Stereospecificity of Reactions Catalyzed by HIV-1 Integrase*

The retroviral integrase catalyzes two successive chemical reactions essential for integration of the retroviral genome into a host chromosome: 3′ end processing, in which a dinucleotide is cleaved from each 3′ end of the viral DNA; and the integration reaction itself, in which the resulting recessed 3′ ends of the viral DNA are joined to the host DNA. We have examined the stereospecificity of human immunodeficiency virus type 1 integrase for phosphorothioate substrates in these reactions and in a third reaction, disintegration, which is macroscopically the reverse of integration. Integrase preferentially catalyzed end processing and integration of a substrate with the (R p)-phosphorothioate stereoisomer at the reaction center and disintegration of a substrate with an (S p)-phosphorothiate at the reaction center. These results suggest a model for the architecture of the active site of integrase, and its interactions with key features of the viral and target DNA.

The retroviral integrase catalyzes two successive chemical reactions essential for integration of the retroviral genome into a host chromosome: 3 end processing, in which a dinucleotide is cleaved from each 3 end of the viral DNA; and the integration reaction itself, in which the resulting recessed 3 ends of the viral DNA are joined to the host DNA. We have examined the stereospecificity of human immunodeficiency virus type 1 integrase for phosphorothioate substrates in these reactions and in a third reaction, disintegration, which is macroscopically the reverse of integration. Integrase preferentially catalyzed end processing and integration of a substrate with the (R p )-phosphorothioate stereoisomer at the reaction center and disintegration of a substrate with an (S p )-phosphorothiate at the reaction center. These results suggest a model for the architecture of the active site of integrase, and its interactions with key features of the viral and target DNA.
Integration of the retroviral genome into a host cell chromosome is required for viral replication (1). To accomplish integration, integrase, a viral enzyme, catalyzes two chemical reactions. In the first reaction, end processing, integrase cleaves the terminal dinucleotide from each 3Ј end of the doublestranded viral DNA (Fig. 1A). In the second chemical reaction, integration or DNA joining, integrase joins the recessed 3Ј ends of the viral DNA to target DNA (Fig. 1B). Integrase and short sequences at each end of the viral DNA are required for these two reactions. After the viral DNA 3Ј ends are joined to host DNA, the 5Ј ends of the viral DNA are trimmed and joined to host DNA by a mechanism that remains to be defined (3).
In vitro assays, using duplex oligonucleotides that mimic viral DNA ends and highly purified, recombinant integrase, have been developed to allow detailed analysis of the end-processing and integration reactions (4). Both the end-processing and integration reactions carried out by HIV-1 1 integrase produce (S p )-phosphorothioate products from model substrates containing an (R p )phosphorothioate at the reaction center, implying that both reactions occur by a one-step transesterification mechanism (5).
The available evidence supports a model in which a single active site is responsible for both reactions (6)(7)(8)(9)(10)(11)(12)(13). The orientation and alteration of the DNA substrates for these two different reactions thus presents an interesting puzzle, as elaborated below.
End processing and integration are both polynucleotidyl transfer reactions, but the identity and organization of the reactants relative to the attacking hydroxyl nucleophile and the leaving group are different. In end processing (Fig. 1A), a specific phosphodiester bond in the viral DNA is attacked by a hydroxyl derived from water (HOH) and the leaving group is the 3Ј hydroxyl of the conserved, subterminal adenosine; in integration ( Fig. 1B), this newly formed 3Ј hydroxyl is the attacking nucleophile. Moreover, in end processing, the phosphate group undergoing substitution is initially part of the viral DNA end and is released as part of a dinucleotide product after cleavage, while in integration, the phosphate that is attacked is initially part of a separate target DNA substrate that then becomes covalently joined to the viral DNA end.
The differences between end processing and integration make it difficult to envision how the DNA substrates in the two reactions could retain a normal double-helical structure and still be accommodated by the same active site. Experimental evidence suggests that the terminal 3 nucleotides of model viral DNA substrates are indeed unpaired prior to end processing (14). Highly bent sites in DNA are preferred targets for integration, a finding consistent with a need for distortion of the DNA duplex to accommodate the reacting phosphate group in the active site (15)(16)(17)(18)(19).
