The Single Ligand-binding Repeat of Tva, a Low Density Lipoprotein Receptor-related Protein, Contains Two Ligand-binding Surfaces*

The receptor for avian sarcoma/leukosis virus subtype A (ASLV-A), Tva, is the simplest member of the low density lipoprotein receptor family containing a single ligand-binding repeat (LBR). Most LBRs contain a central Trp (Trp33 in Tva) that is important for ligand binding and, for the low density lipoprotein receptor, is associated with familial hypercholesterolemia. The Tva ligand-binding module contains a second Trp (Trp48) that is part of a DEW motif present in a subset of LBRs. Trp48 is important for ASLV-A infectivity. A soluble Tva (sTva) ligand-binding module is sufficient for ASLV-A infectivity. Tva interacts with the viral glycoprotein, and a soluble receptor-binding domain (SUA) binds sTva with picomolar affinity. We investigated whether Tva, a retroviral receptor, could behave as a classic LBR by assessing sTva interactions with the universal receptor-associated protein (RAP) and comparing these interactions with those between sTva and its viral ligand (SUA). To address the role of the two Trp residues in Tva function, we prepared sTva harboring mutations of Trp33, Trp48, or both and determined the binding kinetics with RAP and SUA. We found that sTva behaved as a “normal” receptor toward RAP, requiring both calcium and Trp33 for binding. However, sTva binding to SUA required neither calcium nor Trp33. Furthermore, sTva could bind both RAP and SUA simultaneously. These results show that the single LBR of Tva has two ligand-binding sites, raising the possibility that other LBRs may also.

The low density lipoprotein receptor (LDLR) 2 family is a large group of evolutionarily conserved proteins found in species as varied as roundworms, insects, birds, and mammals (1). The proteins were originally recognized for their ligand uptake capabilities (2). LDLR family members also have important roles as transducers of extracellular signals in embryonic development, maintenance of the nervous system, and epithelial cell function (reviewed in Refs. [3][4][5][6][7]. They also serve as receptors for a variety of pathogens (8 -13). LDLRs bind their ligands through a series of ligand-binding repeats (LBRs). One or more clusters of tandem LBRs are found on the different LDLR ectodomains. Ligand/receptor specificity appears to be achieved by the employment of different combinations of LBRs by different ligands (14 -17).
The structures of many individual LBRs have been solved by either x-ray or NMR analysis (18 -24). Although the LBR sequences vary significantly except for a small number of highly conserved residues, they exhibit a remarkable conservation of tertiary structure. The overall structure, formed by three conserved disulfide bonds, consists of two loops linked by a bridge containing a one-turn helix, with variably oriented N-and C-terminal chains. An acidic cluster chelates a Ca 2ϩ ion, and coordination of this Ca 2ϩ ion is critical for module structure (22,(25)(26)(27). A conserved Trp resides on the bridge and coordinates the calcium ion through its backbone carbonyl (22,23,28). Each LBR also has a hydrophobic bundle that helps to stabilize the N-terminal loop. Mutations of the Cys residues, acidic cluster, and/or bridge Trp of LDLR LBRs are found in patients with familial hypercholesterolemia, suggesting the importance of these residues in receptor function (29 -31). Although many of these mutations affect folding and/or processing of the LDLR, the Trp mutations appear to function at the level of binding and recycling (31). Each LBR folds independently and does not interact with adjacent LBRs (18,21,32,33), allowing the study of individual LBRs.
Recent structural studies have revealed a role for the LBR bridge Trp and a conserved Lys in its respective ligand in receptor interaction. For example, in the crystal structure of the LDLR ectodomain at the pH of ligand release, a Lys from the ␤-propeller domain is in contact with the bridge Trp of LBR5 (33). A similar ligand-receptor contact is observed between a Lys of the receptor-binding domain of human rhinovirus-2 viral protein-1 and the bridge Trp of a model LBR (34), demonstrating the use of this LBR-binding surface by a viral pathogen. An interaction between a Lys from apoE and a Trp from an LBR-mimetic antibody has been observed (35) and has been modeled for the N-terminal region of apoE and LBR5 of the LDLR (36). The involvement of the bridge Trp in ligand binding has also been demonstrated for other ligand/LBR pairs (37,38).
Lys residues critical for LBR binding have been identified in a variety of ligands (39 -44).
Receptor-associated protein (RAP) is a universal chaperone of LDLR family proteins. It is required for proper folding of nascent receptors and inhibits premature binding of their natural ligands (reviewed in Ref. 45). When applied extracellularly, RAP inhibits ligand interaction and pathogen uptake (46,47) and can be used as a receptor agonist (46). RAP has three LBR-binding domains that bind two to three LBRs within a given receptor in a non-cooperative manner (48,49). The ability of RAP to bind the LBRs of all LDLR family members and to inhibit their natural ligand binding suggests that, despite the variations in specific sequence and surface charge among the various LBRs and the differential specificities of each receptor for its cognate ligand, individual LBRs fold similarly and interact with ligands by a common mechanism.
