|Home | About | Journals | Submit | Contact Us | Français|
Two strains of the parvovirus minute virus of mice (MVM), the immunosuppressive (MVMi) and the prototype (MVMp) strains, display disparate in vitro tropism and in vivo pathogenicity. We report the crystal structures of MVMp virus-like particles (MVMpb) and native wild-type (wt) empty capsids (MVMpe), determined and refined to 3.25 and 3.75 Å resolution, respectively, and their comparison to the structure of MVMi, also refined to 3.5 Å resolution in this study. A comparison of the MVMpb and MVMpe capsids showed their structures to be the same, providing structural verification that some heterologously expressed parvovirus capsids are indistinguishable from wt capsids produced in host cells. The structures of MVMi and MVMp capsids were almost identical, but local surface conformational differences clustered from symmetry-related capsid proteins at three specific domains: (i) the icosahedral fivefold axis, (ii) the “shoulder” of the protrusion at the icosahedral threefold axis, and (iii) the area surrounding the depression at the icosahedral twofold axis. The latter two domains contain important determinants of MVM in vitro tropism (residues 317 and 321) and forward mutation residues (residues 399, 460, 553, and 558) conferring fibrotropism on MVMi. Furthermore, these structural differences between the MVM strains colocalize with tropism and pathogenicity determinants mapped for other autonomous parvovirus capsids, highlighting the importance of common parvovirus capsid regions in the control of virus-host interactions.
Viral tissue tropism and pathogenesis are highly dependent on well-orchestrated and defined interactions that occur between the viral pathogen and its host. The single-stranded Parvoviridae infect a broad range of natural hosts that include invertebrates and mammals, and pathogenic members cause serious disease in the young and immunocompromised adults. However, disparities in tissue tropism and in vivo pathogenicity can be observed between highly homologous strains for several members of the Parvoviridae. The combination of molecular analysis with disease outcomes for a number of strains of the Aleutian mink disease parvovirus (AMDV) (9, 10), canine parvovirus (CPV), and feline panleukopenia virus (FPV) (46, 61, 62), and for porcine parvovirus (PPV) (8, 41), has provided models for probing the determinants of parvovirus tissue tropism and pathogenicity. In the parvovirus minute virus of mice (MVM) model adopted in this study, the prototype strain (MVMp) replicates in vitro in mouse fibroblast cell lines, while the immunosuppressive strain (MVMi) replicates in mouse T lymphocytes and mouse hematopoietic precursors. In vivo, MVMp infection of newborn mice is asymptomatic, while MVMi infection of newborn mice as well as of adult mice with severe combined immunodeficiency syndrome (SCID) is lethal (11, 12, 47, 51, 52).
Parvovirus capsids are ~260 Å in diameter and contain 60 copies (in total) of viral protein 1 (VP1) to VP4, in a T=1 icosahedral capsid arrangement, with the smallest VP, depending on the virus member, being the major capsid viral protein. The number of capsid VP species per virion differs among parvoviruses. For example, the AMDV capsid contains only two polypeptides, VP1 and VP2; MVM has three, VP1 to VP3; and members of the subfamily Densovirinae have four, VP1 to VP4. The VPs are overlapping, with the entire sequence of VP4 contained within VP3; VP3 is contained in VP2, which is in turn contained within VP1; VP1 has a unique N-terminal domain (58). The VPs are translated from the same mRNA or result from posttranslational cleavage. For example, in MVM, VP3 is formed by postassembly cleavage of approximately 18 to 20 amino acids from the N terminus of VP2 in full (DNA-containing) infectious virions. This cleavage event is not seen in empty (no DNA) particles digested in vitro with trypsin, unless exposure of VP2 is induced by heat (19, 27). The molecular sizes of the MVM proteins are 83,000, 64,000, and 61,000 Da for VP1, VP2, and VP3, respectively.
