INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Integrase is the retroviral protein responsible for inserting a double-stranded DNA copy of the viral genome into the host chromosome. To accomplish this essential step in viral replication, integrase catalyzes two chemical reactions. In the first chemical step, a dinucleotide is cleaved from the 3' ends of the newly synthesized viral DNA. The newly exposed hydroxyls on the 3' ends are used in the second chemical step to attack phosphodiester bonds on opposite strands of the target DNA, joining the 3' viral DNA ends to the target DNA. The gapped structure that results is repaired by an essentially uncharacterized process. The virally encoded integrase protein and phylogenetically conserved sequences present at the ends of the double-stranded DNA genome are required for integration.

In vitro assays have been developed to allow detailed analysis of integrase-catalyzed reactions. A synthetic oligonucleotide model of the viral DNA end can serve as a substrate for 3'-end processing and integration (16). In vitro, integrase can also catalyze the reverse of integration, or disintegration (15). The substrate for this reaction resembles an intermediate in the integration reaction: a single viral DNA end joined to target DNA. Integrase can catalyze a cleavage-ligation reaction on this substrate in which the 3' hydroxyl on target DNA 5' to the site of joining is used as a nucleophile to attack the junction between the viral DNA and target DNA 3' to the site of joining. The products are a continuous target DNA strand and a free viral DNA end. Disintegration substrates have proven to be extremely useful for determining the catalytic requirements of integrase and for analyzing mutant integrase proteins that have defects in requirements for the forward reaction (11, 27, 37, 43, 44, 69). End processing, integration, and disintegration all require a divalent metal ion but no exogenous energy source (5, 15).

Mutational analysis has led to the definition of three functional domains of integrases. The N terminus of integrase contains a phylogenetically conserved zinc finger motif and binds zinc in vitro (8, 11, 72). Zinc binding promotes tetramerization of human immunodeficiency virus type 1 (HIV-1) integrase and enhances end-processing and integration activities (72). The enzymatic properties of proteins with mutations in the conserved histidine or cysteine pairs, or with a deletion of this domain, suggest that this domain may be involved in integrase-integrase and/or integrase-substrate interactions (26, 35, 37, 69, 72). The C terminus of integrase possesses sequence- and metal ion-independent DNA binding activity (28, 55, 70). Biochemical methods have demonstrated that this isolated domain exists as a dimer in solution (1, 46), and in the case of HIV-1 integrase, the solution structure has been solved as a dimer (24, 46). The core domains of all retroviral integrases contain a phylogenetically conserved DD(35)E motif. Mutations in any of the three acidic residues composing this motif have parallel detrimental effects on 3'-end-processing, integration, and disintegration activities, arguing for the existence of a single active site that is directly involved in catalysis of all three reactions (20, 27, 39, 44, 67). The proximity of these residues in the crystal structures of the avian sarcoma virus (ASV) and HIV-1 integrase core domains supports the hypothesis that they form a catalytic center (6, 7, 22, 23). This isolated domain was crystallized as a dimer with an extensive hydrophobic interface in the case of both HIV-1 and ASV integrase. Biochemical methods have demonstrated that this domain exists as a dimer in solution (34, 36). Upon soaking the ASV integrase core domain crystal in Mg²⁺ (or Mn²⁺), an Mg²⁺ (or Mn²⁺) ion was bound by the two aspartic acid residues, suggesting that these residues coordinate a divalent metal ion required for catalysis (7). The core domain is the most conserved of the three functional domains of integrase, having some amino acid homology, but much more striking structural homology, with the RNase H domain of HIV-1 reverse transcriptase, Escherichia coli RNase H, Mu transposase, and the Holliday junction resolvase RuvC (for a review, see reference 57), all of which are proteins that catalyze substitution reactions on phosphodiester bonds.

