Relationship between the Oligomeric Status of HIV-1 Integrase on DNA and Enzymatic Activity*

The 3′-processing of the extremities of viral DNA is the first of two reactions catalyzed by HIV-1 integrase (IN). High order IN multimers (tetramers) are required for complete integration, but it remains unclear which oligomer is responsible for the 3′-processing reaction. Moreover, IN tends to aggregate, and it is unknown whether the polymerization or aggregation of this enzyme on DNA is detrimental or beneficial for activity. We have developed a fluorescence assay based on anisotropy for monitoring release of the terminal dinucleotide product in real-time. Because the initial anisotropy value obtained after DNA binding and before catalysis depends on the fractional saturation of DNA sites and the size of IN·DNA complexes, this approach can be used to study the relationship between activity and binding/multimerization parameters in the same assay. By increasing the IN:DNA ratio, we found that the anisotropy increased but the 3′-processing activity displayed a characteristic bell-shaped behavior. The anisotropy values obtained in the first phase were predictive of subsequent activity and accounted for the number of complexes. Interestingly, activity peaked and then decreased in the second phase, whereas anisotropy continued to increase. Time-resolved fluorescence anisotropy studies showed that the most competent form for catalysis corresponds to a dimer bound to one viral DNA end, whereas higher order complexes such as aggregates predominate during the second phase when activity drops off. We conclude that a single IN dimer at each extremity of viral DNA molecules is required for 3′-processing, with a dimer of dimers responsible for the subsequent full integration.

The 3-processing of the extremities of viral DNA is the first of two reactions catalyzed by HIV-1 integrase (IN). High order IN multimers (tetramers) are required for complete integration, but it remains unclear which oligomer is responsible for the 3-processing reaction. Moreover, IN tends to aggregate, and it is unknown whether the polymerization or aggregation of this enzyme on DNA is detrimental or beneficial for activity. We have developed a fluorescence assay based on anisotropy for monitoring release of the terminal dinucleotide product in realtime. Because the initial anisotropy value obtained after DNA binding and before catalysis depends on the fractional saturation of DNA sites and the size of IN⅐DNA complexes, this approach can be used to study the relationship between activity and binding/multimerization parameters in the same assay. By increasing the IN:DNA ratio, we found that the anisotropy increased but the 3-processing activity displayed a characteristic bell-shaped behavior. The anisotropy values obtained in the first phase were predictive of subsequent activity and accounted for the number of complexes. Interestingly, activity peaked and then decreased in the second phase, whereas anisotropy continued to increase. Time-resolved fluorescence anisotropy studies showed that the most competent form for catalysis corresponds to a dimer bound to one viral DNA end, whereas higher order complexes such as aggregates predominate during the second phase when activity drops off. We conclude that a single IN dimer at each extremity of viral DNA molecules is required for 3-processing, with a dimer of dimers responsible for the subsequent full integration.
The integration of a DNA copy of the HIV-1 2 genome into the host genome is a crucial step in the life cycle of the retrovirus. Integrase (IN) is responsible for the two consecutive reac-tions that constitute the overall integration process. The first of these two reactions is 3Ј-processing, which involves cleavage of the 3Ј-terminal GT dinucleotide at each extremity of the viral DNA. The hydroxyl groups of newly recessed 3Ј-ends are then used in the second reaction, strand transfer, for the covalent joining of viral and target DNAs, resulting in full-site integration. IN is sufficient for catalysis of the 3Ј-processing reaction in vitro, using short-length oligodeoxynucleotides (ODNs) that mimic one viral long terminal repeat (LTR) in the presence of the metallic cofactor Mg 2ϩ . This reaction generates two products: the viral DNA containing the recessed extremity and the GT dinucleotide. One of the two products, the processed viral DNA, as well as the target DNA serve as substrates for the subsequent joining reaction.
IN belongs to the superfamily of polynucleotidyl transferases. Its catalytic core domain contains a triad of acidic residues constituting the D,D-35-E motif, which is strictly required for catalysis. The catalytic core establishes specific contacts with the viral DNA and, together with the C-terminal domain, is involved in DNA binding (1)(2)(3)(4). The 3Ј-processing reaction is highly specific, and the terminal 13 bp of the LTR play a key role for Mg 2ϩ -dependent 3Ј-processing in terms of reaction specificity (4,5). In particular, the CA sequence preceding the GT dinucleotide cleaved by IN is strictly required. Reaction specificity seems to depend on the catalytic step, because no significant difference in affinity is observed in vitro for different DNA sequences (2,3,5,6). The catalytic mechanism of IN has been extensively studied, but the structural determinants of IN activity remain unclear, and controversy remains concerning the multimeric status of active IN. Singleand double-domain crystallographic structures have been produced for IN, alone or complexed with the IN-binding domain of the cellular partner lens epithelium-derived growth factor, and all these structures display a conserved dimeric structure of the catalytic core (7)(8)(9)(10)(11). Based on topological considerations concerning the distance separating the active sites of the two protomers, which is more than double the distance separating the two joining sites on the target DNA (separated by 5 bp), a multimeric state of an order higher than dimers, tetramers at least, is required for complete integration. However, it remains unclear whether such a higher order multimeric state is also required for 3Ј-processing activity and how this activity is modulated by the self-association properties of IN. Moreover, most enzymatic studies are carried out with an excess of IN over DNA substrate. Taking into account the low solubility of IN, we investigated whether high order multimers or aggregated forms of IN favored by high protein concentrations in solution were detrimental or beneficial for activity.
