Pathogenic hantaviruses bind plexin–semaphorin–integrin domains present at the apex of inactive, bent αvβ3 integrin conformers

Abstract

The αvβ3 integrins are linked to human bleeding disorders, and pathogenic hantaviruses regulate the function of αvβ3 integrins and cause acute vascular diseases. αvβ3 integrins are present in either extended (active) or dramatically bent (inactive) structures, and interconversion of αvβ3 conformers dynamically regulates integrin functions. Here, we show that hantaviruses bind human αvβ3 integrins and that binding maps to the plexin–semaphorin–integrin (PSI) domain present at the apex of inactive, bent, αvβ3-integrin structures. Pathogenic hantaviruses [New York-1 virus (NY-1V) and Hantaan virus (HTNV)] bind immobilized β3 polypeptides containing the PSI domain, and human (but not murine) β3 polypeptides inhibit hantavirus infectivity. Substitution of human β3 residues 1–39 for murine β3 residues directed pathogenic hantavirus infection of nonpermissive CHO cells expressing chimeric αvβ3 receptors. Mutation of murine β3 Asn-39 to Asp-39 present in human β3 homologues (N39D) permitted hantavirus infection of cells and specified PSI domain residue interactions with pathogenic hantaviruses. In addition, cell-surface expression of αvβ3 locked in an inactive bent conformation conferred hantavirus infectivity of CHO cells. Our findings indicate that hantaviruses bind to a unique domain exposed on inactive integrins and, together with prior findings, suggest that this interaction restricts αvβ3 functions that regulate vascular permeability. Our findings suggest mechanisms for viruses to direct hemorrhagic or vascular diseases and provide a distinct target for modulating αvβ3-integrin functions.

Pathogenic hantaviruses cause two human diseases with striking changes in vascular permeability: hemorrhagic fever with renal syndrome and hantavirus pulmonary syndrome (1–3). Hantaviruses infect endothelial cells throughout the body, and pulmonary or renal manifestations may accompany either disease (1–3). Unlike some hemorrhagic viruses, hantaviruses do not lyse endothelial cells, and alternate processes are required to explain vascular changes that direct pathogenesis. There are no therapeutic approaches for hantavirus diseases, and the cause of acute thrombocytopenia, edema, or vascular hemorrhage after hantavirus infections have not been defined.

Hemorrhagic diseases are caused by permeability changes within the vascular endothelium. β3 integrins on platelets (α_IIbβ3) and endothelial cells (αvβ3) regulate vascular permeability (4–8), and several human hemorrhagic diseases result from dysregulating β3-integrin functions (9–14). Glanzmann's disease is a human genetic β3-integrin-specific bleeding disorder hallmarked by thrombocytopenia and microvascular hemorrhage (9). At least three autoimmune hemorrhagic diseases are caused by Ab responses to N-terminal residues of β3-subunits or to an atypical β3-integrin ligand (10–14). Among these diseases, Goodpasture's syndrome is hallmarked by lethal pulmonary edema, hemorrhage, and glomerulonephritis (11, 13, 14). Pathogenic hantaviruses also dysregulate β3-integrin functions (15) by means of atypical interactions with β3 (16), and the similarity of β3-integrin disorders to hantavirus pathogenesis suggests a further role for β3 integrins in viral hemorrhagic diseases (9).

We have reported that Abs to β3-integrin subunits, as well as the high-affinity αvβ3-integrin ligand vitronectin, selectively inhibit infection by diverse pathogenic hantaviruses [hantavirus pulmonary syndrome: Sin Nombre virus (SNV) and New York-1 virus (NY-1V); hemorrhagic fever with renal syndrome: Hantaan virus (HTNV), Seoul virus (SEOV), and Puumala virus (PUUV)] (17, 18). In contrast, nonpathogenic hantaviruses [Prospect Hill virus (PHV) and Tula virus (TULV)] were not inhibited by β3-integrin reagents, and the cellular entry of these viruses is consistent with the use of α5β1 integrins (16–18). Recombinant αvβ3 integrins expressed in CHO cells were shown to confer infection by only pathogenic hantaviruses, and infectivity was selectively blocked by β3 Abs. Also, the ability of endothelial cells to migrate on vitronectin is also selectively inhibited by infection with pathogenic, but not nonpathogenic, hantaviruses and results indicate that the function of αvβ3, rather than αvβ3 receptor levels, were selectively blocked by pathogenic hantaviruses (15).