Integrase can catalyze a third reaction in vitro, termed "disintegration" (Fig. 1C; Refs. 10 and 20 -26). The substrate for this reaction mimics one end of viral DNA joined to target DNA. Disintegration is the macroscopic reverse of the integration reaction: the adjacent 3Ј hydroxyl of the target DNA attacks the phosphodiester bond joining viral and target DNA, releasing the model viral DNA end and concomitantly restoring the continuity of the target DNA. In order to interpret results of experiments using disintegration substrates, it is important to determine whether the arrangement of viral and target DNA near the active site is the same for the disintegration and integration reactions.
We used phosphorothioate-containing substrates to investigate the orientation of the reacting phosphodiester and the attacking nucleophile in end processing, integration, and disintegration. We found that substrates with an (R p )-phosphorothioate at the reaction center were preferred by HIV-1 integrase as substrates for end processing and integration. In contrast, HIV-1 integrase preferentially catalyzed disintegration of substrates with an (S p )-phosphorothioate at the reaction center. The stereospecificity of integrase in the three reactions provides information regarding the likely arrangement of the key substrate features in the active site and the relationship between viral and target DNA bound to the enzyme.

Enzymes and Reagents
Integrase was purified as described previously (26). Oligonucleotides were purchased from Operon Technologies, Inc. (Emeryville, CA) and purified by electrophoresis on a 15% or 20% denaturing polyacrylamide gel prior to use. T4 polynucleotide kinase, T4 DNA ligase, and exonu-Substrates Dumbbell Disintegration Substrates-A "dumbbell" disintegration substrate is composed of a single oligonucleotide that folds into a structure resembling the product of an oligonucleotide integration reaction ( Fig. 2A).The sequence of the dumbbell disintegration substrate (DB1) and preparation of this substrate were as described previously (22).
To obtain a nearly pure population of the less preferred stereoisomer for disintegration, duplicate 500-l reactions, each containing 300 nM integrase and 1 M unlabeled dumbbell substrate with a racemic phosphorothioate at the reaction center (DB2), were incubated overnight at 37°C, in a standard reaction mixture (see below) to convert most of the preferred stereoisomer substrate to product. The reaction products were extracted with an equal volume of phenol:chloroform, then ethanol-precipitated. After resuspension in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA (TE), an equal volume of formamide loading buffer (95% formamide, 50 mM EDTA) was added and the sample was heated to 90°C for 3 min before loading onto a 20% denaturing polyacrylamide gel. After electrophoresis, the gel was stained with ethidium bromide, and the band corresponding to residual substrate was excised, eluted, and precipitated with ethanol. This DNA was 5Ј end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. Y-mer Disintegration Substrates-The disintegration substrate used to determine the stereochemistry of disintegration was radiolabeled at the 3Ј end of the target DNA 5Ј to the site of joining using exonucleasefree Klenow enzyme and [␣-32 P]dGTP (Fig. 3A), as follows: 25 M amounts of the complementary strands TT1 and TB1 were annealed in 20 l of TEN (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 50 mM NaCl). Next, 25 l of a 3.3 M solution of radiolabeled nucleotide, 5 l of 10ϫ Klenow buffer (New England Biolabs), and 1 l (10 units) of exonuclease-free Klenow enzyme were added, and the reaction was incubated at 37°C for 30 min. The oligonucleotides were then separated from unincorporated radionucleotides by passing the reaction mixture over two sequential Sephadex G-15 (Sigma) spin columns and concentrated in a Microcon-3 ultrafiltration device (Amicon). The disintegration substrate has a branched or "Y-shaped" structure that is formed by annealing four oligonucleotides together. For preparation of this "Y-mer" disintegration substrate, a 5-fold molar excess of each of the three unlabeled strands (TT2, TT3, and TB2) was annealed to the strand labeled at its 3Ј end (TT1). The disintegration substrate was annealed by heating to 90°C for 2 min and slowly cooling to room temperature. The correctly annealed product was gel-purified as described previously (20).
