GRL-09510, a Unique P2-Crown-Tetrahydrofuranylurethane -Containing HIV-1 Protease Inhibitor, Maintains Its Favorable Antiviral Activity against Highly-Drug-Resistant HIV-1 Variants in vitro

We report that GRL-09510, a novel HIV-1 protease inhibitor (PI) containing a newly-generated P2-crown-tetrahydrofuranylurethane (Crwn-THF), a P2′-methoxybenzene, and a sulfonamide isostere, is highly active against laboratory and primary clinical HIV-1 isolates (EC50: 0.0014–0.0028 μM) with minimal cytotoxicity (CC50: 39.0 μM). Similarly, GRL-09510 efficiently blocked the replication of HIV-1NL4-3 variants, which were capable of propagating at high-concentrations of atazanavir, lopinavir, and amprenavir (APV). GRL-09510 was also potent against multi-drug-resistant clinical HIV-1 variants and HIV-2ROD. Under the selection condition, where HIV-1NL4-3 rapidly acquired significant resistance to APV, an integrase inhibitor raltegravir, and a GRL-09510 congener (GRL-09610), no variants highly resistant against GRL-09510 emerged over long-term in vitro passage of the virus. Crystallographic analysis demonstrated that the Crwn-THF moiety of GRL-09510 forms strong hydrogen-bond-interactions with HIV-1 protease (PR) active-site amino acids and is bulkier with a larger contact surface, making greater van der Waals contacts with PR than the bis-THF moiety of darunavir. The present data demonstrate that GRL-09510 has favorable features for treating patients infected with wild-type and/or multi-drug-resistant HIV-1 variants, that the newly generated P2-Crwn-THF moiety confers highly desirable anti-HIV-1 potency. The use of the novel Crwn-THF moiety sheds lights in the design of novel PIs.


Results
Antiviral activity of GRL-09510 against HIV-1 LAI and HIV-2 ROD . We first examined the antiviral potency of GRL-09510 against a wild-type HIV-1. The activity of GRL-09510 against HIV-1 LAI was greater with an EC 50 (50% effective concentration) value of 0.0014 µM compared to other clinically available PIs examined including DRV (Table 1), as assessed with the MTT assay using MT-2 target cells. GRL-09510 was also active against HIV-2 ROD with an EC 50 value of 0.0018 µM (Table 1). GRL-09510's cytotoxic profiles were favorable with CC 50 (50% cytotoxic concentration) being 39.0 µM and the selectivity indices proved to be 27,857 and 21,667 for HIV-1 LAI and HIV-2 ROD , respectively ( Table 1). The activity of GRL-09610, an isomer of GRL-09510, against   Table 1. Antiviral activity of GRL-09510, -09610 against HIV-1 LAI , HIV-2 ROD and cytotoxicity against MT-2 cells. MT-2 cells (10 4 /ml) were exposed to 100 TCID 50 of HIV-1 LAI or HIV-2 ROD and cultured in the presence of various concentrations of each PI, and the EC 50 values were determined by the MTT assay. All assays were conducted in duplicate, and the data shown represent mean values derived from the results of two or three independent experiments. *Each selectivity index denotes a ratio of 50% cytotoxicity (CC 50 ) to EC 50 against HIV-1 LAI or HIV-2 ROD .

GRL-09510 exerts strong activity against highly PI-resistant clinical HIV-1 isolates.