The 3-dimensional structures of several parvoviruses, adeno-associated virus serotype 2 (AAV2), human parvovirus B19 (B19), CPV, Galleria mellonella densovirus (GmDNV), FPV, MVMi, and PPV, have been determined using X-ray crystallography (1-3, 33, 54, 55, 63, 67, 68). Crystal structures of host range and antigenic mutants of CPV, as well as of CPV and FPV under various pH and ionic conditions, have also been determined (26, 36, 53). The structures of AMDV strain G (AMDV-G), AAV2, AAV4, AAV5, B19-globoside receptor complex, CPV-Fab complex, and Junonia coenia densovirus (JcDNV) have been determined by cryoelectron microscopy and image reconstruction (13, 18, 34, 39, 44, 64, 65). The parvovirus capsid VP topology is highly conserved, even for members that are only ~20% identical at the amino acid sequence level, such as AAV2 and B19 (33), with variations localized to surface loop regions between strands of a core β-barrel domain. The characteristic parvovirus capsid surface features include protrusions at or surrounding the icosahedral threefold axes and depressions at the icosahedral twofold axes and around the fivefold axes. Exceptions to this general surface topology are seen in the smoother GmDNV and JcDNV capsids as a result of smaller loop insertions between their β-strands (13, 54).
Structural mapping of amino acids reported to control tissue tropism and pathogenicity onto the capsid proteins of AMDV-G, CPV, FPV, MVMi, and PPV localize these regions on or close to variable capsid surface loops (2, 3, 6, 8, 9, 17, 24, 26, 28, 30, 31, 38, 45, 53, 55, 61, 62, 69). In a continued effort to elucidate the role of the parvovirus capsid structure in tropism and pathogenicity by utilizing the MVM model, we report the structures of native wild-type (wt) empty capsids of MVMp containing VP1 and VP2 (MVMpe) and of baculovirus-expressed VP2 virus-like capsids (VLPs) (MVMpb) determined and refined to 3.75 and 3.25 Å resolution, respectively. The MVMpb structure is further compared to that of MVMi (3), also refined in the present study. We observe local surface conformational differences between the capsids of the MVM strains occurring at the icosahedral fivefold axis, on the “shoulder” of protrusions at the icosahedral threefold axis, and in and surrounding the depression at the icosahedral twofold axis. Comparison of the MVM VP2 structures to those of CPV, FPV, and PPV identified similar hot spots of surface loop structure variations, some of which colocalize with CPV/FPV, PPV, and ADV strain tissue tropism and pathogenicity determinants. These observations suggest the common utilization of variable parvovirus capsid regions by highly homologous viral strains with different tropisms and pathogenicities.
MVMpe, MVMpb, and empty MVMi capsids were produced and purified as previously described (27, 37). The crystallization and preliminary characterization of the X-ray diffraction data for MVMpb have been reported elsewhere (27). MVMpe crystals were grown under the same conditions used for MVMpb (27), with the virus at a concentration of 10 mg/ml in 10 mM Tris-HCl (pH 7.5)-8 mM CaCl2, and with 0.5 to 1.0% polyethylene glycol 8000 as a precipitant.
Data on MVMpb and MVMpe crystals were collected at 4°C on a MAR30 image plate and on an ADSC Quantum 4 CCD detector at beamline 9.6 at the Daresbury Synchrotron Radiation Source (United Kingdom), operating at a wavelength (λ) of 0.870 Å, and on a MAR30 image plate detector at the EMBL X31 beamline at DESY (Hamburg, Germany), operating at a λ of 1.071 Å. Oscillation images were indexed and processed with the DENZO program (43) and were scaled, merged, and postrefined using SCALEPACK (43). The space groups for the crystals were monoclinic C2, with postrefined cell parameters (Table (Table1)1) that were pseudo-isomorphous to that reported for MVMi (a = 448.7 Å, b = 416.7 Å, c = 305.3 Å, and β = 95.8°) (37). The data processing and scaling of images from a total of 30 crystals for MVMpb and 26 crystals for MVMpe followed procedures utilized for MVMi (37). The data collection and processing statistics are summarized in Table Table11.
