Asymmetric binding of transferrin receptor to parvovirus capsids
- Susan Hafenstein*,
- Laura M. Palermo†,‡,
- Victor A. Kostyuchenko*,
- Chuan Xiao*,
- Marc C. Morais*,§,
- Christian D. S. Nelson†,
- Valorie D. Bowman*,
- Anthony J. Battisti*,
- Paul R. Chipman*,
- Colin R. Parrish†, and
- Michael G. Rossmann*,¶
- *Department of Biological Sciences, Purdue University, 915 West State Street, West Lafayette, IN 47907-2054; and
- †The James A. Baker Institute, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853
-
Contributed by Michael G. Rossmann, February 22, 2007 (received for review February 18, 2007)
Abstract
Although many viruses are icosahedral when they initially bind to one or more receptor molecules on the cell surface, such an interaction is asymmetric, probably causing a breakdown in the symmetry and conformation of the original infecting virion in preparation for membrane penetration and release of the viral genome. Cryoelectron microscopy and biochemical analyses show that transferrin receptor, the cellular receptor for canine parvovirus, can bind to only one or a few of the 60 icosahedrally equivalent sites on the virion, indicating that either canine parvovirus has inherent asymmetry or binding of receptor induces asymmetry. The asymmetry of receptor binding to canine parvovirus is reminiscent of the special portal in tailed bacteriophages and some large, icosahedral viruses. Asymmetric interactions of icosahedral viruses with their hosts might be a more common phenomenon than previously thought and may have been obscured by averaging in previous crystallographic and electron microscopic structure determinations.
Successful viral infection begins with capsid attachment to a cellular surface and penetration into the interior of the host cell. Viruses have adapted to use a wide variety of cellular surface proteins or glycans as receptors (1). In the available structural studies of icosahedral viruses interacting with the soluble ectodomain of their cellular receptors, the receptor molecules bind to each symmetry equivalent unit, remaining faithful to the original symmetry of the virion (2–4). Nevertheless, the initial interaction of such a virus with a membrane-bound receptor is inherently asymmetric (5). Canine parvovirus (CPV) and feline panleukopenia virus bind to transferrin receptor (TfR) with specific interactions, allowing the viruses to use TfR for cell attachment and infection (6).
Parvoviruses are small, 260-Å-diameter, icosahedral, nonenveloped, single-stranded DNA viruses with a genome of ≈5 kb. Each of the 60 subunits consists of an eight-stranded, antiparallel, β-barrel motif (7) that is found in numerous viral capsid structures (8–10). In CPV (7) and feline panleukopenia virus (11), the β-barrel has large insertions between β-strands that form most of the capsid surface and create small, protruding “spikes” around the icosahedral threefold axes. CPV emerged as a natural variant of feline panleukopenia virus in 1978 (12). Whereas both viruses can use feline TfR to enter cells, CPV has gained the ability to bind canine TfR, albeit at a lower affinity (13). Residues that are involved in host-range control and specific recognition of TfR are on the viral surface in the vicinity of the spikes (6, 14).
TfR is a type II membrane protein that protrudes ≈30 Å from the cell surface. The structure of human TfR (15, 16), which has 79% amino acid identity to feline TfR, consists of a large, butterfly-shaped, dimeric molecule with a span of ≈100 Å and a molecular mass of 180 kDa. Each monomer has a carboxypeptidase-like domain, an apical domain, and a helical domain. Mutagenesis of feline and canine TfR indicates that CPV and feline panleukopenia virus bind to the apical domain of the receptor, distal from the membrane-binding region (14).
Results and Discussion
Attempts to obtain icosahedrally averaged cryoelectron microscopy (cryoEM) image reconstructions of CPV incubated with the soluble ectodomain of feline TfR failed to detect any additional density that might correspond to receptor. The lack of TfR density in the reconstruction could not be explained by lack of occupancy due to steric hindrance. If a TfR molecule at one binding site were in steric conflict with a TfR molecule bound to one or more neighboring binding sites, there would have been density at the sterically overlapped regions. Hence, the result of icosahedral averaging would be to produce density equal in height to that of the viral capsid itself in the overlapped region. Furthermore, micrographs of negatively stained virus–receptor complexes showed capsids bound to only one or two receptors in a background of unbound TfR protein (Fig. 1). Apparently, the absence of TfR on the icosahedrally averaged cryoEM maps was because the TfR occupied only one or a few of the 60 icosahedrally equivalent sites on the virus. Therefore, the receptor density would have been averaged to a level indistinguishable from noise when using icosahedral symmetry in the reconstruction process. Thus, to visualize the TfR molecules binding to the virion surface, it was necessary to perform an asymmetric reconstruction that did not obliterate the receptor as a result of icosahedral averaging.