In our previous work, we isolated four highly multi-PI-resistant primary HIV-1 strains (HIV-1 MDR/B , HIV-1 MDR/C , HIV-1 MDR/G , HIV-1 MDR/TM ) from patients with AIDS, who had failed then-existing cART regimens after receiving 9 to 10 anti-HIV-1 drugs over 34 to 83 months 17,18 . These primary strains contained 11 to 15 amino acid substitutions in the PR-encoding region, which have been reportedly associated with HIV-1 resistance against various PIs (see Supplemental Table 1). The four different multi-drug resistant clinical isolates (HIV MDR ) used in the assays shown in Table 2 contain various resistance-associated amino acid mutations in the reverse transcriptase (RT) as well as in the PR. All four patients from whom these variants were isolated had received 6 different NRTIs and 1 of the four had received 1 NNRTI. The activity of APV, ATV, and LPV against such HIV MDR strains was significantly compromised as examined in PHA-PBMC as target cells using p24 production inhibition as an endpoint (Table 2). However, GRL-09510 exerted strong antiviral activity with its EC 50 values against those HIV MDR ranging 0.0032~0.0040 μM, the same range of the activity of GRL-09510 against HIV-1 ERS104pre , a wild-type HIV-1 strain isolated from a drug-naïve individual ( Table 2). The antiviral activity of GRL-09510 proved to be the most effective against those HIV MDR s examined compared to the four Food and Drug Administration (FDA)-approved PIs (APV, ATV, LPV, and DRV). We also examined antiviral activity of GRL-09510 against a highly DRV-resistant variant (HIV-1 DRV R 20P ) 19 . This variant was generated using the mixture of 8 highly multi-PI-resistant clinical isolates as a starting HIV-1 source and selected with increasing concentrations of DRV. GRL-09510 maintained its activity against HIV-1 DRV R 20P (EC 50 : 0.025 μM), being more effective than DRV by 12-fold (Table 2). In addition, we determined antiviral activity of GRL-09510 against an ×4-tropic subtype-A strain, R5-tropic subtype-B strain, and a dual tropic subtype-B strain in PHA-PBMCs. As can be seen in Supplemental Table 2, GRL-09510 effectively inhibited the replication of all of the strains employed. GRL-09510 is active against various PI-selected laboratory HIV-1 variants. We also examined GRL-09510 against an array of HIV-1 NL4-3 variants, which had been selected by propagating HIV-1 NL4-3 as a starting strain, in the presence of increasing concentrations (up to 5 µM) of each of 3 FDA-approved PIs (ATV, LPV and APV) in MT-4 cells 20,21 . Such variants had acquired major HIV-1 PI-resistance-associated amino acid substitutions in their PR-encoding region of the viral genome (see Supplementary Table 1). Each variant was highly resistant to its corresponding PI and showed significant resistance with the EC 50 value of >1 µM. GRL-09510 maintained its antiviral activity against all the variants with EC 50 values of 0.0037~0.0048 μM (fold-differences were 1~2 compared to against HIV-1 NL4-3 ) (Table 3). Overall, GRL-09510 generally exerted significantly favorable antiviral activity against various wild-type HIV-1 strains, drug-resistance variants, and HIV-2 strain than other conventional PIs examined (Tables 1-3). The antiviral activity of GRL-09610 was also found compromised to all the PI-resistant variants employed in the present study as compared to that of GRL-09510 (Tables 2 and 3).
In vitro selection of HIV-1 variants resistant to GRL-09510. We next attempted to select HIV-1 variants resistant to GRL-09510 by propagating HIV-1 NL4-3 in MT-4 cells in the presence of increasing concentrations of GRL-09510 as previously described [22][23][24] . As shown in Fig. 2a, HIV-1 NL4-3 almost immediately started to replicate in the presence of APV and RAL and the selection concentration reached at 5 µM at passage 20 and beyond. HIV-1 NL4-3 was also selected in the presence of GRL-09610, the isomer of GRL-09510. As shown in Fig. 2a, HIV-1 NL4-3 started to replicate at around passage 30 and the selection concentration of GRL-09610 reached 5 µM at around passage 43. Under the same condition, HIV-1 NL4-3 was first exposed to 0.003 µM GRL-09510 and underwent 24 passages for selection concentration to reach a 35-fold greater concentration (0.105 µM) than the initial concentration. By around passage 27, the amounts of p24 Gag protein produced in the culture medium was rather modest (up to ~265 ng/ml); however, at around passage 30, at which the virus began continuously failing to propagate in concentrations >0.27 µM and we discontinued the selection at passage 37. Taken together, the emergence   of GRL-09510-resistant variants was significantly delayed, strongly suggesting that GRL-09510 has a substantially high