Multimerization and DNA Binding Properties of INI1/hSNF5 and Its Functional Significance*

INI1/hSNF5/BAF47/SMARCB1 is an HIV-1 integrase (IN)-binding protein that modulates viral replication in multiple ways. A minimal IN-binding domain of INI1, S6 (amino acids 183–294), transdominantly inhibits late events, and down-modulation of INI1 stimulates early events of HIV-1 replication. INI1 both stimulates and inhibits in vitro integration depending on IN concentration. To gain further insight into its role in HIV-1 replication, we purified and biochemically characterized INI1. We found that INI1 forms multimeric structures. Deletion analysis indicated that the Rpt1 and Rpt2 motifs form the minimal multimerization domain. We isolated mutants of INI1 that are defective for multimerization using a reverse yeast two-hybrid system. Our results revealed that INI1 residues involved in multimerization overlap with IN-binding and nuclear export domains and are required for nuclear retention and co-localization with IN. Multimerization-defective mutants are also defective for mediating the transdominant effect of INI1-S6-(183–294). Furthermore, we found that INI1 is a minor groove DNA-binding protein. Although IN binding and multimerization are required for INI1-mediated inhibition, the acceptor DNA binding property of INI1 may be required for stimulation of in vitro strand transfer activities of IN. Binding of INI1 to IN results in the formation of presumably inactive high molecular weight IN-INI1 complexes, and the multimerization-defective mutant was unable to form these complexes. These results indicate that the multimerization and IN binding properties of INI1 are necessary for its ability to both inhibit integration and influence assembly and particle production, providing insights into the mechanism of INI1-mediated effects in HIV-1 replication.

HIV-1 3 replication is a dynamic process that is modulated by the interaction of several host cellular proteins (1). A genome-wide siRNA-mediated knockdown indicated that hundreds of host factors are involved in the stimulation or inhibition of HIV-1 replication (2). Understanding the interplay between the host proteins and the HIV-1 viral proteins is essential to fully comprehend the dynamic relationship between the virus and the host.
INI1/hSNF5/BAF47/SMARCB1 is a core component of the SWI/SNF chromatin-remodeling complex. It interacts directly with the HIV-1-encoded integrase (IN) required for the integration of the viral DNA into the host chromosome (3,4). IN mediates the insertion of viral cDNA into host chromosomal DNA by sequential steps of 3Ј processing and strand transfer (or joining) (4,5). INI1 binds directly to HIV-1 IN in vitro and in vivo and modulates several steps of HIV-1 replication (3, 6 -8). The ectopically expressed minimal IN-binding domain of INI1 transdominantly and potently inhibits HIV-1 assembly and particle production (8). The inhibitory effect is dependent on IN-INI1 interaction and is abrogated when an IN mutant defective for interaction with INI1 is used (8). Furthermore, particle production is minimal in cells lacking INI1, and reintroduction of INI1 into these cells can partially correct the defect (6). These results indicate that INI1 is required for HIV-1 late events. Additional studies have indicated that INI1 is selectively incorporated into HIV-1 but not other retroviral and lentiviral particles (9). Virally encapsidated INI1 is required for post-entry early events of HIV-1 replication prior to integration (6). These studies indicate that producer cell-associated as well as virionassociated INI1 is required for HIV-1 replication. Contrary to these proviral functions of INI1, siRNA-mediated knockdown studies indicate that INI1 in the target cells inhibits early events of HIV-1 replication (7). These studies indicate that whereas INI1 in the target cells may act as an antiviral host protein, HIV-1 may subvert the INI1 antiviral effect, and HIV-1 may utilize this host factor for late events in the producer cells and for early preintegration events in the target cells. Interestingly, in an earlier study, we demonstrated that partially purified INI1 both inhibits and stimulates in vitro integration in a manner dependent on IN concentration (3). Although INI1 stimulates in vitro strand transfer reactions at low IN concentrations, it inhibits the reaction at high concentrations (3). Further structure-function analysis of INI1 is required to understand this complex and dual role of INI1 during HIV-1 replication.
