The A128T Resistance Mutation Reveals Aberrant Protein Multimerization as the Primary Mechanism of Action of Allosteric HIV-1 Integrase Inhibitors

Background: The A128T substitution in HIV-1 integrase (IN) confers resistance to allosteric integrase inhibitors (ALLINIs). Results: The A128T substitution does not significantly alter ALLINI IC 50 values for IN-LEDGF/p75 binding but confers marked resistance to ALLINI-induced aberrant integrase multimerization. Conclusion: Allosteric perturbation of HIV-1 integrase multimerization underlies ALLINI antiviral activity. Significance: Our findings underscore the mechanism of ALLINI action and will facilitate development of second-generation compounds.

HIV-1 integrase (IN) 2 is an important therapeutic target, as its function is essential for viral replication (1). IN catalyzes the insertion of the reverse-transcribed RNA genome into human chromatin in a two-step reaction (2). During the initial step (termed 3Ј-processing), IN removes a GT dinucleotide from each 3Ј terminus of the viral DNA. The subsequent transesterification reactions (termed DNA strand transfer) covalently join the recessed viral DNA ends into the host genome. To carry out these reactions, highly dynamic IN subunits assemble in the presence of cognate DNA to form the stable synaptic complex or intasome (3)(4)(5). Premature multimerization of HIV-1 IN by various ligands in the absence of cognate DNA restricts the functionally essential dynamic interplay between individual subunits and DNA and thus impairs IN catalytic activities (3,4,6,7).
A cellular protein, lens epithelium-derived growth factor (LEDGF)/p75, markedly enhances the integration process in vitro and in infected cells. LEDGF/p75 tethers stable synaptic complexes to chromatin via direct interactions of its N-terminal chromatin-binding domain with nucleosomes, whereas its C-terminal IN-binding domain links to HIV-1 IN (8,9). Structural studies (10,11) have elucidated the principal interacting interfaces between the LEDGF/p75 IN-binding domain and the HIV-1 IN catalytic core domain (CCD). The cornerstone of this protein-protein interaction is a hydrogen bonding network between the side chain oxygen atoms of LEDGF/p75 Asp-366 and the IN backbone amides of Glu-170 and His-171 (10,11). These findings have opened up new venues for antiviral drug discovery.
Using a structure-based rational drug design approach, Christ et al. (12) developed 2-(quinolin-3-yl)acetic acid derivatives, which inhibit the IN-LEDGF/p75 interaction in vitro and HIV-1 replication in cell culture. Paradoxically, the identical class of compounds has emerged from a high-throughput screen for IN 3Ј-processing activity (13). Subsequent studies from our group and others demonstrated that 2-(quinolin-3yl)acetic acid derivatives exhibit a multimodal mechanism of action by allosterically modulating the IN structure, which affects both IN-LEDGF/p75 binding and catalytic activity (14 -16). Accordingly, we have proposed to name this class of inhibitors as allosteric IN inhibitors (ALLINIs). Structural studies have shown that the carboxylic acid of ALLINIs hydrogen bonds with the backbone amides of Glu-170 and His-171 and thus occupy the principal LEDGF/p75-binding interface (12,14,16). At the same time, the quinoline core and the substituted phenyl group of the inhibitor bridge the two IN subunits and promote allosteric multimerization of the protein. As a result, ALLINIs potently inhibit both IN-LEDGF/p75 binding and LEDGF/p75-independent IN catalytic activities.
The ability of ALLINIs to impair steps in the viral replication cycle that extend beyond IN catalytic function results in a highly cooperative inhibition of HIV-1 replication (14). Along these lines, it was recently reported that HIV-1 particles made in the presence of ALLINIs are noninfectious (15). Whereas the Food and Drug Administration-approved HIV-1 IN strand transfer inhibitor raltegravir (RAL) exhibits a Hill coefficient of 1, which is consistent with only a single or non-cooperative mode of action, ALLINIs impair viral replication with a Hill coefficient of ϳ4, indicating a highly cooperative mode of action (14). Compounds with a high cooperativity are particularly desirable for superior clinical outcomes because they enable stronger viral suppression at clinical drug concentrations (17,18).
