crystallizations, except when otherwise noted (Table 2A). All three sites are in close proximity to each other on the capsid surface, near the ¯exible loop. One of these sites (site 1) was previously recognized by Agbandje et al. (1993) as stabilizing the ends of the ¯exible loop. The two other sites (sites 2 and 3) are also observed in CPV (ID8 and 9) and in FPV (ID7) at

600

Calcium Binding in Canine and Feline Parvoviruses

Figure 2. Stereodiagrams showing the ¯exible loop in various parvovirus structures. (a) Ca trace of the ¯exible loops (blue, CPVAlabama, ID1; red, FPV at pH 7.5, ID3; yellow, CPV-d (A300D), ID4; green, FPV at pH 6.2, ID7). The Ca2‡ at site 1 in FPV (pH 7.5) is shown in black. The loop conformations can be roughly differentiated into two types, identi®ed as A and B. (b) Detailed structure of the ¯exible loop in the A conformation in CPV (ID1) at pH 7.5. (c) Detailed structure of the ¯exible loop in the B conformation in FPV (ID3) at pH 7.5. The position of the calcium ion at site 1 is indicated in the FPV pH 7.5 structure. These Figures were prepared using the programs XTALVIEW (McRee, 1993) and Raster3D (Merritt & Bacon, 1997).

low pH. Double Ca2‡-binding sites, such as sites 2 and 3, are common. There are at least 215 entries in the Protein Data Bank that have two Ca2‡-bindÊ apart. A histogram (Figure 4) ing sites less than 6 A Ê , which shows that the optimal separation is 4.1 A Ê distance between sites is similar to the 4.6(0.1) A 2 and 3 in CPV and FPV. The electron density of FPV (ID3) crystals, soaked in GdCl3 at pH 7.5, showed very large difference density peaks at sites 2 and 3, with site 3 having a higher occupancy (Table 3), consistent

with the replacement of bound Ca2‡ (M. Agbandje-McKenna, C. R. P. & M. G. R., unpublished results). On the other hand, when a CPV-d crystal was soaked with GdCl3 at pH 6.2, only site 3 was occupied (ID11). Furthermore, FPV crystals grown in the presence of 5 mM EDTA showed no electron density at these sites (ID10). The lower occupancy of Ca2‡ in the CPV A300 ! D (ID4) structure suggests that the mother liquor used during crystal mounting may have contained a lower concentration of CaCl2.

Table 2. Crystallographic data for CPV and FPV A. Atomic resolution structures of mammalian parvoviruses ID Virus Full/Empty HAa

pH

Cryst. temp. (K)

Bufferb

Resol. limit Ê) (A

R-factorc

Ref.d struc. det.

PDB accession Conf. of Ref.d refinement number flexible loope

1 2 3 4 5 6 7 8 9 10

Yes Yes No Yes Yes No Yes Yes Yes No data

7.5 7.5 7.5 7.5 7.5 7.5 6.2 6.2 5.5 6.2

277 277 277 277 277 100 100 100 100 100

2.8 3.0 3.3 3.25 3.5 3.5 3.0 3.5 3.5 3.0

28.3 21.1 28.5 (39.5) 21.4 (28.1) 24.5 (31.7) 27.6 (37.4) 28.8 (35.7)

1 1, 4 5 6 7 This work This work This work This work This work

2, 3 This work This work This work This work This work This work This work

4DPV 2CAS 1C8F 1C8D 1MVM 1C8G 1C8H 1C8E

A* A* B B C Disordered A Disordered Disordered Disordered

Yes

6.2

100

I I I I I I0 I0 (soaked low pH) I0 I0 I0 ‡ 5 mM EDTA without CaCl2 I0 ‡ 22 mM GdCl3

3.5

-

This work

This work

-

Disordered

CPV-Alab. Full and empty CPV-Alab. Empty FPV Empty CPV-d, A300 ! D Empty MVM Full FPV Empty FPV Empty CPV-d Empty CPV-d Empty FPV Empty

11

CPV-d ‡ Gd

B. Space groups ID 1 2 3 4 6, 7 8, 9, 11 10

Empty

Virus

Space group

Ê) a (A

Ê) b (A

Ê) c (A

a (deg.)

b (deg.)

g (deg.)