INI1 gene is also a tumor suppressor that is biallelically deleted in aggressive pediatric cancers known as rhabdoid tumors (10). INI1 mutations have been found in other soft tissue cancers (11)(12)(13). The mechanism of INI1-mediated tumor suppression is not fully understood. INI1 protein has two highly conserved domains that are imperfect direct repeats (termed Rpt1 and Rpt2) of each other and a third conserved coiled coil domain (termed homology region 3 or HR3) at the C terminus. The Rpt1 and Rpt2 domains appear to be involved in proteinprotein interaction with various cellular and viral proteins (3, 14 -18). Additionally, the Rpt2 domain harbors a masked nuclear export signal, and the C-terminal domain is involved in inhibiting the nuclear export of the protein in the steady state. INI1 exhibits nonspecific DNA binding activity (18). The cancer-associated mutations occur throughout the open reading frame of the INI1 gene, suggesting that mutation in any one of the INI1 domains may inactivate the protein and that multiple domains are required for its function (19 -21).
To gain further insight into the mechanism of its action, we purified INI1 protein to homogeneity and characterized it biochemically. Here we report, for the first time, that INI1 forms dimeric and higher order multimeric structures. We have characterized the multimerization domain of INI1 and found that multimerization and IN binding activities of INI1 are required for inhibition of in vitro integration. Furthermore, we found that the multimerization, IN binding, and nuclear export properties of INI1 are important for transdominant effects. In addition, we found that INI1 possesses a minor groove DNA binding activity and that the nonspecific acceptor DNA binding activity of INI1 may be required for stimulation of in vitro integration. Finally, we found that multimerization of the fulllength protein is necessary for its ability to be retained in the nucleus and to co-localize with HIV-1 IN in the nucleus. Thus, our studies provide novel insights into the mechanism by which INI1 regulates HIV-1 replication. The plasmids pGEX-INI1, pGEX-IN, pSH 2 -INI1,  and pGADNotINI1 and deletion fragments pGADNotINI1-(183-294), pSH 2 IN, pBABEpuro-INI1, pCGNINI1, and pQE32-INI1 have been described previously (18). Details of the construction of pQE32-INI1-(141-304) and generation of the random mutagenesis library of pGADNotINI1-(183-294) are as described in the supplemental material. CFP-INI1, DD5, and DD6 constructs were generated by PCR amplification and subcloning of the subsequent fragments at the EcoRI-BamHI sites of pECFP vector (Clontech).

Plasmids-
Purification of Proteins-His 6 -INI1 and His 6 -INI1-(183-294) were purified to homogeneity as described in the supplemental material. His 6 -LEDGF was purified as reported (27) except that His 6 -LEDGF was not digested with PreScission protease, and two purification steps, viz. Ni-NTA and Mono-S Sepharose, were used. The partially purified protein was dialyzed against IN dialysis buffer (3). C-terminal His-tagged IN was purified as described (3).
Yeast Two-hybrid Analysis and GST Pulldown Assays-These tests were performed as described, and details are provided in the supplemental "Materials and Methods" (3).
Glycerol Gradient Centrifugation and Gel Filtration Chromatography-The details of the methods used are provided in the supplemental "Materials and Methods." Briefly, Mono-Q Sepharose eluate of His-INI1 was separated on a 15-35% glycerol gradient at 40,000 rpm (SW41 Ti rotor) for 24 h. Fractions (400 -450 l) were collected and subjected to trichloroacetic acid precipitation. The precipitated proteins were subjected to Western blot analysis using an anti-His monoclonal antibody as probe. To compare the oligomeric status of wild type and mutant INI1 proteins, the protein was incubated in 500 l of the buffer for 3 h on ice and then subjected to glycerol gradient centrifugation as described above.
Gel filtration chromatography was performed in 20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM dithiothreitol, with 200 g of His-INI1-(183-294) eluate from the Mono-Q Sepharose column. Fractions (500 l) were collected and subjected to Western blot analysis using anti-His antibody as probe.
Joining Assay, DNA Binding Assay, and EMSA-The joining assays were performed with 32 P-labeled 3Ј preprocessed duplex DNA (U5.5 and U5.4) as donor DNA and pcDNA as target DNA as described previously (3) using the indicated amounts of proteins except that ϳ7 ng of radiolabeled substrate and 100 ng of target DNA were used. Salt concentrations were varied between 10 and 150 mM NaCl.
For agarose-based gel retardation assay about 8 and 16 pmol of hydroxylapatite eluate of the wild type INI1 or mutant INI1 were incubated with 200 ng of pcDNA in buffer containing 20 mM HEPES-KOH (pH 6.8), 100 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 3 mM MnCl 2 , and protease inhibitors at 30°C for 1 h. Protein-DNA complexes were resolved by 1% agarose gel electrophoresis and stained with ethidium bromide.
