Mechanism of Inhibition of HIV-1 Integrase by G-tetrad-forming Oligonucleotides in Vitro *

The G-tetrad-forming oligonucleotides T30177 and T30695 have been identified as potent inhibitors of human immunodeficiency virus type 1 integrase (HIV-1 IN) activity J. Biol. Chem. 273, 34992–34999). To understand the inhibition of HIV-1 IN activity by the G-quartet inhibitors, we have designed the oligonucleotides T40215 and T40216, composed of three and four G-quartets with stem lengths of 19 and 24 Å, respectively. The fact that increasing the G-quartet stem length from 15 to 24 Å kept inhibition of HIV-1 IN activity unchanged suggests that the binding interaction occurs between a GTGT loop domain of

Combination therapy for AIDS, which uses two or more drugs simultaneously to inhibit human immunodeficiency virus (HIV) 1 replication, was developed to lower toxicity by decreasing the dosage of individual compounds. This approach reduces the risk of developing drug resistance and maintains synergistic antiviral activity (1). Although combination therapy can depress the HIV virus to undetectable levels in the blood of many HIV-positive patients, the HIV viruses within T cells are still fully capable of replicating and infecting other cells (2)(3)(4). The development of new agents against human immunodeficiency virus type 1 integrase (HIV-1 IN) (5), which may eliminate HIV-1 from intracellular sites, would be a major advance in the treatment of HIV infection.
A new class of oligonucleotides with only G and T residues in their sequences has been discovered to inhibit HIV-1 IN (6,7). The analyses have been performed on the oligonucleotide 5Ј-G*TGGTGGGTGGGTGGG*T, synthesized with a phosphorothioate linkage at the two end G residues (G*) and referred to as T30177. This G-rich oligonucleotide is capable of forming a stable intramolecular G-quartet fold (6). IC 50 (50% inhibitory concentration) values of T30177 for HIV-1 IN are in the nanomolar range in vitro (7).
A 16-residue oligonucleotide (5Ј-G*GGTGGGTGGGTGGG-*T) referred to as T30695 has been designed and synthesized to improve both the structural stability and inhibition of HIV-1 IN activity (8). Compared with T30177, T30695 forms an even more stable and orderly G-quartet fold. Our NMR and kinetic data demonstrated that in response to K ϩ binding, T30695 folds into a stable and symmetric G-tetrad complex (8 -10). The folding is a two-step process, dependent on the nature of the alkaline metal ion. The first step of the process involves the coordination of one K ϩ ion, which competes with a Li ϩ ion to bind within the core of two G-quartets. The second step involves the binding of two additional K ϩ ions to the loop domains. NMR results have shown that in the absence of K ϩ , T30695 forms an intramolecular fold with a pair of distorted G-quartets flanked by extended, unfolded loop domains (called the Li ϩ form). When coordinated with 3 K ϩ eq, T30695 folds into a more compact and symmetric structure with an ϳ15-Å width and a 15-Å length (called the K ϩ form). This structure resembles a cylinder with positive charges inside and negative charges on the outer surface (10). The IC 50 of T30695 for inhibition of HIV-1 IN without K ϩ (Li ϩ form structure) is 530 nM. In the presence of K ϩ , the IC 50 of T30695 (K ϩ form structure) decreases from 530 to 31 nM, corresponding to a 20-fold increase in the inhibition of HIV-1 IN activity. Kinetic data demonstrated that this increase in inhibition of HIV-1 IN activity for T30695 is correlated with the folding of its loops into a stable, compact structure by the binding of 2 additional K ϩ eq. The loop structures of T30695 appear to play a key role in the inhibition of HIV-1 IN activity.
To investigate the binding interaction between the catalytic site of HIV-1 IN and the G-quartet inhibitors, we have designed G-tetrad-forming oligonucleotides with two, three, and four G-quartets. The lengths of these G-quartet stems are 15, 19, and 24 Å, respectively. The results demonstrated that these G-tetrad-forming oligonucleotides with different stem lengths have the same ability to inhibit HIV-1 IN activity in vitro.
