Determination of the Affinity and Kinetic Constants for the Interaction between the Human Virus Echovirus 11 and Its Cellular Receptor, CD55*

The biochemical properties of the molecular interactions mediating viral-cell recognition are poorly characterized. In this study, we use surface plasmon resonance to study the affinity and kinetics of the interaction of echovirus 11 with its cellular receptor decay-accelerating factor (CD55). As reported for interactions between cell-cell recognition molecules, the interaction has a low affinity (K D ∼3.0 μm) as a result of a very fast dissociation rate constant (k on∼105 m −1·s−1,k off ∼0.3 s−1). This contrasts with the interaction of soluble ICAM-1 (sICAM-1, CD54) with human rhinovirus 3 which has been reported to have a similar affinity but 102–103-fold slower kinetics (Casasnovas, J. M., and Springer, T. A. (1995) J. Biol. Chem.270, 13216–13224). The extracellular portion of decay-accelerating factor comprises four short consensus repeat domains (domains 1–4) and a mucin-like stalk. By comparison of the binding affinity for echovirus 11 of various fragments of decay-accelerating factor, we are able to conclude that short consensus repeat domain 3 contributes ∼80% of the binding energy.

Echovirus 11 (EV11) 1 is a member of the genus Enterovirus, family Picornaviridae. While most enterovirus infections either are asymptomatic or result in mild symptoms, EV11 infection has been associated with disorders of the central nervous, cardiac, respiratory, and gastrointestinal systems (2). The interaction of the virus with a specific cell surface receptor is a prerequisite for cell infection, and individual members of the Picornaviridae have evolved to use many different cell surface molecules for attachment and entry (3)(4)(5)(6). EV11 strain 207 (hereafter referred to as EV11) is one of several enteroviruses that use decay-accelerating factor (DAF, CD55) as a cell surface receptor 2 ; anti-DAF antibodies block infection of cells by these viruses in vitro (7,8). DAF is a widely expressed glycosylphosphatidylinositol-anchored cell surface glycoprotein that protects cells from damage by autologous complement-mediated lysis (reviewed in Ref. 9) by accelerating the decay of the C3/C5 convertases (hence DAF). DAF is also commonly used as a cell surface receptor by various pathogenic viruses and bacteria, thereby providing an initial site of host-pathogen interaction. Both the complement-regulatory and pathogen recognition functions are located within the membrane distal portion of the molecule which contains four contiguous, 60-amino acid-long short consensus repeats (SCR) (10 -12). The SCR domains are connected to the glycosylphosphatidylinositol anchor by a Ser/ Thr-rich stalk containing many O-linked oligosaccharides (13,14). The position of the virus-binding site on DAF appears to vary, even between closely related viruses. For example, SCR domains 2 to 4 (DAF 234 3 ) are required for binding and infection by most DAF-utilizing picornaviruses (e.g. Coxsackie B3 (15) and echovirus 7 (7,11,16)) whereas domains 1 and 2 are sufficient for Coxsackie A21 binding (17).
In the present study, we have used surface plasmon resonance to measure the affinity and kinetics of the binding of soluble DAF fragments, expressed in the yeast Pichia pastoris (16), to EV11. Our results indicate that DAF binds to EV11 with a low affinity and very fast kinetics and that most of the binding energy is contributed by the DAF SCR domain 3.

