Characterization of the Self Association of Avian Sarcoma Virus Integrase by Analytical Ultracentrifugation*

Retroviral integration protein (IN) has been shown to be both necessary and sufficient for the integration of reverse-transcribed retroviral DNA into the host cell DNA. It has been demonstrated that self-assembly of IN is essential for proper function. Analytical ultracentrifugation was used to determine the stoichiometry and free energy of self-association of a full-length IN in various solvents at 23.3 °C. Below 8% glycerol, an association stoichiometry of monomer-dimer-tetramer is observed. At salt concentrations above 500 mm, dimer is the dominant species over a wide range of protein concentrations. However, as physiological salt concentrations are approached, tetramer formation is favored. The addition of glycerol to 500 mm NaCl, 20 mm Tris (pH 8.4), 2 mm β-mercaptoethanol significantly enhances dimer formation with little effect on tetramer formation. Furthermore, as electrostatic shielding is increased by increasing the ionic strength or decreasing the cation size, dimer formation is strengthened while tetramer formation is weakened. Taken together, the data support a model in which dimer formation includes favorable buried surface interactions which are opposed by charge-charge repulsion, while favorable electrostatic interactions contribute significantly to tetramer formation.


Retroviral integration protein
has been shown to be both necessary and sufficient for the integration of reverse-transcribed retroviral DNA into the host cell DNA. It has been demonstrated that self-assembly of IN is essential for proper function. Analytical ultracentrifugation was used to determine the stoichiometry and free energy of self-association of a full-length IN in various solvents at 23.3°C. Below 8% glycerol, an association stoichiometry of monomer-dimer-tetramer is observed. At salt concentrations above 500 mM, dimer is the dominant species over a wide range of protein concentrations. However, as physiological salt concentrations are approached, tetramer formation is favored. The addition of glycerol to 500 mM NaCl, 20 mM Tris (pH 8.4), 2 mM ␤-mercaptoethanol significantly enhances dimer formation with little effect on tetramer formation. Furthermore, as electrostatic shielding is increased by increasing the ionic strength or decreasing the cation size, dimer formation is strengthened while tetramer formation is weakened. Taken together, the data support a model in which dimer formation includes favorable buried surface interactions which are opposed by chargecharge repulsion, while favorable electrostatic interactions contribute significantly to tetramer formation.
A distinguishing characteristic of retroviruses is the integration of their genome into host DNA. Integrase (IN) 1 is the enzyme responsible for many of the processes leading to genomic insertion. Because genomic insertion is an essential step in the life cycle of retroviruses, IN has been the focus of intense study (1,2).
Both the full-length primary structure (3) and the crystallographic structures for the catalytic portion of IN from the avian sarcoma virus (ASV) and human immunodeficiency virus-I (HIV-I) have been reported (4,5). Full-length, wild type HIV-I IN exhibits limited solubility in aqueous solvents, making it difficult to conduct solution studies with this protein (6). However, wild type, full-length ASV IN is relatively soluble, making it a more practical protein for studying its aqueous solution behavior (7). ASV IN has a monomer molecular weight of 31,750 (8). Structural analyses suggest that its catalytic core is similar to that of HIV-I IN (5). Furthermore, ASV IN and HIV-I IN are both members of a superfamily of polynucleotidyl transferases that show similar structure and, presumably, use similar mechanisms of catalysis (3). All studies to date indicate that self-assembly is required for IN activity (3). Studies by Jones, et al. (7) showed that ASV IN formed dimers and tetramers in 50 mM Tris (pH 8.0), 750 mM NaCl, and 2 mM ␤ME. Other studies have demonstrated that enzymatically active fragments of ASV IN and HIV-I IN also self-associate (3) and that assembly is required for activity. In addition, a more soluble, mutant, fulllength HIV-I IN has been isolated that exhibits a dimer-tetramer equilibrium when examined in 25 mM HEPES (pH 7.5), 1 M NaCl, 1 mM dithiothreitol, 1 mM EDTA, and 10% (w/v) glycerol (6).