The A128T substitution in HIV-1 IN has been identified from cell culture assays as a primary mechanism for resistance to ALLINI compounds (12,16,19). Ala-128 is located at the IN dimer interface in the pocket occupied by ALLINIs or LEDGF/ p75. Here, we have investigated the structural and mechanistic properties for the resistance of A128T IN to ALLINIs. Strikingly, the A128T substitution only modestly affected ALLINI IC 50 values for IN-LEDGF/p75 binding but markedly altered the multimerization of IN in the presence of the inhibitors. As a result, the catalytic activities of the WT protein were potently inhibited by ALLINIs, whereas A128T IN exhibited significant resistance to the inhibitor. Furthermore, considerably higher concentrations of ALLINIs were required to inhibit the infectivity of the A128T mutant virus compared with the WT counterpart. Taken together, our studies highlight that aberrant IN multimerization is the primary target of this class of inhibitors and thus provide the structural foundations for the development of second-generation ALLINIs with increased potency and decreased potential to select for drug resistance.

EXPERIMENTAL PROCEDURES
Antiviral Compounds-ALLINI-1, referred to as BI-1001 previously, was synthesized as described (14). The synthesis of ALLINI-2 is described in supplemental Figs. S1 and S2. RAL and saquinavir were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program.
Expression and Purification of Recombinant Proteins-LEDGF/p75 and WT and A128T HIV-1 IN recombinant proteins with His or FLAG tags were expressed in Escherichia coli and purified as described previously (14).
Protein-Protein Interaction Assays-Homogeneous time-resolved fluorescence (HTRF)-based IN-LEDGF/p75 binding and IN multimerization assays were performed as described previously (14). The HTRF signal was recorded using a PerkinElmer EnSpire multimode plate reader.
Solubility Assays-WT IN was diluted to a final concentration of 100 nM in buffer containing 25 mM Tris (pH 7.4), 2 mM MgCl 2 , 0.1% Nonidet P-40, 1 mg/ml BSA, and either 150 or 750 mM NaCl. Increasing concentrations of ALLINI-1 or ALLINI-2 were then added to the mixture and incubated for 1 h at room temperature. The mixture was subjected to centrifugation for 2 min at 2000 ϫ g. The supernatant was collected, and the pellet was washed three times with the same buffer. The supernatant and pellet fractions were analyzed by SDS-PAGE, and IN was detected with anti-His antibody (Abcam).
3Ј-Processing, Strand Transfer, and LEDGF/p75-dependent Integration Assays-Gel-based LEDGF/p75-dependent and LEDGF/p75-independent integration activity assays were performed as described previously (14). A recently reported timeresolved fluorescence assay (16) was used to quantify IN 3Ј-processing and strand transfer activities. The HTRF-based LEDGF/p75-dependent integration activity was measured by adding recombinant LEDGF/p75 protein to the assay mixture prior to the incubation with labeled DNA substrates. The timeresolved fluorescence signal was recorded using a PerkinElmer EnSpire multimode plate reader.
Crystallization and X-ray Structure Determination-The HIV-1 IN CCD (residues 50 -212 containing the F185K substitution) and A128T CCD (with an extra substitution of A128T) were expressed and purified as described (20). The CCD was concentrated to 8 mg/ml and crystallized at 4°C using the hanging drop vapor diffusion method. The crystallization buffer contained 10% PEG 8000, 0.1 M sodium cacodylate (pH 6.5), 0.1 M ammonium sulfate, and 5 mM DTT. Crystals reached 0.1-0.2 mm within 4 weeks. The A128T CCD was concen-trated to 8.5 mg/ml and crystallized at room temperature (20°C) using the hanging drop vapor diffusion method. The crystallization buffer contained 0.1 M sodium cacodylate (pH 6.5), 1.4 M sodium acetate, and 5 mM DTT. The crystals reached 0.2-0.4 mm within 1 week.