Based upon the experimental evidence and modeling study, we provide some critical information regarding the interaction and structure-activity between HIV-1 IN and the G-quartet inhibitors, which could be of benefit in the design of novel anti-HIV therapeutic drugs.

MATERIALS AND METHODS
Oligonucleotide Synthesis-All of the G-rich oligonucleotides were obtained from Midland Certified Reagent Co. (Midland, TX). These oligonucleotides were synthesized using cyanothyl phosphoramidite chemistry. After removal of the protecting groups by hydrolysis with concentrated ammonium hydroxide, the products were purified by anion-exchange high pressure liquid chromatography. The qualities of materials were then measured by mass spectroscopy. We used these oligonucleotides without further treatments.
Melting and Annealing Measurements-Oligonucleotides at 5 M strand equivalents (20 mM Li 3 PO 4 , pH 7) were heated to 90°C for 5 min and then cooled at 4°C for 30 min in the presence of KCl at a designated concentration. Subsequent to the incubation step, thermal denaturation profiles of the oligonucleotides were obtained over a range from 20 to 90°C for melting and from 90 to 20°C for annealing. Absorbance was monitored at 240 nm by a Hewlett-Packard 8452A diode array spectrophotometer using a Hewlett-Packard 89090A temperature regulator.
The thermal denaturation curves of the oligonucleotides were analyzed with an intramolecular folding equilibrium (8,11): where K eq is the constant for the random coils to folded oligonucleotide equilibrium, ␣ is the fraction of folded strands, 1 Ϫ ␣ is the fraction of random coils, A(T) is the absorbance at temperature T, A rc is the absorbance when all strands are random coils, A st is the absorbance when all strands are folded, and is the cooperativity of the melting transition (referred to as the helix interruption constant) defined by ϭ exp(⌬S i /R), where ⌬S i is in units/mol of interruption. In our analysis study, the values of are in the range of 0.9 -0.999, determined by an optimized fitting program. Values for T m and ⌬G were obtained on the basis of the fitting procedure, which inputs the values of ⌬H 0 , ⌬S 0 , A rc , and A st , estimated from the experimental measurements, and then uses an optimized fitting program to search for the best fit.
Gel Electrophoresis-The G-rich oligonucleotides, T30695 and its derivatives, plus 10 mM KCl in 20 mM Li 3 PO 4 , pH 7, were labeled with 32 P using a 5Ј-end labeling procedure and purified using Microspin G-25 columns. The oligonucleotide solution was heated at 90°C for 5 min and then cooled at 4°C for 30 min. 20% nondenaturing polyacrylamide gels containing 1ϫ Tris/borate acid/EDTA buffer, 10% ammonium persulfate, and 30 l of TEMED in 1ϫ TBA buffer were precooled in a 4°C cold room for 1 h. The prepared samples were run on 20% nondenaturing polyacrylamide gels in a 4°C cold room for 6 h. Gels were stained in a 0.01% Stains-All/formamide solution.
Circular Dichroism-CD spectra of the G-quartet oligonucleotides were obtained in 15 M strand concentration plus 10 mM KCl in 20 mM Li 3 PO 4 , at pH 7, on a Jasco J-500A spectropolarimeter at 24°C. Each spectrum represents five average scans. Data are presented in molar ellipticity (degrees⅐cm 2 ⅐dmol Ϫ1 ).
HIV-1 IN Protein-The pET-15b-IN-(1-288)/F185K/C280S plasmid was expressed in Escherichia coli strain BL21 as described previously (12) with the following modifications. Cells were grown in 1000 ml of LB medium (Digene, Beltsville, MD) containing 5 g/ml ampicillin until an optical density of 0.8 was reached at 600 nm. Protein expression was induced for 3 h with the addition of 0.4 mM isopropyl-␤-D-thiogalactopyranoside. Cells were harvested and resuspended in lysis buffer containing 25 mM HEPES, pH 7.5, 1 M NaCl, 5 mM imidazole, 2 mM ␤-mercaptoethanol, and 0.3 mg/ml lysozyme. After 30 min on ice and sonication, lysed cells were centrifuged for 20 min at 30,000 ϫ g, and the supernatant was applied to a nickel-Sepharose column. Integrase retained on the column was washed with buffer containing 25 mM HEPES, pH 7.5, 0.5 M NaCl, 2 mM ␤-mercaptoethanol, and an increasing imidazole concentration from 20 to 250 mM. The protein was then eluted with the same buffer containing 750 mM imidazole and dialyzed overnight against 25 mM HEPES, pH 7.5, 1 M NaCl, 2 mM EDTA, 10 mM dithiothreitol, 2 mM ␤-mercaptoethanol, 100 mM imidazole, and 10% glycerol.