To date, there is no crystal structure available for full-length ASV IN. However, partial proteolysis experiments, deletion mutation results, and sequence similarity with other retroviral integrases, retrotransposons, and bacterial insertion elements suggest that there are three distinct domains in full-length ASV IN: the N-terminal, central core, and C-terminal domains (3). Structural analysis reveals that both the central core and C-terminal domains form dimers (9,10). All three domains have been implicated in the self-association of the full-length proteins. The N-terminal domain contains an HCC, Zn 2ϩ finger motif, and there is evidence that Zn 2ϩ promotes self-association and alters the assembly stoichiometry of mutant, full-length HIV-I IN (11). Deletion mutations indicate that the C-terminal domain is necessary for tetramer formation (6). However, there is no convincing evidence for which domain interfaces are responsible for dimer formation and which are responsible for tetramer formation.
The aim of this study is to better characterize the selfassociation of wild type, full-length ASV IN, and in so doing, to contribute to a better understanding of IN function. Here we report the effects of pH, ionic strength, cation type, and glycerol concentration on full-length, wild type ASV IN dimer and tetramer formation. All of these solvent properties affect the selfassociation significantly, indicating that care must be taken when comparing published results acquired under different conditions. Furthermore, the results suggest that the electrostatic contributions to dimer and tetramer formation are very different, with charge-charge repulsion opposing dimer formation, and favorable electrostatic interactions (e.g. dipole-dipole) favoring tetramer formation.

EXPERIMENTAL PROCEDURES
Materials-ASV integration protein (IN) was purified as described previously (12). Protein was stored at Ϫ20°C in a buffer of 50 mM Tris (pH 7.5), 0.5 M NaCl, 2 mM dithiothreitol, 0.1 mM EDTA, and 40% glycerol. Sterile Centrex cellulose acetate columns for the centrifugal gel exchange were purchased from Schleicher and Schuell, Keene, NH. Sephadex G-25 beads were purchased from Amersham Pharmacia Bio-tech. Anhydrous glycerol, 99.6%, was purchased from the Ultrapure Bioreagent line of J. T. Baker. All other buffer components were ultrapure or reagent grade.
Buffer exchanges for all experiments were performed by centrifugal gel chromatography (13), with the initial eluant used as the optical reference. Protein dilutions were made with the buffer used to prepare the gel column.
Sedimentation Equilibrium-Sedimentation equilibrium experiments were conducted in the Beckman Model E analytical ultracentrifuge at a temperature of 23.3°C (14) with an on-line Rayleigh interferometer (15) using a pulsed laser diode light source at 670 nm. Rotor speeds for all experiments were 20,000, 24,000, 28,000, and 36,000 rpm. Either an AN-D (2 hole), AN-F (4 hole), or AN-G (6 hole) rotor was used. Four solution/solvent pairs were examined in each cell using a 12-mm thick charcoal-filled epon or Kel-F short column centerpiece (16, 17) with sapphire windows.
Experiments at 4 and 37°C were conducted in a Beckman XLI analytical ultracentrifuge. Protein was prepared by centrifugal buffer exchange into 500 mM KCl, 20 mM Tris (pH 8.0), and 2 mM ␤ME. Equilibrium experiments were carried out using the absorbance optical system (280 nm), 4-hole titanium rotor, 6-channel charcoal epon centerpieces, and fused silica windows. Data were acquired as the average of five absorbance measurements at a radial spacing of 0.002 cm and at rotor speeds of 20,000, 24,000, and 28,000 rpm.
All protein solutions were spun in a high speed Beckman microfuge for 5 min to remove any particulate matter prior to cell loading, and serial dilutions for short column experiments were performed as described (17). Protein concentrations were estimated either using the extinction coefficient at 280 nm of 1.83 ml/mg⅐cm as determined by the tyrosine and tryptophan content of the protein (18) or using the "hinge point" method (19).
Equilibrium was established according to the method of Yphantis (20) by subtracting consecutive interference scans of the cell containing the highest viscosity sample. The average equilibrium time for solutions without glycerol was 1.5 h using short column centerpieces and 15 h using 6-channel centerpieces. When high viscosity samples were examined, centrifugation at each speed was conducted for 24 h (short column centerpieces) before equilibrium was checked using consecutive scans made at least 1 h apart.
