Rotavirus spike protein VP5* binds {alpha}2beta1 integrin on the cell surface and competes with virus for cell binding and infectivity

Rotaviruses are members of the family Reoviridae and are major gastrointestinal pathogens of humans and animals. The virions are formed by three concentric protein layers, with the outermost layer comprising VP7 and protruding VP4 spikes (Yeager et al., 1990). In group A rotaviruses, VP4 and VP7 dictate the P and G serotypes, respectively. VP4 is the major viral cell-attachment protein, whereas VP7 has a lesser role (Bass et al., 1991; Crawford et al., 1994; Kirkwood et al., 1998; Ludert et al., 1996). Proteolytic cleavage of VP4 by trypsin at residue 247 results in greatly increased infectivity and produces two subunits, VP5* (aa 248776) and VP8* (aa 1247) that remain associated with the virion (Espejo et al., 1981; Estes et al., 1981; Gilbert & Greenberg, 1998). Trypsin treatment does not appear to alter viruscell binding, but increases the rate of cell entry by an unknown mechanism (Clark et al., 1981; Kaljot et al., 1988). However, it is known that trypsin-treated, but not untreated, virions can induce cellular fusion from without (Falconer et al., 1995). Also, expressed VP5* permeabilizes membranes (Denisova et al., 1999; Dowling et al., 2000; Golantsova et al., 2004). Trypsin cleavage of VP4 results in dimeric VP4 spikes with a well-ordered structure (Crawford et al., 2001), suggesting the induction of a conformational change in VP4. Crystal structures of most of the portions of VP8* and VP5* that project from the virion show that VP4 also undergoes a second rearrangement that may facilitate VP4membrane interactions and results in a stable trimer (Dormitzer et al., 2002, 2004). At high pH, VP4 spikes on the virion undergo conformational change to a trilobed structure (Pesavento et al., 2005).

Rotaviruses recognize several cell-surface molecules, and the processes of rotavirus cell attachment and entry appear to be multifactorial. Binding of a minority of animal rotaviruses to terminal sialic acids has been demonstrated, although this interaction is not essential (Ludert et al., 2002; Mendez et al., 1993). Terminal sialic acid usage relates to P serotype (Ciarlet et al., 2002b), and VP8* of the neuraminidase-sensitive rotavirus strain RRV contains a sialoside-binding region involved in cell binding and haemagglutination (Dormitzer et al., 2002; Fiore et al., 1991). Other rotaviruses, including some human strains, may recognize subterminal sialic acids that are not removed by neuraminidase treatment, or other sugars, on glycoproteins or glycolipids including gangliosides (Delorme et al., 2001; Fukudome et al., 1989; Guo et al., 1999; Jolly et al., 2000, 2001a; Liakatos et al., 2006; Rolsma et al., 1998).

Several human and monkey rotaviruses bind recombinant, cell-surface-expressed α4β1 and α4β7 integrins and require the same α4 subunit regions for binding as natural α4 ligands. Rotaviruses may recognize α4 integrins on cells of the immune system, possibly facilitating virus spread or host immune response modulation (Graham et al., 2005; Halasz et al., 2005; Hewish et al., 2000). The heat-shock cognate protein 70 (Hsc70) has also been proposed to form a component of a rotavirus receptor complex that is recognized through aa 642659 of VP5* (Zárate et al., 2003).

Infection of permissive cells by many human and animal rotaviruses depends, to a significant extent, on VP4 recognition of α2β1 integrin and VP7 interactions with integrins αxβ2 and αvβ3 (Ciarlet et al., 2002a; Coulson et al., 1997; Graham et al., 2003, 2004; Guerrero et al., 2000). These rotaviruses are classified as integrin-using strains and include monkey virus strains RRV and SA11 and human strain Wa. Integrin usage relates to P serotype independently of terminal sialic acid usage (Graham et al., 2003). Almost all group A rotaviruses have the AspGlyGlu (DGE) sequence in VP5* at aa 308310, a motif that has been implicated in α2β1 recognition by type I collagen, and the GlyProArg (GPR) sequence in VP7 that is a ligand in fibrinogen for αxβ2 (Coulson et al., 1997). Monomeric and polymeric peptides containing the DGE and/or GPR sequences inhibit the infectivity of integrin-using rotaviruses by 3090 % (Coulson et al., 1997; Graham et al., 2004; Zárate et al., 2000a). Integrin-using rotaviruses have been reported to use αxβ2 and αvβ3 at a post-binding stage to facilitate infection (Graham et al., 2003; Guerrero et al., 2000).