Avian sarcoma/leukosis virus is the prototype oncovirus. Avian sarcoma/leukosis virus subtype A (ASLV-A) infects avian species that express its receptor, Tva (tumor virus A). Tva is an LDLR family member that contains a single LBR. The natural ligands and cell function of Tva remain unknown. For ASLV-A, binding to Tva at a target cell surface via its fusion glycoprotein, EnvA (the envelope glycoprotein of ASLV-A), unleashes a series of conformational changes in EnvA that embed the EnvA fusion peptide in the cell membrane, thereby initiating the infection process (50 -52). The isolated LBR of Tva, soluble Tva (sTva; the 47-residue LBR of Tva), is sufficient for this activation (51)(52)(53). EnvA consists of a trimer of heterodimers. Each dimer is composed of a receptor-binding domain, SUA (the receptor-binding subunit of EnvA), and a fusion-mediating domain. A soluble form of SUA binds sTva with picomolar affinity, and binding induces a conformational change in SUA (54). sTva contains two Trp residues in its LBR, Trp 33 (the bridge Trp) and Trp 48 in the C-terminal loop. Mutations of the bridge Trp (Trp 33 ) increase sTva misfolding, but do not abolish SUA binding (55). On the other hand, mutagenesis of Trp 48 revealed the requirement for an aromatic amino acid at this position for ASLV-A infectivity (56). A subsequent study showed that mutation of Trp 48 to Ala renders Tva unable to induce the conformational changes in EnvA that lead to fusion (51). Although hydrophobic interactions (54) and charge interactions (56,57) have been proposed to be important for binding, there is no structural information on the SUA/sTva binding interface. Furthermore, it has not yet been established whether sTva exhibits normal LBR function.
To assess the ability of sTva to function as a normal LBR and to further characterize SUA/sTva binding, we compared the binding of RAP and SUA to the isolated Tva LBR, sTva. Because of the importance of Trp residues in both SUA function and familial hypercholesterolemia pathogenesis, we also prepared sTva mutants in which one or both Trp residues were mutated. We first characterized the properties of these mutants by CD and intrinsic tryptophan fluorescence (ITF) and then compared their binding to RAP and SUA. We found that the Trp mutations did not significantly alter overall sTva structure. Our data also revealed that RAP binding to sTva required both calcium and the bridge Trp (Trp 33 ), mirroring conditions for RAP binding to other LBRs. In contrast, SUA binding was best in the absence of added calcium and did not require Trp 33 . We further found that RAP and SUA could not compete for sTva binding and also that they could bind sTva simultaneously. These results indicate that RAP and SUA use independent binding sites. Thus, a simple single LBR has two ligand-binding sites, raising the possibility that other LBRs may as well.

MATERIALS AND METHODS
Preparation of SUA, sTva, and RAP-SUA was prepared as described previously (54). Briefly, S2 cells stably expressing SUA were maintained in serum-free insect medium-1 (Sigma) supplemented with 10% heat-inactivated fetal bovine serum, 1ϫ penicillin/streptomycin, 1ϫ L-glutamine, 1ϫ pyruvate, and 300 g/ml hygromycin (all from Invitrogen). For expression of SUA proteins, cells were removed from the dishes; washed twice with phosphate-buffered saline containing calcium and magnesium (MediaTech, Inc.); and diluted in serum-free insect medium-1 supplemented with 1ϫ penicillin/streptomycin, 1ϫ L-glutamine, 1ϫ pyruvate, and 1% EX-CYTE (Serologicals Corp.) at 1-2 ϫ 10 6 cells/ml. Two days later, CuSO 4 was added to a final concentration of 1 mM. Cells were fed with additional medium as needed to keep them from reaching stationary phase and harvested 5 days post-induction. SUA was purified from culture supernatants by consecutive TALON (Clontech) and then SP-Sepharose (Amersham Biosciences) chromatography. Samples were concentrated to ϳ1 mg/ml and stored at 4°C.
The preparation of sTva has been described previously (28). Briefly, cultures of pMAL-sTva were grown in LB/ampicillin medium, induced with 0.3 mM isopropyl ␤-D-thiogalactopyranoside, and harvested 19 h later. Cells were lysed in sonication buffer (50 mM Tris-HCl, 1 mM EDTA, and 200 mM NaCl, pH 8.0) using a French press; the soluble material was diluted 1:4 with water; and proteins were separated by Q-Sepharose chromatography. The pMAL-sTva-containing fractions were pooled, bound to amylose (New England Biolabs Inc.), and eluted with 10 mM maltose. pMAL-sTva fractions were pooled and refolded in oxidation buffer (50 mM Tris-HCl, 100 mM NaCl, and 20 mM CaCl 2 ) to which 3 mM oxidized and 0.3 mM reduced glutathione were added just prior to use. The refolded protein was then concentrated, cleaved with Factor Xa, and purified by reverse-phase chromatography. Lyophilized samples were stored at room temperature until needed. sTva as used here refers to the LBR of the quail homolog of Tva.
cDNA for the expression of RAP as a glutathione S-transferase (GST) fusion protein and anti-RAP polyclonal antibody were the generous gifts of Dr. Dudley K. Strickland. GST-RAP was purified as described previously (58). Briefly, BL21 cells were transformed with pGEX-RAP DNA. Cultures were grown in 2X YT/ampicillin medium; harvested; and lysed in 0.02 M HEPES, 0.1 M KCl, 1 mM dithiothreitol, and 0.5 mM EDTA, pH 7.5. GST-RAP was isolated by GSH-Sepharose chromatography and subsequent purification over Q-Sepharose. This GST-RAP fusion protein will be referred to as RAP throughout, except where relevant to the discussion that it is the full-length GST-RAP fusion protein we are using.