For EMSA ϳ7 ng of radiolabeled 3Ј preprocessed duplex DNA (U5.5 and U5.4) was incubated with 5 pmol of hydroxylapatite eluate of wild type INI1 in strand transfer reaction buffer and protease inhibitors at 37°C for 10 min. 10ϫ cold viral LTR DNA was added to the reaction mixture. Minor groove inhibitors were added to the reaction mixture as indicated. About 10 l of reaction mixture was run on a 6% native polyacrylamide gel made in Tris borate-EDTA, and the gel was run at 10 mA for 2 h. The gel was dried and subjected to autoradiography.

Purified
During the course of these studies, we observed that INI1 forms a higher order structure and has a tendency to aggregate at high concentration. This and additional observations using in vitro and in vivo binding studies suggested that INI1 is a multimer, as detailed below. To determine whether INI1 exists as a multimer in solution, purified INI1 eluted from the Mono-Q Sepharose column was subjected to glycerol gradient centrifugation. Purified marker proteins were run as a control to estimate the apparent molecular weight relative to the fractions. At low concentrations (Ͻ5 nM), purified INI1 fractionated in a single peak after BSA (67 kDa) in glycerol gradients (Fig. 1E). However, at higher protein concentrations (Ͼ100 nM), INI1 migrated in multiple peaks (Fig. 1F). The first peak of INI1 was observed just after the 79-kDa marker, indicating that it is a dimer. Furthermore, other peaks of INI1 migrated with approximate molecular masses indicative of tetramers and octamers and higher order forms (Fig. 1F). Taken together, these studies demonstrated that recombinant INI1 is a dimer in solution at low concentrations but forms higher order structures at high concentrations.
To further confirm that INI1 self-associates, we carried out other in vitro and in vivo interaction assays. A GST pulldown assay was carried out using GST-INI1 and His 6 -INI1 expressed in bacteria. The bacterial lysates expressing His 6 -INI1 were treated with DNase I, and the binding reaction was carried out in the presence of ethidium bromide to avoid DNA-mediated association (22). In this assay, although the control GST bound to beads failed to pull down His-INI1, GST-INI1 was able to interact robustly with His-INI1, confirming that INI1 interacts with itself in vitro (  bodies against both the N-terminal His epitope tag and the C terminus of INI1 recognized the two polypeptides in the purified INI1-(141-304) preparation (Fig. 2C, middle and right panels), and (ii) mass spectrometric analysis revealed a single polypeptide species in purified preparations (data not shown). We found that the faster migrating polypeptide was an artifact of detergent solubilization, as it was absent when the protein was purified in the absence of IGEPAL, and the single band stained positively with INI1 antibody (Fig. 2D and data not shown). These results indicated that we were able to purify the deletion fragment to apparent homogeneity.
To investigate the multimeric form of INI1-(141-304) we subjected the purified protein to gel filtration chromatography using a Superdex 200 HR 10/30 column. INI1-(141-304) eluted from the column soon after ovalbumin (43 kDa) and before chymotrypsinogen A (25 kDa), demonstrating that it is a dimer (ϳ42 kDa) (Fig. 2E). Western blot analysis of fractions eluted from the gel filtration column using an antibody against the His epitope tag confirmed that the eluted protein was INI1-(141-304) (Fig. 2E).
Isolation and Characterization of Multimerization-defective Mutants of INI1-To investigate the functional significance of multimerization of INI1 in HIV-1 replication, we isolated a panel of mutants of the minimal multimerization domain of INI1, S6 (amino acids 183-294), using a reverse yeast two-hybrid system. A random mutation library of INI1-S6-(183-294) fused to GAL4AD (8) was screened against LexADB-INI1 to isolate a panel of multimerization-defective mutants with one or two amino acid substitutions (Table 1). These mutations were in either the Rpt1 or the Rpt2 motif and were mostly hydrophobic (Fig. 3A). These results suggested that hydrophobic interactions are likely to be important for INI1 selfassociation. Furthermore, among the residues mutated, amino acids Phe-204, Trp-206, Phe-233, Ile-237, Ile-263, Ile-264, and Ile-268 were partially or fully conserved phylogenetically (data not shown). In addition, several clusters of mutations were observed in Rpt1, and mutations in the Rpt2 motif overlapped with the nuclear export signal (NES) of INI1 (Fig. 3A) defective mutants were more frequent in the Rpt1 domain, multimerization-defective mutants were found in both the Rpt1 and Rpt2 regions.
