Fluorine Modifications Contribute to Potent Antiviral Activity against Highly Drug-Resistant HIV-1 and Favorable Blood-Brain Barrier Penetration Property of Novel Central Nervous System-Targeting HIV-1 Protease Inhibitors In Vitro

ABSTRACT

To date, there are no specific treatment regimens for HIV-1-related central nervous system (CNS) complications, such as HIV-1-associated neurocognitive disorders (HAND). Here, we report that two newly generated CNS-targeting HIV-1 protease (PR) inhibitors (PIs), GRL-08513 and GRL-08613, which have a P1-3,5-bis-fluorophenyl or P1-para-monofluorophenyl ring and P2-tetrahydropyrano-tetrahydrofuran (Tp-THF) with a sulfonamide isostere, are potent against wild-type HIV-1 strains and multiple clinically isolated HIV-1 strains (50% effective concentration [EC₅₀]: 0.0001 to ∼0.0032 μM). As assessed with HIV-1 variants that had been selected in vitro to propagate at a 5 μM concentration of each HIV-1 PI (atazanavir, lopinavir, or amprenavir), GRL-08513 and GRL-08613 efficiently inhibited the replication of these highly PI-resistant variants (EC₅₀: 0.003 to ∼0.006 μM). GRL-08513 and GRL-08613 also maintained their antiviral activities against HIV-2_ROD as well as severely multidrug-resistant clinical HIV-1 variants. Additionally, when we assessed with the in vitro blood-brain barrier (BBB) reconstruction system, GRL-08513 and GRL-08613 showed the most promising properties of CNS penetration among the evaluated compounds, including the majority of FDA-approved combination antiretroviral therapy (cART) drugs. In the crystallographic analysis of compound-PR complexes, it was demonstrated that the Tp-THF rings at the P2 moiety of GRL-08513 and GRL-08613 form robust hydrogen bond interactions with the active site of HIV-1 PR. Furthermore, both the P1-3,5-bis-fluorophenyl- and P1-para-monofluorophenyl rings sustain greater contact surfaces and form stronger van der Waals interactions with PR than is the case with darunavir-PR complex. Taken together, these results strongly suggest that GRL-08513 and GRL-08613 have favorable features for patients infected with wild-type/multidrug-resistant HIV-1 strains and might serve as candidates for a preventive and/or therapeutic agent for HAND and other CNS complications.

INTRODUCTION

Approximately 38 million people are living with human immunodeficiency virus type 1 (HIV-1) globally, 1.8 million individuals have been newly infected, and over 690,000 individuals have passed away due to AIDS-associated disorders, reported by the World Health Organization in 2019 (1). In addition, 26 million HIV-1-positive people were treated with combination antiretroviral therapy (cART) in 2019, which has brought about a substantial impact on the universal prevalence of HIV-1 infection and AIDS. Recent analyses have revealed that the life expectancy of the individual who is infected with HIV-1 has been found to be close to that of uninfected people (2–5). Nevertheless, both the eradication of HIV-1 and the cure of HIV-1 infection/AIDS yet remain extremely difficult challenges because of the reservoir of HIV-1 latent in the patient’s tissues/organs, including the central nervous system (CNS) and lymph nodes. Over the years, a number of trials have been conducted toward improving the longstanding deficits of cART (6–10). They include (i) drug-related side effects, e.g., myopathy, lactic acidosis, and lipodystrophy; (ii) the development of HIV-1 infection-related malignancies; (iii) persisting inflammation in patients who are receiving cART; (iv) the incidence of the immune reconstruction syndrome (IRS); and (v) the increasing costs of cART drugs (11–15). Furthermore, we have encountered an additional challenge in recent years: HIV-1-associated neurocognitive disorders (HAND) and other complications in the CNS due to the prolonged life of the patient and insufficient CNS penetration of conventional cART drugs (16–18).

HAND, which encompasses mild neurocognitive disorders (MND), HIV-associated dementia (HAD), and asymptomatic neurocognitive impairment (ANI), is characterized clinically by the following triad: (i) cognitive impairment, (ii) behavioral impairment, and (iii) motor impairment. HAND has been constantly increasing despite the achievement of cART in decreasing the HIV-1 viral load in organs/tissues where cART agents succeed in reaching therapeutically sufficient concentrations (16). This is certainly the case with CNS complications, including HAND, seen in ∼50% of people living with HIV-1 throughout their lives (17, 18), and specialized treatments do not exist for the HAND and HIV-associated CNS complications. The problems related to CNS disorders include compromise of a patient’s quality of life (QOL) and poor adherence to cART drugs. Poor cART adherence leads to increased risk of emergence of drug-resistant HIV-1 variants and patient morbidity/mortality. At the same time, CNS HIV-1 infection may also lead to the spread of the HIV-1 reservoir in the CNS, where the conventional cART drugs cannot reach because of their low blood-brain barrier (BBB) penetration properties (19, 20). Moreover, subtherapeutic concentrations of cART drugs in the CNS may also expedite the development of drug-resistant HIV-1 strains (21, 22). A number of complex factors are thought to contribute to HAND pathogenesis, such as (i) neurotoxins, chemokines, or cytokines secreted by HIV-1-infected monocytic cells, macrophages, microglia, and astrocytes in the CNS (23–26); (ii) neurotoxicity of viral proteins Tat and gp120, which activate caspases and promote the upregulation of the death receptor Fas, resulting in neuronal apoptosis (27–29); (iii) microvascular abnormalities, with infarcts and BBB alterations induced by chronic HIV-1 infection in the CNS (30); and (iv) synthesis of nitric oxide and reactive oxygen species (ROS) produced by HIV-1-infected or -activated cells in the CNS (31, 32). Chronic HIV-1 infection and the following persisting inflammation in the CNS are thought to be the major contributors to HAND pathogenesis. Therefore, the development of novel anti-HIV-1 drugs possessing strong anti-HIV-1 activity and favorable CNS penetration property is critically required.

