Mapping the Epitope of an Inhibitory Monoclonal Antibody to the C-terminal DNA-binding Domain of HIV-1 Integrase*

Integrase (IN) catalyzes the insertion of retroviral DNA into chromosomal DNA of a host cell and is one of three virus-encoded enzymes that are required for replication. A library of monoclonal antibodies against human immunodeficiency virus type 1 (HIV-1) IN was raised and characterized in our laboratory. Among them, monoclonal antibody (mAb) 33 and mAb32 compete for binding to the C-terminal domain of the HIV-1 IN protein. Here, we show that mAb33 is a strong inhibitor of IN catalytic activity, whereas mAb32 is only weakly inhibitory. Furthermore, as the Fab fragment of mAb32 had no effect on IN activity, inhibition by this mAb may result solely from its bivalency. In contrast, Fab33 did inhibit IN catalytic activity, although bivalent binding by mAb33 may enhance the inhibition. Interaction with Fab33 also prevented DNA binding to the isolated C-terminal domain of IN. Results from size-exclusion chromatography, gel electrophoresis, and matrix-assisted laser desorption ionization time-of-flight mass spectrometric analyses revealed that multiple Fab33·IN C-terminal domain complexes exist in solution. Studies using heteronuclear NMR showed a steep decrease in1H-15N cross-peak intensity for 8 residues in the isolated C-terminal domain upon binding of Fab33, indicating that these residues become immobilized in the complex. Among them, Ala239 and Ile251 are buried in the interior of the domain, whereas the remaining residues (Phe223, Arg224, Tyr226, Lys244, Ile267, and Ile268) form a contiguous, solvent-accessible patch on the surface of the protein likely including the epitope of Fab33. Molecular modeling of Fab33 followed by computer-assisted docking with the IN C-terminal domain suggested a structure for the antibody-antigen complex that is consistent with our experimental data and suggested a potential target for anti-AIDS drug design.

is one of three virus-encoded enzymes that are required for retroviral replication (1). IN catalyzes insertion of the linear double-stranded viral DNA into the chromosomal DNA of a host cell. The IN protein comprises three distinct regions known as the N-terminal, catalytic core, and C-terminal domains (see Fig. 1A) (2,3). The isolated N-terminal domain (residues 1 to ϳ50) assumes a three-helix bundle structure with a helix-turn-helix motif stabilized by Zn 2ϩ coordination that stimulates enzymatic activity in vitro (4,5). The N-terminal domain contributes to formation of tetrameric or higher multimeric forms of the protein (6 -8). Crystallographic analysis of the catalytic core domains of human immunodeficiency virus type 1 (HIV-1) IN and avian sarcoma virus IN revealed that each subunit binds at least one divalent cation cofactor, Mg 2ϩ or Mn 2ϩ , mediated by acidic residues in the highly conserved D,D(35)E motif that comprises the active site of the enzyme (9 -12). Binding of the metal cofactor to this motif activates HIV-1 IN (13,14) and stimulates preferential attachment of the protein to its viral DNA substrate (15). The Cterminal domain, comprising amino acids 220 -270, binds DNA. Its NMR structure resembles that of the SH3 domain, a motif known to promote protein-protein interactions (16 -18).
Although the isolated C-terminal domain forms a homodimer in solution, only the catalytic core domain forms a dimer in the crystal structure of a two-domain derivative of HIV-1 IN that includes the core and C-terminal domains (19). In such crystals, the C-terminal domain moieties are separated from each other by 55 Å, but are involved in a variety of contacts with C-terminal domains in adjacent unit cells.
IN proteins must recognize and bind to both viral and host cell target DNAs. In the processing reaction, two nucleotides are removed from the 3Ј-ends of the viral DNA. These new 3Ј-ends are then joined to phosphorus atoms in the backbone of both strands of the target DNA in a concerted cleavage and ligation reaction. The strongest DNA-binding determinants of HIV-1 IN have been localized to the C-terminal domain, but such binding is sequence-independent. Despite intense efforts in many laboratories, the detailed mechanism by which IN interacts with host and viral DNAs remains unknown. To address this and other questions of structure and function, a library of monoclonal antibodies (mAbs) was raised against HIV-1 IN and characterized in our laboratory (20). Several of these mAbs inhibit the enzymatic activities of IN in vitro. Among these, mAb17 binds to the N-terminal domain, mAb4 binds to the catalytic core, and mAb32 and mAb33 bind to the C-terminal domain of IN (see Fig. 1A). Details of the binding of the N-terminal domain to the inhibitory mAb17 have been reported (21); its epitope was mapped to a relatively neutral surface of the dimeric N-terminal domain, likely to be involved in protein-protein interaction. Here, we describe detailed binding studies and structural analyses of the interaction of the C-terminal domain-specific mAb33. We show that binding of Fab33 prevented the interactions of the C-terminal domain with DNA substrates and also inhibited the enzymatic activity of full-length IN, whereas binding of Fab32 produced neither effect. The epitope recognized by Fab33 was mapped to the three-dimensional structure of IN-(220 -270) using heteronuclear NMR spectroscopy and computer-assisted molecular modeling.  (13)(14)(15). For NMR experiments, the protein expression vectors pET28b/IN-(220 -270) and pET28b/IN-(220 -288) were prepared by inserting sequences encoding amino acids 220 -270 and 220 -288 of HIV-1 IN, respectively, into plasmid pET28b (Novagen, Madison, WI) as described (21). The products include four expression plasmid-encoded amino acids (GSHM) at their N termini, which remain after thrombin cleavage to remove the His 6 tag. 15 (21,22).