A soaking buffer containing 5 mM ALLINI-1 or ALLINI-2 was prepared by dissolving the compound in crystallization buffer supplemented with 10% Me 2 SO. The protein crystal was soaked in the buffer for 8 h before flash-freezing in liquid N 2 . Diffraction data were collected at 100 F on a Rigaku R-AXIS 4ϩϩ image plate detector at the Ohio State University Crystallography Facility. Intensity data integration and reduction were performed using the HKL2000 program (21). The molecular replacement program Phaser (22) in the CCP4 package (23) was used to solve the structure, with the HIV-1 IN CCD structure (Protein Data Bank code 1ITG) (20) serving as starting model. The Coot program (24) was used for the subsequent refinement and building of the structure. Refmac5 (25) of the CCP4 package was used for the restraint refinement. TLS (26) and restraint refinement was applied for the last step of the refinement. The crystals belonged to space group P3 1 21 with cell dimensions a ϭ b ϭ 73 and c ϭ 65 Å, with one 18-kDa monomer in the asymmetric unit. The data collection and refinement statistics are listed in supplemental Table 1. Coordinates have been deposited in the Protein Data Bank with accession numbers 4JLH, 4GW6, and 4GVM (supplemental Table 1).
HIV-1 Virion Production and Infectivity Assay-HEK293T and HeLa TZM-bl cells were cultured in Dulbecco's modified Eagles medium (Invitrogen), 10% FBS (Invitrogen), and 1% antibiotic/antimycotic (Invitrogen) at 37°C and 5% CO 2 . Cultures of HEK293T cells (2 ϫ 10 5 cells/well of a 6-well plate in 2 ml of complete medium) were transfected with 2 g of pNL4 -3 (WT or A128T mutant) at a 1:3 ratio of DNA to X-tremeGENE HP (Roche Applied Science) following the manufacturer's protocol. Twenty-four hours post-transfection, cells were washed once with complete medium, and the culture supernatant was replaced with complete medium containing Me 2 SO, RAL (250 nM), ALLINI-1, or ALLINI-2 (at the indicated concentrations). After 1 h, the culture supernatant was again replaced with fresh complete medium containing either Me 2 SO or the indicated inhibitors. The virus-containing cell-free supernatant was collected after 24 h, and HIV-1 Gag p24 ELISA (ZeptoMetrix) was performed following the manufacturer's protocol. Virions equivalent to 2-4 ng of HIV-1 p24 was used to infect 2 ϫ 10 5 HeLa TZM-bl cells in the presence of 8 g/ml Polybrene (Sigma). HeLa TZM-bl cultures were extracted in 1ϫ reporter lysis buffer (Promega), and virion infectivity was measured using the luciferase assay (Promega).

RESULTS
To examine the effects of the A128T substitution on HIV-1 IN function, we compared the catalytic activities of purified recombinant WT and mutant proteins. The two proteins exhibited comparable levels of LEDGF/p75-independent and LEDGF/p75-dependent integration activities (Fig. 1). Thus, the A128T substitution does not significantly alter the function of IN.
The A128T mutation has been shown to confer resistance to four different ALLINI compounds in cell culture assays (12,15,16,19). To delineate the mechanism of drug resistance, the ALLINI-1 and ALLINI-2 compounds were synthesized (Fig. 2). ALLINI-1 was identified by Boehringer Ingelheim through a high-throughput screen for IN 3Ј-processing activity (13), and its multimodal mechanism of action has been elucidated by our group (14). In resistance studies under selective pressure of ALLINI-1 (19), the A128T substitution in IN was identified in both early and late stage viral passages. This single amino acid change resulted in 32-fold higher ALLINI-1 IC 50 values compared with the WT virus. In this study, we also examined ALLINI-2, a tert-butyl derivative of ALLINI-1. Similar to a published report (16) showing that the tert-butyl group increases the potency of this class of compounds, ALLINI-2 was ϳ10-fold more potent (IC 50 ϭ 0.63 Ϯ 0.3 M) (supplemental Table 2) than ALLIN-1 (IC 50 ϭ 5.8 Ϯ 0.1 M). Although we have not selected for the A128T mutation by serially passaging HIV-1 in the presence of ALLINI-2, the ability of this mutation to confer relative pan-tropic resistance to a number of different compounds, including those that harbor the tert-butyl moiety (15,16), gave us confidence that A128T would likely confer resistance to ALLINI-2. Supplemental Table 2 indeed shows that A128T conferred 19-fold resistance to ALLINI-2 in a spreading HIV-1 replication assay.