Arginine–glycine–aspartic acid (RGD) motifs direct ligand binding to several integrins, and RGD-containing peptides competitively inhibit ligand binding. Interestingly, RGD peptides do not inhibit hantavirus infection, and β3-integrin mutants, which are incapable of binding RGD or vitronectin, still direct the infection of CHO cells by pathogenic hantaviruses (17, 18). These findings are consistent with the lack of RGD sequences within hantavirus G₁ and G₂ surface glycoproteins and are similar to Goodpasture's antigen, which was recently shown to be an RGD-independent β3-integrin ligand (19, 20). These findings suggest that pathogenic hantaviruses recognize unique RGD-independent αvβ3-integrin domains (17, 18) and provide a compelling rationale for vascular diseases to emanate from hantavirus interactions with β3 integrins.

There has been little rationale for how hantaviruses might interact with αvβ3 integrins to explain the RGD independence of the interaction and hantavirus dysregulation of αvβ3 functions (17, 18). However, recent crystal and EM structures of αvβ3 integrins demonstrate the presence of two dramatically different active and inactive αvβ3-integrin conformations, which are dynamically regulated (21–23). An activated αvβ3 conformation, directed by manganese ions, binds ligands through adjacent globular heads present at the apex of extended αvβ3-integrin heterodimers (21). In contrast, calcium ions direct the formation of an inactive, bent αvβ3-integrin structure with heterodimeric integrin subunits folded nearly in half and masking ligand-binding head domains, which face the cell surface (21). Bent αvβ3-integrin functions have not been reported, although the inactive conformation may present a unique RGD-independent target that could mediate hantavirus binding and regulate αvβ3-integrin functions.

In this article, we demonstrate that hantaviruses bind to residues within the plexin–semaphorin–integrin (PSI) domain (residues 1–53) present at the apex of the bent, inactive αvβ3 integrin and that recombinant bent integrins direct hantavirus entry into cells. We define β3-integrin residues that direct binding to human αvβ3 integrins and demonstrate that expressed human β3 polypeptides block hantavirus infection. These findings demonstrate hantavirus binding to αvβ3, explain β3-integrin regulation by pathogenic hantaviruses, and they provide a mechanism for hantavirus-directed vascular diseases. Our results suggest the utility of PSI domains on bent β3 integrins as targets for inhibiting hantavirus infection.

Materials and Methods

Cells and Virus. VeroE6 and CHO cells were grown in DMEM/10% FCS as described (17). Sf-9 cells were maintained in Grace's medium with 10% FCS or SF-900 (GIBCO). NY-1V, HTNV strain 76–118 (18), and PHV were propagated on VeroE6 cells as described (0.5–1 × 10⁶ cells per ml) in a BSL-3 facility (17, 18), and they were determined by PCR (Roche) to be mycoplasma-free.

Ligands, Peptides, and Abs. Vitronectin, fibronectin, and BSA were obtained from Sigma, and GRGDSP peptide was obtained from GIBCO. Rabbit sera to αvβ3 was kindly provided M. Roivainen (National Public Health Institute, Helsinki). Anti-human β3 (Ab 1932 and mAb1976) and αv mAb 1978 (clone LM-142) were obtained from Chemicon, and anti-murine β3 (CD61) was obtained from BD Biosciences. Rabbit anti-nucleocapsid polyclonal Abs, described in ref. 17, were used to detect hantavirus-infected cells by immunoperoxidase staining. Horseradish peroxidase and FITC conjugates were obtained from The Jackson Laboratory.

Plasmids. Human β3 and human αv in pcDNA3.1-/ZEO were provided by Barry Baxt (U.S. Department of Agriculture, Plum Island, NY), and murine β3 in pREP9 was provided by Eric Brown (University of California, San Francisco). αv and β3 pCDM8 clones were provided by Mark Ginsberg (The Scripps Research Institute, La Jolla, CA). Murine β3 was subcloned into pcDNA3.1-/ZEO and chimeric β3-integrin constructs were generated as described in Table 1, which is published as supporting information on the PNAS web site. Recombinant αv and β3 expression plasmids (Hαv-G307C and Hβ3-R563C) were provided by Tim Springer (Harvard University, Boston) (21). Hβ3-C5A, Mβ3-V22M, Mβ3-S³²Q³³-P³²L³³, Mβ3-T³⁰S³²Q³³-A³⁰P³²L³³, and Mβ3-N39D were generated by oligonucleotide-directed mutagenesis (Stratagene), sequenced, and evaluated for surface expression of αvβ3 by FACS analysis (15). N-terminal fragments containing residues 1–136 and 1–53 of Hβ3 and Mβ3 were subcloned from human or murine β3 plasmids into pET6His (24) at the BamHI site. The ectodomains of human αv (amino acids 1–969) and β3 (amino acids 1–692) (25) were PCR-amplified from pCDM8 plasmids (17) and cloned into pAcUW51 (Pharmingen) at the EcoRI (αv; p10 promoter) and BamHI (β3; polyhedron promoter) sites.