genetic barrier to the emergence of resistant HIV-1 variants. The PR-encoding region of the proviral DNA isolated from infected MT-4 cells was isolated, sequenced at passage 43 under 5 µM GRL-09610 selection, cloned and sequenced at passages 8, 15, 28, and 36 under GRL-09510 selection. The sequences of the region cloned and the % frequency of identical sequences at each passage are depicted in Fig. 2c. HIV-1 NL4-3 at passage 43 (5 µM of GRL-09610) had 5 amino acid substitutions in its PR region; L10F, L33F, M46I, I47V, I50V (Fig. 2b), and 4 substitutions in Gag region; H87Q, R229K, V230I in the Capsid-encoding region, and L1F in p6 region. It was thought probable that most of these amino acid substitutions were associated to HIV-1 NL4-3 's acquisition of resistance to GRL-09610. We then examined GRL-09510-exposed HIV-1 NL4-3 's acquisition of amino acid substitutions in its PR-and gag-encoding regions in more details. By passage 8, I47V substitution had been acquired in 18 of 21 clones. At passage 15, I54V was seen in 14 of 20 clones and I47V, L10F and I84V were seen in 6 of 20 clones. However, at passage 28, I54V was found to have mutated back to the wild-type and all clones had L10F, I47V, and I84V. At passage 28 and beyond, most clones had T91, which is rarely identified in other PI-resistant HIV-1 variants. By passage 36, all clones had acquired L10F, I47V, I84V, and T91A substitutions in the PR-encoding region (Fig. 2c), H87Q, R143K in its Capsid (p24)-encoding region, and L1F in its p6-encoding region. The locations of all of these amino acid substitutions seen in HIV-1 9510 R 36P in its PR are illustrated in Supplemental Fig. 1. It was suggested the presence of these amino acid substitutions might relate to the acquisition of viral resistance to GRL-09510. Thus, we further examined the effects of each of the 4 substitutions, by generating single substitution-carrying HIV-1 variants (HIV-1 L10F , HIV-1 I47V , HIV-1 I84V , and HIV-1 T91A ). Each single substitution-carrying HIV-1 variant showed no significant changes in their susceptibility to GRL-09510 (≤2-fold increase), while the fold-difference of the activity of GRL-09510 against HIV-1 9510 R 36P was 11 as compared to that against HIV-1 NL4-3 (Table 4). These data, taken together, show that the amino acid substitutions might have made HIV-1 resistant to GRL-09510; however, the level of the resistance endowed seemed to be only modest.

HIV-1 9510
R P36 could grow with a high concentration of GRL-09510. Next Table 3).
X-ray crystallographic analysis of PR WT in complex with GRL-09510. The X-ray crystal structure of PR WT in complex with GRL-09510 was solved in the space group P6 1 22 with one PR monomer per asymmetric unit. GRL-09510 was found to bind in two alternate orientations (separated by 180°) to the active site of PR WT dimer as evident from the difference-electron density map shown in Fig. 3a. In order to analyze the hydrogen bonds (H-bonds), hydrogen atoms were added and their orientations were optimized sampling the crystallographic water molecules through the protein preparation wizard in Maestro (v9.0 Schrodinger LLC). As shown in Fig. 3b, the P2-moiety of GRL-09510 forms three strong H-bonds with the backbone amide hydrogen atoms of D29 (two H-bonds with inter-atomic distances: 1.8 Å and 2.9 Å) and D30 (one H-bond with inter-atomic distance: 2.2 Å). One strong H-bond was seen with the backbone carbonyl oxygen atom of G27 with an inter-atomic distance of 2.6 Å. The transition-state-mimic hydroxyl group of GRL-09510 showed one H-bond each with the side chain δ-oxygen atoms of D25 and D25′ with inter-atomic distances of 2.2 Å and 1.5 Å, respectively. The P2′-methoxybenzene moiety of GRL-09510 forms one strong H-bond with the backbone amide hydrogen atom of D30′ with an inter-atomic distance of 2.7 Å. One conserved crystallographic water molecule was seen bridging between GRL-09510 and the backbone amide hydrogen atoms of I50 and I50′ (inter-atomic distances ranging between 1.5 and 2.0 Å). GRL-09510 had multiple VdW contacts with the active site residues of the PR. Especially,   Table 3. Antiviral activity of GRL-09510 and -09610 against laboratory highly conventional-PI-resistant variants. a Amino acids sequence of each variant was described in Supplemental Table 1. b The EC 50 (50% effective concentration) values were determined by using MT-4 cells as target cells. MT-4 cells (10 5 /ml) were exposed to 100 TCID 50 s of each HIV-1, and the inhibition of p24 Gag protein production by each drug was used as an endpoint. All assays were conducted in duplicate or triplicate, and the data shown represent mean values (±1S.D.) derived from the results of two to four independent experiments.

