Andes Virus Recognition of Human and Syrian Hamster β3 Integrins Is Determined by an L33P Substitution in the PSI Domain

ABSTRACT Andes virus (ANDV) causes a fatal hantavirus pulmonary syndrome (HPS) in humans and Syrian hamsters. Human αvβ3 integrins are receptors for several pathogenic hantaviruses, and the function of αvβ3 integrins on endothelial cells suggests a role for αvβ3 in hantavirus directed vascular permeability. We determined here that ANDV infection of human endothelial cells or Syrian hamster-derived BHK-21 cells was selectively inhibited by the high-affinity αvβ3 integrin ligand vitronectin and by antibodies to αvβ3 integrins. Further, antibodies to the β3 integrin PSI domain, as well as PSI domain polypeptides derived from human and Syrian hamster β3 subunits, but not murine or bovine β3, inhibited ANDV infection of both BHK-21 and human endothelial cells. These findings suggest that ANDV interacts with β3 subunits through PSI domain residues conserved in both Syrian hamster and human β3 integrins. Sequencing the Syrian hamster β3 integrin PSI domain revealed eight differences between Syrian hamster and human β3 integrins. Analysis of residues within the PSI domains of human, Syrian hamster, murine, and bovine β3 integrins identified unique proline substitutions at residues 32 and 33 of murine and bovine PSI domains that could determine ANDV recognition. Mutagenizing the human β3 PSI domain to contain the L33P substitution present in bovine β3 integrin abolished the ability of the PSI domain to inhibit ANDV infectivity. Conversely, mutagenizing either the bovine PSI domain, P33L, or the murine PSI domain, S32P, to the residue present human β3 permitted PSI mutants to inhibit ANDV infection. Similarly, CHO cells transfected with the full-length bovine β3 integrin containing the P33L mutation permitted infection by ANDV. These findings indicate that human and Syrian hamster αvβ3 integrins are key receptors for ANDV and that specific residues within the β3 integrin PSI domain are required for ANDV infection. Since L33P is a naturally occurring human β3 polymorphism, these findings further suggest the importance of specific β3 integrin residues in hantavirus infection. These findings rationalize determining the role of β3 integrins in hantavirus pathogenesis in the Syrian hamster model.

Hantaviruses persistently infect specific small mammal hosts and are spread to humans by the inhalation of aerosolized excreted virus (41,42). Hantaviruses predominantly infect endothelial cells and cause one of two vascular leak-based diseases: hemorrhagic fever with renal syndrome (HFRS) and hantavirus pulmonary syndrome (HPS) (41). Hantavirus diseases are characterized by increased vascular permeability and acute thrombocytopenia in the absence of endothelial cell lysis (36,41,42,54). In general, hantaviruses are not spread from person to person; however, the Andes hantavirus (ANDV) is an exception, since there are several reports of person-toperson transmission of ANDV infection (11,37,47,52). ANDV is also unique in its ability to cause an HPS-like disease in Syrian hamsters and serves as the best-characterized hantavirus disease model with a long onset, symptoms, and pathogenesis nearly identical to that of HPS patients (20,21,50).
Hantavirus infection of the endothelium alters endothelial cell barrier functions through direct and immunological responses (8,14). Although the means by which hantaviruses cause pulmonary edema or hemorrhagic disease has been widely conjectured, the mechanisms by which hantaviruses elicit pathogenic human responses have yet to be defined. Hantaviruses coat the surface of infected VeroE6 cells days after infection (17), and this further suggests that dynamic hantavirus interactions with immune and endothelial cells are likely to contribute to viral pathogenesis. Hantavirus pathogenesis has been suggested to involve CD8 ϩ T cells, tumor necrosis factor alpha or other cytokines, viremia, and the dysregulation of ␤ 3 integrins (7, 8, 13-16, 25-28, 32, 34, 38, 44-46). However, these responses have not been demonstrated to contribute to hantavirus pathogenesis, and in some cases there are conflicting data on their involvement (18, 25-28, 34, 35, 44, 45, 48). Immune complex deposition clearly contributes to HFRS patient disease and renal sequelae (4,7), but it is unclear what triggers vascular permeability in HPS and HFRS diseases or why hemorrhage occurs in HFRS patients but not in HPS patients (8,36,54). Acute thrombocytopenia is common to both diseases, and platelet dysfunction resulting from defective platelet aggregation is reported in HFRS patients (7,8).