The capsid structures of MVMpb and MVMpe were determined by molecular replacement (50), involving the determination of particle orientations and positions within the crystal unit cell. The MVMpb and MVMpe C2 cells contained two half-particles with different orientations in the asymmetric unit as previously described for MVMi (37). The orientations of the two half-particles were determined initially by a self-rotation function (59) computed using ~11% of the data in the 10 to 5 Å resolution range as large terms to represent the second Patterson function and then, more accurately, by the locked self-rotation function (59) computed using ~16% of the data in the 4 to 3.25 Å resolution range for MVMpb and ~16% in the 5 to 3.75 Å resolution range for MVMpe. Packing considerations suggested that the centers of the complete T=1 MVMp particles are located on the crystallographic twofold axes at (0, 0, 0) (particle 1) and (0, ~1/2, 1/2) (particle 2), as was also reported for MVMi (37). Molecular replacement calculations, applying 60-fold noncrystallographic symmetry (NCS), were performed using the Purdue suite of programs (50), with the structure of CPV as the initial phasing model, for data in the 20 to 3.25 and 20 to 3.75 Å resolution ranges for MVMpb and MVMpe, respectively. The inner and outer radii of the particles were set at 70 and 145 Å, respectively, to define solvent boundaries. The calculations were performed in an artificial “H-cell,” a unit cell containing a single particle in a known standard orientation, as defined by Rossmann et al. in 1992 (50), and were utilized as described in the structure determination of MVMi (37). The MVMi structure (3) was not considered as a starting model to avoid phase bias due to the high degree of identity (97%) between the amino acid sequences of MVMi and MVMp. The “climb” procedure of the ENVELOPE program (50) was used to refine the orientations and positions of the particles several times during phase refinement cycles.
After 75 cycles of molecular replacement calculations, the final orientations (ψ, , and κ) for the two half-particles were (0, 0, 29.83°) and (0, 0, −106.42°) for MVMpb and (0, 0, 30.06°) and (0, 0, −106.78°) for MVMpe. The final particle positions (in fractional coordinates) were (0, 0, 0) and (0, 0.49985, 0.50000) for MVMpb and (0, 0, 0) and (0, 0.50000, 0.50000) for MVMpe. The final averaging correlation coefficients were 0.74 and 0.71 for MVMpb and MVMpe, respectively. Residues 39 to 587 were built into the H-cell electron density map. The resulting model coordinates were transformed to the crystallographic unit cell by using the [P] matrix, which defines a rotational relationship between the structure in the H-cell and the reference particle in the crystallographic unit cell (50). These models were used for all subsequent refinement steps of the MVMpb and MVMpe structures.
Crystallographic model refinement and all further electron density map calculations were carried out using the CNS program (15). All reflections in the 20.0 to 3.5 Å resolution range for MVMi (PDB accession no. 1MVM) (3), in the 20.0 to 3.25 Å resolution range for MVMpb, and in the 20 to 3.75 Å resolution range for MVMpe were used during the refinement, with 5% of the data sets partitioned in a test set for monitoring the refinement process (14). Several alternating cycles of manual model rebuilding using the interactive molecular graphics program O (32) and refinement improved the quality of the models. The refinement protocol consisted of bulk solvent correction, geometry regularization, and least-squares conjugate-gradient refinement, followed by simulated annealing, conventional positional refinement, and individual restrained B-factor refinement. Strict NCS was applied to generate symmetry-related subunits from the coordinates of a single VP2 subunit. The MVM VP2 models were manually inspected with simulated-annealed omit maps and sigma-weighted 60-fold averaged 2Fo-Fc and Fo-Fc electron density maps and were adjusted to fit the density. N-terminal residues 1 to 38 were not built into the MVMp structures due to poor density, and residues 29 to 38, modeled in the MVMi structure (3), remain unrefined. Water molecules were added into unassigned positive electron density (at 1.5 σ) in difference Fourier maps that were within hydrogen bond donor or acceptor distances. The VP2 final models, residues 39 to 587, were examined for main-chain torsion angles using the PROCHECK program (35).
The MVMpb, MVMpe, and MVMi VP2 structures were compared using the LSQ subroutine in the O program (32), the Homology program (49), and GRASP (42). Figures were generated using the programs BOBSCRIPT (22), Excel (Microsoft, Inc.), GRASP (42), Raster3D (40), and PyMol (21).
