A soluble active mutant of HIV-1 integrase: involvement of both the core and carboxyl-terminal domains in multimerization.

Structural studies of human immunodeficiency virus type 1 (HIV-1) integrase have been impeded by the low solubility of the protein. By systematic replacement of hydrophobic residues, we previously identified a single amino acid change (F185K) that dramatically improved the solubility of the catalytic domain of HIV-1 integrase and enabled the structure to be determined by x-ray crystallography. We have introduced the same mutation into full-length HIV-1 integrase. The resulting recombinant protein is soluble and fully active in vitro, whereas, HIV-1 carrying the mutation is replication-defective due to improper virus assembly. Analysis of the recombinant protein by gel filtration and sedimentation equilibrium demonstrate a dimer-tetramer self-association. We find that the regions involved in multimerization map to both the catalytic core and carboxyl-terminal domains. The dramatically improved solubility of this protein make it a good candidate for structural studies.

Structural studies of human immunodeficiency virus type 1 (HIV-1) integrase have been impeded by the low solubility of the protein. By systematic replacement of hydrophobic residues, we previously identified a single amino acid change (F185K) that dramatically improved the solubility of the catalytic domain of HIV-1 integrase and enabled the structure to be determined by x-ray crystallography. We have introduced the same mutation into full-length HIV-1 integrase. The resulting recombinant protein is soluble and fully active in vitro, whereas, HIV-1 carrying the mutation is replication-defective due to improper virus assembly. Analysis of the recombinant protein by gel filtration and sedimentation equilibrium demonstrate a dimer-tetramer self-association. We find that the regions involved in multimerization map to both the catalytic core and carboxyl-terminal domains. The dramatically improved solubility of this protein make it a good candidate for structural studies.
Integration of a DNA copy of the retroviral RNA genome into a host chromosome is a crucial step in viral replication (1)(2)(3). Two specific enzymatic reactions are involved in integration. In the first reaction, 3Ј processing, two nucleotides are removed from each 3Ј end of the linear viral DNA made by reverse transcription. A subsequent DNA strand transfer reaction then splices these 3Ј ends into the host chromosome. The 5Ј ends of the viral DNA and the 3Ј ends of the host DNA remain unjoined in the resulting integration intermediate (4,5). Completion of integration requires only removal of the unpaired bases at the 5Ј ends of the viral DNA and repair of the single-strand gaps between the viral and host DNA, steps which may be completed by cellular DNA repair enzymes.
HIV-1 1 integrase catalyzes both the 3Ј processing and DNA strand transfer reactions in vitro. Short oligonucleotides that mimic the ends of HIV-1 DNA serve as the substrate for 3Ј processing, and a second oligonucleotide acts as a target DNA for the subsequent strand transfer reaction (6 -8). Integrase will also catalyze an apparent reversal of the strand transfer reaction, termed disintegration (9). In this reaction the viral DNA segment of a branched substrate is liberated and the target DNA segment is resealed.
HIV integrase is comprised of three functional domains. Although the central core domain alone can catalyze the disintegration reaction, both the amino-and carboxyl-terminal domains are necessary for catalysis of 3Ј processing and DNA strand transfer (10 -12). Site-directed mutagenesis has revealed that a triad of highly conserved acidic residues within the core domain is essential for all three catalytic activities of integrase (13)(14)(15). These residues, Asp-64, Asp-116, and Glu-152, comprise the D,D-35-E motif that is conserved in all retroviral and retrotransposon integrase proteins and is also found in some bacterial transposable elements (13, 16 -18).
The limited solubility of HIV-1 integrase has impeded structural studies. However, the mutation of phenylalanine to lysine at amino acid 185 (F185K) within the catalytic core domain of HIV-1 integrase resulted in a soluble protein (19) that enabled the core domain to be crystallized and the structure solved to 2.5-Å resolution (20). We have now introduced the same mutation (F185K) into full-length HIV-1 integrase, together with a cysteine to serine substitution at position 280. The resulting protein is dramatically more soluble than full-length wild-type HIV-1 integrase and retains full activity for both the 3Ј processing and DNA strand transfer reactions. The mutant protein exists in an equilibrium between dimeric and tetrameric species in buffer containing 1 M NaCl. Both the core and carboxylterminal domains are involved in multimerization.