␣ v ␤ 3 and ␣ IIb ␤ 3 integrins exist in two conformations: an active extended conformation where the ligand binding head domain is present at the apex of the heterodimer and a basal, inactive bent conformation where the globular head of the integrin is folded toward the cell membrane (30,53,55). Pathogenic HTN and NY-1 hantaviruses bind to the N-terminal plexin-semaphorin-integrin (PSI) domain of ␤ 3 integrin subunits and are selective for bent, inactive ␣ v ␤ 3 integrin conformers (38). Pathogenic hantavirus binding to inactive ␣ v ␤ 3 integrins is consistent with the selective inhibitory effect of hantaviruses on ␣ v ␤ 3 function and endothelial cell permeability (14,15,38). Although the mechanism of hantavirus induced vascular permeability has yet to be defined, there is a clear role for ␤ 3 integrin dysfunction in vascular permeability deficits (5,6,22,29,39,40,51) which make an understanding of hantavirus interactions with ␤ 3 subunits important for both entry and disease processes.
The similarity between HPS disease in humans and Syrian hamsters (20,21) suggests that pathogenic mechanisms of ANDV disease are likely to be coincident. Curiously, other hantaviruses (Sin Nombre virus [SNV] and Hantaan virus [HTNV]) are restricted in Syrian hamsters and fail to cause disease in this animal, even though they are prominent causes of human disease (50). Although the host range restriction for SNV and HTNV in Syrian hamsters has not been defined (33), the pathogenesis of ANDV in Syrian hamsters suggests that both human and Syrian hamster ␤ 3 integrins may similarly be used by ANDV and contribute to pathogenesis.
We demonstrate here that ANDV infection of the Syrian hamster BHK-21 cell line and human endothelial cells is dependent on ␣ v ␤ 3 and inhibited by ␣ v ␤ 3 specific ligands and antibodies. Further, polypeptides expressing the N-terminal 53 residues of human and Syrian hamster ␤ 3 subunits block ANDV infection. This further indicates that ANDV interaction with the N-terminal 53 residues of both human and Syrian hamster ␤ 3 integrins is required for viral entry. We also demonstrate that ANDV recognition of human and Syrian hamster ␤ 3 integrins is determined by proline substitutions at residues 32/33 within the ␤ 3 integrin PSI domain. These results define unique ANDV interactions with human and Syrian hamster ␤ 3 integrins.
Cloning and sequencing of Syrian hamster ␤ 3 integrin subunit. Total RNA was extracted from Syrian hamster cell line BHK-21 and Syrian hamster liver tissue (RNeasy; Qiagen). RNA was reversed transcribed into cDNA (Transcriptor First Strand cDNA Synthesis; Roche) (25°C for 10 min, 42°C for 60 min, and 95°C for 5 min) by using primers containing consensus ␤ 3 integrin subunit sequences derived from human, murine, rat, and rabbit ␤ 3 integrins (GenBank accession numbers NM_00012, NM_016780, NM_153720, and NM_001082066, respectively): antisense primers 5Ј-ATCACAKACTGTAGCCTGCATGATGG C-3Ј (ending at bp 777 of the human ␤ 3 integrin sequence) and 5Ј-AGCACRT GTTTGTAGCCAAACATGGG-3Ј (ending at bp 659 of the human ␤ 3 integrin sequence) (R ϭ A or G; K ϭ T or G). cDNA was subjected to nested PCR by using the forward primer 5Ј-CTGGCGCTGGGGGCGCTGGCGGGCGT-3Ј starting at bp 43 of the human ␤ 3 integrin sequence and the reverse primer 5Ј-TCCACRAAKGCCCCRAAGCCAATCCG-3Ј starting at bp 526 of the human ␤ 3 integrin sequence (cycle 95°C for 1 min, 60°C for 1 min, and 72°C for 30 s). Amplified cDNA fragments were ligated into pCRII-TOPO (Invitrogen), transformed into XL1-Blue cells, and sequenced. The region corresponding to the N-terminal 53 residues of human ␤ 3 integrin was subcloned into pET6His vector at the BamHI/EcoRI site, transformed into BL21(DE3) cells, and subsequently expressed as described above.