We developed an assay for monitoring DNA binding and subsequent 3Ј-processing in the same sample. This assay makes possible the separation of binding and catalytic parameters and the study of real-time kinetics. It is based on steady-state fluorescence anisotropy (r), which is sensitive to rotational diffusion, and thus suitable for studies aiming to identify structural modifications leading to a significant change in molecular size (12)(13)(14)(15). Using a fluorescent probe covalently linked to the GT dinucleotide, this makes it possible to follow DNA binding and dinucleotide release, because both steps strongly influence molecular size of the fluorescent moiety. The anisotropy-based technique is also highly suitable for studies of the relationship between the overall size of IN⅐DNA complexes and activity. We found that, for low IN:DNA ratios, the r values obtained after DNA binding and before catalysis were fully predictive of subsequent IN activity, according to the fractional saturation function. For high IN:DNA ratios, anisotropy continued to increase, but 3Ј-processing activity decreased. Because r depends on both fractional saturation and the molecular size of complexes at saturation of DNA sites, our results show that high order multimers or aggregated states of IN are detrimental to 3Ј-processing activity. Activity levels were highest for non-aggregative smaller species. A more precise characterization of catalytically competent complexes by time-resolved fluorescence anisotropy (TFA) (16 -20), confirmed steady-state data and led to the identification of dimeric forms as the most active forms for 3Ј-processing. Our results also highlight that only the IN:DNA ratio but not the IN concentration per se determines the aggregation properties and thus the activity of IN and that DNA binding stimulates the self-organization of IN to give a catalytically competent non-aggregative form.

EXPERIMENTAL PROCEDURES
Oligonucleotides and Nomenclature-Unlabeled and fluorescein-labeled single-stranded (ss) ODNs were purchased from Eurogentec (Liege, Belgium) (except for the 2Ј-aminouridinecontaining ODN) and purified by electrophoresis in acrylamide gels. The specific HIV sequence was 5Ј-GTGTGGAAAATCT-CTAGCAGT-3Ј (the bases removed by IN are underlined). This sequence was denoted a, and the complementary nonprocessed strand was denoted b. Fluorescein (F) was attached at the 5Ј-or the 3Ј-end of strand a or b, via a 6-carbon linker. As an example, the specific double-stranded (ds) ODN used to monitor activity was called HIV-a3F, indicating that fluorescein was attached to the 3Ј-end of strand a. HIV-GTGT-a3F was identical except that the four terminal bases at the 3Ј-end of strand a were GTGT rather than the canonical sequence CAGT. The nonspecific sequence, NS-TTCC, was 5Ј-ACCTATGCGCCG-CTAGATTCC-3Ј (strand a). Two other nonspecific ODNs with different 3Ј-ends of strand a were derived: NS-CACC and NS-CAGT. ds ODNs were obtained by mixing equimolar amounts of complementary strands in 20 mM Tris-HCl (pH 7.2), 100 mM NaCl. The mixture was heated to 85°C for 5 min, and annealing was allowed by slow cooling to 25°C.
The uridine-containing ODN (corresponding to strand a of the DNA substrate HIV-aUF) contains a fluorescein attached to the 2Ј-amino group of the 3Ј-terminal 2Ј-aminouridine. It was synthesized as follows: the solid support was prepared as previously described (21). Briefly, 2Ј-deoxy-5Ј-O-4,4Ј-dimethoxytrityl-2Ј-O-trifluoracetamidouridine (0.2 mmol) was co-evaporated with pyridine (3 ϫ 5 ml) and dissolved in dry pyridine (2 ml). Succinylated long-chain alkylamine-controlled pore glass (500 Å) (200 mg), 2,4,6-triisopropylbenzenesulfonyl chloride (0.6 mmol), and 1-methylimidazole (1.2 mmol) were added, and the mixture was incubated for 3 h at 25°C. The support was then filtered and washed successively with pyridine, CH 2 Cl 2 , and ether. Starting from 2Ј-deoxy-5Ј-O-4,4Јdimethoxytrityl-2Ј-O-trifluoracetamidouridine-derivatized long-chain alkylamine-controlled pore glass (500 Å), a 21-mer ODN was assembled on an ABI394B DNA synthesizer, by the phosphoramidite method, according to the manufacturer's recommendation. Protected 2Ј-O-deoxyribonucleoside phosphoramidites and S-ethylthiotetrazole were purchased from Glen Research. Ammonia was used for cleavage from the support and deprotection overnight at 55°C. The reaction mixture was then analyzed by reverse-phase high-performance liquid chromatography in ion-pair mode. For fluorescein labeling, 1 mg of fluorescein isothiocyanate dissolved in 20 l of dimethylformamide was added to a solution of 40 nmol of 2Ј-amino ODN in 130 l of sodium carbonate-bicarbonate (1 M, pH 9):water (v/v, 5:8) (22,23). The mixture was incubated at 25°C in the dark for 18 h and then loaded onto an NAP5-Sephadex G-25 column (Amersham Biosciences) pre-equilibrated in water. The ODN was eluted with water and analyzed by reversed-phase high-performance liquid chromatography in ion-pair mode.