No. particles per asym. unit

CPV-Alab. CPV-Alab. FPV CPV-d, A300 ! D FPV CPV-d FPV

P21 P43212 P212121 P1 P21 P1 P21

263.1 245.5 380.1 267.6 266.2 254.9 254.2

348.9 245.5 379.3 268.4 250.1 254.6 254.0

267.2 795.0 350.8 274.3 379.1 453.8 380.3

90.0 90.0 90.0 61.9 90.0 77.6 90.0

90.8 90.0 90.0 62.6 94.1 74.9 92.8

90.0 90.0 90.0 60.2 90.0 69.3 90.0

1 1/2 1 1 1 2 1

Hemagglutination: hemagglutinates Rhesus macaque erythrocytes at 4  C. Buffer: I was Tris-HCl (pH 7.5), 0.75 % PEG 8000 and 8 mM CaCl2; I0 was I with 30 % ethyl glycol added as cryoprotectant. c R ˆ (jFoj) ÿ (jFcj)  100/jFoj, where Fc is computed based on the interpreted atomic structures. Values in parentheses are for the highest-resolution shell. In all cases, the deviation of bond Ê and 1.7  from idealized values. The proportion of non-glycine residues outside the most favored regions of the Ramachandran plot never lengths and bond angles were better than 0.01 A exceeded 75.1. d References: 1, Tsao et al. (1991); 2, Chapman & Rossmann (1996); 3, Xie & Chapman (1996); 4, Wu & Rossmann (1993); 5, Agbandje et al. (1993); 6, Llamas-Saiz et al. (1996); and 7, AgbandjeMcKenna et al. (1998). e A, B, and C are different conformations of the ¯exible loop; A* implies a partially disordered structure of conformation A. f Soaked low pH. a

b

602

Calcium Binding in Canine and Feline Parvoviruses

Figure 3. Plot of temperature factors. b Strands in the central barrel are indicated by A to I. Other secondary structural elements are also shown. Blue, CPV-Alabama at pH 7.5, ID1; red, FPV at pH 7.5, ID3; yellow, CPV-d (A300D) at pH 7.5, ID4; green, FPV at pH 6.2, ID7; grey, CPV at pH 5.5, ID9; black, FPV at pH 6.2 ‡ EDTA, ID10. Large temperature factors are primarily associated with the ¯exible loop, the b cylinder, and the b cylinder around the 5-fold axis. Notice the wide variation of thermal parameters for residues in the loops 1 and 2. These loops are adjacent to each other on the viral surface, and, in FPV, residue Lys93 makes hydrogen bonds with the backbone carbonyl groups of residues 225 and 226. Residue 93 is replaced with an Asn in CPV, a mutation that is one of the important determinants of host range. The pH-dependent changes in the ¯exibility of these two loops implies that their mobility may be important for interactions with a host residue.

Site 1, closely associated with the ¯exible loop, is present only in FPV at pH 7.5 (ID3; Figure 2(c)). In contrast, it is unoccupied in FPV at pH 6.2 (ID7), showing that this site is sensitive to pH (Figure 5(a)). The site is coordinated by Asp373, Asp375, and the carbonyl oxygen atoms of Arg361 and Gly362 (Figure 2(c)). However, Asp375 is replaced by Asn in CPV, causing site 1 to be unoccupied in all the CPV structures and the ¯exible loop to be either partially or completely disordered. Sites 2 and 3 are separated from each other by Ê and are about 10 A Ê from site 1. Site 2, only 4.6 A on the surface of the virus, is coordinated by Asp237 and Asp239, and site 3 is coordinated by Asp237, Asp240, and Asp405 (Figure 5(b)). Site 3, which has the largest number of aspartic acid ligands, has the highest occupancy in every structure (Table 3A). As calcium ions normally have

four to six ligands, it is possible that solvent molecules transiently coordinate the ions at the vacant ligand positions. The occupancy of the calcium ions affects the conformation of the ¯exible loop. In FPV (ID3), the carbonyl groups of Arg361 and Gly362 are ligands to site 1, and Arg361 also forms a hydrogen bonding network connecting Gln365 on the ¯exible loop to the carbonyl oxygen atom of Ser401 and Gly402 (Figure 5(b)). However, in CPV (ID1 and 2), site 1 is unoccupied and Arg361 points away from sites 2 and 3, making a salt-bridge with Glu411 (Figure 5(c)). In the CPV mutant A300 ! D (ID4), site 2 is unoccupied (Table 3A; Figure 5(d)), allowing Arg361 to make a salt-bridge with Asp405 and Asp237, hence stabilizing the ¯exible loop in conformation B, even in the absence of Ca2‡ at site 1. Histidine residues 277 and 403 form a second layer around site 3 (Figure 5(b)). The Ne atom of

603

Calcium Binding in Canine and Feline Parvoviruses

Figure 4. Histogram showing distribution of interCa2‡ distances in structures in the PDB. The broken line shows the distance between sites 2 and 3 in CPV and FPV.

The functional importance of the Ca2‡-binding sites was veri®ed by site-directed mutagenesis of some of the Ca2‡ ligands. Both the FPV singlemutant virus D237A and the FPV double-mutant virus D239A and D240A assembled, but were noninfectious (see Materials and Methods). Both CPV and FPV hemagglutinate erythrocytes by binding the sialic acid N-glycolylneuraminic acid (Neu5Gc) (Barbis et al., 1992; Tresnan et al., 1995). However, CPV hemagglutinates at low and high pH, whereas FPV is active only at high pH. Site-speci®c mutations of residues that ligate Ca2‡ site 1 in FPV, or that are located elsewhere in the ¯exible loop, alter the HA properties (Figure 7). The functional signi®cance of CPV and FPV binding to sialic acid is not known, but this interaction is host-speci®c. Cats express Neu5Gc on their cells, whereas most dog cells express only N-acetylneuraminic acid and, therefore, would not bind virus by interaction with their sialic acids (Yasue et al., 1978). Structural differences between CPV and FPV are con®ned to the ¯exible loop and its immediate environment on the capsid surface. That region has been shown to be important in HA and affects canine and feline host range (Chang et al., 1992; Horiuchi et al., 1992). Data presented here identify calcium ions that are essential for mediating the conformational changes. It is shown that the structure of the ¯exible loop is in¯uenced by the charge distribution created by the Ca2‡-binding sites, the occupancy of which is sensitive to pH and the concentration of free Ca2‡ in solution. This suggests a common mechanism for the observed HA behavior and the postulated host-speci®c step after endocytosis, both of which require pH-dependent structural changes.