Ligand and antibody pretreatment of cells. HUVECs, VeroE6 cells, and BHK-21 cells were pretreated with antibodies (0.1 to 1 g/ml) or potentially competitive ligands (5 to 20 g/ml) for 1 h at 4°C. Antibodies and ligands were preadsorbed to cells in 50 l of DMEM with 2% FCS in duplicate wells of a 96-well plate on ice. Sera or ligands were removed, and the cells were washed with PBS. Approximately 1,000 focus-forming units (FFU) of hantavirus were adsorbed to monolayers for 1 h at 37°C. Unbound virus was removed, monolayers were washed three times with PBS, and infected monolayers were incubated 24 h prior to methanol fixation (100% methanol, 30 min at Ϫ20°C).
Polypeptide Inhibition of ANDV and NY-1V Infection. Increasing amounts (5 to 20 g/ml) of human, murine (negative control), S32P murine mutant, bovine, P33L bovine mutant, Syrian hamster, or N39D Syrian hamster mutant ␤ 3 polypeptides (residues 1 to 53) were incubated with approximately 1,000 FFU of ANDV or NY-1V (2 h at 4°C) and then adsorbed to HUVECs, VeroE6 cells, or BHK-21 cells in duplicate wells (96-well plate, 100 l per well) for 1 h at 37°C. Monolayers were washed with PBS, and the cells were incubated for 24 h prior to methanol fixation. Infected cells were detected by immunoperoxidase staining using anti-nucleocapsid antibody and quantitated as previously described (16). The geometric mean titer of ␤ 3 integrin fluorescence on CHO cells was used as a measure of integrin expression levels (15), and only cells expressing comparable levels of human or bovine ␤ 3 integrins were used in ANDV infection experiments. Transfected and mock-transfected CHO cells were infected with ANDV 48 h posttransfection as described above. Cells were methanol fixed after 24 h and immunoperoxidase stained for hantavirus nucleocapsid protein, and the number of infected cells was quantitated by microscopy as previously described (16).

RESULTS
Syrian hamsters infected with ANDV develop a lethal disease with onset, symptoms, and respiratory distress similar to that of HPS patients (20,21,50). To date, all pathogenic hantaviruses have been demonstrated to use ␤ 3 integrins for cellular entry and pathogenic hantaviruses inhibit ␤ 3 functions days after infection (13)(14)(15)(16)38). However, ANDV interactions with human or Syrian hamster ␤ 3 integrins have not been defined, and the role of ␤ 3 integrins in regulating endothelial barrier functions provides a rationale for integrin regulation to contribute to viral pathogenesis. We investigated here whether ANDV interacts with human or Syrian hamster ␤ 3 integrins and defined residues required for ANDV infection. Initially, we evaluated the ability of ANDV to infect BHK-21 cells that are derived from Syrian hamsters. Figure 1 indicates that BHK-21 cells are highly infected by ANDV, although at a slightly reduced level from the identical infection of VeroE6 cells. However, there was a nearly 3-log reduction in NY-1V infection of BHK-21 cells compared to VeroE6 cells (Fig. 1). This result suggests a fundamental difference in ANDV infection of Syrian hamster cells from that of NY-1V, and this could result from differences in adherence to ␤ 3 integrins.
In order to determine whether ANDV uses ␤ 3 integrins, we evaluated whether integrin ligands or antibodies inhibited ANDV infection of VeroE6 and BHK-21 cells. Pretreating cells with increasing amounts of collagen, chondroitin sulfate, laminin, or fibronectin had no effect on ANDV infection. In contrast, pretreating VeroE6 cells ( Fig. 2A) and BHK-21 cells ( Fig. 2B) with vitronectin, a high-affinity ␣ v ␤ 3 ligand, inhibited ANDV infection by 80 and 70%, respectively. Since antibodies that recognize Syrian hamster ␤ 3 integrins are not available, we investigated the ability of antibodies to human ␤ 3 integrins to inhibit ANDV infection. Figure 3 shows that pretreating cells with antibodies to human ␤ 3 or ␣ v ␤ 3 dose dependently inhibited ANDV infection of HUVECs and VeroE6 cells (85 and 70%, respectively; Fig. 3). Antibodies to the ␣ v integrin subunit also reduced ANDV infection of HUVECs and VeroE6 cells (40 to 60%) (Fig. 3), whereas antibodies to ␣ 1 , ␣ 5 ␤ 1 , ␣ 2 , ␤ 2 , and ␤ 5 had no inhibitory effect on ANDV infection. These results indicate that ligands or antibodies specific to ␣ v ␤ 3 integrins mediate ANDV entry into human, simian, and Syrian hamster cells.
