Contact Residues and Predicted Structure of the Reovirus Type 3-Receptor Interaction*

Sequence similarity between the reovirus type 3 hemagglutinin (HA3) and a anti-idiotypic monoclonal an- tibody (87.92.6) has been shown to define the site of interaction with a neutralizing (idiotypic) monoclonal antibody (9B.G5) and the cellular receptor for the virus. A synthetic peptide (V, peptide) derived from the anti-idiotypic sequence inhibits viral binding to the receptor. In this study, variants of the VL peptide were utilized to probe specific amino acid residues involved in binding the neutralizing antibody and the receptor. These studies indicate that the " O H groups of several residues are involved in contacting the reovirus type 3 receptor, including Tyr4', Ser50, Ser5', and ThrS3 in the anti-idiotypic sequence, corresponding to Tyr326, Ser3", Ser3", and Ser326 in HA3, respectively. In contrast, only Ser" of the anti-idiotypic sequence, corresponding to Ser3" of HA3, significantly altered neu- tralizing antibody binding. type 3 cells

mary step in the infectious process. The reovirus type 3 cellattachment site is associated with the crl polypeptide, which is also the reovirus type 3 hemagglutinin (HA3).' This HA3 epitope for cell attachment is defined by neutralizing monoclonal antibody 9B.G5 (1,2). A murine anti-idiotypic monoclonal antibody developed against 9B.G5, termed 87.92.6, competes with reovirus type 3 for binding to its specific cellular receptors (reovirus type 3 receptors (Reo3R)) (3,4). The deduced amino acid sequence of HA3 was found to share sequence similarity with a combined determinant comprised of the 87.92.6 heavy and light chain variable region second complementarity-determining regions (CDR 11) ( 5 ) . Synthetic peptides have been utilized to determine whether the sequence similarity 87.92.6 and HA3 defines amino acids essential for epitope recognition. Prior studies (6) indicate that a peptide corresponding to the 87.92.6 light chain CDR I1 (VL peptide) inhibits the interactions between 9B.G5 and HA3,9B.G5 and 87.92.6, 87.92.6 and the Reo3R, and HA3 and Reo3R. A corresponding peptide derived from the HA3 sequence (reo peptide) is capable of eliciting a neutralizing immune response specific for reovirus type 3 (7). These studies strongly suggest that this epitope is directly involved in reovirus binding to the Reo3R.
The epitope defined by the two peptides appears to have additional biological functions. Cross-linking of the Reo3R by either virus or anti-receptor antibodies results in inhibition of cellular growth and down-modulation of the Reo3R (8). These effects were reproduced by dimeric forms of the VL peptide (9). Thus, the interaction between VI. peptide/87.92.6/ HA3 and the Reo3R is also capable of transducing transmembrane signals in Reo3R-bearing cells. Three-dimensional models of the corresponding epitopes on HA3 and in 87.92. 6 have been developed (6).
Studies characterizing the Reo3R on murine L cells implicate sialic acid as a potential site for HA3 binding (10,11). Data consistent with sialic acid participation in reovirus type 3 binding include decreased binding of reovirus type 3 following neuraminidase treatment of L cells as well as inhibition of reovirus type 3 binding by sialylated glycoproteins and sialic acid-containing carbohydrates ( 1 0 , l l ) . Identification of HA3 as the viral polypeptide for sialic acid binding was accomplished by analysis of reovirus reassortment clones derived from reovirus type 1 and 3 genetic crosses, suggesting a direct interaction between HA3 and sialic acid.

9241
In this study, we utilize variants of the VL peptide to determine specific amino acid side chains potentially involved in contacting the Reo3R on L cells. We also utilize sialylated glycoproteins to explore potential interactions between sialic acid and the HA3187.92.6 binding epitope. By combining these observations, we have developed preliminary models of the interactions between sialic acid and HA3 and between sialic acid and 87.92.6.

MATERIALS AND METHODS
Peptides and Proteins-Bovine submaxillary mucin (BSM) and bovine serum albumin (BSA) were purchased from Sigma. All peptides were synthesized by solid-phase methods, deprotected, and released from the resin utilizing anhydrous HF. Peptides were lyophilized and further purified by passage over a Sephadex G-25 superfine column and lyophilized. Purity was assessed by high performance liquid chromatography utilizing a C8 column and a 0-70% acetonitrile gradient. All peptides purified were >90% homogeneous.
