Structure of Parvovirus B19 Decorated by Fabs from a Human Antibody

Parvovirus B19 is a common human pathogen and a particular threat to children, pregnant women, and patients with sickle cell disease or AIDS. Currently, neutralizing antibody is the most efficient treatment for acute B19 infections. Research on the antigenic properties of B19 will guide the usage of these antibodies and facilitate vaccine development. We have determined and report here the high-resolution structure of B19 virus-like particles (VLPs) complexed with the Fab of a human neutralizing antibody. The structure shows a quaternary structure epitope formed by three VP2 proteins and provides details on host recognition of human B19 virus.

ABSTRACT

Parvovirus B19, one of the most common human pathogens, is a small DNA virus that belongs to the Parvoviridae. As a result of previous infections, antibodies to B19 are present in most adults. B19 has a strong tropism to erythroid progenitor cells and is able to cause a series of medical conditions, including fifth disease, arthritis, myocarditis, hydrops fetalis, and aplastic crisis. No approved vaccine is currently available for B19, and there is a lack of structural characterization of any B19 epitopes. Here we present the first cryo-electron microscopy (cryo-EM) structure of a B19 virus-like particle (VLP) complexed with the antigen-binding fragment (Fab) of a human neutralizing antibody, 860-55D. A model was built into the 3.2-Å-resolution map, and the antigenic residues on the surface of the B19 capsid were identified. Antibody 860-55D bridges the capsid of B19 by binding to a quaternary structure epitope formed by residues from three neighboring VP2 capsid proteins.

IMPORTANCE Parvovirus B19 is a common human pathogen and a particular threat to children, pregnant women, and patients with sickle cell disease or AIDS. Currently, neutralizing antibody is the most efficient treatment for acute B19 infections. Research on the antigenic properties of B19 will guide the usage of these antibodies and facilitate vaccine development. We have determined and report here the high-resolution structure of B19 virus-like particles (VLPs) complexed with the Fab of a human neutralizing antibody. The structure shows a quaternary structure epitope formed by three VP2 proteins and provides details on host recognition of human B19 virus.

INTRODUCTION

B19 is a human parvovirus in the genus Erythrovirus, which primarily infects erythroid progenitor cells, and it is the etiological agent of childhood erythema infectiosum (fifth disease) (1, 2). B19 infection is also associated with rheumatoid arthritis and myocarditis (3–7) and infections in pregnant women which may lead to hydrops fetalis or even fetal loss (1, 8, 9), and it can cause severe medical problems if the patient has hematological or immunological deficiencies. For example, in patients with AIDS or sickle cell disease, B19 infection interferes with hematopoiesis and causes acute anemia (10–15).

The B19 virus has a linear, single-stranded DNA genome packaged into a T=1 icosahedral capsid. The capsid has a diameter of about 260 Å and is composed of 60 capsid proteins, a mixture of VP1 and VP2. VP2 accounts for about 95% of the viral capsid proteins, whereas VP1 accounts for about 5%. VP1 is essentially identical to VP2, with the exception of a unique region of 227 amino acids (VP1u) at the N terminus of VP2. VP1u contains a phospholipase A2 (PLA₂) domain that is necessary for B19 infection (16). The structure of the mature virus is unknown, whereas the structure of a virus-like particle (VLP) that only contains VP2 has been determined previously (17). A few receptors have been suggested for B19. Erythrocyte P antigen has been shown to play a central role in B19’s tropism to bone marrow (18). The initial receptor binding induces the externalization of the VP1u, which then binds to a coreceptor and initiates the viral uncoating process inside the host cell (19–23).

Human monoclonal antibody (MAb) 860-55D, derived from an HIV-positive patient, diminishes B19 infectivity by 50% at an antibody concentration of about 0.73 μg/ml or 4.9 nM (24). MAb 860-55D binds only to assembled viral particles and therefore has been used to detect mature B19 virions or VP2 VLPs (21, 22, 25–27).