Binding of the NY-1V to ␤ 3 integrins was previously shown to be RGD independent, mediated by the N-terminal PSI domain present on human but not murine ␤ 3 integrin subunits, and dependent on an aspartic acid at position 39 (16,38). In order to determine whether ANDV interacts with human ␤ 3 integrin PSI domains, we determined whether the expressed ␤ 3 integrin PSI domain (residues 1 to 53) or antibody to the PSI domain (residues 17 to 48) inhibited ANDV infection. Pretreating VeroE6 cells ( Approximately 1,000 FFU of ANDV were adsorbed for 1 h at 37°C to duplicate wells of a 96-well plate. After viral adsorption, inocula were removed, and the cells were washed and incubated 24 h at 37°C before methanol fixation. ANDV-infected cells were identified by immunoperoxidase staining of the viral nucleocapsid protein and quantitated by microscopy as previously described (38). The results were reproduced in at least two separate experiments and are presented as a percentage of the controls.
in a 70% decrease in ANDV infection compared to control antibodies. Similarly, pretreating ANDV with increasing amounts of the expressed human ␤ 3 PSI domain reduced ANDV infection of HUVECs by 75% (Fig. 5A) and ANDV infection of BHK-21 cells by 85% (Fig. 5B). In contrast, the murine ␤ 3 integrin PSI domain had no apparent effect on ANDV infection of either cell type. In order to determine whether Syrian hamster ␤ 3 integrins also blocked ANDV infection, we cloned and expressed the ␤ 3 integrin PSI domain from BHK-21 cells. Five independent clones of the Syrian hamster ␤ 3 integrin were sequenced and found to be identical to the cDNA sequenced from Syrian hamster tissue (submitted to GenBank). Figure 6 compares Syrian hamster ␤ 3 integrin coding sequences with human, murine, and bovine ␤ 3 PSI domains. The Syrian hamster ␤ 3 integrin PSI domain differs from the human sequence by eight residues. In contrast, murine and bovine PSI domains, which do not confer hantavirus infectivity, differ by 9 and 4 residues, respectively, from human PSI domains, whereas murine sequences differ by five residues from the Syrian hamster PSI domain (positions 30, 32, 42, 43, and 50).
We further evaluated whether the expressed Syrian hamster ␤ 3 integrin PSI domain inhibited ANDV infection. Figure 7A indicates that pretreating ANDV with the Syrian hamster ␤ 3 integrin PSI domain (1 to 53) reduced ANDV infection of BHK-21 cells by 80%, similar to the human ␤ 3 integrin PSI domain (1 to 53). In contrast, the murine ␤ 3 PSI domain had no effect on ANDV infection of BHK-21 cells at any concentration. This indicates that both the Syrian hamster and human ␤ 3 integrin interact with ANDV. Since we previously demonstrated that the presence of an N or D residue at position 39 of ␤ 3 differentiated NY-1V interaction with ␤ 3 (38), we evaluated the effect of mutating residue 39 of the Syrian hamster ␤ 3 PSI domain from N to D. Figure 7A indicates that there was no difference in PSI domain inhibition of ANDV by the N39D mutant. This suggested a fundamental difference in the interaction of ␤ 3 with ANDV and NY-1V.
In order to determine whether the Syrian hamster ␤ 3 or the N39D Syrian hamster ␤ 3 mutant inhibited NY-1V infectivity, we comparatively evaluated the ability of the NY-1V and ANDV to infect HUVECs in the presence or absence of ␤ 3 PSI domains. Figure 7B demonstrates that the human and Syrian hamster ␤ 3 PSI domains inhibited ANDV infection of human endothelial cells irrespective of the N39D mutation. In contrast, the Syrian hamster wild-type ␤ 3 PSI domain had no effect on NY-1V infection of endothelial cells, whereas the Syrian hamster N39D mutant blocked NY-1V infectivity sim-

FIG. 3. ANDV infectivity is inhibited by integrin-specific antibodies. Duplicate wells with VeroE6 cells (A) and HUVECs (B)
were pretreated for 1 h at 4°C with 0.01 to 1 g of antibodies to the indicated integrins or integrin subunits prior to viral adsorption. Monolayers were washed with PBS, and ϳ1,000 FFU of ANDV were adsorbed to cells for 1 h at 37°C. The inocula were removed, and the number of ANDV-infected cells was quantitated as previously described (16). The results were reproduced in at least two separate experiments and are presented as a percentage of the controls.

