Structure of an Enteric Pathogen, Bovine Parvovirus

ABSTRACT

Bovine parvovirus (BPV), the causative agent of respiratory and gastrointestinal disease in cows, is the type member of the Bocaparvovirus genus of the Parvoviridae family. Toward efforts to obtain a template for the development of vaccines and small-molecule inhibitors for this pathogen, the structure of the BPV capsid, assembled from the major capsid viral protein 2 (VP2), was determined using X-ray crystallography as well as cryo-electron microscopy and three-dimensional image reconstruction (cryo-reconstruction) to 3.2- and 8.8-Å resolutions, respectively. The VP2 region ordered in the crystal structure, from residues 39 to 536, conserves the parvoviral eight-stranded jellyroll motif and an αA helix. The BPV capsid displays common parvovirus features: a channel at and depressions surrounding the 5-fold axes and protrusions surrounding the 3-fold axes. However, rather than a depression centered at the 2-fold axes, a raised surface loop divides this feature in BPV. Additional observed density in the capsid interior in the cryo-reconstructed map, compared to the crystal structure, is interpreted as 10 additional N-terminal residues, residues 29 to 38, that radially extend the channel under the 5-fold axis, as observed for human bocavirus 1 (HBoV1). Surface loops of various lengths and conformations extend from the core jellyroll motif of VP2. These loops confer the unique surface topology of the BPV capsid, making it strikingly different from HBoV1 as well as the type members of other Parvovirinae genera for which structures have been determined. For the type members, regions structurally analogous to those decorating the BPV capsid surface serve as determinants of receptor recognition, tissue and host tropism, pathogenicity, and antigenicity.

IMPORTANCE Bovine parvovirus (BPV), identified in the 1960s in diarrheic calves, is the type member of the Bocaparvovirus genus of the nonenveloped, single-stranded DNA (ssDNA) Parvoviridae family. The recent isolation of human bocaparvoviruses from children with severe respiratory and gastrointestinal infections has generated interest in understanding the life cycle and pathogenesis of these emerging viruses. We have determined the high-resolution structure of the BPV capsid assembled from its predominant capsid protein VP2, known to be involved in a myriad of functions during host cell entry, pathogenesis, and antigenicity for other members of the Parvovirinae. Our results show the conservation of the core secondary structural elements and the location of the N-terminal residues for the known bocaparvovirus capsid structures. However, surface loops with high variability in sequence and conformation give BPV a unique capsid surface topology. Similar analogous regions in other Parvovirinae type members are important as determinants of receptor recognition, tissue and host tropism, pathogenicity, and antigenicity.

INTRODUCTION

Bovine parvovirus (BPV), the type member of the Bocaparvovirus genus of the Parvoviridae family, was discovered in 1961 in the gastrointestinal tract of diarrheal calves (1). In addition to acute gastroenteritis, it is associated with reproductive disorders such as spontaneous abortions and stillbirths (2). The rate of occurrence of BPV infections in herds is high, at 83 to 100% worldwide (2–5). Other known genus members include the human bocaviruses (6–8), canine minute virus (9), porcine bocavirus (10), gorilla bocavirus (11), California sea lion bocavirus (12), feline bocavirus, and canine bocavirus (13), which all have similar disease phenotypes. Currently, no treatment or preventive measures are available for Bocaparvovirus infections. The majority of the data available for these viruses is epidemiological. Hence, there is a need to study these viruses at the cellular, molecular, and structural levels.

BPV is an autonomously replicating virus with a linear single-stranded DNA (ssDNA) genome of ∼5.5 kb flanked by nonidentical palindromic terminal hairpins, similar to other bocaparvoviruses (14). The genome consists of three open reading frames (ORF1 to ORF3). ORF1 encodes a nonstructural protein, NS1, important for DNA replication. ORF2 encodes a nuclear phosphoprotein, NP1, unique to the Bocaparvovirus genus, containing nuclear localization signals (NLSs), and plays a role in viral RNA processing during gene expression (15, 16). Lastly, ORF3 encodes two capsid viral proteins (VP1 and VP2), which are generated as a result of alternative splicing events (73). VP1 (75 kDa) and VP2 (61 kDa) share a C-terminal end, but VP1 contains an additional N-terminal region (VP1u). VP1u has a phospholipase A₂ (PLA₂) motif and an NLS essential for infectivity by facilitating the release of the virus from the endocytic pathway during trafficking and entry into the nucleus for initiation of viral replication, respectively. A total of 60 copies of VP1 and VP2 assemble a T=1 icosahedral capsid. However, 60 copies of Bocaparvovirus VP2 alone are also able to assemble a capsid (17).

The parvovirus capsid is involved in a myriad of functions, including cell entry, endosomal trafficking to the nucleus, and cell egress (18, 19). These capsids have also been reported to be subjected to the selective pressures of the environment, host cells, and the host immune system (20). Several structural “hot spots” have been identified on the capsid surface, especially near the icosahedral axes of symmetry, which are important for different functions, including receptor attachment, antigenicity, and pathogenicity (reviewed in reference 21). For the bocaparvoviruses, while there is some information on requirements for cellular infection, nothing is known about the capsid determinants of these interactions. BPV, in particular, has been shown to bind to N- and O-linked sialic acid moieties, which may serve as primary receptors for cell entry (22, 23). Cell-based studies with human bocavirus 1 (HBoV1) virions predict that primary receptors and coreceptors on the apical membrane of human airway epithelial cells influence efficient virion infection and cellular transduction (24). Information on Bocaparvovirus capsid structure will provide a platform to begin the elucidation of sites on the capsid that serve as important determinants of host range, tissue tropism, pathogenicity, and bocaparvovirus-related disease emergence.