Mutagenesis-The sTva mutations W33G, W33F, W48A, W48F, and 2W2F were made by site-directed mutagenesis using the QuikChange mutagenesis kit (Stratagene) according to the manufacturer's instructions. The existence of the correct mutation and the fidelity of the rest of the gene were confirmed by sequencing. Each mutant was prepared according to the protocol described above for sTva.
Endoglycosidase H (Endo H) Treatment of SUA-Concentrated fractions of SUA were treated with Endo H (New England Biolabs Inc.) for 3 h at 37°C in buffer A (25 mM HEPES, 5% (v/v) glycerin, 250 mM NaCl, and 0.02% sodium azide, pH 7.2). The extent of the reaction was followed by silver staining. Endo H-treated SUA was separated from Endo H by size exclusion chromatography on a Superdex G-200 column (Amersham Biosciences).

Determination of the Calcium Ion Concentration of Buffer Solutions-
The calcium concentration of various buffers prior to calcium addition was determined by ratiometric analysis of the excitation fluorescence characteristics of Fura-2 pentapotassium salt (catalog no. F-1200, Molecular Probes). For these measurements, a 2 M stock solution of Fura-2 was prepared in 10 mM EGTA, 100 mM KCl, and 30 mM MOPS, pH 7.2, and added to samples to a final concentration of 2 nM. The fluorescence was then measured at 510 nm while scanning excitation wavelengths from 250 to 450 nm in a Jobin Yvon Fluorolog-3 spectrofluorometer equipped with an F-3004 Peltier sample cooler maintained at 20°C. Both excitation and emission monochrometors were set to 2-nm band pass to optimize signal strength, and a Schott KV450 filter was used in the emission beam to eliminate second-order diffraction effects and to increase the signal-tonoise ratio. The K D for Fura-2 was empirically determined using the calcium calibration buffer kit 2 (catalog no. C-3009, Molecular Probes) according to the manufacturer's instructions and using Equation 1, where R max is the F 340 /F 380 ratio at 39 M free calcium, R min is the F 340 /F 380 ratio at 0 M free calcium, R is the F 340 /F 380 ratio of the sample, F 380(max) is the fluorescence at 380 nm excitation of 0 M free calcium, and F 380(min) is the fluorescence at 380 nm excitation of 39 M free calcium. A double log plot of log(K D ⅐((R Ϫ R min )/(R max Ϫ R)))⅐(F 380(max) /F 380(min) ) versus log[Ca 2ϩ ] was constructed yielding a linear relationship where the x intercept represented the log(K D ). Using this method, the K D for calcium binding to Fura-2 was determined to be 2.07 ϫ 10 Ϫ7 M. The free calcium content of the relevant buffer solutions was then determined from the ratio of their emission at 510 nm after excitation at 340 versus 380 nm using Equation 1 in conjunction with the empirically determined value for the K D .
Surface Plasmon Resonance-Surface plasmon resonance biosensor data were collected on a Biacore 3000 optical biosensor (Biacore AB, Uppsala, Sweden). Biotin-labeled sTva was manually injected over one well of a streptavidin chip. A parallel well was activated and then deactivated but left unconjugated as a control. Samples were diluted two times either with running buffer (0.01 M HEPES, 0.15 M NaCl, and 0.005% polysor- The brackets denote the three conserved disulfide bonds. Residues forming the hydrophobic patch are green. Ser 19 and Arg 45 , which help stabilize the N-and C-terminal loops, are red. Residues chelating calcium are pink, except Trp 33 , which is blue. Pro 30 , which packs against Trp 33 in the structure of Tonelli et al. (28), is gray. Trp 48 is cyan. The C-2-C-3 sequence, except for Ser 19 , is white. All other residues are orange. Those residues coordinating the Ca 2ϩ ion with their side chain carbonyls are indicated by blunted arrowheads; those coordinating Ca 2ϩ with their backbone carbonyls are indicated by arrowheads. The mutants used in this study are shown below the sequence. B, representation of the sTva structure (Protein Data Bank accession code 1K7B) (28) showing the general features of the sTva LBR. The Trp residues are shown in space-filling format. Coloring is as described for A. The figure was generated using PyMOL. The N and C termini and the two Trp residues are labeled. bate 20) or with running buffer containing 3 mM CaCl 2 . Kinetic studies were performed on duplicate injections at 25°C at a flow rate of 50 l/min. Samples were injected at 90 s and dissociated at 300 s. The chip was regenerated by injection of 30 l of 50 mM NaOH and 1 M NaCl. The real-time data were fit using a 1:1 binding model after subtraction of buffer effects and any binding to the control chip. For GST-RAP binding, a control with GST alone was tested under each condition; no binding between GST and sTva was observed. Where S.D. values are given, the reported values represent the average of two to five independent experiments. The values for each kinetic parameter were averaged independently.