HIV-1 IN Proteins and Anti-IN
Oligonucleotide Activity Assay-The effect of mAb32 and mAb33 on the enzymatic activity of full-length HIV-1 IN was determined by measuring both the processing and joining reactions using a 21-base pair oligonucleotide duplex that represents the viral U5 DNA end as substrate (13). Assay conditions are described in the relevant figure legends.
Kinetic Analysis and Efficiency of IN Binding to DNA Substrates-Surface plasmon resonance (SPR; BIACORE) was employed to analyze the interactions between the isolated C-terminal domain of IN and DNA substrates. Double-stranded oligonucleotides representing either the viral U5 DNA end (20-mer) or a target DNA substrate (24-mer) with no sequence match with viral DNA ends were immobilized on the surface of a chip as described (15). To allow association, solutions of the IN protein were applied to a chip containing immobilized DNA. Dissociation of IN from the nucleoprotein complex was monitored in real-time after application of buffer to wash the chip. The kinetic rate constants for dissociation (k off or k d ) and for apparent association (k on or k a ) were obtained by fitting the real-time data, and the apparent dissociation constant was calculated as K d ϭ k off /k on (15).
Enzyme-linked Immunosorbent Assay (ELISA)-The HIV-1 IN-3CS protein was immobilized on 96-well high binding microtiter plates by applying a solution containing 1 g IN/well in a total volume of 50 l of Tris-buffered saline (20 mM Tris-HCl (pH 7.5) and 150 mM NaCl). After overnight incubation, 50 l of 1 mg/ml bovine serum albumin in the above buffer was added to each well, and the plates were incubated for 2 h to block the remaining binding sites. The plates were subsequently washed with 200 l of Tris-buffered saline four times, followed by addition of primary antibodies and secondary antibodies labeled with horseradish peroxidase. The standard protocol was then followed, and the relative binding efficiency of monoclonal antibody to the immobilized IN protein was determined by measurement of absorbance at 405 nm (13,20,21).
NMR Spectroscopy-NMR spectra were recorded at 37°C on a Bruker DMX 600-MHz spectrometer equipped with a 5-mm x,y,zshielded pulsed-field gradient triple-resonance probe. Typically, each sample contained ϳ0.5 mM IN-(220 -288) dissolved in 95% H 2 O and 5% D 2 O at a final buffer concentration of 50 mM NaH 2 PO 4 (pH 6.5), 100 mM NaCl, and 0.5 mM EDTA, which was the same as that used by Lodi et al. (16). Protein-mAb interaction was studied by recording 1 H, 15 15 N HSQC spectrum was recorded as a 2048 ( 1 H) ϫ 128 ( 15 N) data matrix with acquisition times of 285 ms in t 2 and 78 ms in t 1 . The data were acquired with 32 scans for each hypercomplex t 1 /t 2 increment with an interscan delay of 1 s. The total acquisition time was 3.5 h for each spectrum. The NMR data were processed using XWINNMR2 (Bruker) and analyzed using Felix2000 NMR processing software (MSI). The 1 H, 15 N HSQC spectrum of IN-(220 -270) was assigned on the basis of the work of Lodi et al. (16) and Clore. 2 Thirty-nine well resolved peaks in the HSQC spectrum of IN-(220 -288) were assigned by comparison.
(The remaining resonances were obscured by the signals from the C-terminal tail (residues 271-288).) As a qualitative measure of the line broadening due to addition of Fab33, we determined peak heights for resolved cross-peaks in the base line-corrected HSQC spectra. Peak heights show a more linear dependence on the concentration ratio of Fab33 to IN-(220 -288) than peak volumes. Solvent-accessible surface areas were calculated using GETAREA Version 1.1 (Sealy Center for Structural Biology, University of Texas Medical Branch, Galveston, TX) (25).
Mass Spectrometry-MALDI-TOF mass spectrometric analysis was carried out as described previously (21,22). Slight modifications in experimental conditions are described in the relevant figure legends.