Discussion
GRL-09510, which contains a newly-generated polycyclic non-peptide P2-crown-tetrahydrofuranylurethane (Crwn-THF) and a sulfonamide isostere, potently suppressed the replication of wild-type HIV-1 and HIV-2 with favorable EC 50 values. GRL-09510 maintained its potent antiviral activity against a variety of clinical HIV MDR isolates with EC 50 values ranging from 0.0032 to 0.0040 μM, while the existing FDA-approved HIV-1 PIs examined either failed to suppress the replication of those isolates or required much higher concentrations to inhibit their replication. GRL-09510 also potently inhibited the replication of laboratory-selected HIV-1 PIs-resistant variants with similarly low EC 50 values.
Of note, we have previously demonstrated a few prototypic PIs (Supplemental Fig. 2) that were highly potent against two wild-type HIV-1s (HIV-1 NL4-3 and HIV-1 WT/ERS104pre ) and a multi-drug resistant clinical strain (HIV-1 MDR/G )(Supplemental Table 4) with EC 50 values ranging from 0.5 to 20 nM. However, compared to GRL-09510 that has an EC 50 value of as low as 3.3 nM against an HIV-1 variant (HIV-1 APV R 5µM ) resistant to APV that structurally resembles DRV, all these prototypic PIs were much less potent against HIV-1 APV R 5µM with EC 50 values ranging from 38 to 560 nM. Moreover, GRL-09510 exerted potent activity against a highly DRV-resistant HIV-1 variant (HIV-1 MDRmixDRV R 20P ) 19 with an EC 50 value of 25 nM (only 7-fold difference to its EC 50 against a wild-type HIV-1)( Table 2); however, one (GRL-0519) of those prototypic PIs and DRV lost their activity against HIV-1 MDRmixDRV R 20P by 50-and 68-fold, respectively (Table 2 and Supplemental Table 4). Although in vitro cell culture may not be much reflective of viral selective pressure imposed by PIs seen in clinical settings, our HIV-1 NL4-3 -based selection experiment using GRL-09510 showed that the emergence of GRL-09510-resistant variants was clearly delayed compared to that of APV-or widely-used integrase inhibitor RAL-resistant variants. APV-resistant HIV-1 variants have been reported to contain V32I, I50V, I54L/M, L76V, I84V, and L90M substitutions at their PR region 26,27 . However, except for I84V, none of such APV-resistance-associated amino acid substitutions were identified during the GRL-09510′s selection (Fig. 2a,c). V11I, V32I, L33F, I47V, I50V, I54M, I54L, T74P, L76V, I84V and L89V are known as DRV-resistance-associated substitutions. Among such substitutions, only two substitutions, I47V and I84V, were seen in our selection experiment using GRL-09510 (Fig. 2c). When we generated and examined HIV-1 I47V and HIV-1 I84V , both variants were as sensitive to GRL-09510 as HIV-1 NL4-3 (Table 4). These data strongly suggest that the combination of these two substitutions is associated with the observed moderate level resistance to PIs of HIV-1 9510 R P36 . In our previous studies, selection experiments with certain potent PIs such as TMC-126 and GRL-1398 28 containing a P2′ para-methoxy moiety led to the selection of resistant variants containing a unique A28S substitution in their PR region 17,26 . It is of particular note that the A28S amino acid substitution did not emerge in the present selection experiment with GRL-09510. Intriguingly, when HIV-1 NL4-3 was selected with GRL-0519, which contains the same P2′ para-methoxy group, the A28S substitution did not emerge over 37 passages 20 . GRL-0519 has tris-tetrahydrofuranylurethane (tris-THF) moiety as the P2 ligand and the para-methoxy moiety at the P2′ site, suggesting that the presence of tris-THF prevented the emergence of the A28S substitution. In this regard, the combination of the novel P2 Crwn-THF and P2′ para-methoxy in GRL-09510 appears to have prevented the selection of A28S substitution. 