Here, we report the structure of BPV determined by X-ray crystallography as well as cryo-electron microscopy and three-dimensional (3D) image reconstruction (cryo-reconstruction) to 3.2- and 8.8-Å resolutions, respectively. The capsid conserves parvovirus features: a channel at and depressions surrounding the 5-fold axes and protrusions surrounding the 3-fold axes. Uniquely, the 2-fold axes have a flat surface flanked by a depression. A comparison to the previously reported HBoV1 structure, determined to a 7.9-Å resolution by cryo-reconstruction (17), revealed conserved density extending the 5-fold channel radially inwards in both viruses, while differences were located on the capsid surface. Surface loop differences, clustered at or close to 2-, 3-, and 5-fold axes in addition to the capsid surface region between the 2- and 5-fold axes (the 2/5-fold wall), gave each Bocaparvovirus capsid a distinctive surface topology. These regions may play a role in governing host- and tissue-specific interactions. Comparison to known crystal structures of other type members of the Parvovirinae subfamily (adeno-associated virus 2 [AAV2], the prototype strain of minute virus of mice [MVMp], and human parvovirus B19) or a pseudoatomic model built into a cryo-reconstructed density (Aleutian mink disease virus [AMDV]) identified similarities and differences in VP2 and the capsid. While maintaining the VP2 secondary structural elements that are likely important for capsid assembly, unique surface features are conferred on BPV due to variations in VP2 surface loop lengths and conformations. These surface variations occur in analogous regions known to control infectious functions, including host range, receptor attachment, antigenicity, and pathogenicity, for other Parvovirinae type members. This structure serves as a template to begin structure-function annotations of bocaparvoviruses.

MATERIALS AND METHODS

VP2 virus-like particle expression and purification.

The VP2 gene of BPV (HADEN strain; GenBank accession number ABC69731.1) was cloned into a pFastBac vector to produce a recombinant baculovirus using standard Bac-to-Bac technology (Invitrogen). Spodoptera frugiperda (Sf9) cells, maintained in Grace's medium with 10% fetal calf serum and antibiotics, were infected with the recombinant baculovirus and harvested at 4 to 7 days postinfection. The purification of BPV virus-like particles (VLPs) was carried out using sucrose cushion and sucrose density gradients, as previously described for HBoV1 (17). The final sample was isolated from the 25% fraction of the sucrose gradient and concentrated to 10 mg/ml using Amicon concentrators (EMD Millipore). The purity and integrity of the VLPs were determined using SDS-PAGE (data not shown) and negative-stain electron microscopy (EM), respectively. A polyclonal primary antibody to the BPV capsid generated in guinea pigs (generously provided by Brent F. Johnson, BYU) and a secondary anti-guinea pig horseradish peroxidase-conjugated protein A antibody (Invitrogen) were used for denatured Western blots and native dot blots to confirm the presence of protein and VLPs, respectively (data not shown).

X-ray diffraction data collection and processing.

The purified BPV VLP sample was used to screen for suitable crystallization conditions using the hanging-drop vapor diffusion method (25). Each drop contained 2 μl of VLPs and 2 μl of the reservoir solution incubated over 1 ml of reservoir solution. Crystallization screens included various polyethylene glycol 8000 (PEG 8000) (0.5 to 4.0%), NaCl (75 to 500 mM), MgCl₂ (0 to 8 mM), and Li₂SO₄ (0 to 50 mM) concentrations and pHs ranging from 6.0 to 8.5 (Bis-Tris and Tris-HCl). The crystals used for data collection were flashed-cooled in the precipitant solution with 10% PEG 8000 and 30% glycerol added as cryoprotectants.

Diffraction images were collected at the Advanced Photon Source (Argonne National Laboratory) using the SER-CAT 22-ID-D beamline on a MAR300 charge-coupled-device (CCD) detector with a λ value of 0.9724 Å. The crystal-to-detector distance, oscillation width, and exposure time were 400 mm, 0.3°, and 4 s, respectively. The measured reflections, with a maximum resolution of 3.2 Å, were indexed, integrated, and scaled using the HKL2000 suite of programs (26). The space group was assigned as C222₁ with the following unit cell parameters: a = 323.4 Å, b = 381.2 Å, and c = 376.6 Å. The Matthews coefficient was calculated using the MATTHEWS_COEF subroutine in the CCP4 suite of programs to be 3.23 Å/Da³, with a corresponding solvent content of 61.9%, assuming a VLP molar mass of ∼3.66 × 10⁶ Da (27, 28). A self-rotation function was calculated using the GLRF program, with 10% of the data between 10- and 5-Å resolution, a 120-Å radius of integration, and κ values of 72°, 120°, and 180° to search for the 5-fold, 3-fold, and 2-fold noncrystallographic symmetry (NCS) elements, respectively, of the icosahedral virus capsid (29). The rotation function was consistent with four VLPs in the unit cell with the NCS 2-fold and crystallographic 2-fold axes being coincident, resulting in 30 VP2 monomers per asymmetric unit.

Molecular replacement and structure refinement.

