Foot-and-Mouth Disease Virus Virulent for Cattle Utilizes the Integrin α_vβ₃ as Its Receptor

Abstract

Adsorption and plaque formation of foot-and-mouth disease virus (FMDV) serotype A₁₂ are inhibited by antibodies to the integrin α_vβ₃ (A. Berinstein et al., J. Virol. 69:2664–2666, 1995). A human cell line, K562, which does not normally express α_vβ₃ cannot replicate this serotype unless cells are transfected with cDNAs encoding this integrin (K562-α_vβ₃ cells). In contrast, we found that a tissue culture-propagated FMDV, type O₁BFS, was able to replicate in nontransfected K562 cells, and replication was not inhibited by antibodies to the endogenously expressed integrin α₅β₁. A recent report indicating that cell surface heparan sulfate (HS) was required for efficient infection of type O₁ (T. Jackson et al., J. Virol. 70:5282–5287, 1996) led us to examine the role of HS and α_vβ₃ in FMDV infection. We transfected normal CHO cells, which express HS but not α_vβ₃, and two HS-deficient CHO cell lines with cDNAs encoding human α_vβ₃, producing a panel of cells that expressed one or both receptors. In these cells, type A₁₂ replication was dependent on expression of α_vβ₃, whereas type O₁BFS replicated to high titer in normal CHO cells but could not replicate in HS-deficient cells even when they expressed α_vβ₃. We have also analyzed two genetically engineered variants of type O₁Campos, vCRM4, which has greatly reduced virulence in cattle and can bind to heparin-Sepharose columns, and vCRM8, which is highly virulent in cattle and cannot bind to heparin-Sepharose. vCRM4 replicated in wild-type K562 cells and normal, nontransfected CHO (HS⁺ α_vβ_3⁻) cells, whereas vCRM8 replicated only in K562 and CHO cells transfected with α_vβ₃ cDNAs. A similar result was also obtained in assays using a vCRM4 virus with an engineered RGD→KGE mutation. These results indicate that virulent FMDV utilizes the α_vβ₃ integrin as a primary receptor for infection and that adaptation of type O₁ virus to cell culture results in the ability of the virus to utilize HS as a receptor and a concomitant loss of virulence.

Many viruses initiate infection by attaching to cell surface molecules which are normal components of the plasma membrane. The molecules which viruses use are diverse, and closely related viruses can use different receptors (74, 88). A well-characterized example of the use of multiple receptors by structurally similar viruses can be found in the Picornaviridae. Within this family, the polioviruses (63) and the major group of human rhinoviruses (40, 80, 82) utilize receptors which are members of the immunoglobulin superfamily, while the minor group of human rhinoviruses utilize the low-density lipoprotein receptor (42). Different variants of encephalomyocarditis virus have been reported to use either the immunoglobulin-like molecule vascular cell adhesion molecule 1 (43) or a 70-kDa cell surface sialoglycoprotein (48) as a receptor, and hepatitis A virus (HAV) has recently been shown to use a unique membrane glycoprotein containing immunoglobulin and mucin domains (3, 49). The coxsackie B viruses have been reported to use either decay-accelerating factor (11, 77), a 100-kDa nucleolin-related membrane protein (33, 52), or a 46-kDa membrane-bound immunoglobulin-like protein (9, 24, 83) either as an initial cell binding protein or in combinations that form functional receptor complexes. Interestingly, this 46-kDa protein has also been shown to be a receptor for two members of the human adenovirus family (9, 83), demonstrating that viruses from different families, which do not share common structural features, can use the same receptor. A coxsackie A virus (type A21) appears to require both decay-accelerating factor and intercellular adhesion molecule 1 (ICAM-1) for productive infection (78).

The picornaviruses echoviruses 1, 8, 9, and 22 (12, 13, 68, 91) and coxsackievirus A9 (CAV9) (73) use cell surface integrins as receptors, although it has been suggested that echovirus 9 and CAV9 can also utilize nonintegrin receptors to infect cells (72, 91). Integrins have also been implicated as receptors for rotaviruses (28) and papillomaviruses (36). They have been shown to play a role in the internalization of some human adenoviruses (4, 86) and in the binding and internalization of a number of other nonviral pathogens (45).