Most evidence indicates that integrin-using human and animal rotaviruses bind to cell-surface α2β1. Infectious SA11, RRV and Wa binding to recombinant α2β1 on the K562 cell surface was specifically inhibited by DG-containing peptides and a function-blocking antibody to the α2 I domain (Graham et al., 2003, 2004; Hewish et al., 2000). SA11 and RRV precipitated two cell-surface proteins with characteristics of α2 and β1 integrin subunits and bound recombinant human α2β1 on Chinese hamster ovary cells to a greater extent than human α2 combined with hamster β1. This binding was inhibited by antibody to the α2 I domain, but not anti-α2 antibodies that mapped outside the I domain, and was eliminated by I domain deletion (Londrigan et al., 2003). VP5* (aa 248474, representing the N-terminal half of VP5*), expressed as a glutathione S-transferase (GST) fusion protein, retained the ability to bind a conformation-dependent neutralizing antibody and bound expressed α2 I domain. This I domain binding depended on the presence of the D308 and G309 residues in the DGE sequence of VP5*, as alanine mutagenesis of these residues in VP5* did not affect neutralizing-antibody recognition, but did abolish VP5* binding to the α2 I domain (Graham et al., 2003). In contrast to the above findings, binding of naturally occurring, integrin-using rotaviruses to cellular α2β1 was not detected by two other groups (Ciarlet et al., 2002a; Zárate et al., 2000a, b). One reason for this was their use of normally adherent MA104 cells as cell suspensions during viruscell binding assays (Graham et al., 2003). However, recent studies by one of these groups using adherent MA104 cells still failed to demonstrate α2β1 binding by RRV (Zárate et al., 2003, 2004).

To examine the mechanism of rotavirus binding to cellular α2β1 further and resolve this difference, we aimed here to analyse directly the importance of VP5* (aa 248474) containing the DGE sequence in rotavirus cell binding and infection mediated by α2β1. These studies provide the first direct experimental evidence that purified recombinant VP5* binds cell-surface α2β1 using DGE and inhibits the infection of homologous and heterologous rotaviruses that use α2β1 as a receptor.

Cell lines, viruses and antibodies.
The origins and maintenance of MA104 and K562 cells and the derivation of the K562 cells transfected with cDNA encoding empty vector (PBJ-K562), α2 (α2-K562), α3 (α3-K562) and α4 (α4-K562) used in this study have been described previously (Coulson et al., 1997; Graham et al., 2003; Hewish et al., 2000). By flow cytometry, MA104 cells express moderate levels of surface α2β1 (Coulson et al., 1997; Londrigan et al., 2000). Monitoring of the surface expression of α2β1, α3β1 and α4β1 on K562 cell lines was carried out by flow cytometry as described previously (Graham et al., 2005; Hewish et al., 2000). The origins, cultivation in MA104 cells and characterization of Rhesus monkey rotavirus P5B[3], G3 strain RRV; human rotavirus P1A[8], G1 strain Wa and porcine rotavirus P9[7], G3 strain CRW-8 have been described previously (Coulson & Kirkwood, 1991; Coulson et al., 1985, 1986). Rabbit antiserum to purified GST and RRV were each produced and obtained as described previously (Londrigan et al., 2003; Warner et al., 2001).