Circular Dichroism-The CD spectra of SUA were measured on an AVIV 215 CD spectrophotometer scanning at 0.5-nm intervals with an averaging time of 0.5 s/data point. The temperature was maintained at 24°C by a circulating water bath, and the cell was maintained under nitrogen purge. The values from three scans were averaged. The contributions of the buffer were subtracted from all spectra.
Fluorescence Spectroscopy-All fluorescence measurements were taken using a Jobin Yvon Fluorolog-3 spectrofluorometer equipped with a F-3004 Peltier sample cooler controlled by a Wavelength Electronics LFI-3751 temperature controller. The excitation slits were set to a 1-nm band pass, whereas the emission slits were set to 3 nm. All samples were prepared at 5 M, and measurements were made using 4-mm path length quartz cuvettes. All spectra were measured at 22°C. The samples were excited at 295 nm, and spectra were taken from 305 to 410 nm with three-scan averaging. The spectrum of each respective buffer under the respective condition was subtracted before plotting the data. For experiments in the presence of calcium, 1 M CaCl 2 was added to a final concentration of 3 mM in a 5 M solution of protein in buffer A. The solution was stirred in the sample chamber for 4 min and then scanned as described above.
Competition Assays-Competition assays were performed in both directions. Immediately prior to initiation of the experiment, the stock protein solutions were pelleted for 30 min at high speed in a tabletop centrifuge to remove any aggregates that might have formed during storage. (No visible pellets were observed in any of these samples.) sTva that had been biotinylated and separated from free biotin as described above was ] was reported to be 0.9 mM.) C and D, ITF spectra were measured between 305 and 450 nm after excitation at 295 nm in the absence and presence of added calcium, respectively. All spectra were measured at 22°C. The contributions of the buffer were subtracted from all spectra. C, the background [Ca 2ϩ ] was determined to be 0.29 mM. D, CaCl 2 was added to each buffer to a 3 mM final concentration.
bound to avidin-agarose beads. As a control, sTva was omitted from one tube of each set. To assess the ability of RAP to compete for SUA binding, one set of the sTva-bound beads was preincubated for 10 min with increasing concentrations of RAP, and then constant amounts of SUA were added to each tube in the set. To assess the ability of SUA to compete for RAP binding, another set of beads was preincubated for 10 min with increasing concentrations of SUA, and then constant amounts of RAP were added to each tube. Incubations were allowed to proceed for 1 h at 4°C, and the beads were washed four times, boiled in SDS-PAGE sample buffer, separated by SDS-PAGE, transferred to nitrocellulose, and probed with antibodies as described in the figure legends.
Co-immunoprecipitation Assay-SUA and RAP stock solutions were pelleted for 30 min at top speed at 4°C immediately prior to beginning the experiment to remove any potential aggregated material. Anti-V5 antibody was bound to protein G-agarose beads and used to capture SUA. The beads were then washed to remove any unbound SUA. Solutions of increasing amounts of RAP in RAP buffer (20 mM HEPES, 50 mM NaCl, and 3 mM CaCl 2 , pH 7.5) containing a constant amount of sTva were preincubated and then added to the washed SUA-bound beads. Incubations were allowed to proceed for 1 h at 4°C with rotation, and the beads were washed four times, boiled in SDS-PAGE sample buffer, separated by SDS-PAGE, transferred to nitrocellulose, and probed with either horseradish peroxidaseconjugated streptavidin (to visualize sTva) or anti-RAP polyclonal antibody followed by horseradish peroxidase-conjugated anti-rabbit antibody.
Cross-linking of sTva Complexes-SUA, sTva, and RAP (or GST as a control) were mixed in various combinations; 0.1 volume of 10ϫ 1-ethyl(-3-[dimethylaminopropyl]carbodiimide hydrochloride (EDC) buffer (1 M MES and 30 mM CaCl 2 , pH 6.0) was added; and the proteins were allowed to interact during 20 min of rotation at room temperature. Half of the sample was transferred to a clean tube; freshly prepared EDC was then added to a final concentration of 1 mM; and the reaction was allowed to proceed during 2 h of rotation at room temperature. The sample in the original tube was treated with buffer only. Samples were resolved by SDS-PAGE, transferred to nitrocellulose, and probed for sTva with horseradish peroxidase-conjugated streptavidin.

RESULTS
To determine the importance of Trp 33 and Trp 48 in sTva structure and interaction with RAP and SUA, we generated the point mutations depicted in Fig. 1A. The positions of Trp 33 and Trp 48 in the NMR structure are shown in Fig. 1B (28). We expressed and purified the resulting proteins as described under "Materials and Methods." Our refolding protocol gave single symmetrical peaks for all mutants as analyzed by reversephase chromatography (data not shown), suggesting that each mutant folded into a single uniform structure.