Size-exclusion Chromatography-The assay was performed on a Superdex 200/PC3.2 column (3.2 mm ϫ 30 cm) equilibrated with 50 mM Hepes (pH 7.5) containing 300 mM NaCl. The column was calibrated with standard proteins of known molecular masses (Bio-Rad), thyroglobulin (670 kDa), ␥-globulin (158 kDa), bovine serum albumin (68 kDa), ovalbumin (44 kDa), and myoglobin (17 kDa), using linear regression on a plot of log molecular mass versus elution time. Samples to be analyzed (each protein at 120 M) were incubated for 30 min on ice prior to injection and eluted at 0.2 ml/min. The eluted proteins were monitored by measuring the absorbance at 280 or 220 nm for IN-(220 -288).
Chemical Cross-linking-IN-(220 -288) was mixed with Fab33 and incubated in the presence of 2.5 mM glutaraldehyde cross-linker in 20 mM Hepes (pH 7.5) and 500 mM NaCl for 30 min at 25°C. Each protein was present at a final concentration of 28 M in the 1:1 ratio reaction. Reactions were terminated by quenching with 40 mM glycine for 10 min prior to addition of loading buffer for analysis by SDS-PAGE. Equal samples of each reaction were loaded on a 16% Tris/glycine gel prior to visualization by silver staining.
Modeling of mAb33-Modeling of the Fv fragment of mAb33 was performed as described previously (21) using PSI-BLAST to compare the sequences of the heavy and light chain variable domains with sequences of known structures in the Protein Data Bank (codes 1IBG and 1A7O, respectively). Sequence identity was 49% for the heavy chain and 62% for the light chain. Side chain coordinates were predicted with the SCWRL program (26).
Docking Fab33 with IN-(220 -270)-As described previously (21) the program HEX Version 2.3 (27) 3 was used to identify a reasonable structure for the complex. The calculations were started in a conformation with the antibody combining site and residues of IN-(220 -270) (Protein Data Bank code 1QMC, Ref. 18) identified as potential points of contact, pointed toward one another. The proteins were then allowed to move in a 60°arc (Ϯ30°) in each direction and Ϯ8 Å along the intermolecular axis to produce the best fit. The solvent-accessible surface area buried upon Fab⅐IN complex formation is calculated as the difference between the accessible surface area of complexed IN-(220 -270) and free IN-(220 -270).

RESULTS
Binding Properties of mAb32 and mAb33-Our previous studies showed that the activity of HIV-1 IN is stimulated by preincubation with the divalent metal ion (Mg 2ϩ or Mn 2ϩ ) required for catalysis (13). Such stimulation is correlated with a conformational change that is blocked by the binding of mAb33 to the apoenzyme. mAb33 also inhibits the processing, joining, and disintegration activities of HIV-1 IN (14). Our previous analyses also showed that mAb33 and mAb32 compete for binding to the C-terminal region of the HIV-1 IN protein (20). Furthermore, the binding of Fab32 has been mapped to specific residues in the C-terminal domain based on MALDI-TOF mass spectrometric analysis combined with timelimited proteolysis (22). To further investigate the binding of  (Fig. 1B). The results show that mAb33 bound only to fragments containing the C-terminal domain (amino acids 220 -270) and that neither the C-terminal region of the core domain (IN-(50 -212)) nor the tail of the C-terminal domain (IN-(271-288)) contributed significantly to recognition of mAb33, confirming and extending our previous observations (20). Thus, for further analysis of the mAb33binding site, we used IN-(220 -270) and the more soluble IN-(220 -288) interchangeably.
To determine whether the effects on IN enzymatic activity are related to the ability of the bivalent antibodies to cross-link the protein, we compared the binding affinities and inhibitory activities of the intact antibodies (mAb32 and mAb33) with the corresponding Fab fragments (Fab32 and Fab33). ELISA studies revealed that mAb33 and Fab33 formed stable complexes with full-length HIV-1 IN-3CS, with half-saturation concentrations of 3.0 ϫ 10 Ϫ9 and 3.0 ϫ 10 Ϫ8 M, respectively (Fig. 1C). In contrast, there was no significant difference in the binding of mAb32 and Fab32 to HIV-1 IN-3CS; half-saturation concentrations were 5.0 ϫ 10 Ϫ7 and 4.0 ϫ 10 Ϫ7 M, respectively (Fig. 1D). However, these values are ϳ10-fold higher than the dissociation constant determined for Fab33 and ϳ100-fold higher than that of mAb33. Kinetic analysis using SPR showed that Fab33 formed a stable complex with IN-(220 -270). The off-rate constant was 4.0 ϫ 10 Ϫ3 s Ϫ1 ; the on-rate constant was 5.0 ϫ 10 5 s Ϫ1 M Ϫ1 ; and the dissociation constant (K d ) was ϳ8 nM. These data are consistent with results from ELISAs shown in Fig. 1C.