Materials and Methods His277 forms a hydrogen bond to the carboxy groups of Asp240. Lowering of the pH, therefore, could alter the charge on the imidazole rings and indirectly affect the binding af®nity of Ca2‡ at sites 2 and 3. The FPV structure in the presence of EDTA (ID10) lacks bound ions at all three sites, resulting in a major conformational change of His403 (Figure 5(e)). Ca2‡ binding was assayed by incubation of capsids in 0.14 mM 45CaCl2. Both FPV and CPV bound 45Ca2‡ at pH 7.5, but at pH 6.0 the amount of bound 45Ca2‡ was reduced by 90 % (Figure 6). This is consistent with the decreased occupancy of Gd3‡ at site 2 at lower pH values (see above). In contrast, sites 2 and 3 were approximately fully occupied in the low-pH crystal structures of both FPV and CPV. The crystal structures were determined in the presence of 8 mM CaCl2. This concentration was 50 times higher than that used in the 45 Ca2‡-binding assays, resulting in a corresponding increase in occupancy.

Virus preparation, purification, and crystallization Prototype strains CPV-d and FPV-b were recovered from infectious plasmids by transfection of Norden laboratory feline kidney (NLFK) cells. Viruses were propagated in NLFK cells in roller bottles and puri®ed as described (Agbandje et al., 1993). Viruses were equilibrated in 20 mM Bis-Tris (either pH 5.5 or 6.2) or against 20 mM Tris-HCl (pH 7.5) at a concentration of 10 mg/ml. Sitting drop crystallizations were set up in microbridges (Hampton Research, Laguna Niguel, CA, USA) by adding 5 ml of virus suspension to 5 ml of mother liquor, which contained 0.75 % to 1.50 % (w/v) PEG 8000 and either 8 mM CaCl2 or 5 mM EDTA. The bridges were placed in the wells containing 0.5 ml of mother liquor at the same pH as the drop. The wells were sealed with clear tape and incubated at 19  C for 2-5 months. Hemagglutination assays Erythrocytes were obtained from Rhesus monkeys, cats, or horses in Alsever's citrated saline. HA assays were performed using 0.5 % (v/v) erythrocytes in either Bis-Tris-HCl buffer (pH 5.5 or 6.2) or Hepes-NaOH

604

Calcium Binding in Canine and Feline Parvoviruses

Table 3. Cation-binding sites A. Calcium-binding sites occupancy

Outer Inner

Site 1 Site 2 Site 3

Room temperature CPV (ID1)a None

B. Heights of peaks in Gd difference maps pH 7.5a Site 2 0.51 Site 3 1.00 Highest noise peak 0.19 a

FPV (ID3) 0.67 0.88 0.93

CPV (ID4) None None 0.50

FPV ‡ Gd pH 7.5a None 0.6 1.0

Frozen CPV-d ‡ Gd pH 6.2 None None 1.0

FPV (ID6)a None 1.0 1.0

FPV (ID7) None 0.97 1.01

CPV (ID8, 9)a None 1.0 1.0

FPV (ID10) None None None

pH 6.2a 1.00 0.33

Normalized to give an occupancy of 1 for site 3.

(pH 7.5), with 150 mM NaCl and 0.5 % (w/v) bovine serum albumin. All HA assays were set up in 25 ml volumes in microtiter plates with 50 ml of erythrocyte sample added to each well. Plates were incubated at 4  C. Mutagenesis The sequences of CPV-d and FPV were mutagenized using uracilated ssDNA in M13 vectors by the method of Kunkel (1987). Mutations were re-introduced into the complete plasmid clones, between the SpeI and EcoR V sites in the viral genomes (Parrish, 1991). NLFK cells were transfected with these constructs and screened for virus by immuno¯uorescence with antibodies against the viral capsid and against the NS1 protein. In cases where viruses were recovered, they were tested for pHdependence of HA and their ability to infect feline and canine cells using a TCID50 assay. Formation of capsids after transfection was assayed by immuno¯uorescence with a monoclonal antibody (Mab 8), that recognizes only assembled capsids. Assay for calcium binding by the capsids Aliquots (60 mg) of CPV or FPV capsid were incubated with 40 mCi of 45CaCl2 (0.14 mM) in either Bis-Tris-HCl (pH 6.0) or Tris-HCl (pH 7.5). After incubation for three hours at room temperature, the viruses were centrifuged into 10-40 % glycerol gradients at 100,000 g for seven hours in the same buffer with 150 mM NaCl. The gradients were collected into fractions; for each, the 45Ca radioactive emission was measured by scintillation counting, and the virus was examined by SDS-PAGE. Refinement of previously determined structures (see Table 2A) CPV-d (A300!D) at 277K and pH 7.5 (ID4 in Table 2A) The previously determined structure of CPV-d (A300 ! D) (Llamas-Saiz et al., 1996) was used as a starting model and was re®ned using the CNS program suite (BruÈnger et al., 1998). Some 150 cycles of restrained positional re®nement were carried out, using data from Ê resolution. The atomic temperature factors were 7-3 A ®xed, and an overall anisotropic temperature factor was ®t to the data. The icosahedral symmetry was strictly imposed at all stages. A calcium ion was built into the structure (site 3), coordinated by residues 237, 240, and 405. After minor modi®cations to the structure, the over-