MATERIALS AND METHODS
Construction of Site-directed Mutations-Site-directed mutagenesis was done by overlapping PCR (21) using a two-step procedure as described (19). Plasmid DNA encoding the mutations F185K/C280S within full-length integrase, IN 1-288 /F185K/C280S, was prepared using pINSD (13) as the PCR template DNA. After the second round of PCR, the full-length fragment was digested with NdeI and BamHI and ligated with NdeI-BamHI-digested pET-15b (Novagen, Madison, WI). This placed IN 1-288 /F185K/C280S under the control of a T7 promoter (22) and also encoded a 20-amino acid histidine tag (HT) at the amino terminus of the protein to facilitate rapid purification on a nickel chelating column. The inclusion of a thrombin cleavage site within the HT sequence allowed the removal of 17 amino-terminal residues containing the polyhistidine motif following purification. F185K was also introduced into the pol gene of pNL4 -3, an infectious plasmid DNA clone of HIV-1, using overlapping PCR. Sequential PCR with pNL4 -3 as the starting DNA template produced a 1832-base pair fragment. The fragment was digested with AgeI and PflMI and then ligated with AgeI-PflMI-digested pNL4 -3 DNA. The sequences of all PCR-generated regions were confirmed by DNA sequencing (23).
Expression and Solubility of F185K/C280S Integrase from E. coli-Plasmid encoding IN 1-288 /F185K/C280S was expressed in Escherichia coli strain BL21(DE3) (24) as described (6,13). The solubility of the mutant protein was examined in a crude cell lysate, as follows. Cells were grown in 2 liters of Super broth (Biofluids) containing 100 g of ampicillin per ml at 37°C until the optical density of the culture was between 0.8 and 1.0 at 600 nm. Protein expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.4 mM. After 3 h, the cells were harvested and resuspended in 12 ml of 25 mM HEPES (pH 7.5), frozen in liquid N 2 , and stored at Ϫ80°C. Cells (100 l) were lysed by the addition, to final concentration, of 0.15 M, 0.5 M, or 1 M NaCl, 2 mM dithiothreitol, and 0.3 mg/ml lysozyme, in a final volume of 170 l. After 30 min at 4°C, cells were frozen in liquid N 2 and then thawed at 4°C. Following ultracentrifugation in a Beckman TL-100 Ultracentrifuge for 45 min at 100,00 ϫ g, 10 l of each supernatant was analyzed by SDS-PAGE.
Protein Purification-Frozen resuspended cells expressing IN 1-288 / F185K/C280S from 24 liters of Super broth were thawed and resuspended in lysis buffer (1 M NaCl, 20 mM HEPES, pH 7.5, 2 mM ␤-mercaptoethanol, 0.3 mg/ml lysozyme, 5 mM imidazole) to a final volume of 1 liter. After a 30-min incubation at 4°C, the lysed cells were homogenized with a SDT-1810 Tissumizer (Tekmar), sonicated, and centrifuged in a Beckman Ti-15 batch rotor for 1 h at 40,000 ϫ g. The supernatant was filtered through a 0.25-m filter and applied to a nickel-affinity (Chelating Sepharose Fast Flow, Pharmacia Biotech Inc.) column (5 ϫ 7 cm). After loading, the column was sequentially washed with 2 liters of 20 mM imidazole buffer containing 25 mM HEPES, pH 7.5, 2 M NaCl, and 2 mM ␤-ME, and 1 liter of 60 mM imidazole buffer containing 25 mM HEPES, pH 7.5, 1 M NaCl, and 2 mM ␤-ME. Protein was eluted with a linear gradient of 60 mM to 1 M imidazole, containing 25 mM HEPES, pH 7.5, 1 M NaCl, 2 mM ␤-ME, and 10% (w/v) glycerol. Fractions containing integrase were pooled, and EDTA was added to a final concentration of 5 mM. This protein was then dialyzed against 25 mM HEPES, pH 7.5, 0.5 M NaCl, 0.3 M imidazole, 2 mM ␤-ME, 1 mM EDTA, and 10% (w/v) glycerol. To cleave the HT from integrase, thrombin (Sigma) at 10 NIH units/mg of integrase was added and incubated at 26°C. After 30 min, additional thrombin at 10 NIH units/mg of integrase was added and the incubation continued for