EXPERIMENTAL PROCEDURES
The expression of soluble DAF in the yeast P. pastoris has been described in detail elsewhere (16). In brief, soluble recombinant DAF (hereafter termed DAF) and DAF domain deletion mutants were expressed with a carboxyl-terminal oligohistidine tag and purified to Ͼ95% purity using nickel-nitrilotriacetic acid columns (Qiagen, Dorking, United Kingdom). The proteins migrated at the expected size on reducing and nonreducing SDS-PAGE analysis, indicating that they did not form disulfide-linked multimers (data not shown). The concentrations of the DAF constructs were calculated from the absorption at 280 nm using extinction coefficients (DAF 1234 ϭ 36,840 M Ϫ1 cm Ϫ1 ; DAF 234 ϭ 29,870 M Ϫ1 cm Ϫ1 ; DAF 12 ϭ 17,780 M Ϫ1 cm Ϫ1 ; DAF 23 ϭ 17,780 M Ϫ1 cm Ϫ1 ; DAF 34 ϭ 19,060 M Ϫ1 cm Ϫ1 ) computed from the amino acid composition (Trp, Tyr, and Cys) on the ExPASY server (18). The recombinant DAF fragments were judged to be correctly folded by two criteria: they had NMR spectra typical of proteins with SCR domains; they were able to inhibit the binding of a range (ϳ30 in total) of different CD55 monoclonal antibodies to erythrocytes (19).
EV11 was grown in confluent monolayers of HT29 human intestinal epithelial cells. Cells were infected at a multiplicity of infection of 5-10, and infection was allowed to proceed for ϳ16 h. Cells remaining attached * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  to the flasks were removed by scraping, and the medium was frozen and thawed three times. Debris was removed by low speed centrifugation, and the virus was pelleted through a 30% sucrose layer. The pelleted material was then subjected to two rounds of rate zonal centrifugation on 15-45% sucrose gradients. All sucrose solutions contained 10 mM Tris-Cl, pH 7.4. Virus was pelleted from gradient fractions following dilution in 10 mM Tris-Cl, pH 7.4, and finally resuspended in the same buffer. Purified virus was concentrated to 8 mg/ml using Centricon 10-kDa cutoff microconcentrators (Millipore, Watford, UK) Surface plasmon resonance experiments were performed on a BIA-core2000 (BIAcore AB, Stevenage, UK). EV11 was covalently immobilized to the carboxylated dextran matrix on the surface of CM5 sensor chips via primary amino groups using the amine-coupling kit (BIAcore AB) as directed (20) with the following modifications. After the activation step, purified virus was injected at 80 g/ml in 10 mM sodium formate (pH 3.0) for 5 min. In separate experiments, we showed that incubation at low pH does not affect virus infectivity (data not shown), indicating that such treatment does not disrupt virion structure. This is not unexpected since the virus replicates in the intestine, having passed through the very low pH environment of the stomach. Different levels (1,300 -16,200 response units or RU) of virus were immobilized by varying the length of the activation step from 30 s to 5 min. Unless otherwise indicated, all experiments were performed at a flow rate of 40 l/min at 25°C using as running buffer commercially obtained (BIAcore AB) HEPES-buffered saline (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20).

RESULTS
Binding of recombinant DAF to EV11 was measured by injecting the proteins, such as DAF 234 , through a BIAcore flow cell with EV11 covalently coupled to the sensor surface as well as through a mock-coupled control flow cell in which the sensor surface had been subjected to the coupling reaction in the absence of EV11 (see, e.g., Fig. 1A). The traces, or sensorgrams, reveal that a background response is observed when DAF 234 is injected (Fig. 1A, bar) over the control (mock-coupled) surface. This background response is seen because the BIAcore measures the refractive index near the sensor surface and therefore detects any changes in bulk refractive index of the injected sample. It does not represent nonspecific binding. A much larger response is observed when DAF 234 is injected over the EV11-coupled flow cell. Subtraction of the sensorgram obtained in the control flow cell from the latter sensorgram gives the actual binding response (Fig. 1A, inset). Inspection of these corrected sensorgrams reveals that the kinetics of DAF 234 binding is rapid; binding reached equilibrium within 10 s of the start of the injection, and dissociation was complete within 20 s of the end of the injection. When DAF 12 was injected, an identical sensorgram was observed in both the EV11 and control flow cells, indicating that this construct does not bind EV11 (data not shown). These results establish that soluble recombinant DAF can bind directly and specifically to EV11 covalently coupled to the BIAcore sensor surface. The ratio of the maximal level of bound DAF 234 to the level of covalently immobilized EV11 (both measured in RUs) varied from 0.05 to 0.08. This can be calculated to be 30 -50% of the theoretical maximum if it is assumed that there are 60 binding sites per virion and that the virus and DAF 234 have molecular masses of 8 ϫ 10 3 and 21 kDa, respectively. A somewhat lower binding stoichiometry (14 -18%) was observed by Casasnovas and Springer (1) for sICAM-1 binding to human rhinovirus 3 (HRV3) immobilized on the BIAcore surface. Submaximal bind-ing is likely to be a consequence of occlusion or disruption of binding sites when the viral particles are covalently coupled to the sensor surface.
Similar sensorgrams were obtained when DAF 1234 , DAF 23 , and DAF 34 were injected over EV11 (data not shown). The equilibrium binding responses were measured during injection of at least five concentrations of the construct varying by at least one order of magnitude, and the data were analyzed using BIAevaluation 3.0 software (BIAcore AB). Nonlinear curve fitting of a simple Langmuir model of the association (A ϩ B 3 AB) to these data (Fig. 1, B and C) yielded values for the dissociation constant, K D (ϳ3 M for the longer constructs). The K D was also determined from Scatchard plots of the data (Fig. 1C, inset). Identical results were obtained irrespective of the order of injections (low to high concentrations or vice versa), indicating that the coupled EV11 was stable for the duration of the experiment (data not shown). Table I summarizes the mean and range of between 5 and 10 K D determinations for each construct that bound EV11. Using the relationship Ϫ⌬G ϭ RT ln K D we calculated the corresponding binding energy for each construct and used these values to estimate the contributions of each domain to the total binding energy. DAF 234 and DAF 1234 had the same binding energy, indicating little or no contribution from domain 1 (Table I). DAF 23 interacted with a ϳ0.7 kcal⅐M Ϫ1 lower binding energy than DAF 234 , indicating that domain 4 contributes at most ϳ0.7 kcal⅐M Ϫ1 . By similar comparison of DAF 34 with DAF 234 , it is evident that domain 2 contributes at most ϳ1.1 kcal⅐M Ϫ1 . Thus, domain 1, 2, and 4 together contribute at most ϳ1.8 kcal⅐M Ϫ1 (or ϳ20%) of the total binding energy, whereas SCR domain 3 contributes the remaining ϳ80%.
Because the full binding activity resided in DAF 234 (Refs. 11 and 16 and this study), this construct was used for a more detailed analysis. Measurement of the affinity at different temperatures revealed that there was a decrease in the affinity as the temperature increased from 15 to 40°C (Table II). This indicates that the interaction is exothermic at physiological temperatures. A (van't Hoff) plot of ln(1/K D ) versus 1/T was not linear (data not shown), indicating that the enthalpy varies with temperature.
Binding on the BIAcore is frequently affected by mass transport limitations which can lead to an underestimate of the kinetic constants (e.g. see Schuck (21)). We sought to avoid this by performing the experiments at high flow rates (40 l/min) with low levels of coupled EV11. Several different DAF concentrations were injected (Fig. 1B). The association (k on ) and dis-   Fig. 1B), providing the k on and k off values shown in Table III (once again, these values are the average after five repeated experiments with the errors reflecting the variation between individual experiments). The same k on and k off values were obtained at different flow rates (data not shown) and with two different levels of virus immobilized to the sensor surface, demonstrating that binding under these condition is not limited by mass transport. Furthermore, the calculated K D values agree well with the K D determined by equilibrium binding (Table I).

