Biological and Structural Analyses of New Potent Allosteric Inhibitors of HIV-1 Integrase

ABSTRACT HIV-1 integrase-LEDGF allosteric inhibitors (INLAIs) share the binding site on the viral protein with the host factor LEDGF/p75. These small molecules act as molecular glues promoting hyper-multimerization of HIV-1 IN protein to severely perturb maturation of viral particles. Herein, we describe a new series of INLAIs based on a benzene scaffold that display antiviral activity in the single digit nanomolar range. Akin to other compounds of this class, the INLAIs predominantly inhibit the late stages of HIV-1 replication. A series of high-resolution crystal structures revealed how these small molecules engage the catalytic core and the C-terminal domains of HIV-1 IN. No antagonism was observed between our lead INLAI compound BDM-2 and a panel of 16 clinical antiretrovirals. Moreover, we show that compounds retained high antiviral activity against HIV-1 variants resistant to IN strand transfer inhibitors and other classes of antiretroviral drugs. The virologic profile of BDM-2 and the recently completed single ascending dose phase I trial (ClinicalTrials.gov identifier: NCT03634085) warrant further clinical investigation for use in combination with other antiretroviral drugs. Moreover, our results suggest routes for further improvement of this emerging drug class.

are investigated in this paper (Fig. 1B). The BDM-2 series of INLAIs is based on a benzene scaffold and the compounds contain either the carboxyl and tert-butoxy groups (BDM-2 and MUT871) or a derivative of this motif, with the tert-butoxy side chain replaced by a cyclopropyloxy group (MUT872, MUT884, and MUT916). Compound MUT871 differs from BDM-2 by a methyl substituent on the chromane group.  Table 1). The most potent inhibitor of the IN-LEDGF/p75 interaction of all INLAIs tested here was MUT871 with 50% inhibitory concentration (IC 50 ) of 14 nM, compared to 90 and 820 nM for previously described inhibitors of this class BI-224436 and S-I-82, respectively. The remaining compounds from the BDM-2 series, MUT872, MUT884, MUT916, and BDM-2, were also very potent at disrupting the IN-LEDGF/p75 complex formation, yielding IC 50 values in the range between 17 nM for MUT872 and 62 nM for MUT884 ( Table 1). The interaction between isolated HIV-1 IN CCD and LEDGF/p75 IBD was likewise susceptible to the compounds, although BDM-2, MUT871, and MUT872   appeared to be more active at inhibiting full-length IN-LEDGF/p75 complex formation (Table 1).
A key property of INLAIs is the ability to promote HIV-1 IN homo-multimerization (39)(40)(41)(42). Concordantly, BDM-2 and its analog compounds enhanced IN aggregation, with 50% of maximum activation (AC 50 ) obtained in the concentration range like that required for the inhibition of IN-LEDGF/p75 interaction. The highest activity, 20 nM, was observed for BDM-2, compared to 43 nM for MUT872, 34 nM for BI-224436, and 47 nM for S-I-82 (Table 1). For all compounds studied here, except MUT871, the AC 50 of IN multimerization was lower than the respective IC 50 of the IN-LEDGF/p75 interaction. MUT871, for which the AC 50 is slightly higher than that of IC 50 , is our most potent inhibitor of IN-LEDGF/p75 interaction. This result confirms that INLAIs tend to be more potent as IN-IN molecular glues than as inhibitors of IN-LEDGF interaction (12,16). The maximal signal (plateau) in our IN aggregation assay corresponds to the highest HTRF signal increase under saturating INLAI concentration, relative to baseline level. Thus, MUT916, MUT884, MUT871, and MUT872 increased HIV-1 IN aggregation 3.7 to 4.4fold, while BDM-2, BI-224436, and S-I-82 did so 5 to 6-fold ( Table 1).