another 30 min at 26°C. Thrombin was removed by adsorption to a benzamidine-Sepharose 6B column (Pharmacia). Cleaved protein was dialyzed against 20 mM sodium phosphate buffer, pH 6.2, 0.3 M NaCl, 1 M urea, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol. This protein was loaded on to a Mono S HR 10/10 column (Pharmacia) and eluted with a linear gradient of 0.25 M to 0.6 M NaCl, containing 1 M urea, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol. Fractions containing fulllength protein were pooled and diluted with an equal volume of 1 M urea, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol and concentrated by readsorption to the Mono S column. Protein was then eluted with 1 M NaCl, 1 M urea, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol, dialyzed against storage buffer (1 M NaCl, 20 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM DTT, and 10% (w/v) glycerol), and frozen in liquid N 2 before storing at Ϫ80°C. The concentrations of purified IN 1-288 /F185K/ C280S were determined using a calculated extinction coefficient of 50,460 M Ϫ1 cm Ϫ1 at 280 nm, based on the amino acid composition (25). The corresponding extinction coefficient at 296 nm was determined based on the experimental ratio of A 280 /A 296 . Wild-type integrase was purified as described (13).
Determination of the Multimeric State of IN 1-288 /F185K/C280S-The multimeric state of IN 1-288 /F185K/C280S was determined by both gel filtration and analytical ultracentrifugation. For gel filtration, purified protein was applied to a calibrated Superdex 200 PC 3.2/30 column (Pharmacia) on a Pharmacia Smart TM System. The column was equilibrated with 25 mM HEPES, pH 7.5, 1 M NaCl, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol, and operated at a flow rate of 50 l/min at 4°C. Analytical ultracentrifugation was performed using a Beckman XL-A analytical ultracentrifuge. Data were acquired as an average of 25 absorbance measurements at nominal wavelengths of 280 nm or 296 nm and a radial spacing of 0.001 cm. All sedimentation equilibrium experiments were run in 25 mM HEPES, pH 7.5, 1 M NaCl, 1 mM DTT, 1 mM EDTA, and 10% (w/v) glycerol, using double-sector charcoal-filled epon centerpieces and column lengths of approximately 4 mm. The density of the solvent was measured at 20.00°C (Anton Parr DMA58 density meter) and corrected using standard tables. Sedimentation equilibrium experiments were conducted at various rotor speeds, from 10,000 to 16,000 rpm, at 4.0°C, using loading concentrations of 0.1 (280 nm), 0.3 (280 nm), 1.0 (296 nm), and 1.25 (296 nm) mg/ml. The time required for the attainment of equilibrium was established by running at a given rotor speed until successive scans, taken 12 h apart, were invariant. Typical times taken to reach equilibrium varied from 48 to 72 h when samples were loaded using a precooled rotor and centrifuge. In the course of the experiment, no significant decrease in the concentration of soluble protein was noted. Such losses were observed only at loading concentrations above 5 mg/ml. Experiments were performed at two different rotor speeds for each sample because it has been demonstrated that the simultaneous fitting of data represents a stringent criterion for establishing that the system indeed is reversibly associating (26). Data analyses by mathematical modeling were performed using Sigma Plot 4.16 (Jandel Scientific). Simultaneous weighted nonlinear least squares fitting of the data sets at each concentration were performed using mathematical models of the following form: where A o,A is the absorbance of the protein at a reference point r o , A r is the absorbance at a given radial position r, H represents 2 /2RT ( is the angular speed in rads s Ϫ1 , R is the gas constant, T the