We have been concentrating on the improvement of novel HIV-1 protease (PR) inhibitors (PIs) exhibiting favorable antiviral potency against highly PI-resistant HIV-1 strains. bis-Tetrahydrofuranylurethane (bis-THF)-containing darunavir (DRV) (33–35) is one such compound and has been approved by the Food and Drug Administration (FDA) since 2006. We have described novel HIV-1 PIs (GRL-04810, -05010, -0739, -10413, -083-13, -084-13, and -087-13) that show both high anti-HIV-1 potency and promising penetration into the CNS estimated by in vitro reconstruction of the BBB (36–40). In the present study, we newly generated and evaluated two CNS-targeting HIV-1 PIs, GRL-08513 and GRL-08613, which include P1-3,5-bis-fluorophenyl or P1-para-monofluorophenyl moiety, respectively, and P2-tetrahydropyrano-tetrahydrofuran (Tp-THF) with a sulfonamide isostere.

RESULTS

Antiviral activity of the CNS-targeting PIs against wild-type HIV-1_LAI and HIV-2_ROD and cytotoxicity.

We evaluated the antiviral activity of GRL-08513 and GRL-08613 (two-dimensional [2D] structures are illustrated in Fig. 1) toward HIV-1_LAI, a laboratory wild-type HIV-1 strain. GRL-08513 and GRL-08613 definitely disrupted the replication of HIV-1_LAI, with EC₅₀s (50% effective concentrations) of 0.0001 and 0.0002 μM, respectively, compared to four Food and Drug Administration (FDA)-approved HIV-1 PIs—amprenavir (APV), atazanavir (ATV), lopinavir (LPV), and DRV (Table 1)—as evaluated with the 3-(4,5-dimetylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay employing MT-2 cells as target cells. Although the cytotoxicity of GRL-08513 and GRL-08613 was moderate, with 50% cytotoxic concentrations (CC₅₀s) of 2.7 μM, the selectivity indices (SI; ratio of CC₅₀/EC₅₀) of GRL-08513 and GRL-08613 proved to be highly desirable, at 27,000 and 13,500, respectively, in MT-2/HIV-1_LAI cells. GRL-08513 and GRL-08613 also showed strong antiviral activity against HIV-2_ROD in the MTT assay, with EC₅₀s of 0.00047 and 0.0026 μM, respectively (Table 1), with greater or comparable SI of 5,745 and 1,038 in MT-2/HIV-2_ROD cells compared to those with other conventional PIs, including DRV (Table 1).

FIG 1.

Structures of CNS-targeting PIs: GRL-08513, GRL-08613, amprenavir, and darunavir.

TABLE 1.

Antiviral activities of GRL-08513 and GRL-08613 against HIV-1_LAI and HIV-2_ROD and cytotoxicity against MT-2 cells^a

Compound	EC50 (μM)		CC50 (μM)	Selectivity index (CC50/EC50)b
	Against HIV-1LAI	Against HIV-2ROD		MT-2/HIV-1LAI	MT-2/HIV-2ROD
GRL-08513	0.0001 ± 0.0001	0.00047 ± 0.00004	2.7 ± 0.3	27,000	5,745
GRL-08613	0.0002 ± 0.0003	0.0026 ± 0.0002	2.7 ± 0.4	13,500	1,038
APV	0.05 ± 0.01	0.44 ± 0.05	45.5 ± 6.0	910	103
ATV	0.005 ± 0.003	0.027 ± 0.009	31.5 ± 2.3	6,300	1,167
LPV	0.032 ± 0.005	0.018 ± 0.006	25.9 ± 3.0	809	1,439
DRV	0.0050 ± 0.0002	0.031 ± 0.002	96.7 ± 9.9	19,340	3,119

^aMT-2 cells (10⁴/mL) were exposed to 100 TCID₅₀s of HIV-1_LAI and cultured in the presence of various concentrations of each PI, and the EC₅₀s were determined by the MTT assay. All assays were conducted in duplicate, and the data represent mean values derived from the results of two or three independent experiments.

^bEach selectivity index represents a ratio of 50% cytotoxicity (CC₅₀) to EC₅₀ against HIV-1_LAI or HIV-2_ROD.

The CNS-targeting PIs exhibit potent antiviral activity against highly multi-PI-resistant clinically isolated HIV-1 strains.