Currently, there is no approved vaccine for B19, and B19 infection is still a threat in certain circumstances. Structural investigations of virus-antibody interaction have been reported for many parvoviruses, but not for B19 (28–32). To study the antigenic properties of B19, we determined the structure of B19 VP2-only VLPs complexed with Fabs of MAb 865-55D to a 3.2-Å resolution using cryo-electron microscopy (cryo-EM) single-particle analysis (Table 1). The electron potential map shows that 860-55D recognizes a conformational epitope composed of three neighboring VP2 proteins close to the “canyon” around the 5-fold axes. By analogy with adeno-associated virus type 2 (AAV-2), the bound antibodies might be blocking the binding of the α5β1 integrin coreceptor, thereby preventing infection. Alternatively, the Fabs that interlock the capsid proteins might prevent uncoating.

TABLE 1.

Statistics of cryo-EM data collection, processing, and model building

Data collection and refinement parameter	Value
No. of movies used	1,759
Defocus range (μm)	1.2–2.5
Electron doses (e−/Å2)	38
No. of frames per movie	40
Exposure time per frame (ms)	200
Pixel size (Å/pixel)	1.30
No. of particles to start with	9,120
No. of particles used for final map	7,395
Symmetry imposed	Icosahedral
Resolution of final map (Å)	3.22
Model refinement with NCS applied
Map-model correlation	0.799
All-atom clash	6.69
Ramachandran outliers	0

RESULTS AND DISCUSSION

Overall structure of the B19-Fab complex.

Purified VLPs were mixed with an excess of Fab molecules at a 1:120 ratio (2 Fab molecules per VP2). The mixture was vitrified after 1 h incubation at room temperature. Due to the bound Fabs, the otherwise smooth VLPs became “fuzzy” in the electron micrographs (Fig. 1A). The cryo-EM data were processed using the jspr package (33) and assuming icosahedral symmetry. A resolution of 3.2 Å was achieved with about 8,000 particles, as calculated by a gold-standard Fourier shell correlation (FSC) curve with a cutoff value of 0.143.

FIG 1.

Cryo-EM processing of the B19-Fab complexes. (A) Raw micrograph of complexes embedded in vitreous ice (bar, 100 nm). (B) Averages of representative 2D classes of the Fab-bound viral particles. (C) Resolution evaluation using a gold-standard FSC curve (dark blue) and an FSC curve between the model and the map (red).

The B19 VLPs were found to be decorated by 60 Fab molecules, one in each icosahedral asymmetric unit (Fig. 2A), and the diameter of the complex is about 390 Å. The light and heavy chains of the Fab molecules bind close to the 5-fold vertices of the virus, while the pseudo-2-fold axis of the variable domain was almost perpendicular to the surface of the capsid. Most of the side chains on VP2 could be built with confidence (Fig. 2B), with the exception of loops 64 to 77, 300 to 312, 357 to 369, 397 to 400, and 525 to 535, which are disordered. The quality of the Fab electron potential density was best closest to the viral capsid. There is no obvious difference in density height between the capsid and the Fab molecules, indicating a high occupancy of the Fab molecules. Given the 1:2 molecular ratio used in the sample preparation, the Fab molecules must be binding tightly to the B19 VLPs.

FIG 2.

The structure of B19 VLP complexed with 860-55D Fab molecules. (A) Surface-rendered cryo-EM map of the B19-Fab complex at 3.2-Å resolution. Yellow (100 Å), green (125 Å), and red (140 Å) coloring indicate increasing distances from the center of mass. (B and C) Representative densities from the map of the B19-Fab complex with the fitted atomic model. (B) Residues 447 to 461 of VP2. (C) The CDR loops on the heavy chain of MAb 860-55D.

The good quality of the potential density map allows model building for the heavy chain of the Fab with a sequence determined from hybridoma cells. However, the sequence of the light chain was not successfully determined. Available structures of homologous Fab molecules (IgG type 3 and light chain type λ) were fitted into the density of the light chain. It was found that the light chain of PDB identifier (ID) 5FHB (34) (except for residues 33 and 44) fits the density very well. Therefore, a pseudomodel generated from PDB ID 5FHB was built into the density and used for the interpretation of the map. Since most of the contact with the epitope was made by the heavy chain, the lack of sequence information for the light chain should not significantly interfere with the determination of the epitope.

Conformational epitope across three neighboring VP2 molecules.