Foot-and-mouth disease virus (FMDV) is a member of the aphthovirus genus of the Picornaviridae. The three-dimensional structure of the virus has revealed a prominant surface protrusion made up of a loop between the βG and βH strands of the capsid protein VP1 (G-H loop [1, 59]). This loop contains major immunodominant epitopes of the virion (22) and a highly conserved sequence, arginine-glycine-aspartic acid (RGD), which has been shown to be essential for virus interactions with its cellular receptor, both by RGD peptide inhibition of virus adsorption (7, 38) and by direct mutation or deletion of the sequences encoding these amino acid residues from infectious cDNA clones of FMDV (55, 61, 62). These data, along with the well-characterized interaction of extracellular matrix proteins containing the RGD sequence with integrin receptors (44), suggested that FMDV could use an integrin to attach to cells.

We have previously shown that FMDV type A₁₂ can compete for cellular receptors with CAV9, which bind to the integrin α_vβ₃ (73), and also demonstrated that antibodies to α_vβ₃ inhibited binding and plaque formation of type A₁₂ (14). Recently, Jackson et al. (47) reported that FMDV type O₁ utilizes cell surface heparan sulfate (HS), a ubiquitous glycosaminoglycan (GAG), as a coreceptor for viral infection. HS has been shown to mediate attachment of several herpesviruses to susceptible cells (65, 67, 79, 89), although a cofactor may be required for viral entry (66). HS has also been suggested to be an attachment molecule for respiratory syncytial virus (53) and has been used as an unnatural target receptor for engineered adenovirus to increase the efficiency of gene delivery (87). We have suggested, however, that binding of FMDV type O viruses to cell surface HS is a consequence of tissue culture adaptation (75).

In this work, we show that cells which are not normally susceptible to FMDV type A₁₂ can be infected with this serotype if they are transfected with cDNAs encoding human α_vβ₃. In addition, we show that a tissue culture-adapted type O₁ virus which is attenuated for disease in cattle replicates in the presence of cell-surface HS, independent of α_vβ₃ expression. Finally, we show that a type O₁-derived virus which is highly virulent in cattle requires the α_vβ₃ integrin for infection in cell culture. These data indicate that the integrin, rather than HS, is most likely the receptor involved in viral replication and pathogenesis in the livestock host.

MATERIALS AND METHODS

Viruses.

FMDV type A₁₂, strain 119ab, was derived from the infectious cDNA clone pRMC₃₅ (71). The cDNA was derived from virus which had an unknown high-passage history in bovine kidney and BHK-21 cells, and following recovery from transfected BHK-21 cells, the virus used in these studies was passaged at least five times in this cell line. An antigenic variant of type A₁₂, harboring VP1 sequences present in a bovine tongue tissue-propagated type A₁₂ (vRM-SSP), has also been described (69). This virus was initially derived by transfection of cDNA-transcribed RNA into BHK-21 cells (69) and amplified by passage into CHO cells expressing a chimeric single-chain antibody–ICAM-1 receptor (CHO-11.1 [70]). Two variants derived from infectious cDNAs containing capsid sequences represented in a seed stock of type O₁Campos have been described elsewhere (75). These viruses are designated vCRM8, a highly cattle-virulent virus which is unable to bind to heparin-Sepharose, and vCRM4, a cattle-avirulent virus which binds to heparin-Sepharose (75). A derivative of vCRM4 virus, vCRM48 (75), with an RGD→KGE mutation in the G-H loop of VP1 (vCRM48-KGE) was derived by site-directed mutation of the vCRM48 cDNA by using methods similar to those used for the derivation of a type A₁₂ KGE mutant (61). Virus derived from BHK-21 cells, transfected with the cDNA-transcribed RNAs of these three viruses, was used to prepare large stocks of virus used in these studies following three or four passages in BHK-21 cells. A tissue culture-propagated type O₁BFS from the Animal Virus Research Institute, Pirbright, United Kingdom, was supplied by Fred Brown (Plum Island Animal Disease Center). Prior to coming to Plum Island, the virus was serially passaged 14 times in BHK-21 cells, plaque purified, and passaged once in BHK-21 cells, followed by a single passage in a swine kidney cell line (IBRS2). After arriving at Plum Island, the virus was plaque purified twice followed by at least five serial passages in BHK-21 cells.

Cells.