Inhibition of HIV-1 IN Activity-The effects of the mAbs and Fab fragments on HIV-1 IN processing activity were tested by quantitation of the Ϫ2 cleavage (processing) product from a 21-base pair oligodeoxynucleotide duplex that represents the viral U5 DNA end. The results show that both mAb33 and Fab33 inhibited HIV-1 IN processing ( Fig. 2A) and joining (data not shown) activities. The half-inhibition concentrations (IC 50 ) of mAb33 and Fab33 were 0.18 and 1.8 M, respectively. The observation that mAb33 exhibited ϳ10-fold higher inhibitory activity than its Fab fragment is consistent with the difference observed in the binding affinities of mAb33 and Fab33 (Fig. 1C). We conclude that bivalent binding is not required for inhibition by mAb33, but may enhance the inhibitory effect. These results are consistent with our previous observations that intracellular expression of the single-chain variable region of this antibody (single-chain Fv33) in a target cell prevents HIV-1 infection (28). Thus, the in vivo activity of single-chain Fv33 is correlated with the ability of Fab33 to inhibit IN enzymatic activity in vitro. On the other hand, mAb32 had a weak inhibitory effect (IC 50 ϳ 2.0 M), and Fab32 had no detectable effect on the processing activity of IN ( Fig. 2A), although the binding affinities of these two reagents for IN were approximately equal (Fig. 1C). Thus, the inhibitory activity probably reflects the ability of the bivalent mAb32 to hold two IN molecules together in a nonproductive complex.
mAb33 Blocks DNA Binding to IN-Binding of the isolated catalytic core domain of HIV-1 IN to duplex oligodeoxynucleotide DNA substrates was not detectable in our previous studies using SPR (15). In contrast, the quantitative analysis shown in Fig. 2B indicated a dissociation constant (K d ) of 1.5 M for a duplex oligodeoxynucleotide representing the viral U5 DNA end in complex with the isolated C-terminal fragment IN-(220 -288). The same K d value was observed with IN-(220 -270) (data not shown). These results indicated that as with mAb33 binding (Fig. 1B), the C-terminal 18-amino acid tail (amino acids 271-288) made no significant contribution to the stability of the DNA⅐IN-(220 -288) complex. Furthermore, when a nonviral DNA substrate was immobilized, the K d value measured was 1.7 M, confirming that DNA sequence does not A, a linear model of the HIV-1 integrase protein and its domains recognized by four mAbs. The N-terminal domain with a helix-turn-helix structure includes the conserved residues HHCC that bind Zn 2ϩ . The catalytic core domain with polynucleotidyl transfer activity includes a highly conserved D,D(35)E motif and binds Mg 2ϩ or Mn 2ϩ . The C-terminal domain with an SH3like structure 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 (2). mAb17, mAb4, and mAb33/32 recognize N-terminal, catalytic core, and C-terminal domains, respectively (20). B, immunoblot analysis with mAb33 and truncated HIV-1 IN proteins. The assay was performed as described previously (20). Proteins (1.2 g/lane) were subjected to electrophoresis on a 12% polyacrylamide gel (with SDS) and then transferred to a nitrocellulose membrane. mAb33 (1:2000 dilution or 0.04 g/ml) was added to test its recognition of the indicated truncated proteins as determined by chemiluminescence with a Pierce Supersignal CL-HRP substrate system. C, ELISA analyses showing mAb33 and Fab33 interaction with HIV-1 IN-3CS. A high binding microtiter plate was coated with full-length IN-3CS, blocked with bovine serum albumin, and treated as described under "Experimental Procedures." Either mAb33 (Ⅺ) or Fab33 (f) was subsequently added as primary antibody, followed by a standard ELISA protocol using a secondary antibody against the -chain. D, ELISA analyses showing the interaction of mAb32 (E) and Fab32 (q) with HIV-1 IN-3CS. The experimental conditions were as described for C. affect the affinity of the C-terminal domain interactions with DNA (29 -31).
