Dolutegravir (S/GSK1349572) Exhibits Significantly Slower Dissociation than Raltegravir and Elvitegravir from Wild-Type and Integrase Inhibitor-Resistant HIV-1 Integrase-DNA Complexes

ABSTRACT The integrase inhibitor (INI) dolutegravir (DTG; S/GSK1349572) has significant activity against HIV-1 isolates with raltegravir (RAL)- and elvitegravir (ELV)-associated resistance mutations. As an initial step in characterizing the different resistance profiles of DTG, RAL, and ELV, we determined the dissociation rates of these INIs with integrase (IN)-DNA complexes containing a broad panel of IN proteins, including IN substitutions corresponding to signature RAL and ELV resistance mutations. DTG dissociates slowly from a wild-type IN-DNA complex at 37°C with an off-rate of 2.7 × 10−6 s−1 and a dissociative half-life (t1/2) of 71 h, significantly longer than the half-lives for RAL (8.8 h) and ELV (2.7 h). Prolonged binding (t1/2, at least 5 h) was observed for DTG with IN-DNA complexes containing E92, Y143, Q148, and N155 substitutions. The addition of a second substitution to either Q148 or N155 typically resulted in an increase in the off-rate compared to that with the single substitution. For all of the IN substitutions tested, the off-rate of DTG from IN-DNA complexes was significantly slower (from 5 to 40 times slower) than the off-rate of RAL or ELV. These data are consistent with the potential for DTG to have a higher genetic barrier to resistance, provide evidence that the INI off-rate may be an important component of the mechanism of INI resistance, and suggest that the slow dissociation of DTG may contribute to its distinctive resistance profile.

Improvements in antiretroviral therapy have resulted in major advances in longevity and quality of life for HIV-infected patients. However, there is still a need to augment the anti-HIV armament: for example, by increasing tolerability and ease of dosing, maximizing potency, providing forgiveness in the face of adherence difficulties, and improving resistance profiles. The latest addition to anti-HIV therapy is inhibition of the essential (31) HIV integrase (IN) enzyme (reviewed in references 1, 4, 7, 17, and 33). IN carries out two key catalytic processes: 3Ј end processing, which removes the final 2 nucleotides (e.g., 5Ј-GT) from each end of the viral cDNA, and insertion of the two viral 3Ј DNA ends into opposite strands of the host DNA (strand transfer). Several compounds that specifically block strand transfer of the viral cDNA into the host DNA have been shown to be efficacious in vivo (reviewed in references 34 and 36). Raltegravir (RAL), which was approved by the FDA in 2007, was the first marketed IN inhibitor (INI), and elvitegravir (ELV) and dolutegravir (DTG; S/GSK1349572) are in late stages of development (Fig. 1). All three molecules have a common two-metal-binding motif by which the two essential metals at the IN catalytic site are bound by the inhibitor.
Resistance is typically the primary chink in an antiretroviral agent's armor. Resistance mutations have been identified for both RAL and ELV during in vitro passaging experiments (20,21,30,40,43,47) and in clinical studies (5,35,36,45). Primary resistance mutations observed within the IN open reading frame include substitutions at T66 (ELV), E92 (ELV), Y143 (RAL), Q148 (both drugs), and N155 (both drugs) (reviewed in references 3 and 36). Multiple secondary mutations which may increase resistance and/or compensate for fitness defects caused by the primary mutations have also been identified. Overall, both RAL and ELV largely share resistance profiles (3,30,36,37), and it was demonstrated that subjects experiencing virologic failure during therapy with ELV did not respond to treatment with RAL (10). Understanding resistance mechanisms may practically assist in optimizing the design and development of new drugs with improved resistance profiles, help define how to best use anti-HIV agents in the clinic, and inform on the potential that HIV might require multiple mutations to achieve resistance for a drug which may in turn be predictive of a higher barrier to resistance in vivo.