Integration Substrates-The 3Ј end of the viral DNA substrate that was used to determine the stereochemistry of integration was annealed (VT2 and a 2-fold molar excess of VB1) and prepared as described above except that [␣-32 P]dATP was used in the end-labeling reaction (see Fig.  4A). The target DNA was a 21-mer duplex that contained all phosphodiesters in one version (TT4 annealed to TB3) or phosphorothioates 5Ј to all C and G nucleotides in a second version (TT5 annealed to TB4).
End-processing Substrates-For preparation of the substrate used to analyze end processing, an oligonucleotide (VT3) was synthesized with a single (racemic) phosphorothioate (see "Oligonucleotides"). The stereoisomers were separated by reverse phase HPLC on a C-18 column (5-m particle size, 25 ϫ 0.45 cm) using a gradient of 3-13% acetonitrile in 0.1 M acetic acid/imidazole, pH 6.5. The peak fractions were lyophilized and resuspended in TE. Reapplying aliquots from each peak fraction to the C-18 column gave the expected single peak.
To determine the stereochemistry of the phosphorothioate in each of the two major peak fractions, an aliquot of each was digested with either 200 g/ml SV at 44°C for 100 min, or 100 g/ml P1 at room temperature for 60 min, and analyzed by HPLC. These samples were chromatographed using a gradient consisting of 3-16% acetonitrile in 0.1 M acetic acid/triethylamine, pH 6.5. The dinucleotide AG was chromatographed under the same conditions for use as a standard. The early eluting peak contained a P1-resistant AG dinucleotide but was completely digested to mononucleotides by SV, while the late-eluting peak had a SV-resistant AG dinucleotide, but was completely digested to mononucleotides by P1, indicating that the early peak contained an oligonucleotide with an (R p )-phosphorothioate and the late peak contained an oligonucleotide with an (S p )-phosphorothioate (27,28).
To make the end-processing substrate, each oligonucleotide containing a homogeneous stereoisomer was phosphorylated at its 5Ј end with T4 polynucleotide kinase and cold ATP. After heat inactivation of T4 polynucleotide kinase, the phosphorothioate-containing oligonucleotide was annealed to VT4 and VB2. Ligase buffer and ligase were added to the annealed oligonucleotides, and the reaction mixture was incubated overnight at 16°C. Sequenase 2.0 and [␣-32 P]TTP were then used to fill in the 3Ј nucleotide of the VT4 strand ligated to the VT3 strand (see Fig.  5A). The phosphodiester substrate was made by annealing VT5 to VB1 and radiolabeling the 3Ј end of VT5 in the same manner. EDTA (25 mM) was added to stop the reaction, and the reaction products were passed over a Sephadex G-15 column to remove unincorporated [␣-32 P]TTP. The 21-mer product was purified by denaturing gel electrophoresis, quantitated, and annealed to VB1.

Reaction Conditions
Disintegration-The reaction buffer for kinetic experiments performed with a dumbbell disintegration substrate ( Fig. 2A) contained 20 mM HEPES, pH 7.5, 10 mM dithiothreitol, 10 mM MnCl 2 , 0.05% Nonidet P-40, and 10 mM NaCl. Reactions were performed in duplicate and with 1 M disintegration substrate and 200 nM integrase and were incubated at 37°C. Aliquots were removed at specified times, quenched with formamide-containing gel-loading buffer, and substrates and products were separated by denaturing electrophoresis and quantitated using a Molecular Dynamics PhosphorImager. Results from duplicate experiments were averaged.
The disintegration reactions from which product was isolated for stereochemical analysis were carried out with 200 nM integrase and 100 nM substrate using the above buffer and conditions. Products were purified by denaturing gel electrophoresis, eluted in 0.5 M ammonium acetate plus 10 mM magnesium acetate, precipitated with ethanol, and resuspended in TE for digestion by the stereospecific phosphodiesterases.
Integration-Integration reactions were performed for stereochemical analysis in the same buffer used for the disintegration reaction by preincubating 20 nM radiolabeled viral end DNA and 300 nM integrase in 100 l for 5 min at 37°C and then adding 1 M target DNA (either TT4 annealed to TB3 or TT5 annealed to TB4). After incubation for 1.5 h at 37°C, reactions were quenched with formamide loading buffer, and the products purified by electrophoresis as described above for the disintegration product analysis.