all temperature factor was removed and 40 cycles of individual, restrained, isotropic temperature factor re®nement were carried out. The Ca2‡ at site 3 was assigned a temperature factor equal to the mean of the temperature factors of the coordinating atoms, and its occupancy was re®ned to a ®nal value of 0.5. FPV at 277K and pH 7.5 (ID3 in Table 2A) The previously determined structure of FPV (Tsao et al., 1991; Wu & Rossmann, 1993) was used as a starting model for re®nement. Some initial rebuilding of the ¯exible loop was necessary, and calcium ions were built into sites 1, 2 and 3. Restrained positional and individual temperature factor re®nement was performed in the same way as for the CPV-d (A300 ! D) structure. Structure determinations and their refinements FPV at 100K and pH 6.2 (ID7 in Table 2A) A crystal of FPV grown at pH 7.5 was soaked for ten minutes in 20 mM Bis-Tris buffer (pH 6.2) containing 5 % (w/v) PEG 8000 and 8 mM CaCl2, as well as 30 % (v/v) ethylene glycol as a cryo-protectant. Diffraction data were collected from a single frozen crystal at the F1 beamline of the Cornell High Energy Synchrotron Source (CHESS), using a Princeton 2 K CCD detector (Table 4). Oscillation photographs were recorded using a crysÊ wavetal-to-detector distance of 135 mm, with 0.928 A length radiation and an exposure time of 30 seconds. Ê resolThe images were indexed and processed to 3.0 A ution using DENZO (Otwinowski & Minor, 1997). The data from 266 frames were scaled and merged using partially recorded re¯ections corrected to their full intensity values with the SNP program (Bolotovsky et al., 1998). Re¯ections whose partiality was greater than 0.05 were used. The space group was P21 (Table 2B), and the MatÊ 3/Da, with thews coef®cient (Matthews, 1968) was 2.1 A one particle per crystallographic asymmetric unit. The orientation of the particle in the asymmetric unit was found by computing a self-rotation function using the program GLRF (Tong & Rossmann, 1990). A translation function was calculated using the program TRANSF (Argos & Rossmann, 1980) with the known structure of FPV (Agbandje et al., 1993) as a search model. The center of the particle was found to be at (0.2307, 0.0000, 0.2584). Non-crystallographic symmetry (NCS) electron density averaging was performed for 22 cycles using a spherical mask with inner and outer Ê , respectively. Calculated structure radii of 70 and 140 A

Table 4. Crystal data collection ID

Virus

pH

Resolution Ê) (A

No. of crystals

No. of frames

6 7 8 9 10 11

FPV FPV CPV-d CPV-d FPV CPV-d

7.5 6.2 6.2 5.5 6.2 ‡ EDTA 6.2 ‡ GdCl3

3.5 3.0 3.5 3.5 3.0 3.5

3 1 1 1 1 1

208 266 198 557 599 142

a

Rmerge

a

17.9 (35.5) 15.4 (20.4) 11.9 (12.1) 9.5 (12.1) 9.2 (22.1) 11.3 (none)

Rdiff

b

19.2 (22.9) 25.0 (31.4) 28.0 (37.7)

Completeness 31.3 38.3 27.7 63.7 85.5 8.0

(10.3) (36.8) (28.3) (40.3) (47.3) ( 4.9)

Osc. angle (deg.)

Effective mosacityc (deg.)