We were interested in identifying amino acids that interact with DNA substrates and control the specificity of the chemical reactions catalyzed by integrase. Specificity for certain substrate features certainly exists. (i) In vivo, mutations in viral DNA end sequences, especially the phylogenetically invariant CA/TG dinucleotide at the viral DNA end, result in a replication-incompetent virus (58). Furthermore, mutations in these base pairs in model viral DNA substrates result in up to a 100-fold reduction in end processing (9, 16, 40, 43, 61, 63, 64, 66). (ii) Viral DNA end substrates that lack the terminal two nucleotides on the 5' end are not joined processively after 3' end processing in vitro, indicating that this dinucleotide is critical for stable enzyme-viral DNA end interactions (10, 25). (iii) Integrase prefers to integrate into bent regions on target DNA (3, 49, 51-53). The core domain influences the target site preference of the integration reaction (38, 50, 65). The determinants of integrase's substrate specificity have proven elusive. The finding that the isolated HIV-1 integrase core domain could recognize key features of both viral and target DNAs led us to focus our attention on this domain (31). The three-dimensional information from the crystal structure of the HIV-1 integrase core domain (22, 23) in conjunction with homology alignments of related integrases allowed us to identify conserved residues near the DD(35)E motif that could potentially be involved in substrate interactions. Using site-directed mutagenesis, we evaluated 13 residues for participation in specificity: 10 near the DD(35)E motif, as well as the 3 acidic residues themselves. More than one substitution was made at most positions. The specificities of these mutant enzymes for viral and target DNAs were evaluated by using in vitro assays.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Construction of mutant integrase genes. Mutant integrase genes were constructed in a synthetic integrase gene (GenBank accession no. AF029884) in a pUC19 vector. The nearest two unique restriction sites flanking the site to be mutagenized were used to make a replacement with duplex oligonucleotides containing the appropriate ends and the mutation. After transformation into DH5 electrocompetent cells, plasmid DNA was prepared from single colonies. The region containing the replacement and its junctions was sequenced, and a larger portion of the gene containing the mutation was subcloned into a vector containing the entire synthetic gene under the control of the T7 promoter. This parent gene contained an F185K mutation and an N-terminal hexahistidine tag to facilitate purification of the mutant proteins. All of the mutant proteins had an N-terminal hexahistidine tag. The only protein that lacked the F185K mutation was IN(D116N). After sequencing to ensure that the appropriate mutation had been transferred to the expression construct, the construct was transformed into the BL21 expression strain.

Integrase expression. Single colonies were tested for expression of integrase by growing them to late log phase in Luria broth (LB) and then inducing integrase expression by the addition of 0.25 mM isopropyl--D-thiogalactopyranoside (IPTG). A sample of the culture before induction and after 3 h of induction at 37°C was boiled in sodium dodecyl sulfate protein gel loading buffer for 3 min and electrophoresed on a sodium dodecyl sulfate-10% polyacrylamide gel. Western analysis was performed on nitrocellulose electroblots of these gels by using polyclonal serum from an HIV-1-infected patient as the primary antibody and either horseradish peroxidase- or alkaline phosphatase-conjugated goat anti-human antibody (Bio-Rad) as the secondary antibody. Cultures expressing integrase were frozen at 80°C in 50% glycerol.

Large-scale growth for purification. Integrase-expressing strains were streaked from the glycerol stock onto LB plates containing 50 µg of ampicillin per ml. An individual colony was used to inoculate 500 ml of LB containing 50 µg of ampicillin per ml. After overnight growth at 37°C, 250 ml of this culture was added to 2 liters of LB containing 50 µg of ampicillin per ml. The culture was grown at 30°C for 3 h, and then IPTG was added to a final concentration of 0.4 mM. After 3 h of induction, the culture was centrifuged for 20 min at 5,000 × g at 4°C to pellet the bacteria. Cell pellets were frozen at 20°C.

