An Inhibitory Monoclonal Antibody Binds at the Turn of the Helix-Turn-Helix Motif in the N-terminal Domain of HIV-1 Integrase*

With the increase in our understanding of its structure and enzymatic mechanism, HIV-1 integrase (IN) has become a promising target for designing drugs to treat patients with AIDS. To investigate the structure and function of IN, a panel of monoclonal antibodies (mAbs) directed against HIV-1 IN was raised and characterized previously in this laboratory. Among them, mAbs17, -4, and -33 were found to inhibit IN activity in vitro . In this study, we investigated the interaction of N-terminal-specific mAb17 and its isolated Fab fragment with full-length HIV-1 IN(1–288) and its isolated N-terminal, Zn 2 1 binding domain IN(1–49). Our results show that binding of Zn 2 1 to IN(1–49) stabilizes the mAb17-IN complex and that dimer dissociation is not required for binding of the Fab. To identify the epitope recognized by mAb17, we developed a protein footprinting technique based on controlled proteolysis and matrix-assisted laser desorption ionization time-of-flight mass spectrometry. Binding was mapped to a region within amino acids Asp 25 – Glu 35 . This peptide corresponds to the end of a helix-turn-helix motif in the IN(1–55) NMR structure and contributes to the dimerization of the N-terminal domain. Antibody binding also appears to destabilize the N-terminal helix in this domain. A molecular model of the [IN(1–49)] 2 z (Fab) 1 complex shows Fab binding across the dimer protein and suggests a potential target

is one of three viral-encoded enzymes that are necessary for retroviral replication (1). IN catalyzes integration of the viral DNA into the chromosomal DNA of a host cell. Extensive genetic and biochemical studies have identified three distinct regions in IN known as the N-terminal, catalytic core, and C-terminal domains (2,3). Although the structure of the full-length protein has not been solved, the three-dimensional structures of the three isolated domains have been determined (Fig. 1A). Crystallographic analysis of the catalytic core domains of human immunodeficiency virus type-1 (HIV-1) IN and avian sarcoma virus (ASV) IN revealed that each subunit binds one divalent cation cofactor, such as Mg 2ϩ or Mn 2ϩ , mediated by acidic residues in the highly conserved D, D(35)E motif that composes the active site of the enzyme (4 -7). Previous studies in this laboratory showed that binding Mg 2ϩ or Mn 2ϩ to full-length HIV-1 IN activates the protein by inducing a conformational change (8,9). Metal binding was also shown to stimulate preferential binding of this protein to its viral DNA substrate and to stabilize the resultant nucleoprotein complexes (10). The C-terminal domain, comprising amino acids 220 -270, binds DNA in a sequence-independent manner. Its structure, revealed by NMR spectroscopy (11)(12)(13), resembles that of Src homology region 3 (SH3), a motif known to promote protein-protein interactions. The crystal structures of two two-domain derivatives containing the core and C-terminal domain of ASV IN (14) and simian immunodeficiency virus IN (15) have been solved recently, and possible DNA binding paths that span both domain surfaces have been suggested in these models. The structures of isolated N-terminal domains of both HIV-1 and HIV-2 IN proteins, comprising amino acids 1 to approximately 50, have also been solved by NMR spectroscopy (16,17) and have been shown to include a three-helix bundle with a common helix-turn-helix (HTH) component. Two His and two Cys residues (the HHCC motif) in this domain coordinate a zinc ion, and Zn 2ϩ binding stabilizes the HTH structure (16,17). Previous studies also showed that binding of Zn 2ϩ to the HIV-1 IN N-terminal domain promotes formation of tetramer or higher multimers of the protein and stimulates enzymatic activity in vitro (18 -20). Thus, in contrast to transcription factors in which a similar constellation of Zn 2ϩ -binding residues stabilizes a DNA-binding structure, the Zn 2ϩ -binding motif in retroviral IN appears to facilitate protein-protein interactions.
IN proteins catalyze two distinct reactions, denoted processing and joining. Although the isolated core domain can catalyze an apparent reversal of the joining reaction, called disintegration or splicing (21,22), two-domain derivatives containing core and C-terminal domains (residues 49 -286) of ASV integrase are capable of carrying out the processing and joining reactions (14,23,24). However, deletion of either the N-terminal or C-terminal domain of HIV-1 IN abolishes both the processing and joining reactions in vitro and viral replication in vivo (25)(26)(27). With the increase in our understanding of the structure and function of retroviral IN, this enzyme has become an attractive target for new drug development to treat patients with AIDS (28). Several types of compounds identified by in vitro screening assays were recently found to be potential inhibitors of IN. Some of the compounds bind to the catalytic core domain (28 -31); the binding sites of others have not been identified. However, because they are essential for activity, both the N-terminal and the C-terminal domains of the HIV-1 IN may be considered as possible targets for anti-AIDS drug design (32,33).