We previously isolated highly multidrug-resistant primary HIV-1 strains (MDRs), termed HIV-1_MDR/B, HIV-1_MDR/C, HIV-1_MDR/G, and HIV-1_MDR/TM, from AIDS patients who revealed failure of a variety of remedies, including 9 or 10 anti-HIV-1 drugs over periods extending from 34 to 83 months (41). These primary MDRs had 11 to 15 PI resistance-associated amino acid (AA) mutations in the PR-encoding region together with multiple reverse transcriptase (RT) inhibitor-resistance-associated AA mutations in the RT-encoding gene (Table 2). Three FDA-approved PIs (APV, ATV, and LPV) exhibited a decreased activity against such MDRs in phytohemagglutinin-activated peripheral blood mononuclear cell (PHA-PBMC)-employing assays using the suppression of p24 (HIV-1 capsid) production in the culture supernatants as an endpoint (Table 3). However, GRL-08513 and GRL-08613 exhibited comparably potent antiviral activities, with EC₅₀s against the MDRs of 0.0009 to ∼0.0032 μM, as they did against the wild-type strain HIV-1_ERS104pre, which was isolated from a treatment-naive patient (Table 3). The fold differences of the potencies of GRL-08513 and GRL-08613 against the MDRs were 0.3 to 1 compared to those against HIV-1_ERS104pre (Table 3). We also examined the antiviral activities of GRL-08513 and GRL-08613 against a highly DRV-resistant variant, HIV-1_MDRmixDRV^R_20P. HIV-1_MDRmixDRV^R_20P was previously generated by us using a mixture of 8 highly multi-PI-resistant clinical isolates as an initial source of HIV-1 and a selection experiment with rising concentrations of DRV (42). Of note, GRL-08513 and GRL-08613 maintained their activity against HIV-1_MDRmixDRV^R_20P (EC₅₀s: 0.0033 and 0.009 μM, respectively), although DRV and other tested FDA-approved drugs remarkably lost their antiviral activity (e.g., 97-fold less activity was seen in DRV [Table 3]).

TABLE 2.

Identified amino acid substitutions in the protease-encoding region of PI-resistant strains

Strain	Amino acid substitution(s) in PRa
HIV-1ERS104pre (wild type)	L63P
HIV-1MDR/B	L10I, K14R, L33I, M36I, M46I, F53I, K55R, I62V, L63P, A71V, G73S, V82A, L90M, I93L
HIV-1MDR/C	L10I, I15V, K20R, L24I, M36I, M46L, I54V, I62V, L63P, K70Q, V82A, L89M
HIV-1MDR/G	L10I, V11I, T12E, I15V, L19I, R41K, M46L, L63P, A71T, V82A, L90M
HIV-1MDR/TM	L10I, K14R, R41K, M46L, I54V, L63P, A71V, V82A, L90M, I93L
HIV-1MDRmixDRVR20P	L10I, I15V, K20R, L24I, V32I, M36I, M46L, L63P, A71T, V82A, L89M
HIV-1ATVR5μM	L23I, E34Q, K43I, M46I, I50L, G51A, L63P, A71V, V82A, T91A
HIV-1LPVR5μM	L10F, M46I, I54V, V82A
HIV-1APVR5μM	L10F, V32I, M46I, I54M, A71V, I84V

^aThe amino acid substitutions identified in the protease-encoding region compared to the consensus type B sequence cited from the Los Alamos National Laboratory database in HIV-1_ERS104pre, HIV-1_MDR/B, HIV-1_MDR/C, HIV-1_MDR/G, HIV-1_MDR/TM, and HIV-1_MDRmixDRV^R_20P are shown. The amino acid substitutions identified in the protease-encoding region compared to the wild-type HIV-1_NL4-3 in HIV-1_ATV^R_5μM, HIV-1_LPV^R_5μM, and HIV-1_APV^R_5μM are also shown.

TABLE 3.

Antiviral activities of GRL-08513 and GRL-08613 against multidrug-resistant clinical isolates in PHA-PBMCs

Virusa	EC50, μMb
	GRL-08513	GRL-08613	APV	ATV	LPV	DRV
HIV-1ERS104pre (wild type)	0.003 ± 0.001	0.0025 ± 0.0004	0.031 ± 0.005	0.0025 ± 0.0004	0.032 ± 0.005	0.0033 ± 0.0004
HIV-1MDR/B	0.0022 ± 0.0007 (1)	0.0032 ± 0.0004 (1)	0.48 ± 0.04 (15)	0.467 ± 0.007 (187)	>1 (>31)	0.037 ± 0.007 (11)
HIV-1MDR/C	0.0029 ± 0.0001 (1)	0.0025 ± 0.0006 (1)	0.41 ± 0.04 (13)	0.039 ± 0.003 (16)	0.42 ± 0.04 (13)	0.026 ± 0.001 (8)
HIV-1MDR/G	0.003 ± 0.002 (1)	0.002 ± 0.002 (1)	0.39 ± 0.02 (13)	0.03 ± 0.01 (12)	0.48 ± 0.09 (15)	0.025 ± 0.003 (8)
HIV-1MDR/TM	0.0009 ± 0.0005 (0.3)	0.0026 ± 0.0006 (1)	0.46 ± 0.03 (15)	0.13 ± 0.05 (52)	0.42 ± 0.05 (13)	0.025 ± 0.006 (8)
HIV-1MDRmixDRVR20P	0.0033 ± 0.0001 (1)	0.0090 ± 0.0002 (4)	>1 (>32)	>1 (>400)	>1 (>31)	0.32 ± 0.01 (97)

^aHIV-1_ERS104pre served as a source of wild-type HIV-1. The amino acid sequence of each variant is described in Table 2.

^bThe 50% effective concentrations (EC₅₀s) were determined by using PHA-PBMCs as target cells, and the inhibition of p24 Gag protein production by each drug was used as an endpoint. The numbers in parentheses represent the fold changes of EC₅₀ for each isolate compared to the EC₅₀ for HIV-1_ERS104pre. All assays were conducted in duplicate or triplicate, and the data represent mean values (±1 SD) derived from the results of two to four independent experiments. PHA-PBMCs were derived from a single donor in each independent experiment.