BHK-21 cells were maintained on minimum essential medium (MEM) containing 10% calf sera and 10% tryptose phosphate broth. The human erythroleukemia cell line K562 transfected with α_vβ₃ cDNAs (K562-α_vβ₃) as well as cells transfected with the parental plasmid (K562-pRSV) have been described elsewhere (16, 17). These cells were maintained in RPMI medium containing 10% fetal calf serum (FCS), an additional 2 mM l-glutamine, gentamicin (10 μg/ml), amphotericin B (Fungizone; 12.5 μg/ml), and the neomycin derivative G-418 (1,200 μg/ml; Life Technologies, Gaithersburg, Md.). CHO-K1 cells and the GAG-deficient mutants pgsA-745 (35) and pgsD-677 (57) were cultured in Ham’s F-12 medium supplemented with 10% FCS.

Derivation of CHO cells expressing human α_vβ₃.

The human α_v cDNA subcloned into pcDM8 (Invitrogen, Carlsbad, Calif.) (17) was used as a template for 10 rounds of PCR amplification with specific dUMP-containing human α_v primers, and the product was annealed to pAMP1 (Life Technologies). The resulting human α_v-encoding cDNA was then subcloned into pcDNA3.1/Zeo(−) (Invitrogen), which contains a Zeocin resistance marker, using the restriction enzymes NotI and EcoRI. The human β₃ cDNA cloned into plasmid pIAP58 (58), containing a neomycin resistance marker, has been described elsewhere (17).

CHO-K1 cells and the two GAG-deficient mutant cell lines were cotransfected with α_v/pcDNA3.1/Zeo(−) and the β₃/pIAP58 clone, using Lipofectin (Life Technologies). Twenty-four hours after transfection, G-418 (600 μg/ml) and Zeocin (500 μg/ml; Invitrogen) were added to the growth medium. Transfected cells expressing human α_vβ₃ were enriched from the antibiotic-resistant population by a combination of single-cell cloning and fluorescence-activated cell sorting (FACS), using the α_vβ₃-specific monoclonal antibody (MAb) LM609 (26). Transfectants were sorted based on their fluorescence intensity, and the 10% highest-staining cells were gated, collected under sterile conditions, and expanded in culture. Periodically, cells were reanalyzed by FACS and resorted if the level of cells expressing the integrin had significantly changed.

Viral replication assays.

K-562 cells, at a concentration of 2 × 10⁶ to 3 × 10⁶ cells/ml, were infected with various FMDV serotypes at multiplicities of infection (MOIs) indicated in the figure legends. Virus was allowed to adsorb by rotating the cells at 37°C for 1 h. Cells were then washed with MEM with 1/20 the normal amount of methionine, 25 mM HEPES (pH 7.5), and 0.5% FCS and resuspended to 10⁶ cells/ml in the same medium. One milliliter of cells per well was added to a 12-well tissue culture dish; at various times after infection (as indicated in the figure legends), [³⁵S]methionine (50 to 75 μCi) was added to each well, and the cells were incubated overnight at 37°C. Cells were lysed in 1% Triton X-100, and cell debris was removed by centrifugation in a microcentrifuge for 1 min. Trichloroacetic acid (TCA)-precipitable counts per minute were determined, and radioimmunoprecipitation (RIP) was performed as described previously (5), using a bovine hyperimmune anti-FMDV serum. Equal amounts of TCA-precipitable counts per minute and protein, prepared by the addition of an unlabeled uninfected cell extract, were immunoprecipitated and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using a 12% polyacrylamide gel. CHO cells were infected in a similar manner except that they were grown in 24-well tissue culture dishes at a concentration of 1.5 × 10⁵ to 2.0 × 10⁵ cells per well, and viral adsorption was for 2 h.

RESULTS

Replication of FMDV in transfected and nontransfected K562 cells.