Monoclonal Antibodies-Neutralizing anti-reovirus type 3 monoclonal antibody 9B.G5 (murine lgGPa,k) or isotype-matched monoclonal antibody A l l was isolated from the culture supernatant by 50% ammonium sulfate precipitation; dialyzed against phosphatebuffered saline (PBS); bound to a staphylococcal protein A column followed by elution with 0.1 M citric acid, pH 3.5; neutralized with 0.5 M Tris-HCI, pH 8.5; dialyzed against PBS; and concentrated on an Amicon protein concentrator. Purified antibody was radioiodinated by the chloramine-T method (6). Monoclonal antibodies 87.92.6 and H013.4 (both murine IgM,k) were grown as ascites from hybridoma cells. For 87.92.6, the ascites was filtered and stored a t -70 "C prior to use. This was necessary due to the observed instability of this antibody when stored in purified form. For H013.4, the antibody was immunoaffinity-purified as previously described ( 6 ) and utilized at a concentration of 10 pg/ml in PBS, which gave similar binding on fluorescence-activated cell sorting analysis as the similar volume of 87.92.6 ascites.
Radioimmunoassay (RIA)-RIA plates (Dynatech Laboratories, Inc., Alexandria, VA) were coated with peptides by evaporation of varying amounts of peptides in distilled water overnight a t 37 "C. The wells were washed with PBS, blocked with 2% BSA in PBS with 0.1% NaN:%, and washed with PBS. '2sII-Labeled 9B.G5 was added at 50,000-100,000 cpm/well in 1% BSA in PBS for 1-2 h a t 37 "C. The wells were decanted and washed extensively, and the counts/minute bound was determined. Specific counts/minute hound was determined by subtracting counts/min bound to uncoated wells from counts/ minute bound to peptide-coated wells.
Competitive RIA-RIA plates were coated with staphylococcal protein A (Sigma) by incubation of 50 pl/well of a 5 pg/ml solution overnight at 4 "C. The wells were washed with PBS and blocked with 2% BSA, PBS, 0.1% NaN:,; and purified 9B.G5 was adsorbed to the wells by incubation of 50 p1 of a 10 pg/ml solution in 1% BSA, PBS, NaN,, for 1-2 h a t 37 "C. The wells were washed; and competitors were added a t various concentrations in 100 p1 of 0.5% BSA, 0.45% NaC1, phosphate buffer for 1 h a t 37 "C. 12slI-Labeled reovirus type 3 was added for an additional 45 min at 5-10 X lo5 cpm/well. The wells were washed extensively, and counts/minute hound was determined. Specific counts/minute bound was determined by subtracting counts/ minute bound to antibody All-coated wells from counts/minute bound to 9BVG5-coated wells. Percent inhibition of binding was determined by the formula: ((specific cpm bound without inhibitorspecific cpm bound with inhibitor)/specific cpm bound without inhibitor) X 100.
Inhibition of Viral Binding-Murine L cells (fibroblasts) were grown in Joklik's modified minimal essential medium with 10% fetal calf serum and added sodium bicarbonate, penicillin/streptomycin, and L-glutamine (GIBCO). The cells were centrifuged and washed twice in 1% BSA, PBS, 0.1% NaN3, and 5 X 10' cells in 50 p1 were distributed in 2% BSA, PBS, NaN:r-blocked RIA wells. For peptide studies, 50 pl of inhibitor was added in distilled H20 to the cells. For studies of sialylated glycoproteins, 12sII-labeled reovirus type 3 (5-10 X IO5 cpm/well) was preincubated with the proteins (50 pl) dissolved in distilled H20. Following a 30-min incubation, L cells and ' "1labeled reovirus type 3 were combined for an additional 30 min a t 37 "C. The cells were centrifuged and washed three times in ice-cold PBS; and specific counts/minute hound was determined as noted above. Percent inhibition of binding was calculated by the formula noted above.