Before beginning our ligand binding studies, we assessed the structures of wild-type sTva and each of the mutants by CD and ITF ( Fig. 2 and Table 1). An important contribution to LBR structural stability is provided by chelation of a Ca 2ϩ ion (20,32,38,59), and Ca 2ϩ coordination is a common requirement for ligand binding (59 -62). The final step in the purification protocol, reverse-phase chromatography, removes the Ca 2ϩ ion from sTva (27). This enabled us to resuspend the lyophilized protein in buffers (with or without added calcium) to examine the effect of calcium on each sTva. We performed the CD and ITF experiments in both the absence and presence of added calcium. The CD spectra of all sTva proteins were similar, with a minimum at ϳ206 -208 nm, except for 2W2F in added calcium, the minimum of which shifted to near 200 nm (Fig. 2, A and  B). The presence of a positive peak in the CD spectra at 228 nm, most obvious in added calcium, precluded meaningful analysis of the tertiary structure by this technique. The calcium-induced increase in this peak is most likely due to bound  a Values are the difference in the molar ellipticity at 288 nm between the spectra obtained for protein reconstituted in phosphate-buffered saline with Ca 2ϩ and that reconstituted in phosphate-buffered saline without Ca 2ϩ . b The contribution to the amplitude of the positive elipticity at 228 nm observed for 2W2F (assumed to represent the contribution to this peak from effects other that Trp asymmetry) was subtracted from the value for the amplitude of the increase at 228 nm for each sTva, and the resulting values were normalized to wild-type sTva (100%). c The fluorescence at max for each mutant was normalized to that for wild-type sTva (100%).
calcium inducing a more rigid structure that places both Trp 33 and Trp 48 in more asymmetric environments (63). As expected from this interpretation, 2W2F had a minimal 228 nm peak. The ITF spectra for all sTva proteins except 2W2F (which remained at the base line as expected for a protein with no Trp residues) were nearly symmetrical, with max values between 350 and 359 nm (Fig. 2, C and D; and Table 1). In general, the max values for the Trp 33 mutants were more red-shifted than those for the Trp 48 mutants ( Table 1), suggesting that Trp 48 is in a more polar environment compared with Trp 33 . The addition of calcium to the buffer caused an increase in the amplitude of all ITF spectra and the appearance of a positive peak at 228 nm in the CD spectra (Fig. 2, B and D), indicating that the addition of calcium decreases the flexibility of both Trp residues. Our CD and ITF data suggested that all of our sTva proteins were similarly folded, able to chelate calcium, and suitable for further characterization.
Both Calcium and sTva Trp 33 Are Required for RAP Binding-To assess whether or not sTva can function as a normal LBR, we evaluated its interaction with RAP using surface plasmon resonance spectroscopy. RAP did not bind sTva or any of the sTva mutants unless CaCl 2 was added to the binding buffer. Fig. 3 shows a representative trace of RAP binding in the absence (light gray trace) and presence (dark gray trace) of calcium. (The background free [Ca 2ϩ ] in the "calcium-free" buffer was determined to be 0.29 M; the added [Ca 2ϩ ] was 3 mM (final concentration) for all surface plasmon resonance measurements.) Table 2 gives the kinetic parameters from the data obtained at multiple concentrations as described under "Materials and Methods." RAP bound wild-type sTva and mutants W48F and W48A in 3 mM CaCl 2 , but could not bind any of the Trp 33 mutants regardless of whether calcium was present or not. In control experiments, no binding was observed for GST at either calcium concentration, confirming that binding was through the RAP portion of our GST-RAP fusion protein. Thus, RAP binding to sTva required both calcium and Trp 33 . These requirements are similar to those observed for the binding of RAP to other LBRs (37,48,59,60). They further support the importance of the bridge Trp (Trp 33 in sTva) in binding to a traditional ligand.
Neither Calcium nor sTva Trp 33 Is Required for SUA Binding-We next investigated how the Trp mutations and calcium would affect the binding of sTva to a soluble form of its viral ligand, SUA. SUA was expressed and purified as described previously (54). In striking contrast to RAP binding to sTva and its mutants (Table 2), SUA bound sTva whether or not calcium was present (Table 3). In fact, the addition of calcium actually decreased the affinity between SUA and wild-type sTva.

TABLE 4 Effect of partial deglycosylation of SUA on its binding to sTva and its mutants
Experiments were performed and data were fit as described under "Materials and Methods." Values were calculated as described for Table III. Values without standard errors are from a representative experiment. NB, no binding observed.  Table 3.