K562-pRSV and K562-α_vβ₃ cells were infected with A₁₂, O₁BFS, vCRM4, and vCRM8 at an MOI of 10 PFU/cell and vRM-SSP at an MOI of 1 PFU/cell. Cells were labeled overnight, and lysates were analyzed by RIP and SDS-PAGE as described in Materials and Methods. The results in Fig. 1a show the presence of viral proteins only in A₁₂- or vCRM8-infected α_vβ₃-expressing K562 cells, indicating that these two viruses require α_vβ₃ to initiate infection. In contrast, types O₁BFS and vCRM4 were able to infect both K562-pRSV and K562-α_vβ₃ cells. The results in Fig. 1b show that the vRM-SSP variant, which binds to BHK-21 cells more poorly than the prototype tissue culture-adapted type A₁₂ (69), grew only in cells transfected with α_vβ₃ cDNAs. These results correlate with those obtained by measuring plaque titer at 24 h after infection of K562 cells by plaque assay in BHK-21 cells (not shown). They also correlate with double-immunofluorescence analysis of infected K562 cells probed with MAb LM609 and a bovine polyclonal anti-FMDV serum (not shown) and confirm our previous findings that antibodies to this integrin inhibit virus adsorption and plaque formation for type A₁₂ (14). In addition, these results indicate that both animal-propagated and tissue culture-adapted type A₁₂ can utilize α_vβ₃ in cell culture. Furthermore, replication of O₁BFS and vCRM4 was not dependent on the presence of this integrin on K562 cells. The only integrin endogenously expressed on the surface of K562 cells is α₅β₁ (17), which is also dependent on the RGD sequence for ligand binding (44). We were unable to inhibit binding or replication of type O₁BFS or vCRM4 in K562 cells with rabbit anti-α₅β₁ serum (AB1905; 1:10; Chemicon International, Temecula, Calif.), purified rabbit anti-α₅ immunoglobulin G (IgG) (AB 1928; 10 μg/ml; Chemicon International), purified anti-β₁ MAb P5D2 (20 μg/ml) (17), and purified anti-α₅ MAb 16 (20 μg/ml) (17). Taken together, these results suggest that the two tissue culture-adapted type O₁ viruses use a nonintegrin receptor to infect K562 cells.

FIG. 1.

Analysis of viral proteins synthesized in FMDV-infected K562 cells transfected with α_vβ₃ cDNAs. K562-pRSV (C) or K562-α_vβ₃ (T) cells were infected with type A₁₂, O₁BFS, vCRM4, or vCRM8 at an MOI of 10 PFU/cell (a) or type A₁₂ or vRM-SSP at an MOI of 1 PFU/cell (b). Cells were labeled between 3 and 24 h after infection with [³⁵S]methionine, and extracts were analyzed by RIP and SDS-PAGE as described in Materials and Methods. Viral proteins synthesized in infected and labeled BHK-21 cells are included as markers (M), and the positions of major viral proteins are indicated on the left. Type A₁₂ was used as the marker for panel b.

Viral replication in CHO cells transfected with α_vβ₃ cDNAs and deficient in HS synthesis.

Jackson et al. (47) reported that O₁BFS had enhanced binding and infectivity in cells expressing surface HS, by showing that this virus was not able to plaque on mutant CHO cells which were deficient in HS synthesis. We recently showed that O₁BFS and vCRM4, but not vCRM8, were able to bind to heparin-Sepharose columns, probably as a result of basic amino acid residues present at the VP2/VP3 interface (75).

To further examine the relationship between the α_vβ₃ receptor and cell surface HS, we engineered wild-type and HS-deficient CHO cells to express human α_vβ₃. The CHO mutant, pgsA-745, synthesizes little, if any, GAG (35), while mutant pgsD-677 is unable to synthesize HS but expresses normal amounts of chondroitin sulfate (57). Figure 2 shows FACS analyses of cells transfected with human α_vβ₃ cDNAs two to three passages after single-cell cloning and FACS using the α_vβ₃-specific MAb LM609. These analyses show that after only a few passages, a portion of the wild-type CHO-K1 cells no longer expressed α_vβ₃, whereas a very high proportion of the transfected pgsA-745 and pgsD-677 cells maintained expression, although the pgsD-677 cells exhibited higher levels of expression than pgsA-745 cells.

FIG. 2.

Analysis of CHO cells transfected with α_vβ₃ cDNAs. CHO-K1 cells and the two GAG-deficient mutants pgsA-745 and pgsD-677 were transfected with α_vβ₃ cDNAs and selected as described in Materials and Methods. After two to three passages, transfected and nontransfected cells were incubated with MAb LM609 for 30 min at 4°C in phosphate-buffered saline, washed, and incubated with fluorescein isothiocyanate (FITC)-labeled goat anti-mouse IgG for an additional 30 min at 4°C. After being washed with phosphate-buffered saline, cells were analyzed on a Becton Dickson FACSCaliber analyzer. Nontransfected cells are represented by open curves, and transfected cells are represented by closed curves.