A library of monoclonal antibodies was raised against purified HIV-1 IN and characterized in our laboratory (34). Several of these mAbs inhibit the enzymatic activities of IN in vitro (8). Among these, mAbs17, -4, and -33 bind to N-terminal, catalytic core, and C-terminal domain of HIV-1 IN, respectively (Fig.  1A). An understanding of the basis for such inhibition may suggest new approaches for the development of antiviral drugs. We report here the results from analysis of mAb17 binding to the N-terminal domain of HIV-1 IN. Its epitope was mapped to determinants that include the turn of the helix-turn-helix structure that is stabilized by coordination of Zn 2ϩ by the HHCC motif in this domain. Our results support the interpretation that these determinants play a critical role in the formation of the catalytically active protein. They also suggest a possible target for drug design.

EXPERIMENTAL PROCEDURES
Expression and Purification of HIV-1 IN and Derivatives-The construction of plasmid pET29b encoding wild type HIV-1 IN(8) and mutants IN-F185KC2805 (8) and IN-3CS (C56S, C65S, and C280S) (10) has been reported previously. The HIV-1 IN-F185KC2805 or IN-3CS proteins, which have much higher solubility than wild type IN, but wild type activity in vitro (48,10), were overexpressed in Escherichia coli BL21(DE3) and purified from soluble extracts (8,10). The construction of plasmid pET28b that included a gene encoding the N-terminal domain fragment (amino acids 1-49) with a six-histidine tag and a thrombin cleavage site on the N-terminal end of the protein was also described in a previous report (8). This plasmid was also expressed in E. coli BL21(DE3) cells. The N-terminal domain fragment was bound to a HiTrap Ni 2ϩ -chelating column (Amersham Pharmacia Biotech) and eluted with a linear gradient of 0.05-1.0 M imidazole in 20 mM HEPES buffer (pH 7.5) containing 1 M NaCl. The pool of protein-containing fractions was dialyzed in 20 mM HEPES buffer (pH 7.5), 0.5 M NaCl, 20% glycerol and 0.5 mM CaCl 2 . The His 6 tag was removed by cleavage with thrombin, added at a ratio of 1000:1 (mg IN/mg thrombin), and incubated with shaking for 4 -6 h at room temperature. Proteolysis was terminated by the addition of 5 mM EDTA. The mixture was passed through a p-aminobenzamidine-agarose (Sigma) column to remove the thrombin. The protein was then loaded on a Ni 2ϩ -chelating column to remove the His 6 tag and any remaining His 6 -tagged proteins. The fraction not bound to the column was collected and concentrated. As this purified protein includes four amino acid residues, GHSM, fused to its N terminus it is here denoted G t -IN .
Purification of Anti-HIV-1 Integrase Monoclonal Antibody 17 and Isolation of Its Fab Fragments-The preparation of ascites from mice and the purification of the anti-HIV-1 IN monoclonal antibodies from the ascites were described previously (34). To isolate Fab fragments, purified monoclonal antibodies mAb17, -4, or -33 (1-2 mg/ml) were incubated at room temperature with papain (Sigma) at a ratio of mAb/ papain ϭ 200:1 (mg/mg) in phosphate-buffered saline (25 mM NaPO 4 , 150 mM NaCl (pH 7.4)) containing 10 mM cysteine. To monitor the proteolytic process, a 2-l aliquot was withdrawn every 30 min and analyzed by SDS-polyacrylamide gel electrophoresis. After digestion, iodoacetamide (Sigma) was added to final concentration 20 mM, and the mixture was incubated for an additional 1 h in the dark to inhibit the papain. To remove the undigested monoclonal antibody and Fc fragments, the preparation was dialyzed in 1 M glycine buffer (pH 8.9) containing 0.5 M NaCl and then loaded on a HiTrap Protein A column (Amersham Pharmacia Biotech) pre-equilibrated with the above buffer. The fractions containing Fab fragments passed through the column and were pooled. The Fab preparations were purified further by loading the protein on a Resource Q column (Amersham Pharmacia Biotech) adapted to a high pressure liquid chromatographic system and eluting with a linear gradient of 0.05-0.5 M NaCl in 50 mM Tris/HCl buffer (pH 8.9). The two most abundant isomeric forms of the mAb 17 Fab frag-ments were also separated in this step. Both isoforms were equally active in our assays. The collected Fab-containing fractions were dialyzed in phosphate-buffered saline, concentrated by ultrafiltration in a Centricon 10 (Amicon), and stored in a Ϫ20°C freezer until use. The concentration of Fabs was determined by Bio-Rad protein assay with IgG as a standard or by UV measurement with an extinction coefficiency constant 1.5 g of IgG/liter.