Although the total change in the K D was only 10-fold, the k off was increased by 1000-fold; this large change in the off-rate was partially compensated by a 100-fold increase in the on-rate. We next examined the binding kinetics of each sTva mutant with SUA (Table 3). Again, in contrast to RAP binding, all of the sTva mutants bound SUA whether or not calcium was present. In particular, the W33G, W33F, and 2W2F mutants were all able to bind SUA, and the W33F mutant displayed nearly wild-type binding kinetics in the absence of calcium. The affinities of the other mutants for SUA were 10 -100-fold lower than that of wild-type sTva, except for W48A in the absence of calcium, which was 10 4 -fold lower. Interestingly, in the presence of 3 mM calcium, the decrease in W48A was only 100-fold, i.e. unlike wild-type sTva, the SUA binding affinity of W48A increased upon the addition of calcium to the buffer. The binding kinetics for the other mutants were not significantly altered by the addition of calcium. In summary, SUA not only binds sTva in the absence of added calcium, but also does not require Trp 33 for binding. Thus, SUA displays unique LBR binding characteristics. Partial Deglycosylation of SUA Alters sTva Binding Kinetics-The folding and binding of RAP to LBRs do not require its single glycosylation site (64). In contrast, specific glycosylation sites in SUA are required either for proper folding or for sTva binding (65). Peptide N-glycosidase F treatment of SUA rendered the protein insoluble (data not shown). However, partial deglycosylation by treatment of either SUA or an SUA-IgG construct with Endo H does not abolish sTva binding (66). 3 We therefore treated SUA with Endo H, purified the Endo H-treated SUA by size exclusion chromatography, and determined its binding kinetics with sTva and the sTva mutants by surface plasmon resonance (Table 4). In general, with the exception of W33G and, to a lesser extent, W33F, there was little effect of calcium on the binding kinetics of Endo H-treated SUA with any of the sTva mutants. In contrast, no binding between Endo H-treated SUA and W33G was seen in the absence of calcium. The other significant result was that Endo H-treated SUA lost the high affinity binding to wild-type sTva.   (Table 3). RAP bound to beads in the absence of sTva (lane 1). Immunoprecipitated proteins were resolved by SDS-PAGE, and RAP was visualized with anti-RAP polyclonal antibody followed by horseradish peroxidase-conjugated anti-rabbit antibody. RAP and SUA Can Bind sTva Simultaneously-We next investigated whether SUA and RAP could compete for binding to sTva. For these experiments, we took an immunoprecipitation inhibition approach (Fig. 4A). We bound biotinylated sTva to avidin-agarose beads and investigated whether increasing amounts of RAP could prevent binding of SUA (Fig. 4B) or whether increasing amounts of SUA could prevent RAP binding (Fig. 4C). As shown in Fig. 4B, no competition of SUA binding could be observed even at RAP concentrations as high as 130 M (100-fold above its K D ). When increasing concentrations of SUA were used to "compete" for RAP coprecipitation by sTva, a similar result was obtained (Fig. 4C). SUA was unable to inhibit RAP coprecipitation by sTva even at 1 M (10 5 -fold above its K D ). These data suggested that SUA and RAP might be binding simultaneously to sTva. To test this possibility, we performed a "sandwich" immunoprecipitation (Fig. 5A). We prebound SUA by its V5 tag to anti-V5 antibody-bound protein G-agarose beads. We then mixed a constant amount of biotinylated sTva with increasing concentrations of RAP and added these solutions to the SUA-bound beads (Fig. 5A). The beads were washed extensively to remove any unbound material. As shown in Fig. 5B (upper panel), a constant amount of sTva bound to the SUA beads regardless of the RAP concentration. We also observed RAP coprecipitation by SUA-bound sTva. The amount of RAP that could be coprecipitated increased in a concentration-dependent manner (Fig. 5B, lower panel). Thus, not only do RAP and SUA fail to compete for binding to sTva, they can both bind sTva simultaneously.
To confirm the observation that both RAP and SUA can bind sTva simultaneously, we used a crosslinking approach. As sTva does not contain any lysines, we were limited in our choice of cross-linking reagents. EDC is a bifunctional reagent that can interact at one end with carbonyl groups and at the other with amines. We prepared mixtures of various combinations of SUA, biotinylated sTva, and GST-RAP (our fusion protein) or GST in buffer containing 3 mM CaCl 2 and allowed them to interact. We kept GST-RAP between 19 and 35 M (14 -25-fold its K D for sTva, the maximum concentration we could achieve in these mixtures) and varied the SUA concentration between 100 nM (2 ϫ 10 3 -fold its K D for sTva) and 10 M (2 ϫ 10 6 -fold its K D for sTva, but closer to a 1:1 molar ratio with GST-RAP) with similar results. sTva was the limiting reagent in all of these mixtures. After co-incubation, the mixtures were treated with EDC or buffer, and the proteins and cross-linked protein complexes were resolved by SDS-PAGE, transferred to nitrocellulose, and probed for biotinylated sTva with horseradish peroxidase-conjugated streptavidin. A representative blot is shown in Fig. 6. Although we did not observe the binary complex in the SUA/ sTva mixture (lane 6) and only a small amount of crosslinked sTva and GST-RAP (lane 10), we did observe bands consistent with the ternary SUA⅐sTva⅐GST-RAP as well as both the SUA⅐sTva and sTva⅐GST-RAP binary complexes when all three components were mixed together and crosslinked (lane 8, bands indicated by asterisks; marked S/T/gR, S/T, and T/gR, respectively, to the left of the blot). Interestingly, more of the binary sTva⅐GST-RAP complex could be cross-linked in the ternary mixture than in the binary mixture (compare lanes 8 and 10). No complexes were observed when GST was substituted for GST-RAP (lanes 4 and 12,  respectively). Although we chose a concentration of EDC that minimized nonspecific cross-linking, a cross-linked GST-RAP oligomer was observed in these experiments (lanes 2 and 8, upper band). The heavily staining band migrating between 10 and 15 kDa on this blot (lanes 3-12) is uncross-linked sTva. All of the RAP-containing samples contained a number of cross-linker-independent bands that reacted with horseradish peroxidase-conjugated streptavidin, including a densely staining band at ϳ23 kDa that does not correspond to a Coomassie Blue-stained band in the GST-RAP preparation (Fig. 6B). The ability to cross-link the ternary SUA⅐sTva⅐GST-RAP complex (lane 8) confirms that SUA and RAP can bind sTva simultaneously. The inability to cross-link a SUA⅐sTva⅐GST complex (lane 4) confirms that GST-RAP binding is through the RAP portion of the protein. The ability to cross-link the binary complexes more efficiently in the context of the ternary complex than in either binary complex further suggests that sTva may adopt different conformations in these different contexts.