To further investigate the relationship between IN binding and multimerization, we analyzed two panels of mutants of S6 INI-(183-294) for their ability to bind to both INI1 and IN, using a yeast two-hybrid system scoring for ␤-galactosidase activity (Fig. 3, Table 1). The first panel consisted of mutants isolated as defective for binding to INI1 using the reverse twohybrid system. The second panel included mutants of S6 INI1-(183-294) that were previously isolated as defective for IN binding (8). We found that the IN binding activity of S6 was stronger than its ability to bind to INI1 (Fig. 3, B-F). In general, the mutants that were defective for multimerization (DD1-DD6) were also defective for IN binding (Fig. 3, B and C). However, there was an inverse correlation between the residual multimerization and IN binding activities of these mutants. The mutants DD1(F233L), DD2(F204S), DD3(V234A,I237T), and DD4(I195T,W206R), which retained partial multimerization (ϳ30% when compared with that of control INI1-(183-294), were completely defective for binding to IN (compare B and C in Fig. 3). Conversely, the mutants DD5 (I263T) and DD6 (I264T,I268T), which harbored mutations in Rpt2 that were completely defective for INI1 binding, retained a partial ability to bind to IN (ϳ10%) ( Table 1; Fig. 3, B and C).
Similarly, in the reciprocal experiment in which IN bindingdefective mutants were tested for INI1 binding, we found that many of the S6 mutants retained their ability to bind to INI1 ( Table 2; Fig. 3, E and F). Although the mutants E4 (T214A) and E6 (E183G,D191G) were able to bind INI1 efficiently, the mutant E3 (D225G) bound to INI1 partially, and the mutant E5 (V185A,I264T) was defective (10% activity compared with wild type) in binding to S6 (Fig. 3, E and F; Table 2). These results indicated that although the multimerization domains overlapped with IN-binding domain, there were clear differences between the two interactions. Amino acid residues Glu-183, Asp-191, and Thr-214 in the Rpt1 motif, which are not involved in multimerization, were critical for IN binding. These results indicated that specific residues in the Rpt1 region are necessary for IN binding and are distinct from those required for multimerization. Western blot analysis of yeast extracts containing wild type or mutant INI1-(183-294) proteins using an anti-GAL4AD antibody showed that the proteins accumulated stably in yeast cells (Fig. 3, D and G).
Effect of Multimerization on Nuclear Export of S6 INI1-(183-294) and on Inhibition of HIV-1 Particle Production-Previously, we demonstrated that INI1-S6-(183-294) harbor-ing the minimal IN-binding domain potently inhibits HIV-1 particle production (8). This fragment is ectopically expressed due to the unmasking of an NES in INI1 (23). In many transcription factors, the nuclear export domains often overlap with multimerization domains (24). Similarly, we observed that the multimerization-defective mutations found in the Rpt2 region of INI1-(183-294)/S6 overlap with the NES region. The NES of INI1 resides within amino acids 263-276, and the hydrophobic residues within this region that match the NES consensus are essential for nuclear export (Ref. 23 and Fig. 4A). We noted that the residue mutated in DD5 (I263T), was one of the five hydrophobic residues essential for nuclear export within this region ( Fig. 4A and Ref. 23). On the other hand, the residues mutated in DD6 (I264T,I268T) fell outside the NES consensus sequence (Fig. 4A and Ref. 23). Therefore, we hypothesized that mutation in DD5, but not in DD6, may disrupt nuclear export. The two mutants DD5 (I263T) and DD6 (I264T,I268T), were expressed as GFP fusion proteins in HeLa cells, and the subcellular localization was determined via confocal microscopy. Although GFP-INI1-(183-294)/S6 localized primarily to the cytoplasm (Fig. 4B), the DD5 mutant exhibited reduced cytoplasmic localization and was nuclear in the majority of cells (59%) (Fig. 4, B and C); the DD6 mutant exhibited cytoplasmic localization in the majority of the cells similar to that of S6 (Fig. 4, B and C). These results confirmed our previous observations of nuclear export of INI1 and demonstrated that some residues required for multimerization and nuclear export overlap.