Integrase purification. Bacterial pellets were thawed on ice. Pellets were resuspended in 50 ml of lysis buffer (20 mM Tris [pH 8.0], 0.1 mM EDTA, 2 mM -mercaptoethanol, 0.5 M NaCl, 5 mM imidazole, and 0.2 mg of lysozyme per ml). The resuspended pellets were sonicated on ice for 25 s five times at intervals of 2 min and then incubated on ice for 30 min. This lysate was centrifuged in a Sorvall centrifuge for 45 min at 39,000 × g in an SS-34 Sorvall rotor at 4°C. The pellet was Dounce homogenized 30 times in TNM (20 mM Tris [pH 8.0], 1 M NaCl, 2 mM -mercaptoethanol) containing 5 mM imidazole. The resuspended pellet was then stirred at 4°C for 1 h. The mixture was centrifuged in a Beckman ultracentrifuge for 1 h at 85,500 × g in a Beckman SW-28 swinging-bucket rotor at 4°C. The supernatant was gravity loaded onto a 1-ml Ni²⁺-nitrilotriacetic acid agarose (Ni-NTA resin; Qiagen) column preequilibrated with TNM containing 5 mM imidazole. The column was then washed with 20 column volumes of TNM containing 5 mM imidazole, followed by 20 column volumes of TNM containing 40 mM imidazole. Integrase was eluted with a step gradient. The Bradford protein assay (Bio-Rad) was used to determine that integrase eluted in the fractions containing 100 and 200 mM imidazole. Peak fractions were pooled, diluted 1:10 with TNM buffer, and gravity loaded onto a second 0.5-ml Ni-NTA column. Washing and elution were carried out as described for the 1-ml column. Peak fractions were pooled, 3-[(3-cholamidopropyl)-dimethyl-ammonio]-1-propanesulfonate (CHAPS) was added to a final concentration of 10 mM, and the fractions were dialyzed against buffer containing 20 mM HEPES (pH 7.5), 10 mM dithiothreitol (DTT), 300 mM NaCl, 10 mM CHAPS, and 10% glycerol. The protein concentration was determined by the Bradford method with purified integrase as a standard. The concentration of the integrase used as a standard had been determined by amino acid analysis. Aliquots were frozen in liquid nitrogen and stored at 80°C. The concentrations of integrase cited throughout this report are concentrations of integrase promoters.

Construction of substrates. All oligonucleotides were purchased from Operon Technologies, Inc. (Emeryville, Calif.). Oligonucleotides were purified by electrophoresis on a 15 or 20% denaturing polyacrylamide gel prior to use. T4 polynucleotide kinase (New England Biolabs) and [-³²P]ATP (Amersham; 3,000 Ci/mmol) were used for all 5' end labelling. Ymer disintegration substrates were made from four oligonucleotides annealed together to form the structure of one viral DNA end joined to target DNA; a Ymer substrate contained 19 bp of viral DNA joined to 30 bp of target DNA. Dumbbell disintegration substrates consisted of a single oligonucleotide folded upon itself to form the structure of one viral DNA end joined to target DNA in which the viral DNA portion was 5 bp in length and the target DNA portion was 10 bp in length. Disintegration substrates were 5' end labelled and annealed as described by Chow et al. (15). For mutant Ymer disintegration substrates, oligonucleotides with the indicated sequence were used in the construction. Viral DNA end substrates, both blunt ended and prerecessed, were 5' end labelled and annealed as described by Ellison and Brown (25).

Activity assays. Activity assays were performed under a variety of conditions. For Table 1 and Fig. 2, 3' end processing was measured by using a blunt-ended U5 viral DNA end substrate. The reaction mixtures included 300 nM integrase and 50 nM viral DNA end substrate incubated in a solution containing 7.5 mM MnCl₂, 12 mM DTT, 50 µg of bovine serum albumin (BSA) per ml, 2 mM CHAPS, 2% glycerol, 22 mM HEPES (pH 7.5), and either 25 or 90 mM NaCl, as indicated. All end-processed and integrated products were included in the quantitation of 3'-end-processing activity. The amount of product for each reaction analyzed was divided by the amount of product for a parallel reaction containing IN(F185K) and is therefore expressed as a percentage in Table 1. The 3'-end-processing reaction mixtures were incubated for 30 min, the integration reaction mixtures were incubated for 10 min, and the disintegration reaction mixtures were incubated for 4 min, all at 37°C. All incubation times were determined to be in the linear range for IN(F185K) in kinetic experiments performed under similar conditions. A prerecessed viral DNA substrate was used in the integration reactions, and a Ymer disintegration substrate was used in the disintegration reactions. The disintegration reaction conditions were identical to those described above except that 150 mM integrase and 50 mM Ymer substrate were used and the final NaCl concentration was 50 mM. Reactions with a Ymer substrate without the 5' dinucleotide were done in parallel (data not shown). Reactions were stopped by the addition of an equal volume of formamide loading buffer (95% formamide, 50 mM EDTA, 0.1% bromophenol blue, and 0.1% xylene cyanol). Samples were heated to 90°C for 2 min before electrophoresis on a 15% denaturing polyacrylamide gel. Substrate and product were quantitated with a Molecular Dynamics Phosphorimager.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