Steady-state Fluorescence Anisotropy Assay-Steady-state anisotropy values were recorded on a Beacon 2000 instrument (PanVera, Madison, WI), in a cell thermostatically held at 25 or 37°C, for the DNA-binding step and 3Ј-processing reaction, respectively. Unless otherwise stated, we studied the formation of IN⅐DNA complexes by incubating fluorescein-labeled ds ODNs with IN in 20 mM Tris (pH 7.2), 1 mM dithiothreitol, 20 mM NaCl, 5 mM MgCl 2 (the sample volume was 200 l). The fractional saturation function (Y) was calculated as follows, where r max and r ODN are the anisotropies of IN-bound and free ODN, respectively (no significant concomitant change in fluorescence intensity was observed). After the DNA-binding step, the temperature was raised from 25 to 37°C for monitoring of where r NP and r dinu are the anisotropy values for pure solutions of non-processed ds ODN and dinucleotide, respectively (fluorescence did not change significantly during the reaction). We used the 5Ј-GT-3ЈF dinucleotide (Eurogentec) to determine r dinu . (ii) In real-time conditions, an additional fluorescent population corresponding to IN complexed with the unprocessed ds ODN, is present in the sample. In this case, F dinu was calculated as follows, where r max is the characteristic r value obtained for optimal activity, and r t ϭ 0 is the r value obtained at the end of the DNAbinding step (before the start of the reaction). The 3Ј-processing activity obtained with Equations 2 and 3 is referred to as Activity SDS and Activity real-time , respectively. Activity real-time was not used if r t ϭ 0 was higher than 0.22 (aggregation of IN on DNA not negligible). We analyzed single-turnover kinetics using the Equations 4 and 5, where k chemistry is the single-turnover rate constant, and K d,app is the apparent K d (25). Reactions were conducted in the presence of Mg 2ϩ (not Mn 2ϩ ) to limit nonspecific hydrolysis products (cleavages at positions Ϫ3, Ϫ4, etc.), because anisotropy cannot discriminate between these small products (which are minor products with Mg 2ϩ but not with Mn 2ϩ ) and the specific GT product. The r values for the TAMRA-labeled DNA substrate ( ex ϭ 562 nm and em ϭ 582 nm) were recorded on a Cary Eclipse spectrofluorometer (Varian, Mulgrave, Australia) in polarization mode and were compared with values obtained with the same instrument and HIV-a3F as the substrate.
K d,app as a function of ODN length were determined by competition experiments. Fluorescein-labeled ds ODN HIV-a5F (4 nM) was preincubated with various concentrations of unlabeled ss or ds HIV ODN (from 10-to 45-mers) in 20 mM Tris buffer (pH 7.2) supplemented with 1 mM dithiothreitol, 20 mM NaCl, and 5 mM MgCl 2 . IN (150 nM final concentration) was then added, and r values were recorded. ⌬r (ϭ r Ϫ r ODN ) was plotted against competitor concentration to determine K d,app (concentration of competitor decreasing the initial ⌬r value by 50%).
Time-resolved Fluorescence Experiments-Time-resolved fluorescence parameters (lifetimes and correlation times) were obtained from the two polarized fluorescence decays I Ќ (t) and I ʈ (t), using the time-correlated single photon counting technique. The instrumentation setup was essentially similar to those previously described (16,17), with modifications: the time scaling was 19.5 ps per channel and 4096 channels were used. The excitation light pulse source was a Ti:sapphire laser (Millennia-pumped Tsunami femtosecond laser, Spectra Physics) (repetition rate: 8 MHz) associated with a second harmonic generator tuned to 490 nm. The emission monochromator (ARC SpectraPro-150) was set to 530 nm (⌬ ϭ 15 nm). The two polarized components were collected alternately over a period of 30 s (total count of I ʈ : 15,000,000). The reaction mixture contained 20 mM Tris, pH 7.2, 20 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol. The correlation time () distributions of free ss or ds ODNs were obtained at ODN concentrations of 10 nM or 0.5 M. All ODNs (from 10-to 45-mer) mimicked the U5-end of the HIV-1 DNA and were fluorescein-labeled at the 5Ј-end of strand a. IN⅐DNA complexes were analyzed using ds HIV-a5F and various IN:DNA ratios, from 40 to 400. We analyzed both decays, I Ќ (t) and I ʈ (t), by the maximum entropy method (26). Fluorescence anisotropy decay is described by Equations 6 and 7, where i is the individual rotational correlation time, and i is the associated amplitude. 0 was found to lie between 0.36 and 0.37. Normalization of for a given temperature was performed using, where is the viscosity, V is the volume of the rotating unit, k is the Boltzmann constant, and T is the temperature (K).

RESULTS
Monitoring 3Ј-Processing Activity by Steady-state Fluorescence Anisotropy-Fluorescence anisotropy measurements are based on the principle of photoselective excitation of a fluorophore by a polarized light, providing information about rotational motions of the fluorophore or fluorescently labeled molecule between photon absorption and emission. Some events such as overall rotational diffusion or flexibility are major causes of light depolarization. High levels of anisotropy are generally associated with large molecules or complexes characterized by slow rotational diffusion or low flexibility level. In this study, we used an extrinsic fluorophore covalently linked to DNA to monitor the binding of IN to viral DNA substrate and the subsequent 3Ј-processing reaction, in the same assay. Both DNA binding and 3Ј-processing would be expected to have a significant effect on the anisotropy parameter, because each of these steps has a major effect on the molecular size of the fluorescent moiety. The principle underlying the anisotropy-based assay is summarized in Fig. 1. The binding of IN to a ds ODN mimicking one end of the viral DNA increases the steady-state anisotropy value (r), most likely by restricting flexibility, allowing the calculation of fractional saturation function. Following DNA binding, the same sample can be shifted to a permissive temperature for 3Ј-processing activity. When the fluorophore is linked to the 3Ј-terminal GT dinucleotide, release of the dinucleotide product should significantly decrease r. The fraction of dinucleotides released can be determined from the real-time decrease in r observed during the reaction (⌬r real-time , corresponding to the difference between IN⅐DNA complex and free dinucleotide) or after the disruption of all IN⅐DNA complexes by the addition of SDS (⌬r SDS , corresponding to the difference between unprocessed ds ODN and free dinucleotide). Unlike most standard 3Ј-processing assays, which are based on quantification of the processed strand (first reaction product), the anisotropy-based assay monitors the released dinucleotide (second reaction product).