Equivalent amounts (0.4 μg) of native MVMi (37), MVMpb, and MVMpe (27) empty capsids in 40 μl of 50 mM Tris-HCl at pH 6.0, 7.5, or 8.8 were simultaneously heated for 10 min in water baths at the temperatures of 25, 70, 75, 80, and 85°C and shock frozen in dry ice. Heated samples were diluted to a final volume of 100 μl with 10× phosphate-buffered saline and H2O and then assayed for hemagglutination (HA) activity with 2% mouse erythrocytes in phosphate-buffered saline (2 h at 4°C) as previously described (27).
The atomic models of wt full and empty CPV, CPV mutants CPV-N93R (full) and CPV-N93D (empty), FPV (empty), and PPV (empty) were obtained from the Protein Data Bank (PDB accession no. 4DPV, 2CAS, 1P5W, 1P5Y, 1C8E, and IK3V). These structures, plus the refined models of MVMi and MVMpb, were superimposed with the least-squares subroutine in the O program (32) to obtain an overall root mean square deviation (RMSD) for their Cα positions. Differences in individual Cα positions were calculated, relative to the refined structure of MVMpb, using the least-squares Homology program (49).
The refined coordinates for MVMpb and MVMi VP2 have been deposited with the Protein Data Bank (PDB accession no. 1Z14 and 1Z1C, respectively).
The structures of MVMpb and MVMpe have been determined to 3.25 and 3.75 Å resolution, respectively. Residues 39 to 587 (the latter is the last C-terminal residue) of the VP2 capsid sequence were built into the MVMp electron density maps (Fig. 1A to G) resulting from molecular replacement and refinement procedures (Tables (Tables11 and and2)2) (15, 50, 59). The MVMp density maps were not interpretable beyond N-terminal residue 39 of the VP2 amino acid sequence. This is in contrast to the interpretation of the VP2 electron density map for the full MVMi capsid. In this DNA-containing structure, density within the icosahedral fivefold channel was built as 10 additional residues extending from N-terminal residue 39, which is located directly under the fivefold channel (3). The lack of ordered density within the fivefold channel of the empty MVMp capsid is similar to observations for other empty parvovirus capsid structures (2, 55, 67) and is consistent with the postulation that the VP2-to-VP3 cleavage that occurs in full (and not in empty) capsids and VP1 exposure of phospholipase A2 activity (23) are likely facilitated by N-terminal externalization via this channel.
The MVMp and MVMi VP2 models were refined using the CNS program (14, 15). The final Rfactor and Rfree were 29.85 and 30.62% for MVMpb, 32.35 and 32.95% for MVMpe, and 32.55 and 32.93% for MVMi (Table (Table2).2). The similarity of Rfactor and Rfree for virus structures stems from the high noncrystallographic icosahedral symmetry of their capsid. These R values are comparable to those quoted for parvoviruses and other virus capsid structures, as detailed on the VIPER website (48). The stereochemical parameters and geometries of the model (Table (Table2)2) were consistent with those reported for other virus structures at comparable resolution.
Comparison of the electron density maps and models for the MVMpb and MVMpe VP2s showed them to be superimposable (Fig. 1A and B), with an RMSD of 0.08 Å for Cα atoms and 0.25 Å for all 4,317 atoms (for residues 39 to 587). This observation provides structural verification that some parvovirus particles produced in a heterologous system are indistinguishable from native capsids produced in host systems (27). The results also support reports that the major capsid protein of MVM is sufficient to form native-like particles that are similar to wt particles (66). Based on the structural identity of the MVMpb and MVMpe VP2s, the higher-resolution MVMpb structure was used for all subsequent comparisons to the MVMi VP2 structure.