absolute temperature), and E is a small baseline correction term. M A represents the calculated buoyant molecular mass M(1 Ϫ for the integrase dimer, using a calculated partial specific volume, , based on the amino acid composition (27). This model has the equilibrium constant as a global fitting parameter and cell reference concentrations and baseline corrections as local fitting parameters. The equilibrium constant for this dimer-tetramer association, k, is on an absorbance concentration scale. The values of lnk obtained experimentally were converted to lnK values, K now being the association constant on a molar scale, using the extinction coefficients calculated for the integrase dimer (⑀). Assuming that the extinction coefficient of the multimer is the sum of the extinction coefficients of the components, it can be shown that for a given path length l: The model describing a reversible dimer-tetramer self-association yielded excellent fits for both IN 1-288 /F185K/C280S and IN 50 -288 / F185K/C280S. Data analysis in terms of a single ideal solute gave poor data fits with residuals typical for aggregating species and weight average molecular weights corresponding to values between those expected for integrase dimers and tetramers.
Cells and Analysis of HIV-1 Proteins-HeLa cells were grown in Dulbecco's modified Eagle's medium containing 5% fetal calf serum, 100 units of penicillin G sodium, and 0.1 mg/ml streptomycin sulfate as described (30). To prepare virus stocks, 3.6 ϫ 10 5 HeLa cells were transfected with 10 g of plasmid (wild-type-pNL4-3 or F185K-pNL4-3) using the Cell-Phect kit (Pharmacia). After 48 h, the culture supernatants were assayed for the presence of Mg 2ϩ -dependent 32 P-reverse transcriptase activity as described previously (31). Filtered supernatants were equalized for 32 P-reverse transcriptase activity for the subsequent infection of the human T-cell line, CEM-12D7 (32). CEM-12D7 cells were grown in RPMI 1640 containing 10% fetal calf serum, 100 units of penicillin G sodium, and 0.1 mg/ml streptomycin sulfate as described (30). 2.5 ϫ 10 6 cells were infected with 2.5 ϫ 10 5 32 P-reverse transcriptase cpm in 0.5 ml of RPMI 1640 for 1 h at 37°C followed by the addition of 5 ml of RPMI 1640 containing 10% fetal calf serum. Cultures were then monitored for 32 P-reverse transcriptase activity. Cell-and particle-associated HIV-1 proteins were analyzed by radiolabeling transfected HeLa cells and immunoprecipitating with AIDS patient antisera (donated by Dr. Alfred Prince through the NIH AIDS Research and Reference Program) as described (30). Radiolabeled proteins were quantitated using a PhosphorImager and ImageQuant 3.3 (Molecular Dynamics).

RESULTS
The F185K Mutation Increases the Solubility of HIV-1 Integrase-Expression of IN 1-288 /F185K and subsequent purification by Ni 2ϩ -affinity chromatography yielded a protein that was considerably more soluble than wild-type integrase in buffer containing 1 M NaCl (data not shown). However, this protein displayed time-dependent aggregation in the absence of a reducing agent. This aggregation was readily reversed in the presence of 5 mM DTT. Preliminary studies with a carboxylterminal deletion mutant and the soluble catalytic core domain of integrase revealed that these proteins did not aggregate in the absence of DTT. We therefore suspected that the aggregation was caused by disulfide linkages involving cysteine 280, the only cysteine present in the carboxyl-terminal region of the protein. Serine was substituted for cysteine at this position. The resulting protein IN 1-288 /F185K/C280S retained the markedly improved solubility properties compared with wild-type integrase but, unlike IN 1-288 /F185K, remained soluble even in the absence of reducing agent.
