Allosteric Inhibition of Human Immunodeficiency Virus Integrase

Background: New antiviral agents bind to a site on HIV integrase protein also bound by the cellular protein LEDGF/p75. Results: Compound GSK1264 binds to this site, but it has surprising properties; it inhibits late during HIV replication, not early during integration, and it promotes abnormal multimerization. Conclusion: GSK1264 provides new insight into HIV replication. Significance: These observations inform the design of improved antiviral agents. HIV-1 replication in the presence of antiviral agents results in evolution of drug-resistant variants, motivating the search for additional drug classes. Here we report studies of GSK1264, which was identified as a compound that disrupts the interaction between HIV-1 integrase (IN) and the cellular factor lens epithelium-derived growth factor (LEDGF)/p75. GSK1264 displayed potent antiviral activity and was found to bind at the site occupied by LEDGF/p75 on IN by x-ray crystallography. Assays of HIV replication in the presence of GSK1264 showed only modest inhibition of the early infection steps and little effect on integration targeting, which is guided by the LEDGF/p75·IN interaction. In contrast, inhibition of late replication steps was more potent. Particle production was normal, but particles showed reduced infectivity. GSK1264 promoted aggregation of IN and preformed LEDGF/p75·IN complexes, suggesting a mechanism of inhibition. LEDGF/p75 was not displaced from IN during aggregation, indicating trapping of LEDGF/p75 in aggregates. Aggregation assays with truncated IN variants revealed that a construct with catalytic and C-terminal domains of IN only formed an open polymer associated with efficient drug-induced aggregation. These data suggest that the allosteric inhibitors of IN are promising antiviral agents and provide new information on their mechanism of action.

Early steps of HIV-1 replication involve reverse transcription, which produces a double-stranded cDNA copy of the viral RNA genome, and integration, which results in the incorporation of the viral DNA into host chromosomal DNA (1). Both of these steps have been targeted by clinically useful inhibitors (2,3). The integrase strand transfer inhibitors bind to the active site of the virus-encoded integrase (IN) 4 enzyme and block the initial strand transfer step that incorporates viral DNA into the host chromosome. In this report, we describe the properties of a small molecule, GSK1264, that functions as an allosteric inhibitor of IN and inhibits late during the viral replication cycle.
In infected cells, HIV-1 IN binds the cellular host factor, LEDGF/p75, which is a transcriptional co-activator that is the product of the PSIP1 gene (Fig. 1A). Following reverse transcription, IN binds the double-stranded viral DNA ends to form the intasome. LEDGF/p75 primarily binds to the catalytic core domain (CCD) and, to a lesser extent, the N-terminal domain (NTD) of integrase (4) via the LEDGF/p75 C-terminal integrase binding domain (IBD) (5). This interaction promotes efficient integration and the targeting of the intasome to active transcription units (6 -8). LEDGF/p75 acts as a simple tether, as shown by experiments in which the LEDGF/p75 chromatin binding domain is substituted with new chromatin binding domains, resulting in retargeting of HIV-1 integration (9 -11). Additionally, the LEDGF/p75⅐IN interaction protects IN from proteolysis and stimulates catalysis both in vitro and in vivo (4,(12)(13)(14)(15)(16).
The structure of the LEDGF/p75 IBD⅐IN(CCD) complex was determined by Cherepanov and colleagues (17,18). The IBD binds at the interface of the IN(CCD) dimer, occupying a pocket at the interface. Earlier work has demonstrated that this site can be bound by small molecules (19), and subsequent screening efforts have yielded several small molecule classes that interfere with binding of the LEDGF/p75 IBD and display antiviral activity (reviewed in Refs. 20 and 21).
Here we present a detailed study of one such molecule, GSK1264 (Fig. 1B). Co-crystallization with the IN(CCD) showed that GSK1264 indeed bound the LEDGF/p75 binding site. GSK1264 inhibited most potently in the late part of the viral replication cycle, after integration. As this work was being completed, several other groups made similar findings with structurally related inhibitors (20,(22)(23)(24)(25)(26). Potency of GSK1264 early during infection was only modest, and there was little effect of GSK1264 on integration target site selection. In the presence of GSK1264, viral particles were produced normally late during infection, but they were reduced in infectivity. Potent inhibition correlated with increased oligomerization of IN in the presence of drug in vitro. Drug-induced oligomerization of preformed LEDGF⅐IN complexes was not associated with detectable displacement of LEDGF/p75. Study of truncated derivatives of IN in vitro demonstrated that drug-induced polymerization was most potent in variants containing the CCD and CTD only. Thus, compounds that bind the LEDGF/p75 site on IN are effective inhibitors whose primary effects occur at the latest steps of replication, and inhibition correlates with abnormal IN polymerization involving specific protein domains.