EM of CPV incubated with excess TfR negatively stained with uranyl acetate. Although the TfR dimer has a diameter of ≈120 Å, negatively stained molecules are somewhat elongated, resulting in a variable profile that depends on the orientation of the molecule on the EM grid. Upon examination by negative stain and by ultracentrifugation, a minor population (≈5%) of TfR particles has been shown to exist as dimers of dimers. These interacting dimers of TfR still bind virus, as observed by EM. Although receptors do aggregate, especially when stored for extended periods of time or exposed to temperatures of <24°C, care was taken to avoid aggregation. (Scale bar, 500 Å.)
The asymmetric reconstruction of the TfR–virus complex (Fig. 2 A) had good icosahedral symmetry for the EM density representing the virus (Fig. 3). Furthermore, there was additional density, with a maximum height of 0.63 of the capsid itself, situated between one of the adjacent three- and fivefold symmetry axes of the virion. In contrast, a control asymmetric reconstruction, performed by the identical procedure that used CPV capsids that had not been incubated with TfR, had additional density that was only 0.34 of the capsid itself (Fig. 2 B).
A surface-shaded representation of the cryoEM asymmetric reconstructions filtered to ≈25-Å resolution of empty CPV capsids complexed with feline TfR by using 8,500 particles (A) and of uncomplexed CPV by using 2,220 particles (B). The lack of icosahedral averaging restricted the final resolution of the reconstructions to be approximately that which would have been attained had only ≈1/60 of the particles been used in icosahedrally averaged reconstructions. The two maps were scaled to have a maximum height of 100. Based on this arbitrary value, the maximum height of the putative TfR density was 63 in the virus–TfR complex map, and the maximum height of noise density in the native virus map was 30. The contour level used for display was 24, cutting off most of the noise in the native map but showing the putative TfR density in the complex map. A dashed line indicates the location of the spherical mask defining the limits used in computing the correlation between the observed difference image and the projected density of the current model. A representative micrograph is displayed behind each reconstruction. The yellow circle indicates the boundary used for selecting each particle in the image reconstruction. The area was of sufficient size to include bound TfR molecules for those particles in which the projected TfR was in the plane of the ice. The effective resolution of the maps was 25 Å for the complex and 35 Å for the control.
The rotation function shown as a stereographic projection on which the polar angles θ and ϕ represent latitude and longitude, respectively, was calculated by using structure factors between 50- and 25-Å resolution for the asymmetric reconstruction of the CPV–TfR complex (18). Density beyond the radius of 140 Å was not included. The five-, three-, and twofold rotation-function sections are shown at 3σ above average, with a 1.5σ step size, and are red, blue, and green, respectively. Icosahedral symmetry axes are marked in black. The limits of each geometrically designated icosahedral asymmetric unit are outlined by black great circles. The rotation function demonstrates that the asymmetric reconstruction still shows good icosahedral symmetry.
The crystal structure of the human TfR ectodomain dimer (15) (Protein Data Bank accession no. 1CX8) could be roughly positioned into the cryoEM additional density (Fig. 4) in such a way that one of the two apical domains was in contact with the shoulder of one of the spikes on the CPV surface. The surface area of contact between the virus and TfR was 1,000 Å2. CPV surface residues within 5 Å of the fitted TfR structure were identified and mapped to the surface of the virus (Fig. 5). A difference map was calculated between the virus–receptor complex and the native virus (17). The difference density in the shell between 143- and 148-Å radii, corresponding to the area of contact between the virus and the putative TfR density, was superimposed on a stereographic projection, together with the atomic positions on the icosahedral CPV surface structure (18). The contact residues matched the projected difference density in the 143- to 148-Å shell (Fig. 5) and included residues that control specific binding to TfR (6, 13).
The crystal structure of a TfR dimer (15) placed manually into the cryoEM density (Fig. 2 A) shows a reasonable fit. One of the two apical domains (green) of a TfR dimer is shown interacting with the virus surface at the shoulder of a spike near a threefold icosahedral axis, as viewed down a threefold axis. The helical domains and carboxypeptidase-like domains of the surface-rendered TfR molecule are yellow and red, respectively. The unoccupied density might be the result of errors in choosing the correct threefold related orientations for some of the cryoEM images during the reconstruction process.