As both mAb33 and mAb32 bind to the C-terminal domain of HIV-1 IN, we investigated whether these mAbs could affect the ability of IN to bind to DNA substrates. Results of SPR measurements indicated that binding of the C-terminal domain to an immobilized DNA substrate was inhibited by preincubation of the protein with Fab33 (Fig. 2C); the binding profile for Fab33 ϩ IN-(220 -288) was indistinguishable from that of Fab33 alone. Detailed studies at various Fab33 concentrations indicated that this inhibition increased with increasing Fab33/IN ratio (data not shown). Complete inhibition was achieved at a Fab/IN-(220 -288) ratio of 1:1 (as illustrated in Fig. 2C) or higher (data not shown). However, if IN-(220 -288) was preincubated with equimolar amounts of Fab32, we observed an increase in the signal (Table I) (Fig. 3, lanes 6, 7, 9, and 10) in the presence of mAb33 and mAb32, respectively. When Fab33 was present (Fig. 3, lanes 12  and 13), most of the IN-3CS (ϳ93%) was soluble under the same conditions. Fab32 gave a similar effect as Fab33 (data not shown). These observations suggest that binding of these C   (16,17). Addition of Fab33 resulted in a major decrease in peak intensity for most cross-peaks, but negligible changes in peak position, indicating that the dynamics of binding/dissociation between IN and Fab33 are slow on the NMR time scale, as expected from the low K d and dissociation rate constant of the complex (ϳ3.0 ϫ 10 Ϫ8 M and 4.0 ϫ 10 Ϫ3 s Ϫ1 , respectively). The signals from nuclei in rigid parts of the bound IN domain are expected to be very weak because of the relatively large size of the Fab33⅐IN complex (ϳ54 kDa for a 1:1 complex or ϳ110 kDa if two molecules of Fab33 bind per IN domain dimer; see below), resulting in severe line broadening. Indeed, the number of cross-peaks observed in the HSQC spectrum of IN-(220 -288) was greatly reduced in the presence of equimolar Fab33. The remaining peaks can be attributed to mobile regions of the IN domain such as the C-terminal tail (residues 271-288) and exposed loops. When a small single-domain protein binds to a large antibody such as a Fab fragment (ϳ47 kDa), it is possible that some residues in the protein still possess substantial mobility even in the presence of Fab, so the line broadening need not be a global effect over the entire NMR spectrum. Cheetham et al. (32) showed that even for a 28-residue peptide, some cross-peaks from the peptide were still detectable upon Fab binding in double-quantum filtered COSY spectra. Thus, we identified the residues involved in the epitope of the C-terminal domain of HIV-1 IN upon Fab33 binding by quantitative analysis of the decay in peak height for each resolved peak upon "titration" of the protein with Fab33.
In (but not its backbone amide) also shows a slope Ͼ0.75 (Fig. 4A,  right panel), but this residue is in the interface of the NMR dimer (Fig. 5) (16 -18). However, the side chains of all other residues in this group (Phe 223 , Arg 224 , Tyr 226 , Lys 244 , Ile 267 , and Ile 268 ) are at least partially exposed to the solvent in the NMR dimer. The steep decay in cross-peak intensity observed for these residues is likely due to their immobilization upon binding of Fab33. Inspection of the structure of the isolated IN-(220 -270) domain shows that Phe 223 , Arg 224 , Tyr 226 , Ile 267 , and Ile 268 lie on the solvent-exposed face of the domain comprising strands 1 and 5, whereas Lys 244 is in a surface location extending from the dimer interface (Fig. 5) toward the Fab33 epitope. These 6 residues form a contiguous patch on the surface of the protein with a solvent-accessible area of ϳ650 Å 2 .
Other residues experience a more gradual decrease in peak height (Fig. 4, A and B), which can be attributed to varying degrees of local mobility. Slopes Ͻ0.3 were observed for some side chain NH groups (e.g. Gln 221 ) (Fig. 4A). In addition, slopes of approximately 0.5 or less were observed for backbone amides of some surface-exposed residues in loop regions (Ser 230 , Gly 247 , and Ser 255 ), as well as several unassigned peaks that we tentatively attribute to the C-terminal tail (residues 271-288; results not shown), all of which are expected to retain some degree of local mobility even in the presence of Fab33. However, inspection of the NMR structures of IN-(220 -270) shows that 3 residues whose backbone amide slopes are Ͻ0.5 (Trp 243 , Val 250 , and Val 259 ) are buried at the dimer interface (Fig. 5B). These residues are at the center of the tightly packed hydrophobic core of the dimeric structure and are unlikely to be mobile, unless the contact between IN monomers is perturbed by binding of Fab33. These results suggest that binding of  6 and 7), mAb32 (lanes 9 and 10), or Fab33 (10 g) (lanes 12 and 13) were incubated in 30 l of 10 mM Hepes (pH 7.5) containing 50 mM NaCl and 5% glycerol at room temperature for 15 min. The mixtures were then subjected to centrifugation at 17,000 ϫ g for 10 min. Supernatants (S) were separated from pellets (P). The protein distribution was analyzed by SDS-PAGE electrophoresis, followed by quantitation with Quantity One computer software (Bio-Rad). integration (33,34). The isolated C-terminal domain exists as a specific homodimer at concentrations of ϳ1 mM (16 -18). Mass spectrometric analysis at a lower concentration (ϳ1 M) indicated that both IN-(220 -270) and IN-(220 -288) were present as monomers and dimers (Fig. 6, A and B). To test whether these monomers and dimers reach a dynamic equilibrium, we analyzed a mixture of IN-(220 -270) and IN-(220 -288) by mass spectrometry. The results revealed that in addition to two homodimers of IN-(220 -270) and IN-(220 -288), a heterodimeric species, IN-(220 -270)⅐IN-(220 -288), was present (Fig. 6C). Thus, we conclude that a dynamic monomer-dimer equilibrium exists among molecules of the isolated C-terminal domain under these conditions.