The BDM-2 series compounds are potent inhibitors of HIV-1 replication. Antiretroviral activity of BDM-2 and its analogs was first evaluated in lymphoblastoid cell line MT4 and in primary human T lymphocytes using NL4-3 and HXB2 laboratory HIV-1 isolates. The inhibitory potencies were compared to that of the INSTIs RAL and DTG, and to the previously described INLAIs BI-224436 and S-I-82 (Table 2). BDM-2 and MUT871 were the most potent INLAIs of this series, with 50% effective concentrations (EC 50 s) 8.7 nM and 3.1 nM against NL4-3, and 4.5 nM and 1.4 nM against HXB2, respectively. The remaining compounds of the BDM-2 series also displayed high potency, with EC 50 s in the range of 6.3 to 18 nM and 15 to 45 nM against HXB2 and NL4-3, respectively (Table 2). In the absence of human serum, BDM-2 and MUT871 were more potent than RAL, while the antiviral activity of MUT871 approached that of DTG (EC 50 of 1.9 nM and 2.7 nM against HXB2 and NL4-3, respectively). All BDM-2 series compounds displayed higher antiviral activity than the previously reported INLAI BI-224436, while BDM-2 matched S-I-82 in potency. Moreover, the compounds had a favorable cytotoxicity profile with 50% cytotoxic concentrations (CC 50 s) in the range of 46 to 139 mM and consequently very high selectivity indexes (Table 2).
EC 90 values for the compounds, inferred from the respective dose response curves, are given in Table 2. We determined the effect of human serum on the antiviral potency by extrapolation of the EC 90 at 100% human serum. This value corresponding to the protein-adjusted EC 90 (PA-EC 90 ), was estimated from the experimental antiviral activities in MT4/NL4-3 infection assays in the presence of 10 to 50% human serum. Thus, PA-EC 90 values of BDM-2 and MUT871 were estimated as 127 nM (corresponding to 50 ng/mL) and 98 nM (40 ng/mL), respectively (Table 2). PA-EC 90 is an important To dissect the mechanism of action of the BDM-2 series compounds, we compared their antiviral activities in single-and multiple-round HIV-1 NL4-3 infection assays. Single-round infection assays utilize pseudotyped, replication-defective viral particles (produced in the absence of inhibitors) and report on the ability of a compound to block the early steps of the HIV-1 replication cycle (from entry into host cells to provirus expression). In contrast, multiple-round infection assays use a replication competent virus to assess antiviral activity over multiple complete infection cycles. As expected, INSTIs were highly potent at inhibiting HIV-1 both in single-and multiple-round assays, with similar EC 50 values (Tables 2 and 4). By sharp contrast, all INLAIs tested here, including BDM-2 and MUT871, possessed much greater antiviral activity in the multiple-round assays. Thus, the INLAI EC 50 values determined in single-and multiple-round infection assays were in the micromolar (0.63 to 9.2 mM, Table 2) and nanomolar range (3.1 to 51 nM, Table 4), respectively. These results are in line with previous observations that INLAIs predominantly inhibit the late stages of the HIV-1 replication cycle (12,13,23,31).
Intriguingly, the ratios of EC 50 values determined in the single-and multiple-round assays varied considerably between INLAIs, ranging from ;30 for BI-224436 to ;800 for S-I-82 (Table 4). Specifically, for BDM-2 and its analogs, the EC 50 ratios were in the range of 100 to 200 (Table 4). These variations may reflect differences in the dual biochemical activities of these compounds, namely, in their ability to inhibit IN-LEDGF/p75 interaction versus their action as molecular glues promoting IN aggregation. Indeed, there is a striking correlation between the IC 50 value of an INLAI in its ability to disrupt the IN-LEDGF/p75 complex (Table 1) and its antiviral potency in single-round infection assay ( Table 4). The INLAI compound with the best IC 50 in disrupting IN-LEDGF/p75 interaction, MUT871, is also the most potent in the single-round assay. The weakest compound to inhibit IN-LEDGF/p75, S-I-82, is also the weakest in antiviral activity in single-round infection assays.