TfR footprint on the CPV surface. The cryoEM difference density (red contours) between 143- and 148-Å radii, the contact region between the virus and TfR, was projected onto the viral surface and is shown as a stereographic projection on which the polar angles θ and ϕ represent latitude and longitude, respectively (18). By associating the surface atoms within the same amino acid residue, the virus surface could be represented as a quilt of amino acids in a manner similar to the “road map” representation of viral surfaces (35). The residues within 5 Å of the fitted TfR molecule (Fig. 4), shown in green or yellow, create the footprint of the TfR model on the virus surface, which is outlined with a thick black line. Residues 93, 299, and 301, which are associated with TfR binding (6, 13), are yellow. The icosahedral asymmetric unit of the virus is indicated by the triangular boundary. Polar coordinates θ and ϕ are the same as defined in Fig. 3.
A reasonable conclusion derived from the EM studies is that most of the CPV capsids had only one or a few bound TfR molecules, although the virus particles had been incubated with a 3-fold excess of receptor molecules to each of the 60 potential binding sites per virion. However, the procedures used for the cryoEM image reconstruction aimed to superimpose a unique TfR molecule on the surface of the virus. Thus, if one or two additional, symmetrically equivalent sites had also been occupied on each virion, these additional TfR molecules would be unlikely to superimpose and would disappear in the iterative reconstruction process.
Biochemical measurements were made to quantify the number of feline TfR molecules bound per capsid by incubating virions with different relative amounts of receptor to form complexes. After purification by size-exclusion chromatography, a Western blot analysis showed that there were only 5 ± 2 TfR dimers per capsid, irrespective of the ratio (3:1, 10:1, or 30:1) of TfR to virus in the preincubation process. These observations are consistent with the cryoEM results, which suggested that only one or a few receptor molecules bound to each virus capsid.
The effect of TfR binding to capsids was also examined by preincubating CPV with different amounts of TfR before application to cells expressing TfR (see Materials and Methods). As few as only three TfR dimers per CPV particle reduced virus binding to cells by a factor of 20%; and, with 10 TfR molecules per virion, binding to cells was reduced by 60% (Fig. 6). One possible explanation of the limited number of TfR molecules bound to CPV virions, as observed in the cryoEM reconstruction, is that TfR binds only poorly and would require a greater excess of receptor to obtain fuller substitution. However, only 3–10 TfR molecules per virion were sufficient to reduce virus binding to cells by 20–60%, suggesting that, whereas the first TfR molecules can bind to the virus quite readily, subsequent additional binding of TfR is less likely to occur.
FACS analysis of virus attachment. (A) Mean fluorescence intensity (MFI) was used as a basis of comparison to determine how the different ratios of TfR to CPV inhibit subsequent binding to cells. (B) Graph depicting nonlinear reduction of binding of CPV to cells as measured by MFI.
The asymmetric binding of TfR might be due to inherent asymmetry of the virus, with one unique site that has a conformation capable of binding TfR, whereas the other icosahedrally equivalent sites are slightly different. For an infectious virus, this might happen if one end of the genome were protruding or if the final one or two subunits were sterically hindered from perfectly finishing the assembly process. Such events are reminiscent of tailed bacteriophage assembly that is probably initiated around a special icosahedral fivefold vertex, which is the site of subsequent genome entry and exit (19, 20). Furthermore, some large, icosahedral, dsDNA viruses also develop special vertices, presumably for injection of their genome into the host (20, 21).
Alternatively, the binding of TfR might induce asymmetry in the initially icosahedral virus, akin to the cooperative behavior observed in some oligomeric enzymes. The asymmetry of the virus after TfR binding was examined with a difference map between the asymmetric reconstruction of the virus–TfR complex and the icosahedral reconstruction of an empty CPV particle. However, at the available resolution, significance of the asymmetry could not be interpreted with confidence.
Binding of one receptor molecule to a unique, well exposed part of the virus, as observed for TfR binding to CPV empty particles, contrasts with many picornaviruses, in which there are 60 equally functional sites of receptor interaction located in surface depressions (canyons) that are mostly inaccessible to antibodies (2). In the case of picornaviruses, the receptor-binding site is hidden from neutralizing antibodies and, thus, is used as a strategy for escaping from immune surveillance by the host. Although the exposed receptor-binding site in parvoviruses is available to neutralizing antibodies, there could be advantages in limiting cell recognition to a unique site on the viral surface. Parvoviruses are rather inefficient at cell infection, requiring, on average, >1,000 virions to infect a single host cell (22), possibly because parvoviruses must cross the plasma membrane to travel to the cell nucleus before entering the nucleus. It would be a biological advantage in the initial stages of a difficult infection route to form a unique portal near the site of receptor interaction through which delivery of the genome could be initiated. This strategy, used by tailed bacteriophages (23, 24), might also be used by Microviridae (25) and giant DNA icosahedral viruses (9, 21). Indeed, the development of asymmetry in icosahedral viruses during the infection process may be a more general phenomenon than has been previously recognized.