Purification of IN 1-288 /F185K/C280S-HIV-1 IN 1-288 / F185K/C280S was purified from the soluble fraction following lysis of cells in the presence of 1 M NaCl (Fig. 1). The HT at the amino terminus of the protein permitted rapid purification by nickel-affinity chromatography; 280 mg of protein were obtained from 24 liters of cells expressing IN 1-288 /F185K/C280S. The integrase peak was pooled and the HT was removed by cleavage with thrombin; the purity of the pooled integrase was analyzed by SDS-PAGE and staining with Coomassie Blue (Fig. 2). Minor contaminants were successfully removed by subsequent chromatography on Mono S.

IN 1-288 /F185K/C280S Forms Both Dimers and Tetramers in Solution-
We used both gel filtration and sedimentation equilibrium studies to determine the multimeric state of IN 1-288 / F185K/C280S in solution. Gel filtration using a Superdex 200 column, with the protein loaded at high concentration (Ͼ4 mg/ml), resulted in a single eluted peak that migrated at the predicted position of a tetramer relative to globular protein standards (Fig. 3). When IN 1-288 /F185K/C280S was loaded at a lower concentration (Ͻ0.4 mg/ml), a single peak was again eluted; however, this corresponded to the position expected for a dimer of integrase. Nevertheless, both the dimer and tetramer peaks contained elements of tetramer and dimer, respectively (Fig. 3).
Although qualitatively informative, the gel filtration data represent the average behavior of reassociating molecules as they migrate through the column. To more rigorously determine the multimeric state of the protein, a series of protein dilutions were examined by sedimentation equilibrium at 4.0°C and different rotor speeds. A single ideal solute fit gave molecular masses which indicated the presence of at least a dimer, even at integrase concentrations as low as 0.1 mg/ml. At higher concentrations, evidence for the self-association of the integrase dimers was observed. The sedimentation equilibrium data for each loading concentration were fitted using a dimertetramer association model (see "Materials and Methods," Equation 1). Protein concentrations of 0.3, 1.0, and 1.25 mg/ml yielded an excellent fit to this model (Fig. 4A) with residuals normally distributed about zero. For all three concentrations, similar values of lnK were obtained, averaged at 10.7 Ϯ 0.2, leading to an apparent association constant, K a , of 44,000 M Ϫ1 . This corresponds to an effective dissociation constant, K d , of 2.2 ϫ 10 Ϫ5 M. The experimental data are thus consistent with a reversible dimer-tetramer self-association (Fig. 4B).

The Carboxyl-terminal Domain of Integrase Is Required for
Tetramerization-To further investigate the requirements for tetramerization, two deletion mutant proteins were constructed. The first contained the core and carboxyl-terminal domains, IN 50 -288 /F185K/C280S, and was purified as described for the full-length counterpart. The second comprised the core and amino-terminal domains, IN 1-212 /F185K, and was purified as described for IN 50 -212 /F185K (19). Gel filtration experi-ments revealed that IN 1-212 /F185K is exclusively a dimer in solution (data not shown). In contrast, sedimentation equilibrium data showed that IN 50 -288 /F185K/C280S, like IN 1-288 / F185K/C280S, exists in an equilibrium between dimers and tetramers. Data were collected at 0.4 mg/ml and fitted using a dimer-tetramer association model (Equation 1). The data yielded an excellent fit to this model ( Fig. 5) with residuals normally distributed about zero. A lnK value of 10.8 Ϯ 0.04 was obtained, leading to an apparent association constant, K a , of 51,400 M Ϫ1 . This corresponds to an effective dissociation constant, K d , of 2.0 ϫ 10 Ϫ5 M. (Fig. 5), a value very similar to that obtained for IN 1-288 /F185K/C280S.