DISCUSSION
The affinity constants reported here for DAF binding to EV11 are much lower than affinities measured for biological interactions involving soluble macromolecules such as hormones, cytokines, and antibodies but are typical of interactions between cell-cell recognition molecules (22). This is appropriate since the interaction between echovirus virions and cell surface DAF is likely to be highly multivalent; the virus surface presents 60 identical sites due to the symmetry of the capsid.
There is one other reported study of the affinity and kinetics of an intact virus with its receptor (1). Although Casasnovas and Springer reported a similar affinity for sICAM-1 binding to HRV3, a virus belonging to the same family as EV11, the kinetics were much slower (see below). A higher affinity (K D 40 -400 nM) has been reported for the binding of the human immunodeficiency virus envelope glycoprotein gp120 to its cell surface receptor CD4 (23). The affinity of DAF 1234 and DAF 234 for EV11 are essentially the same, whereas DAF 23 and DAF 34 bind with affinities between 3-and 6-fold lower (Table I). Comparison of the calculated binding energies of all these constructs indicates that SCR domain 3 contributes ϳ80% of the binding energy (Table I). In this context, it is noteworthy that two of the antibodies that most effectively inhibit in vitro infection by the closely related echovirus 7 (1H4 and 3D11) have been shown to bind domain 3 (7,11). Furthermore, Escherichia coli Dr and related adhesins have also been shown to bind DAF SCR domain 3 (12).
The k on for the interaction between EV11 and DAF 234 falls within the usual range for macromolecular interactions (22), while the k off is unusually fast, a feature characteristic of interactions between cell-cell recognition molecules (22). These results differ from the findings of Casasnovas and Springer (1), who reported kinetic constants for the sICAM-1/HRV3 interaction that were two to three orders of magnitude slower (k on 130 -2450 M Ϫ1 .s Ϫ1 and k off 0.0013 s Ϫ1 ). Casasnovas and Springer (1) proposed that the unusually slow k off rate was either due to the relatively inaccessible nature of the sICAM-1 binding site on HRV3 (the tip of sICAM-1 binding in a depression on the surface of the virus capsid (24)) or because of a requirement for conformational change in the HRV3 binding site before sICAM-1 binding can occur. In contrast, we show that a domain in the "middle" of DAF is crucial for binding to EV11, making it unlikely that DAF binds in a similar way, i.e. to a deep depression on the surface of EV11. The binding of EV11 to DAF can also be fully described using a simple 1:1 binding model, consistent with the "docking" of two preformed sites. In contrast, the sICAM-1/HRV3 interaction exhibited biphasic binding kinetics, suggesting a more complex binding scheme. Finally, analysis of the temperature dependence of the sICAM-1/HRV3 interaction indicated it was endothermic, which is somewhat unusual for a protein-protein interaction, whereas our analysis of the EV11/DAF interaction indicates that it is exothermic.
In conclusion, we have shown that DAF binds to EV11 with a low affinity and very fast kinetics. These binding characteristics are typical of the molecular interactions mediating cellcell recognition but differ substantially from the sICAM-1/ HRV3 interaction, the only other whole virus cellular-receptor interaction studied to date (1). We also show that the bulk of the binding energy is contributed by the third SCR domain.