Anti-HIV-1 IN Activity Assay-The HIV-1 IN assays were performed as described previously (7) with the following modifications. In the 3Ј-processing and strand transfer assay, HIV-1 IN was preincubated at a final concentration of 400 nM with inhibitors for 15 min at 30°C in a reaction buffer containing 25 mM MOPS, pH 7.2, 25 mM NaCl, 7.5 mM MnCl 2 , 0.1 mg/ml bovine serum albumin, and 14.3 mM ␤-mercaptoethanol. Then, a 5 nM concentration of the 32 P-5Ј-end-labeled oligonucleotide was added to a final volume of 10 l, and incubation was continued for an additional 30 min. Reactions were quenched by addition of 5 l of denaturing loading dye. Samples were loaded on a 20% (19:1) denaturing polyacrylamide gel.
Modeling Computation-The molecular structures of T40215 and T40216 were built up and optimized under AMBER force field by INSIGHT II/DISCOVER. The optimization of the molecular structures followed this procedure: 1) 100 steps of conjugate gradient energy minimization; 2) 1000 steps of restrained molecular dynamics equilibration with a time step of 0.33 fs at 1000 K; 3) 1000 steps of restrained molecular dynamics equilibration with a time step of 0.1 fs at 300 K; and 4) 1000 steps of conjugate gradient energy minimization. The intramolecular hydrogen bonds of G-quartets were used as constrains for the molecular optimization.
The molecular structure of the core domain of HIV-1 IN, HIV-1 IN-(51-209) (14), was obtained from the Protein Data Bank. Next, we docked the NMR structure of T30695 (10) into this x-ray structure using the GRAMM docking program with a high resolution matching mode (15)(16)(17)(18). This program is based on a geometry-based algorithm for predicting the structure of a possible complex between molecules of known structures. It can provide quantitative data related to the quality of the contact between the molecules. The intermolecular energy calculation relies on the well established correlation and Fourier transformation techniques used in the field of pattern recognition. The docking calculation by GRAMM predicts the structure of the complex formed between the two constituent molecules by using their atomic coordination, without any prior information as to their binding sites. To eliminate the terminal effects of the structure of HIV-1 IN-(51-209) in the docking calculation, the docking range of the calculation was set up from amino acids 60 to 160. 1000 different structures of an HIV-1 IN⅐T30695 complex were created by the calculations. The information on the binding interaction and hydrogen bond formation of each docking structure was analyzed. Based upon the hydrogen bond formation, the statistical possibilities of the residues of T30695 interacting with HIV-1 IN and of the amino acid residues of HIV-1 IN binding T30695 were plotted.

RESULTS
Structures of T30923, T40215, and T40216 -T30695 has been determined previously as an intramolecular G-quartet structure with two G-quartets in the center and two folded loop domains on the top and bottom (10). Upon coordination with 3 K ϩ ion eq, the structure of T30695 becomes symmetric and compact with a 15-Å width and a 15-Å length. The distance between the two central G-quartet planes is ϳ3.9 Å. Recently, T30695 and the thrombin-binding aptamer, which also forms an intramolecular G-quartet structure (19,20), were further studied by nondenaturing gel electrophoresis (11). Similar migration further confirmed that T30695 forms an intramolecular G-quartet structure with the same structural size as the thrombin-binding aptamer. T30923 (5Ј-GGGTGGGTGGGTGGGT) has the same sequence as T30695 (Fig. 1). However, the two terminal phosphorothioate linkages of T30695 in the G 1 and G 15 positions are replaced by two phosphodiester linkages. We previously showed that T30923 has the same structure and physical properties as T30695 (11).