3Ј End Processing-Reactions were carried out at 37°C with 20 mM HEPES, pH 7.5, 10 mM DTT, 50 g/ml bovine serum albumin, and 7.5 mM of either MgCl 2 or MnCl 2, and 300 nM dithiothreitol and 2.75 nM substrate (all phosphodiester, (R p )-phosphorothioate at the reaction center, or (S p )-phosphorothioate at the reaction center, see "End-processing Substrates"). In reactions containing 7.5 mM MgCl 2 , 5% polyethylene glycol and 10% dimethyl sulfoxide were included (29). Glycerol (10% v/v) was added in one of the two sets of reactions performed with 7.5 mM MnCl 2 . Aliquots were removed at specified times and quenched. Samples were loaded onto a 24% polyacrylamide plus 8 M urea denaturing gel in 1ϫ TTE (89 mM Tris, pH 8.5, 29 mM taurine, and 0.5 mM EDTA) with 0.5ϫ TTE gel running buffer. Duplicate reactions were quantitated using a Molecular Dynamics PhosphorImager and averaged.
Digestion with Snake Venom and P1 Phosphodiesterase-P1 phosphodiesterase (P1) was resuspended at 1 mg/ml in 10 mM sodium acetate, pH 5.2. Digestions with P1 were performed in 10 mM sodium acetate, pH 5.2, at room temperature with 17 g/ml P1 for the disintegration products and at 37°C with 100 g/ml P1 for the integration products. SV was resuspended at 2 mg/ml in 110 mM Tris-HCl, pH 9.0, containing 110 mM NaCl, 15 mM MgCl 2, and 50% glycerol. Digestions with SV (200 g/ml) were performed in 100 mM Tris-HCl, pH 9.0, 100 mM NaCl, and 15 mM MgCl 2 at 37°C for the integration reaction products and 44°C for the disintegration product. The samples were electrophoresed as described for 3Ј end processing.

RESULTS
Stereospecific Disintegration of Phosphorothioate Substrates-A dumbbell disintegration substrate, composed of a single oligonucleotide that folds on itself to form a structure mimicking one viral DNA end joined to target DNA ( Fig. 2A; Ref. 22), was chemically synthesized with either a standard phosphodiester or a racemic phosphorothioate at the junction between viral and target DNA. The kinetics of disintegration of these two substrates were compared (Fig. 2B). Approximately 85% of the substrate with the standard phosphodiester bond at the junction was converted to product by 90 min, but only 45% of the substrate with the racemic phosphorothioate at the junction was converted to product even after 300 min. As chemically synthesized phosphorothioates are typically composed of a roughly 1:1 mixture of the two stereoisomers (along with a small amount of contaminating phosphodiester material) (30), the leveling-off of product formation at approximately 50% suggested that integrase preferentially disintegrated one of the two stereoisomers.
To test and quantitate this preference for one stereoisomer, we purified the putative less preferred stereoisomer by isolating the residual substrate following overnight reaction of the dumbbell disintegration substrate with integrase. This mate-rial was then radiolabeled and used as a substrate in a second disintegration reaction (Fig. 2C). Parallel reactions were carried out with otherwise identical substrates containing a standard phosphodiester bond or a racemic phosphorothioate at the reaction center. The rate of disintegration of the phosphodiester substrate was at least 50-fold greater than that of the less preferred phosphorothioate stereoisomer. This value is a lower limit because of potential contamination of the less preferred stereoisomer substrate with either the phosphodiester substrate or the preferred phosphorothioate stereoisomer. If the substrate containing the racemic phosphorothioate was not substantially contaminated with phosphodiester substrate, and the stereoisomers were present initially at a ratio near 1:1, then the disintegration rate of the preferred phosphorothioate stereoisomer was about 2-fold lower than that of the phosphodiester substrate.