Molecular replacement methods were used to obtain the initial phases, using the crystal structure of B19 VP2 (PDB accession number 1S58) as the phasing model to 7-Å resolution using the AutoMR subroutine within the PHENIX program (30–32). A polyalanine model of B19 VP2 was generated with the MAPMAN program, and an oligomer of 30 VP2 monomers was built in VIPERdb2 (calculated by icosahedral matrix multiplication) (33, 34). Phases were extended to the final high resolution of 3.2 Å using the density modification subroutine in the CNS program suite, which includes solvent flattening and NCS averaging (35). Subsequent model coordinate refinement was conducted with the CNS program using simulated annealing, energy minimization, position, and individual temperature factor refinement (B_factor) while applying 30-fold NCS. The refinement process was monitored using 5% of the total data set for the calculation of R_free (defined in Table 1). This step was followed by real-space electron density averaging using a VP2 molecular mask generated with the CNS program. The electron density was interpretable for residues 39 to 536 (C-terminal end according to VP2 numbering) of BPV VP2 in the averaged sigma-weighted 2F_o-F_c electron density map (F_o and F_c are defined in Table 1). The VP2 model was built into this density by substitution of the alanine residues, in the initial B19 model, with specific BPV residues using the COOT program (36). This was followed by alternating cycles of refinement and model building guided by the calculated averaged density map until no further improvement was observed in the model, as monitored by the R_factor (defined in Table 1). Density below a threshold value of 0.5 σ extending from residue 39 below the icosahedral 5-fold axis was not modeled because the backbone direction could not be unambiguously determined. The quality of the BPV VP2 model was assessed with the COOT and MOLPROBITY programs (36, 37). The CNS program was used to calculate the root mean square deviations (RMSDs) from ideal bond lengths and angles, while the average B_factor values for the VP2 model were determined by the MOLEMAN program (35, 38). Data collection, processing, and refinement statistics are summarized in Table 1. Figures were generated with the UCSF-Chimera and PyMOL programs (39, 40).

TABLE 1.

Crystallography data collection, processing, and refinement statistics

^aValues in parentheses are for the highest-resolution shell.

^bR_sym = (Σ|I − 〈I〉|/ΣI) × 100, where I is the intensity of an individual reflection with indices h, k, and l and 〈I〉 is the average intensity of all symmetry equivalent measurements of that reflection; the summation is over all intensities.

^cR_factor = (Σ|F_o| − |F_c|/Σ|F_o|) × 100, where F_o and F_c are the observed and calculated structure factor amplitudes, respectively.

^dR_free is calculated similarly to R_factor, except that it uses 5% of the reflection data partitioned from the refinement process.

Cryo-electron microscopy and image reconstruction.

Three-microliter aliquots of purified BPV VLPs (∼0.2 mg/ml) were applied to C-flat holey carbon grids (Protochips, Inc.) and vitrified using a Vitrobot Mark IV instrument (FEI Co.). The sample was examined using a 16-megapixel CCD camera (Gatan, Inc.) in a Tecnai (FEI Co.) G2 F20-Twin transmission electron microscope operated at a voltage of 200 kV under low-dose conditions (∼20 e/Ų). Images were recorded with a defocus range of ∼1.2 to 4 μm at a magnification of ∼×69,000, resulting in a pixel size of 2.24 Å. The RobEM subroutine within the AUTO3DEM software suite (41) was used to extract 1,414 particles from 63 cryomicrographs. Estimations for the defocus values of the micrographs were made using the ctffind3 subroutine in AUTO3DEM, and corrections for the microscope-related contrast transfer functions were applied during the search and refine modes in AUTO3DEM (42). An initial low resolution (30-Å) cryo-reconstructed model map was generated using an ab initio random-model method and imposing icosahedral symmetry with 150 particle images (43). This map was used for the initial determination of the orientations and origins followed by refinement of these parameters for the entire set of particle images using AUTO3DEM (41). The final 3D map was reconstructed from 1,131 particle images to an estimated resolution of ∼8.8 Å based on a Fourier shell correlation (FSC) threshold criterion of 0.5. An inverse temperature factor of 1/100 Ų was used to improve the high-resolution features at 8.8-Å resolution. To avoid amplification of noise in addition to signal, the structure factors were multiplied by a noise suppression factor postrefinement in AUTO3DEM (44).

Difference map calculation.

A 60-mer (generated in VIPERdb2 [33]) from the BPV VP2 crystal structure was docked into the cryo-reconstructed density map using the UCSF-Chimera program with standard cross-correlation as a criterion to measure the fit (39). The docked crystal structure coordinates were back-filtered and used to calculate a map to the resolution of the reconstructed map using the “pdb2mrc” subroutine in the EMAN2 program (45) for a difference map calculation. The two maps (model and cryo-reconstructed) were normalized using the MAPMAN program prior to use (46). A difference map was calculated by subtracting the crystal structure map from the cryo-reconstructed map using the “vop” (volume of operation) subroutine in the UCSF-Chimera program (39). Positive density observed below the 5-fold axis was modeled as 10 additional N-terminal residues, residues 29 to 38 (VP2 numbering).

Revision of the pseudoatomic model built into the 7.9-Å cryo-reconstructed HBoV1 structure.