Protein Expression and Purification. Recombinant baculoviruses were generated by using pAcUW51 constructs and BaculoGold DNA (BD Biosciences), and expressed αvβ3 was secreted from Sf-9 cell supernatants as described (25). Secreted αvβ3 was purified by Con A–Sepharose 4B chromatography, elution with 30 mM α-methyl-d-mannoside and binding to RGD–Sepharose (0.5 mg of GRGDSPK coupled to cyanogen-bromide-activated Sepharose 4B). αvβ3 was eluted with 0.2 M EDTA, dialyzed against PBS, and analyzed by SDS/PAGE and silver staining (Bio-Rad) or Western blotting with a rabbit anti-αvβ3. Control preparations were ConA- and RGD-purified from supernatants of Sf-9 cells expressing a nonsecreted rotavirus protein VP5 (BacRec-VP5) (26).

Proteins were expressed from pET6His plasmids (24). Isopropyl β-d-thiogalactoside (IPTG) induction (1 mM for 3 h) was performed at 37°C for β3 1–136 and at 30°C for β3 1–53 polypeptides. Bacteria were resuspended in 0.1 M NaH₂PO₄/10 mM Tris·HCl/1 M urea, sonicated, and purified by using Ni²⁺–nitrilotriacetic acid resin (Qiagen, Valencia, CA) (24). Proteins were eluted with 0.5 M EDTA, dialyzed overnight in PBS (3.5-kDa cutoff), and quantitated by bicinchoninic acid (BCA) assay (Pierce).

Inhibiting Infectivity. NY-1V, HTNV, or PHV [≈1,000 focus-forming units (FFU)] were incubated with polypeptides (2 h on ice), and then adsorbed to duplicate wells (100 μl per well) of VeroE6 cells in 96-well plates (1 h at 37°C). Monolayers were washed with DMEM/2% FCS, and cells were incubated for 24 h, methanol-fixed, immunoperoxidase-stained for nucleocapsid protein, and quantitated (17).

Transfection and Infection. CHO cells were transfected (FuGene 6, Roche) with equal amounts of human αv and β3 plasmids (pcDNA3.1, 1.5 μg) and selected with G418 and Zeocin (500 μg/ml). Cells expressing αvβ3 integrins were incubated with mAb 1976 or anti-CD61 (murine β3) anti-mouse FITC, and they were sorted by using a FACSCalibur cell sorter (BD Biosciences) as described (15). Cells expressing similar levels of αvβ3 on their surface were collected (GeoMean 12–16) and used for infectivity experiments (Table 2, which is published as supporting information on the PNAS web site). Transfected and mock-transfected cells were infected with hantaviruses as described above, fixed at 24 h after infection, immunoperoxidase-stained for hantavirus nucleocapsid protein. Experiments using Hαv-G307C/Hβ3-R563C (locked bent) and Hβ3-C5A (locked extended) were performed doubly blinded, with coded samples transfected and quantitated independently.

Viral Binding. Protein A/G beads (Santa Cruz Biotechnology) were preincubated with mAb 1978 (anti-αv, LM-142, nonblocking) and adsorbed to 1 mg of purified secreted human αvβ3 in DMEM/2% FCS or with BSA (1 mg) in control purified Sf-9 supernatants (SF-900). NY-1V (≈5,000 FFU) was adsorbed to αvβ3 or control resin (2 h at 25°C) and washed with DMEM/2% FCS, and resin containing bound virus was adsorbed to duplicate wells of VeroE6 cells and virus-infected cells were quantitated 48 h after infection. Purified β3 polypeptides containing residues 1–136 (4 μg per 15 μl of resin) were bound to Ni²⁺–nitrilotriacetic acid beads for 30 min at 25°C. NY-1V or HTNV (≈5,000 FFU) were bound to resin and washed, and the amount of infectious virus bound to resin was quantitated as described above.