EXPERIMENTAL PROCEDURES
Cell Lines-The TZM-bl, 293T, and U373/CD4/CCR5 (27) cell lines were obtained through the National Institutes of Health AIDS Research and Reference Reagent Program (ARRRP) and grown as directed (28). A1953 chronic HIV producer cells were a gift from James Hoxie.
HIV-1 Infection and Integration Target Site Analysis-Infections were carried out in TZM-bl cells using standard methods and the HIV-1 strain HIV 89.6 (29). Analysis of HIV-1 integration targeting was carried out as described previously (6, 30 -32). All sites common among samples (including the reporter construct in the TZM-bl cells) were removed prior to analysis.
For the study of LEDGF/p75 knockdown cells, an shRNA construct (Sigma-Aldrich, TRCN0000074819) was transduced into a 293T-derived cell line, and cells were subjected to puromycin selection (1 g/ml), yielding KD19 cells. In parallel, a matched construct encoding a GFP-targeting shRNA was introduced into the 293T cell line and compared. Knockdown was confirmed to reduce LEDGF/p75 mRNA levels by 92%, and protein was undetectable by Western blot analysis.
The detergent was confirmed to be at submicellar concentrations at this ionic strength (450 mM NaCl) using both a colorimetric assay and small-angle x-ray scattering (SAXS) analysis (34, 35) (data not shown). It has been reported that detergents such as CHAPS can attenuate IN oligomerization (36). Thus, this model system provides a hypo-oligomeric background against which drug-induced multimerization is measured.
IN-LEDGF FRET Binding Assay (48)-GSK1264 was dissolved in DMSO to 10 mM and serially diluted in DMSO for assays. Reactions were performed at a 10-l final volume in a 384-well plate format (Greiner Bio-One, San Diego, CA). The reaction buffer contained 50 mM HEPES, pH 7.5, 150 mM NaCl, 20 mM MgCl 2 , 0.1 mg/ml bovine serum albumin (BSA), 50 M CHAPS, and fresh 2 mM DTT. After the addition of drug in a 0.1-l volume, hexahistidine-tagged IN F185K (CCD) was added to a final concentration of 5 nM using a Multidrop Combi reagent dispenser (Thermo Fisher Scientific). After 30 min of incubation at room temperature, another addition was made with the dispenser to deliver GST-tagged LEDGF(IBD) (residues 347-429) to a final concentration of 5 nM alongside the time-resolved FRET reagents allophycocyanin-conjugated ␣-hexahistidine monoclonal antibody and ␣-GST europium-labeled monoclonal antibody (PerkinElmer Life Sciences) at 5 nM final concentrations. After an additional 1 h of incubation at room temperature, the time-resolved FRET signal at 665 nm was recorded with a ViewLux microplate imager (PerkinElmer Life Sciences).
Size-exclusion Chromatography and Multiangle Light Scattering (SEC-MALS)-Absolute molecular weights were determined by multiangle light scattering coupled with refractive interferometric detection (Wyatt Technology Corp., Santa Barbara, CA) and a Superdex 200 10/300 GL column (GE Healthcare) at room temperature, as described previously (33).
Sedimentation Equilibrium Analysis-Sedimentation equilibrium analytical ultracentrifugation experiments were performed at 4°C with an XL-A analytical ultracentrifuge (Beckman-Coulter, Brea, CA) and a TiAn60 rotor with two-channel charcoal-filled Epon centerpieces and quartz windows. Data were collected at 4°C with detection at 280 nm for 5, 7.5, and 10 M samples. Analysis was carried out using global fits to all concentrations with data acquired at 18,000, 22,000, and 24,000 rpm, with strict mass conservation using the program SEDPHAT (70).