Source

Final corr. coeff.d

0.20 0.20 0.20 0.20 0.30 0.25

0.25 0.14 0.25 0.35 0.15 0.25

CHESS F1 CHESS F1 CHESS F1 CHESS F1 APS 14BM-c CHESS F1

0.55 0.90 0.57 0.58 0.86 -

Rmerge ˆ hijhIhi ÿ hIhii  100j/hihIhi. When there were insuf®cient data, the individual re¯ections were scaled against a reference data set and assessed by Rdiff as de®ned in the text. Mosaicity as de®ned by Rossmann (1979). d Correlation coef®cient ˆ h(hFci ÿ Fc)(hFoi ÿ Fo)/{[h(hFci ÿ Fc)2]1/2[h(hFoi ÿ Fo)2]1/2}, where Fc is the structure factor obtained by Fourier back-transformation of the averaged cell. b c

606

Calcium Binding in Canine and Feline Parvoviruses

Figure 5. Stereodiagrams showing the environment of the double Ca2‡-binding sites 1, 2, and 3 in relation to the ¯exible loop. (a) FPV, pH 6.2 (ID7). (b) FPV, pH 7.5 (ID3). (c) CPV, pH 7.5 (ID1). (d) CPV A300D, pH 7.5 (ID4). (e) FPV, pH 6.2 EDTA (ID10). These ®gures were prepared using the programs XTALVIEW (McRee, 1993) and Raster3D (Merritt & Bacon, 1997). factors were used for unobserved re¯ections to compute new electron density maps. The position and orientation of the particle were re®ned by minimizing the difference between electron densities at NCS-related positions

using the program ENVelope (Rossmann et al., 1992). The ®nal overall correlation coef®cient was 0.90 (Table 4). The rotation matrix [P], which maps the orientation in the unit cell to the standard orientation of an icosahe-

607

Calcium Binding in Canine and Feline Parvoviruses

pH 6.2 structure determination (ID7 in Table 2A). The already merged FPV pH 6.2 data were used as a reference for scaling the data because of the low multiplicity of the re¯ections. In the case of data sets with very low multiplicity, the re®nement of image scale factors is less stable in SCALEPACK (Otwinowski & Minor, 1997) than with the SNP or SCALA (Collaborative Computational Project Number 4, 1994) programs. SCALEPACK sums adjacent partial re¯ections to obtain the total re¯ection intensity, whereas the latter two programs use each partially recorded re¯ection separately and obtain the full re¯ection intensity by correcting for partiality. The latter technique effectively constrains the scales of adjacent images relative to each other by scaling different parts of the same re¯ection to each other, assuming a standard rocking curve for the crystal. The Rdiff factor between the reference (Fref) and observed (Fo) measurements was 19.2 % (Table 4), where: 

Rdiff

Figure 6. Binding of 45Ca2‡ to CPV and FPV at pH 6.0 or pH 7.5. Aliquots (60 mg) of puri®ed CPV and FPV capsids were incubated with 40 mCi of 45Ca2‡ at pH 6.0 or 7.5, and then centrifuged into 10-40 % glycerol gradients. The amount of 45Ca2‡ was determined in the virus peak fractions.

…Fref ÿ Fo †2 ˆ F2o

1=2

NCS electron density averaging followed the same procedure as described for the FPV pH 6.2 data (ID7 in Table 2A). The relatively low ®nal correlation coef®cient (Table 4) is possibly due to the high mosaicity of the crystal, as a result of the freezing procedure. This caused the overlap of re¯ections on the image and the distortion of the re¯ection pro®les. The structure was not re®ned.

CPV-d at 100K, pH 6.2 and pH 5.5 (ID8 and 9 in Table 2A) dron (taken to be that given by the orientation of the PDB entry 1FPV), is de®ned as: X ˆ ‰PŠY where X and Y are the Cartesian coordinates with respect to the standard orientation and the crystal unit cell, respectively. The orthogonalization convention was de®ned as having the Y2 axis along b and the Y1 axis along a*. It was found that: 0 1 ÿ0:5403 ÿ0:8131 0:2169 ‰PŠ ˆ @ 0:3037 0:0519 ÿ0:9513 A 0:7848 ÿ0:5798 0:2189 The structure was re®ned with the CNS program in the same manner as for CPV-d (A300 ! D) (ID4). Calcium ions were placed at sites 2 and 3. FPV at 100K and pH 7.5 (ID6 in Table 2A) Crystals of FPV were grown at pH 7.5 with 8 mM CaCl2 and then soaked for ten minutes in 20 mM Tris buffer containing 5 % PEG 8000 (w/v) with 8 mM CaCl2, and 30 % (v/v) ethylene glycol as a cryo-protectant. Diffraction data were collected from three frozen crystals using a crystal-to-detector distance of 350 mm, with Ê radiation and an exposure time of 60 seconds. 0.917 A Ê resolution using the The data were processed to 3.5 A HKL system (Otwinowski & Minor, 1997). The crystals were isomorphous with the FPV crystals used for the