FIG. 4. Antibodies to human ␤ 3 integrin PSI domain inhibit ANDV infection. Duplicate wells of VeroE6 cells (A) or HUVECs
(B) were pretreated for 1 h at 4°C with 0.01 to 1 g of preimmune antibody or rabbit antibodies generated to a peptide containing residues 17 to 48 of the human ␤ 3 integrin PSI domain. Subsequently, the cells were washed with PBS, and ϳ1,000 FFU of ANDV were adsorbed to monolayers in duplicate wells of a 96-well plate for 1 h at 37°C. After viral adsorption, the inocula were removed, and the cells were washed and incubated 24 h at 37°C before methanol fixation. The number of ANDV-infected cells was quantitated by immunoperoxidase staining of the hantavirus nucleocapsid protein as previously described (16). The results were reproduced in two separate experiments and are presented as a percentage of the untreated controls.  (38) and indicates that ANDV interactions with ␤ 3 integrins require discrete PSI domain residues from that of NY-1V. Specific PSI domain inhibition of ANDV infection further suggested that PSI domain sequences determine ANDV attachment and infectivity. We previously determined that substituting residues 1 to 43 of murine ␤ 3 with human ␤ 3 sequences conferred hantavirus infection (38). When this finding is considered, only two residues in the murine ␤ 3 (T30 and S32) and one in the bovine ␤ 3 (P33) are completely discrete from human or Syrian hamster PSI domain sequences (Fig. 6). Murine and bovine ␤ 3 PSI domains contain unique proline changes in adjacent residues (32 and 33, Fig. 6 [boxed]) that differ from the Syrian hamster and human ␤ 3 PSI domains and suggest that they are determinants of ANDV binding to ␤ 3 integrins.
In order to define the role of proline 33 in ANDV recognition, we mutated the bovine ␤ 3 integrin proline 33 to the leucine (P33L) present in human homologues and then determined whether the bovine P33L PSI domain mutant was capable of inhibiting ANDV infection. We similarly mutated the murine ␤ 3 PSI domain to contain a proline residue at position 32 (S32P) in order to mimic the human ␤ 3 PSI domain. Figure  8 indicates that the murine ␤ 3 PSI domain fails to inhibit ANDV infectivity, whereas both human and Syrian hamster ␤ 3 PSI domains block ANDV infection. However, the murine S32P mutant ␤ 3 PSI domain also inhibited ANDV infection by Ͼ50%. Similar to murine ␤ 3, the wild-type bovine ␤ 3 PSI domain failed to inhibit ANDV infection of BHK-21 cells and VeroE6 cells (Fig. 9). However, the P33L mutant bovine ␤ 3 PSI domain inhibited ANDV infection of VeroE6 and BHK-21 cells by 60 and 55%, respectively. These findings suggest that the absence of proline 32 or the addition of a second proline at position 33 in the ␤ 3 integrin PSI domain dramatically reduces ANDV recognition of PSI domain polypeptides.
To further evaluate the effect of P33 on ANDV infection, we reciprocally mutated the full-length bovine ␤ 3 subunit to contain P33L and the full-length human ␤ 3 to contain L33P. Subsequently, we transfected CHO cells with the ␤ 3 mutants and determined the susceptibility of CHO cells expressing mutant ␤ 3 subunits to ANDV infection. Equal expression of the ␣ v ␤ 3 integrins on the surface of CHO cells was confirmed by flow cytometry, as measured by comparable geometric mean titers of ␣ v ␤ 3 cell surface fluorescence. CHO cells expressing either bovine ␤ 3 or the human ␤ 3 L33P mutant failed to confer ANDV infectivity (Fig. 10). In contrast, CHO cells expressing either human ␣ v ␤ 3 or a bovine ␤ 3 P33L mutant enhanced ANDV infection by Ͼ5-fold (Fig. 10). These results establish that a proline residue at position 33 inhibits ANDV recognition of ␤ 3 integrin PSI domains and further demonstrate the ␤ 3 integrin sequence specificity of ANDV recognition.