The (S p )-Phosphorothioate Stereoisomer Is the Preferred Substrate for Disintegration-Two stereospecific phosphodiesterases were used to determine the stereochemistry of products of the integration and disintegration reactions. P1 has a high FIG. 2. Disintegration reactions with a dumbbell disintegration substrate. A, the dumbbell substrate. B, time courses for reactions of substrates with a phosphodiester (q) or a racemic phosphorothioate (ࡗ) at the junction between viral and target DNA (the reaction center). Reactions were carried out as described under "Experimental Procedures" with 10 mM NaCl. C, time courses for reactions of substrates with a phosphodiester (q), a racemic phosphorothioate (ࡗ), or the less preferred phosphorothioate (छ) at the reaction center. The NaCl concentration in the reaction buffer for these reactions was 30 mM instead of 10 mM, which presumably accounts for the difference in the kinetics relative to the reactions shown in panel B (26). stereospecificity for (S p )-phosphorothioates (28), whereas SV has a high stereospecificity for (R p )-phosphorothioates (27).
A Y-mer disintegration substrate was constructed to determine the stereochemistry of the product of the disintegration reaction (Fig. 3A, see also "Y-mer Disintegration Substrates" under "Experimental Procedures"). After incubation at 37°C for 4 h, approximately 65% of the phosphodiester substrate and 15% of the phosphorothioate substrate had been converted to product (data not shown).
The ligated target DNA strands from these disintegration reactions were isolated by gel electrophoresis and analyzed by digestion with P1 and SV. With the racemic phosphorothioate substrate, the ligated target DNA strand resulting from the disintegration reaction (Fig. 3B, lane 10) contained a dinucleotide that was resistant to digestion with P1 (lanes 11-14), but was completely digested to mononucleotides by SV (lanes [15][16][17][18]. This cleavage pattern is expected for the (R p )-phosphorothioate isomer. The product originating from the phosphodiester substrate (Fig. 3B, lane 1) was completely digested by both P1 (lanes 2-5) and SV (lanes 6 -9), under conditions identical to those used to digest the phosphorothioate-containing product. Control reactions with purified phosphorothioate stereoisomers validated the stereospecificity of these preparations of P1 and SV (see "Experimental Procedures"). Assuming that disintegration, like 3Ј end processing and integration (5), proceeds with inversion about the reactive phosphoryl group, integrase preferentially catalyzed disintegration of the substrate containing the S p stereoisomer at the reaction center. Densitometry of the dinucleotide remaining in the final time point of the SV digestion (Fig. 3B, lane 18) gave a ratio of ϳ50 for the mononucleotide relative to the dinucleotide digestion product, suggesting a strong preference for the (S p )-phosphorothioate substrate over the (R p )-phosphorothioate substrate.
The (R p )-Phosphorothioate Stereoisomer Is the Preferred Target DNA Substrate for Integration-Phosphorothioate substrates were designed to monitor the stereochemistry of the products of the integration reaction (Fig. 4A). The phosphodiester bond immediately 3Ј to the terminal adenosine residue of the viral DNA substrate was radiolabeled. The purified products from integration into either all-phosphodiester target DNA (Fig. 4B, lane 8) or the phosphorothioate-containing target DNA (lane 1) were digested separately with P1 or SV. The products of integration into the all-phosphodiester target DNA were completely digested to mononucleotides by both SV (lanes 9 -11) and P1 (lanes 12-14). After extensive digestion of the products of integration into the phosphorothioate-containing target DNA with SV, residual AC and AG dinucleotides persisted (lanes 2-4), suggesting that these dinucleotides had the S p stereochemistry expected from attack of the viral 3Ј hydroxyl on (R p )-phosphorothioates 5Ј to a C or G. The mononucleotides were the products expected from SV digestion of integration products in which the viral ends had joined 5Ј to an A or T, residues without phosphorothioate substitution. Because the target oligonucleotide contained runs of up to six consecutive G and C nucleotides, each containing phosphorothioate linkages, SV digestion would also be expected to yield smaller quantities of products larger than a dinucleotide, as was observed (Fig. 4B, lanes 3 and 4). In contrast, the absence of an accumulation of dinucleotide and longer products with P1 digestion suggests that there was no significant integration reaction with the (S p )-phosphorothioates.