The epitope on B19 recognized by 860-55D is formed by residues distributed across three VP2 proteins, here referred to as P1, P2, and P3 (Fig. 3A). P1 and P2 are related by a 5-fold axis, and P2 and P3 are related by a 3-fold axis.

FIG 3.

The quaternary structure epitope on B19 consists of residues from three VP2 proteins. (A) VP2 proteins around an icosahedral 5-fold vertex. The black pentagon represents an icosahedral 5-fold axis perpendicular to the plane of the figure. The black dashed line outlines the footprint of one of the 60 bound Fab molecules. This Fab recognizes a quaternary epitope formed by three neighboring VP2 proteins, which are colored in blue, green, and red. They are referred to as P1, P2, and P3, respectively. The VP2 proteins that do not interact with the same Fab molecule are colored in gray. (B) “Roadmap” showing the surface residues of VP2 capsid projected onto a planar surface. Each of the residues is colored according to its distance from the center of the capsid. The black triangle outlines an asymmetric unit. The 2-fold axis is at the middle bottom, perpendicular to the image. The boundaries of different VP2 proteins are represented by thick solid lines (red, P1; pink, P2; yellow, P3). Residues that form the conformational epitope (within 4 Å from Fab) are surrounded by white dotted lines.

P2 is located right beneath the bound Fab, whereas the P1 HI extends above P2 and is inserted under the Fab molecule (the nomenclature of the VP2 loops follows that of Kaufmann et al. [17]). P2 and P3 are interlocked, and P3 loop 3 extends between the Fab and P2, almost making contact with P1 loop HI (Fig. 3B). Loop 3 (P255-Y257 and L276-H281) of P3 is surrounded by the three complementarity-determining region (CDR) loops on the heavy chain (Fig. 4A). The HI loop in P1, on the other hand, makes contact with CDRH3 and CDRL1 simultaneously.

FIG 4.

Interactions between the CDR loops of 860-55D and the epitope on B19. (A) Interface of B19-Fab interaction. Three CDR loops on the heavy chain (magenta) and the first CDR loop on the light chain (purple) recognize the epitope formed by three neighboring VP2 proteins (P1, blue; P2, green; P3, red). (B to E) Local interactions between different regions of the quaternary epitope and the CDR loops. The corresponding sites have been highlighted in panel A with circled letters.

Residues that make contact with the Fab molecules were identified using the ncont function in the CCP4 suite with a cutoff distance of 4 Å (35) (Table 2). All of the residues identified by the program were visually confirmed. Notably, I470, L256, R274, H278, and E279 each interact with multiple residues from the Fab and are considered more important in the binding (Fig. 4B to E).

TABLE 2.

Residues of VP2 that form the epitope

VP2 copya	B19 residue(s)	MAb 860-55D residue(s)b
P1	I470, K471, M473	Light chain CDR loop 1
	I470	A105
P2	F57	R103
	S58, P59	Y60
	A60, S62	V59
	A61	G57, T58
	I197	S67
	I197, S198	S70
	D200	Y61
P3	L256	V102, R103
	R274	S31, Y35, Y55
	H278	Y35, D100, R103
	E279	Y35, Y54, S56, T58, Y60
	D280	Y54
	I283	Y55

^aAs in Fig. 3.

^bAll the specific residues in this column refer to the heavy chain.

In summary, P1 loop HI, P2 loop 1 and loop 2, and P3 loop 3 together form a quaternary structure epitope on the surface of the B19 capsid. As a result, when a B19 virus is recognized and bound by antibody 860-55D, the capsid proteins are bridged. The number of copies of antibodies required to neutralize one virus and whether the cross-linking does prevent the uncoating process remain to be studied.

Structural alteration of VP2 by Fab binding.