Production and characterization of purified recombinant VP5* proteins.
The RRV VP4 plasmid pBS/VP4 was provided by Dr E. Mackow (Department of Medicine and Department of Molecular Genetics and Microbiology, Stony Brook University, NY, USA). Soluble RRV VP5* containing aa 248474 and the D308A, G309A mutant of this VP5* generated by site-directed mutagenesis were produced as purified GST fusion proteins GSTVP5* and GSTVP5*D308A/G309A, respectively, by Escherichia coli expression as described previously (Graham et al., 2003). Most of the GSTVP5* and GSTVP5*D308A/G309A protein produced was insoluble, as reported previously (Dormitzer et al., 2004), but under the expression conditions used, a fraction that co-purified with the heat-shock bacterial chaperonin GroEL was soluble, as determined previously (Graham et al., 2003). By enzyme immunoassay (EIA), the soluble GSTVP5* and GSTVP5*D308A/G309A produced for the studies here were bound by anti-VP5* monoclonal antibody 2G4 (indicating preservation of this conformational epitope), and GSTVP5* but not GSTVP5*D308A/G309A bound to the α2 integrin I domain fusion protein, as described previously (Graham et al., 2003). The soluble protein obtained therefore was functional for 2G4 recognition and α2 integrin I domain binding. Plasmid insert identity was verified by DNA sequencing. Fusion protein identity was confirmed by Western blotting using anti-GST monoclonal antibody CH-1, as described previously (Graham et al., 2003).

Flow cytometric assay of recombinant VP5* binding to cells.
The ability of GSTVP5*, GSTVP5*D308A/G309A and GST to bind PBJ-K562, α2-K562, α3-K562 and α4-K562 cells was determined by flow cytometry as described previously (Warner et al., 2001). In brief, equimolar amounts of GSTVP5*, GSTVP5*D308A/G309A or GST (0·110 µg) were incubated with 5x10⁵ washed cells for 45 min at 4 °C unless stated otherwise. Cell numbers were quantified carefully to ensure that results were comparable between cell lines. All the following steps were carried out at 4 °C. Cell-bound protein was detected with rabbit antiserum to GST diluted 1 : 500. Similarly diluted, normal rabbit serum negative for rotavirus antibodies by EIA and with a neutralization titre against SA11 of <1 : 100 was used as a negative control. Bound rabbit antibodies were detected using fluorescein isothiocyanate-conjugated sheep anti-rabbit F(Ab')₂ (Chemicon) and cells were analysed by flow cytometry, as described previously (Graham et al., 2003). The relative level of recombinant protein bound to cells was quantified by calculation of the relative linear median fluorescence intensity (RLMFI) from the flow cytometric histograms, as described previously (Graham et al., 2003). An RLMFI value of 1·20 was considered to indicate cell binding by the recombinant protein, as has been determined previously for monoclonal antibody binding (Graham et al., 2003).

Indirect immunofluorescence assays of viruscell binding and infectivity and recombinant VP5* inhibition of viruscell binding and infectivity.
Measurement of the binding of infectious rotavirus to cells was determined using 5x10⁵ cells and infectivity assays were carried out using 1x10⁴ cells, as described previously (Coulson et al., 1997; Graham et al., 2003; Hewish et al., 2000; Londrigan et al., 2000). Assays were carried out using clarified viruscell harvests at an m.o.i. of 3·5 (viruscell binding) or 0·02 (infectivity). The determination of these optimum m.o.i. values, and demonstration that the titres of purified virus and clarified viruscell harvests bound to cellular integrins are indistinguishable, have been described previously (Coulson et al., 1997; Graham et al., 2003, 2005). Titres are given as fluorescent cell-forming units (f.f.u.) ml¹.

Briefly, for binding assays, trypsin-activated virus was adsorbed to cells at 4 °C for 1 h. Bound virus was harvested by two cycles of freezing and thawing in the presence of 1 µg porcine trypsin ml¹ and activated with 10 µg trypsin ml¹ at 37 °C for 1020 min, depending on the virus strain. In conjunction with vigorous vortex mixing, these trypsin digestions reduced levels of virion-associated protein and virion aggregation. Cell-bound virus titres were determined by incubation of MA104 cells with activated virus at 37 °C for 1 h, followed by indirect immunofluorescent staining of infected cells after 15 h. In experiments analysing recombinant protein inhibition of viruscell binding and infectivity, cells were treated with GSTVP5*, GSTVP5*D308A/G309A or GST at 37 °C for 1 h prior to virus addition, unless otherwise stated. Assays were then completed as described above. None of the proteins caused aggregation of PBJ-K562 or integrin-transfected K562 cells. Cell viability (measured by trypan blue exclusion) and the microscopic appearance of MA104 and K562 cell lines were unaltered by protein treatment.