Flow Cytometry Analysis-The ability of sialylated glycoproteins to inhibit antibody binding to cells was determined by preincubation of the antibody with varying amounts of inhibitor in (in PBS) for 30 min to 1 h a t 23 "C. Cells (either L cells grown as described above or R1.l cells grown in RPMI 1640 medium with 10% fetal calf serum and added antibiotics and L-glutamine from GIBCO) was washed in 1% BSA, PBS, 0.1% NaN:< and resuspended a t 107/ml. Cells (100 pl) were then added in 1% BSA, PBS, 0.1% NaNa; and the incubation was continued for 20-30 min. Ice-cold 1% BSA, PBS, 0.1% NaNa was added, the cells were centrifuged and washed prior to counterstaining with a 1:100 dilution of fluorescein isothiocyanate-conjugated goat anti-mouse lg (Fisher) in 1% BSA, PBS, 0.1% NaN3. The cells were washed twice, and fluorescence intensity was determined as previously described (6). Inhibition of binding was calculated as noted above with A mean channel number utilized in place of counts/ minute.
Neuraminidase Treatment-L cells were centrifuged, washed twice in PBS, and resuspended a t 4-5 X lO5/rnl in PBS. Type VlII neuraminidase from Clostridiumperfringens (Sigma) was dissolved in PBS a t 5 units/ml and added to a final concentration of 20-25 milliunits/ lo6 cells in tissue culture flasks. Following a 1-h incubation, the cells were centrifuged, washed twice in 1% BSA, PBS, 0.1% NaN:<; and processed for flow cytometry as described above. Control cells were treated identically without the addition of neuraminidase.
Modeling of HA3 and V,, Peptides Interaction with Sialic Acid-The strategy for the modeling studies was comparative model building based on immunoglobulin hypervariable geometries followed by molecular mechanics and energy optimization. The development of starting geometries followed two knowledge base approaches: (i) consideration of sequence similarity between fragments and (ii) consideration of loop size and side-chain orientation of hypervariable loops. Starting geometries for the VI, peptide and HA3 epitopes were derived as previously described (6,12). The starting geometry for sialic acid was derived from that of Weis et al. (13). Energy parameters were those of Kollman and co-workers (14). The program DISCOVER (BIOSYM Technologies) was used for the calculations. A sliding scale dielectric was used in the calculations. Only nonbonded interactions for atom pairs closer than 7.5 A were calculated, with typical convergence criteria assumed in optimization. Model building was performed using the program INSIGHT (BIOSYM Technologies).

RESULTS
Interaction of 9B.G5 with V L Peptide and Variants-To determine the relative affinity of binding of neutralizing anti-HA3 monoclonal antibody 9B.G5 to the variants of the VL peptide, a simple solid-phase RIA was performed utilizing peptides dried onto microtiter wells. Binding of directly radioiodinated 9B.G5 to increasing amounts of peptides was quantitated ( Fig. 1). First, binding to peptides of increasing length was determined (Fig. 1, upper). This revealed significant binding of 9B.G5 to a 10-residue peptide (VI,10 peptide) corresponding to the carboxyl-terminal decamer of the VL peptide, without significant binding to similar nonamer, OCtamer, and heptamer peptides ( V L~, V L~, and v1. 9 peptides). This indicates that the 10 amino acids in the VI,10 peptide are the minimal length that elicits significant 9B.G5 binding in this assay.
To examine the contributions of individual amino acid side chains in this interaction, binding to the VL peptide variants noted in Table I was examined (Fig. 1, lower). This revealed that a Ser + Ala substitution at position 1 2 (VLA~B) significantly diminished binding by 9B.  a decamer corresponding to the region of greatest sequence similarity between HA3 and VL CDR 11. It was found that some of the substitutions appeared to enhance binding in the solid-phase assay. This may not have been due to a direct interaction with 9B.G5, but instead may have been due to enhanced binding of the particular peptides to the RIA plates. To eliminate this variable, aqueous-phase competition assays were utilized to determine the ability of these peptides to compete with lZ51-labeled reovirus type 3 for binding to 9B.G5. Peptides were added to SB.G5-coated microtiter wells, and inhibition of subsequent '261-labeled reovirus type 3 binding was determined (Fig. 2). The VLlO decamer was an effective inhibitor in this assay with the shorter heptamer, octamer, and nonamer peptides being ineffective (Fig. 2, upper). This confirms the ability of VLlO to bind to 9B.G5 and infers binding at the site of interaction with HA3.