The MVM VP2 structures have the general parvovirus capsid viral protein topology (4). An eight-stranded β-barrel motif (βB to βI) forms the core contiguous capsid, decorated by loop insertions between the β-strands (Fig. (Fig.1K).1K). Small stretches of antiparallel β-strands are observed in the loops between the core strands, as was reported for CPV (69). A small α-helix (αA) spanning residues 125 to 135 that is conserved in all the parvovirus structures determined so far and lies close to the icosahedral twofold axis is also present in the MVM VP2 structure (Fig. (Fig.1K).1K). A channel at the parvovirus icosahedral fivefold axis formed by the clustering of five symmetry-related β-ribbons (residues 153 to 171) between βD and βE (Fig. 1F to K) is conserved (Fig. 2A and B). A protrusion is centered at the icosahedral threefold axes (Fig. (Fig.2C),2C), resulting from the clustering of six large surface loops, two from each threefold-symmetry-related VP2 subunit. These loops are between βE and βF (residues 217 to 239) and between βG and βH (residues 405 to 455). The MVM capsid radii at the icosahedral threefold axis are ~135 Å, while the three apexes that surround it are at ~150 Å, leading to a slight depression at the center of the protrusions (Fig. 2A and C). A difference in the side chain conformation of surface residue E229 (torsion angles for MVMpb/MVMi are as follows: chi1 = 69/−78°, chi2 = −69/−178°, and chi3 = −54/−54°), located at the three vertices of the threefold protrusion, results in a more “pointed” appearance in the MVMi capsid than in the MVMp capsid (Fig. 2A and C). The shoulder of the protrusion is formed by loops between βB and βC (residues 83 to 108) and between βG and βH (residues 284 to 360). Depressions are observed at and surrounding the icosahedral twofold axes, with capsid radii of ~105 Å (Fig. 2A and D), and surrounding the cylindrical structure, with a channel at icosahedral fivefold axes. The capsid radii at the icosahedral fivefold axis are ~130 Å (Fig. 2A and B).
The VP2 structures of MVMi and MVMpb are almost identical, with the Cαs of residues 39 to 587 superimposing with an RMSD of 0.48 Å (for 546 out of 549 residues) (Fig. (Fig.1K).1K). A total of 14 amino acids (residues 10, 160, 232, 317, 321, 362, 366, 368, 388, 402, 410, 440, 455, and 551) differ between the MVMi and MVMp VP sequences (5); all of these are within VP2. N-terminal residue 10 (glycine in MVMi and serine in MVMp) is not ordered in the current VP2 structures. At the resolution of the electron density maps calculated, the densities for the remaining 13 amino acids were consistent with the amino acid types (examples are given in Fig. 1A to G). Twelve of the 13 residues (except for residue 160, discussed below) are clustered, from symmetry-related VP monomers, in the depression at the icosahedral twofold axes, on the walls of the twofold depression, and at the shoulder of the protrusions at the icosahedral threefold axes that surround the depression (Fig. 3A and B). Eight of these 12 residues (residues 232, 321, 362, 366, 368, 388, 410, and 440) are surface exposed (Fig. 3A and B), and residue 317 is solvent accessible. Residues 399, 460, 553, and 558, which confer fibrotropism on MVMi (3), are also located in the twofold depression or on the wall surrounding it, with residues 399, 553, and 558 on the capsid surface (Fig. 3A and B). Forward mutations are selected at these four positions in the MVMi sequence when this virus, with a site-directed mutation at allotropic residue 317 or 321 changing the amino acid type to that of MVMp, is used to infect mouse fibroblasts.