To determine the contribution of the multimerization and/or nuclear export properties of INI1-S6-(183-294) for its ability to inhibit HIV-1 particle production, the mutants were co-expressed along with HIV-1 viral vectors in 293T cells, and particle production was assessed by measuring the intracellular and virion-associated p24 levels by enzyme-linked immunosorbent assay (Fig. 4D). These results indicated that INI1-(183-294) robustly inhibited virion-associated p24 levels, consistent with previous reports (Fig. 4D, top right panel, and Ref. 8). This decrease was accompanied with a decrease in intracellular p24 levels (Fig. 4D, top left panel). In addition, we found that there was a decrease in the ratio of virion associated to intracellular p24 levels (Fig. 4D, bottom panel), suggesting that S6 inhibits viral particle production by both decreasing the intracellular p24 and by further affecting particle release. Interestingly, the mutations in DD5 (I263T) and DD6 (I264T,I268T) partially rescued the inhibition of particle production (Fig. 4D, top and  bottom panels). Wild type and mutant proteins were expressed at similar levels, indicating that rescue in particle production is not due to a defect in expression (Fig. 4E). These results demonstrate that a defect in the multimerization of DD5 and DD6 and a defect in the nuclear export properties of DD5 significantly abrogated their ability to mediate the dominant negative effect.
Because multimerization and IN-binding residues were overlapping, we investigated whether DD5 and DD6 were defective for binding to IN in vivo. We tested the ability of DD5 and DD6 to bind to YFP-IN in 293T cells by coimmunoprecipitation assay. Although HA-tagged INI1-(183-294)/S6 robustly associated with YFP-IN (Fig. 4F), DD5 (I263T) showed an ϳ2-fold reduction and DD6 (I264T,I268T) demonstrated a 10-fold reduction in YFP-IN binding activity as compared with the wild type fragment (Fig. 4F). These results suggested that multimerization, IN binding, and cytoplasmic localization were necessary for INI1-(183-294)/S6 to inhibit viral particle production. These studies, taken together, implicate a role for INI1 in maintaining viral protein levels and in the assembly/release of viral particles in the cytoplasm of the producer cells.

Co-localization of IN with INI1-
To determine whether the mutations that affect properties of the truncated version of INI1/S6 also affect the full-length protein, the mutations DD5 (I263T) and DD6 (I264T,I268T) were introduced into full-length INI1 by site-directed mutagenesis. To test the ability of IN to co-localize with INI1 and its mutants, INI1, S6, and full-length DD5 and DD6 mutants were expressed as fusion to cyan fluorescent protein, and IN was expressed as fusion to yellow fluorescent protein. We had demonstrated previously that INI1 is primarily nuclear and that the truncated version of the protein (S6) is cytoplasmic (Ref. 8 and Figs. 4 and 5). Here, we first tested the localization properties of DD5 and DD6. As expected, DD5 mutants, within the context of fulllength protein, localized primarily to the nuclear compartment with some diffuse localization (Fig. 5A). Interestingly, the DD6 mutant within the context of the full-length protein, with an intact NES, was able to localize to the cytoplasm (Fig. 5A). These results confirm the presence of NES in INI1 and further suggest that multimerization may be important for keeping the protein in the nucleus. We then tested the co-localization of IN with various proteins. IN is primarily nuclear. We found that when IN was co-expressed with INI1, it tightly co-localized with INI1 (Fig.  5B). Furthermore, as long as INI1 mutant proteins retained the ability to bind to IN, they appeared to dictate the subcellular localization of IN (Fig. 5B). For example, expression of S6 (a cytoplasmic mutant of INI1) resulted in cytoplasmic localization of IN (Fig. 5B). IN also colocalized with the DD5 mutant, as it retained the ability to bind to IN (Fig. 5B). Interestingly, IN only partially co-localized with the DD6 mutants (Fig. 5, B and C). These results are consistent with our observation that the DD6 mutant of INI1 is partially defective for binding to IN (Fig. 4F). Together, these results are consistent with our findings that IN and INI1 co-localize in the nucleus in vivo and that this co-localization is dependent on INI1 multimerization and IN binding.  lapatite columns. These proteins were tested for their ability to multimerize, bind to IN, and modulate in vitro strand transfer reactions.

Modulation of in Vitro Integration by INI1 Is Mediated by Its Ability to Multimerize and Bind to IN-To
To determine the multimerization status, the purified proteins were subjected to glycerol gradient centrifugation at a low concentration (ϳ5 nM) to separate various oligomeric forms of INI1. Although wild type INI1 migrated as a dimer and a higher molecular weight species, the mutant DD5 (I263T) was partially defective for multimerization forming both monomeric and dimeric and a higher molecular weight species (Fig. 6A). The DD6A and DD6C mutants with I264T and I268T substitutions, respectively, were defective in multimerization. Although I264T formed less than 10% dimers, I268T formed only monomers (Fig. 6A). Furthermore, the mutant DD6 (I264T,I268T) formed only monomers at this concentration. These results indicate that these mutants indeed are defective for multimerization.