Antiretroviral activity of the BDM-2 series on polymorphic recombinant and primary HIV-1 isolates. Located on the HIV-1 IN CCD dimerization interface, the principal INLAI binding pocket overlaps with a hot spot of amino acid sequence polymorphism involving IN residues 124 and 125 (16). To test if the new INLAIs maintained antiviral activity against major polymorphisms, we introduced all major combinations of residues found in circulating HIV-1 strains at positions 124 and 125 into NL4-3 HIV-1 molecular clone. In total, we evaluated 15 such combinations that collectively covered 98% of all HIV-1 clade  New Highly Potent Allosteric Inhibitors of HIV-1 Integrase Antimicrobial Agents and Chemotherapy polymorphisms involving these positions (48). To gauge effects on the antiviral activity, we calculated the EC 50 fold change (FC) relative to wild type NL4-3 HIV-1 strain that features Thr124 and Thr125 (TT; specific polymorphisms are referred to by a single letter amino acid code). BDM-2 and its analogs maintained their potency against almost all 15 polymorphisms tested, with the corresponding FC values of 1 (Table 5). Particularly, BDM-2 displayed an FC of 1 for all polymorphisms except AV, which increased the FC to 2. MUT871 likewise performed well, with only the GT, GA, and NA polymorphisms increasing the FC to 2, 3, and 2, respectively. For MUT916, the only polymorphism with FC .1 was NA, with an FC of 4. The FC of 3-fold or less can be considered modest, demonstrating that this BDM-2 series is associated with broad polymorph coverage. By sharp contrast, S-I-82 lost most of its antiviral activity against HIV-1 strains with an Asn residue at position 124, displaying an FC of 10 for NA, and an FC of 5 for NT and NV viruses. Next, we evaluated the potency of the BDM-2 series of INLAIs against several primary HIV-1 isolates with polymorphisms at IN positions 124 and 125. As shown in Table 6, BDM-2 compounds maintained excellent antiviral activity against all 14 primary isolates tested, with corresponding FCs of 1 to 3, except for primary isolate vGA, which was inhibited by BDM-2 and analogs with FCs of 4 to 5. In addition, MUT916 lost substantial activity against vNA_2. In agreement with the results obtained with recombinant NL4-3 variants, S-I-82 displayed more severe losses against polymorphic primary HIV-1 isolates, particularly against primary isolates featuring NA or NV polymorphisms with FCs of 210, 23, 23, 34, respectively, and a FC of 12 for the primary isolate vSA_1.
BDM-2 is fully active on viruses resistant to all classes of current clinical antiretrovirals. Next, we wished to confirm that BDM-2, as a representative lead compound of this new series of INLAIs, retained activity against HIV-1 isolates that have acquired     Table 7). Selection of HIV-1 resistance to BDM-2. To evaluate the genetic barrier to BDM-2 resistance, we propagated the NL4-3 HIV-1 strain under escalating compound levels. In parallel, we followed selection of resistance to three reference compounds: a clinical INSTI (RAL), a clinical NNRTI (nevirapine [NVP]), and an unrelated INLAI (BI-224436). The progress of virus replication in the presence of stepwise increasing concentrations of inhibitors was monitored over time (Fig. 2). The emergence of phenotypic resistance to BDM-2 and BI-224436 was similar to that of NVP and occurred slightly faster than to RAL. These results confirmed that INLAIs display relatively low genetic barrier to resistance, which is like that of NNRTIs and likely lower than INSTIs.
To monitor emergence of genotypic resistance, we sequenced entire IN-coding regions within viral genome populations at various time points during the selection process ( Table  8). The occurrence frequency of individual missense mutations was estimated as a percentage of the total number of missense mutations observed within the viral culture at each time point. Early changes detected at day 18 were amino acid substitutions close to the INLAI binding site in the IN-CCD: A128T, the most frequent one found in 50% of IN sequences, and Y99H found in 36% of IN sequences. Interestingly, a low level (7%) of N222K mutation in the CTD domain was also detected at that time point. At the intermediate time point, 42 days postinfection, the landscape of INLAI resistance mutations did not change drastically. The most detrimental mutations such as T174I, and to a lesser degree A129T, occurred late during the selection process, between days 66 and 73, while previously detected mutations A128T and N222K were still present in IN sequences. Identical mutations with similar kinetics were also detected during the selection process with BI-224436 ( Fig. 2 and Table 8).