We and others have shown that ALLINIs inhibit multiple functions of WT IN, including LEDGF/p75-independent catalysis and IN-IN multimerization, with similar IC 50 values (14 -16). Such a mechanism has been attributed to the fact that these compounds occupy the LEDGF/p75-binding pocket at the IN dimer interface. As a result, ALLINIs inhibit IN-LEDGF/p75 binding and also bridge two IN subunits and allosterically modulate their multimerization. In turn, the latter impairs the catalytic functions of WT IN.
Strikingly, we observed that the A128T substitution had markedly different effects on the IC 50 values for different assays in comparison with WT IN (Table 1). For example, the mutation conferred only an ϳ2-fold increase in IC 50 values for inhibiting the IN-LEDGF/p75 interaction, whereas the IC 50 values for the 3Ј-processing reaction increased by 287fold and 1112-fold for ALLINI-1 and ALLINI-2, respectively. Thus, A128T IN was markedly resistant to ALLINIs in 3Ј-processing reactions, whereas these compounds remained potent inhibitors of A128T IN binding to LEDGF/ p75 ( Fig. 3 and supplemental Fig. S3). The A128T substitution resulted in ϳ11.5and 5-fold resistance to ALLINI-1 and ALLINI-2, respectively, in the strand transfer reactions (Table 1). In LEDGF/p75-dependent integration assays, the mutant protein exhibited ϳ12and 25-fold resistance to ALLINI-1 and ALLINI-2, respectively (Table 1).
Interestingly, IN multimerization assays have revealed further striking differences between WT and mutant INs ( Fig. 4 and supplemental Fig. S4). Both WT and A128T INs exhibited a characteristic biphasic dose-response curve upon the addition of ALLINIs. The HTRF signal increase with increasing concentrations of inhibitor was due to the inhibitor-induced protein multimerization, yielding higher FRET. Although the exact nature of the descending curve was not clear, it could be explained by reduced accessibility of the fluorescent antibodies to their respective tags in the context of higher order IN oligomers. The assay uses anti-His 6 -XL665 and anti-FLAGeuropium-cryptate antibodies to monitor fluorescence energy transfer (HTRF signal) between His-IN and IN-FLAG proteins. During the initial multimerization of IN, when dimers and tetramers form, the affinity tags are sufficiently exposed to readily engage the antibodies. However, these interactions may be limited in higher order IN oligomers (see Fig. 5 and supplemental Fig. S5) due to structural hindrance of the affinity tags and could thus account for the drop in HTRF signal. Fig. 4 and supplemental Fig. S4 show that the dose-dependent addition of ALLINIs to WT and A128T INs yielded different peak heights. These differences could be explained by WT and mutant INs adopting different oligomeric states (see below) or alternative conformations in the presence of the inhibitor.
To delineate between these possibilities, we performed size exclusion chromatography experiments ( Fig. 5 and supplemental Fig. S5). Due to the reduced sensitivity of this approach compared with the HTRF-based assays, elevated concentrations of IN and ALLINIs were necessary. Tetramer and monomer peaks were detected with both WT and A128T INs in the absence of inhibitor, demonstrating that the substitution does not affect IN multimerization ( Fig. 5 and supplemental Fig. S5) or catalytic activities (Fig. 1). Upon the addition of ALLINIs, the tetramer peak of WT IN was markedly reduced, and instead, new peaks corresponding to higher order oligomers were detected. In sharp contrast, the tetramer peak persisted upon the addition of ALLINI-1 or ALLINI-2 to A128T IN (Fig. 5 and supplemental Fig. S5 and Tables 3-6). These findings are consistent with the results of the HTRF-based multimerization assays: the formation of higher order structures upon ALLINI addition resulted in greater HTRF signal strength compared with mutant IN and could also account for the downward slope of the WT IN curves at high compound concentrations ( Fig. 4 and supplemental Fig.  S4).