On graphs, results were expressed as a percentage of the virus titre in the absence of any treatment and given as mean±SD of at least three experiments.

RRV GSTVP5* (aa 248474) binds K562 cells by both α2β1-dependent and -independent mechanisms
Previously, it was shown by EIA that RRV GSTVP5* binds recombinant α2 integrin I domain protein (Graham et al., 2003). The ability of GSTVP5* to bind cell-surface-expressed α2β1, α3β1 or α4β1 was examined at 4 and 37 °C by flow cytometry, using K562 cells transfected with the empty vector (PBJ-K562) or expressing recombinant human α2β1 (α2-K562), α3β1 (α3-K562) or α4β1 (α4-K562). The only endogenous β1 integrin expressed by K562 cells is α5β1. GSTVP5* bound to PBJ-K562, α2-K562, α3-K562 and α4-K562 cell lines at both 4 °C (Fig. 1a) and 37 °C (Fig. 1b). However, α2-K562 cells supported substantially more GSTVP5* binding than PBJ-K562, α3-K562 and α4-K562 cells at both 4 and 37 °C (Fig. 1ad). GSTVP5* binding to α2-K562 and α3-K562 cells was concentration-dependent over a 100-fold range (0·110 µg GSTVP5*) and was saturated at 5 µg GSTVP5* per 5x10⁵ cells (Fig. 1c and d). RLMFI values calculated from the experiments using GSTVP5* indicated that α2-K562 cells supported 1·32·2-fold more GSTVP5* binding than PBJ-K562, α3-K562 and α4-K562 cells (Table 1). Overall, at 4 and 37 °C, GSTVP5* bound to K562 cells using α2β1 and also via an additional mechanism that was independent of α2β1, α3β1 or α4β1.

(40K): Fig. 1. Binding of recombinant RRV GSTVP5* to cells occurs by both α2β1-dependent and α2β1-independent mechanisms and is concentration-dependent and saturable. Flow cytometric histograms show the binding of 5 µg GSTVP5* to K562 cell lines at 4 °C (a) and 37 °C (b). One histogram trace, which was indistinguishable from the other histograms, is given for the bracketed cell lines for clarity. At 37 °C, GSTVP5* binding to α2-K562 and α3-K562 cells was concentration-dependent and saturable, using 0·1, 5 and 10 µg VP5* (c) and 1 µg VP5* (d). Histograms of GSTVP5* reacted with cells and detected with anti-GST serum are shown as unbroken lines. In all panels, control histograms (dotted lines) indicate GSTVP5* binding to each K562 cell line detected with negative-control rabbit serum and GST binding to each cell line detected with anti-GST or negative-control serum. Concentrations of GST and GSTVP5* in controls were matched to the test protein concentrations and were equimolar for GST.

Table 1. Relative levels of GSTVP5* and GSTVP5*D308A/G309A binding to K562 cell lines

RRV GSTVP5*D308A/G309A only binds to K562 cells independently of α2β1
RRV GSTVP5* binding to recombinant α2 integrin I domain protein has been demonstrated to depend on the presence of D308 and/or G309 in VP5*, as RRV GSTVP5*D308A/G309A did not bind the α2 I domain (Graham et al., 2003). The ability of RRV GSTVP5*D308A/G309A to bind K562 cells by α2β1-dependent and α2β1-independent means was examined. As shown in Fig. 2 and Table 1, GSTVP5*D308A/G309A binding to α2-K562 cells was detected (RLMFI=1·4) at a lower level than GSTVP5* binding to α2-K562 cells (RLMFI=3·1). GSTVP5*D308A/G309A and GSTVP5* bound α3-K562 cells indistinguishably, with identical RLMFI values of 1·4, at levels indistinguishable from VP5*D308A/G309A binding to α2-K562 cells (RLMFI=1·4). These results showed that mutation of the DG sequence in VP5* abolished VP5* binding to cell-surface-expressed α2β1, but that VP5* binding to cells independently of α2β1 remained detectable.