To confirm that the observed genetic changes are responsible for the observed phenotypic resistance, we introduced major single point mutations identified during selection into NL4-3 HIV-1 clone. The resulting mutants were tested for sensitivity to the compound ( Table 9). The most detrimental mutation for all INLAIs, including BDM-2, was T174I with FC values of 253, 706, and 15 for BDM-2, BI-224436, and S-I-82, respectively. The remaining single point mutations detected during the selection process, namely, Y99H, A128T, H171Q, and N222K conferred much weaker resistance with the corresponding FC values in the range of 2 to 4. All INLAIs investigated shared a similar resistance profile to that of compounds of the BDM-2 series. These results are consistent with the resistance profile for previously reported INLAIs. As expected, EVG as an INSTI representative of current drugs used in clinic, maintained full activity against all INLAI-resistant mutants.
We also analyzed the impact of the INLAI resistance mutations on viral fitness. The most detrimental mutation for INLAI activity, T174I, also had the greatest negative impact on replication fitness, reducing it to ;29% of that of wild-type control (Fig. S1). All remaining mutations, Y99H, N222K, A128T, and H171Q, that resulted in moderate INLAI resistance had moderate impact on replication fitness (Fig. S1).
No antagonism between BDM-2 and antiretroviral drugs currently used in clinic. All efficient treatments against HIV-1 infection are combination therapies of several (typically three for the highly active antiretroviral therapy, HAART) antiretroviral drugs with orthogonal modes of action. Therefore, it was important to confirm that BDM-2, as the lead compound of the new series of INLAIs, did not antagonize other antiretroviral drugs. To this end, we evaluated the activity of BDM-2 in the presence of approved antiretroviral drugs against NL4-3 HIV-1 clone replicating on MT4 cells. Examples of MacSynergy plots obtained for the combinations of BDM-2 with the protease inhibitor    (45), the asymmetric units of the two-domain cocrystal structures each contained a single CCD dimer with a pair of associated CTDs. Every CCD-CTD interface buried a molecule of INLAIs, well-defined in an electron density map (Fig. 4C).
As shown for all other INLAIs (10,12,14,21,31), the small molecules engage the wellcharacterized pocket on the CCD dimerization interface, also implicated in the interaction with LEDGF/p75 IBD (38). The contacts made by all compounds within the CCD pocket are very similar, regardless of the IN construct used (Fig. S3). The INLAIs are anchored through hydrogen bonds with main chain amides of Glu170 and His171, and additionally with Thr170 side chain (Fig. 5, Fig. S2 and S3). In addition, the INLAIs engage in extensive hydrophobic interactions with IN CCD residues Leu102, Thr124, Thr125, Ala128, Ala129, Trp132, Ala169, Gln168, and Met178 (Fig. 5, Fig. S2 and S3). In particular, the tert-butoxy groups of BDM-2 and MUT871 make van der Waals contacts with Thr174, explaining why a substitution to a bulkier Ile residue at this position causes high-level resistance to these compounds (Table 9). In contrast, in MUT872, MUT884, and MUT916, the tert-butoxy group is replaced by a more compact cyclopropyloxy group (Fig. 5), which may allow sufficient space to accommodate the Ile side chain, explaining higher activity of these three compounds against T174I HIV-1 variants compared to BDM-2 and MUT871 ( Table 9).
The molecules protrude from the HIV-1 IN CCD cleft to function as molecular glue for the recruiting of the CTD (Fig. 4A and B and Fig. 5). As recently shown for BI-D, STP0404, and BI-224436 (42,45), BDM-2 and the analogs make hydrophobic contacts with the side chains of the CTD residues Trp235, Tyr226, Lys266, and Ile268. The carboxylate groups of the INLAIs form salt bridges with the Lys266 side chain (Fig. 4). However, the interactions of the BDM-2 INLAIs with Tyr226 are mediated by cyclopropyl sidechains, and therefore lack the aromatic stacking component that was prominent in the case of BI-D, which features a bicyclic aromatic scaffold (Fig. 1) (45). Moreover, the compact nature of the benzene scaffold and small size of the cyclopropyl side chain allow the side chain of Trp235 to approach closer to the CCD than in complexes stabilized by STP0404 and BI-D, resulting in a shift of Trp235 Ca atom by 1.4 and 1.9 Å, respectively, concluding in a pronounced pivoting of the entire CTD domain (Fig. S4, Movie S1).