Sedimentation Velocity Analysis-Sedimentation velocity ultracentrifugation experiments were performed at 4°C with an XL-A analytical ultracentrifuge (Beckman-Coulter, Brea, CA) and a TiAn60 rotor with two-channel charcoal-filled Epon centerpieces and quartz windows. Samples were analyzed at an A 280 of 0.5-1.2. Complete sedimentation velocity profiles were recorded every 30 s for 50 -200 boundaries at 45,000 rpm. Data were fit using the c(S) distribution model of the Lamm equation as implemented in the program SEDFIT (37). After optimizing meniscus position and fitting limits, the sedimentation coefficient (S) and best fit frictional ratio (f/f 0 ) was determined by iterative least squares analysis, and final values were corrected to 20°C in water (s 20,w ). The partial specific volume (v ) of the protein studied, including molar extinction coefficients, solvent density ( ϭ 1.01 g/ml), and viscosity ( ϭ 0.01002 poise) were derived from chemical composition by the program SEDNTERP (38).
SAXS-Data were recorded on a Rigaku S-Max3000 smallangle x-ray scattering system equipped with Osmic mirror optics (Osmic Inc., Troy, MI), a three-pinhole enclosed preflight path, an evacuated sample chamber with customized sample holder cryostated at 4°C, and a gas-filled multiwire detector; the instrument is served by a Rigaku MicroMax-007 HF microfocus rotating anode generator (Rigaku America, Woodland, TX). Protein samples were spun at 45,000 RPM for 5 min at 4°C in a tabletop centrifuge prior to the addition of DMSO or GSK1264 and immediate 10-min x-ray exposures. The forward scattering from the samples studied was recorded on a CCD detector and circularly averaged to yield one-dimensional intensity profiles as a function of q (q ϭ 4sin/, where 2 is the scattering angle, in units of Å Ϫ1 ). Data were reduced using SAXSGui version 2.05.02 (JJ X-Ray Systems ApS, Lyngby, Denmark), and matching buffers were subtracted to yield the final scattering profile. The sample-to-detector distance and beam center were calibrated using silver behenate and intensity converted to absolute units (cm Ϫ1 ) using a known polymer standard.
SAXS Data Analysis-All of the preparations analyzed were monodisperse, as evidenced by linearity in the Guinier region of the scattering data and agreement of the I(0) and R g values determined with inverse Fourier transform analysis by the programs GNOM (39). Guinier analyses were performed where qR g Յ 1.4. Molecular mass was derived from I(0) measurements, using the forward scatter from the following series of protein standards of known mass at a 5 mg/ml concentration for a standard curve: cytochrome c (12.2 kDa), RNase A (13.7 kDa), myoglobin (17.7 kDa), soybean trypsin inhibitor (20.1 kDa), chymotrypsin (25 kDa), horseradish peroxidase (44 kDa), catalase (dimer of 125 kDa), ␥-globulin (151 kDa), and thyroglobulin (a dimer of 670 kDa).
When fitting manually, the maximum diameter of the particle (D max ) was adjusted in 5-10-Å increments in GNOM to maximize the goodness of fit parameter, to minimize the discrepancy between the fit and the experimental data, and to optimize the visual qualities of the distribution profile.
Turbidity Assays-Assays were performed using a Tecan 96-well plate reader (Tecan Group Ltd., Männedorf, Switzerland), which monitors absorbance at 405 nm in 1-min intervals over 10 -60 min at 27°C. Reactions were initiated by adding protein solutions to a final concentration of 9 -30 M to a buffer containing 20 mM HEPES, pH 7.5, 450 mM NaCl, 0.1 mM EDTA, 10 M ZnOAc 2 , 5 mM CHAPS, 10 mM DTT, and inhibitor at 5-100 M final concentrations or DMSO control. Experiments performed on preformed LEDGF⅐IN complexes were performed in the absence of CHAPS.
Structure Determination-The CCD of HIV-1 IN F185K was diluted to 10 mg/ml in 10 mM HEPES, pH 7.0, 500 mM NaCl, and 3 mM DTT. Apocrystals were grown by hanging drop vapor diffusion at 4°C. 2-l drops were set by combining 1 l of protein and 1 l of 7% PEG 8000, 0.2 M ammonium sulfate, 0.1 M sodium cacodylate, pH 6.5, 5 mM manganese chloride, 5 mM magnesium chloride, and 5 mM DTT. The apocrystals were harvested and soaked with 2.5 mM GSK1264 for 3 days at 4°C. The soaked crystals were transferred to 30% ethylene glycol in mother liquor and subsequently flash-frozen in liquid nitrogen.
