|Home | About | Journals | Submit | Contact Us | Français|
Many entero-, parecho-, and rhinoviruses use immunoglobulin (Ig)-like receptors that bind into the viral canyon and are required to initiate viral uncoating during infection. However, some of these viruses use an alternative or additional receptor that binds outside the canyon. Both the coxsackievirus-adenovirus receptor (CAR), an Ig-like molecule that binds into the viral canyon, and decay-accelerating factor (DAF) have been identified as cellular receptors for coxsackievirus B3 (CVB3). A cryoelectron microscopy reconstruction of a variant of CVB3 complexed with DAF shows full occupancy of the DAF receptor in each of 60 binding sites. The DAF molecule bridges the canyon, blocking the CAR binding site and causing the two receptors to compete with one another. The binding site of DAF on CVB3 differs from the binding site of DAF on the surface of echoviruses, suggesting independent evolutionary processes.
Coxsackieviruses belong to the family Picornaviridae, which contains some of the most common viral pathogens of vertebrates (39, 45, 48, 54, 56, 62). Picornaviruses are small, icosahedral, nonenveloped animal viruses. Capsids have 60 copies of each of four viral proteins, VP1, VP2, and VP3 (Fig. (Fig.1)1) and VP4, that form an icosahedral shell ~300 Å in diameter filled with a positive-sense, single-stranded RNA genome. A prominent feature of the capsid surface is a depression around the fivefold axes of symmetry called the “canyon” (41). The results of both genetic and structural studies have shown that the canyon is the site of receptor binding for many of these viruses (2, 9, 19, 32, 43), including coxsackieviruses which utilize the coxsackievirus-adenovirus receptor (CAR) (13). Receptor molecules that bind into the canyon have been found to belong to the immunoglobulin superfamily (43). When these receptor molecules bind into the canyon, they dislodge a “pocket factor” within a pocket immediately below the surface of the canyon whose shape and environment suggest that it might be a lipid (19, 52). The flexible floor of the canyon is depressed into the pocket when receptor binds. In contrast, the roof of the pocket is pushed upwards when occupied by pocket factor. Hence, the receptor molecule and pocket factor have overlapping binding sites, thus competing with each other for binding to the virus. Absence of the hydrophobic pocket factor destabilizes the virus and initiates transition to altered “A” particles (10), a likely prelude to uncoating of the virion, possibly during passage through an endosomal vesicle (31, 59).
Not all receptors of picornaviruses bind in the canyon. A group of minor rhinoviruses bind to very-low-density lipoproteins (16, 30, 61), and some other picornaviruses, including certain coxsackie- and echoviruses, utilize decay-accelerating factor (DAF) (CD55) as a cellular receptor (7, 22, 36, 50). In general, coxsackievirus serotype B2 (CVB2), CVB4, and CVB6 do not bind DAF, whereas CVB1, CVB3, and CVB5 do bind DAF (49), although there are variations in DAF binding within the CVB3 serotype. Passage of CVB3 through human rhabdomyosarcoma (RD) cells, a cell line that is normally resistant to CVB3 infection, gives rise to a variant form of the virus, CVB3-RD. This variant is able to grow well on RD cells while retaining the ability to infect HeLa cells, albeit with less efficiency and by forming smaller plaques. The binding site of DAF on the surfaces of echovirus 7 (EV7) and EV12 crosses the icosahedral twofold axes of symmetry (4, 13, 35). Mutational analysis has shown that two amino acids, residues 2108 and 2151 (residues in VP1, VP2, VP3 and VP4 are numbered sequentially starting with 1001, 2001, 3001, and 4001, respectively), determine the CVB3-RD phenotype in some strains (24). These residues are located near the twofold axes of symmetry (29). However, subsequent work showed that additional surface residues, 1080, 3137, 3184, and 3234, might contribute to defining the CVB3-RD phenotype (51).
DAF is a member of a family of proteins that regulate complement activation by binding to and accelerating the decay of convertases, the central amplification enzymes of the complement cascade (5, 20, 21, 27). As DAF is expressed on virtually all cell surfaces, the host's cells are protected from attack by the host's own immune system. The functional region of DAF consists of four short consensus repeats (SCR1, -2, -3, and -4), each about 60 amino acids long. The four domains rise about 180 Å above the plasma membrane on a serine- and threonine-rich stalk of 94 amino acids, 11 of which are O glycosylated and attached to the plasma membrane by a glycoslphosphatidylinositol anchor. The entire DAF molecule is relatively rigid, forming an extended rod with dimensions of 160 by 50 by 30 Å (25). Each of the SCR domains is folded into a β structure stabilized by disulfide bridges. The solution structures of the SCR2 and -3 domains and the crystal structures of the SCR3 and -4 domains as well as the full-length human DAF have been determined (25, 60, 64).