Variations in the side chains of the 12 differing MVMi/p amino acids discussed above and in the side chains of the 4 residues that are selected in forward fibrotropic MVMi mutations (but are the same in the wt viruses) result in distinct local surface topologies (Fig. 3A and B) and intersubunit contacts close to and in the depression at the icosahedral twofold axes of the MVMi and MVMpb capsids (Fig. 3C to F). A chain of weak intra and intersubunit amino acid interactions from the wall (E321) toward the floor (R368, D399) of the twofold depression is observed in MVMi (Fig. (Fig.3F).3F). These interactions involve MVMi/p differing and forward mutation residues, with a weak ionic contact between R368 and E321 (allotropic residue) from a threefold-symmetry-related VP2 monomer, and a hydrogen bond between residues R368 and D399 (Fig. (Fig.3F).3F). These interactions are not possible in MVMpb, which contains G321 and K368, and the terminal oxygens of the D399 side chain are too far from the terminal amino group of K368 for a hydrogen-bonding interaction (Fig. (Fig.3E).3E). In addition, MVMi/p differences are observed in the side chain conformations of residue E400 (immediately under the capsid surface beneath residue D399) (Fig. 3D to F). This residue is involved in hydrogen bond interactions with residue S460 (a forward fibrotropic mutant residue that is nonsurface) in the MVMi and MVMpb structures, although the interaction is slightly stronger in MVMi (Fig. (Fig.3F)3F) than in MVMpb (Fig. (Fig.3E).3E). Interestingly, the site-directed/forward mutation combinations (A317T/D399G, A317T/D399A, A317T/D553N, E321G/A317T, E321G/S460A, and E321G/Y558H) that confer fibrotropism on MVMi (3) eliminate the potential for ionic interactions by the forward mutant residues (Fig. 3E and F) or alter the surface charge within the vicinity of the twofold depression. These observations and the refined MVMpb structure, in which the number of ionic interactions involving MVMi/p differing residues is lower than that in the MVMi capsid, support previous suggestions that fibrotropism is likely controlled by a less stable MVM capsid (see also below) and/or a reduced acidic environment at or close to the icosahedral twofold axis (3). Coincidently, altered intra- and intersubunit interactions involving tropism and pathogenicity determinants are also observed in comparisons of CPV to FPV and to host range mutants with altered receptor binding properties (26). A predicted main-chain distortion in the MVMp capsid due to the allotropic amino acid difference (A317T) between MVMi and MVMp (3) was not observed in the MVMpb structure.
The remaining MVMi/p differing residue, residue 160 (of the 13 residues ordered within VP2), which is a serine in MVMi and a leucine in MVMp, is located close to the top of the loop between the β-strands making up the β-ribbons that form the fivefold channel (Fig. 1F to J). This single-amino-acid change causes a drastic conformational change in this loop region, from residue 157 to 164, an area that was clearly interpretable in 2Fo-Fc electron density maps contoured at a 1.8 σ level (Fig. 1F and G). The different loop conformations create a topology at the top of the fivefold cylinder in MVMpb that is distinct from that in MVMi (Fig. 2A and B). The loop rearrangement increases the diameter at the top of the fivefold channel in MVMpb (16.7 Å) relative to MVMi (8.5 Å) (Fig. 1I and J). This dramatic conformational rearrangement was unexpected for a serine/leucine side chain difference. There is a possibility that the topology difference is due to a capsid structural rearrangement resulting from VP2 N-terminal externalization for cleavage to VP3 in the full MVMi capsid (which would not occur in the empty MVMpb capsid) rather than to the difference in residue 160. Comparisons of other full and empty parvovirus structures have shown variation in the topology of this fivefold loop, postulated to facilitate its predicted role in enabling VP externalization (3, 23, 26, 63, 69), although the differences are generally less dramatic (26). A structural study of DNA-full MVMp capsids has been initiated to enable a more complete interpretation of the observed MVMi/p fivefold differences.
Parvovirus capsids are generally stable over wide temperature (25 to 70°C) and pH (pH 3.0 to 9.0) ranges (19, 27). A comparative study of the stability of the MVM capsids was prompted by the observation that intra- and intersubunit interactions, involving clustered MVMi/p differing residues, differ between the two viruses, with MVMi predicted to be more stable. The stabilities of empty MVMi, MVMpe, and MVMpb capsids were measured based on their abilities to hemagglutinate mouse erythrocytes, a simple but reliable assay for MVM capsid disassembly, as supported by other techniques including tryptophan fluorescence and differential scanning calorimetry (16). Capsids were incubated at three different pHs (pH 6.0, 7.5, and 8.0) over a temperature range of 25 to 85°C (Fig. (Fig.4)4) and assayed. As previously described (27), both MVMp capsids were equally stable during the assay, with complete disassembly occurring only at high temperatures, although the MVMpb capsids consistently showed slightly lower HA activity. The MVMi capsid, however, was clearly more stable than both types of MVMp capsids at any pH (Fig. 4A to C). In general, MVM capsids were less stable at increasing pHs, and the differences between the strains were more pronounced at pH 8.8 (Fig. (Fig.4C).4C). At this pH and temperatures of ≥70°C, the stability of MVMp capsids was dramatically lower than that of the MVMi capsid. Both MVMp capsids were completely disassembled by 75°C at this pH (Fig. (Fig.4C),4C), while the MVMi capsid still retained approximately 50% of the activity observed at the lower pHs at 80°C. Complete disassembly of the MVMi capsid at pH 8.8 occurred at 80°C, rather than the 85°C observed at pHs 6.0 and 7.5.