X-ray diffraction data were collected at the Advanced Photon Source at Argonne National Laboratories on beam line 21-ID-G. Data collection statistics are summarized in Table 1. The structure was determined by molecular replacement using the program Phaser (40). AutoBuster (41) was used for the initial rounds of refinement and map calculations. The structure was rebuilt using the program Coot (42). Final refinements were done using the program Refmac (43).  (23, 44 -46).

GSK1264 Inhibits Binding of LEDGF(IBD) to IN F185K (CCD)-
GSK1264 was tested in an IN F185K (CCD) versus LEDGF(IBD) time-resolved FRET binding assay (47). The pIC 50 was determined to be 8.2 Ϯ 0.14 nM (n ϭ 7, where pIC 50 is the negative log of the IC 50 in moles/liter; IC 50 is the concentration required to achieve 50% inhibition). In multicycle viral replication assays (MT4 cell assay), GSK1264 inhibited HIV-1 replication with an IC 50 of ϳ38 nM.
X-ray Crystallography Reveals That GSK1264 Binds to the IN(CCD) Dimer Interface-Crystals of IN F185K (CCD) were soaked with GSK1264, and the structure was determined by x-ray crystallography (Fig. 1C). In this structure, electron density consistent with bound GSK1264 could be identified and unambiguously modeled. As is observed in other crystal structures of IN(CCD) and inhibitors of this class, the drug is bound within a pocket defined by helices ␣1 and ␣3 from one monomer subunit and ␣4 and ␣5 from another. The carboxylate moiety of the compound interacts with the backbone amides of IN residues Glu-170 and His-171, with additional hydrogen bonds to the carboxylate moiety provided by the hydroxyl oxygen of Thr-174. As was observed with tert-butoxy-(4-phenyl-quinolin-3-yl) acid inhibitor co-crystallized with the IN(CCD) (44), the tert-butoxy moiety binds to a complementary pocket lined by residues Gln-95, Tyr-99, Thr-125, and Thr-174. The substituent at R4 of the isoquinoline core is pointed toward Trp-132 and adjacent to Ala-128, almost at a 45º angle with respect to the plane of the dimer interface and parallel to ␣3. The positioning of the allosteric inhibitor GSK1264 spatially coincides with that observed for other multimodal inhibitors studied by x-ray crystallography (22-24, 44, 48).
Inhibition of Early Steps of HIV-1 Infection by GSK1264- Fig.  2A shows the effects of GSK1264 on the early steps of HIV-1 infection. TZM-bl cells (28) were infected with HIV-1, and the efficiency of infection was read out as luciferase production from a long terminal repeat (LTR)-luciferase reporter. Thus, the assay measures early infection steps through gene expression from the viral LTR. Efficient infection was seen at concentrations of 250 nM GSK1264 and lower, but inhibition was seen at 1,250 nM. Thus, inhibition as measured in the TZM-bl assay, which requires only infection steps through transcription, was much less potent than that measured in the multicycle viral replication assay (which was ϳ38 nM).
Lack of Effect of GSK1264 on Integration Targeting-The LEDGF/p75 protein binds both HIV-1 IN and chromatin at active transcription units, thereby directing HIV-1 DNA integration to these sites (6, 8, 49 -51). We thus investigated whether integration targeting was altered in the presence of GSK1264. For this study, we used the infected cells generated in the study in Fig. 2A. Integration site distributions were mapped in genomic DNA using ligation-mediated PCR and 454/Roche deep sequencing. Results are summarized as a heat map in Fig. 2B.
Each integration site data set was matched with random sequences for statistical analysis, and departures from random by the experimental data relative to genomic features were scored using the ROC (receiver operating characteristic) area method (30). Samples were then compared in the presence and absence of GSK1264. The samples tested were from the infections in Fig. 2A, so at 1,250 nM, efficient inhibition was observed.
As can be seen by comparing rows in the heat map in Fig. 2B, only slight differences were seen between samples. No comparisons achieved statistical significance. The study was limited by the relatively modest number of integration sites available due to the low multiplicity of infection used in Fig. 2A but is still sufficient to conclude that concentrations of GSK1264 that are sufficient to achieve significant inhibition in the TZM-bl assay (1,250 nM) did not have major effects on integration site targeting at this depth of analysis.