Crystals of CPV-d were grown at pH 6.2 and 5.5 with 8 mM CaCl2. The cryoprotectant was the same as described above. In both cases, the data were collected from a single frozen crystal with a crystal-to-detector (CCD Quantum 4, ADSC Co.) distance of 380 mm and Ê wavelength radiation. 0.92 A The pH 5.5 data set was processed with the DPS data processing suite (Rossmann & van Beek, 1999) using MOSFLM (Leslie, 1992) for intensity integration and the program SCALA for scaling (Table 4). Re¯ections with partiality greater than 0.05 were used by separately correcting measurements to the equivalent full intensity value, similar to the procedure used in SNP. The lesscomplete pH 6.2 data set was scaled using the already merged pH 5.5 data as a reference (see above and Table 4). The initial structure determination was performed with the better and more complete pH 5.5 data set. The pH 5.5 and 6.2 crystals were found to be isomorphous, in space group P1, with two particles in the unit cell, Ê 3/Da. Particle giving a Matthews coef®cient of 2.2 A orientations were determined as described above. The ®rst particle center was arbitrarily set to (0,0,0), and the second particle was located relative to the ®rst with the translation function program TFFC (Collaborative Computational Project Number 4, 1994). CPV-Alabama (ID1) was used as a search model and subsequently for initiating phasing, and the position and orientation of the particles were further re®ned as described above. The ®nal rotation matrices used to orient the two particles in the unit cell were (see above for de®nition of [P]):

608

Calcium Binding in Canine and Feline Parvoviruses

Figure 7. Histogram showing variability of HA for cat, Rhesus monkey, and horse erythrocytes. Whereas CPV can hemagglutinate all types of erythrocytes at pH 6.2 or 7.5, FPV hemagglutinates only at pH 7.5. However, mutations of residues in the region of the ¯exible loop (D367A, D375N, D373N, and D323N) diminish the pH dependence in FPV. Similarly, mutation of N375D can partially create pH dependence in CPV.

0

ÿ0:36572 Particle 1 ‰PŠ ˆ @ 0:26911 0:89097

ÿ0:86570 0:25316 ÿ0:43182

1 ÿ0:34176 ÿ0:92924 A 0:14038

ÿ0:70754 0:15976 ÿ0:68837

1 ÿ0:38857 ÿ0:90158 A 0:19015

and: 0

ÿ0:59025 Particle 2 ‰PŠ ˆ @ 0:40202 0:69999

with a ®nal position of the second particle at (0.4662, 0.5259, 0.4944). After 23 cycles of NCS electron density averaging, the ®nal correlation coef®cient was 0.58 for the pH 5.5 data set. The pH 5.5 structure was re®ned with the CNS program as described above, except that the structure amplitudes obtained by back-transformation of the ®nal averaged map were used as the target for re®nement. This gave better results because of the poor quality of the original observed data. The pH 5.5 CPV structure was then used as a starting model to solve the pH 6.2 CPV structure, giving a ®nal correlation coef®cient of 0.57 after 50 cycles of NCS electron density averaging. FPV at pH 6.2 in the presence of EDTA (ID10 in Table 2A) A crystal of FPV was grown in 20 mM Bis-Tris-HCl (pH 6.2), in the presence of 5 mM EDTA. Data were collected on beamline 14BM-c at the Advanced Photon Source (APS), using an ADSC Quantum 4 CCD detector Ê wavelength at a distance of 350 mm, with 1.000 A radiation and 0.3  oscillations. A total of 599 frames was Ê resolution using the collected and processed to 3.0 A HKL system. A self-rotation function was calculated using GLRF, with a kappa angle of 72  ; the directions of the 5-fold axes were used to orient the particle. The particle position was found by calculating a translation function with the program TFFC, using FPV (accession number 1FPV) as a search model. Initial rigid body re®nement was carried out using the program Ê resolution. Restrained CNS, using data from 9 to 3.5 A

individual positional and temperature factor re®nements were then performed with NCS constraints (Table 4). This model was used to calculate phases that were re®ned using NCS averaging, as described above, which Ê resolgave a ®nal correlation coef®cient of 0.865 to 3 A ution after 20 cycles. Upon examination of the electron density map, the calcium ions were removed from sites 2 and 3, and the ¯exible loop was removed from the structure because it was disordered. Residues 402 to 403 were rebuilt. An (Fo ÿ Fc) map was calculated using the re®ned atomic model, with the variable regions of the structure omitted. This map showed the same structural changes as the averaged map (Table 4). The ®nal particle orientation was found to be: 0 1 0:69595 0:38880 0:60373 ‰PŠ ˆ @ ÿ0:71119 0:25687 0:65439 A 0:09934 ÿ0:88479 0:45528 with the particle center at (0.2426, 0.0000, 0.2486). CPV-d at 100K, pH 6.2 with GdCl3 (ID11 in Table 2A) Crystals used for the Gd studies were grown at pH 6.2, as described for the other experiments above, and soaked in 22 mM GdCl3. The X-ray diffraction data were collected from a single frozen crystal using a CCD Quantum 4 detector, with a crystal-to-detector distance Ê wavelength radiation. The data of 350 mm and 0.923 A were scaled using the already merged CPV-d pH 5.5 data set as a reference because of the incompleteness of the GdCl3 data. Difference map coef®cients were calculated using the program SIGMAA (Collaborative Computational Project Number 4, 1994). The heights of peaks in the Gd3‡ difference map are shown in Table 3B. Deposition of coordinates Coordinates and structure factors here have been deposited with the PDB. Accession numbers are recorded in Table 2A.