The (R p )-Phosphorothioate Stereoisomer Is the Preferred Substrate for 3Ј End Processing-Phosphorothioate-containing oligonucleotides mimicking a viral DNA end were designed to monitor the stereochemistry of the products of the end-processing reaction (Fig. 5A). To determine integrase's stereospecific-ity for end processing, we compared the rates of end processing of stereochemically pure viral DNA end substrates containing either an (S p )-or (R p )-phosphorothioate at the reaction center.
phodiester substrates were tested in parallel end-processing reactions. The end-processing rate for the R p stereoisomer substrate was at least 50-fold greater than the rate for the S p stereoisomer substrate, when Mn 2ϩ was used as the metal ion cofactor (Fig. 5B). The rate of end processing for the phosphodiester substrate was only slightly higher than the rate observed for the (R p )-phosphorothioate substrate. The same stereospecificity was observed with Mg 2ϩ as the metal ion cofactor (Fig.  5C); the (R p )-phosphorothioate substrate was processed at a rate 40-fold higher than that observed with the (S p )-phosphorothioate substrate, and the phosphodiester substrate was processed slightly faster than the (R p )-phosphorothioate substrate.
Integrase also preferred the (R p )-phosphorothioate substrate for end processing in reactions carried out in the presence of 7.5 mM MnCl 2 and 10% glycerol (data not shown). It has been previously shown that end processing can give three chemically distinct dinucleotide products, depending on the nucleophile used (5,31). When Mg 2ϩ is included as the divalent cation, water is the primary nucleophile, and the main product is the hydrolysis product, a linear GT dinucleotide. However, in the presence of Mn 2ϩ , a variety of nucleophiles can be used, including glycerol, generating the dinucleotide (GT) product linked at its 5Ј end to glycerol, or the terminal 3Ј hydroxyl of the viral DNA, producing a 3Ј-5Ј cyclic dinucleotide. Furthermore, the (R p )-phosphorothioate substrate and the phosphodiester substrate yielded similar proportions of cyclic product (data not shown). Thus, the substitution of sulfur for a non-bridging oxygen did not appear to alter the propensity of the terminal 3Ј hydroxyl to reach and attack the reactive subterminal viral phosphodiester relative to attack by water. In contrast, the proportion of product formed from the use of glycerol as a nucleophile was 4-fold lower for the (R p )-phosphorothioate substrate than for the phosphodiester substrate (Fig. 5D). Since sulfur is a much poorer hydrogen bond acceptor than oxygen, this result suggests that hydrogen bonding between the cishydroxyl of the glycerol and the oxygen of the reactive phosphate helps to position the glycerol for nucleophilic attack. Such an interaction could account for the preference for cisdiols relative to simple alcohols as nucleophiles in the endprocessing reaction (31). DISCUSSION The observed stereospecificity of HIV-1 integrase in catalyzing end processing, integration, or disintegration of model substrates with phosphorothioates at the reaction center places new constraints on models for the orientation of reactants in the enzyme's active site. A model for the arrangement of substrate features in the active site of HIV integrase, based on these results and previous observations, is illustrated in Fig. 6. This model begins with three assumptions. 1) The transition state is a trigonal bipyramid, with the nucleophile and leaving groups occupying the apical positions for an in-line attack and the other three, equatorial substituents coplanar with the central phosphorus, as expected from chemical and enzymological precedents (32,33). 2) Integrase uses a single active site to catalyze end processing, integration, and disintegration, as suggested by mutational and structural results (6 -13). 3) Certain specific interactions in the active site are the same for all three reactions (see below). In developing this model, we first consider groups occupying the apical positions of the trigonal bipyramid transition state, basing our discussion on previous results. We then turn to results from the present study that pertain to the placement of groups in the equatorial positions of the transition state.