Oligonucleotide Activity Assay-Enzymatic activity of HIV-1 IN was determined as reported previously (8).
Enzyme-linked Immunosorbent Assay (ELISA)-HIV-1 IN proteins (either wild type or IN-3CS or G t -IN ) was applied to 96-well, high-binding microtiter plates by overnight incubation of 1 g/ml IN protein and 0.2 mM ZnCl 2 in TBS (20 mM Tris-HCl buffer (pH 7.5), 150 mM NaCl) with a total 50 l volume in each well. Subsequently, 50 l of 1 mg/ml bovine serum albumin in the above buffer was added to each well and incubated for 2 h to block the remaining binding sites. In order to investigate the effect of Zn 2ϩ binding on the recognition of IN protein by mAb17, the plates were washed once with 200 l of TBS containing various concentrations of EDTA to remove the metal bound to the protein ( Fig. 2A). The plates were further washed with 200 l of TBS buffer four times, followed by addition of primary antibodies, and second antibodies labeled with horseradish peroxidase. The standard protocol was then followed, and the relative binding efficiencies of monoclonal antibody to the immobilized IN protein were determined by measurement of absorbance at 405 nm (8,34).  Fig. 3 in order to catch the signals of species with high molecular weight). The drops were allowed to dry at 25°C prior to mass analysis. The mass spectra were recorded in a linear mode at an acceleration voltage of 20 kV. Each spectrum was acquired by averaging 256 scans. The calibration was first carried out using two internal standards, CHCA dimer (MH ϩ , 390.37) and undigested G t -IN(1-49) (MH ϩ , 5933.17). Several assigned IN-derived peptides in each spectrum were used subsequently as internal standards for the final reported results. The calibrated masses were analyzed with the Paws computer software (ProteoMetrics) and mapped to G t -IN(1-49)-derived peptides. A match with an error no greater than 1.0 mass unit was considered to be acceptable.
Modeling of Antibody mAb17-PSI-BLAST (35) was used to compare the sequence of the heavy and light chain variable domains of mAb17 with sequences of known structure in the Protein Data Bank (PDB) (36). For the light chain, the closest sequence in the PDB was chain L of PDB entry 1A0Q, a 2.30-Å crystal structure of catalytic antibody 29G11 in complex with [1-(1-N-succinylamino)pentyl]-2-phosphonate (37). The alignment had no gaps and gave a sequence identity of 88%. For the heavy chain, we used a composite of two antibody crystal structures. The majority of the chain, except for the H2 CDR loop, was taken from the 2.50-Å crystal structure of antitumor antibody R24, PDB entry 1BZ7, chain B (38). The H3 loop of the R24 antibody is the same length as that in mAb17, and they share three aromatic amino acids that are not present in the H3 loops of other antibodies of known structure. These residues are likely to determine the conformation of the loop. However, this antibody has an H2 loop one amino acid longer than the mAb17 H2 loop. For this loop we used the 2.7-Å crystal structure of the anti-digoxin antibody 40 -50 in PDB entry 1IBG, chain H (39). The structures of the heavy chain variable domain from PDB entries 1BZ7 and 1IBG were aligned with the MIDAS program (40), such that 3-residue anchors on either side of the H2 loop superimposed with 0.1-Å root mean square deviation. The composite coordinates were compiled to form a single heavy chain variable domain consisting of the H2 loop from 1IBG-H and the rest of the domain from 1BZ7-B. The heavy chain model was docked onto the light chain model by superimposing the backbone coordinates of the frame onto those in PDB entry 1A0Q chain H using MIDAS (40) and by compiling a single file with the models of the light and heavy chain variable domains of mAb17. We used the side chain conformation prediction program SC-WRL (41,42) to model the side chain coordinates of the Fab17 sequence onto the backbone model produced from PDB entries 1A0Q, 1BZ7, and 1IBG. SCWRL uses a backbone-dependent rotamer library (43,44) to place side chains onto a fixed backbone, followed by combinatorial optimization to remove steric clashes, and is more accurate than alternative methods (42). No further energy minimization was performed, as such minimization usually does not improve a model (45).