DISCUSSION
The LBR sTva contains two Trp residues: Trp 33 , which is important for folding (55), and Trp 48 , which is important for ASLV-A binding and infectivity (51,52,56). Trp 33 corresponds to the conserved bridge Trp present in the majority of LBRs. The bridge Trp is part of the ligand-binding surface in all LBR structures determined so far (33,34,36,38). Trp 48 is part of a DEW motif present in a subset of LBRs. We set out to determine whether Tva, a retroviral receptor, behaves as a classic LBR and whether the common ligand-binding face is important for viral binding. We therefore assessed Tva interactions with a classic ligand (RAP) and compared these interactions with those between Tva and its viral ligand (SUA). To address the role of the two sTva Trp residues in these interactions, we mutated each of the them singly and together (Fig. 1A) and compared the interactions of these mutants with RAP and SUA.
Because the natural receptor for Tva remains unknown and because SUA binding does not conform to the canonical LBR ligand-binding requirements (the presence of calcium and the bridge Trp), the ability of Tva to function as a normal LBR has been uncertain. We have shown here that RAP could bind sTva and that this binding required both calcium and the bridge Trp, Trp 33 ( Fig. 3 and Table 2), as has been seen for RAP binding to other LBRs (37, 59 -61, 67). Our findings that RAP binds sTva under classic conditions places Tva firmly in the LDLR family and supports the possibility that its natural ligand binds at the conventional LBR ligand-binding surface.
Association of SUA with sTva exhibited very different properties than did RAP binding to sTva (Table 3). In particular, SUA binding required neither calcium nor Trp 33 . In fact, SUA bound wild-type sTva with higher affinity in the absence of added calcium than when excess calcium was present. Furthermore, the W33F mutant bound SUA with almost wild-type affinity. The high affinity binding between SUA and sTva observed in the absence of calcium was due primarily to a remarkably slow off-rate that was not observed in the presence of calcium. With 3 mM CaCl 2 in the buffer, the off-rate was increased by 1000-fold, but the on-rate was increased by 100fold. A similar although less profound trend in the effects of calcium on the SUA/sTva binding kinetics was obtained using an SUA-IgG construct (68). Our finding that the affinity of SUA for W48A in 3 mM CaCl 2 was 100-fold lower than that for wildtype sTva correlates well with the reported difference between the apparent affinities of wild-type sTva and W48A (in calciumcontaining buffers/media) for trimeric EnvA (51), suggesting that our data are physiologically relevant. The difference in the on-and off-rates for SUA binding to sTva in the absence or presence of added calcium (Table 3) and the decrease in flexi-bility of sTva upon the addition of calcium (Fig. 2) strengthen a previous suggestion (68) that, in contrast to other LBRs with structures that are not altered by ligand binding (33), SUA may require flexibility in sTva to achieve tight association. Ca 2ϩ coordination may increase the population (lifetime) of sTva in the preferred conformation for binding SUA but prevent conversion to the conformation that binds SUA tightly (releases SUA slowly). Indeed, our ability to cross-link SUA to sTva in the context of the ternary SUA⅐sTva⅐GST-RAP complex, but not in the binary SUA/sTva mixture (Fig. 6A), may be due to alterations in the sTva structure in these different complexes. Surprisingly, the oligosaccharide chains that are removed by Endo H treatment may be involved in molding the SUA-binding conformation of sTva, as their removal rendered SUA unable to bind W33G or to achieve the slow release conformation of wildtype sTva in the absence of Ca 2ϩ (Table 4). Clearly, however, any conformational change in sTva induced by SUA binding must not destroy the RAP-binding surface, as RAP was fully able to associate with SUA-bound sTva (Figs. 5 and 6).
In contrast to SUA/sTva interactions, RAP/sTva interactions displayed a small (Ͻ10-fold) but real increase in affinity for sTva with the Trp 48 mutation (Table 2). Interestingly, mutation of Trp 48 to Phe decreased the off-rate, whereas mutation of Trp 48 to Ala caused an increase in the on-rate (Table 2). We suspect that the improved affinity between RAP and sTva caused by the Trp 48 mutations in sTva is more likely due to adjustments in sTva structure on the Trp 33 face of the molecule than to any direct involvement of Trp 48 in RAP binding. Our CD and ITF data ( Fig. 2 and Table 1) show that, even in the presence of added calcium, the environment of Trp 33 was different for W48F and W48A (note the difference in the height of the 228 nm peak in the CD spectra (Fig.  2B) and the difference in the quantum yield for Trp 33 in the ITF spectra (Fig. 2D)).