Results

Commercial preparations of αvβ3 integrins are extracted from membranes, contain hydrophobic anchor sequences, and are present in detergent that is deleterious to enveloped viruses. To define hantavirus interactions with αvβ3 integrins, we expressed and RGD-purified the ectodomains of the human αvβ3 as secreted heterodimers (Fig. 1A) (25). As reported (25), secreted αvβ3 contained two αv bands that are consistent with the presence of partially cleaved posttranslational αv products. Incubating increasing amounts of purified αvβ3 with virus before adsorption inhibited infection by pathogenic NY-1V and HTNV but had no effect on PHV (Fig. 1B). These data support our previous findings that only pathogenic hantaviruses use αvβ3 integrins (17, 18), and they suggest that purified αvβ3 might serve as a platform for viral-binding experiments.

Fig. 1.

β3-Integrin hantavirus interactions. (A) αvβ3 heterodimers were secreted from baculovirus infected Sf-9 cells as described (25). Secreted αvβ3 was purified by using ConA, followed by RGD-affinity chromatography (25), and detected by silver stain or Western blot analysis with rabbit anti-αvβ3. (B) Purified αvβ3 was incubated with NY-1V, HTNV, and PHV (DMEM/2% FCS) before adsorption to VeroE6 cells. Infected cells were methanol-fixed at 24 h after infection, immunoperoxidase-stained for hantavirus N-protein, and quantitated (17). ConA- and RGD-purified material from Bac-Rec-VP5-infected (26) Sf-9 cell supernatants was used as a control. Rabbit anti-αvβ3 (10 μg) binding to αvβ3 prevented αvβ3 from inhibiting infection. Results are presented as a percentage of infected cells (n ≈ 100) in the absence of αvβ3 and represent three independent experiments. (C) mAb LM-142 was bound to protein A/G resin and coated with secreted purified αvβ3 (500 ng). Vitronectin, fibronectin, RGD peptides (20 μg/μl), or BSA were adsorbed to αvβ3 resin in DMEM/2% FCS. NY-1V (5000 FFU) was adsorbed to untreated or protein-treated resin (2 h at 25°C). After washing (five times with DMEM/2% FCS), resin and bound virus were applied to VeroE6 cells, monolayers were washed, and at 48 h after infection, hantavirus-infected cells were quantitated, as shown in B. In two samples, NY-1V was pretreated with either secreted αvβ3 (200 ng) or control Sf-9 supernatent (Fig. 1B) before adsorption to resin containing αvβ3. Data are expressed as the percentage of control (resin without αvβ3). At least three independent experiments were performed with similar results. (D) β3-Integrin homologue swaps CHO cells expressing full-length Huαv. Indicated human or murine β3 integrins or β3-integrin chimeras were analyzed by FACS for αvβ3 and infected with pathogenic NY-1V, HTNV, or PHV (≈1,000 FFU) as described above, and cells were quantitated as in B at 24 h after infection. Chimeric β3 integrins (1–762) contain indicated human β3 residues inserted into a murine β3-integrin background. Data are expressed as percentage of mock-transfected CHO cells, and experiments were performed at least three times with similar results.

Low-hantavirus titers have made binding assays using conventional viral-labeling approaches difficult. To bypass this limitation, viral binding was assayed by monitoring binding of infectious virus to αvβ3 immobilized on protein A/G resin. Resin was incubated with NY-1V in the presence of indicated proteins in DMEM/2% FCS, washed, and bound infectious virus was quantitated. Immobilized αvβ3 bound NY-1V, and similar to infection studies, neither RGD peptides nor fibronectin inhibited NY-1V binding to purified αvβ3. However, NY-1V binding was inhibited specifically by the high-affinity αvβ3 ligand vitronectin (17, 18), as well as by soluble αvβ3. These results support our previous studies (17, 18), indicating that NY-1V, HTNV, Seoul virus (SEOV), and Puumala virus (PUUV) infections were blocked by vitronectin but not by RGD or fibronectin ligands. These findings demonstrate that NY-1V binds directly to αvβ3, and as a result, αvβ3 is a cellular receptor for pathogenic hantaviruses.