Docking Fab17 with the N-terminal Domain of HIV-1 IN-The program HEX version 2.3 (46) was used to identify a reasonable structure for the complex between the antibody, Fab17, and the integrase domain, G t -IN . The structure of this domain was taken from the PDB (1WJC) (16). The structure of the Fab was modeled as described above.
The electrostatic surface potentials of the N-terminal domain dimer and Fab17 were studied using GRASP (47). No obvious complementarity was observed. A large, relatively neutral surface was found on the "top" of the dimer (see "Results"). The top is described by a surface passing through the N terminus and Asn 27 of each monomer. Binding at this surface was also consistent with experimental proteolysis data (see "Results"). Hence, we chose an initial binding configuration that oriented the complementarity determining region (CDR) binding loops of the antibody with the identified surface of the integrase. In the rigid docking calculation, the dimer and Fab were allowed to sample points over a wide range of orientations and displacements starting from the original position (HEX parameters contact ϭ 45, samples ϭ 492, ␣ samples ϭ *, distance samples ϭ 31). In the lowest energy configuration (Ϫ572 kJ⅐mol Ϫ1 ), the CDR loops of the antibody were involved in the binding of the mAb17 to the integrase. This was not the case in any of the next four configurations with low binding energies (Ϫ559, Ϫ541, Ϫ518, and Ϫ511 kJ⅐mol Ϫ1 ). Hence, the lowest energy structure was accepted and is presented below.

Monoclonal Antibody 17 and Its Isolated Fab Fragment
Inhibit IN Activity-Our initial HIV-1 IN activity assays indicated that several monoclonal antibodies that we had isolated inhibited HIV-1 IN activity in vitro (data not shown). Of these, mAb17 bound to the N-terminal region of HIV-1 IN (Fig. 1A) (34). To determine if this inhibition was due to cross-linking of IN via bivalence of the mAb, we isolated its monovalent Fab derivative, and we tested its effect on the HIV-1 IN activity. As shown in Fig. 1B, Fab17 inhibited the processing reaction of wild type IN approximately one-tenth the efficiency of mAb17, with IC 50 ϳ4.5 and 0.4 M, respectively. Both mAb17 and Fab17 also inhibited the joining reaction (data not shown). Under the same conditions, a nonspecific mAb (mouse IgG 1 ,

FIG. 1. Structural and functional domains of HIV-1 IN and its interaction with mAb17.
A, linear model of HIV-1 integrase protein and its domains recognized by three mAbs. The N-terminal domain with a helix-turn-helix structure contains the conserved residues HHCC that bind Zn 2ϩ (16,17). The catalytic core domain, with polynucleotidyl transfer activity, includes a highly conserved D,D(35)E motif and binds Mg 2ϩ or Mn 2ϩ (4,5). The C-terminal domain with an SH3-like structure (11)(12)(13) has nonspecific DNA binding activity. The numbers above the map indicate the approximate borders of domains, and numbers below the map show the positions of conserved residues of HIV-1 IN. mAbs17, -4, -33 recognize N-terminal, catalytic core, and C-terminal domains, respectively (34). B, inhibitions of the processing activity of HIV-1 IN by mAb17 (Ⅺ) and Fab17 (f) and control mAb (␣MOPC12) (F). 1 M wild type IN protein was incubated with mAb17 or Fab17 in HEPES buffer (pH 7.5) containing 10 mM MnCl 2 , 6.67% Me 2 SO, 10% glycerol. The reaction was initiated by addition of 1 M 32 P-labeled 21-base pair model viral DNA substrate and incubation was for 30 min. Processing was measured by quantitation of the Ϫ2 cleavage product after exposure of the radioactive gel to an imaging plate and analysis on a Fuji MacBAS2000 imaging system. C, ELISA results showing mAb17 and Fab17 interaction with HIV-1 IN. A high binding microtiter plate was coated with full-length IN(IN-3CS), blocked with bovine serum albumin, and washed as described under "Experimental Procedures." Either mAb17 (Ⅺ) or Fab17 (f) were subsequently added as primary antibody, followed by a standard ELISA protocol using a secondary antibody against chain (8,34). ␣MOPC21) that has the same isotype as mAb17 shows no significant inhibitory effect (Fig. 1B). These results indicated that bivalent binding was not required for the inhibition by mAb17 but enhanced the inhibitory effect. Independent experiments (to be reported elsewhere) showed that mAb17 does not block the binding of IN to its DNA substrate(s). Results from an ELISA summarized in Fig. 1C indicate that the intact mAb17 binds more tightly to HIV-1 IN-3CS than the Fab fragment. The half-saturation concentrations of mAb17 and Fab17 are 2.14 ϫ 10 Ϫ9 and 2.  , coated the wells, the same effect of EDTA was observed (Fig. 2B). These results suggest that Zn 2ϩ is required for mAb17 binding to IN, presumably because the complex of Zn 2ϩ with the HHCC motif stabilizes the mAb17 epitope. As Zn 2ϩ facilitates folding of N-terminal domain of HIV-1 IN (16,17), these observations also suggest that Fab17 recognizes the properly folded protein preferentially.