We tested 21-mer DNA substrates labeled with fluorescein at one end, in all possible combinations (5Ј-or 3Ј-extremity of the a or b strand), or labeled on the 2Ј-amino group of a 3Ј-terminal 2Ј-aminouridine ( Fig. 2A). The free ds ODNs were characterized by r values of 0.060 -0.130 at 25°C (0.045-0.105 at 37°C), depending on the location of fluorescein, whereas the fluorescein-labeled dinucleotide was characterized by r ϭ 0.02. DNA binding on the addition of IN was monitored at 25°C and led to a significant increase in r value (Ͼ0.2 in the experimental conditions of Fig. 2). Equilibrium was typically reached after 15 min, as previously reported (12,27). The processing reaction was started by shifting the sample to a temperature of 37°C, and the initial r values obtained therefore correspond to r t ϭ 0 . In the presence of the divalent cationic cofactor Mg 2ϩ , all tested ODNs gave similar r t ϭ 0 values (0.22), indicating that fluorescein position had no effect on DNA binding by IN (Fig. 2B).
Experiments carried out in the absence of Mg 2ϩ gave reproducibly larger r values (0.25). This result confirms that DNA-binding activity of IN is not strictly dependent on Mg 2ϩ , consistent with other studies (2,28,29). Nevertheless, the larger r value suggests that IN aggregation is favored by an absence of Mg 2ϩ , as previously reported (17).
The time dependence of r and ⌬r under real-time conditions at 37°C is shown in Fig. 2 (B and C): only the DNA substrates HIV-a3F and HIV-aUF, with the fluorescein directly attached to the small GT reaction product, displayed significant  decreases in r or ⌬r, and these decreases were strictly related to the presence of the metallic cofactor. In contrast, HIV-a5F, HIV-b5F, and HIV-b3F displayed no significant decrease in r or ⌬r value, under either real-time (Fig. 2) or fixed-time conditions (data not shown) (3Ј-processing activity was normally detected by a gel-electrophoresis method for HIV-b5F and HIV-b3F; HIV-a5F was not tested). The result obtained with these three ODNs indicates that anisotropy is not sensitive enough to differentiate between the DNA substrate (ds ODN 21/21) and the first reaction product (ds ODN 19/21). Furthermore, under real-time conditions, this result suggests that the processed DNA product remains tightly bound to the enzyme after 3Ј-processing. This tight binding together with the slow catalytic step could be responsible for the observed single-turnover property of IN, even under conditions of excess DNA substrate (25). In contrast, the terminal dinucleotide is released normally from IN⅐DNA complexes after 3Ј-processing, and our results indicate that the decrease in r is related to the formation of the GT product. Thus, anisotropy is reliable for monitoring the 3Ј-processing reaction, but only if fluorescein is attached to the GT dinucleotide.
The 3Ј-processing reaction is strongly sequence-dependent, and the endonucleolytic site includes the crucial conserved CA dinucleotide immediately preceding the GT dinucleotide. We assessed sequence specificity, using the anisotropy-based assay and the following ODNs: HIV-a3F corresponds to the wildtype (wt) sequence, whereas HIV-GTGT-a3F is a variant in which the 3Ј-terminal CAGT sequence is replaced by GTGT. We also tested three nonspecific sequences, NS-TTCC-a3F, NS-CACC-a3F, and NS-CAGT-a3F. We found that the DNAbinding step was not influenced by sequence. Indeed, the final r values obtained after DNA binding were similar for all ODNs, irrespective of sequence context, confirming that in vitro assays primarily reveal the nonspecific DNA binding mode of IN. In contrast, sequence had a major effect on ⌬r value in catalysis conditions (Fig. 3A). Only HIV-a3F gave a large decrease in ⌬r. This decrease was abolished by replacement of the 3Ј-terminal sequence CAGT by GTGT or the use of nonspecific sequences, even with the CAGT sequence at the 3Ј-end. These results confirm that the CA dinucleotide is strictly required but not sufficient for activity, particularly in the presence of Mg 2ϩ , which gives more stringent conditions than Mn 2ϩ (4,5). We also investigated the effect of a styrylquinoline compound, FZ41, which inhibits 3Ј-processing activity (30). The anisotropybased approach has been shown to be suitable for assays of the effects of inhibitors on the formation and stability of IN⅐DNA complexes, independently of catalysis, and has been successfully used in studies of competitive inhibitors preventing IN⅐DNA interactions (12). Here, we investigated both steps, DNA binding and catalysis, in the presence of inhibitor. The IC 50 value for 3Ј-processing was found to be 0.74 M (Fig.  3B), consistent with the value determined by standard gelelectrophoresis (30). In the same experiment, the study of the DNA-binding step revealed the competitive nature of FZ41: r t ϭ 0 decreased with increasing FZ41 concentration (not shown), suggesting that FZ41 primarily prevents the IN-DNA interaction.