To examine the composition of Fab33⅐IN complexes present in solution, we performed three types of analyses. In the first, Fab33/IN mixtures were subjected to mass spectrometric analysis as described above. Two additional peaks were detected in this analysis, with masses expected for 1:1 and 1:2 complexes of Fab33 and IN-(220 -270) (Fig. 7A). Under the same conditions, no such complexes were observed between a non-cognate Fab (Fab17; Fig. 1) and IN-(220 -270) (Fig. 7B), nor were complexes formed between Fab33 and a non-cognate antigen of similar size and pI (trypsin inhibitor, mass ϭ 6517 Da; data not shown.) Due to the difficulty in resolving high molecular mass complexes by this method, a 2:2 species, if present, is unlikely to be detected. However, these results indicate that Fab33 can bind to both monomers and dimers of the C-terminal domain.
The composition of Fab33⅐IN-(220 -288) complexes was next examined by SDS-PAGE analysis after cross-linking with glutaraldehyde (Fig. 8A). No bands were observed at the position expected for such complexes if either component was omitted (Fig. 8A, lanes 1 and 7, respectively), and no Fab33⅐IN-(220 -288) complexes were observed in the absence of the cross-linker (lane 2). In the presence of cross-linker, IN-(220 -288) alone formed large aggregates that migrated only to the edge of the stacking gel (Fig. 8A, lane 7). Under these experimental conditions, we expect that only a small faction of the Fab33⅐IN-(220 -288) complexes formed in the mixtures will be linked covalently (especially species that require multiple cross-links for detection on a denaturing gel). Nevertheless, the results showed that both the 1:1 and 1:2 complexes persisted when increasing concentrations of Fab33 were present; in fact, the complexes were detected most clearly at the higher Fab ratios (Fig. 8A, lanes  3-6). Thus, results from these cross-linking analyses indicate that Fab33 binding does not abrogate IN C-terminal domain multimerization.
As a final method to examine their composition, the Fab33⅐IN-(220 -288) complexes were subjected to size-exclusion chromatography (Fig. 8B). In this analysis, the complex(es) eluted at a position consistent with an apparent molecular mass of 80 kDa. This value falls between that expected for 1:1 (55 kDa) and 2:2 (110 kDa) species and could represent the average mobility of multiple complexes in dynamic equilibrium during migration through the column. Fig. 6 provides evidence for such an equilibrium among IN C-terminal domain monomers. In addition to a mixture of 1:1 and 1:2 species, complexes comprising one Fab molecule and an IN-(220 -288) oligomer would also be eluted at this position. In any case, elution earlier than the molecular mass standard of 68 kDa indicates that the Fab33⅐IN complex contains a multimer of IN-(220 -288). Taken together, these three methods of analysis show that several Fab⅐IN complexes exist in solution and that Fab33 binding has no detectable effect on the multimerization of the C-terminal domain.

Modeling of Fab33 Bound to the C-terminal Domain of IN-
Using the sequence determined for the Fab33 heavy and light chain complementarity-determining region (28) and the structure of homologous antibodies in the Protein Data Bank, a molecular model of Fab33 was constructed and then docked with the NMR structure of IN-(220 -270) (18) as described under "Experimental Procedures." The best model was obtained by starting docking after Fab33 was positioned close to the experimentally determined residues of the Fab33 epitope on IN-(220 -270). Docking initiated at other positions either did not converge on a solution or did not produce a better fit. In the model, epitope residues are neatly bound by the complementarity-determining region loops of the antibody, with each loop contributing some residues to the complex (Fig. 9B). The surfaces are complementary in shape (Fig. 9A), with no large gaps between IN and the antibody. As this is a rigid model docking procedure, this model represents initial complementarity at the interface; further changes in side chain conformation of FIG. 7. Mass spectrometry of the Fab33⅐-IN-(220 -270) complex. Prefix G t -is defined in Fig. 6. 2 M G t -IN-(220 -270) and 2 M Fab33 (A) or Fab17 (B) were incubated in 5 mM Hepes (pH 7.5) with 50 mM NaCl for 30 min at room temperature. A 0.4-l sample was mixed with 0.4 l of sinapinic acid on a target plate and air-dried before the plate was inserted into the Voyager TM MALDI-TOF BioSpectrometer. The spectrum was recorded as described in the legend to Fig. 6. The peaks at 47.11 and 23.60 kDa correspond to the singly and doubly charged Fab fragments, respectively. The peaks at 53.55 and 59.80 kDa were assigned as 1:1 (one Fab molecule/one G t -IN-(220 -270) monomer (MH ϩ )) and 1:2 (one Fab molecule/one G t -IN-(220 -270) dimer (MH ϩ )) complexes, respectively. ϳ10 l of protein sample (1 mg/ml) was injected into a pre-equilibrated column and eluted as described under "Experimental Procedures." The column was characterized with standard globular protein markers whose positions are indicated by arrows above the figure (left to right, 670, 158, 68, 44, and 17 kDa). The apparent molecular masses of mAb32 and Fab33 are consistent with known mass spectrometry-determined masses. The broad peak of IN-(220 -288) likely reflects association and dissociation of multimers, which eluted slightly sooner in the mixture with Fab33 due to some partial association with Fab early in the run, prior to migrating independently. both the IN epitope and Fab33 would be expected upon actual binding.