To investigate the interactions between HIV-1 IN and the G-quartet oligonucleotides, T40215 (5Ј-(GGGGT) 4 ) and T40216 (5Ј-(GGGGGT) 4 ) were designed to form an intramolecular Gquartet structure, composed of three and four G-quartets with stem lengths of ϳ19 and 24 Å, respectively (Fig. 1). CD spectra were employed to demonstrate that T30923, T40215, and T40216 form the same molecular structure as T30695. The CD spectra of T30695, T30923, T40215, and T40216 are characterized by nonconservative spectra, with maxima at 264 and 210 nm and minima at 240 nm (Fig. 2), which demonstrated that all four oligonucleotides form the same G-quartet structure. Compared with the CD spectrum of T30695, it appears that a longer G-quartet stem corresponds to weaker CD ellipticity at 264 and 240 nm.
Further evidence to support intramolecular G-quartet formation for T30923, T40215, and T40216 was obtained from nondenaturing gel electrophoresis (Fig. 3A) and from melting and annealing measurements ( Fig. 3B and Table I). Fig. 3A shows that the bands of T30923, T40215, and T40216 have the same migration compared with that of T30695. T30695 was used as a control since its molecular structure has been determined by NMR to form an intramolecular G-quartet structure in the presence of K ϩ (10). The migration rate on nondenaturing gels depends on the molecular structure of the G-quartet oligonucleotides. The same migration of all four G-quartet oligonucleotides demonstrates that they form the same molecular structure with a similar size. Therefore, T30923, T40215, and T40216 also form an intramolecular G-quartet structure in the presence of 10 mM KCl. However, electrophoresis cannot resolve the small difference in the length of these G-quartets from 15 to 24 Å.
A melting curve corresponds to a disordered state, called hyperchromicity. The annealing curve measures the base pairing and stacking of the secondary structure, called hypochromicity. The two curves (heating and cooling) of the same oligonucleotides are expected to be identical if the oligonucleotide forms an intramolecular, self-associated G-quartet. Compared with the melting T mH (where "H" is heating) values in Table I, the annealing T mC (where "C" is cooling) values of all three oligonucleotides at 5 M strand concentration have a slight shift to a lower temperature. The ⌬T values of T30923, T40215, and T40216 are 1.4, 2.4, and 3.4°C, respectively, showing that the molecules with longer G-quartet stems have greater ⌬T values. Interestingly, the ⌬T values of T40215 at strand concentrations of 2 and 20 M are 0 and 4.8°C, respectively, which suggests that ⌬T may also depend on oligonucleotide concen-tration and that higher concentrations induce greater ⌬T. At 2 M, there was no measurable temperature shift (⌬T ϭ 0) between the melting and annealing curves of T40215, which provides evidence for an intramolecular G-quartet formation. The slight temperature shifts observed at 5 and 20 M were considered to be induced by the disruption of structure refolding caused by higher molecular aggregation. The same phenomenon also applied when the G-quartet stem length was increased.
The constant T mH values of T40215 from 2 to 20 M (72.7°C) also confirm that this oligonucleotide form an intramolecular G-quartet structure. T m depends on the oligonucleotide concentration when a G-quartet structure is formed by dimeric or tetrameric oligonucleotides. Higher oligonucleotide concentrations increase the possibility of forming dimeric or tetrameric G-quartet structures, corresponding to higher T m . Only the T m of an intramolecular G-quartet structure is independent of the concentration of oligonucleotides. The conclusion obtained from these results is consistent with that from Fig. 3A.