We next tested the ability of these mutants to bind to INI1 and IN in a GST pulldown assay (Fig. 6, B and C). As a control, we also tested D225G (a mutant originally isolated as IN binding-defective S6) within the context of full-length INI1 (8 6D), although the mutant I268T exhibited reduced (ϳ50%) activity (Fig. 6D). Similar results were obtained at 150 mM salt, although the overall efficiency of integration activity was reduced (data not shown). We then tested the mutants for their ability to inhibit IN joining activities at high IN concentrations. The results indicated that although wild type INI1 showed robust inhibition, the DD6 (I264T,I268T) mutant, which is most defective for both multimerization and IN binding, was unable to inhibit integrase activity (Fig. 6E). These results were in contrast to the lack of effect of DD6 mutant on the stimulation of IN activity. These studies taken together demonstrate that the ability of INI1 to stimulate IN joining activity is independent of multimerization and that IN binding and the ability to inhibit strand transfer activity are dependent on these protein-protein interactions. Purified IN and INI1 proteins were mixed together at a molar ratio of 1:4 and then subjected to glycerol gradient centrifugation. As a control, IN alone was also subjected to glycerol gradient centrifugation. The results of these experiments indicated that IN separated as monomeric, dimeric, and higher molecular weight forms upon glycerol gradient centrifugation (Fig. 7A, top  panel). As indicated previously, at these concentrations, INI1 also forms dimers (refer to Fig. 1E). However, when IN was mixed with INI1, lower order forms of these proteins were either reduced (as for IN, Fig. 7A, second panel) or completely eliminated (as for INI1, third panel). These results suggested that the interaction of IN and INI1 may result in a presumably inactive, large, multimeric, high molecular weight complex that sediments at the bottom of the gradient (Fig. 7A) binding-deficient mutant of  INI1 DD6 (I264T, I268T), which was most defective for the inhibition of joining activities, was not defective for stimulation of IN joining activities in vitro (Fig. 6, D and  E). These results suggest either that the residual IN binding activity of these mutants is sufficient to stimulate integration reaction or that the ability of INI1 to stimulate IN joining activity is due to some other function of INI1. Previously, we had demonstrated that INI1 has the ability to bind to DNA nonspecifically (18). We surmised that the nonspecific DNA binding activities of INI1 could modify the DNA such that it would be more accessible for IN-mediated strand transfer activities and that the mutants defective for proteinprotein interactions may not be defective for DNA binding activities. To investigate this further, we tested the DNA binding properties of wild type and mutant INI1 proteins at two different protein concentrations (8 and 16 pmol). The wild type protein bound to plasmid DNA, as indicated by the lack of migration of DNA into the gel from the wells (Fig. 7B, lanes  2 and 3). As a control, the addition of BSA had no effect on the DNA migration (Fig. 7B, lane 1). When INI1 was depleted from the extract (Fig. 7B, lane 15), there was a lack of DNA binding activity in the extract, as indicated by the appearance of DNA migration into the gel, similar to that of the control BSA (Fig. 7B, compare lanes 1 and 15), suggesting that DNA binding activity in the extract was due to INI1. Interestingly, we found that all of the mutants except mutant I268T were able to retard the DNA similar to findings in the wild type protein (Fig. 7B). The mutant I268T exhibited a partial defect in DNA binding, in that although the plasmid DNA entered the gel in the presence of this mutant, it was retarded to some extent (Fig. 7B, lane 8).
These results are consistent with our hypothesis that the DNA binding activities of INI1 may be required to stimulate IN strand transfer reactions.