Human, but not murine, β3-integrin subunits permit hantavirus infection of CHO cells, and infection does not depend on the source of αv (16). These findings and the conservation of β3-subunits permitted us to map domains of human β3 that permit hantavirus infection when substituted for murine β3 residues (Fig. 1D). Plasmids expressing chimeric human/murine β3-subunits (Fig. 6, which is published as supporting information on the PNAS web site) were cotransfected into cells along with human αv expression plasmids. Cells were sorted for similar levels of αvβ3-integrin expression (Table 2) as described (15), and hantavirus infection was evaluated. Fig. 1D indicates that all chimeras containing residues 1–43 of the human β3 integrin direct the cellular entry of NY-1V and HTNV, but not PHV (Fig. 6). These findings map integrin requirements for hantavirus infection to residues outside the RGD-binding site and within the PSI homology domain at the N terminus of β3 (27).

We bacterially expressed and purified N-terminal β3-integrin polypeptides (1–136 and 1–53) (Fig. 7, which is published as supporting information on the PNAS web site) to determine whether they were capable of binding hantaviruses (Fig. 2) or inhibiting hantavirus infectivity (Fig. 3). Similar to αvβ3-binding experiments, residues 1–136 of human, but not murine, β3-subunits were capable of binding pathogenic NY-1V and HTNV (Fig. 2). In addition, when used as a potential competitor of hantavirus infectivity, pretreatment of NY-1V or HTNV with increasing amounts of human, but not murine, β3 polypeptides (1–53 shown) blocked viral infection of VeroE6 cells (Fig. 3). These findings indicate that human polypeptides of N-terminal β3-integrin domains bind hantaviruses and are capable of competitively inhibiting NY-1V and HTNV infection.

Fig. 2.

Binding to residues 1–136. Ni²⁺–NTA resin was incubated with 4 μg of a human or murine β3 polypeptides (residues 1–136; Fig. 7). NY-1V, HTNV, or PHV (≈5,000 FFU) were incubated with resin in DMEM/2% FCS (2 h, 25C). Resin was washed (five times, DMEM/2% FCS), and resin and bound virus were applied to VeroE6 cells and analyzed as described for Fig. 1C. The mean number of infected cells in the control (resin alone) was ≈100, and experiments were performed at least three times with similar results.

Fig. 3.

β3 polypeptides inhibit infection. Increasing amounts of human or murine β3 polypeptides (residues 1–53, Fig. 7) were incubated (2 h at 4°C in DMEM/2% FCS) with NY-1V, HTNV, and PHV, and they were subsequently adsorbed to VeroE6 cells. Polypeptide quantities are indicated as follows: none, black; 0.75 μM, white; 1.5 μM, dark gray; 3.0 μM, hatched; and 6 μM, light gray). Cells were washed, and infected cells at 24 h after infection were quantitated as described for Fig. 1B. The mean number of infected cells in controls (no added polypeptides) was ≈60, and experiments were performed at least three times with similar results.

There are only eight amino acid differences between residues 1–53 of human and murine β3 integrins and all changes reside within residues 1–39 (Fig. 4A). Bovine β3-subunits also fail to permit hantavirus infection (Fig. 8, which is published as supporting information on the PNAS web site), and there are only three residue differences between human and bovine β3-subunits in this region. Murine and bovine residue differences coincide, and as a result, we mutagenized coincident and adjacent murine β3 residues to homologous human residues (V22M, SQ32/33PL, and N39D). Mutant polypeptides (residues 1–53) were evaluated as inhibitors of hantavirus infection and compared with WT human or murine polypeptides (Figs. 4B and 9, which is published as supporting information on the PNAS web site). Changing murine residues 32/33 resulted in a polypeptide that reduced NY-1V infection by 25%, with little if any effect on HTNV. Altering residue 22 did not affect infection by any hantavirus, and the infectivity of PHV was unaffected by any of the tested polypeptides. In contrast, the N39D change within the murine β3 1–53 polypeptide inhibited NY-1V (75%), identically to the human β3 polypeptide, although they inhibit infection of HTNV by ≈56% (Fig. 4B). This result identified a single residue difference between human and murine β3 that is required for inhibiting infection by NY-1V and HTNV.

Fig. 4.

PSI-domain residue interactions. (A) Human and murine β3-integrin homologues (residues 1–43) are aligned. Residue differences are shown in bold. Only three residues differ between human and bovine β3 (underlined). (B) Indicated murine β3-subunit residues were mutated to homologous human residues and expressed in 6-His-tagged 1–53 polypeptides. Mutant polypeptides (6 μM) were added during adsorption and assayed for their ability to inhibit infection as in Fig. 3. The mean number of infected cells in the controls (media alone) was ≈60, and experiments were repeated at least three times with similar results. (C) CHO cells (expressing Huαv and WT or mutated murine β3 integrins substituted with homologous human β3 residues at indicated positions) were analyzed by FACS for αvβ3 and assayed for their ability to confer hantavirus infection as in Fig. 1D. The mean number of infected cells in control CHO cells was ≈15, and results are representative of three independent experiments.