The premise of these experiments was that phylogenetically conserved residues near the active site of HIV-1 integrase were likely candidates to participate in recognizing critical substrate features. We chose to mutate the amino acids shown mapped onto the structure of the HIV-1 integrase core domain in Fig. 1. The three acidic residues of the phylogenetically conserved DD(35)E motif are depicted in red. Nine of the 10 residues mutagenized are shown; the 10th, Q148, is not resolved in this structure but is predicted to be near the DD(35)E residues. The selection of residues potentially involved in specificity was based on the existing crystal structure for HIV-1 integrase (22, 23); any conformational change that occurs upon DNA or divalent metal ion binding might cause us to miss residues whose position relative to the DD(35)E motif is affected by such a change. Moreover, potential interactions with other integrase promoters in a multimeric complex could not be predicted in this way.

View larger version (103K): [in this window] [in a new window]   FIG. 1.   Amino acids targeted for mutagenesis. The three-dimensional structure of the HIV-1 integrase catalytic core domain is shown (22, 23). The amino acids chosen for mutagenesis are in color: Q62 is grey, T66 is yellow, H67 is dark blue, E92 is green, N117 is light blue, N120 is magenta, N155 is orange, K156 is purple, and K159 is light green. The residues that define the DD(35)E active-site motif (D64, D116, and E152) are red. A Q148 mutant was also analyzed in this work; this residue has not been resolved in the crystal structure. This picture was generated by using the program GRASP.

All mutations were made in the context of full-length integrase with the F185K mutation and an N-terminal hexahistidine tag. Mutant enzymes were overexpressed in E. coli and purified by Ni²⁺ affinity chromatography. Mutant proteins were assayed for 3'-end-processing, integration, and disintegration activities under a variety of conditions. Parameters varied included protein concentration, substrate concentration, and salt concentration.

3'-end-processing, integration, and disintegration activities of mutants. Table 1 summarizes the activities of the mutant integrase proteins relative to those of the F185K parent protein, which is considered to be wild type in vitro (35). The only tested position at which substitution did not appear to have an effect on catalysis of 3' end processing, integration, or disintegration under any conditions was H67. Serine was the only residue substituted for histidine at this position. This residue is therefore probably not involved in enzyme-substrate interactions (or protein-protein interactions) that are critical for catalysis. Substitutions at position E92 resulted in mutant proteins that displayed parallel reductions in end-processing, integration, and disintegration activities. The identity of the substitution (A or N) did not have a large effect on the phenotype of the mutant.

View larger version (23K): [in this window] [in a new window]   FIG. 2.   Salt sensitivities of mutant integrase proteins. The end-processing and integration activities of mutant integrase proteins were measured in buffers containing 25 or 90 mM NaCl. The y axis represents the total amount of substrate that had been converted to product. IN(F185K) and IN(H67S) (A and B, respectively) had similar activities at both tested NaCl concentrations. However, the activities of a subset of the mutant integrase proteins were lower at the higher NaCl concentration: IN(Q62A) (C), IN(N117S) (D), IN(K156E) (E), and IN(K159N) (F). Other mutants with similar salt sensitivities are noted in the text and in Table 1.