Materials and Methods
Cells and Viruses.
Trichoplusia ni (High Five) cells were grown in shaking flasks in Express5 serum-free medium, and Spodoptera frugiperda (Sf9) cells were grown in TNM-FH medium (fully supplemented Grace's medium, including trace metals) with 10% FBS. Feline NLFK cells were grown in a 1:1 mixture of McCoy's 5A and Liebovitz L15 media with 5% FBS. TfR-deficient TRVb cells were grown in Ham's F12 medium with 10% FCS. The feline TfR was expressed from plasmids in TRVb cells by transfection as previously described (13). CPV type 2 strain was derived from an infectious plasmid clone of the viral genome (26). Capsids were prepared by growing virus in NLFK cells, concentrated by polyethylene glycol precipitation, followed by sucrose gradient centrifugation to purify full and empty capsids, and dialyzed against either PBS or 20 mM Tris·HCl (pH 7.5) and stored at 4°C.
Empty particles of CPV were used in these studies (7). Conformational differences between fully infectious, DNA-filled CPV particles and empty particles are small and confined to the interior of the capsid (7, 27). Thus, whereas the present study concerns itself exclusively with the interaction of feline TfR with empty particles of CPV, it is reasonable to expect similar results for receptor interaction with the corresponding infectious virions.
Feline TfR Ectodomain Expression.
The ectodomain of the feline TfR was expressed from a baculovirus vector (28). Purified feline and canine TfR molecules reveal complex interactions with the capsids of canine and feline parvoviruses that correspond to their host ranges (28). The clone contained an N-terminal sequence coding for a baculovirus gp67 secretion signal, a cleavage site, and a 6-His nickel-binding tag fused to residue 121 of the feline TfR. After infection of High Five cells for 3 days, the culture medium was enzymatically treated, clarified, filtered, and dialyzed against 50 mM Tris·HCl (pH 7.5) and 150 mM NaCl. The TfR ectodomain was isolated by binding to Ni2+–N-nitrilotriacetic acid resin (Qiagen, Valencia, CA) in either 0 or 10 mM imidazole/10% (vol/vol) glycerol/300 mM NaCl. The TfR was eluted with 25–100 mM imidazole. The TfR then was purified by chromatography in a Sephacryl S300 column (GE Healthcare Life Sciences, Piscataway, NJ) in 50 mM Pipes–HCl (pH 7.5) with 0.15 M NaCl. The TfR protein was concentrated at 22°C by filtration in Ultra 15 10,000-Da cutoff filters (Millipore, Billerica, MA) and stored at 4°C.
CryoEM and Image Reconstruction.
Purified empty CPV particles were mixed with soluble feline TfR ectodomains at a molar ratio of 1:180 CPV/TfR and incubated at 24°C for 1 h. Aliquots (≈3.5 μl) of the mixture were applied to carbon-coated electron microscope grids and negatively stained or vitrified in liquid ethane as described by Baker et al. (29). Electron micrographs were recorded on SO-163 films (Kodak, Rochester, NY) in a CM200 FEG, transmission electron microscope (Philips Electronic Instruments, Eindhoven, The Netherlands) at a magnification of ×50,000 (complex) or ×38,000 (native). Micrographs were digitized on a PHODIS microdensitometer (ZI Imaging Intergraph Corp., Englewood, CO) at 7-μm intervals and locally averaged to correspond to a 2.8- or 3.68-Å pixel size, respectively. Selection of particles used for the image reconstruction effectively separated the CPV–TfR complexes from aggregates.
A procedure was developed by using a combination of the SPIDER program (30) and a modified version of XMIPP (31) (available upon request); the procedure used the particle orientations found in an icosahedral reconstruction of the CPV–feline TfR complex. The density of the icosahedrally reconstructed virus was subtracted from each image (32). A spherical mask of radius 95 Å, centered 185 Å from the origin of the virus, was defined to roughly correspond to one icosahedral asymmetric unit. The area of each of the 60 possible projections of the mask, given the angles that define the virus orientation for each specific difference image, was examined to determine the best correlation between the density of the difference image and the projected density of the mask. These orientations were used to compute an asymmetric reconstruction. The procedure was iterated by using the resultant map to reselect from the 60 possible orientations the preferred angle that defined the TfR position relative to the orientation of the virus. Convergence of the particle orientations was reached after approximately six cycles. The final orientations were then applied to the original images of the virus–TfR complex to compute a reconstruction.