As the extracellular milieu is high in calcium, we expect the in vivo binding kinetics between ASLV-A and Tva to be more like those we observed in 3 mM CaCl 2 for SUA and sTva. Indeed, trimeric EnvA/Tva K D values have been reported in the 0.1-50 nM range (51-53, 56, 69), and calcium-containing buffers have been shown to lower the activation barrier for achieving the conformational changes in EnvA associated with activating EnvA for fusion (70). The ability of EnvA to bind sTva quickly to induce the fusion-activating change and to release it quickly to allow subsequent steps in fusion may be critical to EnvA function.
As has been noted (51,56), a Trp at position 48 of Tva, although not absolutely required for SUA binding, enhances this binding significantly (Table 3). In particular, the W48A mutant, a mutant that cannot support infection by ASLV-A (56), had the lowest binding affinity for fully glycosylated SUA in the absence of calcium of any of the mutants tested here (Table 3). In 3 mM CaCl 2 , however, binding was similar to that of 2W2F. It has been observed that, although an aromatic residue is required at position 48 for ASLV-A infectivity, infectivity cannot be directly correlated with binding affinity (56). We found a similar lack of correlation between binding affinity and fusioninducing activity using a liposome floatation assay (52). In this assay, the 2W2F mutant was fully able to trigger membrane bind-ing, whereas the W48A mutant was not. 4 These results suggest that residue 48 resides at the SUA-binding interface.
It has long been recognized that basic residues in LBR ligands are essential for binding (71). For example, Lys residues that are essential for LBR binding have been identified in RAP, human rhinovirus-2 viral protein-1, and apoE (36,39,40). LBR requirements include the LBR fold, Ca 2ϩ chelation, and the bridge Trp. All structural studies to date reveal ligand binding to a common face of LBRs, the face harboring the bases of the two loops and the bridge between them (33,34,36,38). In these structures, Ca 2ϩ coordination orients acidic residues in the LBR for interaction with basic residues in the respective ligands, and the bridge Trp of at least one LBR within a multiple-module motif interacts with a specific Lys in the respective ligand (33,34,36,38). In accordance with these principles, RAP binding to sTva requires both Trp 33 and calcium ( Table 2). On the other hand, SUA/sTva interactions require neither calcium nor the bridge Trp (Table 3) and can neither be competed by RAP (Fig. 4) (73) have minimal effects on ASLV binding and infectivity provided properly folded conformers are tested. Furthermore, extensive mutational analysis has thus far not revealed an essential Lys in SUA, although Arg 213 may be important (72). Given that 1) high affinity binding and functional activity require a bulky hydrophobic residue at position 48 of sTva (Refs. 51, 56, and 68 and this study), 2) are modulated by the length of the loop between C-2 and C-3 (74), and 3) occlude a hydrophobic surface (54); that 4) Trp 48 , the C-2-C-3 loop, the hydrophobic patch, and other residues that have been shown to modulate ASLV-A infection (55, 74 -76) occur on the opposite face of sTva from the normal ligandbinding face (Fig. 7A); that 5) RAP binding is not affected by mutations of Trp 48 ( Table 2); and that (6) SUA and RAP can bind simultaneously to sTva (Figs. 5 and 6), we propose a model in which sTva has two independent binding sites: a common ligand-binding surface used by RAP and a novel ligand-binding surface used by SUA (Fig. 7B). Although our data are consistent with the surface shown in Fig. 7A harboring the SUA interface, confirmation of the actual SUA-binding surface awaits structural data.  The space-filling model of sTva was generated from the atomic coordinates from Protein Data Bank code 1K7B (28) using PyMOL. A, surface of sTva opposite the Trp 33 face. The hydrophobic patch and residues shown to modulate ASLV-A infection are labeled. Coloring is as described in the legend to Fig. 1, except for Pro 12 and Gln 31 , which are salmon. These residues (which are Ser and Leu, respectively, in the chicken isoform of sTva) have been recently shown to alter the infectivity of chicken versus quail cells by ASLV-A harboring specific mutations in SUA (75). B, the two ligand-binding faces of sTva. The common ligand-binding face, predicted to bind RAP, is shown on the right; the novel ligand-binding face, predicted to bind SUA, is shown on the left. Coloring is as described in the legend to Fig. 1.
Our findings identify a novel LBR-binding motif that is independent of both Ca 2ϩ and Trp 33 but involves Trp 48 . Because Trp 48 is part of a DEW motif present in a subset of LBRs, our findings raise the intriguing possibility that other LBRs may also use an alternate face (i.e. not containing the bridge Trp) for ligand binding and may be able to simultaneously bind more than one ligand. If this is so, the SUA/sTva complex may provide a new paradigm for the study of ligand/LBR interactions that are not competed by RAP.