Altogether, our data show that the decrease in r during incubation at 37°C is related to specific 3Ј-processing activity and that the anisotropy-based method is reliable for inhibitor characterization. In our experiments, strand transfer products, as quantified by gel-electrophoresis, represent 5-8% of the total products, regardless of fluorescein labeling (data not shown). Only, the strand transfer of one processed DNA into a fluorescently labeled DNA (not yet processed) may underestimate the 3Ј-processing activity determined by anisotropy. Consequently, the half-transfer reaction is more problematic at the beginning of the reaction, and its influence on r tends to disappear during 3Ј-processing. However, this effect is probably small, particularly in the real-time assay, in which the high r values characterizing the IN⅐DNA complex cannot be severely influenced by strand transfer.  Fig. 2). All the ODNs tested were fluorescein-labeled on the 3Ј-extremity of strand a. HIV-a3F (black circles) corresponds to the wt sequence, whereas HIV-GTGT-a3F (white triangles) is a variant in which the CAGT 3Ј-terminal sequence was replaced by GTGT. Three nonspecific sequences (NS) were also studied (see "Experimental Procedures") and differed in the last four bases at the 3Ј-end, which are explicitly mentioned in their names: NS-TTCC-a3F (black squares), NS-CACC-a3F (white circles), NS-CAGT-a3F (white squares). The experimental conditions were similar to those described in the legend for Fig. 2. B, inhibition of 3Ј-processing activity by FZ41 compound, as monitored by fluorescence anisotropy. Activity calculations were based on measurements of ⌬r SDS .
Quantification Analysis and Characterization of the 3Ј-Processing Reaction-The 3Ј-processing activity of IN can be calculated by real-time assay or by quenching of the reaction with SDS. Both these approaches allow quantification of the fraction of released dinucleotides. Fig. 4A displays one example of the time course of GT release in real-time conditions. Activity increased linearly with time over ϳ180 min. This apparent linearity is compatible with single-turnover conditions (excess of IN over DNA substrate), when the rate constant of cleavage is low (25). Under single-turnover conditions, the time course of product formation provides a measure of the actual chemical cleavage step, which is not affected by subsequent turnovers. The kinetic profile, as determined by anisotropy measurement, gave k chemistry ϭ 0.0035 min Ϫ1 , consistent with the value obtained with the standard gel-electrophoresis method (k chemistry ϭ 0.004 min Ϫ1 ) (25). A similar result was obtained using the anisotropy method with SDS (data not shown). Indeed, Activity real-time and Activity SDS were well correlated (Fig. 4B). These results demonstrated that (i) all the GT dinucleotides produced by the cleavage reaction were released from the IN⅐DNA complexes in the real-time assay and were therefore not trapped within the complexes, (ii) the decrease in r in real-time conditions was due exclusively to the release of dinucleotides from the IN⅐DNA complexes, rather than another phenomenon, such as dissociation of the enzyme from the DNA substrate or product. The single-turnover rate constant, as determined with the anisotropy-based test, was similar to that determined by gel electrophoresis, suggesting that the presence of fluorescein at the DNA 3Ј-extremity has no effect on DNA-binding or the subsequent catalytic step. Comparison of the 3Ј-processing activity of non-fluorescent and fluorescein-labeled U5-substrates by gel electrophoresis confirmed that fluorescein has no influence on activity (Fig. 4C). This result sharply contrasts with that obtained using a TAMRA-labeled ODN: this ODN was also tested by both anisotropy and gelelectrophoresis methods and, although TAMRA did not influence the number of IN⅐DNA complexes, 3Ј-processing activity was significantly decreased by a factor of 5 (data not shown). Note that the r parameter is not sensitive to the precise positioning of IN on the DNA, and this positioning may be disturbed by the TAMRA probe in the absence of a strong effect on overall DNA-binding isotherms.
The Initial Steady-state Anisotropy Value Is Not Predictive of 3Ј-Processing Activity-The final r value obtained after the DNA-binding step corresponds to the initial r value for activity (r t ϭ 0 ) and is related, in first approximation, to the fractional  Fig. 2B). This activity was named "Activity real-time " in contrast to "Activity SDS ," which was calculated from ⌬r SDS when all the native IN⅐DNA complexes were disrupted by SDS after various incubation times. B, correlation between Activity real-time and Activity SDS values. The various activity values were obtained by varying either the incubation time or the IN:DNA ratio. We used r as measured under real-time conditions or after adding SDS to calculate the fraction of released dinucleotides, giving Activity real-time or Activity SDS , respectively. The linear fit gives a slope of 0.97 (R 2 ϭ 0.96). C, 3Ј-processing activity, as detected by radioactivity on a denaturing acrylamide/urea gel. Left, non-fluorescent DNA substrate. Right, fluorescein-labeled DNA substrate (HIV-a3F). The upper and lower bands correspond to the 21-mer DNA substrate and the 19-mer processed DNA product, respectively. The slower migration of HIV-a3F is due to the presence of fluorescein at the 3Ј-extremity. saturation function (Y). Under single-turnover conditions, activity is expected to increase with Y, and the fraction of released dinucleotides should therefore be proportional to the number of IN⅐DNA complexes initially formed in the sample. Interestingly, a plot of 3Ј-processing activity against r t ϭ 0 gave a clear bell-shaped curve (Fig. 5A). The various r t ϭ 0 values were obtained by varying the initial IN:DNA ratio from 4.2 to 400, corresponding to different mixtures of IN (from 50 to 800 nM) and ODN (from 2 to 12 nM). The optimal ratio was found to be ϳ40 (Fig. 5B) and corresponds to an r t ϭ 0 value of 0.226. Larger values (up to 0.272) led to a significantly decrease in the 3Ј-processing activity. Increases in r t ϭ 0 may occur for two reasons: (i) the number of free DNA molecules decreases with increasing IN:DNA ratio and (ii) self-assembly of IN on DNA may further increase the size of the complexes. The first phase (r t ϭ 0 Ͻ 0.226) most likely corresponds to an increase in the number of complexes, increasing both r t ϭ 0 and product formation. In the second phase (r t ϭ 0 Ͼ 0.226), the activity drops off suggesting that high order multimeric forms of IN on DNA, possibly aggregates, formed under conditions of high IN:DNA ratio and characterized by high r values, are less active than lower order multimeric forms. Furthermore, the r max value cor-responding to saturation ([DNA] free Ϸ 0) must be below 0.272, and, thus, 0.226 is the best estimate for r max .