Buffer densities were calculated from the density increments and corrected to 23.3°C as described (21). A calculated partial specific volume, v , of ASV IN of 0.73 21 ml/g based on amino acid composition was used (21,22). However, for multicomponent solvents, the isopotential partial specific volume, Ј, and not v must be used for data analysis. There was insufficient protein to determine Ј experimentally. Accordingly, the effects of salt and glycerol on the v of other proteins were used to estimate Ј. An approximately linear increase in Ј is expected with both salt (0 -1 M) and glycerol (0 -40% by volume) (22). The worst case in the experiments reported here (40% glycerol, 0.5 M NaCl) would be expected to increase v by 0.01 5 ml/g. The effect of protein concentration on Ј is negligible at the concentrations used and was ignored.
Data Analysis-Data from the ultracentrifuge experiments were edited as described previously (19) and analyzed using nonlinear least squares (NONLIN) methods (23) to provide fitting parameters and estimates of the 95% confidence interval. A model was considered to be correct if it provided an acceptable variance of fit and random distribution of residuals. If more than one model fit these criteria, the model chosen was the simplest one (23,24). Because of the lower precision afforded by the short solution columns used in these studies, it was not possible to extract the monomer molecular weight and multiple association constants routinely. Accordingly, association constants were determined using a model that assumed a known monomer molecular weight of 31,700 and a v of 0.73 21 ml/g, with the buoyancy adjusted for the solvent density. For the cases where the monomer molecular weight and association constants could be estimated simultaneously, the monomer molecular weights were within 12% of the expected value. Furthermore, adjusting the assumed buoyant monomer molecular weight for an uncertaintly in v of Ϯ 0.015 ml/g resulted in changes in the association constants that fell within the confidence interval when the analysis was conducted using v ϭ 0.73 21 ml/g.
Conversion from fringes to weight concentration was performed using a specific refractive increment of 3.333 fringes/mg/ml or an extinction coefficient of 1.83 ml/mg⅐cm. Association constants were calculated as described previously (24).
The dimer-tetramer association constant (K 274 ) was calculated as where K 174 is the measured association constant describing the monomer-tetramer assembly and K 172 is the corresponding constant for the monomer-dimer assembly. The confidence intervals for K 274 were calculated using the worst case combination of the intervals for K 174 and K 274 . The reciprocal of the association constants was used to calculate dissociation energies, where n7m is either 271 or 472, as appropriate.

RESULTS
Previous studies have shown that full-length ASV IN assembly can be modeled as a reversible monomer-dimer-tetramer equilibrium (7). Except in solutions where the glycerol concentration exceeded 8%, this same model provided the best fit to the data (25). At glycerol concentrations exceeding 8%, only the dimer-tetramer equilibrium could be characterized (below).

Solubility of ASV IN as a Function of Ionic Strength-
The solubility of IN was determined by the "hinge-point" method (19) in solvents containing varying concentrations of either Na ϩ or K ϩ . The solubility increases as the ionic strength increases for both cations (Fig. 1). However, at all ionic strengths, ASV IN is more soluble in K ϩ -containing than in Na ϩ -containing buffers. These results suggest that electrostatic contributions are important to ASV IN aggregation. To explore this possibility further, sedimentation equilibrium experiments were undertaken to assess the effect of varying salt concentration and pH on ASV IN self-association.
The Effect of Salt Concentration on ASV IN Self-association at pH 8.4 -Short column sedimentation equilibrium experiments were conducted to determine ⌬G 472 and ⌬G 271 as a function of salt concentration, using either Na ϩ or K ϩ -containing solvents (Fig. 2). For both cations, there is an overall strengthening of dimer formation and weakening of tetramer formation as the salt concentration is increased. These effects appear to saturate, with both dissociation energies becoming nearly constant at salt concentrations above 0.5 M for either cation.
The solubility limit of ASV IN precluded analysis of selfassociation in the ultracentrifuge at physiological salt concentrations. At the lowest ionic strength, however, tetramer formation is the energetically favored species for either cation. These results suggest that tetramer would be the favored species at physiological salt concentrations.