DISCUSSION
␤ 3 integrins are prominent cell surface receptors on endothelial cells and regulate vascular integrity, permeability, and hemostasis (2,6,19,23,39,40). Interestingly, all pathogenic hantaviruses analyzed thus far use ␤ 3 integrin receptors on human endothelial cells and cause vascular permeability-based diseases (HTNV, SNV, Puumala virus [PUUV], Seoul virus [SEOV], and NY-1V), whereas ␣ 5 ␤ 1 integrins are used by nonpathogenic hantaviruses (13,16,38). ANDV infection of Syrian hamsters is a model of HPS pathogenesis that closely mimics human disease (20,21,50); however, the use of ␤ 3 integrins by ANDV has not been established. We performed here a series of experiments to determine how ANDV interacts with ␤ 3 integrins. Our results indicate that ANDV infection of either human endothelial cells or Syrian hamster BHK-21 cells was inhibited by the high-affinity ␣ v ␤ 3 integrin ligand vitronectin and antibodies against ␤ 3 subunits. These findings demonstrate that ANDV infectivity is dependent on the presence of ␤ 3 integrins. ␤ 3 integrins reportedly form at least two conformations: an active extended conformation which binds vitronectin and an inactive bent conformation which forms the basal integrin state (30,53,55). Pathogenic hantaviruses have been shown to bind PSI domains present at the apex of bent inactive ␤ 3 integrin FIG. 5. Human ␤ 3 integrin PSI domain polypeptides inhibit ANDV infection. Increasing amounts (5 to 20 g) of expressed, and purified polypeptides containing residues 1 to 53 of human or murine ␤ 3 integrins were incubated for 2 h at 4°C with ϳ1,000 FFU of ANDV. Virus was subsequently adsorbed to HUVECs (A) or BHK-21 cells (B) in 96-well plates for 1 h at 37°C. Monolayers were washed with PBS, and at 1 day postinfection the number of ANDV-infected cells was quantitated after methanol fixation and immunoperoxidase staining of the hantavirus nucleocapsid protein as previously described (38). The results were reproduced in at least two separate experiments and are presented as a percentage of the mock-treated controls.
FIG. 6. Alignment of human, Syrian hamster, murine, and bovine ␤ 3 integrin PSI domains. Amino acid sequences of residues 1 to 53 from human, Syrian hamster, murine, and bovine ␤ 3 integrin (GenBank accession no. XM_616376) subunits are comparatively presented. Residue differences observed in ␤ 3 integrin homologues that differ from human ␤ 3 sequences are indicated, and dashes indicate identical residues. Residues 17 to 48 are delineated by a line, and proline residue differences at positions 32 and 33 of the ␤ 3 integrin homologues are boxed.
subunits, and the binding is species specific (38). The results presented here demonstrate that antibodies against the ␤ 3 integrin PSI domain, as well as PSI domain polypeptides derived from either human or Syrian hamster origin, inhibit ANDV infectivity. In contrast, ANDV failed to recognize PSI domain polypeptides of murine or bovine origin. These findings indicate that ANDV selectively recognizes human and Syrian hamster ␤ 3 integrin PSI domains which direct ANDV infectivity. These findings are consistent with ANDV pathogenesis in both humans and Syrian hamsters. The species-specific use of ␤ 3 integrin PSI domains permitted us to further analyze critical residues within the PSI domain required for ANDV recognition. A comparison of species specific ␤ 3 subunit PSI domains shows that murine and bovine ␤ 3 sequences contain proline substitutions at positions 32 and 33 (P32S and L33P, respectively) resulting in the absence of proline (murine) or the presence of two adjacent prolines (bovine) in this domain. Mutagenesis demonstrated that ANDV recognition of human and Syrian hamster ␤ 3 integrins is abolished by substituting leucine for proline at position 33 of the human ␤ 3 subunit (L33P). Similarly, substituting proline for serine (P32S) in the murine ␤ 3 PSI domain permitted the FIG. 8. Murine ␤ 3 integrin PSI domain mutants inhibit ANDV infection. Increasing amounts (5 to 20 g) of human, murine, murine S32P, and Syrian hamster ␤ 3 integrin PSI domains were incubated with ϳ1,000 FFU of ANDV for 2 h at 4°C. Mixtures were adsorbed to VeroE6 cells in a 96-well plate for 1 h at 37°C. After a washing step, the cells were incubated for 24 h, and the number of ANDV-infected cells was quantitated by immunoperoxidase staining of the nucleocapsid protein as previously described (38). The results were reproduced in at least two separate experiments and are presented as a percentage of the mock-treated controls. Conversely, when proline 33 of the