To get deeper insight into the bell-shaped phenomenon, we investigated the relationship between the fraction of product obtained at the end of the reaction and the Y parameter, which is related to the number of DNA-substrate molecules initially bound to IN (at t ϭ 0). The initial number of complexes was varied, and activity at t ∞ (corresponding to the maximum amount of released product) was determined and plotted as a function of Y (Fig. 6, A-C). The fraction of product was found to increase linearly with the initial number of IN⅐DNA complexes, as expected under single-turnover conditions. Optimal conditions, corresponding to an IN:DNA ratio of ϳ40 (characterized by r ϭ 0.226), gave a final activity of 92%. This result confirms that complete saturation of the DNA substrate can be achieved at anisotropy values much lower than 0.272, and is compatible with r max ϭ 0.226. The activity per complex, specific activity (Â ϭ activity/Y), was then calculated for each IN:DNA ratio (Fig. 6D). Specific activity decreased as a function of IN:DNA ratio (Fig. 6D), suggesting that aggregation occurs as soon as IN is present in excess over DNA and that the activity of large complexes (IN agg ⅐DNA) is lower than that of the lower order multimeric form (IN act ⅐DNA). At ratios exceeding 200:1, the samples contained only IN agg ⅐DNA complexes (r ϭ 0.272; see also TFA experiments below). However, these complexes retained significant levels of 3Ј-processing activity (Ϸ35% the activity of IN act ⅐DNA complexes). The number of aggregated forms of IN increased continuously with the IN:DNA ratio, suggesting that, at the peak of the bell-shaped curve, the sample already contained some IN agg ⅐DNA complexes, together with the IN act ⅐DNA complexes. From Fig. 6, we can estimate that IN agg ⅐DNA complexes in these conditions accounted for ϳ11-15% of total complexes. We made use of the additivity law of anisotropy to calculate the theoretical r max value (0.218) characterizing the fully active complex (IN act ⅐DNA), in the absence of both free DNA and aggregated complexes. Thus, the bellshaped curve shown in Fig. 5A Fig. 5A, independently by TFA. This approach allows the determination of rotational correlation times (), which are directly related to hydrodynamic volume according to Equation 8. We first characterized ss and ds ODNs of various lengths (10-to 45-mer) in the absence of IN (Table 1). All ODNs were characterized by short values ( 1 and 2 ), which can reasonably be assigned to rotation of the fluorescein moiety around the linker at the 5Ј-end of the DNA and the flexibility of the linker itself. Unlike 1 and 2 , longer values ( 3 and 4 ) strongly depend on length and thus account for the overall rotational motion of the ODN. For each length, the longest correlation time was 2-2.7 times longer for ds than for ss ODNs. For example, the 4 values for   (Fig. 7). We then applied TFA to different samples corresponding to various IN:DNA ratios. We used HIV-a5F to prevent analytical problems due to the activity-dependent decrease in r when fluorescein is linked to the GT dinucleotide. The activity observed at t ∞ for an IN:DNA ratio of 40:1 (92%) suggested that only a small amount of free ODN was present in this sample. We therefore restricted our analysis to samples corresponding to the peak of the bellshaped curve and to the decreasing phase, so as to avoid studying mixtures of free and IN-bound ODNs. At 37°C, two short (Ͻ2.5 ns) were resolved upon the addition of IN to ODN ( Table  2). The distribution in the short domain was not dependent on the IN:DNA ratio. In contrast, the characteristic long 25°C of 8.6 ns found for free DNA was replaced by a 25°C Ն 37.8 ns (26.8 ns at 37°C), depending on the IN:DNA ratio ( Fig. 8 and Table 2). The sample corresponding to the peak of the bellshaped curve (IN:DNA 40:1; r ϭ 0.226) was rather monodisperse, as suggested by the small difference between the maximum (26.8 ns at 37°C) and the center of gravity (31.4 ns at 37°C) of the peak in the distribution. As discussed above, samples corresponding to the TFA distribution shown in   A and B) and was estimated by fitting: Act ϭ Act max ϫ (1 Ϫ e Ϫkobsϫt ). Y was calculated using r max ϭ 0.226. D, specific activity (Â ϭ activity/Y) as a function of the IN:DNA ratio. Total activity was normalized using fractional saturation function (Y), giving 3Ј-processing activity per complex.

ODN
Single-stranded a Double-stranded distribution toward longer values (Fig. 8, B-E). Our data confirmed that the sample corresponding to the highest r value (0.272) consisted essentially of aggregated forms of IN on DNA (Fig. 8E), as demonstrated by the broad distribution above 100 ns (17).

DISCUSSION
We used a continuous fluorescence assay to monitor the HIV-1 IN 3Ј-processing reaction. We found that the processed DNA product remained tightly bound to IN, thereby limiting enzyme turnover, whereas the other reaction product (GT dinucleotide) was rapidly released from the processed DNA⅐IN complex. This assay made it possible to characterize standard equilibrium/kinetic parameters and to obtain insight into the relationship between the size of IN⅐DNA complexes and catalytic activity. We studied 3Ј-processing under various IN:DNA ratio conditions: optimal activity was obtained at a ratio of 40:1. Higher ratios, associated with high r values, led to a decrease in product formation. The oligomeric status of active IN was further investigated by TFA. The best condition for 3Ј-processing activity corresponds to a majority of complexes in the sample characterized by a long correlation time of ϳ38 ns. This value is compatible with a dimeric form of IN bound to one short DNA substrate mimicking a single viral DNA end. The aggregation of IN on DNA was found to be responsible for the decrease in the 3Ј-processing activity.