The ion binding stoichiometry of the G-quartet structures has been established previously (8). First, the thermal denaturation curves of T30923, T40215, and T40216 in 0.1, 0.5, 1.0, and 5.0 mM KCl were obtained from UV absorbance at 240 nm  and then analyzed using a single-strand denaturation equilibrium (see "Materials and Methods" for details). The fitting procedure is to input the constants ⌬H 0 , ⌬S 0 , A rc and A st , which were estimated from the experimental data, and then to use an optimizing program to search for the best fit. As shown in Table  II, the fitting coefficient for each melting curve is ϳ0.99 or higher. The slopes of the lines for T30923, T40215, and T40216 were 3.8, 5.1, and 6.5, respectively, obtained by fitting the data points of the calculated ⌬G 0 versus log[K ϩ ] (Fig. 4). Based upon ⌬n ϭ ⌬⌬G 0 /2.3RT⌬log 10 [K ϩ ] and k slope ϭ ⌬⌬G 0 /⌬log 10 [K ϩ ], the ⌬n K ϩ ion eq released from unfolding was calculated as 2.8, 3.8, and 4.8 for T30923, T40215, and T40216, respectively. NMR data previously confirmed that T30695 coordinates three K ϩ ions. One is bound between the two G-quartets, and two additional K ϩ ions are bonded between a G-quartet and a neighboring TGTG loop domain: one at the top and the other at the bottom (9, 10). The ⌬n values obtained here indicate that T30923 also coordinates three K ϩ ions, whereas T40215 and T40216 coordinate four and five K ϩ ions between three and four G-quartets and two loop domains, respectively. These results confirm that T30923, T40215, and T40216 are composed of two, three, and four G-quartets, respectively, as shown in Fig. 5.

Inhibition of HIV-1 IN Activity by T30923, T40215
, and T40216 -The inhibition of HIV-1 IN by G-quartet oligonucleotides has been measured in a dual assay for 3Ј-processing and strand transfer activities (7). As shown in Fig. 6A, T30923, T40215, and T40216 were tested for the effect on both 3Јprocessing and strand transfer using a 21-mer duplex oligonucleotide. The 3Ј-processing reaction cleaves the 3Ј-terminal dinucleotide to a 19-mer oligonucleotide. The strand transfer reaction then results in an integration of two oligonucleotides together, which links the precleaved 3Ј-end of one 19-mer oligonucleotide to another 21-mer noncleaved or 19-mer precleaved oligonucleotide (Fig. 6A). The strand transfer products yield larger molecular species with slower migration compared with the 21-mer substrate. The inhibition of HIV-1 IN activity by T30923, T40215, and T40216 is shown in Fig. 6B. Compared with control lanes 2 and 21 (with only DNA plus integrase), the intensities of the 19-mer oligonucleotide band and of the strand transfer product bands decreased when the concentrations of T30923, T40215, and T40216 were increased. This result demonstrated that the G-quartet oligonucleotides blocked HIV-1 IN to yield the products of 3Ј-processing and strand transfer. IC 50 values for inhibition of 3Ј-processing and strand transfer for . These data were fitted by a straight line yielding a slope (k slope ϭ ⌬⌬G 0 /⌬log 10 [K ϩ ]) of 3.8, 5.1, and 6.5 for T30923, T40215, and T40216, respectively. According to the simple model of the transition between the folded and unfolded states for an intramolecular G-quartet (7), the values of the released K ϩ equivalent, ⌬n (ϭ⌬⌬G 0 / 2.3RT⌬log 10 [K ϩ ]), for T30923, T40215, and T40216 are ϳ2.8, 3.8, and 4.8, which correspond to the release of 3, 4, and 5 K ϩ ion eq. FIG. 5. Molecular structures of T30695, T40215, and T40216. The structure of T30695 was determined by NMR (10). The structures of T40215 and T40216 were obtained from molecular modeling (see "Materials and Methods" for details). T30923 has the same structure as T30695 (not shown). A, top view, which shows that the three structures have similar GTGT loop structures; B, side view, which shows different G-quartet stem lengths for the three oligonucleotides: T30695, T40215, and T40216 with two, three, and four G-quartets, respectively. T30923, T40215, and T40216 were obtained from plots of percentage inhibition versus drug concentration (Fig. 6C). IC 50 values for T30923, T40215, and T40216 were 70, 100, and 80 nM in 3Ј-processing and 85, 90, and 60 nM in strand transfer, respectively (Table III). The results of the assay of anti-HIV IN activity demonstrated that all three G-quartet oligonucleotides had comparable ability to inhibit HIV-1 IN. However, based upon the structure determination (see above), T30923, T40215, and T40216 are composed of two, three, and four G-quartets with stem lengths of ϳ15, 19, and 24 Å, respectively. The same range of IC 50 values for the three oligonucleotides indicated that the increase in length of the G-quartet stem did not appear to disrupt the binding interaction between the catalytic sites of HIV-1 IN and the G-quartet