The residues with the steepest antibody-induced resonance intensities are all located at or near the binding interface predicted by the model. Another IN residue in this site is Arg 262 . The Arg 262 cross-peak is obscured by other cross-peaks in the crowded region of the HSQC spectrum containing residues of the disordered tail (amino acids 271-288). As with the unambiguously assigned cross-peaks already listed as part of the Fab33 epitope, the tentative assignment of the Arg 262 cross-peak shows a steep drop in relative intensity (slope of Ϫ0.78). The solution structure of IN-(220 -270) places Arg 262 adjacent to Lys 244 , with both residues positioned to directly interact with Fab33 in the model shown (Fig. 9). The buried surface area of the 1:1 complex found in this model is calculated to be 994 Å 2 , consistent with the high affinity interaction of Fab33 with IN-(220 -270) we reported here. As illustrated in Fig. 9B, Fab33 binding is unlikely to affect several residues implicated in DNA binding (Ser 230 , Arg 231 , Leu 234 , Lys 258 , Pro 261 , and Lys 264 ) either in the saddle formed by loops from each monomer or along the outer edge of each monomer (18). Although the antibody contacts some residues close to the dimer interface, it is not apparent in the model that binding of Fab33 should interfere with IN-(220 -270) dimerization. The model suggests that two Fab molecules can bind simultaneously to one IN dimer without significant steric problems. Thus, the modeled complex is fully consistent with the biochemical and NMR results. DISCUSSION Two monoclonal antibodies developed in our laboratory (mAb32 and mAb33) compete for binding to the C-terminal domain of HIV-1 IN (20), but affect its properties differently (14). Thus, these reagents provide valuable tools for probing the structure, dynamics, and function of this important enzyme. We previously mapped the binding of Fab32 to a specific region in the C-terminal domain using MALDI-TOF mass spectrometric analysis combined with time-limited proteolysis (22). Results from immunoblot studies in the present work confirmed that mAb33 also binds to residues in the C-terminal domain (amino acids 220 -270) of HIV-1 IN. Although these two monoclonal antibodies bind to the same domain of IN, the affinities of mAb32 and Fab32 for the full-length HIV-1 IN protein are 100-and 10-fold lower than those of mAb33 and Fab33, respectively. The 10-fold difference in the binding affinities of mAb33 and its Fab fragment is consistent with the 10-fold difference in the ability of the two reagents to inhibit the catalytic activities of full-length IN. On the other hand, mAb32 and Fab32 have similar binding affinities, yet only the mAb shows a weak inhibition of IN catalysis, whereas the Fab fragment has no effect. SPR analyses show that Fab33 binding inhibits protein-DNA interaction, but Fab32 does not. Thus, these results indicate that the inhibitory activity of mAb32 may result solely from cross-linking due to its bivalency, whereas inhibition by mAb33 is likely due to direct interference with enzyme function.
Our previous studies showed that mAb17 and Fab17 binding to the N-terminal domain of HIV-1 IN can enhance the solubility of full-length HIV-1 IN in low ionic strength buffer (21). Here, we show that mAb32 and mAb33 and especially their Fab derivatives can also enhance IN protein solubility; neither catalytic core-specific Fab4 nor Fab19 displayed this activity. 4 Such activity could result from masking of the hydrophobic residues in the epitope or from inhibition of oligomerization/ aggregation by antibody binding. In either case, as the Fab33⅐IN complex is considerably more soluble than IN alone, it may prove to be a useful reagent for crystallization of IN.
Using limited proteolysis in conjunction with mass spectrometric analyses, residues in the HIV-1 IN C-terminal domain that are protected by binding of Fab32 were previously localized to two strands of the ␤-sheet, ␤ 1 (Phe 223 , Arg 224 , Tyr 226 , and Arg 228 ) and ␤ 5 (Lys 264 , Lys 266 , and the side chain of Arg 262 ) (22). However, efforts to map the epitope for Fab33 using the same method were unsuccessful because Fab33 binding renders the entire C-terminal domain (residues 220 -270) resistant to proteolytic digestions (22). Thus, NMR was employed as an alternative method to identify residues that compose the epitope for Fab33. Although most of the residues in the Cterminal domain ( 270) are affected by addition of Fab33, a more detailed analysis of the backbone amide NH peak heights as a function of antibody concentration revealed substantial differences in the response to antibody binding among individual residues. Amino acids at or near the binding interface and interior residues are expected to show the most pronounced line broadening (and therefore decrease in peak height) as they are immobilized in the large Fab⅐IN complex. The steepest rate of this decrease is observed for some core residues, as well as a group of solventaccessible residues that map to a contiguous area on the surface of the C-terminal domain, including Phe 223 , Arg 224 , and Tyr 226 on ␤ 1 ; Lys 244 on ␤ 2 ; and Ile 267 and Ile 268 on ␤ 5 (Fig. 5). As these residues are largely immobilized in the complex with Fab33, they are likely to be located at or near the binding interface.