When the substituted VL peptide analogs were tested, the VLA12 variant demonstrated diminished ability to inhibit the 9B.G5-HA3 interaction (Fig. 2, center and lower). None of the other substitutions demonstrated diminished interaction with the binding site of 9B.G5 by this assay. The Ser "-* Ala substitution at position 5 (VLA5), which is outside the region of sequence similarity, while enhancing binding on solidphase RIA (Fig. 1, lower), had no effect on the aqueous-phase competition assay. Similarly, the Tyr + Phe substitution at position 11 (VLF11) enhanced 9B.G5 binding in the solid phase ( Fig. 1, lower) without affecting competition in the aqueous phase (Fig. 2, lower). Of interest was the increased ability of the Gly + Ala-substituted V&3 and dimeric VLSH peptides to compete for 9B.G5 binding in the aqueous-phase assay. The dimeric VLSH peptide is likely to have increased avidity for the cell-bound Reo3R, whereas altered conformational properties of the VLA13 peptide may affect its apparent increased affinity for 9B.G5 (see below).
Together, these results suggest that deletion of the " O H group from position 12 (corresponding to SersO in the CDR I1 and Ser"' in HA3) diminishes the ability of the peptide to inhibit HA3 binding by 9B.G5. Dimerization of the VL peptide (VLSH peptide) or addition of a "CH, group to position 13 (VLA13 corresponding to Gly" in the CDR I1 and Gly328 in HA3) enhances the ability of the peptide to inhibit HA3 of peptide/well. The mean f S.E. is shown for duplicate determinations on replicate wells. Upper, results for peptides of increasing length; B, results for peptides with specific substitutions (see Table  counts/minute hound is shown as a function of increasing amounts I). The VH peptide sequence is CQGLEWIGRIDPANG. binding by 9B.G5. A decamer (VLlO, corresponding to amino acids 323-332 of HA3 and amino acids 46-55 of the 87.92.6 VL CDR 11) is sufficient to inhibit HA3 binding by 9B.G5.
Interaction of Reovirus Type 3 Receptor with V, Peptide and Variants-The capacity of these peptides to interact with the Reo3R was next determined. This was quantitated by their ability to inhibit binding of lZ51-labeled reovirus type 3 to murine L cells in aqueous-phase binding assays. The short hepta to decapeptides tested were ineffective in inhibiting reovirus type 3 binding to L cells (data not shown). When the variant peptides were evaluated, the Ser + Ala substitution at position 5 (VLA5) had no effect on binding inhibition compared with the unsubstituted VL peptide (Fig. 3, upper). The dimeric form of the VL peptide (VLSH) was more effective on a molar basis than the VL peptide. Thus, these substitutions had similar effects on interactions with the Reo3R and 9B.G5.
In contrast, loss of " O H groups from position 11, 12, 14, or 15 (VLF11, VLA12, VLA14, and VA15) resulted in diminished ability of the peptides to inhibit lZ51-labeled reovirus type 3 binding (Fig. 3, lower). The loss of " O H groups at positions 14 and 15 had a more pronounced effect than that at positions 11 and 12. This is in contrast to binding to 9B.G5, where position 12 seemed to make the major contribution (Figs. 1 and 2). Also at variance with the 9B.G5 results was the diminished ability of the VA13 peptide to inhibit the reovirus type 3-Reo3R interaction, where this peptide had a greater apparent affinity for 9B.G5 in a similar assay (compare Figs. 2 (center) and 3 (upper)).
These results suggest that deletion of " O H groups from position 11, 12, 14, or 15 (corresponding to Tyr4', Ser50, S e P , and Thrs3 in the CDR I1 and to Tyr326, Ser327, Ser3", and S e P 5 in HA3, respectively) reduces the ability of the peptides to inhibit HA3 binding. Deletion of the -OH group at position 5 has no effect, whereas addition a of "CH, group at position 13 reduces the ability of the peptide to inhibit HA3 binding. Dimerization of the VL peptide (VLSH) enhances its ability to inhibit HA3 binding to L cells. Thus, some of the peptide substitutions affect 9B.G5 and Reo3R binding similarly (dimerization and the Ser 4 Ala substitution at position 5), whereas others have different effects (substitutions at positions 11-15).