DISCUSSION
In this report, we described biochemical and antiviral properties of BDM-2 and its analogs MUT871, MUT872, MUT884, and MUT916 that collectively comprise a new series of INLAIs. Our results show that these small molecules possess strong anti-HIV-1 activity in T-lymphoblastoid cell lines as well as in primary human T-lymphocytes. Encouragingly, the potency of the BDM-2 series INLAIs is comparable to that of second-generation INSTIs with very high selectivity ratios, demonstrating negligible cytotoxicity (Table 2). Importantly, the compounds retained full activity against a panel of HIV-1 variants resistant to currently used antiretroviral drugs (Table 7). We observed no antagonism between the BDM-2 series INLAIs and the clinically approved antiretroviral drugs and detected synergy with the protease inhibitor lopinavir and INSTIs (Fig. 3).
INLAIs can be affected by naturally occurring amino acid sequence polymorphisms between circulating HIV-1 strains. While the interaction with the CTD involved highly conserved and invariant HIV-1 IN residues (Fig. 5) (45), the lining of the CCD pocket is subject to naturally occurring variability, which can affect susceptibility to INLAIs. In particular, the widespread polymorphisms at HIV-1 IN amino acid positions 124 and 125 can be problematic for INLAIs. Thus, S-I-82 was severely compromised in its ability to inhibit HIV-1 NA variants (Table 5 and 6). Likewise, we previously showed that HIV-1 strains with an Ala residue at IN position 125 were much less susceptible to MUT-A (16). In contrast, BDM-2 and its analogs maintained antiviral activity against an extended panel of 124/125 polymorphic strains (Table 7). Although INLAIs, including the BDM-2 series, form weak hydrophobic interactions with the HIV-1 IN residues 124 and 125 (Fig. 5, Fig. S2C and S3), these contacts are unlikely to be critical for binding to the CCD dimer. Indeed, the 124/125 polymorphisms did not affect the ability of MUT-A to inhibit IN-LEDGF/p75 interaction, which requires strong binding to the CCD pocket (16). Instead, the mutations reduced hyper-multimerization of IN in the presence of MUT-A. Our cocrystal structures with the two-domain HIV-1 IN construct revealed considerable pivoting of the CTD with respect to the CCD dimer depending on the small molecule mediating the interface (Fig. S4, Movie S1). Thus, for some INLAIs, changes at the 124/125 positions may be detrimental to stability of the CTD-CCD interface. It would be of interest to study the details of the CTD-CCD interfaces mediated by INLAIs that are sensitive to 124/125 polymorphisms.
Our results highlight that the ability to inhibit the IN-LEDGF/p75 interaction does not fully correlate with their potency as HIV-1 inhibitors (Tables 1, 2, and 4). We note that such dichotomy can be expected for molecular glues. Indeed, while the former property requires only strong binding of the compound to the IN CCD pocket, the latter depends on formation of a quaternary complex and fitness of the molecular surfaces. The INLAI cocrystal structures with two-domain HIV-1 IN constructs reported here ( Fig. 4 and 5) and elsewhere (42,45) will inform further development of INLAIs with improved antiviral properties. Although INLAIs predominantly function as maturation inhibitors, their ability to inhibit formation of the IN-LEDGF/p75 complex may lead to novel   (1). Debyser and colleagues proposed that inhibitors of the IN-LEDGF/p75 interaction may be useful in a block-and-lock strategy by retargeting HIV-1 integration to sites that are less susceptible to reactivation (29,49). INLAIs which are particularly potent at inhibiting IN-LEDGF/p75 interactions, such as MUT871 and BDM-2 could be considered potential candidates for this approach. Indeed, MUT871 dissociated the IN-LEDGF/p75 complex with an IC 50 of 14 nM (Table 1) and displayed a comparatively high antiviral activity in the single-round assay, with an EC 50 of 630 nM (Table 4).