Structural and genetic studies have shown that closely related viruses have adapted to bind to DAF at different sites on the receptor surface (Table (Table1)1) (7, 18, 26, 37, 38, 50, 63). Furthermore, different regions on the same SCR can be utilized by different viruses for attachment (3, 35). Although DAF binding is likely to facilitate viral adsorption and mediate tropism, the availability of DAF receptor molecules on the host is normally not sufficient for CVB3 or EV7 to be able to enter cells (37, 47). Presumably, viral adaptation to bind DAF provides additional advantages. In order to investigate DAF interactions with CVB3, a cryoelectron microscopy (cryoEM) reconstruction of DAF bound to CVB3-RD was made. The binding site of DAF was found to cross the canyon on the viral surface near the “puff” region on VP2 (41) that defines the southern canyon rim.
Human DAF was expressed in Pichia pastoris as a C-terminally His6-tagged protein (14). The DAF construct consisted of the full-length ectodomain, containing SCR1, -2, -3, and -4 (amino acids 1 to 254) but lacking the S/T-rich linker domain and the glycoslphosphatidylinositol anchor. The ectodomain of CAR protein (D1D2) used in the plaque reduction assays was kindly provided by Paul Freimuth (Brookhaven National Laboratory) and was produced and purified as described previously (13).
The CVB3 M strain (11) was propagated in HeLa cells and purified as described previously (29). CVB3-RD was propagated in RD cells from an inoculum kindly provided by Jeff Bergelson (University of Pennsylvania). RD cells were brought to confluence in 150-mm-diameter plates at 37°C in Dulbecco minimal Eagle medium (Invitrogen) with 10% fetal bovine serum. The cells were rinsed with phosphate-buffered saline (PBS), followed by the addition of 5 ml of CVB3-RD stock inoculum diluted in Dulbecco minimal Eagle medium per dish, for a multiplicity of infection of 1 to 5. After incubation at 37° for 1 h, 7 ml of fresh medium was added per dish, and infection was allowed to progress for 48 h. Cells were harvested, pooled, and stored at −80°C.
Virus was purified by freezing and thawing the cells three times before adding NP-40 (1%). After homogenization, the preparation was centrifuged at 5,000 rpm for 10 min. MgCl2 (to 5 mM), DNase (0.05 mg ml−1), and sodium dodecyl sulfate (to 0.5%) were added to the supernatant and incubated for 30 min at room temperature. Trypsin was added (0.5 mg ml−1) and incubated for 10 min at 37°C. EDTA (10 mM) and sarcosine (1%) were added, and the pH was adjusted to neutral with NH4OH. The virus was pelleted through 30% sucrose in 50 mM MES (morpholineethanesulfonic acid) (pH 6.0) by centrifugation at 48,000 rpm for 2 h using a 50.2 Ti rotor. The pellets were resuspended in 50 mM MES, loaded onto 10 to 40% K-tartrate-50 mM MES gradients, and centrifuged at 36,000 rpm for 90 min using an SW41 rotor. The virus bands were collected, diluted in buffer, and pelleted by centrifugation at 48,000 rpm for 90 min using a 50.2 Ti rotor. Pellets were resuspended in buffer to estimate the virus concentration by measuring the absorbance at wavelengths of 260, 280, and 310 nm.
Full-length DAF molecules were incubated with CVB3-RD at room temperature for 1 h at a ratio of four DAF molecules per potential binding site on the virus (240:1). Small aliquots of this mixture were applied to carbon-coated grids and vitrified in liquid ethane. Electron micrographs were recorded on Kodak SO-163 film by using a Phillips CM300 FEG microscope. Micrographs were digitized with a Zeiss PHODIS microdensitometer at 7-μm intervals. The scans were averaged in boxes of 2 by 2 pixels. The final averaged pixel size was 3.11 Å. Approximately 3,500 particles were selected and corrected for contrast transfer function of the microscope using the program RobEM (http://cryoem.ucsd.edu/programs.shtm). The defocus distance ranged from 1.1 to 4.6 μm. The EM reconstruction processes were performed using icosahedral averaging with the programs EMPFT and EM3DR (1). The cryoEM density of native CVB3 was used as an initial starting structure. The final resolution was estimated by using maps of the reconstructed cryoEM density representing the viral capsid between radii 75 and 160 Å and then determining where the Fourier shell correlation fell below 0.5 using the CUTPIFMAP, FFT, and EMRESOL programs written by Chuan Xiao (http://bilbo.bio.purdue.edu/~viruswww/Rossmann_home/river_programs/). A similar cryoEM reconstruction procedure was used for a reconstruction of the native virus (Table (Table22).