To improve the original pseudoatomic model built for HBoV1, the BPV crystal structure coordinates were used as the template to build a new VP2 homology model with the UCSF-Chimera program (39). This exercise was carried out because the original HBoV1 pseudoatomic model utilized B19, with ∼23% sequence identity to HBoV1, while BPV shares ∼46% identity. In lieu of an HBoV1 crystal structure, the revised model served as a better representative of the structure. Docking of a 60-mer of the revised HBoV1 homology model into the previously reported HBoV1 cryo-reconstructed map was carried out using the “fit in map” subroutine of the UCSF-Chimera program (39). Further interactive model adjustments to improve model fitting guided by the density map envelope, followed by real-space refinement, was carried out using the COOT program (36). Readjustments were made mainly in the following surface variable regions (VRs) (defined in Table 3): VR-II (DE loop), VR-III, VR-IV, VR-V, and VR-VIIIB (HI loop).

TABLE 3.

Sequence and structure comparison of members of the Parvovirinae subfamily

Virus	Total no. of VP2 residues	% sequence identity to BPV determined using:		RMSD from BPV-1 (Å)c
		Full-length VP2 sequencesa	Residues ordered in the crystal structure or pseudoatomic modelb
B19	523	23.0	23.1	1.9
AAV2	598	26.0	29.0	1.6
MVMp	587	22.0	18.5	3.2
AMDV	638	22.3	18.0	2.2

^aPercent sequence identity to BPV determined using full-length VP2 sequences.

^bPercent sequence identity determined using only residues ordered in the crystal structures (AAV2, B19, and MVMp) or the pseudoatomic model (ADMV) compared to the BPV VP2 crystal structure.

^cThe crystal structures were superposed based on secondary structures, and root mean square deviation (RMSD) values were calculated for the main-chain α-carbons between the known crystal structures or pseudoatomic models and the BPV crystal structure.

Sequence and structural comparison to other members of the Parvovirinae with available 3D structures.

BPV, AMDV, AAV2, MVMp, and human parvovirus B19 VP2 sequences (GenBank accession numbers ABC69731.1, AAA96023.1, YP_680427.1, CAB46508.1, and AAV35058.1, respectively) were aligned using the CLUSTALW-2 online server (47). The C_α atoms of the VP2 atomic coordinates for MVMp, B19, and AAV2 (PDB accession numbers 1Z14, 1S58, and 1LP3, respectively) and pseudoatomic coordinates modeled into the cryo-reconstructed density maps of AMDV (48) and HBoV1 (EMDB accession number 1739) were superposed onto the BPV atomic coordinates using the secondary structure matching (SSM) tool found in the PDBefold online server (49). The server generates RMSDs for the superposed structures, and the distances between the aligned C_α positions were used to identify VRs. The definition of VRs for the autonomously replicating viruses, regions with two or more consecutive residues with an RMSD of >2.0 Å between the superposed structures (50), was used. Cartoon and surface representations were generated with the PyMOL and UCSF-Chimera programs, respectively (39, 40).

Accession numbers.

The coordinates and structure factors for the crystal structure of BPV VP2 have been deposited in the PDB database (http://www.rcsb.org/) under accession number 4QC8. The cryo-reconstructed map was deposited in the EMDB database under accession number EMD-6168.

RESULTS AND DISCUSSION

Crystal structure of BPV VP2.

Purified and homogeneous BPV VLPs, which formed two-dimensional crystalline arrays when viewed by EM, produced flat crystals in a solution containing 10 mM Tris-HCl (pH 7.5), 4% PEG 8000, and 8 mM MgCl₂ with or without 25 to 50 mM Li₂SO₄ (Fig. 1A and B). The crystals diffracted X rays 3.2-Å resolution (Table 1). Residues 39 to 536 of VP2 were ordered in the structure determined using molecular replacement and assigned to the averaged density map (Fig. 2A and B). Atomic detail comparison of the BPV map and the refined model with the coordinates of the B19 phasing model showed no phase bias at both the residue level and the secondary structure level, wherein the density was consistent with the BPV sequence (Fig. 2A and B). The first 38 residues at the N terminus are not present in the refined model due to the lack of interpretable density. This lack of density is consistent with previously reported observations for most other parvoviruses, except B19 (21, 52). This could be a result of multiple conformations of the VP2 N terminus or inherent disorder in this region, which is incompatible with the 30-fold NCS applied during model refinement and density averaging. The refined parameters for the crystal structure are provided in Table 1 and are within the range reported previously for other virus structures (including parvoviruses) determined to similar resolutions. The similarity of R_factor and R_free values for the virus structures stems from the noncrystallographic icosahedral symmetry of the capsid that does not permit random selection of unique reflections required for the R_free calculation.

FIG 1.

Purification, crystallization, and freezing of BPV VLPs. (A) Micrograph of negatively stained BPV VLPs imaged at a magnification of ×97,000. (B) Flat crystals of BPV (indicated by white arrows) obtained using a solution containing 10 mM Tris (pH 7.5), 4% PEG 8000, 8 mM MgCl₂, and 25 mM LiSO₄. (C) Cryomicrograph of unstained BPV VLPs.

FIG 2.

Crystal structure of BPV. (A and B) Sections of the 2F_o-F_c electron density map, contoured at a threshold of 1.5 σ (gray mesh), highlighting the unbiased BPV structure at the residue (A) and polypeptide chain (B) levels. The stick and cartoon representations show the superposed coordinates of BPV (magenta) and B19 (orange). (C) Cartoon representation of the BPV VP2 crystal structure (magenta) with the following secondary structure elements highlighted: β-strands (red), loops (magenta), and helices (cyan). The conserved core consisting of the β-strands (βA-BIDG-CHEF) and α-helix (αA) is labeled accordingly. The approximate positions of the icosahedral 2-, 3-, and 5-fold symmetry axes are indicated as a filled oval, triangle, and pentagon, respectively. The N and C termini of the polypeptide chain are also labeled. (D) Radially depth-cued surface representation of the BPV capsid shown approximately along the 2-fold axis of symmetry. The approximate positions of the 2-, 3-, and 5-fold axes as well as the 2/5-fold wall are indicated. A color key corresponding to the depth cue radii (in Å) is shown. Panels A to C were generated with PyMOL, and panel D was generated with UCSF-Chimera (39, 40).