Sialylated Glycoproteins Interact with Reovirus Type 3 and 87.92.6-Prior studies (10, 11) implicate sialic acid as an important carbohydrate component of the L cell receptor for reovirus type 3. Studies utilizing heavily sialylated glycoprotein BSM demonstrate that BSM binds reovirus type 3 and inhibits binding of reovirus type 3 to L cells (10,11). This inhibition was dependent on the presence of sialic acid moieties in BSM as treatment with agents that removed sialic acid (such as sodium borohydride) diminished the ability of BSM to bind reovirus type 3. In addition, the binding of BSM was dependent on the HA3 molecule as a reovirus type 3 genetic reassortant containing the hemagglutinin of reovirus type 1 (reovirus type 3.HA1) did not bind BSM. Thus, a specific interaction between the sialic acid moieties of BSM and HA3 was inferred.
To confirm the prior studies, and to determine the capacity of sialic acid to bind to the reovirus type 3 analog 87.92.6, the ability of BSM to inhibit binding of reovirus type 3 and 87.92.6 to L cells was assessed. In Fig. 4, inhibition of '"1labeled reovirus type 3 binding to L cells by increasing amounts of BSM is compared to inhibition by BSA, which wells is shown uersus increasing amounts of peptide inhibitors. Upper, results for peptides of increasing length; center and lower, results for peptides with specific substitutions (see Table I). bears fewer sialic acid moieties. Inhibition of reovirus type 3 binding by BSM is seen in a dose-dependent fashion, with BSA having no significant effect on binding. These studies confirm those of Pacitti and Gentsch (10, l l ) , indicating the ability of BSM to inhibit reovirus type 3 binding to L cells. This supports the contention that sialic acid is a component of the cellular receptor for reovirus type 3 on these cells.
Whereas these studies defined a likely HA3 sialic acid interaction, it remained unclear if the 87.92.6 antibody (which mimics HA3) or if the VL peptide derived from 87.92.6 possessed similar interactions. To begin to address these points, the ability of BSM to inhibit 87.92.6 binding was assessed. In Since an isotype-matched IgM monoclonal antibody binding a distinct surface molecule on L cells was not available for comparison, the effect of BSM on the binding of H013.4 to murine R1.1 cells was utilized to control for nonspecific effects. Fig. 5 demonstrates that BSM has no effect on binding of H013.4 to Thy1 molecules present on R1.l cells. In contrast, BSM markedly inhibited 87.92.6 binding to L cells. As an additional control, the ability of BSM to inhibit increasing amounts of antibody binding was also assessed (Fig. 6). Preincubation of increasing amounts of antibodies with a fixed amount of BSM was compared to preincubation with BSA to determine inhibition of binding. Again, specific inhibition was obtained with BSM, and this effect could be overcome by addition of increasing amounts of 87.92.6. This implies an interaction between BSM and 87.92.6 that inhibits 87.92.6 binding to L cells. The ability of increasing amounts of 87.92.6 to overcome the inhibition by BSM suggests that BSM is interacting with 87.92.6 and not with the Reo3R, as the number of L cells was held constant. These studies indicated that whereas BSM specifically inhibited 87.92.6 binding to L cells, it did not inhibit H013.4 binding to R1.1 cells. Thus, a specific interaction of BSM with the binding site of 87.92.6 is implied.
The effects of neuraminidase treatment on L cell binding of 87.92.6 was evaluated by flow cytometry (Fig. 7). A marked decrease in fluorescence intensity followed neuraminidase treatment of cells. This implies a direct interaction between a neuraminidase-sensitive substance on the surface of L cells and the 87.92.6 antibody. Sialic acid is a likely target for the effect of neuraminidase in this experiment.
Molecular Modeling of HA3/87.92.6-Sialic Acid Interaction-In light of the above results, a direct interaction between the VL peptide-defined epitope and sialic acid seems unlikely. The data from variants of the VL peptide (Fig. 3) implicate the hydroxyl groups of residues 11, 12, 14, and 15 of the VL peptide in the binding interaction of the VL peptide to sialic acid. Therefore, models can be developed utilizing Competitive binding analysis was performed as described for Fig. 5. BSM and additional BSA were utilized at a final concentration of 720 pg/ml. Percent inhibition of binding is shown versus increasing volume of antibodies added. structural analysis of potential intermolecular interactions and the peptide with sialic acid that will allow a molecular understanding of the binding sites on 87.92.6 and HA3.