Characterizing the mechanisms of HIV drug resistance can involve many methods, including passage of virus in cell culture in the presence of a drug to select resistance, examination of resistance profiles observed in virus isolates from subjects experiencing virologic failure in the clinic, evaluation of the relative contributions of mutations to resistance and to resto-ration of viral fitness, and finer-level studies involving biochemical measurements and analysis of molecular structure using crystallography. Of note, biochemical studies have suggested that binding of INIs to HIV-1 IN proceeds by a two-step mechanism with a slow second step, and mutations that increase resistance may alter this mechanism (19,32). In addition, fast INI dissociation has been proposed to contribute to INI resistance with the N155H, T66I, and Q148R IN substitutions (14,23).
DTG was designed to have excellent potency, low-mg dosing given once daily without pharmacokinetic boosting, and an improved resistance profile with a potential for a higher barrier to resistance (29). DTG has shown potent antiviral activity in multiple cell types and cell-based assay formats (25,30) and has demonstrated potent in vivo efficacy in clinical studies to date (39,42). Extensive in vitro antiviral profiling has shown that DTG has a distinct resistance profile compared to other INIs, with significant antiviral activity against virus with RAL and ELV resistance IN mutations (30). For example, DTG was shown to have wild-type or near wild-type activity in vitro against site-directed mutants with the single primary resistance mutations observed in vivo at Q148 (including Q148H/K/R), at N155, and at Y143 (25,30) and against clinical isolates harboring genotypes with the Y143 and N155 pathway mutations (46) alone or with additional secondary IN mutations. DTG has also demonstrated in vivo activity in subjects with RAL resistance in the VIKING study (16). As described herein, we investigated the dissociation of DTG, RAL, and ELV from wild-type and mutant IN proteins complexed with DNA to obtain a better understanding of INI dissociation kinetics and the relationship between dissociation rates and INI resistance.

MATERIALS AND METHODS
Reagents. DTG, RAL, and ELV were synthesized and iodinated at Glaxo-SmithKline and then tritium labeled by Quotient Bioresearch (Cardiff, United Kingdom) or Tritec (Teufen, Switzerland), resulting in specific activities of 27 Ci/mmol, 23 Ci/mmol, and 18 Ci/mmol, respectively. Streptavidin-coated scintillation proximity assay (SPA) imaging beads were purchased from PerkinElmer (Boston, MA). Full-length HIV-1 BH10 wild-type and mutant IN enzymes were expressed and purified as described previously (19) with or without 2-mercaptoethanol in the lysis and extraction buffers. Site-directed mutagenesis was used to make mutations in the wild type HIV-1 BH10 IN sequence (41). The oligonucleotides 5Ј-biotin-ACCCTTTTAGTCAGTGTGGAAAATCTCTAGCA (plus strand, mimics 3Ј processed end with GT removed) and 3Ј-GAAAATCA GTCACACCTTTTAGAGATCGTCA (minus strand) were obtained from Invitrogen (Carlsbad, CA). Dissociation studies. Dissociation experiments with INIs and IN-DNA complexes were performed by methods similar to published methods (32). Viral long terminal repeat DNA duplexes were prepared by heating 50 M 5Ј-biotinylated, 3Ј-processed plus-strand DNA and 50 M minus-strand DNA at 95°C for 5 min in 50 mM 3-(N-morpholino)propanesulfonic acid (MOPS [pH 7.2]), 50 mM NaCl, and 10 mM MgCl 2 , followed by cooling at room temperature for several hours. The DNA duplexes were attached via the 5Ј-biotin linker on the plus strand to streptavidin-coated SPA imaging beads by incubating 2 M biotinylated DNA and 20 mg/ml beads in 25 mM MOPS (pH 7.2) on a Nutator mixer for 80 min at room temperature. Unbound DNA was removed by two rounds of centrifugation and resuspension of the DNA-bead complex in 50 mM MOPS (pH 7.2), 50 mM NaCl, and 10 mM MgCl 2 . The final bead concentration was 40 mg/ml, and the DNA-bead complex was stored at 4°C until needed. 