CryoEM Map Analyses.
The cryoEM map of the CPV–TfR complex obtained from an asymmetric reconstruction was Fourier-transformed to produce structure factors that were input into the GLRF program (33). The resultant rotation function showed that the orientations of the observed five-, three-, and twofold axes were consistent with icosahedral symmetry. TfR difference density was calculated by subtracting the icosahedrally averaged cryoEM reconstruction of CPV from the map of the complex. Both maps were limited to 25-Å resolution, and their effective “temperature factors” were equalized before scaling to a map calculated from x-ray diffraction data (Protein Data Bank ID code 1C8D) by comparing the densities within the protein shell.
Biochemical Determination of Stoichiometry of Receptor and Capsid After Complex Formation.
Gel chromatography and quantitative Western blot were used to determine the stoichiometries of TfR relative to capsid in complexes. Capsids were incubated for 30 min at 22°C with the purified feline TfR ectodomain at molar ratios of 1:15, 1:30, and 1:60 and then passed through a 60- × 0.6-cm Sepharose CL2B column in Pipes–HCl (pH 7.5) and 0.15 M NaCl. Fractions corresponding to aggregation (void volume), CPV–TfR complex, and control fractions to establish background were probed for CPV and TfR by using quantitative Western analysis of slot blots on nitrocellulose filters (34). The band intensity of each sample was determined, and the quantities of the unknown samples were estimated with respect to a calibration curve produced by using predetermined amounts of CPV and TfR.
The Effect of Soluble Feline TfR on the Virus Capsid.
Feline TfR was expressed from plasmids in TRVb cells by transfection as previously described (13), and the transfected cells were incubated for 48 h. CPV capsids (10 μg/ml) were incubated with soluble TfR at molar ratios of 1:0, 1:3, 1:10, and 1:60 for 60 min at 22°C. The mixtures were then placed on cells along with Cy5-labeled canine transferrin and incubated for 60 min at 37°C. The cells were fixed with 4% paraformaldehyde and permeabilized, and the virus was detected with Cy2-labeled Mab8. The amounts of cell-associated virus and transferrin were determined by using a FACScalibur flow cytometer (Becton–Dickinson, San Jose, CA).
Acknowledgments
We thank John Burgner for ultracentrifugation analyses and Sheryl Kelly, Cheryl Towell, and Sharon Wilder for help in preparation of the manuscript. Some of the codes used to produce the various stereographic projections incorporated in the program RIVEM were derived from the GLRF program written by Liang Tong (Columbia University, New York, NY). This work was supported by National Institutes of Health Grants AI 11219 (to M.G.R.) and AI 33468 (to M.G.R. and C.R.P.). S.H. was supported by National Institutes of Health Postdoctoral Fellowship AI 060155.
Footnotes
- ¶To whom correspondence should be addressed. E-mail: mr{at}purdue.edu
-
Author contributions: S.H., L.M.P., V.A.K., M.C.M., C.D.S.N., C.R.P., and M.G.R. designed research; S.H., L.M.P., V.A.K., M.C.M., C.D.S.N., V.D.B., A.J.B., and P.R.C. performed research; S.H., V.A.K., and C.X. contributed new reagents/analytic tools; S.H., L.M.P., V.A.K., C.X., M.C.M., C.D.S.N., V.D.B., A.J.B., P.R.C., C.R.P., and M.G.R. analyzed data; and S.H. and M.G.R. wrote the paper.
-
↵ ‡Present address: Pediatrics/Infectious Diseases, Weill Medical College, Cornell University, New York, NY 10021.
-
↵ §Present address: Department of Biochemistry and Molecular Biology, University of Texas Medical Branch, 6.614B Basic Science Building (Mail Route 0647), 301 University Boulevard, Galveston, TX 77555-0647.
-
The authors declare no conflict of interest.
-
Data deposition: The atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org [PDB ID code 2NSU (structure of the ectodomain of human TfR fitted into a cryoEM reconstruction of CPV and feline TfR complex)] and cryoEM reconstructions have been deposited in the Electron Microscopy Data Bank [EMDB reference nos. 1287 (CPV) and 1288 (CPV complexed with feline TfR)].
- Abbreviations:
- CPV,
- canine parvovirus;
- cryoEM,
- cryoelectron microscopy;
- TfR,
- transferrin receptor.
- © 2007 by The National Academy of Sciences of the USA