The program EMfit (42) was used to calibrate the exact magnification of the cryoEM native CVB3 map by comparing it with a map derived from the X-ray crystallographically determined coordinates of CVB3 (Protein Data Bank [PDB] accession number 1COV). A difference map was calculated by subtracting the native map from the cryoEM map of the CVB3-RD and DAF complex. The SCR2 and -3 fragment was fitted into the cryoEM difference density and refined by using the program EMfit (42) (PDB accession codes for X ray, 1OJV, 1OJW, and 1OJY; PDB accession code for nuclear magnetic resonance, 1NWV). The fitted SCR2 and -3 fragment was then removed from the cryoEM difference density by setting to zero the density within a radius of 4.5 Å of each Cα atom, prior to the independent fitting of SCR1 and SCR4. Residues in the virus/receptor interface were identified as those in CVB3 that had any atoms less than 3.6 Å from any atom in the fitted DAF structure. The buried surface area was calculated using CCP4 programs Areaimol and Surface (Table (Table3)3) (6, 8, 23).
Monolayers of HeLa or RD cells were grown to confluence in six-well 35-mm culture dishes. Virus inoculum was diluted in PBS plus 0.1% (wt/vol) bovine serum albumin to a titer that would allow formation of 50 to 100 plaques. Soluble receptor or an equal volume of PBS plus 0.1% bovine serum albumin was added to the diluted virus and incubated for 1 h at room temperature before being applied to cells. After incubation for 45 min at room temperature to allow virus to attach to cells, the cultures were fed with medium containing 0.8% agar and incubated for 24 to 48 h. The agar overlay was removed, and cells were stained with crystal violet to visualize plaques.
Accession numbers for data bank deposition are as follows: EMD-1412, cryoEM reconstruction of CVB3 RD strain complexed with DAF; 2QZD, 2QZF, and 2QZH, structures of the ectodomain of the DAF fitted into a cryoEM reconstruction of CVB3 RD strain and DAF complex; EMD-1411, cryoEM reconstruction of CVB3 M strain.
The cryoEM reconstruction of CVB3-RD complexed with full-length DAF (Table (Table2;2; Fig. Fig.2)2) showed that the receptor does not bind into the canyon as does CAR but rather lies across the surface of the virus, crossing over the canyon to interact with the southern rim of the canyon at the “puff” region of VP2 (loop EF, residues 2129 to 2180) (Fig. (Fig.2).2). The cryoEM density corresponding to DAF was approximately equal in magnitude to that of the virion capsid, indicating that nearly all of the 60 possible DAF binding sites were fully occupied. The length and shape of the DAF density corresponded to an extended DAF molecule consisting of four SCR domains. This contrasts with the binding of DAF to the surfaces of EV7 and EV12, which showed only partial occupancy because of low affinity of receptor for virus and steric clashes at the twofold symmetry axes (4, 13, 35).
Three different crystal structures of DAF (25) generated eight different models with variations of twist and tilt between SCR1 and -2 and between SCR3 and -4. Each of these eight models of full-length DAF was fitted into the DAF difference density using all nonhydrogen atoms (see Materials and Methods for the difference map calculation). The general shape of the DAF molecule, in particular the ~30 to 40° “kink” between SCR1 and SCR2, readily determined the overall orientation. Although the DAF molecule is generally long and cylindrical in shape, the kink or elbow angle modified its shape to permit an accurate orientation of the DAF molecule about its long axis. The asymmetry of the DAF molecule determined the placement of SCR1 closer to a fivefold axis and SCR4 closer to a threefold axis (Fig. (Fig.3).3). However, none of the eight different crystal structures used in their entirety exactly filled the DAF density, indicating that the structure of DAF bound to the surface of the virus had a significantly different tilt and twist between SCR1 and -2 and between SCR3 and -4 than observed in any of the crystal structures. Consequently, a stepwise fitting and refinement operation was used, which first fitted the structure of the central part of the DAF molecule, SCR2 and -3, and subsequently fitted SCR1 and SCR4 separately into the DAF density, by using the EMfit program (40) (Table (Table4).4). The best fit of SCR2 and -3 was obtained by using the nuclear magnetic resonance structure (60). The density corresponding to SCR2 and -3 was subtracted from the map, and the full-length structure of DAF (25) was used to initiate the placement of SCR1 and SCR4 into corresponding densities. The position and orientation of SCR1 were refined while restraining the distance between the C terminus of SCR1 and the N terminus of the fitted SCR2 and -3 to be less than 8 Å. Similarly, the SCR4 position and orientation were refined while restraining the N-terminal end of SCR4 to be no more than 8 Å from the C-terminal end of SCR2 and -3. Throughout the stepwise fitting procedure, the full-length DAF structure was used to check that the individual SCRs retained the correct spatial and angular relationships to each other.