2), 23 mM NaCl, 10 mM MgCl 2 , and 8% DMSO was added to the high-signal and high-signal-control wells. Dissociation of 3 H-labeled INIs at 37°C was monitored for up to 3 weeks using a ViewLux charge-coupled-device imager (PerkinElmer), with the plate being maintained in a 37°C incubator between time points. The signals for duplicate wells were averaged, and the signal from control wells, which represents nonspecific binding of the 3 H-labeled INI to the DNA-bead complex, was subtracted from the high signal (the high-signal control was DNA-bead, 40 nM 3 H-labeled INI, and 0.5% DMSO) and the dissociation wells (the dissociation control was DNA-bead, 40 nM 3 H-labeled INI, 40 M INI, and 0.5% DMSO). Relative binding was calculated as the ratio of the background-subtracted dissociation signal to background-subtracted high signal and was fit with the equation RB ϭ EP ϩ ⌬RB (e Ϫk off t ), using the SigmaPlot (version 8.0) program (Systat Software, Chicago, IL), where RB is the relative binding at time t, EP is the relative binding endpoint, ⌬RB is the change in relative binding, and k off is the rate constant for dissociation. The endpoint was allowed to float, except for cases of extremely slow dissociation, when it was fixed at 0.1. Dissociative half-life (t 1/2 ) values were calculated from (ln2)/k off using the mean k off value from multiple experiments. Statistical analysis was performed with the JMP (version 9) program (SAS Institute, Cary, NC). P values were calculated from the log 10 (k off ) using all of the individual off-rate determinations and the least-squares means differences Student's t test (␣ ϭ 0.05, t ϭ 1.9835).
Strand transfer assay. To determine the catalytic stability of the IN-DNA complex, IN-DNA-bead complexes were formed as described above for wildtype, E92Q, Y143R, Q148H, N155H, E92Q/N155H, E138K/Q148R, and G140S/ Q148H INs. After resuspension of the beads at 1.5 mg/ml, the IN-DNA-bead solution was put into a 37°C incubator and strand transfer activity was determined at various time points. The strand transfer assay was conducted as described previously (2), except that DMSO was omitted and the reaction was quenched after 30 min with 25 mM MOPS (pH 7), 50 mM EDTA, and 500 mM NaCl.   (Table 1). Both RAL and ELV dissociated more quickly (P Ͻ 0.0001) than DTG (Fig. 3A), with k off values of (22 Ϯ 2) ϫ 10 Ϫ6 s Ϫ1 for RAL and (71 Ϯ 4) ϫ 10 Ϫ6 s Ϫ1 for ELV ( Table 1). The dissociative t 1/2 calculated for DTG was 71 h, which was 8 times longer than that of RAL (t 1/2 , 8.8 h) and 26 times longer than that of ELV (t 1/2 , 2.7 h) (     (30). In these studies, INI potency against HIV-1 isolates harboring site-directed NL432-based IN molecular clones was assayed in HeLa-CD4 cells and the fold change (FC) in the INI 50% effective concentration (EC 50 ) versus that for wildtype HIV-1 was determined. For comparison purposes, the dissociative half-life values calculated from k off and the published in vitro antiviral data are listed together in Table 2.

INI dissociation from wild-type IN
There is no direct correlation between dissociation, antiviral potency, and resistance (Fig. 5) (30). Consistent with the decreased antiviral potency for ELV with the Q148K/R and E92Q/N155H, E138K/Q148R, and G140S/Q148H substitutions (FCs, 240 to Ͼ1,700), increasing the concentration of [ 3 H]ELV 4-fold in the dissociation experiments did not result in higher signals relative to background (data not shown), suggesting that these substitutions may reduce the binding affinity of ELV with these IN-DNA complexes. Overall, there is a qualitatively inverse relationship between the dissociative half-life and in vitro antiviral potency for DTG, RAL, and ELV (Fig. 5). This observation is consistent with results obtained with other INIs (13). Empirical observations based on our data suggest that in vitro resistance (FC in EC 50 , Ն3 versus wild type [30]) was not observed for IN-DNA complexes/INIs with a dissociative half-life of greater than 4 h and that pronounced in vitro resistance was generally observed for IN-DNA complexes/ INIs with a dissociative half-life of less than 1 h.