Using this panel of cell lines with defined HS and α_vβ₃ expression, we tested the abilities of different viruses to utilize these two receptors by analysis of the proteins synthesized in infected cells (Fig. 3). Type A₁₂ was unable to replicate in either wild-type or mutant CHO-K1 cells. When these cells were transfected with α_vβ₃ cDNAs, however, viral proteins were synthesized, indicating that replication of this virus was dependent on α_vβ₃ expression. In contrast, both O₁BFS and vCRM4 were able to replicate in CHO-K1 cells, regardless of whether they were expressing α_vβ₃. In both GAG-deficient CHO mutants infected with O₁BFS, however, no virus-specific proteins were observed in either control or α_vβ₃-transfected cells. In the CHO mutants infected with the vCRM4 virus, some viral replication was detected in both control and α_vβ₃-transfected pgsA-745 cells, although no replication was evident in the pgsD-677 cells (Fig. 3). Since equal amounts of TCA-precipitable counts per minute were used for the immunoprecipitation, comparison of the intensity of the viral proteins in the pgsA-745 cells to the intensity of those present in the CHO-K1 cells indicates that the level of viral replication was low in the mutant cells. FACS analysis of the wild-type and mutant CHO cells with an anti-α₅β₁ MAb showed that these cells express this integrin to comparable levels on their surface (not shown). To rule out the possibility that vCRM4 used another integrin, we engineered an RGD→KGE mutation in the G-H loop of a derivative virus (vCRM48) [75]). The vCRM48-KGE replicated as well in BHK-21 cells as the virus containing the RGD sequence in the G-H loop (not shown). This contrasted with the inability of a type A₁₂ virus, with similar engineered mutations, to either adsorb to or replicate in BHK-21 cells (61). Infection of CHO cells with vCRM48-KGE resulted in high levels of viral replication in the CHO-K1 control and transfected cells, a much lower level of replication in the pgsA-745 cells, and no replication in the pgsD-677 cells (Fig. 3). These results were indistinguishable from those seen with vCRM4. Treatment of the control and transfected CHO cells with 20 U of heparinase III (47) per ml resulted in an almost total inhibition of viral replication by vCRM4, vCRM48-KGE, and O₁BFS, and treatment of vCRM4 with soluble heparin (800 μg/ml) resulted in an almost total inhibition of replication in pgsA-745 cells (not shown). Thus, it appears that the vCRM4 virus and our strain of O₁BFS utilize HS, and not an integrin, as their primary receptor. In addition, these results indicate that the pgsA-745 cells may express low levels of HS on their surface, permitting vCRM4 and vCRM48-KGE replication. At this time it is not clear why type O₁BFS did not exhibit the low-level replication observed for vCRM4 in the pgsA-745 cells.

FIG. 3.

Analysis of viral proteins synthesized in FMDV-infected CHO cells transfected with α_vβ₃ cDNAs. Transfected (T) or nontransfected (C) CHO-K1, pgsA-745, or pgsD-677 cells were infected with FMDV type A₁₂, O₁BFS, vCRM4, or vCRM48-KGE at an MOI of 10 PFU/cell. Cells were labeled between 3 and 24 h after infection with [³⁵S]methionine, and extracts were analyzed by RIP and SDS-PAGE as described in Materials and Methods. Viral proteins synthesized in infected and labeled BHK-21 cells are included as markers (M), and the positions of major viral proteins are indicated on the left.

When a similar experiment was performed with CHO cells infected with either vCRM8 or vRM-SSP, no viral proteins were observed after 24 h of infection (not shown). Since replication of these viruses in K562 cells was dependent on transfection with α_vβ₃ cDNAs (Fig. 1), and both viruses grow poorly in BHK-21 cells, we allowed infection of the CHO cell cultures to proceed for another 24 h and labeled the cells between 24 and 48 h after infection. The results in Fig. 4 show that viral proteins were observed only in α_vβ₃-transfected pgsD-677 cells infected with either virus. The level of viral protein synthesis in the vCRM8-infected pgsD-677-α_vβ₃ cells was very low, probably reflecting the poor tissue culture adaptation of this virus. With both viruses, however, no replication was evident in either control or transfected CHO-K1 or pgsA-745 cells. This is probably a result of the poor α_vβ₃ expression in these cells (Fig. 2) coupled with poor viral replication in tissue culture. Interestingly, a similar 24- to 48-h labeling of O₁BFS-infected pgsA-745 and pgsD-677 cells failed to show any evidence of viral replication of this virus in the GAG-deficient cells (Fig. 4).

FIG. 4.