oligonucleotides. Based on this observation, we propose that the binding interaction occurs between a catalytic site of HIV-1 IN and a GTGT loop domain of the G-quartet oligonucleotides, referred to as a face-to-face interaction. Previous NMR data have demonstrated that a GTGT loop domain of the G-tetrad-forming oligonucleotides folds into a plane when a K ϩ ion is bound between the loop domain and a neighbor G-quartet (8,10). The folded loop domains largely increase the ability of the G-tetrad-forming oligonucleotides to inhibit HIV-1 IN activity because of an enhancement in face-to-face interaction. Therefore, an increase in the G-quartet The histograms were plotted based upon the number of hydrogen bonds formed between T30695 and HIV-1 IN from 1000 random docking structures in the docking range of amino acids 60 -160 (Fig. 8). Based upon the number of hydrogen bonds formed between HIV-1 IN and T30695 in the docking ranges, Fig. 8A shows the statistical distribution of the binding interaction in T30695 residues, and Fig. 8B shows the statistical distribution of the binding interaction in the catalytic domain of HIV-1 IN.
Interestingly, the histogram in Fig. 8A demonstrates that residues G 1 , T 8 , G 9 , G 10 , and T 16   Their functions in the enzyme are still unclear. We also found that the amino acid residues with a high binding possibility such as Glu 69 , Asp 116 , Tyr 143 , and Gln 148 are located in loops of the molecular structure of the catalytic core domain. DISCUSSION HIV-1 IN is composed of three functional and structural domains. Although all three domains are required for 3Ј-processing and strand transfer (21), the central core domain is directly involved in catalysis for the strand transfer reaction, as demonstrated by a disintegration assay (22). Crystal structures show that the catalytic core domain comprises a fivestranded ␤-sheet structure flanked by helical regions (23)(24)(25). The catalytic core contains the three active residues (Asp 64 , Asp 116 , and Glu 152 ) that form an active site required for enzyme activity and is referred to as the D,D-35-E motif. This motif can be found in all retroviral integrases (13,26,27). Recent experimental data and modeling demonstrate that the catalytic core domain residues Lys 156 , Lys 159 , and Lys 160 are also involved in the functional interaction of HIV-1 IN with DNA and mononucleotide inhibitors (28,29). The N-terminal domain of HIV-1 IN is composed of four helices with a zinc coordination site (12) and is also required for full integrase activity (30 -32). The C-terminal domain can bind DNA nonspecifically (33)(34)(35) and is composed of a five-stranded ␤-barrel with an SH3-like fold (36,37). However, the precise roles of the N-and C-terminal domains in the overall integration reaction are not well understood. No crystal structure has been identified for a full-length integrase thus far, although structures of all three HIV-1 IN domains have been determined individually (12, 19, 23-25, 36, 37, 40).
The details of the inhibition of HIV-1 IN activity by T30695 and its derivatives have not been structurally determined yet. However, several critical features for enzyme inhibition have been observed by several laboratories (6 -8, 10, 38). An intramolecular G-quartet structure is required for the inhibition  A and B). Based upon the x-ray structure of the core domain (14) and the NMR structure of T30695 (10), the complex structure was obtained from docking calculation using the GRAMM program. of HIV-1 IN activity. The loop residues of T30695 and its homologues are important for the inhibition of HIV-1 IN activity, as indicated by comparison of IC 50 values of T30695 with those of its derivatives with base substitution within the loops (7,8). The results from a gel-based assay using several different integrase mutants demonstrate that T30695 homologues require the enzyme zinc finger region (in the N-terminal domain of integrase) for inhibitory activity. The zinc finger is considered to assist in stabilizing the binding to the G-tetrad inhibitor (7). Also, the structure-activity results (11) demonstrate a functional relationship between thermal stability and anti-HIV activity for the G-tetrad-forming oligonucleotides, which suggests that the anti-HIV activity of the G-quartet inhibitors depends on their structural stability. The inhibitors with higher thermal stability have a stronger ability to inhibit the HIV-1 IN activity in vitro and HIV-1 in cell culture. Although the G-quartet inhibitor T30177 was proposed to inhibit HIV-1 adsorption in cell cultures (39), the G-quartets show a strong inhibition of HIV-1 replication in cell cultures (11). Understanding the mechanism of interactions between HIV-1 IN and G-quartet inhibitors in vitro can be of benefit in improving anti-HIV drug design.