Binding of Fab17 Does Not Cause Dissociation of the Nterminal Domain Dimers-Previous size exclusion chromatography and NMR structural studies revealed that the isolated N-terminal domain of HIV-1 IN is a dimer in solution (16,17). This subunit-subunit interaction is stabilized by Zn 2ϩ binding to the HHCC motif and is believed to contribute to formation of tetrameric or higher oligomeric forms of the full-length IN protein which may be required for the concerted insertion of two viral DNA ends into the DNA of host cells. To determine if Fab17 inhibits the inter-subunit protein-protein interaction of the N-terminal domains, we analyzed the complex formed between   (48,10). However, even these variant proteins are poorly soluble at low concentrations of NaCl. For example, in 50 mM NaCl and 20 mM HEPES buffer (pH 7.5), IN-F185KC280S precipitates (Fig. 4, lane 3) and the protein is detected mainly in the pellet fraction (lane 3); IN-3CS produces the same protein distribution under these conditions (data not shown). Addition of a nonspecific mAb␣MOPC21 does not change this distribution of the protein (lanes 4 -6). In contrast, in the presence of equimolar mAb17 more than half of the IN protein remains in the supernatant fraction (lanes 7-9 in Fig. 4). Fab17 has even a stronger effect on the solubility IN (lanes 10 -12). This difference may be due to some precipitation of IN by the divalent mAb17. Nevertheless, these results indicate that binding of mAb17 or Fab17 increases the solubility of HIV-1 integrase.
Mapping the mAb17 Epitope-Preliminary ELISAs 2 indicated that mAb17 bound only to HIV-IN-derived peptides that included amino acids 22-35. To map the binding region more precisely, and to investigate whether any other structural alterations could be detected as a consequence of binding, we developed a protein footprint technique using controlled proteolysis of the G t -IN(1-49)-Fab17 complex, combined with peptide mapping by MALDI-TOF mass spectrometry. For these analyses, a specific protease was added to the soluble complex formed between the isolated N-terminal fragment and Fab17, and the digestion products were analyzed by MALDI-TOF mass spectrometry after increasing times of incubation. Considering that the protease may be consumed by the Fab17 fragments during the digestion process, a control antibody, Fab33, which does not interact with the N-terminal domain (see Fig. 1A) of HIV-1 IN, was added to parallel samples. Preliminary tests showed that the presence of this control Fab did not affect the kinetics of proteolysis of IN. The same spectra were obtained with a second control antibody, Fab4 (data not shown). Thus, sites protected by Fab17 could be identified by comparison of the spectra obtained in the presence of Fab17 with those in the presence of Fab33.
Proteolysis by Trypsin-Trypsin cleaves specifically on the carbonyl side (C-side) of residues with positive charges on their side chains, i.e. lysine and arginine. also the same with both Fabs. These observations indicate that binding of mAb17 does not protect residues Lys 7 or Lys 14 from proteolysis. On the other hand, peptide Ala 21 -Ala 49 (peak JЈ in Fig. 5B; MH ϩ , 3035.5), a direct product from G t -IN(1-49), was apparent only for the first 10 min in the presence of Fab33. Disappearance of this peak was due to the cleavage at C-side of Lys 34 which yielded a smaller peptide Ala 21 -Lys 34 (peak LЈ in Fig. 5B; MH ϩ , 1460.7). This new peptide Ala 21 -Lys 34 was then a dominant peak that was stable during trypsin digestion. In contrast, in the presence of Fab17 peptide Ala 21 -Ala 49 (peak J in Fig. 5A) was one of the major products during the first 3 h of digestion. It then disappeared gradually due to cleavage on the C-side of Lys 46 , to give a shorter peptide Ala 21 -Lys 46 (peak K in Fig. 5A, MH ϩ , 2779.3). This peptide was stable for more than 48 h in the presence of Fab17. Therefore, Lys 34 is not accessible to trypsin after binding of Fab17. Another difference observed between the two spectra was that the peptide G t -IN(1-34) (peak QЈ in 5B, MH ϩ , 4358.0), a direct product from G t -IN(1-49), appeared after 1 min of digestion in the presence of Fab33 (Fig. 5B) due to cleavage on the C-side of Lys 34 , and it disappeared after 10 min of digestion due to further cutting on the carbonyl sides of Lys 7 and Lys 14 . However, peptide G t -IN  was not visible at any time in the presence of Fab17. This difference also suggests that Lys 34 is protected by Fab17 binding to IN. From the above comparisons, it is apparent that Lys 34 , but not Lys 7 , Lys 14 , Arg 20 , or Lys 46 , is protected by the binding of Fab17.