Additionally, we determined the antiviral activities of GRL-08513 and GRL-08613 against four distinct HIV-1 subtypes, HIV-1_92UG029 (subtype A, X4 tropic), HIV-1_92UG037 (subtype A, R5 tropic), HIV-1_Ba-L (subtype B, R5 tropic), and HIV-1_97ZA003 (subtype C, R5 tropic). As shown in Table 4, both CNS-targeting PIs showed the most potent anti-HIV-1 activity against these four subtypes among tested drugs, such as APV, ATV, LPV, and DRV.

TABLE 4.

Antiviral activities of GRL-08513 and GRL-08613 against X4- or R5-tropic subtype A, subtype B, and subtype C clinical isolates in PHA-PBMCs

Virusa	EC50, μMb
	GRL-08513	GRL-08613	APV	ATV	LPV	DRV
HIV-192UG029 (subtype A, X4)	0.0022 ± 0.0001	0.0032 ± 0.0002	0.047 ± 0.005	0.0049 ± 0.0007	0.027 ± 0.006	0.0047 ± 0.0003
HIV-192UG037 (subtype A, R5)	0.00033 ± 0.00003	0.0013 ± 0.0002	0.051 ± 0.007	0.0038 ± 0.0005	0.12 ± 0.03	0.0073 ± 0.0008
HIV-1Ba-L (subtype B, R5)	0.0002 ± 0.0001	0.0009 ± 0.0002	0.028 ± 0.003	0.0037 ± 0.0005	0.017 ± 0.003	0.003 ± 0.001
HIV-197ZA003 (subtype C, R5)	0.0006 ± 0.0001	0.0021 ± 0.0002	0.032 ± 0.001	0.0030 ± 0.0006	0.072 ± 0.002	0.0104 ± 0.0003

^aX4 and R5, X4-tropic and R5-tropic HIV-1 strains, respectively.

^bThe EC₅₀s were determined by using PHA-PBMCs as target cells, 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 represent mean values (±1 SD) derived from the results of two independent experiments. PHA-PBMCs were derived from a single donor in each independent experiment.

The CNS-targeting PIs are active against multiple PI-selected HIV-1 variants.

Next, we evaluated GRL-08513 and GRL-08613 against a panel of HIV-1_NL4-3-derived variants which had been selected in the presence of up to 5 μM concentrations of ATV, LPV, or APV (43, 44). Such variants possessed multiple major PI resistance-associated AA mutations (Table 2). Generally, each variant was remarkably resistant to the very PI selected, with an EC₅₀ of >1 μM (Table 5). However, GRL-08513 and GRL-08613 exhibited robust antiviral activities against HIV-1_ATV^R_5μM and HIV-1_LPV^R_5μM, with EC₅₀s of 0.0034 to ∼0.004 μM (2- to 3-fold difference compared to those against HIV-1_NL4-3). All four control PIs tested in the current study had critically reduced activities against HIV-1_APV^R_5μM, with EC₅₀s of 0.4 to >1 μM, although GRL-08513 and GRL-08613 retained their strong activities against HIV-1_APV^R_5μM, with EC₅₀s of 0.003 and 0.006 μM (2- to 3-fold difference compared to those against HIV-1_NL4-3), respectively (Table 5).

TABLE 5.

Antiviral activities of GRL-08513 and GRL-08613 against laboratory highly conventional-PI-resistant variants

Virusa	EC50, μMb
	GRL-08513	GRL-08613	APV	ATV	LPV	DRV
HIV-1NL4-3	0.0013 ± 0.0004	0.0018 ± 0.0004	0.029 ± 0.006	0.0027 ± 0.0009	0.024 ± 0.008	0.0030 ± 0.0005
HIV-1ATVR5μM	0.0036 ± 0.0009 (3)	0.004 ± 0.002 (2)	0.37 ± 0.05 (13)	>1 (>370)	>1 (>42)	0.033 ± 0.006 (11)
HIV-1LPVR5μM	0.0034 ± 0.0002 (3)	0.0034 ± 0.0006 (2)	>1 (>34)	0.038 ± 0.006 (14)	>1 (>42)	0.036 ± 0.002 (12)
HIV-1APVR5μM	0.003 ± 0.001 (2)	0.006 ± 0.003 (3)	>1 (>34)	0.40 ± 0.04 (148)	>1 (>42)	0.43 ± 0.04 (143)

^aThe amino acid sequence of each variant is described in Table 2.

^bThe EC₅₀s were determined by using MT-4 cells as target cells. MT-4 cells (10⁵/mL) were exposed to 100 TCID₅₀s 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 represent mean values (±1 SD) derived from the results of two to four independent experiments.

Overall, GRL-08513 and GRL-08613 exerted favorable antiviral activities against various laboratory/clinical wild-type HIV-1 strains and drug-resistant HIV-1 variants, in comparison with other FDA-approved HIV-1 PIs tested (Tables 1 and 3 to 5).

The CNS-targeting PIs show favorable log P and log D properties.