The calculated pI (25) for both the MVMi and MVMp VP2 amino acid sequences is ~5.8, and the overall charge of both capsids should be the same at the pHs tested. The pKa values of the terminal amide groups of the lysine and arginine MVMp/i difference at amino acid position 368 are ~10 and ~12, respectively. These amide groups will be positively charged at pH 6.0 to 8.8. K368 of MVMp is too far from D399 to engage in a salt bridge interaction (Fig. (Fig.3E).3E). Thus, at the higher pH, 8.8, the interactions of R368 with the adjacent D399 and E321 (Fig. (Fig.3F)3F) will result in a more stable capsid for MVMi than for MVMp, which contains a glycine at amino acid position 321 and a lysine at position 368. Therefore, the differential capsid stability observed between MVMi and MVMp could be due to the difference in local surface interactions across the twofold axes (Fig. 3C to F).
The MVM VP2 structure topology is very similar to those available for the autonomous parvoviruses CPV, FPV, and PPV (Fig. 5A and B). The structures are superimposable (Fig. (Fig.5B),5B), with an overall RMSD of 0.4 to 0.6 Å between structurally equivalent VP2 Cα atoms. These values are within the range calculated for the superimposition of MVMpb and MVMi VP2 (0.48 Å), wt CPV and wt FPV, or wt CPV and CPV mutant VP2 structures (26). However, there are surface loop regions (labeled 1 to 8 in Fig. 5A and B, Fig. Fig.6A,6A, and Table Table3)3) that show Cα differences of as much as ~5 Å (Fig. (Fig.5A).5A). These regions also differ structurally between highly homologous parvovirus strains, such as wt CPV and wt FPV, wt CPV and its host range mutants, and also MVMi and MVMp (2, 26, 53, 61). These local loop differences cluster, from icosahedral-symmetry-related monomers, to create local variations on the characteristic parvovirus capsid features at the fivefold axes, at the shoulder of the threefold axes, and at and surrounding the twofold axes (Fig. 3B and D and and6A).6A). An analogous comparison of dependovirus VP3 structures and 3-dimensional models also identified similar variable regions on their capsids that are clustered from symmetry-related monomers (44). The dependovirus differences are postulated to control receptor recognition and antigenic phenotypes. The suggested functional roles of the variable autonomous virus capsid regions are discussed below.
The role of variable region 7, a loop located between βH and βI, is currently unknown. Variable region 1 contains MVM residue 160 and, as discussed above, structural rearrangements of this capsid region are likely involved in the externalization of VP1 and VP2 (Table (Table3),3), for phospholipase A2 function and VP2-to-VP3 cleavage, respectively. Variable regions 2 and 3 contain CPV and FPV residues that form part of two major antigenic epitopes (57). Residues in the four remaining variable regions (variable regions 4, 5, 6, and 8), plus region 3 (Fig. (Fig.55 and and6A),6A), have been mapped previously as being important for parvovirus tropism, pathogenicity determination, hemagglutination behavior, or receptor attachment (Table (Table3).3). The amino acid residues involved in controlling these functions for AMDV, CPV, FPV, MVM, and PPV are mapped onto homologous capsid regions of the MVMpb structure in Fig. Fig.6B6B.