The ordered BPV VP2 structure (residues 39 to 536) displays the conserved features of the parvovirus VP topology. It maintains the core eight-stranded antiparallel jellyroll motif, organized in two β-sheets, BIDG and CHEF, seen commonly in most virus structures (Fig. 2C and 3). This β-barrel motif constitutes 17% of the VP2 sequence and forms the contiguous capsid shell. Additional antiparallel β-strand stretches are present in the loops between the core β-strands, as reported previously for some of the other parvoviruses (21) (Fig. 2C and 3). A small α-helix (αA) spanning residues 109 to 121 (VP2 numbering), also conserved in all other Parvovirinae structures solved thus far, was observed in BPV. BPV VP2 uniquely contains an additional 2-turn α-helix at residues 211 to 218 not reported for other parvovirus structures (Fig. 2C and 3). Loops of various lengths and conformations, named for their flanking β-strands, for example, the DE loop for the loop between βD and βE, were inserted between the β-strands. These loop regions, often clustered at the capsid surface as a consequence of the icosahedral symmetry, contribute to the BPV capsid surface topology (Fig. 2D).

FIG 3.

Structural alignment of BPV and HBoV1. Secondary structure elements, β-strands and α-helices, are indicated with arrows (blue) and cylinders (red), respectively. The conserved core elements (βA-BIDG-CHEF and αA) are labeled. Insertions, deletions, or VRs (as defined in Materials and Methods) are shown in red in the HBoV1 sequence and labeled. The first N-terminal residues modeled into the BPV reconstructed and crystal structures are indicated with black arrows. Symbols below the alignment indicate standard ClustalW2 nomenclature, where * indicates identity, : indicates high-level conservation, and · indicates low-level conservation.

The characteristic dimple-like depression centered directly at the 2-fold axis of symmetry, common in other parvoviruses (21), is divided by a raised flat surface in BPV (Fig. 2D). Residues 502 to 506 give rise to this raised flat surface appearance, while residues 109 to 121 (of the conserved αA helix) and residues 499 to 519 (of the loop between βI and the C terminus) form the floor/wall of the 2-fold depression. A depression is also centered at the 3-fold axis, surrounded by three distinct protrusions composed of three large loops within the GH loop (comprising 199 residues), contributed by two 3-fold symmetry-related monomers, residues 246 to 297 and residues 385 to 405 from one VP2 protein and residues 304 to 320 from the other (Fig. 2D). A second outer ring of slightly smaller protrusions, formed by residues 81 to 90, 200 to 220, and 321 to 358 of the same VP2 monomer, is located on the 2/5-fold wall surrounding the 3-fold protrusions (Fig. 2D). The DE loop, comprised of 22 residues, clusters to form a cylindrical channel at the 5-fold axis (Fig. 2C and D). This feature is maintained in all other known Parvovirinae structures (21). Surrounding this 5-fold channel is a shallow depression lined by HI loops (residues 455 to 469) from 5-fold-neighboring symmetry-related monomers (Fig. 2C and D). Overall, the BPV capsid diameter is ∼220 Å at the 2-fold axis, ∼275 Å at the peak of the 3-fold protrusions, and ∼240 Å at the 5-fold axis (Fig. 2D).

The density extending the 5-fold channel is interpreted as the N terminus of VP2.

As mentioned above, N-terminal residues 1 to 38 of the BPV VP2 structure determined by X-ray crystallography were disordered (Fig. 2C). This observation is similar to observations made for all structures determined to date for other parvoviruses by X-ray crystallography. This has been postulated to be due to either a low copy number, flexibility in conformation, or intrinsic disorder in this VP region. These properties are incompatible with the NCS applied during structure determination and refinement and the finer spatial frequency of data sampling as resolution is increased in crystallographic structure determination. Cryo-reconstruction, at low or median resolution, is thus often a more applicable method for overcoming this limitation. Consistently, this method was used to visualize the N terminus of VP2 in B19 (52). Thus, VLPs extracted from cryomicrographs were used to generate an 8.8-Å-resolution cryo-reconstructed map for BPV (Fig. 1C and 4A). The map is similar in size and surface features to the crystal structure (Fig. 2D and 4A). The conserved secondary structural elements of the BPV VP2 crystal structure, βA, the βBIDG-CHEF β-barrel, and the αA helix (Fig. 2C), were readily fitted into the cryo-reconstructed density map (4B and C). In addition, although the apexes of most of the surface loops displayed high flexibility and correspondingly high thermal temperature factor values in the VP2 crystal structure, some of these loops, for example, the DE (Fig. 4D) and HI (data not shown) loops, were precisely located in the cryo-reconstructed map. The correlation coefficient for the fit between the cryo-reconstructed map and a model map generated from a 60-mer of the BPV VP2 crystal structure (residues 39 to 536) was 0.91.

FIG 4.