DISCUSSION
Our goal was to compare INI dissociation rates with wildtype and INI-resistant IN-DNA complexes for three INIs that have demonstrated efficacy in clinical studies. Included in this work were DTG (now in phase 3 clinical studies) and the two earlier INIs (RAL, the first FDA-approved INI, and ELV, which is also currently in phase 3 clinical studies). Resistance mutations observed in the clinic during treatment with RAL or ELV were used as a basis to select the IN substitutions for analysis, as no treatment-naïve subject has yet developed resistance to DTG.
For RAL, two main pathways for resistance involve signature mutations which are initially observed at amino acids Q148 and N155 and are almost invariably found with secondary mutations which may increase resistance and impact viral fitness and catalytic efficiency (5,11,18,38,40). With additional viral replication, mutations may be selected at Y143C and Y143R (12,44), which then may outcompete the original mutant viruses. Studies have shown that the Y143, Q148, and N155 primary mutations or the addition of Q148 and N155 pathway secondary mutations has less impact on the in vitro enzyme and antiviral potencies of DTG than those of RAL and ELV (25,30,48). In our experiments, DTG demonstrated significantly slower dissociation from all IN-DNA complexes than RAL and ELV. The Q148H substitution caused the greatest fold increase in k off for all three INIs. While the addition of the Q148 and N155 secondary mutations did cause an additional increase in k off for DTG compared to the primary substitutions, the effect was less pronounced than it was for RAL. DTG maintained prolonged binding even with the IN-DNA complex containing G140S and Q148H, mutations that are frequently observed in RAL resistance during treatment (36) but have limited impact on the in vitro potency of DTG (25,30). The in vitro biochemical and antiviral data generated with DTG suggest that an accumulation of IN mutations may be required in these RAL signature resistance pathways to have effects on DTG binding and potency similar to those observed for RAL and ELV.
The crystal structures of prototype foamy virus (PFV) IN in complex with DNA and RAL, ELV, (24), or DTG (25) and HIV-1 IN structural modeling studies (8,9) provide insight into INI binding modes and how substitutions in the active site could impact INI binding. For RAL, the co-crystal structures show a key -stacking interaction between RAL's oxadiazole and the side chain of Y212 in PFV IN, which corresponds to Y143 in HIV-1 IN. Since substitution of Y143 with His, Cys, or Arg likely compromises this interaction, RAL's dissociation rates are expected to increase, with the Y143R substitution having the greatest effect, given its flexibility and formal charge. The co-crystal structures with ELV and DTG reveal that these INIs make only limited van der Waals contact with Y143, suggesting that ELV and DTG dissociation rates would be minimally impacted by Y143 substitutions, which is consistent with our dissociation results. From a comparison of the PFV IN-DNA structures, the S217H and N224H (Q148H and N155H, respectively, in HIV-1 IN) substitutions appear to alter the architecture of the structural and catalytic components of the IN active site, which, moreover, appear to perturb the binding of the INIs. We speculate that Q148 may play a role in stabilizing the HIV-1 IN active-site loop into a catalytically active state and that His, Lys, or Arg substitutions at this position may alter the loop, perhaps to a greater extent than is evident in the PFV IN structures, given the added flexibility of the HIV-1 IN loop imparted by G140 (8). Also, the N155H substitution may alter the base of the HIV-1 IN catalytic pocket, the placement of at least the Mg 2ϩ ion coordinated to E152, and the structure of that portion of the ␣4 helix forming one side of the pocket, causing a minor displacement of the INIs within the pocket. Altogether, these changes would be expected to increase k off for all three INIs, which is in fact observed. Consistent with our findings, an altered loop conformation or a displacement of the coordination complex would be expected to have the greatest negative impact on RAL binding, given the possibility of disrupting its -stacking interaction with Y143. Additionally, the structural (8) and electronic (9) characteristics of DTG's metal-binding scaffold may contribute to the slower dissociation kinetics of DTG than RAL and ELV. Overall, the dissociation data are consistent with the crystallography and structural modeling results and suggest that key aspects of both the IN active site and the INIs may contribute to the observed differences in INI dissociation rates from the IN-DNA complexes.