The root mean square deviation (RMSD) between the atoms in the bound and unbound VLPs was only 1.20 Å. Therefore, binding of the Fab molecules did not significantly change the structure of the VLPs (17); however, there are a few loop regions that change their structures as much as 10 Å for some atoms (K471) (Fig. 5). A major difference occurs in the HI loop (amino acids 467 to 474), which is a part of the epitope. The HI loop is located close to a 5-fold vertex and extends through the boundary between the capsid proteins. With the Fab molecules bound, the tip of the HI loop is bent by about 60^° toward the Fab and interacts with both the CDRH2 and CDRH3 loops. A similar but less significant conformational change occurs in loop 2, which is slightly bent due to interaction with CDRH2. Among the five disordered loops in the cryo-EM map, only loop 300 to 312 is also disordered in the crystallographic structure of the unbound VLP. This is possibly due to stabilization by crystalline packing.

FIG 5.

Comparison of B19 VP2 with Fab binding (red), B19 VP2 without Fab binding (PDB ID 1S58; green) and AAV-2 VP3 (PDB ID 1LP3, blue). The most significant conformational change caused by Fab binding occurs at the HI loop, with a 10-Å difference in the alpha carbon (C-α) of K471. The residues that form the epitopes of 860-55D (B19; red) and A20 (AAV-2; blue) are represented in transparent spheres.

Mechanism of neutralization.

Antibodies that recognize viral capsid proteins are able to neutralize infections by inhibiting genome release (uncoating for nonenveloped viruses or membrane fusion for enveloped viruses), blocking c0ell attachment, or inducing irreversible conformational changes (or even disassembly) of the virion. Because B19 is structurally similar to AAV-2, with an RMSD of 1.8 Å (17, 36), we therefore evaluated B19 neutralization within the context of studies on AAV-2.

AAV-2 uses a primary receptor, heparin sulfate proteoglycan, for initial cellular attachment and then binds to coreceptors (AAVR; the function of αVβ5 as a coreceptor is in question) for internalization (37–39). Similarly, B19 utilizes blood P antigen for attachment and α5β1 integrin as a cellular coreceptor. A neutralizing antibody of AAV-2, MAb A20, allows viral attachment but inhibits the internalization of the viruses (40). Superimposing the structures of B19 VP2 and AAV-2 VP3 showed that A20 has a very similar conformational epitope with 860-55D (Fig. 5). Four of the five regions on AAV-2 that form the epitope of A20 overlap the epitope of 860-55D in the alignment (31). In contrast, the epitope of 860-55D does not overlap the footprint of heparin sulfate on AAV-2 (41). Therefore, the neutralization of B19 by MAb 860-55D is likely achieved by blocking receptor binding to α5β1 integrin. However, it is also possible that 860-55D abrogates B19’s ability to uncoat inside a host cell by bridging the capsid proteins.

MATERIALS AND METHODS

Generation of B19 VP2-only VLPs and the Fab of 860-55D.

The procedure described here was based on the protocol described in previous studies (17, 42). The VP2 capsids were produced by a baculovirus-based expression system. VP2 protein was expressed in Spodoptera frugiperdes (Sf9) insect cells and assembled into icosahedral (T=1) capsids. The Autographa californica multiple nucleopolyhedrovirus (AcMNPV)-VP2 construct described previously (17) was used to infect Sf9 cells at around 0.5 × 10⁶ cells/ml as a suspension in Sf-900 II medium (Thermo Fisher Scientific) without any antibiotics. The cell culture was shaken continuously at 27°C and harvested 70 h after infection by centrifugation at 4,000 × g. The pallet was resuspended in lysis buffer (10 mM Tris-HCl [pH 7.4]; 10 mM NaCl, 15 mM MgCl₂, and 0.5% Triton X-100, supplemented with protease inhibitors). The cells were lysed by either repeated freezing and thawing or mild sonication. The lysate was centrifuged at 2,000 × g to eliminate the remaining cells and large cellular debris. The recombinant capsids in the lysate were sedimented into a 30% CsCl cushion by centrifugation at 150,000 × g. No band could be seen after the centrifugation, and a mouse antibody (ab64295; abcam) was used to determine the location of the capsids. The lower part of the cushion was extracted and added to a solution containing 36% CsCl, 50 mM Tris-HCl (pH 8.7) and 25 mM EDTA. The mixture was centrifuged again at 150,000 × g for 20 to 24 h until a density equilibrium was reached. An opaque band was extracted and dialyzed against phosphate-buffered saline (PBS)-Mg (8 mM Na₂HPO₄, 1.47 mM KH₂PO₄, 137 mM NaCl, 2.68 mM KCl, and 0.49 mM MgCl₂). The solution was then concentrated to 0.1 to 1.0 mg/ml, depending on the following experiments.