Generally, the addition of a fluorine(s) into certain compounds is recognized to contribute to chemical stability and better lipophilicity of such compounds, including nucleoside analogs (45, 46). Hence, we evaluated the partition coefficient (log P) profiles and distribution coefficient (log D) profiles of GRL-08513 and GRL-08613. 1-Octanol and water were used for the determination of log P, while 1-octanol and Tris-buffered saline (TBS; pH 7.4) were used for the determination of log D. Prior to obtaining actual values, a reference standard curve was generated, and then compound or drug concentrations obtained from each compartment (1-octanol, water, and TBS) were quantified on a light spectrophotometer at a 230-nm absorbance (47). Of note, both GRL-08513 and GRL-08613 reached higher concentrations in the octanol lipid interface (n-octanol^w; 98.86 and 99.11 μM for GRL-08513 and GRL-08613, respectively) than did DRV (15.50 μM) (Table 6). Since the more negative the log P/log D value is, the less lipophilic the material is estimated to be (45), GRL-08513 was thought to be the most lipophilic compound in the log P determination, with a log P value of 0.262, compared to −0.63 for DRV. GRL-08513 was also estimated to be the most lipophilic in the log D determination, with a log D value of −0.08, compared to −1.03 for DRV. Similarly, GRL-08613 was estimated to be more lipophilic than DRV, with a log P value of 0.108 and a log D value of −0.17 (Table 6).

TABLE 6.

Partition and distribution coefficients of GRL-08513, GRL-08613, and DRV using the shake flask method

Compound	Concn (μM)				Log Poctanol/watera	Log Doctanol/Trisa
	n-Octanolw	n-Octanolt	Water	Tris buffer
GRL-08513	98.86	98.41	54.08	65.43	0.262	−0.08
GRL-08613	99.11	96.31	77.27	65.48	0.108	−0.17
DRVb	15.50	11.89	81.49	48.77	−0.63	−1.03

^aThe partition (log P) and distribution (log D) coefficients of CNS-targeting PIs were determined. DRV was used as a control. n-Octanol^w, an organic alcohol, and water were used for log P determination, while Tris buffer (pH 7.40) and n-octanol^t were utilized for the log D assay. Prior to the retrieval of actual values, a standard curve was generated as a reference. The drug concentrations for each compartment (n-octanol^w, n-octanol^t, water, and Tris buffer) were measured at an absorbance wavelength of 230 nm using a light spectrophotometer. Assays were performed following the OECD guidelines for testing of chemicals (72). Log P and log D values were calculated according to the formulas given in Materials and Methods.

^bData were previously reported by us in reference 38.

The CNS-targeting PIs well penetrate across the BBB as evaluated in an in vitro BBB assay.

We also tried to evaluate whether GRL-08513 and GRL-08613 had promising blood-brain barrier (BBB) permeation by employing an in vitro BBB model consisting of a triple-cell-coculture system with rat astrocytes and pericytes and monkey endothelial cells. This model, termed BBB Kit, is thought to serve as an in vitro BBB model for a transporting assay of certain compounds/drugs, allowing a sufficient cross talk of the cells involved and providing a system to estimate the passage of compounds/drugs through the BBB (48). GRL-08513, GRL-08613, or reference materials were initially applied to the blood side (luminal interface) of microtiter culture wells under the optimal conditions decided by transendothelial electrical resistance (TEER) value (greater than 150 Ω/cm²). At 30 min after the addition of GRL-08513, GRL-08613, or reference materials to the blood side of culture wells, the concentrations of GRL-08513, GRL-08613, or reference materials that penetrated into the brain side (abluminal interface) were determined using a spectrophotometer. The concentrations of caffeine and sucrose, used as the most and least lipophilic substances (positive and negative controls), in the brain side were 4.85 and 0.07 μM, respectively (Table 7). Eleven FDA-approved cART drugs, saquinavir (SQV), APV, LPV, ATV, and DRV (PIs), 3′-azido-2′,3′-dioxythymidine (AZT) and abacavir (ABC) (nucleotide reverse transcriptase inhibitors [NRTIs]), nevirapine (NVP) and efavirenz (EFV) (nonnucleotide reverse transcriptase inhibitors [NNRTIs]), and raltegravir (RAL) and elvitegravir (EGV) (integrase strand transfer inhibitors [INSTIs]), were also used as references in the assay, showing brain side concentrations of 0.33, 0.85, 0.97, 1.20, 0.60, 1.04, 0.91, 1.79, 0.95, 0.68, and 0.98 μM, respectively. GRL-08513 and GRL-08613 showed the highest concentrations in the brain side of the microtiter culture wells among PIs and reference materials tested, at 2.52 and 2.46 μM, respectively (Table 7).

TABLE 7.

Determination of apparent BBB permeability coefficients of GRL-08513 and GRL-08613 and other drugs using a novel in vitro model^a

Compound	Class	Final abluminal tracer concn (μM)	Papp (10−6), cm/s
GRL-08513	PI	2.52 ± 0.11	38.1 ± 2.2
GRL-08613	PI	2.46 ± 0.03	37.2 ± 0.8
SQV	PI	0.33 ± 0.03b	4.9 ± 0.4b
APV	PI	0.85 ± 0.07	12.9 ± 1.1
LPV	PI	0.97 ± 0.06	14.6 ± 0.9
ATV	PI	1.20 ± 0.14	18.2 ± 2.1
DRV	PI	0.60 ± 0.18	9.0 ± 2.7
AZT	NRTI	1.04 ± 0.07	15.8 ± 1.0
ABC	NRTI	0.91 ± 0.18	13.7 ± 2.7
NVP	NNRTI	1.79 ± 0.11	27.1 ± 1.7
EFV	NNRTI	0.95 ± 0.35b	14.4 ± 5.4b
RAL	INSTI	0.68 ± 0.23b	10.2 ± 3.5b
EVG	INSTI	0.98 ± 0.08	14.8 ± 1.3
Caffeine	PC	4.85 ± 0.08	73.5 ± 1.3
Sucrose	NC	0.07 ± 0.05	1.1 ± 0.4

^aIn the in vitro model using a triple coculture of rat astrocytes, pericytes, and monkey endothelial cells, GRL-08513, GRL-08613, SQV, APV, LPV, ATV, DRV, AZT, ABC, NVP, EFV, RAL, or EVG (all 100 μM) and the positive (PC) and negative (NC) controls (caffeine and sucrose, respectively) were added to the luminal interface (termed blood side) of duplicate wells. The mathematical formula used for the calculation of P_app is described in Materials and Methods. Results show average values ± 1 SD of duplicate determinations.