View larger version (44K): [in this window] [in a new window]   FIG. 3.   (A) IN(Q148L) does not catalyze end processing and integration processively. Mixtures for the reactions analyzed in lanes 1 to 9 each contained a 50 nM concentration of the blunt-ended viral DNA substrate. Samples analyzed in lanes 1 and 5 were taken at 5 min from reaction mixtures containing 300 nM IN(Q148L). Samples analyzed in lanes 3 and 7 were taken at 5 min from reaction mixtures containing 300 nM IN(F185K). After 5 min, either 3 µl of TEN or 3 µl of TEN containing 15 pmol of A227/A228 per µl was added to 7 µl of the reaction mixture, and the incubation was continued for 25 min at 37°C. Samples analyzed in lanes 2 and 6 contain reactions with IN(Q148L) in which 3 µl of TEN or TEN containing 15 pmol of competitor DNA per µl was added, respectively. Samples analyzed in lanes 4 and 8 were taken from reactions with IN(F185K) to which TEN or TEN containing 15 pmol of competitor DNA per µl was added, respectively. Integrase was omitted from the reactions analyzed in lanes 9, 12, and 15. Mixtures for the reactions analyzed in lanes 10 to 12 contained a 50 nM concentration of the prerecessed viral DNA end substrate and were incubated at 37°C for 10 min. Mixtures for the reactions analyzed in lanes 10 and 11 contained 300 nM IN(Q148L) and 300 nM IN(F185K), respectively. Mixtures for the reactions analyzed in lanes 13 to 15 contained 50 nM Ymer disintegration substrate and were incubated at 37°C for 5 min. Mixtures for the reactions analyzed in lanes 13 and 14 contained 150 nM IN(Q148L) and 150 nM IN(F185K), respectively. Integration of 3'-end-processed viral DNA ends by IN(Q148L) is poor in the presence of a molar excess of competitor DNA (compare lanes 2 and 6), whereas the integration of 3'-end-processed viral DNA ends by IN(F185K) proceeds normally under these conditions (compare lanes 4 and 8). (B) Turnover of IN(Q148L) is insensitive to the presence of the 5' dinucleotide of the viral DNA end. Parallel reactions were performed in duplicate with 1.5 µM substrate and 150 nM integrase. Results from duplicate experiments were averaged. A dumbbell disintegration substrate either with (db+) or without (db) the 5' overhang was used. IN(Q148L) and IN(F185K) had approximately the same rates of turnover in reactions with the (db) substrate. However, while IN(F185K) showed a lower rate of turnover in reactions with the db+ substrate, turnover of IN(Q148L) was unaffected by inclusion of the 5' overhang.

Disintegration activities and substrate specificities of mutant integrases with alterations in the putative active-site residues D64, D116, and E152. We evaluated the effects of several substitutions in each of the three acidic active-site residues: D64, D116, and E152. Mutations in these residues have previously been reported to have parallel detrimental effects on 3' end processing, integration, and disintegration, reducing catalytic activity by several orders of magnitude (20, 27, 44, 67). This DD(35)E motif is the most phylogenetically conserved feature in retroviral integrases and bacterial transposases; the subterminal CA/TG base pairs in the viral DNA are also universally conserved in retroviruses and are a common feature of bacterial transposons. It therefore seemed possible that the conserved aspartic acid and glutamic acid residues themselves could be responsible in some part for specificity for the CA/TG sequence. In order to measure the substrate specificities of proteins with mutations in the aspartic acid or glutamic acid active-site residues, substitutions that would generate proteins with measurable catalytic activity had to be found. Since the disintegration assay is the most sensitive to very low levels of integrase activity, we used disintegration substrates with mutations in the CA/TG base pairs to assay these mutant proteins for their substrate specificity. Since integrase shows similar specificities for these base pairs in the integration and disintegration reactions (14, 69) and since the orientations of viral DNA and target DNA with respect to the active site are similar for integration and disintegration (32), the interactions mediating specificity for the CA/TG sequence should be similar for both reactions.