Identical changes were made in full-length murine β3-integrin subunits, which were expressed on CHO cells and analyzed by FACS, as described above, for similar expression levels of αvβ3 (Fig. 4C). Only murine β3 recombinants with the N39D residue change permitted pathogenic NY-1V or HTNV hantavirus infection when expressed on CHO cells (Figs. 4C and 9). There was no apparent affect of other mutations on hantavirus infection, and PHV infectivity was unaffected by expression of any αvβ3 integrins. These findings indicate that aspartic acid 39 within human β3 PSI domains is required for infection by pathogenic NY-1V and HTNV.

The conformation of αvβ3-integrin heterodimers are dynamically regulated by divalent cations, with calcium favoring bent integrin conformers and manganese directing extended αvβ3-integrin conformations (21–23). Structural analysis indicates that β3 PSI domains are present at or near the apex of bent αvβ3 conformers and likely in an inaccessible position half-way down the heterodimeric stalk of the extended integrin (22, 23). Hantavirus interactions with PSI domains suggest that hantavirus infection may similarly be determined by divalent cations and αvβ3-integrin conformations. To address this possibility, we pretreated VeroE6 cells with manganese or calcium ions and evaluated hantavirus infection. Fig. 5A indicates that infection by NY-1V and HTNV were enhanced by calcium and inhibited by manganese ions. These results are consistent with NY-1V and HTNV interactions with apical PSI domains exposed on the surface of bent αvβ3-integrin conformers.

Fig. 5.

Bent and extended αvβ3 integrins. (A) VeroE6 cells were pretreated with Ca⁺² (1 mM) or Mn⁺² (40 μM) for 1 h before infection with NY-1V or HTNV. Virus was adsorbed (DMEM/2% FCS, 1 h), and the number of infected cells was quantitated at 24 h after infection, as described in Fig. 1B.(B) CHO cells transfected with Huαv and WT and locked-bent (αv-G307C and β3-R563C) (13) or locked-extended (β3-C5A) (21) β3 were analyzed by FACS for αvβ3; infected with NY-1V, HTNV, or PHV; and assayed as described for Fig. 1D. The mean number of infected cells in controls was ≈15, and results are representative of three independent experiments. (C) Model of pathogenic hantavirus interaction with bent αvβ3. For integrin, the α-subunit (gold), β-subunit (blue), and PSI domain (black) are indicated. For hantavirus proteins, G₁/G₂ (red and blue), N protein (green), and polymerase (yellow) are indicated.

To determine whether hantaviruses selectively interact with bent integrin conformers, we expressed recombinant αvβ3 integrins that are locked into bent (21) or extended (28) forms on the surface of nonpermissive CHO cells. Cells were sorted for similar levels of expressed αvβ3 integrins on their surface as described above and hantavirus infected. Expression of locked extended αvβ3-integrin conformers failed to confer hantavirus infection (Fig. 5B). In contrast, locked bent αvβ3-integrin conformers enhanced NY-1V and HTNV infection of cells, similar to cells expressing unmodified human αvβ3 integrins (Fig. 5B). Collectively, our findings demonstrate that hantaviruses bind to residues within the β3-integrin PSI domain that is exposed on the surface of bent αvβ3-integrin structures (Fig. 5C).

Discussion

We have demonstrated that β3 integrins direct the infection of cells by several divergent pathogenic, but not nonpathogenic, hantaviruses (16–18). We further determined that infection by pathogenic hantaviruses selectively inhibited the migration of endothelial cells on the high-affinity αvβ3-integrin ligand vitronectin, demonstrating that pathogenic hantaviruses dysregulated αvβ3-integrins functions after infection (15). β3-integrin interactions were RGD-independent, and collectively, these studies tied hantavirus pathogenesis to the use and regulation of β3 integrins (15–18). However, there was no understanding of RGD-independent ligand interactions with αvβ3 or rationales for αvβ3-integrin functions to be inhibited by non-RGD-directed ligands. Recent structural studies of αvβ3 (21–23, 27, 29, 30) have provided a compelling rationale for hantavirus interactions that explain these findings. Dramatic conformational changes of αvβ3 switch the integrin from active (extended) to inactive (bent) ligand-binding states (27), dynamically regulating RGD ligand-binding functions of αvβ3, and exposing new domains on bent integrins that may facilitate integrin regulation. Bent αvβ3-integrin conformations provide a mechanism by which RGD-independent interactions of hantaviruses dysregulate normal ligand-binding functions of αvβ3.