Proteolysis by Endoprotease Asp-N-Asp-N cleaves specifically on the amino side (N-side) of aspartate so that this protease allows us to ask if any aspartate residue can be protected by Fab17. We observed that three peptides, Asp 3 -Ala 49 (peak AЈ, MH ϩ , 5259.9), G t -IN(1-24) (peak GЈ, MH ϩ , 3275.6), and Asp 25 -Ala 49 (peak VЈ, MH ϩ , 2675.1), were produced from G t -IN  in the presence of Fab33 after 1 min of digestion (Fig.  6B). These arose from cleavages at Asp 3 and Asp 25 . The relative abundance of G t -IN(1-49) (peak FЈ) decreased with digestion time, but this substrate was still visible after 1 h of proteolysis. The relative signals of product peptides Asp 3 -Ala 49 and G t -IN(1-24) increased initially and then decreased. Thus, these peptides are intermediates in proteolysis. Asp 3 -Ala 49 was further digested either at Asp 25 to give two peptides Asp 3 -Ser 24 (peak UЈ, MH ϩ , 2602.8) and Asp 25 -Ala 49 (peak VЈ) or at Asp 6 to give Asp 6 -Ala 49 (peak BЈ, MH ϩ , 4974.6), which was subsequently cleaved at Asp 25 . Peptide G t -IN(1-24) coexisted with G t -IN . This peptide is a direct product of G t -IN  and was subsequently cleaved at Asp 3 and/or Asp 6 . However, peptide Asp 25 -Ala 49 was relatively resistant to digestion and was cleaved slowly at Asp 41 to produce a detectable peak of Asp 25 -Cys 40 (peak XЈ, MH ϩ , 1702), only after 4.5 h of exposure to Asp-N.
In contrast, in the presence of Fab17, G t -IN(1-49) disappeared faster than when Fab33 was present; this peak was undetectable after 30 min of digestion in the presence of Fab17 (Fig. 6A), compared with 2 h with Fab33. Thus, Fab17 promoted digestion of the G t -IN(1-49) by Asp-N. The relative abundance of peptides Asp 6 -Ala 49 and Asp 3 -Ala 49 (peaks B and A, respectively, in Fig. 6A) was higher in the presence of Fab17 than Fab33 during the same period of digestion, indicating that proteolysis at residues Asp 3 and Asp 6 was enhanced by the binding of Fab17. On the other hand, peptide G t -IN(1-24), a direct product of digestion G t -IN  at Asp 25 in the presence of Fab33 (peak GЈ in Fig. 6B), was not detected in the presence of Fab17. Peptide product assignments and their relative kinetics of appearance are summarized in Table I. These results suggest that Asp 25 , but not Asp 3 , Asp 6 , and Asp 41 , is included in the epitope of mAb17 binding to IN. In addition, binding of Fab17 appears to induce sufficient molecular distortion to render Asp 3 and Asp 6 more sensitive to digestion by Asp-N.
Proteolysis by Chymotrypsin-Similar analyses were performed after proteolysis by chymotrypsin which cleaves at Cside of aromatic residues Tyr, Phe, and Trp and also cleaves at C-side of Leu, Met, Ala, Glu, and Asp at a low rate. The peptide products identified are listed in Table II Table I. Peaks labeled with question marks could not be assigned.  (Table III) IN(1-49).
The effects of Fab17 on the proteolytic digestion of the Nterminal domain of HIV-1 IN by trypsin, chymotrypsin, Asp-N, and Glu-C are summarized in Fig. 7A. No changes in digestion by these enzymes due to Fab17 binding were observed in the regions Cys 40 to Ala 49 and Arg 20 to Met 22 . However, this binding was found to promote proteolysis in the region from Asp 3 to Trp 19 and, most importantly, to protect completely residues Asp 25 , Phe 26 , and Lys 34 , and to protect partially residue Glu 35 from proteolytic digestion. We conclude, therefore, that the recognition site for Fab17 is included within the region spanning Asp 25 to Glu 35 . The positions of the protected residues in a three-dimensional model of this domain are shown in Fig. 7B. All are located in the helix-turn-helix motif that composes a dominant feature of this domain. Asp 25 and Phe 26 are located at the end of the second helix. Lys 34 and Glu 35 are in the middle of the third helix that contributes the formation of the Nterminal domain dimer. The sites that are rendered more susceptible to proteolysis are included in ␣-helix 1, which stacks against the ␣-helix that includes Asp 25 , in the HX 3 H region that participates in Zn 2ϩ binding, and in the N-terminal portion of ␣-helix 2.