Calcium Binding in Canine and Feline Parvoviruses

Acknowledgments We are grateful to the staff of CHESS (Cornell) and BioCARS of APS (Argonne) for assistance in the data collection. We thank Cheryl Towell and Sharon Wilder for help in the preparation of the manuscript. The work was supported by a National Institutes of Health grant to M.G.R. and C.R.P. (AI33468), as well as a Purdue University reinvestment grant. B.H. was supported by a postdoctoral fellowship from the Fonds pour la Formation de Chercheurs et l'Aide aÁ la Recherche (Fonds FCAR) of QueÂbec.

References Agbandje, M., McKenna, R., Rossmann, M. G., Strassheim, M. L. & Parrish, C. R. (1993). Structure determination of feline panleukopenia virus empty particles. Proteins: Struct. Funct. Genet. 16, 155-171. Agbandje, M., Kajigaya, S., McKenna, R., Young, N. S. & Rossmann, M. G. (1994). The structure of human Ê resolution. Virology, 203, 106parvovirus B19 at 8 A 115. Agbandje-McKenna, M., Llamas-Saiz, A. L., Wang, F., Tattersall, P. & Rossmann, M. G. (1998). Functional implications of the structure of the murine parvovirus, minute virus of mice. Structure, 6, 1369-1381. Arella, M., Garzon, S., Bergeron, J. & Tijssen, P. (1990). Physiochemical properties, production, and puri®cation of parvoviruses. In Handbook of Parvoviruses (Tijssen, P., ed.), pp 11-30, CRC Press, Inc., Boca Raton. Argos, P. & Rossmann, M. G. (1980). Molecular replacement method. In Theory and Practice of Direct Methods in Crystallography (Ladd, M. F. C. & Palmer, R. A., eds), pp. 361-417, Plenum, New York. Ball-Goodrich, L. J. & Tattersall, P. (1992). Two amino acid substitutions within the capsid are coordinately required for acquisition of ®brotropism by the lymphotropic strain of minute virus of mice. J. Virol. 66, 3415-3423. Barbis, D. P., Chang, S. F. & Parrish, C. R. (1992). Mutations adjacent to the dimple of the canine parvovirus capsid structure affect sialic acid binding. Virology, 191, 301-308. Basak, S. & Turner, H. (1992). Infectious entry pathway for canine parvovirus. Virology, 186, 368-376. Bergeron, J., HeÂbert, B. & Tijssen, P. (1996). Genome organization of the Kresse strain of porcine parvovirus: identi®cation of the allotropic determinant and comparison with those of NADL-2 and ®eld isolates. J. Virol. 70, 2508-2515. Bolotovsky, R., Steller, I. & Rossmann, M. G. (1998). The use of partial re¯ections for scaling and averaging X-ray area detector data. J. Appl. Crystallog. 31, 708717. BruÈnger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N. S., Read, R. J., Rice, L. M., Simonson, T. & Warren, G. L. (1998). Crystallography and NMR System: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905-921. Chang, S. F., Sgro, J. Y. & Parrish, C. R. (1992). Multiple amino acids in the capsid structure of canine parvovirus coordinately determine the canine host range

609 and speci®c antigenic and hemagglutination properties. J. Virol. 66, 6858-6867. Chapman, M. S. (1993). Mapping the surface properties of macromolecules. Protein Sci. 2, 459-469. Chapman, M. S. & Rossmann, M. G. (1996). Structural re®nement of the DNA-containing capsid of canine parvovirus using RSRef, a resolution-dependent stereochemically restrained real-space re®nement method. Acta Crystallog. sect. D, 52, 129-140. Collaborative Computational Project Number 4 (1994). The CCP4 Suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760-763. Durham, A. C. H., Henry, D. A. & von Wechar, M. B. (1977). Does calcium ion binding control plant virus disassembly? Virology, 77, 524-533. Giranda, V. L., Heinz, B. A., Oliveira, M. A., Minor, I., Kim, K. H., Kolatkar, P. R., Rossmann, M. G. & Rueckert, R. R. (1992). Acid-induced structural changes in human rhinovirus 14: possible role in uncoating. Proc. Natl Acad. Sci. USA, 89, 1021310217. Haynes, J. I., Chang, D. & Consigli, R. A. (1993). Mutations in the putative calcium-binding domain of polyomavirus VP1 affect capsid assembly. J. Virol. 67, 2485-2495. Horiuchi, M., Ishiguro, N., Goto, H. & Shinagawa, M. (1992). Characterization of the stage(s) in the virus replication cycle at which the host-cell speci®city of the feline parvovirus subgroup is regulated by canine cells. Virology, 189, 600-608. Kim, S., Boege, U., Krishnaswamy, S., Minor, I., Smith, T. J., Luo, M., Scraba, D. G. & Rossmann, M. G. (1990). Conformational variability of a picornavirus capsid: pH-dependent structural changes of Mengo virus related to its host receptor attachment site and disassmebly. Virology, 175, 176-190. Kunkel, T. A. (1987). Rapid and ef®cient site-speci®c mutagenesis without phenotypic selection. Methods Enzymol. 154, 367-382. Leslie, A. G. W. (1992). Recent changes to the MOSFLM package for processing ®lm and image plate data. In CCP4 and ESF-EACMB Newsletter on Protein Crystallography, No. 26, Daresbury Laboratory, Daresbury, UK. Liljas, L. (1986). The structure of spherical viruses. Progr. Biophys. Mol. Biol. 48, 1-36. Llamas-Saiz, A. L., Agbandje-McKenna, M., Parker, J. S. L., Wahid, A. T. M., Parrish, C. R. & Rossmann, M. G. (1996). Structural analysis of a mutation in canine parvovirus which controls antigenicity and host range. Virology, 225, 65-71. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491-497. McRee, D. E. (1993). Practical Protein Crystallography, Academic Press, San Diego. Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505-524. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307-326. Parker, J. S. & Parrish, C. R. (1997). Canine parvovirus host range is determined by the speci®c conformation of an additional region of the capsid. J. Virol. 71, 9214-9222. Parker, J. S. L. & Parrish, C. R. (2000). Cellular uptake and infection by canine parvovirus involves rapid dynamin-regulated clathrin-mediated endocytosis,