^bData were previously reported by us (36, 38, 40).

The apparent permeability coefficient (P_app), thought to denote an index for brain uptake of a material, is a way to evaluate quantitatively and qualitatively the penetration of compounds or reference materials across the BBB model (48). The P_app values of GRL-08513 and GRL-08613 (38.1 × 10⁻⁶ cm/s and 37.2 × 10⁻⁶ cm/s, respectively) were higher than that of DRV (9.0 × 10⁻⁶ cm/s) and those of all the anti-HIV-1 drugs tested (values in parentheses): SQV (4.9 × 10⁻⁶ cm/s), APV (12.9 × 10⁻⁶ cm/s), LPV (14.6 × 10⁻⁶ cm/s), ATV (18.2 × 10⁻⁶ cm/s), AZT (15.8 × 10⁻⁶ cm/s), ABC (13.7 × 10⁻⁶ cm/s), NVP (27.1 × 10⁻⁶ cm/s), EFV (14.4 × 10⁻⁶ cm/s), RAL (10.2 × 10⁻⁶ cm/s), and EGV (14.8 × 10⁻⁶ cm/s) (Table 7). Compounds with P_app values higher than 20 × 10⁻⁶ cm/s are thought to have greater efficient penetration property across the BBB; those with values of 10 × 10⁻⁶ to 20 × 10⁻⁶ cm/s are estimated to have a moderate level of BBB penetration property, whereas those with values lower than 10 × 10⁻⁶ cm/s are thought not to well penetrate the BBB (49). Thus, GRL-08513 and GRL-08613 were estimated to have a significantly more favorable BBB penetration property.

X-ray crystal structures of PR_WT–GRL-08513 and PR_WT–GRL-08613.

The X-ray crystal structures of PR_WT in complex with GRL-08513 or GRL-08613 were solved in the space group P6₁ and were refined to 1.95 Å and 1.90 Å, respectively (Table 8). Both structures were solved with one PR_WT dimer per asymmetric unit based on Matthew’s coefficient analysis followed by the solvent content. The X-ray diffraction data for PR_WT plus GRL-08513 and PR_WT plus GRL-08613 were processed to 1.8 Å and 1.9 Å, respectively (Table 9). Although the cocrystals of PR_WT plus GRL-08513 diffracted to 1.8 Å, due to weaker intensities at higher resolution, the final structure was refined to a resolution of 1.95 Å (Table 8). Contiguous difference-electron density was seen for both GRL-08513 and GRL-08613 (Fig. 2), indicating two binding orientations that are 180° apart. Accordingly, the inhibitors were fit in both orientations and were refined with occupancies of 0.51 and 0.49. The higher-occupancy (0.51) orientation was chosen for structural analysis. The P2-Tp-THF moiety of GRL-08513 and GRL-08613 showed chair conformations. As evident from the Ramachandran plot, no outliers were detected for either structure (Table 8), suggesting good quality of the overall stereochemistry.

TABLE 8.

Refinement statistics for structure solutions of PR_WT in complex with GRL-08513 or GRL-08613

Parameter	PRWT–GRL-08513	PRWT–GRL-08613
PDB code	6UWB	6UWC
Resolution range (Å)	45.40–1.95	45.06–1.90
No. of reflections used	13,353	13,196
R cryst a	0.20	0.21
R free	0.26	0.26
No. of protease dimers per AUb	1	1
No. of protein atoms per AU	1,516	1,516
No. of ligand molecules per AU	2	2
No. of ligand atoms per AU	94	92
No. of water molecules	106	97
Mean temp factors
Protein (Å2)	34.86	20.15
Main chains (Å2)	17.25	9.17
Side chains (Å2)	17.87	10.98
Ligand (Å2)	28.43	11.76
Waters (Å2)	39.91	27.24
RMSD bond length (Å)	0.010	0.010
RMSD bond angle (Å)	1.243	1.186
Ramachandran plot
Most favored (%)	98.45	97.42
Additional allowed (%)	1.55	2.06
Generously allowed (%)	0	0.52
Disallowed (%)	0	0

^aR_cryst = Σ ‖F_obs| − |F_calc‖ / Σ|F_obs|.

^bAU, asymmetric unit.

TABLE 9.