bovine ␤ 3 subunit was replaced by leucine (P33L) the mutant bovine ␤ 3 subunit also conferred cell susceptibility to ANDV infection. These findings indicate that ANDV specificity for ␤ 3 integrins is directed by residues 32 and 33 or altered by proline-induced changes at these positions. Interestingly, required ␤ 3 residues for ANDV binding differ from recognition sequences of NY-1V. In contrast to ANDV, NY-1V binding to ␤ 3 is dependent on D39, blocked by introducing N39 within human ␤ 3 (38) or the N39 containing wild-type Syrian hamster ␤ 3 (Fig. 7), and unaltered by residue 32 or 33 substitutions (38). This is a fundamental difference in ␤ 3 recognition between ANDV and NY-1V; however, residues required for binding of either virus reside within the same ␤ 3 domain. These findings suggest that additional pathogenic hantaviruses may also have unique ␤ 3 residue binding requirements that specify their integrin interactions. It is unclear whether unique ANDV interactions with ␤ 3 integrins are responsible for the discrete ability of ANDV to cause disease in Syrian hamsters or permit person to person transmission. NY-1V is clearly restricted in BHK-21 Syrian hamster cells (Fig. 1), and the data presented here suggest that at one level this difference may be due to species-specific ␤ 3 integrin residues (Fig. 7B). It is unclear whether similar differences in ␤ 3 residue requirements will be observed in SNV and HTNV, which are also nonpathogenic in Syrian hamsters and yet pathogenic in humans. However, there are many reasons for why a virus may be restricted from causing disease in another species. In fact, analysis of an SNV reassortant containing the ANDV M segment, and its encoded attachment proteins, indicates that SNV is restricted from causing disease in Syrian hamsters as a result of its S and L segments (33). In fact, S and L segment RNA transcription by the SAS reassortant is identical to SNV and reduced 1 to 2 logs from that of ANDV (33), suggesting that replication differences may restrict SNV from being pathogenic in Syrian hamsters. Consistent with this, SNV did not cause viremia within infected Syr-ian hamsters, and no S-segment RNA was isolated from peripheral blood mononuclear cells after infection (50). However, it is also possible that SNV and ANDV differ in their ability to regulate other cell or immune response imposed restrictions within Syrian hamsters that both viruses regulate successfully in order to be pathogenic in humans. Although there are many hurdles for a virus to overcome in order to be pathogenic in any species, the use of human and Syrian hamster ␤ 3 integrins by ANDV and the function of ␤ 3 integrins in regulating vascular permeability provide a compelling role for the involvement of ␤ 3 in the hantavirus disease process.
The importance of residues 32 and 33 as determinants of ␤ 3 integrin antigenic specificity furthers the potential role of ␤ 3 in altering vascular barrier functions during hantavirus infection. The L33P substitution is a naturally occurring ␤ 3 integrin polymorphism that differentiates human platelet antigen 1a (HPA-1a) from HPA-1b (24,51). Interestingly, the L33P substitution is sufficient to direct autoimmune responses to ␤ 3 integrins from blood containing a different HPA type, and this immune response to ␤ 3 results in two autoimmune diseases, fetomaternal allo-immune thrombocytopenia (FMAIT) and posttransfusion purpura (PTP) (24,49,51). FMAIT and PTP patients display vascular permeability and acute thrombocytopenia similar to symptoms of hantavirus-infected HFRS and HPS patients (24,49,51). Unfortunately, there are no patient data available on the role of the HPA-1a/1b polymorphism on hantavirus infection or disease. One study has indicated that a polymorphism in the HPA-3b allele (I843S) of ␣ IIb ␤ 3 integrins results in more severe clinical HFRS disease (29). However, since only 1% of the Asian population is homozygous for the FIG. 9. P33L Mutants of the bovine ␤ 3 integrin inhibit ANDV infection. Increasing amounts (5 to 20 g) of human, murine, bovine (Bo) wild-type, or bovine P33L ␤ 3 integrin PSI domain were incubated with ϳ1,000 FFU of ANDV for 2 h at 4°C. Mixtures were adsorbed onto BHK-21 or VeroE6 cells in a 96-well plate for 1 h at 37°C. After a washing step, the cells were incubated for 24 h, and the number of ANDV-infected cells was quantitated by immunoperoxidase staining of the nucleocapsid protein as previously described (38). The results were reproduced in at least two separate experiments and are presented as a percentage of the mock-treated controls.