Our previously published structures for reovirus HA3 (residues 323-332) and the 87.92. 6 VL CDR I1 were utilized as starting geometries (Fig. 8). same face of the reverse-turn structures. These hydrogen bonds pair with sialic acid side chains as noted in Table 11. In Table 11  are predicted to form hydrogen bonds with identical partners on sialic acid. These differ from the IHA structure in two respects. The carbohydrate a t position 4 of sialic acid is utilized in these model structures, but not in the IHA structure. Conversely, the glycerol hydroxyl group at position 9 of sialic acid interacts with IHA, but not with the current model structure. Otherwise, the intermolecular distances of the current models are very similar to those in the IHA-sialic acid pair (Table 11).
In the absence of crystallographic or NMR information, the placement of hydrogens in the donor-acceptor scheme in any model is tentative at best. In our models of the possible binding mode of the epitopes, the models reflect subtle differences in the donor-acceptor pairings. For example, the Tyr is a donor in the VL peptide-sialic acid model, whereas it is an acceptor in the HA3-sialic acid model. The differences are indicative of the model geometries as well as the partial energy parameters estimated for the sialic acid moiety. The structures reflect in uucw explorations of conformational space to generate models that are useful and stereochemically valid.
The molecular models indicate that the HA3 and VL conformers have different backbone conformations. However, the spatial orientations of the side chains that define the geometry of the sialic acid binding conformations (modes) are similar. The conformational energy for this binding mode was evaluated utilizing both the HA3 and 87.92.6 VL CDR I1 epitopes (Table 111). This was approached in a sequential fashion. In constructing the model for the HA3 structure, the reverse turn of the immunoglobin NEW heavy chain CDR I1 was utilized as a template, whereas for the 87.92.6 VL CDR I1 epitope, the REI light chain CDR 11 was utilized (6,12). T o examine the sequence-independent properties of the starting conformations, polyalanine forms of these starting geometries were developed, and their stabilization energies were determined (Table 111, Column 1). The sequence-independent effect of the conformation change in assuming the modes of these epitopes was then calculated for these polyalanine forms (Column 2). The root mean square deviations for the transition of the starting geometry to the respective binding mode is -0.68. However, the poly(A1a) forms of the complex conformation are -4 kcal/mol closer in energy than the starting conformations. The root mean square value for the backbone conformation of the structures in Column 2 approaches 3.5. The poly(A1a) forms of the complex conformation are 4 kcal/ mol closer in energy than the starting conformation. The

TABLE 111
Energetics of the VL and HA3 conformational states For column 1, poly(A1a) forms of NEW and REI structures were optimized to examine the relative sequence-independent nature of the low energy forms of the two starting geometries. Energy is in kilocalories/mole. For Column 2, poly(A1a) forms of the HA3 and VL peptides in their sialic acid binding conformations (SLC) are shown. For Column 3, the energy of the starting conformation of the HA3 epitope using NEW 2H geometry (top) and its conformational energy after binding sialic acid (bottom) are shown. For Column 4, the energy of the starting conformation of the V L CDR I1 epitope using the REI 2L geometry (top) and its conformational energy after binding sialic acid (bottom) are shown. For Column 5, the conformational energies for the HAS-mutated conformation of the V L structure complexed with sialic acid (top) and the V L peptide-mutated conformation of the HA3 structure complexed with sialic acid (bottom) are shown. Energy parameters were those of Weiner et al. (14). The program DISCOVER (BIOSYM Technologies) was used for the calculations. conformational energies of the HA3 epitope before and after binding sialic acid are shown in Column 3, whereas the corresponding energies for the VL CDR I1 epitope are shown in Column 4. In both cases, the modeled sialic acid binding mode is the preferred conformation for the derived structures by -16 kcal/mol, with a 0.68 root mean square for their backbones. This energy difference appears to be due to the side chains as the poly(A1a) forms for all of the conformations are similar (Columns 1 and 2). This implies that there is an energy cost for forcing the side chains into the REI conformation (for the VL peptide) or the NEW conformation (for HA3). The degree of backbone structural variability represented by the relatively low root mean square is in keeping with the comparisons of root mean square values of crystallographically determined CDR I1 regions of light chains (data not shown). In Column 5, the corresponding conformational energies for the HA3 and VL CDR I1 epitopes with sialic acid are shown utilizing each other's starting geometries. In this column, the HA3 epitope is in the VL/SLC conformation, whereas the VL CDR I1 epitope is in the HA3/SLC conformation. Comparison of HA3 (VL/SLC) with HA3 (SLC) (Columns 5 and 3, respectively) indicates that the two conformations are close in energy. The same is true for VL (HA3/SLC) and VL (SLC) (Columns 5 and 4, respectively). This indicates that although the low energy forms of the HA3 and VL epitopes modeled with different starting geometries give vastly different backbone structures (root mean square difference of 3.5), the energies of these conformations complexed with sialic acid differ by 1.5 (for HA3) to 2.5 (for the VL peptide) kcal/mol. Whereas the REI-type conformation may be more preferred for both HA3 and VL structures, the closeness in energy of the two structures may indicate that either conformation may be observed 50% of the time. Thus, whereas it is impossible to distinguish the true binding mode, several potential geometries are suggested by these analyses.