0.024 ± 0.002 (11) HIV-1 L10F 0.0044 ± 0.0002 (2) HIV-1 I47V 0.0036 ± 0.0005 (2) HIV-1 I84V 0.0033 ± 0.0003 (2) HIV-1 T91A 0.0033 ± 0.0004 (2)    The structural analysis using the X-ray crystallographic data of PR complexed with GRL-09510 showed that the P2-Crwn-THF of GRL-09510 forms a strong hydrogen-bond network with the backbone atoms of Asp 29 and Asp 30. In addition, as illustrated in Fig. 3c, the Crwn-THF forms greater VdW and hydrophobic contacts with multiple amino acids in the active site of PR than the bis-THF moiety of DRV. Such greater VdW contacts of GRL-09510 should be at least in part responsible for GRL-09510's more potent anti-HIV-1 activity compared with that of DRV. The two strong CH-O bonds between the Crwn-THF and the backbone carbonyl oxygen atom of G48 with the interatomic distances of 2.4 Å and 2.6 Å (Fig. 3d) should also be responsible for the observed potent activity of GRL-09510 against HIV-1 WT and various drug-resistant HIV-1 variants.
The C β , C γ and C δ atoms of Ile47 (located in the highly flexible flap domain of the protease) are engaged in multiple Van der Waals interactions with the P2-Crwn-THF moiety of GRL-09510 (Fig. 3c). The I47V substitution (mutation of a longer side chain of Ile to a shorter side chain of Val) results in loss of Van der Waals interactions with P2-Crwn-THF due to the absence of C δ atom in Val (HIV-1 9510 R 8P , Fig. 2c). Nevertheless, as shown in Table 4, the HIV-1 I47V did not display resistance against GRL-09510, suggesting that the P2-Crwn-THF moiety of GRL-09510 still restrains the flap domain from major conformational changes that could affect the inhibitor binding. The P2-mono-THF (of APV) or P2-bis-THF (of DRV) due to smaller size compared to P2-Crwn-THF may not retain their contacts with Val47 generated by the I47V substitution. This confirms that the increase in size favors enhanced binding of GRL-09510 compared to APV or DRV. As shown in Table 4, the substitution T91A failed to select resistant strains of HIV-1 against GRL-09510. However, in the presence of I47V, L10F and I84V substitutions, the T91A substitution (HIV-1 9510 R 36P , Fig. 2c) displayed a 11-fold increase in EC 50 . Thr91 is located in the helical domain of the protease (Supplemental Fig. 1), adjacent to Trp6′, and is engaged in multiple polar contacts with Asn88, Arg87 (directly) and Asp29 (indirectly). The T91A substitution may induce structural distortion in the helix generating a conformational "ripple effect" that may destabilize the H-bonds between P2-Crwn-THF moiety of GRL-09510 and the backbone of Asp29. Furthermore, this structural distortion in the helix domain could be worsened by the enhanced Van der Waals interactions between Ala91 (generated by the T91A substitution) and Trp6′. Due to increase in size, GRL-09510 can tolerate both I47V and T91A without any resistance from HIV-1. Thus, the substitutions I47V and T91A require L10F and I84V to cause increased resistance against GRL-09510. Thus, the P2-Crwn-THF moiety of GRL-09510 perhaps not only contributes to the enthalpic component of binding via H-bonds with Asp29 and Asp30 (as seen with DRV) but also contributes to the entropic component via Van der Waals interactions with Ile47 as well as Val47 (I47V) due to its increased size.