Both in vivo and in vitro results suggest that the phylogenetically conserved CA/TG dinucleotide pair adjacent to the cleavage site at the viral DNA end is a crucial feature for recognition of viral DNA by integrase. In vitro, changing the phylogenetically conserved CA/TG dinucleotide pair adjacent to the cleavage site in a model substrate results in up to a FIG. 4. Determination of the stereospecificity of the integration reaction. A, the model viral DNA end substrate was labeled by incorporating [ 32 P]dATP at the 3Ј end. The phylogenetically conserved CA/TG dinucleotide pair is highlighted in bold type. Target DNA contained either all phosphodiesters (TT4 annealed to TB3, see "Experimental Procedures") or phosphorothioates (indicated by (S), TT5 annealed to TB4) 5Ј of all C and G nucleotides. Following the integration reaction, integration products were purified. B, digestion of integration products by stereospecific phosphodiesterases. The integration products from the reactions in which the target DNA contained phosphorothioates 5Ј to the C and G nucleotides (lane 1, undigested) were digested with SV (lanes 2-4) and P1 (lanes 5-7). The integration products from the reactions in which the target DNA contained all phosphodiesters (lane 8, undigested) were digested with SV (lanes 9 -11) and P1 (lanes [12][13][14]. Aliquots were removed from the phosphodiesterase digestions at 2 min (lanes 2, 5, 9, and 12), 30 min (lanes 3, 6, 10, and 13), and 120 min (lanes 4, 7, 11, and 14).
100-fold reduction in end processing or disintegration activity (4, 14, 20, 25, 34 -39). In vivo, integration of viral genomes in which the conserved CA/TG dinucleotide pair has been altered by mutation is impaired by a factor of 10 5 . Since the CA/TG dinucleotide pair is critical for all three reactions, it is reasonable to assume that this substrate feature is bound in the same specific site for end processing, integration, and disintegration. This site is depicted in yellow in Fig. 6.
In the transition state for phosphoryl transfer, the nucleophile and the leaving group typically occupy the apical positions. As noted above, previous results have strongly suggested that the phylogenetically conserved adenosine of the viral DNA end is bound in the same specific site for all three reactions. Because the 3Ј hydroxyl of this adenosine is either the attacking or leaving group in each of the integrase-catalyzed reactions, the model places the fixed binding site for this 3Ј hydroxyl in an apical position, highlighted in yellow in the illustration (Fig. 6). The other attacking or leaving group in each reaction, the identity of which depends upon the reaction, occupies the opposite apical position, highlighted in blue (Fig.  6). According to the proposed model, the "blue" apical position would then be occupied by water or an alcohol hydroxyl acting as the nucleophile in end processing, or a target DNA 3Јhydroxyl acting as the nucleophile or the leaving group in disintegration and integration, respectively.
We now consider the groups that occupy the equatorial positions. The stereospecificity observed for phosphorothioate stereoisomers in the end-processing, integration, and disintegration reactions suggest that, in each of these reactions, the active site interacts with one of the non-esterified phosphoryl oxygen atoms in manner that discriminates against a substituted sulfur atom. Fig. 6 presents the simplest model to account for these observations, in which a single site, depicted in pink, interacts with a non-esterified oxygen atom, but cannot readily accommodate a sulfur atom. Once the occupants of the two apical positions and one of the three equatorial positions in the transition state are specified, the observed stereospecificities of the three integrase-catalyzed reactions dictate the ar- FIG. 5. Determination of the stereospecificity of the integration reaction. A, a model viral DNA end substrate was prepared as described under "Experimental Procedures" by separating a racemic phosphorothioate-containing 5-mer into two stereochemically pure peaks (VT3, R p and S p ), annealing each to the complementary bottom strand (VB1) and the 5Ј portion of the top strand (VT4), ligating the two top strand pieces together, and finally, 3Ј end-labeling the top strand with [␣-32 P]TTP (see "Experimental Procedures"). The 5-mer is shaded, the phylogenetically conserved viral CA/TG base pairs are in bold type, and the arrowhead indicates the site of processing by integrase. The phosphodiester substrate was prepared by 3Ј end-labeling a top strand lacking the final T with [␣-32 P]TTP. B and C, end-processing kinetics for viral DNA end substrates containing an (R p )-phosphorothioate (OE), (S p )-phosphorothioate (छ), or phosphodiester (E) at the cleaved position. Either 7.5 mM MnCl 2 (panel B) or 7.5 mM MgCl 2 (panel C) was included as the divalent metal ion cofactor for the end-processing