Modeling of Fab17 Bound to the N-terminal Domain of Integrase-A molecular model of Fab17 was constructed from the known amino acid sequence of its heavy and light chain combining sites (49) 3 and the structures of antibodies in the Protein Data Bank (PDB) entries 1ADQ, 1B27, and 1IBG (described under "Experimental Procedures"). By using the program HEX, the Fab17 model was then docked with the structure of the N-terminal domain of HIV-1 IN, PDB entry 1WJC (16). Two Fabs could not be docked with a single N-terminal domain dimer in a manner that was consistent with the footprinting data. The best fit was with a complex comprising one N-terminal domain dimer and one Fab17 molecule (Fig. 8). In this complex, the top of the N-terminal domain dimer with a large, relatively neutral surface (Fig. 9) is neatly bound by the CDR loops of the antibody. Each loop has some residues in contact with the IN dimer. The surfaces are complementary in shape with no large gaps between the IN and the antibody. All six residues in Asp 3 -Trp 19 that were identified as being more susceptible to proteolysis remain open to solution after the antibody binds. Two of the four residues, i.e. Lys 34 and Glu 35 , identified as being protected from proteolysis by Fab17 binding are located on the dimer interface. The binding between two N-terminal domain monomers is likely to be enhanced by the binding of the antibody across the top of the dimer, thus providing a possible explanation for the protection at these two sites. The other two protected residues, Asp 25 Fig. 6A; their counterparts are labeled with prime symbols as superscripts in Fig. 6B. b Value of MH ϩ obs Ϫ MH ϩ calc . c A tryptic product due to long time incubation.   the other, again providing a plausible explanation for their protection against proteolysis.

DISCUSSION
In this report we have analyzed the molecular interaction between HIV-1 IN and a monoclonal antibody (mAb17) that was known to bind to its N-terminal domain. We have provided evidence (Fig. 1B) that both the mAb and its derivative Fab fragment can increase the solubility of HIV-IN in low ionic strength (Fig. 4) and inhibit the enzymatic activity of HIV-1 IN in vitro. We have observed that the intact mAb is approximately 10 times more active as an inhibitor than the isolated Fab fragment, even when corrected for valence. This difference might be due to decreased stability caused by dissociating the antigen combining-site from the remainder of the immunoglobulin. The reduced affinity of the Fab for IN relative to the intact mAb (Fig. 1C) is consistent with this interpretation. Alternatively, as suggested by the results in Fig. 4, precipitation or aggregation of IN remaining with the divalent Ab may account for some inhibition by the intact antibody. As the binding of the Fab does not introduce this complication, we presume that its inhibitory activity is a direct consequence of molecular contacts.
By analyzing the mass spectra of peptides produced with increasing time of proteolysis, we were able to monitor the kinetics of cleavage at specific sites and, thus, to detect either enhancement or protection due to binding of a specific mAb or Fab. In this study we showed that binding of Fab17 to the isolated N-terminal domain of HIV-1 IN strongly inhibited proteolytic digestion at residues Asp 25 , Phe 26 , Lys 34 and partially at Glu 35 (Fig. 7). Therefore, we conclude that the epitope is included within the 11-amino acid region spanning from 25 to 35. On the other hand, Fab17 was found to stimulate proteolysis at residues Asp 3 , Asp 6 , Glu 10 , Glu 13 , Tyr 15 , and Trp 19 . This observation suggests that the binding of Fab17 causes or stabilizes a conformational change in which the N-terminal helix is rotated away from the HTH motif. Thus, this method of analysis may also detect structural distortions induced by antibody binding.
