The proteins, gp120 and sCD4, were gifted by the National Institutes of Health Research and Reference Reagent Program. SAXS data were collected at beam line X21 at National Synchrotron Light Source (Brookhaven National Laboratory). The wavelength of the beam was 1.24 Å and the ratio of the sample/detector distance to the diameter of the charge-coupled device detector (D/d) was 9.94M. Samples (30 μl) of the individual proteins, their complex, and phosphate buffer saline pH 7.4 were exposed for 60 s, each at 10°C and at a flow-rate of 27 μL/min. SAXS data collected at several concentrations of hen egg white lysozyme (14.2 kDa; ACROS, Morris Plains, NJ) were run in the same sample cell under identical conditions. Images were circularly averaged, buffer-subtracted, and scaled to obtain relative scattering intensity (I) as a function of q (reciprocal space, q = [4πsinθ]/λ). All experiments were carried out in duplicate.

Guinier analysis, using PRIMUS (6), illustrate a linear fit to the unliganded gp120 data (Fig. 1 A, inset), the sCD4 data (Fig. 1 C, inset), and the SAXS data of their mixture (Fig. 2 A, inset) over the expected q range (where q_max · R_g ≤ 1). This analysis demonstrates monodispersity of scattering particles in solution. An indirect Fourier transform of the SAXS data, using GNOM45 (7), gave the pairwise distance distribution function in real space, P(r). The beam intensity at zero-angle (I₀) determined from GNOM analysis of the lysozyme data, which had been extrapolated to infinite dilution, was measured at 9.7 ± 0.1. This value was used to calculate the concentrations of the individual proteins (0.95 mg/ml of gp120 and 1.25 mg/ml of sCD4). A 1.3:1 molar mixture of sCD4/gp120 was made by adding 10 μl of sCD4 (46 μM) to 47 μl of gp120 (7.8 μM). I₀ analysis of this mixture predicts that 73% of the proteins had formed a 1:1 complex under these conditions. Assuming formation of a complete 1:1 complex, actual protein concentration for the gp120/sCD4 complex would be maximum 0.95 mg/ml (6.5 μM), leaving 1.73 μM (0.04 mg/ml) of sCD4 uncomplexed in solution. (The uncomplexed sCD4 would contribute only 1.8% to the scattering from the mixture). P(r) analysis of the gp120 data (Fig. 1 B) solves for an R_g of 43.6 Å and a D_max of 166 Å. Indirect Fourier transformation of the sCD4 data over a q range of 0.01–0.25 Å⁻¹ solves for an R_g of 22.3 Å and a D_max of 75 Å (Fig. 1 D). This P(r) curve exhibits a peak and a shoulder indicative of the expected bilobal shape for the two domains of sCD4. P(r) analysis of the SAXS data for the mixture (q range of 0.01–0.25 Å⁻¹) solves well for an R_g of 44.4 Å with a D_max of 165 Å (Fig. 2 B). Values for the I₀, R_g, and D_max determined from the SAXS data analysis are summarized in Table 1.

GASBOR22IQW (8) was employed for ab initio shape restorations, whereby the protein structure is represented by an ensemble of dummy residues forming a chain-compatible model and tested for best fit to the I(q) scattering data. The models of unliganded gp120, sCD4, and their complex used 2168, 298, and 2387 dummy residues (packing radius 1.9 Å each), and 1598, 234, and 1598 dummy water molecules, respectively. Each calculation was repeated 10 times, with no shape or symmetry bias (Supplementary Figs. 1–3). For each data set modeled, one representative shape restoration model was selected based on the frequency of occurrence of the major shape features and are shown in Figs. 3 and and4.4.

The shape of the gp120 molecule in solution is characterized by a dense central core and less dense regions on three sides of the core region (Fig. 3 A). Using SUPCOMB13 (9), inertial axes of this gp120 model were overlapped with those for the unliganded SIV gp120 core (Protein Data Bank (PDB) No. 2BF1) (3). The automated superposition of high- and low-resolution models show that SIV gp120 core resides within the dense region of our model (Fig. 3 B). SAXS-based model of unliganded gp120 thus displays the volume, which the variable domains as well as high levels of glycosylation must be occupying in solution. The SIV gp120 protein (44.3 kDa) used for the crystal structure has ~65% smaller molecular weight than that used in our solution studies. Ab initio structure restoration for sCD4 (Fig. 3 C) clearly shows the expected low density between the two domains, also seen in the shape of the P(r) plot. Overlapping the inertial axes of our scattering envelope with those for the unliganded sCD4 (D1-D2) (PDB No. 3CD4) (10) showed excellent agreement in shape and size between the two models (Fig. 3 D).

The scattering envelope computed for the sCD4/gp120 SAXS data (Fig. 4 A) reveals three distinct features: the bilobal shape of the two domains of sCD4, a densely packed region over the top of one lobe of sCD4, and a less dense region that corresponds to the gp120 glycoprotein. The envelope volume is 303,000 ų, which agrees well with the sum total envelope volumes calculated for the individual components (25,900 and 276,000 ų), indicating that flexibility does not change considerably upon binding. Alignment of inertial axes of only the sCD4 domains from the crystal structures (PDB Nos. 1G9M and 1GC1) with the small bilobal arm protrusion in our SAXS-based model places the gp120 core structure from the crystal structures within the dense region of our model (Fig. 4 C). The additional volume around two sides of D1 domain in complex could be arising from the β-strands of gp120 core clamping that lobe as in 1GC1 or 1G9M (1,2). Comparing Fig. 4, B and C, it is clear that the binding of gp120 core to the apex of D1 domain of CD4 drastically reorients the less dense regions of gp120 not seen previously in crystal structures containing only the cores.