We next investigated whether competition between LEDGF and GSK1264 was having any detectable influence early during infection. We compared the IC 50 of GSK1264 in test infections on 293T-derived cells knocked down for LEDGF using shRNA versus control cells. We found that there was a drop in IC 50 from 55.2 Ϯ 10.2 nM in the wild-type cells to 11.0 Ϯ 1.4 nM in the LEDGF knockdown cells. As controls, an integrase strand transfer inhibitor and a nonnucleoside reverse transcriptase inhibitor were tested and found to show no difference between wild-type and LEDGF knockdown cells. This 5-fold differential seen with GSK1264 was less than for BI-D, another compound of this class, which showed a 30-fold differential (23). Thus, competition between LEDGF and GSK1264 early during infection is less pronounced for GSK1264 than for another compound in this class.
Potent Inhibition of Late Replication Steps by GSK1264 in a Virus Prebinding Format-In contrast to the modest potency observed for inhibition of early steps in the single round assay, inhibition of late steps was potent. Two assay formats were used to make this point.
In the first, 293T cells producing HIV NL4 -3 virus were treated with GSK1264 or DMSO carrier, and infectivity was analyzed. Virus production quantified by RT assay (Fig. 3A) or p24 antigen yield (Fig. 3B) showed no significant difference in the presence of GSK1264 up to 5 M. To test the infectivity of these particles, virus produced in the presence of GSK1264 or DMSO control was applied to U373/CD4/CCR5 target cells. Potent inhibition was detected, with an estimated IC 50 of 12 nM. In contrast, infection of the U373/CD4/CCR5 cells in the presence of different concentrations of GSK12654 with HIV NL4 -3 (where virus was not treated with drug) yielded an IC 50 of 557 nM. Thus, there was a 46-fold difference between the efficiency of infection late versus early.
A complication arises in analyzing virus stocks produced in the presence of drug, which is that compound carried over with the viral stock may influence the course of the infection in the newly infected target cells (although the above study was designed to dilute drug in viral stocks extensively before application to target cells). To eliminate this concern, in a second study, HIV-1 particles were prebound to retronectin-coated plates and washed, and then TZM-bl target cells were added. In this format, inhibition by GSK1264 remained effective (Fig.  3C), with complete inhibition of viral replication seen between 10 and 50 nM. As a control, non-drugged and drugged viruses were mixed, bound to the plate, and washed, and then cells were added (Fig. 3C, GSK1264 ϩ DMSO). Infection was efficient, indicating that the drug was not carried over and that the virus produced in the presence of drug did not inhibit infection by non-drug-treated virus.
We thus conclude that GSK1264 exerts its effects primarily in the late steps of the viral replication cycle, resulting in the formation of replication-deficient particles, a conclusion that has also been reached by others in studies of structurally related molecules (20,22,23,25,26). (Fig. 4, A-E). For this, we examined a mutant version of IN Q (C56S, F139D, F185H, and C280S) (33, 52-54) that has more favorable solubility properties than wild-type IN alone. Global fitting of sedimentation equilibrium data with strict mass con-servation obtained from recombinant IN Q at 4°C was best described by a monomer-dimer-tetramer association, with dissociation constants of ϳ5 M for K d1-2 and ϳ86 M for K d2-4 , respectively (data not shown). Attempts to analyze IN Q in the presence of 50 M GSK1264 were stymied by loss of soluble mass. However, the data that were obtained through two speeds (ϳ40 h at 4°C) and across three concentrations could be well described by a dimer-tetramer-octamer association with dissociation constants of ϳ4 M for K d2-4 and ϳ86 M for K d4 -8 , respectively. However, such an analysis cannot reliably discriminate a 2-4-8 association from other models, such as 2-4-16 or 2-5-7.

Effects of GSK1264 on the Solution Properties of Full-length IN in Vitro-We analyzed IN by several biophysical methods to assess the effects of GSK1264 on its solution properties in vitro
Simulations using the monomer-dimer-tetramer association model over the 1-100 M range predicted that IN alone forms a mixture of monomer, dimer, and tetramer species (supplemental Fig. 1). Such heterogeneity precludes structural analysis by small-angle x-ray scattering. However, weight-averaged mass determination is possible using I(0) analysis, where mass is derived by comparison of the incident intensities of known concentration with mass standards of known mass and concentration. We found that the weight-average mass (M w ) of IN was consistent with ϳ128 kDa from 1-5 mg/ml concentrations ( Fig. 4A and Table 2), in line with our simulations. Titration of the GSK1264 compound with IN at 4°C over a 10-min period revealed a concentration-dependent aggregation of IN, where the determined mass did not terminate with tetramer or octamer but instead arrived in the megadalton range. These observations were corroborated by sedimentation velocity analyses (Fig. 4, B-D), which allowed for the discrimination of discrete oligomers stimulated by the compound.