Our HIV-1 IN-INLAI cocrystal reported here ( Fig. 4 and 5) and elsewhere (45) may inform the design of research tools to inhibit the IN-LEDGF/p75 interaction without promoting IN aggregation. Interestingly, most of the BDM-2 series of INLAIs appeared to inhibit the interaction between full-length HIV-1 IN and LEDGF/p75 with higher potency than the interaction between the respective minimal binding domains (Table 1). In contrast to full-length HIV-1 IN, which forms stable tetramers (24,43,44), isolated CCD is only known to dimerize (50). We speculate that tetramerization of full-length IN may be slightly beneficial for the binding of BDM-2 and some of its analogs. Indeed, KF116, a pyridine-based INLAI, was reported to selectively target HIV-1 IN tetramers (51).
Most BDM-2 resistance mutations selected by the dose-escalation method mapped to the principal INLAI binding pocket on the CCD dimer (Table 9, Fig. 5, and Fig. S2). Explained by the interactions within the pocket, these mutations affect all INLAIs to various extents (10,23,33). Of these, T174I is the most detrimental, leading to a substantial loss in the antiviral activity of BDM-2 (Table 9). Thr174 makes hydrogen bonding and van der Waals interactions with the tert-butoxy moiety present in BDM-2 and shared by all potent INLAIs reported to date (Fig. 5, Fig. S2 and S3). A larger Ile side chain in position 174 is expected to clash with the tert-butoxy moiety explaining the loss of compound binding. Therefore, it is very encouraging that substitution of the tert-butoxy group for a more compact cyclopropyloxy in MUT872, MUT884, and MUT916, reduced the level of resistance associated with T174I mutation (Table 9). These results indicate that the CCD binding functionality of INLAIs may benefit from further development.
The surface of the CTD participating in the INLAI-induced interface with the CCD dimer is involved in multiple functions during the lentiviral integration process, explaining its high level of conservation (45). Strikingly, the triad of the CTD residues directly recruited by INLAIs (Tyr226, Trp235, and Lys266) are conserved in most lentiviruses. Therefore, it is not surprising that INLAI resistance mutations in the CTD have not been described. Our selection experiment revealed a single point mutation outside the CCD, N222K, which enabled a low level of resistance to BDM-2 and its analogs, as well as to BI-224436 (Table  9). Likewise, N222K was previously reported for its association with mild resistance to INLAIs (23,39). Positioned ;14 Å from the bound inhibitor in our two-domain cocrystal structures, Asn222 is not involved in the INLAI-induced CTD-CCD interface. More research is required to explain if N222K has an allosteric affect indirectly reducing CTD surface complementarity with the CCD dimer, or possibly by reducing interactions of IN protomers as part of the hinge region between the a-helical CCD-CTD linker and the CTD.
To date, there is still no INLAI in advanced clinical investigation for proof-of-concept studies in patients infected with HIV-1 in phase IIa and phase III clinical trials. BI-224436 from Boehringer Ingelheim originally was introduced in phase I clinical trial, but this trial was interrupted for an unknown reason (52). Gilead reported in 2017 a very active INLAI compound with a benzothiazole core group, GS-9822 (47); however, this compound provoked renal and urinary bladder toxicity, which precluded further clinical investigation (47). VIIV also developed a new series of very active INLAIs, tetrahydronaphtyridines (53); however, these compounds have not yet reached clinical trials. Recently, ST Pharm (Seoul, South Korea) reported STP0404 (Pirmitegravir), a very promising INLAI (21), which has now completed clinical phase I study (22). Well tolerated and with favorable pharmacokinetics suitable for once-daily low dose regimen, the compound is expected to move onto a phase IIa trial.
After completion of preclinical studies (data not shown), BDM-2 has been further investigated in a single ascending dose phase I clinical trial investigating safety, tolerability, and pharmacokinetics in healthy male subjects completed in 2020. The results of this trial were reported on 17 June 2020 without any serious adverse event and few mild adverse events (reference 20 and manuscript in preparation). Therefore, BDM-2 is together with Pirmitegravir the most advanced INLAI in investigation today in humans, which supports further clinical investigation.
Molecular biology and biochemistry. Epitope-tagged proteins used in IN-LEDGF/p75, CCD-IBD interaction and IN multimerization assays were constructed and purified as described previously (12). HTRF-based CCD-IBD interaction, IN-LEDGF/p75 interaction and IN multimerization assays were performed as described previously (12).