The interaction models between the HA3/87.92.6 VI. CDR I1 epitopes and sialic acid are shown in Fig. 9 (A and B ) . The superposition of the contact residues with the two modeled epitopes rotated -90" with respect to one another (similar to Ref. 6) is shown in Fig. 9C. Dynamics run of these structures before sialic acid binding indicate that intramolecular hydrogen bonds involved in stabilizing the reverse-turn structures involve most of the residues involved in binding sialic acid, with the possible exception of Tyr. This may result in a new lower stabilization energy than would otherwise be expected. The pairing of the hydroxyl group corresponding to ThrI6 of the VL peptide with the charged carboxylic acid group of sialic acid may account for its relatively large contribution to the binding interaction (Fig. 3). Similarly, the pairing of the hydroxyl group corresponding to SerI4 of the VL peptide with the carbohydrate hydroxyl group at position 4 of sialic acid would be expected to form a relatively strong hydrogen bond. In contrast, the Tyr"-glycerol -OH pair at position 8 results in entropic loss from the glycerol side chain losing conformational flexibility, whereas the Ser"-acetamido " N H utilizes a less polar " N H group. This could result in weaker interaction energies for these pairs, accounting for the lesser contribution of the VL peptide hydroxyl groups at positions 11 and 12 in binding.
However, binding of synthetic peptides to 9B.G5 and the Reo3R reveals differences in binding strategies to these moieties. A 10-amino acid peptide corresponding to the region of sequence similarity between HA3 and the 87.92.6 VL CDR I1 inhibits binding of HA3 with 9B.G5 (Fig. 2), but does not inhibit HA3 binding to the Reo3R on L cells. Longer versions of these peptides are currently under development to evaluate the minimum-sized peptide necessary to inhibit binding of reovirus type 3 to L cells. Within this region, the " O H from Ser5' of the CDR I1 (corresponding to Ser327 in HA3) is likely to directly interact with 9B.G5, as indicated by the diminished ability of the VLA12 peptide to interact with 9B.G5 (Figs. 1 and 2). In contrast, "OH groups from Tyr49, Ser50, Ser", and Thr"< of the CDR I1 (corresponding to Tyr326, Ser327, Ser3'"', and Ser3'5 in HA3, respectively, are all likely to directly interact with the receptor structure as VLF11, VLA12, VLA14, and V*,A15 peptides all have diminished ability to interact with the Reo3R. Thus, whereas general epitope bound by 9B.G5 and the Reo3R represented by the VL peptide and its analogs seems to encompass the same stretch of amino acids, specific intermolecular interactions involved in binding 9B.G5 may differ from those involved in binding the Reo3R. Since small peptides in solution are thought to exist in a multiplicity of transient conformational states in dynamic equilibrium, it is of interest to examine the conformational possibilities of structural analogs of VH and VL reverse-turn loops that have shown biological activity in an effort to establish design criteria in the development of biologically active peptides derived from antibody templates. Unlike previous studies of the conformational properties of CDR loops that were interested in defining the accuracy of prospective predictions to actual crystallographic structures (17,18), this approach is more concerned with an analysis of the local energy minimum of analogs of well-defined Ig crystallographic templates. In this regard, the differences in binding the V d 1 3 peptide demonstrated in Figs. 2 and 3 are also of interest. The Gly + Ala substitution of this residue results in increased binding to 9B.G5 (Figs. 1 and 2), but diminished binding to the Reo3R (Fig. 3).