X-ray diffraction data processing details for PR_WT in complex with GRL-08513 or GRL-08613

Parameter	PRWT–GRL-08513	PRWT–GRL-08613
PDB code	6UWB	6UWC
Resolution range (Å)	50.0–1.8	50.0–1.9
Unit cell
a (Å)	62.92	62.29
b (Å)	62.92	62.29
c (Å)	82.11	81.98
α (°)	90.00	90.00
β (°)	90.00	90.00
γ (°)	120.00	120.00
Space group	P61	P61
Solvent content (%)	54.43	43.19
No. of unique reflectionsa	16,579 (653)	13,777 (633)
Mean [I/σ(I)]	7.0 (0.3)	15.43 (2.14)
R merge b	0.16 (0.00)	0.09 (0.83)
Data redundancy	3.1 (1.8)	5.6 (4.5)
Completeness (%)	97 (77.6)	95.5 (89.8)

^aValues in parentheses are for the highest-resolution shell.

^bR_merge = Σ |I − | / Σ I.

FIG 2.

Electron density maps for GRL-08513 and GRL-08613, shown in red in panels A and B, respectively. Both maps are contoured at 2.5 σ. Contiguous density can be seen for both inhibitors in the active site of PR_WT. Both inhibitors are shown in stick models, with carbon, nitrogen, oxygen, sulfur, and fluorine atoms in green, blue, red, yellow, and gray, respectively. This figure was generated using CCP4 molecular graphics software.

Profiles of PR_WT inhibitor polar contacts.

GRL-08513 and GRL-08613 exhibit similar binding profiles in the S2-, S1′-, and S2′-binding pockets of PR_WT. However, the P1 modified phenyl moieties of these PIs exhibit distinguished binding profiles. The S2-binding pocket of PR_WT accommodated three strong hydrogen bonds via the backbone amide hydrogens of D29 and D30, with an average bond length of 2.3 Å targeting the P2-Tp-THF moieties of both PIs. Both catalytic aspartic acids from PR_WT contributed to at least one (or two) hydrogen bonds (average bond length of 2.1 Å) via their side chains targeting the hydroxyl groups of the PIs that mimic the transition state during substrate hydrolysis. The P2′-sulfonamide isostere (N-isopropyl-1,3-benzothiazol-2-amine) moiety showed two strong hydrogen bonds, one with the backbone amide hydrogen atom (average interatomic distance of 2.35 Å) and the second one with the side chain δ-oxygen atom (average interatomic distance of 1.75 Å) of D30′ in both inhibitors. The backbone carbonyl oxygen atom of G27 within the vicinity of the catalytic aspartic acids supported a single strong hydrogen bond with both PIs (average bond length of 2.25 Å). Among the crystallographic waters, a conserved water molecule exhibited a hydrogen bond network within the active site of PR_WT. This conserved water molecule displays four hydrogen bonds, two of which target the flaps of PR_WT while the other two target the PI. All four of these hydrogen bonds have an average length of 1.8 Å (Fig. 3). Considering these binding profiles, it can be concluded that both PIs are bound in the active site of PR_WT with high affinities.

FIG 3.

The profiles for polar contacts made by GRL-08513 and GRL-08613 in the active site of PR_WT are shown in panels A and B, respectively. Both inhibitors are displayed in thick stick models. The overall structure of the PR_WT active site is shown in cartoon models in the background with dark green in both panels. The PR_WT amino acid residues involved in polar contacts are shown as thin stick models, with carbon, nitrogen, and oxygen atoms in light green, blue, and red, respectively, in both panels. The two inhibitors display similar profiles in the active site of PR_WT, each involved in 10 polar contacts directly with PR_WT and 2 polar contacts bridged through a conserved water molecule. The average hydrogen bond length in both panels is 1.8 Å, indicating very strong interaction with PR_WT. The PR_WT amino acid residues are labeled as 1 to 99 for the first monomer and 1′ to 99′ for the second monomer. This figure was generated using PyMOL molecular graphics software.

Halogen bonding pattern of GRL-08513 and GRL-08613 in the S1-binding pocket of PR_WT.

The P1-difluorophenyl moiety of GRL-08513 displays an enhanced halogen bonding pattern compared to the P1-para-fluoro moiety of GRL-08613 in the active site of PR_WT, as shown in Fig. 4. GRL-08513 displays tight interactions in the S1 pocket of PR_WT due to the high-affinity halogen interactions from the two fluorine atoms on the P1 group of the PI. Sandwiched between the amide plane of G49-I50 and P81′, one of the fluorine atoms holds the flap and the 80s loop together, while the other fluorine atom is involved in halogen interactions with R8′ and V82′. However, the P1 moiety of GRL-08613, which contains only one fluorine atom, which is not as tightly sandwiched between the flap and the 80s loop of PR_WT as in the case of GRL-08513, displays relatively lower-affinity halogen interactions. In addition to the halogen interactions, the P1 moieties of both PIs are capable of van der Waals (vdW) interactions as well. The halogen bonding pattern of both inhibitors supports their corresponding antiviral data.

FIG 4.

Halogen bonding profile for GRL-08513 and GRL-08613. Panels A and B display the halogen bonding profiles of P1-difluorophenyl (GRL-08513) and para-fluorophenyl (GRL-08613) moieties, respectively, in the active site of PR_WT. In both panels the phenyl group is shown as a white stick model with the fluorine atoms in cyan spheres, while the interacting PR_WT residues are shown as green stick models with the interacting atoms as spheres. The PR_WT carbon, nitrogen, and oxygen atoms are displayed in green, blue, and red, respectively. The P1-difluoro moiety of GRL-08513 (A) shows enhanced halogen bonding profile compared to the P1-para-fluoro moiety of GRL-08613 (B). This figure was generated using PyMOL molecular graphics software.

Profiles of PR_WT-inhibitor van der Waals contacts.