(15K): Fig. 2. Flow cytometric histograms of GSTVP5*, GSTVP5*D308A/G309A and GST bound to α2-K562 (a) and α3-K562 (b) cells show that GSTVP5*D308A/G309A binding to K562 cells was independent of α2β1. Proteins were added at equimolar levels, containing 1 µg VP5* or VP5*D308A/G309A. Completely overlapping histograms in (b) are bracketed. Controls (dotted lines) were GSTVP5* or GSTVP5*D308A/G309A bound to each K562 cell line and detected with negative-control rabbit serum and GST bound to each cell line and detected with anti-GST or negative-control serum. Graphs show data representative of at least two experiments.

Binding of RRV GSTVP5*, but not RRV GSTVP5*D308A/G309A, to cell-surface-expressed α2β1 prevents RRV binding
The ability of cell-bound RRV GSTVP5* and RRV GSTVP5*D308A/G309A to inhibit infectious RRV binding to cellular α2β1 was examined (Fig. 3). In the absence of cellular treatment with recombinant protein, the RRV titre bound to α2-K562 cells was increased 1·9-fold over that bound to PBJ-K562 and α3-K562 cells, as shown previously (Graham et al., 2003). Cellular treatment with RRV GSTVP5* at 37 °C eliminated infectious RRV binding to α2β1 on α2-K562 cells, as the RRV titre bound was indistinguishable from that bound to PBJ-K562 and α3-K562 cells. However, RRV GSTVP5*D308A/G309A and GST had no effect on the titre of RRV bound to α2β1 on α2-K562 cells and did not alter the background level of RRV binding to PBJ-K562 and α3-K562 cells. Thus, RRV binding to cellular α2β1 was prevented by prior cellular treatment with homologous GSTVP5* at 37 °C and this inhibition was completely dependent on VP5* D308 and/or G309.

(31K): Fig. 3. Binding of recombinant RRV GSTVP5* to recombinant α2β1 on the K562 cell surface at 37 °C prevents infectious RRV binding and this blockade depends on the presence of D308 and/or G309 in GSTVP5*. Prior to RRV binding, PBJ-K562, α2-K562 and α3-K562 cells were treated with 50 µg recombinant proteins GST, GSTVP5* or GSTVP5*D308A/G309A ml1 or were mock treated. The mean titre of infectious RRV bound to each cell line is expressed as a percentage of the titre bound to PBJ-K562 cells in the absence of GST fusion protein, which gave a mean±SD of 8·0x104±6·4x103 f.f.u. ml1.

Binding of RRV GSTVP5*, but not RRV GSTVP5*D308A/G309A, to cell-surface-expressed α2β1 inhibits RRV and Wa infection of MA104 cells
As RRV GSTVP5*, but not RRV GSTVP5*D308A/G309A, was able to bind cell-surface α2β1 on K562 cells and so prevent RRV binding to the same integrin, the abilities of these proteins to bind permissive MA104 cells and compete with rotavirus infectivity were examined. Both GSTVP5* and GSTVP5*D308A/G309A bound to MA104 cells, showing RLMFI values of 5·2±0·5 and 4·0±0·4, respectively. These RLMFI values are consistent with the results described above, which demonstrated that GSTVP5* binds K562 cells both via α2β1 and independently of this integrin, whereas GSTVP5*D308A/G309A binds K562 cells independently of α2β1 only. The higher RLMFI value shown by GSTVP5* probably reflects its additional ability to bind α2β1. Application of RRV GSTVP5* to MA104 cells at 4 °C had no effect on the infectivity of RRV added subsequently (Fig. 4a). However, cellular treatment with RRV GSTVP5* at 37 °C inhibited RRV infectivity in a dose-dependent fashion with a mean±SD of 38±4 % at 50 µg ml¹. This infectivity blockade was abrogated by the D308A and G309A mutations in VP5*, as RRV GSTVP5*D308A/G309A did not inhibit RRV infectivity at any concentration tested (Fig. 4b). Importantly, RRV GSTVP5* similarly inhibited infection by the heterologous human rotavirus strain Wa, with a mean of 46±3 % at 50 µg ml¹, whereas RRV GSTVP5*D308A/G309A did not inhibit Wa infectivity (Fig. 4c). Neither RRV GSTVP5* nor RRV GSTVP5*D308A/G309A had any effect on infection by CRW-8 rotavirus (Fig. 4d), which does not use α2β1 for cell binding or growth (Graham et al., 2003, 2004; Hewish et al., 2000). Thus, RRV GSTVP5* bound to cells at 37 °C but not at 4 °C inhibited RRV and Wa, but not CRW-8, rotavirus infectivity in MA104 cells. This blockade required the presence of VP5* D308 and/or G309 and so was dependent on VP5* binding to α2β1.