610

Calcium Binding in Canine and Feline Parvoviruses

followed by slower intracellular traf®cking. J. Virol. 74, 1919-1930. Parrish, C. R. (1991). Mapping speci®c functions in the capsid structure of canine parvovirus and feline panleukopenia virus using infectious plasmid clones. Virology, 183, 195-205. Parrish, C. R., Aquadro, C., Strassheim, M. L., Evermann, J. F., Sgro, J.-Y. & Mohammed, H. O. (1991). Rapid antigenic-type replacement and DNA sequence evolution of canine parvovirus. J. Virol. 65, 6544-6552. Rhode, S. L. I. (1985). Nucleotide sequence of the coat protein gene of canine parvovirus. J. Virol. 54, 630633. Robinson, I. K. & Harrison, S. C. (1982). Structure of the expanded state of tomato bushy stunt virus. Nature, 297, 563-568. Rossmann, M. G. (1979). Processing oscillation diffraction data for very large unit cells with an automatic convolution technique and pro®le ®tting. J. Appl. Crystallog. 12, 225-238. Rossmann, M. G. & Johnson, J. E. (1989). Icosahedral RNA virus structure. Annu. Rev. Biochem. 58, 533573. Rossmann, M. G. & van Beek, C. G. (1999). Data processing. Acta Crystallog. sect. D, 55, 1631-1640. Rossmann, M. G., McKenna, R., Tong, L., Xia, D., Dai, J., Wu, H., Choi, H. K. & Lynch, R. E. (1992). Molecular replacement real-space averaging. J. Appl. Crystallog. 25, 166-180. Simpson, A. A., Chipman, P. R., Baker, T. S., Tijssen, P. & Rossmann, M. G. (1998). The structure of an insect parvovirus (Galleria mellonella densovirus) at Ê resolution. Structure, 6, 1355-1367. 3.7 A Spitzer, A. L., Maxwell, F., Corsini, J. & Maxwell, I. H. (1996). Species speci®city for transduction of cul-

tured cells by a recombinant LuIII rodent parvovirus genome encapsidated by canine parvovirus or feline panleukopenia virus. J. Gen. Virol. 77, 17871792. Tong, L. & Rossmann, M. G. (1990). The locked rotation function. Acta Crystallog. sect. A, 46, 783-792. Tresnan, D. B., Southard, L., Weichert, W., Sgro, J. Y. & Parrish, C. (1995). Analysis of the cell and erythrocyte binding activities of the dimple and canyon regions of the canine parvovirus capsid. Virology, 211, 123-132. Tsao, J., Chapman, M. S., Agbandje, M., Keller, W., Smith, K., Wu, H., Luo, M., Smith, T. J., Rossmann, M. G., Compans, R. W. & Parrish, C. R. (1991). The three-dimensional structure of canine parvovirus and its functional implications. Science, 251, 14561464. Vihinen-Ranta, M., Yuan, W. & Parrish, C. R. (2000). Cytoplasmic traf®cking of the canine parvovirus capsid and its role in infection and nuclear transport. J. Virol. 74, 4853-4859. Weichert, W. S., Parker, J. S. L., Wahid, A. T. M., Chang, S. F., Meier, E. & Parrish, C. R. (1998). Assaying for structural variation in the parvovirus capsid and its role in infection. Virology, 250, 106117. Wu, H. & Rossmann, M. G. (1993). The canine parvovirus empty capsid structure. J. Mol. Biol. 233, 231244. Xie, Q. & Chapman, M. S. (1996). Canine parvovirus Ê resolution. J. Mol. capsid structure, analyzed at 2.9 A Biol. 264, 497-520. Yasue, S., Handa, S., Miyagawa, S., Inoue, J., Hasegawa, A. & Yamakawa, T. (1978). Difference in form of sialic acid in red blood cell glycolipids of different breeds of dogs. J. Biochem. 83, 1101-1107.

Edited by I. A. Wilson (Received 6 March 2000; received in revised form 11 May 2000; accepted 15 May 2000)