FIG. 10. Bovine ␤ 3 integrin P33L confers cell susceptibility to ANDV infection. CHO cells were cotransfected with recombinant human ␣v integrin subunits along with human wild-type ␤ 3 , bovine wild-type ␤ 3 , mutant bovine P33L ␤ 3 , or human L33P ␤ 3 integrin subunits. Cell surface expression of ␣ v ␤ 3 integrins was determined by flow cytometry, and cells were subsequently infected with ϳ1,000 FFU of ANDV for 1 h at 37°C. After viral adsorption, inocula were removed, and the cells were washed and incubated 24 h at 37°C before fixation. Infected CHO cells were quantitated by immunoperoxidase staining of the nucleocapsid protein as previously described (38). The results were reproduced in at least two separate experiments and are presented as fold of mock-transfected controls.
HPA-1b allele, the study did not determine whether HPA-1 influences hantavirus disease (43). Thrombocytopenia is another hallmark of patients infected by pathogenic hantaviruses (8,36,41,54); however, neither the role of platelets in hantavirus disease nor the mechanism of hantavirus induced thrombocytopenia has been defined (31,41). Cosgriff et al. demonstrated that platelets from HFRS patients are defective in platelet activation (aggregation and granule release) and that this defect was not the result of a soluble circulating factor (7). This implies that thrombocytopenia in HFRS patients results from a block in platelet activation rather than from excessive platelet activation. This platelet defect following hantavirus infection, together with an understanding of hantavirus binding to inactive ␣ IIb ␤ 3 integrins, suggests that hantaviruses may bind inactive ␣ IIb ␤ 3 integrin conformers and prevent platelet activation. Since hantaviruses are cell associated at late times after infection (17), this also suggests that cell surface-displayed hantaviruses could potentially recruit quiescent platelets to the endothelial cell surface and simultaneously prevent platelet activation, reduce the level of activated platelets, and mask the presence of hantavirus-infected cells.
Our findings do not exclude additional immunologic mechanisms which may contribute to hantavirus pathogenesis and vascular permeability (25-28, 32, 34, 44, 46). However, these and other findings suggest that there are a number of means by which hantavirus dysregulation of ␤ 3 integrin functions may contribute to vascular permeability and pathogenesis (6, 13-16, 31, 38-40). Pathogenic hantaviruses clearly alter the permeability of infected endothelial cells in a manner consistent with the effects of ␤ 3 integrin dysregulation (2,3,6,14,15,23,(38)(39)(40). Pathogenic hantaviruses selectively inhibit ␤ 3 integrindirected migration (15) and enhance endothelial cell permeability in response to VEGF (14). However, hantavirus regulation of endothelial cell integrin functions is not the result of virus adsorption, since low MOIs (0.1 to 1) are applied to cells, and neither endothelial cell migration nor endothelial cell permeability are altered at early times after infection (14,15). Inhibited migration and enhanced permeability of endothelial cells is only observed days after infection (14,15) when hantaviruses reportedly coat the surface of infected cells (17). In fact, pathogenic hantaviruses may remain cell associated through hantavirus-␤ 3 interactions on the surfaces of endothelial cells and thus contribute to ␤ 3 dysfunction and enhanced endothelial cell permeability following viral emergence. These findings further suggest that ANDV interactions with ␤ 3 integrins could contribute to endothelial cell barrier dysfunction observed in humans and Syrian hamsters after ANDV infection. The studies presented here rationalize the use of the ANDV Syrian hamster HPS disease model to determine whether ␤ 3 integrins contribute to hantavirus pathogenesis.