View larger version (24K): [in this window] [in a new window]   FIG. 4.   Sensitivities of mutant integrase proteins to substitutions for the conserved subterminal A/T base pair of the viral DNA end. The rate of disintegration was calculated by plotting the amount of substrate converted to product at each time point and determining the slope of the line by the least-squares method. The y axis represents the calculated rates. In a few cases where a standard deviation was calculated by determining rates for a substrate in triplicate, this is indicated by an error bar. The base pair substituted at the position of the phylogenetically conserved subterminal A/T base pair is indicated below each bar. Rates given for IN(D64C) and IN(D116C) were determined at pH 8.5; rates given for IN(F185K), IN(D116E), and E(152D) were determined at pH 7.5. Rates were also determined for IN(F185K) at pH 8.5 and were approximately twofold lower for each substrate (data not shown).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The catalytic core domain of HIV-1 integrase is responsible for recognition of the subterminal phylogenetically conserved CA/TG dinucleotide pair near the viral DNA end, for the 5'-terminal dinucleotide, and for target DNA adjacent to the site of joining (31). We used model substrates to analyze the substrate specificities of integrase mutants that contained changes in residues near the critical acidic residues of the phylogenetically conserved DD(35)E motif or in these acidic residues themselves. We identified specific residues involved in recognition of the subterminal A/T base pair and the 5' dinucleotide at the viral DNA ends.

The DD(35)E active-site motif is shared by proteins in a superfamily of polynucleotidyl transferases. The two conserved aspartic acids have been shown to coordinate a divalent metal ion in the crystal structure of the core domain of ASV integrase. Substituting cysteine for one of the aspartic acids in the active site of the related TnsA (59) or MuA (2) transposase changes the metal ion preference from Mg²⁺ to Mn²⁺, strongly suggesting that the aspartic acid residue is part of an essential metal ion binding site in both enzymes. Substituting cysteine for D64 or D116 of HIV-1 integrase (the residues analogous to the aspartic acids coordinating a metal ion in the active site of ASV integrase) resulted in mutant proteins with an altered pH optimum and in 50- or 5-fold-less catalytic activity than in the wild-type protein, respectively. Thus, the ability of each of these side chains alone to provide only a single coordination site for a metal ion is sufficient to coordinate a divalent metal ion in the active site of integrase. In the case of ASV integrase, only one of the two carboxyl oxygens of each of the active-site aspartic acids coordinates the divalent metal ion (7). Although these substitutions of cysteine for aspartate might have been expected to alter the metal ion preference of integrase, sulfur interacts preferentially with Mn²⁺ over Mg²⁺, and wild-type HIV-1 integrase already has an essentially absolute preference for Mn²⁺ as a cofactor for disintegration in vitro.

In contrast to the aspartic acid residues, the glutamic acid residue of the DD(35)E motif of HIV-1 integrase was relatively intolerant of substitutions. Substitution of cysteine at position E152 resulted in a protein with 10⁵-fold-lower disintegration activity than IN(F185K) under substrate-excess conditions, indicating that this unnatural side chain was not compatible with catalysis. The most conservative substitution, that of aspartic acid, for E152 was the only substitution that resulted in a mutant protein with enough activity to allow catalytic rates to be measured for disintegration substrates with mutations in the subterminal viral DNA A/T base pair. In contrast to IN(F185K) and other active-site mutants, IN(E152D) displayed a marked indifference to mutations in the phylogenetically conserved viral A/T base pair. IN(E152D) did not display a gross DNA binding defect (data not shown), arguing that this altered substrate specificity was achieved downstream of DNA binding, perhaps in a subsequent step of docking into the active site. This altered specificity, along with the intolerance to substitution, suggests that in addition to, or instead of, playing a direct role in the chemistry of the reaction, E152 may play a role in the recognition of the A/T base pair that is essential for efficient catalysis. E152 is located in a loop that demonstrates flexibility in the crystal structures of many related proteins, including ASV and HIV-1 integrases and MuA. The finding that a mutation in E152 affects recognition of the A/T base pair suggests that E152 may normally interact with this base pair; the salient effect of this interaction may be to stabilize this peptide in an active conformation. The simplest model for an E152-A/T interaction that stabilizes the geometry of the active site would be a base-specific interaction. However, more indirect interactions, for example, base-specific interactions between other residues and the A/T base pair, which act to specifically position E152 in an active orientation, would also account for these results.