Cryo-EM structure of BPV. (A) Surface representation of the cryo-reconstructed density map, viewed along the 2-fold axis, radially depth cued as per the color key (in units of Å). An equilateral triangle depicting the viral asymmetric unit, with the 3-fold axes at the base vertices separated by the 2-fold axis and the 5-fold axis at its vertex (labeled 3f, 2f, and 5f, respectively), is shown. (B) Zoom-in view of the conserved α-helix (αA) from 2-fold-related monomers showing the fit into the reconstructed density (gray mesh). (C) Cross-sectional view of the 60-mer of the BPV crystal structure docked into the cryo-reconstructed density map (shown as gray mesh and contoured at 1 σ). The dashed lines indicate the icosahedral symmetry axes. (D) Side view of a cross section of the map and fitted crystal structure (shown in rainbow colors), highlighting the extended 5-fold channel. The first N-terminal residues observed in the crystal structure and in the pseudoatomic BPV model built into the cryo-reconstructed map are indicated. Approximate dimensions for the width and height of the channel are also shown. This figure was generated using UCSF-Chimera (39).

Densities, not satisfied by the fitted VP2 crystal structure (residues 39 to 536), were ordered inside the capsid of the cryo-reconstructed map, at the base of each 5-fold vertex, radially extending the channel inwards (Fig. 4C). A difference map (crystal structure subtracted from the cryo-reconstructed map) showed that the largest positive difference was consistent with these densities, which were immediately adjacent to the first residue, residue 39, in the VP2 crystal structure model (Fig. 4D). For each VP2, this difference density was modeled as a stretch of 10 additional N-terminal residues, 29-SVGGGGRGGS-38 (Fig. 4D). The model extends the channel to a total length of ∼60 Å, with a capsid exterior width of ∼14 Å and capsid interior widths of ∼10 Å and ∼18 Å at the neck and innermost edge of the channel, respectively (Fig. 4D). Any additional density radially extending inwards from the innermost end of this channel beyond residue 29 was not modeled due to ambiguity and a low signal-to-noise ratio. The correlation coefficient for the new 60-mer capsid model (with each monomer comprising residues 29 to 536) fitted into the cryo-reconstructed map increased from 0.91 to 0.93. This is consistent with satisfaction of the previously unoccupied interior 5-fold density. Similar VP1/2 N-terminal localizations under the 5-fold axis have been reported for other autonomous parvovirus members, for example, MVMp and canine parvovirus (53, 54).

Bocaparvovirus capsids conserve the extended 5-fold channel but show variability in capsid surface features.

The cryo-reconstructed BPV capsid structure was compared to a median-resolution, 7.9-Å structure of the HBoV1 VP2 capsid, also determined using cryo-reconstruction (17) (Fig. 5A and B). Similar to the BPV cryo-reconstructed map, ordered density was observed below the 5-fold axes in the HBoV1 cryo-reconstructed map (Fig. 5B), and 9 residues were built into this density (17). However, the interior end of the channel in HBoV1 appears to be morphologically different from that in BPV (Fig. 5B). Due to the conservation of this density feature in both Bocaparvovirus capsids, the current hypothesis is that the N termini of their VP2s are bundled under the 5-fold axis and extend the 5-fold channel into the capsid interior. Given the overlap between VP1 and VP2, the anticipation is that the VP1 N termini are similarly localized. VP1u has been postulated to become externalized through this channel for other parvoviruses (55–59, 71). How VP1u is extruded through the channel for its PLA₂ function during infection requires further studies for members of this genus.

FIG 5.

Comparison of Bocaparvovirus cryo-EM structures. (A and B) Surface (A) and cross-sectional (B) views of BPV and previously reported HBoV1 (17) radially depth cued by the color key (in units of Å). (C) Superposition of the C_α of the crystal structure of BPV (magenta) and the revised pseudoatomic model of HBoV1 (red), depicted as smooth loops. The approximate positions of the icosahedral 2-, 3-, and 5-fold symmetry axes are indicated as a filled oval, triangle, and pentagon, respectively. The following VRs are labeled on the VP2 monomer (left) and colored on the surface of the BPV capsid (right) (in gray) in different colors: VR-II (blue), VR-III (yellow), VR-IV (red), VR-VIII (green), VR-VIIIB (wheat), and VR-IX (chocolate). Panels A and B were generated using UCSF-Chimera, and panel C was generated with PyMOL (39, 40).

The BPV and HBoV1 capsids have dissimilar surface topologies. The VP2 proteins for these viruses share a sequence identity of 46%. The revised pseudoatomic model (described in Materials and Methods) of HBoV1 VP2 docked into the HBoV1 cryo-reconstructed map with a correlation coefficient of 0.92. Comparison of the HBoV1 and BPV VP2 shows conservation of the βBIDG-CHEF core and α-helix regions with variability localized to residues in the loop regions, characterized by insertions and deletions within a number of the VRs defined for the Parvovirinae (Fig. 3 and 5C and Table 2). BPV has four regions with deletions of two or more residues compared to HBoV1: VR-III (between βE and βF), VR-VIII (between βG and βH), VR-VIIIB (the HI loop), and VR-IX (between βI and the C terminus). A combination of a 7-amino-acid deletion in VR-III located on the 2/5-fold wall and a 6-amino-acid deletion in VR-IX spanning the 2-fold axis results in a shallow and wide 2-fold depression in BPV compared to that in HBoV1, in which the depression is more delineated (Fig. 3 and 5A and C). In contrast, BPV has an insertion of 3 amino acids (residues 218 to 220) downstream of VR-III, which gives rise to the small, raised morphology seen near the 2/5-fold wall (Fig. 3 and 5A and C).