At higher salt concentrations, both dimer and tetramer for- mation are stronger in Na ϩ -containing than in K ϩ -containing solvents. Taken together with the solubility results, these data suggest that tetrameric ASV IN is less soluble than either monomer or dimer under these conditions.
The Effect of pH on ASV IN Self-association in 500 mM KCl-The charge per monomer of ASV IN estimated from its amino acid composition is about ϩ12e at pH 8.4, rising almost linearly to ϩ20e at pH 6.1. Thus, a change in the association behavior might be expected as the pH is varied. To test this, ⌬G 271 and ⌬G 472 were determined over a pH range of 6.1 to 8.4. The free energy of both dimer and tetramer formation were found to be nearly constant, changing only by a few kJ/mol as the pH is increased from 6.1 to 8.4 (Fig. 3). These results indicate that ionic strengths of 500 mM or greater are sufficient to suppress electrostatic effects.
The Effect of Salt Concentration on ASV IN Self-association at pH 6.1-Previous work at ionic strengths well below 500 mM has shown that both the full-length ASV and its catalytic core are essentially enzymatically inactive at pH 6, probably because of a small structural rearrangement (10). To further determine how the self-association of the IN is affected by salt concentration at lower pH, the salt dependence of ⌬G 231 and ⌬G 472 were determined at pH 6.1 (Fig. 4). In general, the protein is more soluble at lower ionic strengths at this lower pH, possibly because of the higher net charge per IN monomer, allowing studies at lower salt concentrations. While the overall dependence of the self-association on salt concentration (Fig. 4) is similar to that observed at pH 8.4 (Fig. 2), at pH 6.1 tetramer is significantly more stable in KCl than NaCl. Furthermore, at physiological KCl concentrations, tetramer is the clearly the favored oligomer at pH 6.1.
The Effect of Glycerol Concentration on ASV IN Self-association-Many studies of both ASV and HIV-I IN include varying amounts of glycerol in the solvent (6, 26 -34). To determine the effect glycerol might have on ASV self-association, a series of experiments were conducted at pH 8.4 in buffer containing 500 mM NaCl and varying amounts of glycerol (Fig. 5). The monomer-dimer equilibrium could be observed only at glycerol concentrations of less than 8%. At higher glycerol concentrations, there was insufficient monomer to detect the monomer-dimer equilibrium, and only the dimer-tetramer association was characterized. Remarkably, the dimer-tetramer equilibrium is virtually unaffected by glycerol concentrations as high as 40%. The fact that the monomer-dimer and dimer-tetramer equilibria exhibit such different sensitivities to glycerol strongly suggests that solvation effects are very different for these two assembly steps. ASV IN Self-association in Solvents with Different Cations-Distinctly different effects are observed on ⌬G 271 and ⌬G 472 when these dissociation energies are determined in solvents containing different cations at 250 mM (Fig. 6). The cation species are arranged in this figure in increasing order of electrostatic shielding potential. Both ⌬G 271 and ⌬G 472 exhibit clear trends, with dimer formation being strengthened slightly, while tetramer formation is weakened considerably, as the cationic shielding potential increases. These results are in accord with the ionic strength effects (Figs. 2 and 4) and indicate that electrostatic interactions contribute measurably and in an opposite fashion to IN dimer and tetramer formation. The more pronounced effect of shielding potential on tetramer formation suggests a larger electrostatic contribution to this interaction than to dimer formation.
The Effect of Temperature on ASV IN Self-association-When examined in 500 mM KCl at pH 8.4, there were only small increases in ⌬G 271 and ⌬G 472 as the temperature was raised from 4 to 37°C (Fig. 7). This suggests that either enthalpic contributions to assembly are small or that there is significant enthalpic/entropic compensation in both assembly steps.

DISCUSSION
The self-association of ASV IN has been examined in a variety of solvent conditions. The assembly was described satisfactorily as a freely reversible monomer-dimer-tetramer equilibrium, except in solvents containing Ͼ8% v/v glycerol, where only the dimer-tetramer equilibrium was observed. As these data show, there are significant changes in the free energies of both steps of assembly as the solvent conditions are varied. Of particular note is the rather dramatic, and opposite, effect that increasing salt concentration has on ⌬G 271 and ⌬G 472 (Figs. 2 and 4), as well as in the different response of ⌬G 271 and ⌬G 472 to changes in cation shielding potential (Fig. 6).