IN has the ability to bind to both the target and LTR DNAs. Therefore, we tested whether INI1 would bind to the radiolabeled LTR donor in an EMSA (Fig. 7C). The binding of INI1 to radiolabeled LTR DNA was evident in the EMSA, which was competed by an excess of nonspecific cold scrambled DNA, indicating that the binding was not sequence-specific (Fig. 7C, lanes 2  and 3). When the protein preparation was depleted of INI1, the gel shift was not observed, indicating that the mobility shift observed was due to the activity of INI1 protein (Fig. 7C, lane 4). To further characterize the DNA binding property of INI1, we tested to see whether it binds to the minor groove of DNA using DNA intercalating agents such as distamycin A and DAPI. These drugs specifically disrupt the binding of proteins to the minor groove of DNA (25,26). For example, tau, a microtubule-associated protein that is a well known DNA minor groove binder, binds a 13-mer DNA with a protein:DNA molar ratio of 1:1 (26), whereas in our experiments INI1 bound a 23-mer DNA with a protein:DNA molar ratio of ϳ30:1. One explanation for this effect is that INI1 is a higher order multimer. Distamycin A and DAPI inhibit a known DNA minor groove-binding transcription factor, hTBP, with an IC 50 of ϳ1 M but do not inhibit major groove-binding proteins like EGR1 and WT1 (25). When distamycin A and DAPI were added to the reaction containing radiolabeled LTR and INI1 protein, we found that they strongly inhibited the DNA binding activity of INI1, with the drugs at 0.1 M inhibiting more than half of the INI1 DNA binding activity (Fig. 7D). These results established, for the first time, that INI1 binds to the minor groove of DNA. The binding of tau to the minor groove is inhibited by distamycin A at an inhibitor:protein ratio of 10:1 (26), whereas distamycin A and DAPI maximally inhibited INI1 binding to DNA at an inhibitor: protein ratio of 7:1 and 5:1, respectively.
We then tested to determine whether the multimerization and IN-binding mutants of INI1 were defective for LTR binding activity. We found that all of the mutants retained the ability to bind to LTR under the current conditions (Fig. 7E). Although there was some variability in the degree of LTR DNA binding, this variability was not significant in repeated experiments. Overall these results indicated that wild type and mutant INI1 proteins are able to bind to the minor groove of LTR. This result is slightly different from the result that we observed for binding of INI1 and the I268T mutant to plasmid DNA. We found that whereas the mutant I268T was somewhat defective for binding to plasmid DNA, it was not defective for binding to LTR DNA. This difference can be explained by a difference in the binding affinity of the protein to the topologically complex plasmid DNA as compared with the 23-mer LTR DNA. These results taken together indicate that differential ability of the wild type and mutant I268T to bind to plasmid DNA correlate with the ability of INI1 to stimulate IN joining activity.

DISCUSSION
INI1 was the first host protein to be identified as a binding partner for HIV-1 IN. However, the role of this protein in HIV-1 replication is not completely understood, due in part to a lack of biochemical and structural information relating to this protein. In this report, we have attempted to purify and characterize INI1 biochemically. We report here, for the first time, that INI1 forms dimeric, tetrameric, octameric, and higher order structures depending on the concentrations of the protein. We have shown that the Rpt1 and Rpt2 regions of INI1 are necessary for multimerization, and the N-terminal domain of INI1, although not necessary, is required for efficient self-association. It is interesting to note that the purified fragment S7-(141-304), which harbors homomeric interaction domains, is a dimer in solution even at higher concentrations. Therefore, it is possible that although the minimal domain needed for dimerization includes the Rpt1 and Rpt2 regions, the N-terminal domain may contribute to higher order interactions leading to multimerization. Future structural information will likely shed light on these properties of INI1.
Knowledge of the biochemical properties of INI1 is essential to comprehend the role that this protein has in HIV-1 replication. As the multimerization property may be necessary to exhibit its effect, we have isolated mutants of INI1 that are defective for this function. Our effort yielded mutants in the Rpt1 and Rpt2 motifs. Most of the mutated residues were partially or fully conserved and hydrophobic in nature, suggesting that hydrophobic interaction may be critical for self-association. Interestingly, a majority of the mutations in the Rpt2 motif are concentrated near the N terminus of the NES or overlap with it. Comparative analysis of the self-association of multimerization-defective mutants indicated that whereas Rpt1 mutants are ϳ30% as active as wild type INI1-(183-294), Rpt2 mutants are only Ͻ5% as active as the wild type protein, suggesting that amino acid residues in the Rpt2 motif are critical for multimerization. This is contrary to findings in IN bindingdefective mutations of INI1- (183-294), where the majority of the mutations were found in Rpt1. Although the mutants that were found to be defective in self-association were also defective in binding to IN, the degree of the defect was inversely correlated. For example, among the multimerization-defective mutants analyzed, DD5 (I263T) and DD6 (I264T,I268T) mutants were most defective for multimerization and were least defective for IN binding. Furthermore, many of the Rpt1 mutants (T214A,D225G and E183G,D191G) that were defective for binding IN were not defective for multimerization. These results suggest that the two functions of INI1, i.e. multimerization and IN binding, are mediated by overlapping regions but require distinct residues. We suggest that the amino acid residues in the Rpt1 motif, in addition to playing a role in multimerization, may make direct contact with IN. This is consistent with our previous report that there is direct association between the Rpt1 motif of INI1 and integrase (18). Finally, the observation that multimerization of INI1 is required for its association with IN was reproduced with the full-length INI1 protein.