In an effort to explain these observations in terms of molecular structure, molecular dynamics calculations (employing starting geometries suggested by the molecular modeling studies) were performed for the VL and v~A 1 3 peptides. These preliminary calculations (data not shown) indicate that the introduction of Ala at position 13 restricts the in uucuo conformational flexibility of the VLA13 peptide. One structural interpretation of these calculations is that whereas both the VL and VLA13 peptides can populate conformations that represent the 9B.G5 binding mode, the energy cost for conformational adjustment in adopting the 9B.G5 binding mode conformation from the respective low energy "solution" structures may be greater for the VL peptide than for the VLA13 peptide. In other words, whereas the V d 1 3 peptide is less conformationally flexible than the VL peptide, its low energy solution conformation may be close to the 9B.G5 binding conformation, and relatively less conformational adjustment is needed to assume this conformation in comparison with the VL peptide. Conversely, the VLA13 peptide may lack the conformational flexibility needed to easily assume the Reo3R binding mode. This implies that somewhat different conformations of the VI, peptides may be optimally bound by 9B.G5 versus the Reo3R.
Analysis of Reo3R on Different Cells-Previous data from this laboratory biochemically characterized the Reo3R on some cells (19)(20)(21). The receptor of murine R1.l (thymoma) cells has been characterized utilizing anti-idiotypic anti-reovirus type 3 polyclonal serum (anti-Id3), which resembles 87.92.6 in its binding interactions. Anti-Id3 identified a -65-kDa glycoprotein with a PI of 5.8-6.0 on the surface of R1.l cells. This protein was shown to possess binding characteristics similar to those of the P-adrenergic receptor (20,21). Treatment of the R1.l Reo3R with inhibitors of glycosylation (including neuraminidase) did not inhibit the ability of anti-Id3 to immunoprecipitate the receptor (19,20). Interestingly, whereas R1.l cells bind anti-Id3, 87.92.6, and reovirus type 3 in a specific/saturable manner, these cells are not infected by reovirus type 3.' The Reo3R on murine L cells is sensitive to neuraminidase treatment. These cells are easily infected by reovirus type 3, and they possess little if any @-adrenergic receptor-like binding activity (22). Thus, by several criteria, the Reo3R on L cells may be distinct from that on R1.l cells. It is noteworthy that BSM and neuraminidase treatment do not appreciably inhibit 87.92.6 binding to R1.l cells.* Thus, these cells may possess a distinct receptor for reovirus type 3. This receptor may divert virus from sialic acid to a distinct binding site, which may not allow virus internalization, or may divert the virus to a intracellular compartment nonpermissive for subsequent steps in the infectious cycle. This mechanisms could play a role in influencing the tissue tropism of reovirus type 3 or other viruses. Implications of Structural Studies-Sialic acid is a common carbohydrate on the surface of many mammalian cells. Viruses have been shown to utilize sialic acid as both a receptor and substrate. Prior structural analysis of the IHA-sialic acid interaction (13) indicates hydrogen binding to four groups on sialic acid. Our study implicates three of these four groups as potential hydrogen bond partners with reovirus HA3. This D. H. Rubin, manuscript in preparation. may indicate a ready geometric availability of these groups for developing intermolecular interactions with biologically important molecular structures. Whereas this study derives the hydrogen bond partners from model structures, similar amino acid residues are implicated as contact residues in this model, as were found in the crystallographically determined IHA-sialic acid structure (Table 111). This may indicate a propensity for certain amino acid residues to form hydrogen bond partners with the sialic acid groups, suggesting a chemical basis for virallantibody binding. However, the specificity of binding for viruses to sialic acid may also relate to other aspects of the sialylated glycoproteins a s well as additional interaction sites on the virus.
The utility of molecular modeling coupled with examination of peptide analogs of binding sites allows development and verification of specific structural data of biological importance. This approach allows development of testable hypotheses regarding binding interactions. These can be probed by additional experimentation, and more refined models can be developed. By utilizing these complementary techniques, substances with specific binding activity can be developed. This should allow rational development of biologically active compounds.