To study the drug-induced oligomerization of IN further, we employed SEC-MALS coupled with dynamic light scattering (Fig. 4E). In this approach, absolute concentration determina- tion is coupled with 18-angle light scattering to determine the weight-averaged mass of the eluent. At room temperature and at eluted concentrations of ϳ0.1 mg/ml, IN was an apparent monomer. However, the slope of the mass profile and the breadth of the eluent peak were indicative of a self-association (Fig. 4E). The retention properties and mass profiles observed were concentration-dependent, consistent with the modest micromolar dissociation constants determined by centrifugation. After preincubation with saturating amounts of GSK1264 (200 M) for 15 min at room temperature, an increase in apparent mass was observed, consistent with a population of dimers and higher order species (Fig. 4A), similar to previous studies with related molecules (23,55,56). Similar tests of raltegravir and azidothymidine did not show aggregation (data not shown). The R h profile across the monomer and dimer peaks provided by in-line dynamic light scattering were nearly identical, suggestive of the transition from an extended monomer to more compact dimer (data not shown).

The Effect of GSK1264 on Preformed LEDGF⅐IN Complexes in
Vitro-We next examined the effect of GSK1264 on preformed LEDGF⅐IN complexes. By co-expression, monodisperse complexes of LEDGF⅐IN can be obtained in 4:4 or 4:2 stoichiometries, depending on the mutational background and length of the C-terminal domain of LEDGF/p75 used (33).
Co-expression of IN Q ⅐LEDGF(326 -530), which contains the C-terminal domain of LEDGF/p75, provides a 4:4 complex (33) (Fig. 5A). SEC-MALS analysis of the complex preincubated with 200 M GSK1264 at room temperature for 15 min showed stimulated oligomerization, with a heterogeneous mass profile up to at least an LEDGF-bound IN octamer. LEDGF appeared to be mostly retained within the aggregate because relatively little LEDGF monomer was detected in the absorbance, refractive index, and light scattering data. A similar result was obtained after GSK1264 treatment at lower concentrations over a 2-h time course by sedimentation velocity analysis at 4°C, where heterogeneous aggregates of the LEDGF⅐IN com- The key at the bottom shows the frequency of association for integration sites compared with the random distributions. Red, positive association; blue, negative association. The numbers on the left indicate the interval size over which the densities of integration sites were compared with the indicated form of genomic annotation. None of the slight differences seen between samples achieved statistical significance. For the experiment in B, we used cellular DNA from infections in A to maintain linkage between experiments; however, low multiplicity of infection was used in A to minimize cells with multiple proviruses, so as a consequence the number of integration sites recovered was modest (645 total) (supplemental Table S1). RLU, relative light units. Error bars, S.D. A, lack of inhibition of particle production measured using an RT assay of particleassociated RT activity. Infected cells were treated with the indicated concentration of GSK1264, and cell-free particles were assayed for RT activity. Drug concentrations are as indicated on the x axis. B, lack of inhibition of particle production measured using a p24 assay of particle production. Infected cells were treated with the indicated concentration of GSK1264, and cell-free particles were assayed for p24 content. C, potent inhibition late during infection. After chronic virus producer cells were treated with GSK1264, viral supernatants were harvested and applied to retronectin-coated plates. Plates were washed, and then TZM-bl indicator cells were added, and luciferase activity was assayed. As a control, a mixture of drug-treated and control viruses was worked up similarly, and infection was shown to be normal (GSK1264 ϩ DMSO), indicating that drug was not carried over in quantities sufficient to be inhibitory; nor was drugged virus able to inhibit untreated virus. Error bars, S.D.