The crystal structure of CVB3 (CVB3 M strain, PDB accession code 1COV) was placed into the cryoEM density of the CVB3-RD and DAF complex by superimposing the icosahedral symmetry elements. The fitted DAF structure was also placed into the map, and atoms located on the virus surface, as well as atoms in the DAF molecule, were identified if they were within 3.6Å of the two fitted structures (Table (Table5;5; Fig. Fig.4).4). Two loops in the flexible puff region of the CVB3 virus, residues 2134 to 2145 and 2160 to 2168, clashed with the SCR2 of the fitted DAF molecule. Presumably these flexible loops alter their structure when forming the complex. Whereas the SCR1 domain of DAF has no interaction with the virus, DAF residues located in SCR2 (residues 76 to 81, 92 to 98, and 101 to 107) were within the contact region. Residues in the SCR2 and -3 junction make a few interactions with the puff, but the main part of SCR3 crosses over the canyon, without entering into the canyon and without making contact with the viral surface. The C terminus of each DAF molecule interacts with two other DAF molecules related by icosahedral threefold symmetry. The site of interaction corresponds to a Ca2+ ion, coordinated by Glu3200, as observed in the crystal structure of CVB3 (29). In the complex of CVB3-RD and DAF, this Ca2+ ion would also be associated with the carboxy-terminal His6 nickel binding tag from each of three symmetry-related DAF molecules. Thus, the DAF molecules are effectively tethered to the virus at two sites, namely, at SCR2 and at the carboxy end of SCR4, causing SCR3 to block access to the canyon (Fig. (Fig.44).
In agreement with previous affinity measurements and mutational analyses (3, 38), most of the interactions between the receptor and virus occur between SCR2 of DAF and the puff region of the virus. Although the DAF footprints onto the EV7 and EV12 surfaces are similar and involve the puff region on the south side of the canyon, they differ greatly from the footprint of DAF onto CVB3 (Fig. (Fig.55 and and6).6). Furthermore, there is little overlap between the DAF residues involved in binding CVB3 and those that bind EV12 (4) (Fig. (Fig.7,7, left). Most of the DAF residues in the CVB3 contact region are located on the opposite side of DAF than are the residues that interact with convertases, consistent with binding and immunological investigations (12, 20, 21, 53) (Fig. (Fig.7,7, right).
Previously, two residues in VP2, Val108 and Ser151, had been identified as necessary for CVB3 infection of RD cells (24). Although Ser2151 maps to the viral surface near the twofold axes of symmetry, Val2108 is located internally on the βD strand of VP2, suggesting that the RD phenotype may not be solely a function of direct receptor interaction. Binding of DAF to the Nancy-new strain of CVB3 has been observed, even though this particular strain had a Ser at position 2151 (RD genotype) and an Asp at position 2108 (wild-type genotype) (51). Furthermore, the amino acid differences reported between Nancy, Nancy-new, and Nancy-RD indicate that there might be some uncertainty as to a common background between different laboratory strains of CVB3 and CVB3-RD, making it difficult to predict which amino acid differences confer DAF binding ability. In contrast to the situation for EV7 and EV12, the structure presented here shows no interaction between DAF and the 2151 surface residue thought to be responsible for the RD phenotype. Possibly the binding site of DAF on CVB3-RD has greater affinity to DAF than the binding site observed in the cryoEM studies of echoviruses. This is supported by the much higher occupancy of DAF on CVB3-RD compared to the cryoEM studies of DAF binding to EV7 and EV12. As the two receptor binding sites are mutually exclusive, it would be expected that the site reported here, once established, would be dominant.