Qualitatively, the salt-dependent increase in ⌬G 271 ( Fig. 2  and 4) with increasing salt concentration is in accord with dimer formation being opposed by charge-charge repulsion between monomers. At the same time the decrease in ⌬G 472 with increased salt concentration suggests that electrostatic interactions favor tetramer formation. However, the shape of these curves, particularly their nearly mirror image nature and the apparent saturation of ⌬G 271 and ⌬G 472 at high salt concentrations, suggest that either specific ion binding or a less specific electrostatically dependent rearrangement in the protein structure may occur. In an attempt to distinguish between these two possibilities, ⌬G 271 and ⌬G 472 were measured at a fixed ionic strength, but using cations of different size and thus different shielding potentials (Fig. 6). In this figure, the trends in ⌬G 271 and ⌬G 472 as the shielding potential is varied are the same as those seen when the ionic strength is varied. This result suggests that the effects are more a consequence of general electrostatic behavior than they are because of specific ion binding.
Varying the shielding potential has a greater effect on ⌬G 472 than it does on ⌬G 271 , which suggests that there is a greater electrostatic contribution to ⌬G 472 than there is on ⌬G 271 (Fig.  6). Consistent with this view is the minimal effect of glycerol on ⌬G 472 (Fig. 5), as might be expected for an interaction that is driven largely by electrostatics and is less dependent on solvation effects. By contrast, ⌬G 271 increases substantially with glycerol concentration, as might be expected for an interaction in which solvation effects and buried surface area are important (35).
Jenkins, et al. (6) have described the self-association properties of a mutant HIV-I IN. In accord with our results on the effect of glycerol (Fig. 5), they observed only a dimer-tetramer equilibrium at 4°C in 10% glycerol, 25 mM HEPES (pH 7.5), 1 M NaCl, 1 mM dithiothreitol, and 1 mM EDTA. Although none of our solvent conditions exactly match these, comparison with our results is invited. They report a tetramer to dimer dissoci- ation constant of 2.2 ϫ 10 Ϫ5 M, which yields a tetramer dissociation energy of 24.7 kJ/mol. This is within the confidence interval for what we observe, which is 22.0 kJ/mol (with a 95% confidence interval of 18.6 to 24.9 kJ/mol) for ASV IN in 1 M NaCl (pH 8.4), at 23.3°C in the absence of glycerol (Fig. 2). That temperature (Fig. 7) (pH) (Fig. 3) and glycerol (Fig. 5) have little effect on tetramer formation in ASV IN suggests that these variables may have similarly weak effects on HIV-I IN assembly.
Nuclear magnetic resonance analysis reveals that the Cterminal domain of HIV-I IN forms dimers (9). Furthermore, Jenkins, et al. (6) have demonstrated that a mutant HIV-I IN lacking the C-terminal domain is unable to form tetramers. Although it is tempting to ascribe the energy of tetramer formation to the contact area between the C-terminal domains, our data suggest that, for full-length ASV IN, tetramer formation may be driven largely by favorable electrostatic interactions and so may reflect more global properties of the protein.
Both HIV-I and ASV IN contain an HHCC, Zn 2ϩ finger motif near their N terminus (3). Lee, et al. (11) have shown that Zn 2ϩ promotes self-association of mutant, full-length HIV-I IN and that the HHCC motif is required for this effect. The addition of 10 M Zn 2ϩ to wild type, full-length ASV IN in 500 mM KCl, 20 mM Tris, 2 mM ␤ME (pH 8.4) resulted in the formation of aggregates of indeterminate stoichiometry that were not in freely reversible equilibrium with lower order oligomers (data not shown). The aggregation precluded accurate analysis of ⌬G 271 and ⌬G 472 in these samples. Analysis of ASV IN H9N, in which a point mutation was made in the HHCC motif, showed a weakening of ⌬G 271 by ϳ6 kJ/mol and a strengthening of ⌬G 472 by ϳ5 kJ/mol (25). It is not possible to interpret such observations further since self-association is a function of the entire protein structure and is sensitive to the surroundings of the protein.