Because the residues involved in multimerization overlapped with the NES sequence, we also investigated the nuclear export ability of these mutants. Our previous studies had identified a NES consensus sequence in INI1 between residues 263 and 275. We reported that in addition to the residues that exactly match the known consensus for NES (265 ⌿XXX⌿XX⌿X⌿), an additional upstream residue, Ile-263, is required for nuclear export (23). Our current results strengthen these previous results in that the mutant DD5 (I263T) with a mutation in a single residue in the NES consensus was defective for nuclear export, and the mutant DD6 (I264T,I268T), which was mutated in hydrophobic residues outside the NES consensus sequence, was not defective for nuclear export. These results suggest that multimerization and nuclear export are mediated by overlapping residues in INI1. We utilized a pair of mutants that are differentially defective for nuclear export, but commonly defective for multimerization, to assess the contribution of these properties to mediate transdominant inhibition of HIV-1 replication. Our results indicate that both multimerization and nuclear export are necessary for mediating the transdominant effect. Interestingly, we found that expression of the transdominant negative mutant not only inhibited particle production, but it also significantly decreases intracellular p24 protein levels. This latter effect could either be due to a decrease in protein expression itself or to a decrease in the overall stability of Gag and Gag-Pol proteins. Because the two mutants that were defective for either multimerization alone or for both multimerization and IN binding partially rescued the defect in p24 levels, we believe that these functions are important for both p24 protein expression/stability as well as particle production. More experiments are needed to address the effect of these mutants on intracellular viral protein levels. Nevertheless, our results demonstrate the intricate relationship among nuclear export, multimerization, and IN binding, due to the overlapping of the domains responsible for these functions.
To determine the influence of multimerization and/or IN binding on in vitro integrase activity, we tested the ability of the mutants I263T, I264T, I268T, D225G, and I264T,I268T to modulate integrase activity. Overall, we found that multimerization and the ability to bind integrase were critical for INI1mediated inhibition of integrase activity, whereas the ability of INI1 to stimulate integrase activity was independent of these functions. At this point, we cannot rule out the possibility that the residual IN binding properties of the INI1 mutants are sufficient for stimulation function. Interestingly, the mutant I268T was partially defective in its ability to stimulate integrase activity, although it retained IN binding. We surmised that the other possible function of INI1, such as nonspecific DNA binding activity, could be important in the strand transfer reaction. Therefore, we tested the mutants for their ability to bind to both the donor oligo duplex derived from LTR and the acceptor plasmid DNA. Consistent with previous results, we found that INI1 robustly bound to the acceptor plasmid. Among the mutants, only I268T demonstrated a partial defect in DNA binding in an agarose gel retardation assay, consistent with the idea that acceptor DNA binding by INI1 is critical for its ability to stimulate integrase function in the strand transfer reaction. However, it should be noted here that even the most defective IN mutant exhibited residual IN binding activity, and INI1-integrase interaction may still be necessary for INI1 stimulatory activity. Thus, the residual binding of the DD6 (I264T,I268T) to integrase may be sufficient for its ability to stimulate but not to inhibit strand transfer activities.
In this report we have demonstrated that INI1 has the ability to bind to the minor groove of double-stranded DNA. The addition of purified INI1 protein resulted in gel shift of LTR oligonucleotides, which was abrogated when the INI1 protein was depleted from the preparation. Furthermore, distamycin A and DAPI, the two DNA intercalating agents that prevent the interaction of proteins to the minor groove, disrupted the interaction of INI1 with LTR DNA. This property of INI1 could also explain the ability of INI1 to stimulate the joining activity. However, this property alone is not a sufficient explanation, as all of the mutants, including the one that was slightly defective for IN joining activities, bound to LTR DNA. We propose that the ability of INI1 to bind acceptor DNA is required for its ability to stimulate IN joining activities. At this time, we cannot rule out the possibility that INI1 binding to LTR DNA could also be important for its stimulatory activity in vitro. These results may help define the mechanism of INI1-dependent modulation of HIV-1 integrase activity. These studies allow us to assess the role of INI1 in HIV-1 replication, which may ultimately lead to the development of novel therapeutic strategies.