(21K): Fig. 4. Recombinant RRV GSTVP5* inhibits MA104 cell infection by RRV and Wa but not CRW-8 and this blockade depends on the presence of D308 and/or G309 in GSTVP5* and on temperature. (a) Incubation of cells with GSTVP5* at 4 °C prior to RRV infection does not inhibit RRV infectivity. (b) RRV infectivity is inhibited by prior cellular incubation at 37 °C with GSTVP5* but not GSTVP5*D308A/G309A or GST. (c) Wa infectivity is inhibited by prior cellular incubation at 37 °C with GSTVP5* but not GSTVP5*D308A/G309A or GST. (d) CRW-8 infectivity is not affected by prior incubation of cells at 37 °C with GSTVP5*, GSTVP5*D308A/G309A or GST. The mean infectivity titre of each rotavirus strain is expressed as a percentage of the infectivity titre in the absence of any GST fusion protein or GST. The mean infectivity titres±SD in f.f.u. ml1 measured in the absence of any protein were 1·3x104±2·4x102 for RRV, 9·8x103±1·5x102 for Wa and 1·2x104±5·3x102 for CRW-8.

In this study, the mechanisms of rotavirus cell attachment involving α2β1 integrin were elucidated further. The direct relevance to rotavirus cell attachment and infectivity of the previously described binding of RRV spike protein VP5* to α2 subunit I domain protein (Graham et al., 2003) was established. It was shown that binding of cellular α2β1 by infectious rotavirus was inhibited by VP5* (aa 248474), which bound to cells in a DGE-dependent manner. This cell-bound VP5* resulted in a reduction in infectivity by both homologous and heterologous rotaviruses that use α2β1 as a cellular receptor.

Rotavirus VP5* binds cellular α2β1 via D308 and G309 and competes with infectious virus for cell binding and infectivity
RRV VP5* bound K562 cells at 4 and 37 °C, both via α2β1 and independently of α2β1, α3β1 and α4β1 (Fig. 1). These data extend previous findings with a similar expressed GSTVP5* construct that bound MA104 cells at 4 °C, as binding of this construct was partially inhibited by an anti-α2 antibody (Zárate et al., 2003). The requirement for the DGE sequence to mediate VP5* binding to cellular α2β1 was demonstrated by the inability of GSTVP5*D308A/G309A to bind α2β1 (Fig. 2). As GSTVP5*D308A/G309A did show α2β1-independent binding to K562 cells (Fig. 2), mutation of the DGE sequence did not affect α2β1-independent binding functions. This demonstrated the specificity of the effect of DGE mutation in RRV VP5* on α2β1 binding.