Kinetic Analysis of 3Ј-Processing Activity by the Anisotropybased Method-The real-time assay described here is based on the use of a DNA duplex fluorescein-labeled on the terminal-3Ј GT. It allows quantification of the cleavage reaction resulting in physical separation of the processed 19/21-mer DNA (first product) and the fluorescently labeled dinucleotide (second product). IN assays based on fluorescence intensity (fluorescence resonance energy transfer or fluorescence quenching) have been described in previous studies (32)(33)(34)(35). Unlike electrophoresis-based methods, which can be used to analyze discrete time points only, fluorescence methods present the significant advantage of allowing continuous monitoring of the reaction for kinetic studies of IN in the same sample. However, only the anisotropy-based method can be used to study both the DNA-binding step and subsequent activity in the same assay. Using various DNA sequences, we showed that IN does not demonstrate highly selective binding to the viral sequence over other sequences, but, in contrast, displayed strong selectivity for catalysis (Fig. 3). This result is not subject to possible bias due to indirect comparison, as might be the case when DNA binding and catalysis are studied by different methods. It suggests that IN first binds viral DNA in a nonspecific manner and that subsequent relaxation leads to the formation of a specific and catalytically competent complex before catalysis. This relaxation step must be very slow to account for the low singleturnover rate constant as previously reported (25).
The anisotropy-based assay is reliable, because r decreased only in the presence of Mg 2ϩ with the wt DNA sequence and only when fluorescein was attached to the GT 3Ј-extremity and was prevented by a competitive inhibitor of IN. The additivity law of anisotropy renders the quantification of activity easy and both real-time and fixed-time studies can be carried out. The real-time assay is possible, because IN⅐DNA complexes with DNA substrate (21/21-mer) or product (processed 19/21-mer) are highly stable in solution (Fig. 2), and the observed decrease in r is therefore due solely to the cleavage reaction and not to dissociation of the protein from DNA. Unlike many standard methods, which quantify the 19/21-mer product under denaturing conditions, the real-time assay allows quantification of the GT product under native conditions. Our kinetic analysis gave a catalytic rate of 0.2 h Ϫ1 , which is similar to the values published in previous studies (25,36,37). The strong correlation between real-time and fixed-time data (Fig. 4B) and similarities between the time courses of GT and 19/21-mer formation demonstrate that the GT dinucleotide is rapidly released from IN after the 3Ј-processing reaction. This release is consistent with the low apparent affinity expected for such a small ligand (see Fig. 7, K d Ͼ 100 M). This situation is very different to that for the other reaction product (19/21-mer), which remains tightly bound to IN after catalysis. Finally, kinetic analysis showed that the catalytic rate constant was not significantly affected by the presence of fluorescein at the end of the DNA molecule, whereas inhibition was observed with TAMRA. TFA  showed that the local motion of fluorescein or TAMRA was more restricted when these fluorophores were covalently linked to a GT 3Ј-extremity, suggesting a direct interaction between the fluorophore and DNA. In the case of fluorescein, this interaction perturbed neither DNA binding nor cleavage. In contrast, the additional methyl groups and (or) charge effects of TAMRA may prevent the correct positioning of IN on the cleavage site or phosphodiester bond hydrolysis itself. Dimeric Forms of IN Are Optimal for the 3Ј-Processing of One DNA Extremity-We found that the optimal conditions for 3Ј-processing were obtained for a majority of dimeric forms bound to the DNA ( long ϭ 38 ns), consistent with complementation studies between individually inactive IN mutants, suggesting that at least the dimeric form is catalytically active for 3Ј-processing (38,39). The optimal IN:DNA ratio (40:1) reported here corresponds to conditions used in most IN assays: IN in excess over DNA and DNA concentration maintained below the K d value. This ratio therefore does not indicate the final stoichiometry and is highly dependent on experimen-tal conditions. For example, increasing ionic strength shifts the K d toward higher values and influences the optimal ratio (data not shown). Under experimental conditions in which DNA and IN concentrations were maintained above the K d value, an optimal ratio of 2:1 was found, consistent with the stoichiometric binding of two IN protomers per DNA end (25). We and others (12,40) have previously observed cooperative binding of IN to DNA. Consequently, using DNA concentration below the K d value, the IN:DNA ratio determines the number of complexes formed in solution as well as the stoichiometry of these complexes. Therefore, below the optimal ratio for activity, it is very likely that, in addition to free DNA, the solution contains a mixture of monomeric and dimeric forms of IN bound to the DNA substrate. Indeed, we have previously found that, using suboptimal IN:DNA ratios (10:1), the complexes exhibit lower values consistent with a monomer-dimer equilibrium at 37°C (normalized long , 24 -27 ns) and monomeric IN at 25°C (normalized long , 16 -20 ns) (16). Here, it is highly suggested that the optimal activity originates in two critical events: DNA binding and cooperativity. From this point of view, steady-state anisotropy is then particularly helpful to establish direct relationships between DNA binding parameters, the specific activity of the complexes, and their hydrodynamic properties as measured by TFA.