We next examined whether the addition of ALLINIs to IN might promote the formation of insoluble aggregates. WT IN was incubated with increasing concentrations of ALLINI-1 or ALLINI-2 and then subjected to centrifugation. The results in Fig. 6 and supplemental Fig. S6 show that, under our reaction conditions, the IN-ALLINI complexes remained soluble. The solubility ( Fig. 6 and supplemental Fig. S6) and HTRF-based IN multimerization ( Fig. 4 and supplemental Fig. S4) assays were carried out at two different NaCl concentrations (150 and 750 mM) and yielded very similar results, indicating that the changes in the ionic strength of the buffer did not significantly affect the solubility of ALLINI-induced higher order IN oligomers. This supports the notion that higher order IN multimerization and not precipitation led to the decrease in HTRF signal seen in Fig. 4 and supplemental Fig. S4.
To elucidate the structural basis for how the resistance mutation affects ALLINI binding, we solved the crystal structures of the inhibitors bound to the WT and A128T IN CCDs (Fig. 7 and  supplemental Fig. S7). The overlay of these two structures shows that the hydrogen bonding network between ALLINIs and subunit 1 is fully preserved in both the WT and mutant proteins ( Fig. 7 and supplemental Fig. S7). These include the interactions of the carboxylic acid with the backbone amides of Glu-170 and His-171 and the methoxy group of ALLINI-2 with the side chain of Thr-174. Thus, ALLINIs effectively shield the access of the key LEDGF/p75 Asp-366 contact to its cognate

HIV-1 Resistance to Allosteric Integrase Inhibitors
hydrogen bonding partners on both WT and A128T INs. Accordingly, these compounds inhibited the binding of the cellular cofactor to WT and A128T INs with comparable IC 50 values ( Fig. 3 and supplemental Fig. S3).
Interestingly, the A128T substitution affected the positioning of the quinoline group ( Fig. 7 and supplemental Fig. S7). ALLINIs were shifted down and inward (toward the protein) by ϳ2 Å as measured at the common bromine atom. Because the positioning of the carboxylic group remained intact, this resulted in the rotation of the rigid molecule by ϳ18°(as measured by the shift of the bromine atom with respect to the C3 atom (numbering according to Fig. 2) in the quinoline ring).
This change also caused the substituted phenyl group to shift downward by 0.8 Å at the chlorine atom. These structural changes could be explained by the substitution of Ala-128 with the bulkier and polar threonine, which could exert a steric effect and electronic repulsion of the compounds. The shifts of the quinoline and substituted phenyl groups that bridge the two monomers of the CCD could be the reason for the differential multimerization of WT and mutant INs.
A previous study (15) demonstrated that ALLINIs do not affect bulk viral particle production but nevertheless impair the infectivity of HIV-1 progeny virions. Here, we examined how the A128T substitution affects this aspect of ALLINI inhibition ( Fig. 8 and supplemental Fig. S8). For this, HEK293T cells were transfected with pNL4 -3 (WT or A128T mutant) and cultured for 24 h to ensure expression of the provirus. Next, we monitored the production of HIV-1 particles in the presence of increasing concentrations of ALLINIs. Twenty-four hours post-addition of the compounds, the virus-containing cell-free supernatant was harvested, and the amounts of viral particles produced were measured by p24 ELISA. The production of both WT and A128T IN viral particles was not affected by ALLINIs (Fig. 8 (upper panel) and supplemental Fig. S8A). Subsequently, we examined the infectivity of these progeny virions in a HeLa-based reporter cell line (TZM-bl) containing the HIV-1 LTR-luciferase reporter gene. For this, TZM-bl cells were infected with equivalent cell-free virions without any additional inhibitor being added to the target cells. Under these conditions, only 0.1% of the input ALLINIs was carried over from the producer cells to the target cells based on the supernatant volumes used for the infections. At the highest concentration of ALLINIs tested (100 M ALLINI-1 treatment of the producer cells), only 100 nM inhibitor would carry over, which is well below the IC 50 value for the WT virus and thus will have   Table 1. In contrast, the A128T virions exhibited a marked resistance to ALLINIs (Fig. 8 (lower panel) and supplemental Fig. S8B). Control experiments with RAL showed no effects of this inhibitor on viral production or infectivity (data not shown).  (14 -16).