Plaque reduction assays were performed with the uncomplexed wild-type CVB3 (11) and the mutant CVB3-RD, as well as with viruses preincubated with either soluble CAR or DAF (Fig. (Fig.8).8). Reduction of plaques occurred only when wild-type virus was preincubated with soluble CAR and applied to HeLa cells or when mutant virus was preincubated with soluble DAF and applied to RD cells (>90% plaque reduction). In all other combinations, plaque reduction was insignificant. These results can be interpreted by noting that HeLa but not RD cells express CAR, that both types of cells express DAF, and that the structural results show that DAF blocks access to the canyon (Fig. (Fig.4).4). In order to further explain the plaque assay results, it is also necessary to make the following assumptions: adaptation by the mutant virus to bind DAF allows DAF to compete successfully with CAR, and either (i) a canyon-binding receptor molecule displaces the pocket factor or (ii) DAF binding alone is sufficient for the virus to enter and infect cells.
Viral plaque reduction assays indicated that wild-type virus was able to infect HeLa cells but not RD cells (Fig. (Fig.8).8). This is consistent with wild-type virus being able to bind CAR but not DAF. Preincubation of wild-type virus with soluble CAR reduced plaques on HeLa cells, presumably by blocking the CAR binding site. The mutant virus was capable of infecting both HeLa and RD cells. However, compared to wild type, the mutant plaques in HeLa cells were smaller and fewer, suggesting either the mutant virus no longer binds CAR as well as the wild type or that the mutant virus is entering cells in a different, less efficient way. Furthermore, the gained ability of the mutant virus to bind to DAF may be demonstrated by plaque formation in RD cells, although it is unclear whether the virus uses only DAF or an unknown canyon-binding receptor to enter RD cells. Preincubation of the mutant virus with DAF inhibits plaque formation on RD cells but not that on HeLa cells. Either the mutant virus relies on an interaction with DAF to infect RD cells or access to the canyon for a hypothetical unknown canyon-binding receptor was blocked by bound DAF. However, it is possible that CAR molecules on HeLa cells successfully competed with the DAF bound to the surface of the mutant virus, allowing infection. Incubating the mutant virus with CAR caused no reduction in plaque formation on either HeLa cells or RD cells. This lack of inhibition suggests that preincubation of the mutant virus with CAR could have expelled the pocket factor, making the virus susceptible for uncoating. Subsequently, the mutant virus would be able to infect HeLa or RD cells by interacting with DAF.
The above assumption (i) that there might be another molecule on the surface of RD cells that can act as receptor for the mutant CVB3-RD virus would be consistent with the hypothesis that the pocket factor needs to be displaced by a canyon-binding molecule, such as CAR, to initiate uncoating of the virus. It is also consistent with observations that indicate that when CVB3 adapts to use DAF, it retains the ability to bind CAR and remains dependent on a CAR interaction for uncoating (12, 36, 47) and that DAF alone is insufficient to initiate infection (28). Furthermore, CVB3 can bind to at least five different cellular receptors distinct from either DAF or CAR on the surface of human cardiac cells (33). However, the alternative assumption (ii), namely, that DAF alone is sufficient for a successful infection, without the assistance of a canyon-binding molecule to displace the pocket factor, cannot be ruled out. For instance, the minor-group rhinoviruses, which use very-low-density lipoproteins for attachment and infection (16, 61), bind the receptor on an exposed surface outside the canyon (15). Similarly, it has been shown that a DAF-binding strain of echovirus can enter cells by using DAF as a receptor, apparently without forming A particles (55).
Closely related viruses bind different surfaces of DAF to different locations on the virus, suggesting a convergent evolution in the use of DAF that offers some advantage to the virion. Such an advantage may be due to the ubiquitous nature of DAF, which is found on most cell surfaces, including serum-exposed cells that line the lumen through which enteroviruses travel. For instance, DAF is necessary to enter polarized epithelial cells (3), because the initial binding of CVB3 to DAF sets off a cell-signaling cascade which causes transport of the bound virus to the tight junction between cells and allows interaction with CAR molecules that are sequestered there. Similarly, coxsackievirus binding to DAF expressed on pancreatic cells causes recruitment of CAR to the site of the interaction (58), allowing subsequent binding of the CAR and successful infection of the cells.
We thank Sheryl Kelly and Cheryl Towell for help in preparation of the manuscript and Rodney McPhail for assistance with Fig. Fig.11.
Figures Figures44 and and77 were produced using the UCSF Chimera package from the Computer Graphics Laboratory, University of California, San Francisco (supported by NIH grant P41 RR-01081). The work was supported by National Institutes of Health grants to M.G.R. (AI 11219), to D.E.M. (AI 23598), and to F.L. (HL 077319). S.H. was supported by a National Institutes of Health postdoctoral fellowship (AI 060155).
Published ahead of print on 5 September 2007.