Catalytic Activities of IN 1-288 /F185K/C280S-Integrase activity was analyzed using double-stranded oligonucleotide substrates that mimic the sequences found at the U5 end of HIV-1 DNA (6,7). It has been shown that 3Ј processing by wild-type integrase in the presence of Mn 2ϩ generates three specific products of nucleophilic attack on the phosphodiester bond at the site of cleavage (28,33). When water acts as the nucleophile, a simple dinucleotide (D) is generated, attack by glycerol produces a glycerol adduct (G), and attack by the 3Ј-OH end of the DNA strand yields a cyclic dinucleotide product (C). However, in the presence of Mg 2ϩ , the simple dinucleotide (D) is the sole product (33). IN 1-288 /F185K/C280S was compared with wild-type integrase for the preference of nucleophile utilization in the cleavage reaction with either Mn 2ϩ or Mg 2ϩ . No difference could be observed between the two enzymes with regard to the selection of nucleophile or the extent of cleavage under either assay condition (Fig. 6). The strand transfer activities of IN 1-288 /F185K/C280S and wild-type integrase were also compared, and no significant differences were observed (Fig. 7). Disintegration activities were also found to be identical. We conclude that the combined mutations of F185K and C280S do not significantly alter in vitro activities of integrase.
Mutation F185K Is Deleterious to Virion Assembly in Vivo-Following our in vitro characterization of the soluble mutant IN/F185K, but before introduction of the additional mutation C280S, we investigated the in vivo effects of the F185K mutation on the replication of HIV-1 in cell culture. The mutation was introduced into the infectious molecular clone HIV-1 NL4 -3 , and virus particles were generated by transfection of HeLa cells. Supernatants from cells transfected with wild-type or F185K clones were normalized for reverse transcriptase activity and used to infect the human T-cell line, CEM-12D7. To assess infection kinetics, CEM-12D7 cell supernatants were monitored for reverse transcriptase activity for 60 days. Wildtype virus production, as measured by reverse transcriptase activity, peaked at day 8. No replication of the mutant (F185K)-HIV-1 NL4 -3 was detected (Fig. 8A). After failure of the mutant virus to infect CEM-12D7 cells, the virus generated by transfection of the HeLa cells was examined. Transfected cells were radiolabeled, and cell-and virus-associated viral proteins were analyzed by immunoprecipitation and SDS-PAGE. Lysates of cells transfected with the wild-type clone contained prominent precursor and mature forms of HIV-1 envelope and GAG proteins (Fig. 8B), whereas the levels of these proteins were reduced in cells transfected with the mutant clone. Analysis of lysates of virions from cells transfected with the wild-type clone revealed two mature products of the pol gene, the p66 subunit of reverse transcriptase and integrase (p32). However, when normalized for gp120 and p24 content, both p66 and p32 were dramatically reduced in lysates of virions from cells transfected with the F185K mutant clone. Electron microscopy of crosssections of transfected cells revealed that the mutant virus appeared to consist of immature rings devoid of nucleoid material or aberrant particles that contained nucleoid material located near the membrane of the particle (data not shown). The F185K change is apparently another example of an integrase mutation which can affect virion protein composition and morphology (30,34,35). DISCUSSION Structural and biophysical studies of HIV integrase have been impeded by the poor solubility of the protein. We find that a single amino acid substitution that dramatically improves the solubility of the core domain (19) has a similar effect on the solubility of the full-length protein. This protein required the presence of reducing agent to prevent slow time-dependent aggregation, a need which was alleviated by introducing the additional mutation C280S. The double mutant provides a protein that should facilitate structural studies of full-length integrase.
Solubility Studies-Wild-type HIV-1 integrase is poorly soluble except in the presence of reagents that are likely to cause at least partial denaturation. Maximum solubility is approximately 1 mg/ml, even under the most favorable conditions of high ionic strength. The F185K mutation considerably increases the solubility of the full-length integrase, provided high ionic strength is maintained. However, like the wild-type integrase, the solubility is greatly reduced at low ionic strength. The requirement for the presence of a reducing agent to maintain the protein in a nonaggregated state is abrogated by the additional mutation C280S. We speculate that intermolecular disulfide cross-links involving Cys-280 may also contribute to the propensity of wild-type integrase to aggregate, but the presence of reducing agents does not significantly reduce aggregation because other intermolecular interactions dominate the aggregation process.