Analysis of viral proteins synthesized in FMDV-infected CHO cells transfected with α_vβ₃ cDNAs. Transfected (T) or nontransfected (C) CHO-K1, pgsA-745, or pgsD-677 cells were infected with FMDV type vCRM8 or O₁BFS at an MOI of 10 PFU/cell (a) or vRM-SSP at an MOI of 1 PFU/cell (b). Cells were labeled between 24 and 48 h after infection with [³⁵S]methionine, and extracts were analyzed by RIP and SDS-PAGE as described in Materials and Methods. Viral proteins synthesized in infected and labeled BHK-21 cells are included as markers (M), and the positions of major viral proteins are indicated on the left.

These data from the CHO cell studies confirm those obtained with K562 cells (Fig. 1), showing that both vCRM8 and vRM-SSP utilize α_vβ₃ as a receptor and can replicate independently of the presence of HS on the cell surface and that replication of O₁BFS and vCRM4 is dependent on the presence of HS.

Heparin neutralization of O₁Campos variants.

To further delineate the use of HS by vCRM4 and vCRM8, we examined whether soluble heparin was able to neutralize viral infectivity. Decreasing concentrations of heparin were incubated with a fixed number of PFU of either vCRM4, vCRM48-KGE, or vCRM8 and inoculated onto BHK-21 cells. The results in Fig. 5 show that heparin was able to neutralize both vCRM4 and vCRM48-KGE in a dose-dependent manner, indicating that this soluble GAG bound to the viral particles and could prevent viral replication either by direct inhibition of viral adsorption or by viral aggregation. Heparin, however, had no effect on the replication of vCRM8, a result consistent with the ability of this virus to grow on GAG-deficient CHO cells expressing α_vβ₃.

FIG. 5.

Heparin neutralization of FMDV. Different concentrations of heparin (Sigma) were mixed with 35 to 100 PFU of the indicated viruses in basal medium Eagle with 25 mM HEPES (pH 7.5) and 0.05% bovine serum albumin and incubated for 20 min at room temperature. Aliquots (200 μl) were adsorbed to BHK-21 cells for 1 h at 37°C. The inoculum was aspirated, and the plates overlaid with a mixture of MEM and 0.6% gum tragacanth and incubated at 37°C. Plates were stained with crystal violet-formalin at 48 h (vCRM8) or 72 h (vCRM4 and vCRM48-KGE). Results are expressed as plaque numbers relative to the number of plaques appearing in virus incubated under the same conditions without heparin.

DISCUSSION

Using reciprocal cross-competition binding assays (6, 76), we previously reported that representatives of six serotypes of FMDV use a common, low-copy-number receptor to bind to cells in culture. However these same studies suggested that some serotypes, including type O₁, could use alternative receptors which were present in high abundance on cultured cells. Our current experimental approach, using cell types with defined receptors, show that A₁₂ has an absolute requirement for the expression of α_vβ₃ for the initiation of infection in cell culture, while O₁BFS initiates infection following binding to the ubiquitous cell surface molecule HS. These results expand our previous study showing that antibodies to α_vβ₃ inhibit virus adsorption and plaque formation of type A₁₂ (14) and are consistent with results of a more recent study, suggesting that type O₁BFS utilizes HS as a coreceptor (47). However, our studies suggest that tissue culture-propagated O₁BFS may be incapable of utilizing the α_vβ₃ integrin as a receptor to initiate infection and is dependent on HS as its primary receptor in cell culture.

Detailed analysis of cell lines expressing HS in the presence and absence of α_vβ₃ demonstrated that A₁₂ does not require HS to infect cells and that the use of HS by O₁ is exhibited only by viruses that have been propagated extensively in tissue culture. This latter conclusion is based on analyses of two variants of type O₁Campos which differ in plaque phenotype, virulence in cattle, and the ability to bind to heparin-Sepharose (75). We have shown that the small-plaque variant (vCRM4), which is avirulent and binds to heparin (75), cannot replicate in HS-deficient cells, even when presented with α_vβ₃, a result identical to that seen with our tissue culture-propagated strain of O₁BFS (Fig. 3). In contrast, the large-plaque variant (vCRM8), which is highly virulent in cattle and cannot bind to heparin (75), replicates only in cells which express α_vβ₃, independent of the expression of cell-surface HS, a result identical to that seen with type A₁₂ (Fig. 1 and 3). We previously demonstrated that tissue culture growth properties and bovine virulence directly correlated with specific amino acids present at the VP2/VP3 interface of type O₁ virus (75). Specifically, we showed that the presence of positively charged residues at positions 134 of VP2 and 56 of VP3 allowed viruses to replicate more efficiently in BHK-21 cells and extended their host range in vitro to CHO cells (75). In contrast, virus present in lesion fluid of infected bovines contained less positively charged residues at either position, could not bind to heparin, and could not infect CHO cells (75). Furthermore, it had also been shown that upon infection of cattle by vCRM4, virus isolated from lesion fluid resembled vCRM8 in plaque phenotype, sequence at the VP2/VP3 interface, and heparin binding (75). The results presented in this report extend the characterization of these variants to their receptor specificities, clearly demonstrating that bovine virulence is inversely related to ability to utilize HS on cells in culture.