DISCUSSION
The IN A128T substitution has been identified from cell culture assays as a primary mechanism of HIV-1 resistance to numerous ALLINIs (12,15,16,19). Here, we elucidated the structural and mechanistic basis for this resistance. The alanine-to-threonine substitution affects positioning of the core quinoline and substituted phenyl ring of ALLINIs that bridge the two IN subunits, whereas the hydrogen bonding network between the inhibitor and the protein that closely mimics the   IN-LEDGF/p75 interaction remains intact. As a result, the A128T substitution shows marked resistance to ALLINI-induced aberrant multimerization of IN compared with its WT counterpart, whereas the compound remains a potent inhibitor of the A128T IN binding to LEDGF/p75.
We have shown that ALLINIs promote aberrant higher order multimerization of WT IN, but not A128T IN. Although previous studies have attributed the HTRF signal increase to IN dimerization (16), the size exclusion chromatography data in Fig. 5 and supplemental Fig. S5 clarify that the addition of ALLINIs to the WT protein promotes the formation of higher order oligomers. As a result, the catalytic activities of the WT protein are fully compromised. In sharp contrast, A128T IN is remarkably resistant to ALLINIs in the 3Ј-processing assays (287-and 1112-fold for ALLINI-1 and ALLINI-2, respectively) and exhibits 5-11-fold resistance in strand transfer assays. How does one explain the differential levels of resistance of A128T IN for 3Ј-processing and strand transfer activities? The HTRF assays coupled with size exclusion chromatography indicate that ALLINIs stabilize a tetrameric form of A128T IN. Of note, IN tetramers formed in the absence and presence of viral DNA adopt distinct conformations (3). Although preformed tetram-ers are known to be active in 3Ј-processing, the strand transfer reactions require individual IN monomers to assemble in the presence of viral DNA to correctly engage target DNA (3,4). Parallels can be drawn with our earlier results demonstrating the importance of highly dynamic interplay of individual IN subunits for productive integration (3,4). The IN tetramers stabilized by the LEDGF/p75 IN-binding domain are active in 3Ј-processing reactions but fail to catalyze concerted HIV-1 integration. Similarly, IN tetramers prematurely stabilized by ALLINIs are likely to be different from the fully functional tetramers in the intasome formed in the presence of DNA substrate.
To understand the 5-11-fold resistance of A128T IN in the strand transfer assays, we analyzed the HTRF data in Fig. 4 and supplemental Fig. S4. Accurate measurements of ALLINI IC 50 values from these assays were complicated due to biphasic curves and differing peak heights for WT and A128T INs. Still, the analysis of the initial ascending curves enabled us to esti-  Table 1).
The differential multimerization of WT and A128T INs induced by ALLINIs correlates with the differences in infectivity of HIV-1 progeny virions. The treatment of producer cells with the inhibitors impairs WT HIV-1 infectivity, with estimated IC 50 values of ϳ6 M for ALLINI-1 and ϳ0.6 M for ALLINI-2, whereas A128T HIV-1 exhibits marked resistance to these compounds ( Fig. 8 and supplemental Fig. S8). In turn, these results correlate well with the inhibitory activities of these compounds with respect to WT and A128T HIV-1 replication in spreading assays (supplemental Table 2) (14).
In conclusion, our findings that the A128T substitution did not significantly alter ALLINI IC 50 values for IN-LEDGF/p75 binding but substantially affected IN multimerization in the presence of the inhibitors indicate that allosteric IN oligomerization is the primary target of these inhibitors in infected cells. Our structural data showing that the A128T substitution repositions the quinoline ring of ALLINIs at the IN dimer interface provide a path for rationale development of second-generation ALLINI compounds with decreased potential to select for drug resistance.