The clustering of MVMi/p allotropic residues (A317T, E321G) (region 4 in Fig. Fig.5B5B and and6A;6A; red balls in Fig. Fig.6B)6B) and of residues involved in MVMi forward fibrotropic mutations (residues 399, 460, 553, and 558) (regions 5 and 8 in Fig. Fig.5B5B and and6A;6A; blue balls in Fig. Fig.6B)6B) suggests a vital role for this capsid region in MVMi/p in cellular recognition. CPV/FPV residue 323, which plays a role in tissue tropism and pathogenicity and is postulated to form part of the receptor attachment (26, 30), is structurally close to MVM's allotropic residue 321 (Fig. (Fig.6B),6B), suggesting possible utilization of common capsid regions for host cell recognition. For AMDV, VP2 residues 352, 395, 434, and 534 (orange balls in Fig. Fig.6B),6B), implicated in tropism and pathogenicity determinations (39), are mapped to variable regions 4 and 5 in Fig. Fig.5B5B and and6A.6A. The pathogenicity determinants for PPV involve two VP2 amino acids, residues 377 and 388 (55) (pink balls in Fig. Fig.6B),6B), which map close to region 6 in Fig. Fig.5B5B and and6A6A on the wall between the twofold and fivefold axes. A third PPV residue, residue 434 (close to the icosahedral threefold axes [Fig. [Fig.6B]),6B]), previously implicated as a pathogenicity determinant, has been shown to be less important (P. Tijssen, personal communication). The role of receptor recognition in AMDV and PPV tropism and pathogenicity determination requires further investigation, but surface clustering of the determinant residues in analogous structural positions in the CPV/FPV and MVMi/p model systems suggests a commonality in functionality.
Although the primary cell surface receptor utilized for infection by MVM is not known, treatment of susceptible cells with neuraminidase, which removes terminal sialic acids, abolishes MVM reinfection, identifying this carbohydrate as a component of the cell surface receptor (20; A. López-Bueno, M. P. Rubio, N, Bryant, R. McKenna, M. Agbandje-McKenna, and J. M. Almendral, unpublished data). Interestingly, CPV/FPV HA, which involves sialic acid recognition, is dictated by CPV residues R377, E396, and R397 (60) on the wall of the twofold dimple (regions 4 and 6 in Fig. Fig.6A;6A; black balls in Fig. Fig.6B),6B), with R377 being structurally proximate to MVM residues 317 and 321 (Fig. (Fig.6B).6B). The sialic acid binding site for the MVMp capsid, leading to infection, was recently mapped to the icosahedral twofold axes, highlighting the role that receptor recognition, involving differing residues, plays in MVM pathogenesis (López-Bueno et al., unpublished data). In addition, the correlation of slight structural alterations within or close to the icosahedral twofold depression, due to disruption of intra- and intersubunit amino acid interactions (Fig. 3E and F), with mutations that change MVM tropism in vitro likely illustrates the nature of the tight regulations required for successful virus-cell interaction during MVM infection. It is also likely that MVM HA involves the clustered MVMi/p differing residues at the icosahedral twofold axes. Thus, the alteration of the contacts/interactions of the capsid amino acids with sialic acid could lead to the differential MVMi/p capsid HA activity/stability observed (Fig. (Fig.44).
In conclusion, the available high-resolution structures of CPV, FPV, and PPV and those reported here for MVM, in addition to a pseudo-atomic model available for AMDV-G, provide a means to attempt the correlation of the vast amount of biology data available on parvovirus tropism and pathogenicity to the capsid structure. This analysis indicates that tropism and pathogenicity determination for highly homologous parvovirus strains is likely a precisely regulated phenotype dictated by minor changes on the capsid surface, mostly located close to or surrounding the icosahedral twofold axes (Fig. 6A and B). The mapping of a binding site for sialic acid, a component of the productive receptor, in this region of the MVMp capsid (López-Bueno et al., unpublished results) and the colocalization of CPV/FPV tropism and pathogenicity determinants with a proposed receptor attachment footprint (26) warrant reexamination of the roles that receptors and intracellular factors (29, 56) play in parvovirus host range determination.
We are grateful to Jean Rommelaere (DKFZ, Heidelberg, Germany) and Peter Tattersall (Yale University, New Haven, CT) for providing the MVM strains. We also thank the staff at DSRS (United Kingdom) and DESY (Germany) for help during synchrotron data collection. The experimental support of Beatriz Maroto and Noelia Valle in the hemagglutination assays is also acknowledged.
This study was supported by the National Science Foundation (MCB 0212846, to M.A.-M.), CSIC-British Council (Integrated Action, HB1998-0173 to C.F.-F. and 1999/2000-8125 to M.A.-M.), and Comunidad Autónoma de Madrid (ref. 07B/0020/2002, to J.M.A.).