To further show that tissue culture adaptation of type O₁ results in viruses that do not need to utilize α_vβ₃ as a receptor, we have engineered a mutation in a vCRM4-related virus. The resulting virus, vCRM48-KGE, adsorbed and replicated as well as the RGD virus in BHK-21 cells (not shown) and replicated in HS-expressing CHO cells (Fig. 3). It was not able to replicate in cells which did not express HS (Fig. 3). This result contrasted with our previous results for mutants within the RGD region of type A₁₂, which were unable to bind or replicate in BHK-21 cells (61) or infect cattle (62). We have also engineered a similar RGD→KGE mutation in the vCRM8 virus, and that mutant was unable to bind or replicate in BHK-21 cells, although an infectious revertant was shown to have a KGD sequence (not shown). The results with the vCRM48-KGE virus are similar to those recently reported for types O₁Kaufbeuren (O₁K) (55) and type C₁ (60). In the case of O₁K, an RGD→RGE mutation resulted in the production of infectious virus in BHK-21 cells (55). The cDNA used to engineer the mutation was derived from a virus which underwent at least 10 passages in BHK-21 cells (90), and this virus may have acquired the ability to bind to HS. With type C₁, viruses with RGD→RED and RGD→RGG mutations were isolated as MAb escape mutants (60). The virus from which these mutants were isolated, however, was serially passaged 100 times in BHK-21 cells and had acquired three additional mutations to positively charged residues clustered on the surface in VP1 and VP3 that were not related to MAb escape (60). Thus, based on the results presented with our KGE mutant, we believe that the O₁K and C₁ viruses with RGD mutations use HS as a receptor and bypass the integrin as well.

Taken together, the existing data suggest that type A₁₂ virus adapts to tissue culture by a different mechanism than types O₁ and C₁. Specifically, type A₁₂ acquires mutations near the RGD sequence, presumably allowing a tighter binding to integrins available on cultured cells (69). We do not know whether the data for type A₁₂ can be extended to other A subtypes, but in preliminary studies, we have not been able to adapt type A₂₄ to use HS as a receptor by serial blind passage in CHO cells (not shown). In the case of type A₂₂, mutations selected during adaptation of virus from monolayer to suspension cultures could alter conformation of the G-H loop. Specifically, the monolayer-adapted virus bound more poorly to tissue culture cells than the suspension-adapted virus (18). In those viruses, there was a single mutation in VP2 lying near the VP1 G-H loop which may alter the favored conformation of the loop (29). Type O₁ viruses, however, appear to adapt to cell culture by another mechanism. Structural data comparing a low-passage type O₁K virus, a MAb-resistant mutant of this virus, and a high-passage O₁BFS isolate showed that the structure of the G-H loop of VP1 was virtually identical for the three viruses and that the main structural differences resided around residue 56 in VP3 (54), which we have previously identified as conferring the vCRM4 and vCRM8 phenotypes (75), including the ability to utilize HS as a receptor. Sequence analysis of the O₁BFS virus used in this study also showed that residue 56 in VP3 is an arginine (not shown), as in vCRM4 (75). Assays to measure the adsorption of radioactive O₁BFS, vCRM4, and vCRM48-KGE to transfected pgsD-677 cells (8) were unable to detect specific binding due to the low level of adsorption; however, we had shown that unlabeled vCRM4 could inhibit the binding of vCRM8 to BHK-21 cells (75) and that both O₁BFS and vCRM4 could inhibit the binding of type A₁₂ to BHK-21 cells (not shown), suggesting that HS-requiring viruses can bind to α_vβ₃. In addition, it has been recently shown that representatives of six FMDV serotypes, including O₁BFS, are able to bind to purified isolated α_vβ₃ in vitro (46).