Our conclusion that the dimeric form of IN is active for 3Ј-processing under native conditions strongly parallels those obtained by Parissi and coworkers (41) with purified cross-linked multimers. The aggregation of IN on DNA was found to be strongly dependent on the IN:DNA ratio and was detrimental for 3Ј-processing, as previously reported for concerted integration (42). The bell-shaped curve obtained in Fig. 5 most likely originates from competition between the cooperative binding of IN on DNA leading to a dimeric form (specific IN-IN interaction, which is favored by increasing the IN:DNA ratio and enhances 3Ј-processing activity) and aggregation (nonspecific interactions also favored by increasing the IN:DNA ratio but detrimental for 3Ј-processing). The bell-shaped behavior is also compatible with the results of a recent cross-linking study (41) suggesting that monomers and tetramers display neither 3Ј-processing or half-strand transfer (transfer of a single viral DNA end)  Table 2.
activity, whereas dimers are competent for both these reactions, and only tetramers catalyze full-site concerted integration. Our results suggest that, in vitro, IN aggregates do not much hinder 3Ј-processing under IN:DNA ratio conditions leading to the formation of dimers, but that such aggregates may impede tetrameric organization in a highly competitive manner. However, TFA experiments did not indicate whether the tetrameric form on one LTR end actually existed, because the intermediate distributions (Fig. 8, B-D) probably corresponded to mixtures of different multimeric forms (polydisperse solutions). It is therefore difficult to assess 3Ј-processing activity of tetramers, because IN:DNA ratios shifting dimer-tetramer equilibrium toward tetramers also gave rise to IN aggregation. Although recombinant IN alone is sufficient for concerted integration in vitro (41)(42)(43)(44), this reaction remains much less efficient with recombinant IN than with the purified pre-integration complex (45) or virion-derived IN (46). The main limiting factor may be the multimeric organization of IN to yield the correct tetrameric form on DNA compatible with full integration process: aggregation is probably more limiting for DNA integration, which requires tetramers, than for 3Ј-processing, which requires dimers. Moreover, samples containing many aggregated forms displayed significant 3Ј-processing activity (one-third that of dimeric IN). The reason for this residual activity remains unclear.
Li and Craigie (44) recently showed that blunt-ended DNAs are more efficient substrates for concerted integration than preprocessed DNAs, whereas half-integration is more efficient with preprocessed DNA. These data, together with our results and those of Parissi et al. (41), suggest that dimers of IN are necessary and sufficient for 3Ј-processing and half-integration and that the 3Ј-processing step probably promotes a conformational change of IN compatible with concerted integration. This conformational change may play a key role in promoting the IN tetramerization required for concerted integration. IN is known to exist in a monomerdimer-tetramer equilibrium in solution (16,17,24,38). The multimeric form is modified on DNA, and the final quaternary structure depends on the IN:DNA ratio and probably also on the nature of the DNA substrate. From this point of view, the tetrameric organization of IN (a dimer of dimers) may be stabilized by the simultaneous presence of two LTR ends (47) and it highly suggested that the condensation-distortion properties of DNA play a key role in bringing the two ends of the viral DNA close enough in space for a dimer of dimers to form. A dimer at each LTR end, as described here, is consistent with current models based on cross-linking (48) or molecular dynamics data (49), and the 5-bp separation at target level is compatible with a tetrameric complex rather than a dimer. Several models of the tetrameric active complex exist for concerted integration: the 5-bp separation is directly compatible with a tetrameric complex displaying 2-fold symmetry, as modeled by McCammon and coworkers (31) or a tetrameric complex assuming target DNA bending (48,50) or a hinging motion of the dimer interface (7). The formation of a dimer of dimers is probably mediated by the C-terminal domain, as suggested by the x-ray structure of the catalytic core/C-terminal double domain (9). Moreover, residue Trp-235 in the C-terminal domain has a significant effect on concerted integration but not on half-integration (44). The resulting symmetric organization, dispensable for 3Ј-processing but not for integration, may be favored or stabilized by the symmetric base sequence of the target DNA. Such symmetry in integration sites has recently been reported (51,52).
It is noteworthy that the single-turnover rate constant of IN is particularly slow, for example about one thousandth that for restriction enzyme HincII (53). This indicates that a pre-catalysis step has a limiting effect independent from that for product release. We found that dimers were the major oligomeric forms of IN on one LTR end. IN free in solution was mostly tetrameric in the 100 -200 nM concentration range and aggregated at concentrations above 250 nM (24). This suggests that the dissociation of high order multimeric forms precedes or occurs simultaneously to DNA binding, as previously reported (16) (Fig. 9). Parissi and coworkers (41) carried out a kinetic analysis of the IN⅐DNA complexes formation and found that the presence of one LTR induces the binding of one IN monomer (M) prior to formation of the IN dimer (D). The conversion of high order multimeric forms of IN into monomers may account for the rather slow DNAbinding step (12,25,27). We have previously shown that IN binds cooperatively to DNA with a Hill number compatible FIGURE 9. Model of the 3-processing reaction according to binding/multimerization and catalytic parameters. Monomers (M) preferentially bind to DNA in a cooperative manner and form a dimer (D) on the LTR at one end of the DNA (see text). This model involves the conversion of large multimers (IN n Ն 4 ) into smaller species prior to or simultaneously with DNA binding. Once the IN 2 ⅐DNA complex has been formed, a slow D N D* relaxation step, generating a catalytically competent complex, accounts for the very low single-turnover rate constant (k chemistry ). AUGUST 11, 2006 • VOLUME 281 • NUMBER 32 with the binding of monomers and the presence of a dimeric form of IN on DNA (12). We suggest that relaxation of the conformation of the IN⅐DNA complex leading to the formation of a competent and specifically bound dimeric form (D*) is probably the limiting step accounting for the slow single turnover of 3Ј-processing. High IN concentrations lead to aggregation on DNA, as shown here by TFA and in other studies by means of a fluorescence correlation spectroscopy approach (54). However, we found that maximum 3Ј-processing activity was not controlled by absolute IN concentration but instead depended on the IN:DNA ratio.