Our findings demonstrate that pathogenic NY-1V and HTNV bind PSI domains within β3-integrin subunits that are present at the apex of bent αvβ3-integrin conformations. Residues within the PSI domain determine hantavirus-binding interactions and expression of locked bent αvβ3 integrins confer cell susceptibility to NY-1V and HTNV. These findings indicate that pathogenic hantaviruses use domains exposed on bent αvβ3-integrin conformers, and they suggest a mechanism for the regulation of αvβ3-integrin functions by pathogenic hantaviruses. RGD-independent interactions of hantaviruses with β3 PSI domains are likely to lock αvβ3 integrins into a bent conformation and are consistent with hantavirus dysregulation of αvβ3 functions after infection (15).

Hantavirus binding and infectivity are centered on RGD-independent interactions with β3-integrin PSI domains. PSI domains are present at the apex of bent αvβ3 integrins (22, 23, 29) and may play important roles in regulating integrins, although there is little understanding of PSI domain functions or binding interactions (31). PSI-domain homologies are defined by the position of a tryptophan relative to five to eight cysteine residues with highly divergent intervening sequences (8, 27, 31). A greater understanding of PSI domains is complicated by the recent assignment of PSI-domain disulfide bonds from structural data that are entirely different from mutagenic and biochemical assignments (8, 32). Residue 39 is partially exposed on PSI domains (27) and appears to be critical for polypeptides to inhibit hantavirus infectivity and required for hantavirus infection of cells (Fig. 4 B and C). This residue is also adjacent to cysteine 38 and, as a result, changes at this position could cause large changes in exposed surfaces of PSI domains (32). However, the ability of bacterially expressed polypeptides to inhibit hantavirus infection suggests that primary rather than secondary PSI domain structures determine hantavirus β3-binding specificity (8).

β3 integrins are directly linked to acute hemorrhagic diseases, which include Glanzmann's disease, fetomaternal alloimmune thrombocytopenia (FMAIT), posttransfusion purpuria (PTP), and Goodpasture's syndrome (9, 20, 33). Similar to these diseases, hantaviruses block β3-integrin functions (15) and have prominent thrombocytopenia and vascular permeability defects, which include edema and hemorrhage. Genetic mutations in β3 integrins cause hemorrhagic disease (9). Immune recognition of β3-integrin isoforms, differing in a single PSI-domain residue (L33P), appears to be responsible for FMAIT and PTP (10, 33), whereas immune recognition of Goodpasture's antigen, an RGD-independent ligand of αvβ3 (13, 14, 19, 20), similarly directs vascular disease. In autoimmune bleeding disorders, acute disease is followed by long recovery times and coincides with waning Ab responses of patients (10–14, 33). Interestingly, Goodpasture's syndrome and hantavirus diseases have common pulmonary and renal disease tropisms as well as edematous and hemorrhagic effects (1–3, 11, 13, 14). Immune complexes have been suggested to play a role in hantavirus disease and may contribute to long disease onset, increased severity of hantavirus disease in healthy young adults, and long recovery periods that may depend on waning Ab responses (1–3). It is currently unknown whether hantaviruses direct similar autoimmune responses; however, the similarity of hantavirus diseases to other β3-integrin-directed vascular diseases provides a rationale for evaluating hantavirus-directed autoimmune responses, which may contribute to hantavirus pathogenesis.

Pathogenic hantaviruses cause diseases with striking similarity to β3-directed hemorrhagic diseases with pulmonary and renal effects. We have defined αvβ3-specific peptide inhibitors that block hantavirus infection and have therapeutic implications for hantavirus diseases. Our findings further indicate that pathogenic hantaviruses bind to PSI domains present at the apex of bent inactive β3 integrins and these data define an explicit β3 target for the development of highly selective hantavirus therapeutics. Because β3 integrins also play prominent roles in vascular disease, tumor invasion and angiogenesis, PSI domains may also serve as therapeutic targets for a wide variety of αvβ3-directed anomalies.