TABLE 2.

Variable regions across the Parvovirinae subfamily

VR	Location (residues)					Capsid location	Functional role(s)c
	BPV-1b	MVMpb	AAV2a	B19b	AMDVb
I	80–90	85–104 (VR0)	258–276 (VR-I)	57–81	82–107	2/5-fold wall	AAV2 antibody recognition and genome packaging; AMDV tissue tropism and pathogenicity; B19 antibody recognition
II	143–147	157–164 (VR1)	319–333 (VR-II)	133–135	160–167	Top of 5-fold channel	AAV2 VP externalization, genome packaging, and transduction; MVMp VP externalization and genome packaging
IIA	174–179	191–196	355–357	164–171	194–197	Capsid interior
III	200–219	217–238 (VR2)	378–394 (VR-III)	189–208	220–254	2/5-fold wall	AAV2 antibody recognition and transduction; AMDV antibody recognition, tissue tropism, and pathogenicity; MVM antibody recognition
IV	256–294	282–289 (VR3, VR4a)	427–439 (VR-IV)	246–283	290–351	3f shoulder	AAV2 antibody recognition and genome packaging; AMDV antibody recognition
V	298–344	336–356 (VR5)	482–511 (VR-V)	296–297	357–383	3-fold protrusion/2/5-fold wall	AAV2 receptor binding and transduction; AMDV antibody recognition
VI	346–357	358–373	523–541 (VR-VI)	299–352	384–401	Base of 3f protrusion	AAV2 genome packaging and transduction; AMDV antibody recognition, tissue tropism, and pathogenicity; B19 antibody recognition; MVMp receptor attachment and tissue tropism
VII	367–370	391–394	544–560 (VR-VII)	356–368	408–426	2/5-fold wall	AAV2 antibody recognition and transduction; AMDV antibody recognition; B19 antibody recognition
VIII	385–405	407–450 (VR4b)	571–604 (VR-VIII)	374–387	438–510	3f protrusion	AAV2 antibody recognition, receptor attachment, and transduction; AMDV antibody recognition, tissue tropism, and pathogenicity; MVM antibody recognition
VIIIA	433–441	479–486 (VR6)	628–633	437–445	539–546	Capsid interior
VIIIB	458–465	504–516 (VR7)	655–667	466–475	564–575	Floor of 5-fold canyon	AAV2 genome packaging and B19 antibody recognition
IX	502–506	558–562 (VR8)	701–720 (VR-IX)	509–518	617–631	2-fold	AAV2 antibody recognition and transduction; MVMp receptor attachment and tissue tropism

^aResidues in VP1 numbering.

^bResidues in VP2 numbering.

^cFunctional roles are reviewed in references 21 and 63.

BPV also has an insertion of 3 amino acids in VR-VIII, which, along with the alternate conformations of VR-IV and VR-V around the 3-fold axis, give BPV larger protrusions than those in HBoV1 (Fig. 3 and 5A and C). This is consistent with the smaller overall diameter seen at the 3-fold axis for HBoV1 (∼260 Å). A conformational difference at the apex of the DE loop (VR-II) gives the exterior surface of the 5-fold channel a wider appearance in HBoV1 (Fig. 5A and C). Additionally, a conformational difference in the HI loops (VR-VIIIB) lining the floor of the depression around the 5-fold channels in BPV and HBoV1 creates a difference in their surface topologies (Fig. 5A). How these VRs correspond to tissue- and/or host-specific functions needs to be further studied.

Commonalities in subfamily VRs localize to functional regions.

The members of the Parvovirinae show high-level sequence divergence, with percent identities in the range of 22 to 26% in VP2 (Table 3). Shared sequence identities are localized to the residues within the conserved β-barrel core and αA helix (data not shown). Consistently, superposition of the BPV VP2 crystal structure onto the VP2/3 (depending on the virus) structures available for the Parvovirinae type members AAV2, AMDV, B19, and MVMp showed conservation of secondary structure elements (Fig. 6A). The overall RMSDs for the BPV VP2 C_α positions superposed onto the C_α positions for the other structures were in the range of 1.6 to 3.2 Å (Table 3).

FIG 6.

Superposition of Parvovirinae VP2 structures and identification of VRs. (A) Cartoon representation of superposed atomic coordinates of B19 (orange), AAV2 (royal blue), and MVMp (green) as well as pseudoatomic coordinates of AMDV (cyan) with the BPV crystal structure (magenta). The positions of the icosahedral 2-fold, 3-fold, and 5-fold axes of symmetry are labeled as described in the legend of Fig. 2C. (B) Cartoon diagram (gray) of BPV with the following VRs: VR-I (purple), VR-II (blue), VR-IIA (forest green), VR-III (yellow), VR-IV (red), VR-V (gray), VR-VI (pink), VR-VII (cyan), VR-VIII (green), VR-VIIIA (orange), VR-VIIIB (wheat), and VR-IX (chocolate). (C) Surface representation of BPV capsid (gray) with VRs colored as described above for panel B. This figure was generated with PyMOL (40).