plex were observed (Fig. 5B). Similar aggregation was seen in the presence of GSK1264 for IN Q ⅐LEDGF(IBD), which is purified as a 4:2 complex (33) (data not shown). To quantitate the aggregation reaction, we devised a turbidity assay by monitoring absorbance at 405 nm. At this wavelength, there is no contributing absorbance from protein or GSK1264 alone. Over a 60-min time course at room temperature, only the full-length IN and the IN F185H (CCD-CTD) con-  struct showed evidence of drug-dependent aggregation (Fig.  6B). Thus, we conclude that GSK1264-induced aggregation of IN in this assay minimally requires interactions mediated by the CCD and CTD. As controls, aggregation was tested in the presence of raltegravir and azidothymidine (Fig. 6C) (59) and that oligomerization is stimulated by LEDGINs (a group of allosteric IN inhibitors) (23,25,26). We found that the druginduced aggregation of IN Q was increased by the addition of the crowding agent Ficoll, potentially mimicking the crowded environment during assembly and in particles. Ficoll added to IN Q in the absence of drug had no effects on its aggregation rate under the conditions tested (Fig. 6D).

DISCUSSION
Here we report a study of GSK1264, an allosteric inhibitor of HIV-1 IN that blocks viral replication. GSK1264 is a compound of the group variously named ALLINIs (20), NCINIs (26), LEDGINs (48,60,61), and the tert-butoxy-(4-phenyl-quinolin-3-yl) acid molecules (44,46). GSK1264 was isolated in an assay requiring displacement of LEDGF from IN, but surprisingly, these compounds seem to act mainly independently of LEDGF in vivo. Instead, these inhibitors have been linked to post-integration effects associated with the self-association of IN (20,(22)(23)(24)(25)(26). We show that the GSK1264 compound elicits concentration and time-dependent polymerization of IN, ultimately leading to the formation of polydisperse, insoluble aggregates. Inhibition did not block particle release but resulted in production of particles impaired for subsequent steps of replication, consistent with a defect in reverse transcription or other steps in the next round of replication (20,(22)(23)(24)(25)(26). However, purified particles were fully competent for reverse transcription after disruption and assay in vitro (Fig. 3A), indicating that the enzyme is present and functional, implicating defects in other properties, such as higher order assembly.
We did not observe displacement of LEDGF(IBD) from preformed IN⅐LEDGF complexes in the presence of GSK1264 in vitro, despite induction of efficient aggregation. It is unclear whether IN protein is saturated with LEDGF in vivo, but here we show that both 4:2 and 4:4 complexes can show stimulated oligomerization in the presence of GSK1264 without measureable LEDGF displacement. A simple model is that LEDGF remains trapped in the aggregate after compound binding. It is unknown whether this would be the case in the context of preintegration complexes in vivo.
A recent study suggested that competition between an ALLINI (BI-D) and LEDGF played out differently at early and late steps of viral replication. Relatively potent late inhibition by BI-D was not affected by the presence of LEDGF, but weaker inhibition of early replication was diminished (23). This suggests that there was competition between LEDGF and BI-D binding to early replication complexes in vivo. For BI-D, the difference in inhibition early versus late in the presence of LEDGF was 26-fold. For GSK1264, the differential early versus late was greater (46-fold), and the effects of LEDGF knockdown on early steps were more modest (only 5-fold versus 30-fold). Biophysical data showed no detectable displacement of LEDGF by GSK1264, and GSK1264 did not strongly disrupt integration target site selection, all consistent with the idea that GSK1264 may be less effective at displacing LEDGF from IN than is BI-D.
We investigated the domain requirements for aggregation with GSK1264, which suggested that polymerization of CTDs mediates self-association and formation of an uncapped polymer. In support of this, the HIV-1 IN(CCD-CTD) crystal structure (Protein Data Bank entry 1EX4 (53)) reveals a tetramer packing interface that is mediated solely by CTD-CTD interactions and bears no resemblance to the tetramers observed in the prototype foamy virus integrase in complex with DNA (62-66). Thus, we hypothesize that the open polymer formed in the presence of allosteric inhibitors proceeds through polymerization of the CTD as in the 1EX4 crystal lattice.
Inhibition by GSK1264 and related molecules may be promoted by IN protein crowding late during viral replication, as emphasized here by increased precipitation in the presence of Ficoll. A typical HIV-1 virion has an internal volume of ϳ980,000 nm 3 and a total macromolecular concentration exceeding 500 g/liter (67)(68)(69). Allosteric inhibitors have been shown to impair maturation of particle cores (23,25,26). As the nascent virion assembles, the crowded environment that arises may promote aggregation induced by GSK1264 and related compounds. Thus, potency may be amplified by the environment encountered by IN during assembly late during the viral replication cycle.