Multimerization of Integrase-Our results demonstrate that IN 1-288 /F185K/C280S exists in a reversible dimer-tetramer self-association in solution (Figs. 3 and 4). Gel filtration results demonstrate the absence of higher molecular weight aggregates, and that at protein concentrations approaching 10 mg/ml the predominant form of the protein is tetrameric.  (36,37). However, a deletion mutant containing the core and carboxyl-terminal domains, IN 50 -288 / F185K/C280S, exists in a dimer-tetramer self-association in solution (Fig. 5). Thus, multimerization interfaces appear to be located in both the core and carboxyl-terminal domains of HIV-1 integrase.
Complementation studies with HIV-1 and HIV-2 integrases (12,38) have demonstrated that multimerization is required for the 3Ј processing and strand transfer activities, but these experiments cannot discriminate whether the active multimer is dimeric, tetrameric, or higher order. Several retroviral integrases have been reported to exist as dimers (6, 39 -41). Rous sarcoma virus integrase has been reported to form a reversible monomer-dimer-tetramer association in solution, and it was also suggested that the protein functioned as a multimer (41). Collectively, these studies point to a propensity of integrases to form both dimers and tetramers, but do not directly address which multimer is the active species for catalysis.
Effects of the F185K/C280S Mutations on Enzymatic Activity and Multimerization-A major concern when using mutagenesis as a tool to improve the behavior of a protein for structural studies is fortuitous alteration of other properties, especially catalytic activity. We therefore examined IN 1-288 /F185K/ C280S to determine if there were any measurable perturbations in catalytic activity. Previous studies with F185K in the context of the catalytic core domain revealed that the mutant protein was in fact slightly more active than the unmutagenized core (19). However, no significant differences could be observed between the unmutagenized and mutant full-length proteins in assays for 3Ј-processing, strand transfer, or disintegration activities. We conclude that the combined F185K and C280S mutations do not adversely affect the catalytic properties of HIV-1 integrase in vitro.
We were unable to detect any difference in the multimerization properties between IN 1-288 /F185K and IN 1-288 /F185K/ C280S (data not shown) and conclude that this amino acid substitution has little, if any, effect on multimerization. It was not possible to compare directly the multimerization properties of IN 1-288 /F185K/C280S with wild-type protein because the latter exhibits extensive aggregation, except at relatively low concentrations in the presence of detergent. The F185K mutation was previously reported to stabilize dimerization of the isolated core domain (19), and this mutation appears to have the similar effect of stabilizing dimerization of the full-length integrase; at a protein concentration of 0.2 mg/ml, wild-type protein exhibits a monomer-dimer equilibrium in the presence of the detergent CHAPS, whereas IN 1-288 /F185K/C280S was exclusively dimeric (data not shown).
We also wished to determine the effect of the F185K mutation on HIV replication in cell culture. Unfortunately, this mutation disrupted proper virion assembly and blocked replication at a step prior to reverse transcription; its potential effect on integration, therefore, could not be determined. This phenotype has been observed with several other mutations in integrase (30,34,35) and probably reflects effects on proteinprotein interactions involving integrase at earlier stages of the replication cycle.
Utility for Structural Studies-The soluble integrase mutant IN 1-288 /F185K/C280S is an attractive candidate for structural studies since it is well behaved and soluble up to at least 20 mg/ml in a buffer containing 1 M NaCl. Soluble derivatives of HIV-1 integrase are now available that include all three domains. Structures of the core (20) and carboxyl-terminal domains (36,37) have already been determined for HIV-1 integrase, and the structure of the core domain of the closely related Rous sarcoma virus integrase has also been solved (42). These structures should eventually prove invaluable in determining the mechanism of action of inhibitors to HIV integrase and ultimately to guiding the design of therapeutically useful compounds. A major remaining objective is determination of the structure of the intact HIV integrase in a complex with its DNA substrates. The soluble mutant described here may help accomplish of this goal.