The P2-bis-THF moiety of DRV and P2-moieties of the analogs generated based on DRV were structurally found to be within the chemical space of GRL-09510's P2-Crwn-THF moiety. The two fused-ring structures enforce H-bonding with the backbone of both Asp29 and Asp30. However, the P2-mono-THF moiety of APV primarily targets the backbone of Asp30 to favor Van der Waals interactions with Val32 and Ile47. According to the substrate-envelope model proposed by Chellappan et al. 29 , APV, DRV and DRV-analogs (including GRL-09510) are within the substrate-envelope of the wild type protease. But the combination of V32I and I47V substitutions confers resistance against APV, which may not necessarily indicate that the P2-mono-THF moiety of APV is out of the substrate-envelope. Previously, it has been reported by Yedidi et al. 30 , that the accumulation of multiple resistance-associated amino acid substitution leads to an expanded active site cavity of protease. Liu et al. 31 confirmed that such protease variants have an altered and expanded substrate-envelope. The expanded substrate-envelope, due to increase in the chemical space, generates higher penalty on the entropic component of the inhibitor binding but not for substrate binding due to substrate co-evolution. The P2-Crwn-THF moiety of GRL-09510 due to its increase in size as well as improved entropy would be a better inhibitor to probe any of changes in the size of substrate-envelope especially in the S2-binding pocket. In this context, further improvement in the size of the P2′-moiety of GRL-09510 may significantly reduce the selection of resistance mutations by filling the S2′-binding pocket as well.
The present data demonstrate that GRL-09510 has favorable features as a candidate drug for treating patients living with HIV-1 WT and/or HIV-1 MDR , and that the newly introduced Crwn-THF with para-methoxybenzene should be critical for the strong binding of GRL-09510 to PR and should serve as a promising pharmacophore for designing novel PIs. Of note, boosted DRV (DRV/r) monotherapy recently conducted in order to reduce toxicities and costs of the current cART demonstrated that DRV/r monotherapy appears to be a slightly higher risk of serum HIV-1 RNA elevations compared to 2 NRTI plus DRV/r regimens 32 . In this regard, GRL-09510, which has stronger and more favorable antiviral profiles than DRV, may serve as a potential candidate for PI monotherapy.

Cells and viruses.
MT-2 and MT-4 cells were grown in RPMI-1640 culture medium with 10% FCS (JRH Biosciences, Lenexa, MD), 50 unit/ml penicillin, and 100 μg/ml kanamycin. The following HIV-1s were employed for the drug susceptibility assay: HIV-1 LAI , HIV-1 NL4-3 , HIV-2 ROD , HIV-1 ERS104pre 33 , a clinical HIV-1 strains isolated from drug-naive AIDS patients, and HIV-1s which were originally isolated from AIDS patients who had received 9-10 anti-HIV-1 drugs over the 34-83 months and were genotypically/phenotypically characterized as multi-PI-resistant HIV-1 variants 17,18 . Amino acid (AA) substitutions in the PR region of PI-resistant strains compared to the consensus type B sequence (Los Alamos database), or wild-type HIV-1 NL4-3 , were summarized at Supplemental Table 1. Antiviral agents. Two novel PIs, GRL-09510 and GRL-09610, were designed and synthesized by AKG et al. Both compounds were confirmed to be enantiomerically pure. The detailed synthetic methods for the two compounds will be reported elsewhere. Hoffmann-La Roche AG (Basel, Switzerland) kindly provided saquinavir (SQV). Amprenavir (APV) was a gift from GlaxoSmithKline, Research Triangle Park, NC. Lopinavir (LPV) was kindly provided by Japan Energy, Tokyo. Atazanavir (ATV) was a contribution from Bristol-Myers Squibb (New York, NY). Darunavir (DRV) was synthesized as previously described 34 . Raltegravir (RAL) was kindly provided by Dr. Kenji Maeda (NCI/NIH, Bethesda, MD). HIV-1 ME46 and HIV-1 92UG029 were provided by NIH/ AIDS-Reagent Program.