Our findings that viruses which have acquired the ability to utilize HS as a receptor are attenuated in bovines suggest that ability to utilize an integrin as a receptor in vivo is critical for production of this disease. With echovirus 9, an insertion in VP1 containing the RGD sequence appears to be critical for virulence of this virus in mice (91). Furthermore, additional positively charged amino acid residues in the envelope glycoprotein of tissue culture-adapted strains of the alphavirus Sindbis virus resulted in a virus which bound to HS and had an attenuated phenotype in mice (51).

It has been previously suggested that the integrin α₅β₁ might be a receptor for FMDV (2, 25, 85). K562 cells (17) and CHO-K1, pgsA-745, and pgsD-677 cells (not shown) express α₅β₁ endogenously. Since type A₁₂ viruses expressing bovine-derived and tissue culture-propagated RGD-containing loops, and the bovine-virulent type O₁Campos (vCRM8), cannot replicate in these four cell lines in the absence of transfected α_vβ₃ cDNAs (Fig. 1, 3, and 4), it appears that they cannot utilize α₅β₁ as a receptor. Although antibodies to α₅β₁ failed to inhibit replication by O₁BFS and vCRM4 (not shown), we cannot rule out that this integrin may be involved in HS-dependent replication of these two viruses. Recently, both subunits of the α₅β₁ integrin have been shown to contain covalently linked sulfate groups which are part of both HS and chondroitin sulfate, leading to the suggestion that this integrin may be a hybrid proteoglycan (84). It had also been shown that interaction of both the α_vβ₃ and α₅β₁ integrins to the core protein of the membrane proteoglycan, perlecan, is partially mediated by the RGD sequence of the protein and modulated by HS (41).

Based on the data showing that α_vβ₃ is required for tissue culture growth of cattle-virulent viruses, it is interesting to examine the distribution of this integrin in relation to the pathogenesis of foot-and-mouth disease. A number of studies have suggested that lung and pharanyngeal areas are sites of initial viral replication (20, 23, 81), with rapid dissemination of the virus to oral and pedal epithelial areas, possibly mediated by cells of monocyte/macrophage origin (20). Recently, a study was performed with bovines experimentally infected with the prototype tissue culture-adapted type A₁₂ via aerosol (21). By using in situ hybridization, it was shown that within the first 24 h, virus was present in respiratory bronchiolar epithelium, subepithelium, and interstitial areas of the lung. By 72 h, signal was detected in epithelial cells of the tongue, soft palate, feet, tonsils, and tracheobronchial lymph nodes. While, to our knowledge, there are no studies reporting the distribution of α_vβ₃ within the bovine respiratory tract, this integrin is generally found expressed on vascular endothelium and on smooth muscle cells (19, 37, 56). In the human lung, α_vβ₃ appears to be restricted to large-vessel endothelium (30) and was not detected in bronchial epithelial cells (64). The integrin is present, however, in multiple salivary gland cells from a number of different species (32). We have performed a preliminary survey for the presence of α_vβ₃ mRNA by reverse transcription-PCR in tissues susceptible to FMDV removed from bovines at necropsy. These studies showed that mRNA was present in these tissues (not shown), but we have not yet demonstrated protein expression in the relevant cell types. FMDV has been reported to replicate in vitro in bovine keratinocytes (31). We have also detected virus in keratinocytes within lesions in the tongue of infected bovines by electron microscopy (not shown). Human keratinocytes express low levels of α_vβ₃ along with much higher levels of another related integrin, α_vβ₅ (50). This latter integrin also appears to be expressed in undifferentiated epithelia derived from the human airway (39). We previously reported that antibodies to the α_vβ₅ integrin did not inhibit the binding or replication of type A₁₂ (14), and a preliminary experiment suggests that our prototype tissue culture-propagated type A₁₂, vRM-SSP, and vCRM8 cannot replicate in K562 cells transfected with cDNAs of this integrin (not shown).

This study presents another example of how selective pressures applied to an RNA virus quasispecies (34) can select viruses with different receptor specificities for replication in either the natural host or tissue culture. This type of receptor variation in clinical versus laboratory prototype strains has also been shown to occur for the coxsackie B viruses (10). In the case of human immunodeficiency virus type 1, phenotype has been related to differences in the chemokine coreceptors used for virus envelope fusion to target cells (15). It has also been possible to isolate poliovirus mutants which are able to use a mutated poliovirus receptor, and have expanded host range, simply by growing wild-type poliovirus on mutant receptor-expressing cells (27). Thus, in determining the proper types of antiviral intervention, be it either vaccines or non-immune system-based interference, the ability of some viruses to escape by changing the receptors they use for infection should be considered.