We have observed that the binding of mAb17 to either the full-length HIV-1 or the N-terminal domain, G t -IN(1-49), is The positions of these residues are also shown in Fig. 9. The helix and sheet elements are drawn as ribbons, and the coil regions are traced to the positions of the ␣-carbons. All the highlighted residues are in the coil regions except Tyr 96 that belongs to the nearby ␤-strand. Its ␣-carbon atom does not coincide with the strand ribbon. stimulated by addition of Zn 2ϩ (Fig. 2). The HHCC motif in this domain is known to coordinate a Zn 2ϩ , and such coordination stabilizes an HTH structure and promotes formation of tetrameric or higher multimers of IN (16, 18 -20). The fact that a properly folded structure is required for the most stable binding of mAb17 suggests that its epitope is conformational and/or that a multimeric form of IN is required for stable binding by the antibody. Our MALDI-TOF mass spectrometry analyses revealed the presence of complexes that included one molecule each of the isolated N-terminal domain and Fab17, [G t -IN(1-49)] 1 ⅐(Fab17) 1 , as well as two to one complexes [G t -IN(1-49)] 2 ⅐(Fab17) 1 (Fig. 3). The one to one complex could be produced from the two to one complex during the MALDI-TOF analysis. Indeed our molecular model suggests that the epitope may span the dimer in a two to one complex (Fig. 8). In this complex the same IN peptides in two IN protomers contact one Fab molecule asymmetrically, with one peptide contacting the heavy chain and the other the light chain of the Fab. Furthermore, the most neutral surface of N-terminal dimer ( Fig. 9) is covered by Fab17 in this complex, consistent with the fact that mAb17 increases the solubility of HIV-1 IN in low ion strength buffer.
The model of the dimeric N-terminal domain complexed with one Fab also provides an explanation for the protective effects of Fab17 against proteolysis. The protection of Asp 25 and Phe 26 can be attributed to the direct contact of these residues with Fab17. Protection of Lys 34 and Glu 35 may be due to stabilization of the dimer by the binding of the Fab. The model suggests that residues in direct contact also include Asn 27 , Ile 28 , and Pro 29 (green in Fig. 8). These residues contribute mainly to the turn of the HTH structure. The results further imply an important role for this turn. The top of the integrase dimer that interfaces with mAb17 was shown to have a largely neutral (i.e. hydrophobic) electrostatic potential (Fig. 9), as calculated by GRASP (47). Such a surface is likely to be involved in protein binding interactions as a means of shielding these hydrophobic residues from exposure to the solvent. This is confirmed by our docking simulations between the integrase and mAb17, in which mAb17 binding directly to this neutral surface gave the most stable configuration (Fig. 8). However, our model of the complex does not provide insight into the basis for destabilization of the N-terminal ␣-helix, as the docking program does not allow for possible conformational changes induced by binding of the Fab.
What can we infer concerning the mechanism of mAb17 inhibition from the results of our analyses? Our data indicate that residues in the region Asp 25 -Glu 35 , which are highly conserved among retroviral integrases (50) and contribute to dimerization of the N-terminal domain (16,17), are important for IN catalytic function. We have eliminated the possibility that mAb17 dissociates the N-terminal dimer. As mAb17 binds to a neutral surface of the domain dimer and hydrophobic interactions provide the bulk of binding free energy in proteinprotein interfaces, mAb17 binding may block critical interdomain interactions of IN. Such a role would be consistent with previous reports that the N-terminal domain makes direct contact with the core domain in an active conformation of IN (52) and is close to the C-terminal domain in a native protein (34). The importance of amino acids in the peptide Asp 25 -Glu 35 is also consistent with previous mutagenesis studies (51) that show a total loss in processing and joining activities of mutant IN-V32G, A33L, and K34A. In addition, a single substitution of Ala for Pro 29 , which forms a turn between two helices, decreases both processing and joining activities. Residues C-terminal of Glu 35 (51,53). We note that that Fab17 has no effect on the proteolysis near these residues. On the other hand, our results also suggest that Fab17 causes a conformational change in the region Asp 3 -Trp 19 . Thus, it is also possible that this distortion contributes to the inhibitory effect of mAb17.
Targeting the N-terminal region of IN has been considered for anti-AIDS drug design. One type of inhibitor, guanosine quartets, was found to require the N-terminal domain for binding (33). The recognition site of mAb17 may represent a novel target in this domain for the development of inhibitory drugs. The hydrophobic nature of its binding surface on IN and the existence of two cavities toward the center of this surface present a possible target for drug binding. Two symmetrical aromatic groups connected by a carbon chain could conceivably bind to these two cavities and prevent the N-terminal dimer from undergoing protein-protein interactions. More detailed drug design calculations are required to explore this possibility. FIG. 9. Electrostatic surface potential of the N-terminal domain dimer as generated by GRASP (43). The surface of the dimer is colored according to the electrostatic potential at each point on the surface. Blue indicates positive potential, and red indicates negative potential. The surface shown is called the top in the text. It can be seen to be predominantly neutral, i.e. hydrophobic with two indentations toward the center of the surface. The hydrophobic residues of mAb17 contacting the top of integrase within 5 Å are showed as stick models. These residues, including one Trp and four Tyr residues, are labeled in Fig. 8.