Drug susceptibility assays. The susceptibility of HIV-1 LAI or HIV-2 ROD to various compounds was determined as previously described [20][21][22] . Briefly, MT-2 cells (10 4 /ml) were exposed to 100 TCID 50 of HIV-1 LAI or HIV-2 ROD in the presence or absence of various concentrations of compounds in 96-well plates and were incubated at 37 °C for 7 days. After incubation, 100 μl of the medium was removed from each well, 3-(4,5-dimetylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution (10 μl, 7.5 mg/ml in PBS) was added to each well, followed by incubation at 37 °C for 2 h. To dissolve the formazan crystals, which were generated from MTT by cellular oxidoreductase in living cell, 100 μl of acidified isopropanol containing 4% Triton X was added to each well and the optical density measured in a microplate reader (Vmax; Molecular Devices, Sunnyvale, CA). To determine the sensitivity of clinically isolated HIV-1s to compounds, PHA-PBMC (10 6 /ml) were exposed to 50 TCID 50 of each HIV-1 and cultured in the presence or absence of various concentrations of compounds in 10-fold serial dilutions in 96-well plates. In determining the drug-susceptibility of certain laboratory HIV-1 strains, MT-4 cells were employed as target cells as previously described 23,24 . MT-4 cells (10 5 /ml) were exposed to 100 TCID 50 of drug-resistant HIV-1s in the presence or absence of various concentrations of compounds and were incubated at 37 °C. On day 7, the supernatants were harvested and the amounts of p24 were determined by using a fully-automated chemiluminescent enzyme immunoassay system (Lumipulse f: Fujirebio, Tokyo) [35][36][37] . Drug concentrations that suppressed the production of p24 by 50% (EC 50 ; 50% effective concentration) were determined by comparison with the p24 production in drug-free control well. All assays were performed in duplicate, and EC 50 shown in this report represent average values of two to four independent experiments. PHA-PBMCs were derived from a single donor in each independent experiment. The research protocols described in the present study were carried out in accordance with relevant guidelines and regulations, and were approved in Ethics Committee for Epidemiological and General Research at the Faculty of Life Sciences, Kumamoto University.

In vitro selection of PI-resistant HIV-1 variants.
MT-4 cells (10 5 /ml) were exposed to HIV-1 NL4-3 (500 TCID 50 ) and cultured in the presence of compounds at an initial concentration of its EC 50 value. Viral replication was monitored by the determination of the amount of p24 produced by MT-4 cells. The supernatants were harvested on day 7 and were used to infect fresh MT-4 cells for the next round of culture in the presence of increasing concentrations of each compound. Proviral DNAs obtained from the lysates of infected cells were subjected to sequencing. This selection procedure was carried out until the drug concentration reached 5 μM as previously described [20][21][22] ).
Determination of nucleotide sequences. Molecular cloning and determination of the nucleotide sequences of HIV-1 strains passaged in the presence of anti-HIV-1 agents were performed as previously described 21 . In brief, DNA was extracted from HIV-1-infected MT-4 cells by using the InstaGene Matrix (Bio-Rad, Hercules, CA) and was subjected to cloning, followed by sequence determination. The 1 st -round PCR mixture consisted of 1 μl proviral DNA solution, 10 μl Ex-Taq (Takara Bio Inc., Otsu, Japan), and 10 pmol each of the 1 st PCR primers in a total volume of 20 μl. The 1 st PCR products (1 μl) were used directly in the 2 nd -round PCR, and 2 nd PCR products were purified with spin-columns (MicroSpin S-400; Amersham Biosciences., Piscataway, NJ), cloned directly using pGEM-T Easy vector (Promega, Fitchburg, WI), and subjected to sequencing with a 3130 automated DNA sequencer (Applied Biosystems, Foster City, CA). The detailed PCR conditions used have been summarized in Supplemental Table 5.

Generation of recombinant HIV-1 clones.
To generate HIV-1 clones carrying the desired AA substitutions, site-directed mutagenesis was performed with a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), and the AA substitution-containing genomic fragments were introduced into pHIV-1 NL4-3Sma , which had been created to have a SmaI site by changing two nucleotides (2590 and 2593) of pHIV-1 NL4-3 .