Cell culture. MT4 cells were obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. MT4 cells were grown in RPMI 1640 supplemented with 10% heat-inactivated fetal calf serum and 100 IU/mL penicillin, and 100 mg/mL streptomycin (Invitrogen) to obtain RPMI-complete medium.
Virus strains and recombinant HIV-1 molecular clones. HIV-1 NL4-3 and NL4-3Denv-luc molecular clones were obtained from the NIH AIDS Research and Reference Reagent Program.
Antiviral assay in MT4 cells (multiple round infection assay). MT4 cells growing exponentially at the density of 10 6 /mL were infected with HIV-1 strain NL4-3 at a multiplicity of infection (MOI) of 0.001 for 2 h in the presence of different concentrations of compounds, and the CellTiter-Glo luminescent reagent (Promega) was used to quantify cell viability as described previously (12). To evaluate the effect of the human serum on the antiviral potency on INLAIs and other compounds used as references, we measured their EC 50 and EC 90 from the dose response curve in the presence of various concentrations of human serum between 0% to 50% and linearly extrapolated their value at 100% human serum for PA-EC 90 determination.
Replication-defective HIV assay (single round infection assay). MT4 cells (growing exponentially at the density of 10 6 /mL) were infected with VSV-G-pseudotyped NL4-3Denv-luc at an MOI of 0.0001, and Luciferase expression was quantified after 2 days using the One-Glo luciferase assay (Promega), as described previously (12).
Cytotoxicity assays. Growth inhibition was monitored in a proliferating human T-cell line (MT4) with different concentrations of compounds, using the CellTiter-Glo luminescent reagent (Promega), as previously described (12).
Resistant virus selection. MT4 cells infected with HIV NL4-3 isolate were cultured in the presence of BDM-2 at the EC 50 concentration determined earlier. At each passage, cells from original culture in the presence of inhibitor were mixed with equal amount of no-drug control cells to propagate, and viral replication was monitored by the production of p24 antigen in the supernatant. Inhibitor concentration was gradually increased at each passage. At three different times, early (day 18), intermediate (day 42), and final passage (day 66 to 73), viral RNA was extracted using QIAexpress (Qiagen) and IN sequences were determined by RT-PCR as a bulk. The quantitative estimate of each mutation induced at each passage mentioned above was determined as a percentage of total mutations detected in the IN sequences as a bulk.
Viral replication capacity. HIV-1 recombinant viruses harboring various single point INLAI resistant mutations (A128T, Y99H, N222K, T174I, H171T, L102F, and T124D) in an NL4-3 background were used to infect MT4 cells with equivalent p24 quantity and compared with infection of MT4 cells in same conditions with identical p24 quantity of wild type NL4-3 and of some INSTI-resistant viruses (G140S and Q148H). Replication kinetics of the INLAI resistant variants were compared and viral production was determined by p24 assay daily for 5 days.
Combination antiviral activity assays. BDM-2, as lead compound representative of the series, was used in these experiments. Combination studies were performed in MT4 cells infected with NL4-3, as previously described (3). Multiple concentrations of BDM-2 were tested in checkerboard pattern in the presence and absence of dilutions of 16 representative approved anti-HIV drugs of different classes. Compound combinations were analyzed by calculations to quantify deviation from additivity at the 50% level. Data were analyzed as described by Prichard and Shipman by using the MacSynergy II program (54). Synergy volumes in the range of 225 to 125 define additivity; 250 minor antagonism, 150 minor synergy, and up to 100 and .100 moderate and strong synergy, respectively. Combinations of RBV with AZT or ddI were used as controls for antagonism and synergy, respectively.

ACKNOWLEDGMENTS
We thank Wandrille Ract-Madoux for support; Ibrahima Guillard for administrative assistance; Juliette Nguyen, Roxane Beauvoir, and Elodie Drocourt for assistance in virology; Jean-Michel Bruneau for assistance in biochemistry; Nicolas Levy for help in crystallographic data collection; Isabelle Mallet and the NIH HIV Reagent Program (https://www.aidsreagent.org/), Division of AIDS, NIAID, for HIV-1-drug resistant viruses and HIV-1 primary isolates.