RRV VP5* (aa 248474) permeabilizes membranes and directs the formation of transient size-selective pores without membrane lysis. These functions require the hydrophobic and basic domains, but are independent of α2β1 recognition and have been proposed to mediate rotavirus entry into cells (Denisova et al., 1999; Dowling et al., 2000; Golantsova et al., 2004). The α2β1-independent cell binding by GSTVP5* and GSTVP5*D308A/G309A (Figs 1 and 2) may have been mediated by these domains. Other VP5* domains may also be responsible for integrin-independent cell binding, such as the MA104 cell-binding domains identified in CRW-8 VP5* using a gene-targeted phage display library (Jolly et al., 2001b). Hsc70 interactions could not account for the integrin-independent binding of VP5*, as the VP5* amino-terminal region that binds Hsc70 was not present in the VP5* used in our studies.

GSTVP5* almost completely blocked binding of RRV to α2β1 and dose-dependently inhibited MA104 cell infection by RRV and Wa but not CRW-8. These inhibitory functions of VP5* depended on the presence of the DGE sequence and so were specific for α2β1. The reduction of heterologous rotavirus infection by RRV GSTVP5* depended on the ability of the infecting virus to utilize α2β1, rather than the terminal sialic acid dependence or VP7 serotype of the virus (Ciarlet et al., 2002b; Nagesha & Holmes, 1991). Approximately 40 % of rotaviruses in the pool failed to enter cells productively when competed with VP5*. This is similar to the levels of blockade of rotavirus cell binding and infectivity by antibodies to α2 and monomeric DGEA peptides (Graham et al., 2004) and is consistent with VP5* blockade of α2β1 binding being a major mechanism by which VP5* inhibits viruscell interactions.

The blockade of RRV cell binding and infection by GSTVP5* depended entirely upon the presence of VP5* D308 and G309. Thus, recombinant GSTVP5* shares with infectious RRV a DGE-dependence for cell binding. Taken in conjunction with GSTVP5* binding to the α2 I domain, which was also eliminated by the D308A and G309A mutations (Graham et al., 2003), this is clear evidence that VP5* of infectious RRV binds cellular α2β1 via the α2 I domain in an interaction dependent upon the presence of VP5* residues D308 and G309. Infection by RRV and Wa rotaviruses was blocked to a similar degree by RRV VP5* and depended similarly on the presence of VP5* D308 and G309. This is the first time that any competition with Wa for infectivity by a protein from RRV has been demonstrated. Wa blocks infection by the RRV nar3 mutant, but nar3 does not affect Wa infectivity (Méndez et al., 1999). Wa binds recombinant α2β1 on K562 cells and Wa infectivity is substantially reduced by DGE-containing peptides (Graham et al., 2003, 2004). The new findings here are a crucial component of the accumulating evidence that Wa also binds to cellular α2β1 integrin via the α2 I domain utilizing VP5* residues D308 and G309. It is also possible that GSTVP5* may inhibit infectious virus binding or infection through the peripheral membrane domain, the hydrophobic domain or an undefined cell-binding domain, if these functions are dependent on the presence of D308 and G309 or prior VP5* binding to α2β1.

Apparently conflicting findings on the ability of rotaviruses to bind α2β1 are reconciled by the demonstration that VP5* bound to cells at 37 °C, but not 4 °C, inhibits virus infectivity
The blockade of virus binding and infection by RRV GSTVP5* shown in Figs 3 and 4 contrasts with earlier reports that RRV binding or infection in MA104 cells is unaffected by recombinant RRV GSTVP5*. RRV GSTVP5* was reported to inhibit RRV infectivity only when MA104 cells were neuraminidase-treated (Zárate et al., 2000b). As the cell-binding partner(s) for VP5* and virus were not identified, the role of α2β1 was not investigated. These earlier studies were performed using MA104 cells treated with VP5* at 4 °C (Zárate et al., 2000a, b, 2003, 2004). In agreement with these reports, it was found here (Fig. 3) that treatment of MA104 cell monolayers with GSTVP5* at 4 °C had no effect on RRV infection. However, we demonstrated that treatment with GSTVP5* at 37 °C resulted in dose-dependent inhibition of RRV and Wa infection (Fig. 4). RRV binding to α2β1 on α2-K562 cells was also almost completely blocked by GSTVP5* at 37 °C (Fig. 3). This GSTVP5* bound α2β1 on α2-K562 cells at 4 °C by flow cytometry (Fig. 1) and binding of a similar construct to MA104 cells at 4 °C was partially dependent on α2β1 (Zárate et al., 2003); thus, the inability of GSTVP5* to block infection when added at 4 °C was not due to a lack of GSTVP5* binding to cellular α2β1 at this temperature.