Substitution of cysteine at D64 resulted in a hyperspecific phenotype; that is, the activity of this mutant protein on substrates mutated at the conserved A/T or C/G base pair compared to its activity on a wild-type substrate was much lower than would be predicted based on the specificity profile of IN(F185K). One possible explanation for this phenotype is that the mutation, either directly or indirectly, caused a loss of nonspecific interactions with terminal nucleotides of the viral DNA, making the mutant protein more reliant on remaining base-specific interactions to position the phylogenetically conserved viral DNA CA/TG base pairs. If these specific interactions were then removed by mutating the substrate, the activity of IN(D64C) would be disproportionately compromised compared to the activity of IN(F185K).

When Q148 was replaced with a leucine, the result was a mutant protein with reduced processivity for end processing and integration and an apparent loss of interactions with the unpaired 5'-terminal dinucleotide of the viral DNA strand. IN(Q148L) did not display altered specificity for the phylogenetically conserved viral DNA CA/TG base pairs (data not shown), implying that Q148 is not involved in specific recognition of this viral sequence. In a previous mutational analysis of HIV-1 integrase, IN(Q148L) was found to generate more cyclic dinucleotide product in the 3'-end-processing reaction than wild-type integrase when Mn²⁺ was used as a cofactor (68). This residue is located in the same flexible loop as E152, which has proven to be difficult to resolve in the HIV-1 integrase core crystal structure and in the crystal structures of related enzymes. This loop may be flexible for an as-yet-unappreciated mechanistic reason that relates to interactions with DNA and possibly to distortion of the substrate. Substrate distortion has been shown to be important for many polynucleotidyl transferase reactions (56, 60-62). Perhaps Q148 promotes unpairing at the viral terminus by interacting with the 5'-terminal two nucleotides of the viral DNA on the unprocessed strand. IN(Q148L) may therefore produce a preponderance of cyclic dinucleotide product because an Mn²⁺-dependent mechanism that promotes unpairing of the terminal base pairs in a manner that favors cyclization is the only pathway available to the mutant for fraying. Whereas numerous substitutions for E152 have been shown to abolish HIV-1 replication (41, 45, 71), the role of Q148 in HIV-1 growth has yet to be determined.

The salt sensitivity observed in a subset of the mutant proteins may be explained if the wild-type residues at these positions are normally involved in nonspecific interactions with the DNA substrates. Depending on the substitution, a particular mutant amino acid side chain might maintain or disrupt a stabilizing interaction. Disruption of an interaction with the DNA substrate (or creation of a new, unfavorable interaction) would make the mutant more dependent on remaining substrate interactions. If favorable electrostatic interactions were then disrupted by raising the salt concentration, the mutant protein might display disproportionately reduced activity compared to wild-type protein. Although we cannot exclude the possibility that disruptions of protein-protein interactions might explain the salt sensitivity of some of the mutants, the location of the altered amino acids near the active site places them in position to interact with DNA. Furthermore, these residues are located on the outside faces of the core dimer (as opposed to the adjoining faces with the large hydrophobic interface), arguing against their involvement in dimerization of the core domain.

Analysis of critical specific enzyme-substrate interactions may provide information that facilitates the development of enzyme inhibitors that may be useful in the treatment of HIV infection. To date, screens for potential antiviral drugs directed at integrase have sought inhibitors of its catalytic activity (12, 13, 17, 18, 29, 30, 33, 42, 47, 48, 54). An alternative approach to targeting antiviral agents at integrase would be deregulation of substrate specificity. Treatment of retrovirally infected cells with a compound that caused integrase to lose specificity could result in inappropriate suicidal cleavage of viral replication intermediates, preventing successful integration and viral replication (4). The identification of specific residues involved in determining the substrate specificity of HIV-1 integrase helps provide a rational foundation for such a strategy.