Fab fragments of human antibody 860-55D were produced following the instructions of the Pierce Fab preparation kit (Thermo Fisher Scientific). First, 0.5 mg IgG was desalted by passing through a Zeba spin desalting column. Then, the flowthrough was mixed with equilibrated immobilized papain at 37°C for 4 h for digestion. The Fab fragments were purified from the fragment crystallizable (Fc) fragments by passing through an equilibrated protein A column. Additional washing was done to achieve optimal recovery. The flowthrough fractions were combined and concentrated to obtain the purified Fab fragments.

Sample vitrification and electron microscopy imaging.

Purified VLPs at 0.4 mg/ml and Fab at 0.3 mg/ml were mixed at a volume ratio of 11:7, resulting a molar ratio of approximately 1:120 (1.9 Fab molecules for each VP2 protein). The mixture was incubated at room temperature (22°C) for 1 h. Aliquots of 3 μl of the virus-Fab mixture were applied to glow-discharged lacey carbon grids (400 mesh copper, lot no. 200617; Ted Pella, Inc.) before the grids were blotted for 4 to 5 s and plunged into liquid ethane using a Cryoplunge 3 system (Gatan). A total of 1,759 movies (3,838 × 3,710 pixels) of the vitrified complexes embedded in vitreous ice were collected in two sessions using a Titan Krios microscope operated at 300 kV with a Gatan K2 direct electron detector in counting mode. Automated data collection was enabled by Leginon (43). The nominal magnification was ×22,500, which produced a pixel size of 1.30 Å. The defocus range was set to 1.2 to 2.5 μm. Each movie consisted of 40 frames and each frame had an exposure time of 200 ms. The dose rate received by the detector was 8 e⁻/pixel, generating a total dose of 38 e⁻/Ų for each movie.

Image processing.

Relative motion between the frames within a movie was corrected using MotionCorr (44), as modified by Wen Jiang at Purdue University. Contrast transfer function (CTF) estimation was calculated with CTFFIND3 (45). Semiautomatic particle boxing was carried out using the program e2boxer.py in the EMAN2 package (46). A total of 9,120 particles was found and confirmed by visual inspection. A subsequent two-dimensional (2D) classification was calculated by RELION (47) with a 400-Å diameter mask. A total of 7,395 particles was selected and randomly divided into two half subsets. The two subsets were then independently refined against randomly generated initial models using the jspr package (33) and assuming icosahedral symmetry. Parameters, including orientation, center, defocus, astigmatism, scale, beam tilt, and magnification anisotropy, were refined for each of the particles through multiple iterations until convergence was reached. The FSC was calculated between the resulting maps of the two subsets. The resolution was 3.22 Å according to the 0.143 criterion (48). Then the subsets were combined to generate a final map. The map was then low-pass filtered to 3.2 Å.

Model building and refinement.

The previously determined structure of VP2 was fitted into the cryo-EM electron potential density (17). Further modeling was done using the program Coot (49). For the Fab molecule, the variable region from another IgG molecule (PDB ID 5FHB) was first fitted into the density. The heavy chain was then mutated and adjusted residue by residue. For the light chain, if a specific residue did not agree with the density, it was replaced by alanine. The models for the capsid and the Fab molecule were combined and refined together in real space using Phenix (50). The icosahedral symmetry was then applied to the model of one VP2-Fab complex to generate the entire capsid. Finally, the entire capsid-Fab complex was refined to reduce clashing and to maximize correlation with the experimental electron potential density using Phenix with noncrystallographic symmetry (NCS) restraints (50). To evaluate the overall agreement between the model and the electron potential density, a map was calculated based on the model. The FSC curve was then calculated between the model map and the original cryo-EM map.

Data availability.

The final map, two half maps, and the FSC eXtensible Markup Language (XML) file have been deposited with the Electron Microscopy Data Bank (EMDB) under accession code EMD-9110. The coordinates of the VP2-Fab complex were deposited with the Protein Data Bank under ID 6NN3.