Comparison with residues protected by Fab32 (22) indicates that there is a partial overlap of amino acids in the C-terminal domain that are recognized by Fab32 and Fab33. For example, Phe 223 , Arg 224 , and Tyr 226 on ␤ 1 have been implicated in binding to both Fab32 and Fab33. This can explain the observed competition of the two antibodies for binding to the C-terminal domain. Although both antibodies interact with residues on ␤ 5 , the two epitopes differ in detail (i.e. Fab32 binding protects Lys 264 and Lys 266 , whereas Fab33 interacts with Ile 267 and Ile 268 ).
Two additional observations concerning the mobility of side chains are noteworthy. First, the majority of the side chains that are buried within the core of the C-terminal domain retain a higher degree of local mobility in the Fab33 complex compared with the residues assigned to the epitope. For example, the residues on ␤ 1 facing the solvent (Phe 223 , Arg 224 , and Tyr 226 ) are among the least mobile in the complex, whereas the adjacent core residues (Val 225 and Tyr 227 ) remain partially mobile upon complex formation (Fig. 4B). Second, the increased mobility of the backbone amides of Trp 243 , Val 250 , and Val 259 could indicate either that Fab33 binding induces dissociation of the NMR-defined dimer interface or that it induces local conformational changes that affect residues in this region. Using mass spectrometry, SDS-PAGE, and size-exclusion chromatography, we could detect no effect of Fab33 binding on the monomer-dimer equilibrium of the C-terminal domain. Thus, we conclude that the mobility changes noted in interface residues by NMR are probably due to local conformational changes induced by binding of Fab33. This interpretation is consistent with the striking loss of protease sensitivity we previously observed with the Fab33⅐IN-(220 -270) complex (22).
A number of C-terminal domain residues, including Ser 230 , Arg 231 , Leu 234 , Lys 258 , Pro 261 , Arg 262 , and Lys 264 , have been implicated in DNA binding based on mutational analyses (16,18,31,34). As illustrated in the molecular model of Fig. 9B, with the exception of Arg 262 , there is no indication that these residues are contacted directly by Fab33. Recently, Gao et al. (35) reported that substitution of Arg 262 with Cys resulted in significant loss of IN activity in vitro and that this Cys residue could be cross-linked to a DNA substrate. Thus, the Arg 262 side chain may be important for DNA binding, and Fab33 contact with this residue could account for some of the Fab interference with DNA binding. We note, however, that this same residue is protected from proteolysis by Fab32 (22). As Fab32 does not prevent DNA binding to the C-terminal domain (Table I), it is possible either that these properties are not mutually exclusive or that contact with Arg 262 is insufficient to block DNA binding. An alternative explanation for interference by Fab33 with DNA binding is that the local conformational changes induced by the antibody cause sufficient rearrangement in IN-(220 -270) to disrupt even distal DNA-binding determinants or to prevent changes that occur upon DNA binding. Additional experiments will be required to test these hypotheses.
Although it has been known for some time that the C-terminal domain participates in IN protein-protein interaction (33,34) and in sequence-independent DNA binding, its exact role in the integration reaction remains obscure. NMR studies show that the isolated C-terminal domain of HIV-1 IN forms a specific dimer in solution (16 -18), whereas inspection of the x-ray crystal structure of a fragment that includes the catalytic core and C-terminal domains suggests that a variety of proteinprotein interactions may be possible among the C-terminal domains in an active IN multimer (19). We speculate that the IN protein may have to assume different oligomeric configurations to function at specific stages in viral replication or to function in different parts of an asymmetric active complex (36) and that Fab33 binding may interfere with the protein-protein interactions that mediate these changes. Thus, further studies with Fab33 and full-length HIV-1 IN may provide additional insights into the role of the C-terminal domain in DNA binding and the conformational requirements for active multimerization. As Fab33 is a potent inhibitor of HIV-1 IN activity, residues composing its epitope may provide a novel target for the development of small molecule inhibitors. Various lines of evidence indicate that the Fab33 epitope in full-length HIV-1 IN is accessible in vivo (28,37), implying a similar accessibility to possible therapeutic agents designed to target this site.