reactions. D, substitution of a sulfur atom for a non-bridging oxygen has an unfavorable effect on the ability of glycerol to serve as the nucleophile for 3Ј end processing of a model viral DNA end substrate. Reactions were carried out in the presence of 7.5 mM MnCl 2 and 10% glycerol. The graph shows the time course of accumulation of the glycerol-linked dinucleotide product from reactions with the phosphodiester-containing substrate (E), or an (R p )-phosphorothioate substrate (‚), as well as the time course for accumulation of all end-processing products (including the hydrolysis product, the glycerolysis product, and the cyclic product) from the same reactions with a phosphodiester-containing substrate (q), or an (R p )-phosphorothioate substrate (OE). rangements of the remaining substituents. Somewhat surprisingly, the results strongly imply that the site occupied by the target DNA extending 3Ј from the reaction center (indicated in purple in Fig. 6) is not the same as the site that accommodates the 3Ј terminal dinucleotide of the viral DNA end in the endprocessing reaction (indicated in green in Fig. 6). The basis for this specific arrangement is not known. One possibility is that the site that binds the 3Ј-terminal dinucleotide in the endprocessing reaction (green) interacts favorably with a single strand of DNA in a configuration that it can only adopt when not base paired. If double-helical DNA cannot be accommodated in this (green) site, but can be accommodated in the other equatorial site (purple), then the target DNA extending 3Ј from the reaction center would be constrained to occupy this (purple) position. DNA targets can vary widely with respect to the size and structure of the features 3Ј to the reaction center (22). Thus, this (purple) site may better be viewed as a space that can accommodate this portion of the target DNA rather than a site that provides crucial specific binding interactions. Although results of previous photo-cross-linking studies of the integrase-DNA complex were not explicitly used in developing the model shown in Fig. 6, the resulting model is entirely consistent with the architecture of the integrase-DNA complex inferred from those cross-linking studies (43).
Macroscopically, disintegration is the reverse of the integration reaction. Moreover, the substrate specificity of the reactions is homologous (22,26,35,44). Integrase's preference for opposite phosphorothioate stereoisomers for integration and disintegration provides further support for a reciprocal relationship between the two reactions and for a model in which the local arrangement of viral and target DNA in the active site is the same for the two reactions.
Enzymes in the integrase/transposase family of polynucleotidyl transferases all apparently catalyze cleavage and joining of the transposable element DNA by similar mechanisms (5,45,46). Many of these enzymes, including the VDJ recombinase, the bacteriophage Mu transposase, HIV-1 integrase, and murine leukemia virus integrase, appear to disrupt base pairing in the vicinity of the reaction center in their otherwise doublehelical substrates, and these enzyme-induced disruptions of the double-helical structure of their substrates can facilitate catalysis in vitro (14 -19, 47, 48). This local disruption of DNA The other axial position in the transition state (blue) is occupied by a 3Ј hydroxyl of the target DNA molecule, which acts as a leaving group or nucleophile, respectively, in the integration and disintegration reactions, or by the water molecule that acts as the nucleophile in the endprocessing reaction. One of the equatorial positions in the transition state is defined by a site (pink) that binds the non-bridging oxygen, but cannot readily accommodate a sulfur atom. A second equatorial position (green) is defined by a binding site that provides favorable interactions with the 3Ј dinucleotide of the viral DNA end in the 3Ј end-processing reaction, and which can accommodate a sulfur atom of a phosphorothioate substrate in the integration and disintegration reactions, but which cannot accommodate a DNA duplex. The third equatorial position (purple) is defined by a site that can accommodate the target DNA 3Ј to the site of integration/disintegration, or the sulfur atom (in the end-processing reaction with a phosphorothioate substrate), but which does not appear to provide specific favorable binding interactions with either the viral or target DNA substrates. Line segments between substrate features and the binding sites represent binding interactions. structure appears to be a common means by which these enzymes manage to use the same active site for donor cleavage and joining. It will now be interesting to investigate whether the evolutionary conservation of structural and mechanistic features of the enzymes in this family will extend to the arrangement of the analogous DNA substrates in their active sites.