It is clear from the above that rotavirus cell binding and infection should be measured using cells in their natural state of adhesion and at physiologically relevant temperatures when possible, and that results of experiments conducted under other conditions should be interpreted with caution. The inability of RRV VP5* at 4 °C to compete with RRV binding that was reported previously was interpreted to mean that RRV binds to the cell mainly through VP8*, as RRV VP8* at 4 °C was found to reduce RRV cell binding (Zárate et al., 2003). However, the demonstration (Fig. 3) that RRV VP5* competes at 37 °C with RRV for α2β1 binding is evidence that RRV binds to the cell through VP5* recognition of α2β1. To explain the inability of RRV VP5* at 4 °C to compete with RRV infectivity, it was proposed that RRV can attach to the cell via VP8* and then efficiently displace already bound recombinant VP5* (Zárate et al., 2000b). However, it was shown in Figs 3 and 4 that VP5* effectively competes with RRV for cell binding and infection at 37 °C, so VP5* is not displaced by RRV under these conditions. In fact, the reverse is the case. In addition, VP5* competition with RRV is for binding to α2β1 (Fig. 3), definitively establishing cellular α2β1 as a VP5* ligand that is important for rotavirus cell binding and infectivity. The conclusion that trypsin activation is necessary for RRV nar3 mutant cell attachment via VP5* also needs to be re-evaluated, as RRV GSTVP5* was reacted with cells at 4 °C in that study (Zárate et al., 2004).

Implications of the temperature dependence on the ability of VP5* to compete with rotavirus infectivity
The lack of GSTVP5* competition with rotavirus infectivity at 4 °C shows that interaction between GSTVP5* and α2β1, which results in blockade of virus infection, requires energy and/or active cellular processes. A likely scenario is that VP5* bound to cellular α2β1 undergoes conformational change between 4 and 37 °C to a structure presenting the DGE sequence analogously to virions primed to engage α2β1. Virions used in cell-binding studies are trypsin-activated at 37 °C, which primes virus for entry by triggering a rearrangement stabilizing the VP4 spikes as upright dimers (Crawford et al., 2001). A further, unknown event during entry has been proposed to trigger a second structural rearrangement in VP5* to produce the trimeric form seen in a crystal structure of VP5*. This second rearrangement could involve a 180° rotation of the VP5* antigen domain, including the DGE sequence (Dormitzer et al., 2004). GSTVP5* bound to α2β1 may take up the primed conformation, or a conformation involved in the second rearrangement, at 37 °C but not at 4 °C. Also, α2β1 undergoes conformational changes and lateral movement in the cell membrane at 37 °C, including activation, that may facilitate stable VP5* binding in the correct conformation for inhibition of viruscell binding and infection (Luque et al., 1996). Supporting this, incubation of cells with anti-α2 antibodies at 37 °C but not at 4 °C results in antibody blockade of RRV cell binding, and activation of α2β1 at 37 °C increases infectious SA11 rotavirus binding to α2β1 (Graham et al., 2003).

Our results show that virion VP5* binds cellular α2β1, an important receptor for rotavirus infection, and that VP5* additionally binds cells independently of α2β1. Rotavirus cell attachment and entry also involve virus recognition of several other cellular components. Further analysis of these processes is important for an understanding of rotavirus tropism and pathogenesis.

We are most grateful to E. Mackow for provision of RRV VP4 plasmids, S. Warner and B. Crabb for anti-GST serum and Yan Tan for expert assistance in the production of rotavirus GST fusion proteins. This work was supported by Project Grants 208900 and 350252 and Research Fellowship Grants 172305, 251546, 299861 and 350253 (B. S. C.) from the National Health and Medical Research Council of Australia.