To date, parvovirus VP VRs have been defined only for members within the same genus, namely, Protoparvovirus (VR0 to VR8) and Dependoparvovirus (VRI to VRIX), which have sequence identities in the range of 50 to 60 and 60 to 99%, respectively (50, 60, 61). The BPV structure increased the number of type member structures available for the Parvovirinae to five, enabling a subfamily-level definition of VRs. The type member structural comparison identified 12 VRs, 9 of which (VR-I to VR-IX) coincide with those defined at the genus level for the dependoparvoviruses, plus three additional VRs (VR-IIA, VR-VIIIA, and VR-VIIIB), named for their preceding VRs and for consistency with the Dependoparvovirus nomenclature (Fig. 6B and C and Table 2). Of these, VR-VIIIA and VR-VIIIB are equivalent to VR6 and VR7, respectively, defined for the protoparvoviruses (50), while VR-VIIIB, or the difference in the HI loop, was recently observed for the Dependoparvovirus AAV5, following its structure determination (62). VR-IIA is the only novel VR identified but is located in the capsid interior under the 2-fold axis. Similarly, VR-VIIIA is located inside the capsid between the 2- and 3-fold axes. The functional relevance of these buried VRs needs to be further investigated. The Parvovirinae subfamily-level VR definition combines the information reported previously with the addition of VR-IIA. These VRs are localized to loops that display high structural variability (>2-Å RMSD) and mostly do not superpose, consistent with large insertions and/or deletions that cluster at or around the symmetry axes (Table 2 and Fig. 6B and C).

The 2-fold region is characterized by VR-III and VR-IX (Table 2 and Fig. 6C). VR-III has a deletion in AAV2 compared to the other type members, reducing the height of its 2-fold wall (Fig. 6C and 7). VR-IX is shorter in BPV and MVMp, and a conformational difference leads to a raised surface in the middle of the 2-fold axis in BPV (Fig. 7). This 2-fold region has receptor attachment properties in MVMp (reviewed in reference 63). VR-I, VR-III, a portion of VR-V, and VR-VII make up the 2/5-fold wall (Fig. 6C). VR-I is shorter in BPV than in the other type members, resulting in a more pronounced 2/5-fold wall (Table 2 and Fig. 7). VR-I, VR-III, and a portion of VR-V form the protrusions on the 2/5-fold wall (which surround the 3-fold protrusions) in B19 and BPV (Fig. 6B and C and 7). VR-VII is shorter in BPV and MVMp than in AMDV, AAV2, and B19 (Table 2 and Fig. 6A and B). Residues in the 2/5-fold wall have been reported to play a role in AMDV tissue tropism and pathogenicity (48, 64) as well as in antibody reactivity to AAV2 (65).

FIG 7.

Comparison of Parvovirinae capsids. Shown are radially depth-cued surface representations (in units of Å) of the parvovirus capsids viewed along the icosahedral 2-fold axis of symmetry. These images were generated using the atomic coordinates of BPV, AAV2, MVMp, and B19 or pseudoatomic coordinates built into the cryo-reconstructed density of AMDV. This figure was generated with UCSF-Chimera (39).

VR-IV, VR-V, and VR-VIII cluster to form the 3-fold protrusions (Table 2 and Fig. 6B and C). A large insertion in AMDV's VR-VIII combined with another in VR-III creates protrusions that are significantly more pronounced than those of the other viruses (Fig. 6 and 7). Differences in the sizes and topologies of VR-IV, VR-V, and VR-VIII give rise to two main types of 3-fold morphologies: a mound-like protrusion formed by the clustering of the VRs from three 3-fold-related monomers, as seen in MVMp, or three distinct protrusions, as seen in AMDV, BPV, B19, and AAV2, with each protrusion being created by two 3-fold-related monomers (Fig. 7). The BPV capsid, unlike HBoV1 (Fig. 5), does not have features similar to those of B19 in this region but rather more closely resembles AAV2 (Fig. 5A and 7). The 3-fold region contains residues important for receptor attachment (in AAV2 and B19) and antibody recognition functions (in AAV2) (21, 65–70).

Three VRs characterize the 5-fold axis: VR-II, located at the apex of the channel, and VR-IIA and VR-VIIIB (HI loop), which are part of the canyon floor (Table 2 and Fig. 6B and C). VR-II displays conformational variability in all members and is longer in AAV2 than in the other type members (Table 2 and Fig. 6A and 7). B19 has a distinct conformation at the apex of its DE loop, resulting in a closed exterior end of the 5-fold channel compared to those of the other type members (52) (Fig. 6A and 7). The residues in the 5-fold channel have been reported to play a role in VP1/VP2 externalization and endosomal escape for AAV2 and MVMp (55, 71). The conformational differences in VR-VIIIB give the canyon floor distinctive morphologies in each type member (Fig. 7). The HI loop is reported to be important for genome packaging in AAV2 (72).

Summary.

This work presents the first high-resolution structure of BPV, the type member of the relatively new yet continually emerging Bocaparvovirus genus of the Parvoviridae family. The structure conserves the core parvovirus VP2 topology, consistent with functional importance, such as in capsid assembly. Variability in surface loops result in a unique capsid morphology, including the lack of a centered 2-fold depression in BPV. This finding suggests a host-specific role that is yet to be determined. In contrast, the extended channel at the 5-fold axes was found to be conserved in the two Bocaparvovirus members compared, suggesting a genus-specific role, which requires further characterization. Significantly, commonalities in the locations of VRs for the Parvovirinae type members with those defined at the genus level, which overlap functional regions, suggest analogous family-level functions. These functions are yet to be defined for the bocaparvoviruses. The BPV structure provides a framework for studies aimed at filling this gap as well as understanding parvovirus capsid evolution. It also serves as a design platform for strategies to control pathogenic BPV infection, in the form of structure-guided design of peptide vaccines and/or small-molecule inhibitors.