This study on the extension of G-quartet stems demonstrated that an increase in length of the G-quartet stem from 15 to 24 Å did not affect HIV-1 IN inhibition. The binding interaction between a G-quartet inhibitor and an HIV-1 IN monomer was believed to occur between a loop domain of the G-quartet inhibitor and a catalytic site of HIV-1 IN, referred to as a face-to-face interaction. This conclusion was also supported by statistical results obtained from molecular calculation. 1000 molecular structures of HIV-1 IN-(51-209)⅐T30695 complexes were created from a random docking calculation. The statistical distributions of hydrogen bond formation obtained from the large number of docking calculations provide the critical information that the binding interaction between HIV-1 IN and T30695 mostly occurs when the loop residues of T30695 bind to the active site of the catalytic core domain of HIV-1 IN. The statistical distributions also suggest that the strong inhibition of HIV-1 IN activity by G-quartet inhibitors in vitro ( Fig. 6 and Table III) could correspond to the high possibility of the G-quartet inhibitors to bind to the active site of HIV-1 IN-(51-209). In the calculation, we identified some amino acid residues of the catalytic domain with a high possibility to interact with the G-quartet inhibitors, such as His 67 , Glu 69 , Tyr 143 , and Gln 148 , some of which could be critical factors for enzyme function.
The crystallized core domains of HIV-1 IN form a tight dimer (14, 23), HIV-1 IN was also proposed to function as a tetramer (13) or an octamer (25). Our molecular structure of an HIV-1 IN-(51-209)⅐T30695 complex was obtained from docking calculations. Compared with the molecular model of 5-N 3 -AZTMP binding to HIV-1 IN-(50 -212) (29), our model shows that the G-quartet inhibitor binds simultaneously to several key amino residues in the active site of HIV-1 IN, such as Asp 64 , Asp 116 , Glu 152 , and Lys 159 , covering a lager section of the enzyme catalytic site.
This report also provides an explanation for the previous observation (10) that a folded and unfolded GTGT loop domain of T30695 yielded a great difference in the ability to inhibit HIV-1 IN activity. IC 50 values for the folded and unfolded loop structures were 31 and 530 nM, respectively. We propose that this difference is because the unfolded loop structure largely disrupts the face-to-face interaction. A recent observation (11) showed that inhibition of HIV-1 IN activity by derivatives of T30695 with the substitution of 5-amino dU for T 5-methyl (CH 3 ) was decreased two to four times. The decrease was considered to be caused by the charge-charge interaction between the positively charged loop of the derivatives and the positive charge of the lysine residues in (or near) the binding site of HIV-1 IN, such as Lys 159 and Lys 160 . The statistical distribution also shows a high possibility for an interaction between Lys 159 of HIV-1 IN and G-quartets. The face-to-face interaction also provides further evidence to support the functional correlation between structural stability and anti-HIV-1 IN activity of G-quartet oligonucleotides observed in vitro (11). Varying the structure of the loop domains of G-quartet oligonucleotides not only induces a decrease in structure stability, but also disrupts the face-to-face interaction. Critical information for AIDS therapeutic drug design is that disruption of the face-toface interaction strongly influences the inhibition of HIV-1 IN activity. Therefore, searching for a loop structure with a better binding affinity will be one of the key steps for improving the efficiency of these compounds to inhibit HIV-1 IN activity.