Expression, purification and refolding of PR WT . Expression/purification/refolding of PR WT were performed as described previously 38 . Briefly, the inclusion bodies isolated from E.coli containing PR WT were extracted with 3 M guanidine HCl (GnCl), centrifuged and the supernatant was loaded on Sephadex-200 column that was pre-equilibrated with 4 M GnCl. PR containing fractions were pooled and were further purified by reverse phase column. Fractions were analyzed by SDS-PAGE and the purity was determined to be > 95%. Lyophilized PR WT was dissolved in 1 ml of 50% acetic acid solution and was added drop-wise to 29 ml of refolding buffer (50 mM sodium acetate, pH 5.2, 5% ethylene glycol, 10% glycerol, 5 mM dithiothreitol (DTT), and a 2-fold molar excess of GRL-09510) while stirring on ice. Refolding was continued at 4 °C with constant stirring overnight. The refolded PR WT -GRL-09510 complex was concentrated using Amicon filters (3 kDa molecular weight cut off) by centrifugation at 4,800 g. The final PR concentration was determined to be ~2 mg/ml. Crystallization of PR-GRL-09510 complexes. Hanging drop vapor diffusion method was used for co-crystallization. PR WT -GRL-09510 complex (4 μl) was mixed with 4μl of well solution per drop. Grid screens such as ammonium sulfate, sodium chloride and quick screen (Hampton Research, CA) were used to obtain preliminary crystallization hits. Co-crystals of PR WT -GRL-09510 were obtained within one day at room temperature. Clusters of rod shaped PR WT -GRL-09510 co-crystals were obtained using 0.8 M sodium/potassium phosphate buffer at pH6.9. The clusters were carefully dissociated using micro tools and individual crystals were picked up into nylon loops. Glucose (30%) was used as cryo-protectant for all the co-crystals. Cryo-coated co-crystals were instantaneously frozen in liquid-nitrogen. X-ray diffraction data collection and processing. X-ray diffraction data for PR WT -GRL-09510 were collected at the SER-CAT (southeast regional collaborative access team), beam line 22-ID (wavelength: 1.0 Å) at the Advanced Photon Source, Argonne National Labs, IL. A Rayonix MX300HS detector was used to record the diffraction data at a distance of 225 mm from the crystal. The exposure time for each frame was 1 s with a frame width of 0.5°. Diffraction data were processed and scaled using HKL2000 39 . Processing details are given in Supplemental Table 6.
Structure solution and refinement. Structure solution was obtained using Molecular replacement (MR) as described previously 38 . Briefly, MR was performed using MOLREP 40 through CCP4 41,42 interface with PR WT taken from PDB ID:4HLA as a search model. Structure solution was directly refined using REFMAC5 43 . The initial coordinates for GRL-09510 were prepared by modifying the structure of the PI, TMC-126 taken from the crystal structure, PDB ID:2I4U. GRL-09510 was fit into the electron density using ARP/wARP Ligands 44,45 . Initial refinement libraries for GRL-09510 were obtained from REFMAC. Solvent molecules were built using ARP/wARP solvent building module. After building water molecules, the final model was refined using the simulated annealing method from phenix.refine 46 on the NIH-Biowulf Linux cluster. The root mean square deviation in bond lengths and bond angles significantly improved by geometry-optimized libraries for ligands using the semi-empirical quantum mechanical method of refinement, eLBOW-AM1 47 during refinement in phenix.refine. Details are given in Supplemental Table 6.
Structural analysis. The final refined structures were used for structural analysis. Hydrogen bonds were calculated between the heavy atom and the hydrogen atom by using a distance cutoff values of 3.0 Å measured between the donor and acceptor heavy atoms 48 . Cutoff values for angles were minimum donor: 90° and minimum acceptor: 60°. Hydrogen bonds with a distance of >3.0 Å were considered weak interactions. Vander Waals contacts were calculated between two atoms (one from GRL-09510 and one from PR WT ) with a maximum 3.5 Å distance cutoff.