The dissociation rate of a drug from its target typically is a major component of residence time, which can be defined as the period during which the ligand (drug) is bound to its receptor (6). A longer residence time on the targeted receptor theoretically provides beneficial characteristics; this is because the ligand has a greater opportunity to have an effect. However, the relationship between in vitro measurements of dissociation and in vivo efficacy is often qualitative rather than quantitative; ligand and receptor effective concentrations may vary substantially during the course of dosing and as a function of ongoing biological processes. As such, evaluation of the in vitro and in vivo relationship should take into account the window or duration of the pharmacological effect. In the case of HIV, there is a window of opportunity for many processes, including those targeted by current anti-HIV agents. For IN strand transfer inhibitors such as the INIs studied here, the window during which binding most likely occurs is between 3Ј processing (which generates structural components of the catalytic binding pocket [15,26] of the HIV cDNA genome ends) and integration into the host genome. Interestingly, in vitro washout experiments found an association between an extended INI dissociative half-life and a more persistent antiviral effect, apparently via the irreversible generation of unintegrated viral cDNA (13 lead to higher levels of resistance. In the case where these mutations are the first step in a sequence of mutations required to achieve high-level resistance, prolonged INI binding theoretically should provide an improved potential for a higher barrier to resistance. The comparison of dissociative half-life and antiviral potency suggests that there might be a threshold residence time for INIs which impacts antiviral efficacy in the cellbased system used to determine antiviral potency (30). From our data set, this threshold would be a dissociative half-life of between 1 and 4 h, but additional data would be needed to further refine this relationship and to identify exceptions. It is possible that the threshold is related to the length of the window during HIV replication when IN activity is required. One could speculate that, in particular during periods of nonadherence when drug levels may reach suboptimal levels, an INI with a longer dissociative half-life might remain bound and prevent integration. Additional experiments with systems in which this integration window is significantly longer, such as primary blood-derived monocytes, might be illuminating. In addition, time-of-removal or time-of-addition experiments might reveal whether DTG has a prolonged antiviral effect upon washout similar to what was observed for BMS-878397 compared to RAL (13).
The dissociation data presented here provide additional evidence of differential binding of INIs to wild-type and mutant IN-DNA complexes and suggest that prolonged INI binding may contribute to efficacy. While dissociation kinetics and compound residence time are likely to be important for INIs, other factors such as the rate of compound association (48), drug pharmacokinetics, and viral fitness could also impact efficacy and resistance. Additional work is necessary to elucidate the contribution of each of these components to the emergence of INI resistance. However, DTG's long dissociation half-life with the wild-type and INI-resistant IN-DNA complexes may contribute to its distinct resistance profile and highlight the potential for improved activity against wild-type HIV-1 and clinically relevant INI-resistant viruses. Among the potential important predictions from the preceding discussion is that DTG might have a higher capacity to suppress viral replication should drug levels become suboptimal due to adherence problems (i.e., possess "forgiveness") and that the slower dissociation may translate to a higher genetic barrier to resistance in vivo. Ultimately, however, such predictions will need to be verified by data from the phase 2b and phase 3 clinical studies which are currently ongoing.