The overall profiles of vdW contacts made by GRL-08513 and GRL-08613 in the active site of PR_WT look similar. However, GRL-08513 displayed at least twice the number of vdW contacts displayed by GRL-08613 in the active site of PR_WT. As reported previously (40), the P2-Tp-THF moieties of GRL-08513 and GRL-08613 show more vdW contacts in the S2-binding pocket of PR_WT than the P2-bis-THF moiety of their analog, GRL-083-13, and darunavir. As shown in Fig. 5, the P2′-N-isopropyl-1,3-benzothiazol-2-amine moieties of GRL-08513 and GRL-08613 showed enhanced profiles of vdW contacts in the S2′-binding pocket of PR_WT compared to the P2′-O-methoxybenzene moieties of their analogs GRL-084-13 and GRL-087-13. The two fluorine atoms in the P1-difluorophenyl moiety of GRL-08513, besides the halogen bonds, showed enhanced networks of vdW contacts in the S1-binding pocket of PR_WT compared to GRL-08613, which contains only one fluorine atom in its P1-para-fluorophenyl moiety.

FIG 5.

The van der Waals interactions of the inhibitors’ P2′ moiety. The P2′-sulfonamide isostere (N-isopropyl-1,3-benzothiazol-2-amine) moiety of GRL-08513/GRL-08613 (A) is compared with the P2′-O-methoxy benzene moiety of GRL-084-13/GRL-087-13 (B) (40) to show the enhanced van der Waals (vdW) contacts in the S2′-binding pocket of PR_WT. In both panels the inhibitors are shown as spheres, with carbon, nitrogen, oxygen, and sulfur atoms in white, blue, red, and yellow, respectively. The S2′-binding pocket of PR_WT is shown as a green surface diagram in both panels. Both GRL-08513 and GRL-08613 display at least twice the number of vdW contacts as their analogs as previously reported (40). This figure was generated using PyMOL molecular graphics software.

Taken together, the structural analyses suggest that both GRL-08513 and GRL-08613 are strongly bound in the active site of PR_WT, with multiple vdW interactions compared to their analogs reported previously (40).

DISCUSSION

Novel CNS-targeting PIs, GRL-08513 and GRL-08613, generated and assessed in the current study, suppressed the replication of wild-type HIV-1/HIV-2 with very low EC₅₀s (Table 1) and showed promising antiviral activities against a variety of clinically isolated MDR HIV-1 strains, with EC₅₀s ranging from 0.0009 to 0.0032 μM, while the FDA-approved HIV-1 PIs tested either failed to inhibit the replication of those MDR variants or needed much higher concentrations for viral suppression (Table 3). Moreover, GRL-08513 and GRL-08613 showed promising activities against a highly DRV-resistant HIV-1 variant (HIV-1_MDRmixDRV^R_20P) (50), with EC₅₀s of 0.0033 and 0.009 μM, respectively (only 1- and 4-fold differences from its EC₅₀ against wild-type HIV-1, respectively) (Table 3), although other FDA-approved PIs, including DRV, lost their antiviral activity against HIV-1_MDRmixDRV^R_20P, by >31- to >400-fold, respectively (Table 3). GRL-08513 and GRL-08613 also inhibited the replication of laboratory-selected highly PI-resistant HIV-1 variants, with low EC₅₀s (0.003 to 0.006 μM) (Table 5).

A variety of novel PIs that we have reported previously (43, 51, 52), including DRV, were designed based on the structure of APV and other approved PIs. Most such PIs were generally less active against HIV-1 variants highly resistant to APV, probably by reason of the resulting structural similarity (43, 51, 52). However, GRL-08513 and GRL-8613 robustly inhibited the replication of HIV-1_APV^R_5μM, with EC₅₀s of 0.003 and 0.006 μM, respectively (only 1- and 2-fold differences from the EC₅₀ against wild-type HIV-1, respectively), suggesting that GRL-08513 and GRL-8613 should have a distinct antiviral property compared to such previously reported PIs structurally APV based.

The CNS penetration property of conventional cART drugs has been rated by the CNS penetration effectiveness (CPE) score. The CPE score consists of 4 ranks and is generally used for patients who have HIV-1-related complications of the CNS (53). However, it was made by the relative evaluation based on the viral load in cerebrospinal fluid (CSF) obtained from cART-receiving patients (54). Thus, a quantitative/rational procedure for the estimation of the CNS penetration property of each drug is much more desirable. In the present study, we assessed the CNS penetration properties of GRL-08513 and GRL-8613 as well as most of the clinically available anti-HIV-1 drugs using an in vitro BBB reconstruction model. The two CNS-targeting PIs we tested showed considerably higher P_app values, ranging 37.2 × 10⁻⁶ to ∼38.1 × 10⁻⁶ cm/s, whereas other conventional cART agents showed much lower P_app values, ranging 4.9 × 10⁻⁶ to ∼27.1 ×10⁻⁶cm/s (Table 7), strongly suggesting that GRL-08513 and GRL-8613 may work as promising therapeutic candidates for patients who suffer with HAND and other CNS complications.

Although the aim of the present study was to optimize anti-HIV-1 agents, which have a high CNS penetration profile, neurotoxicity associated with high drug concentrations in the CNS should be avoided. Therefore, we examined cytotoxicity of GRL-08513 and GRL-08613 against three different lines of brain-derived cells (HKBMM, KG-1-C, and TM-31). While some control drugs, such as SQV and EFV, showed moderate cytotoxicity against these cell lines, it is noteworthy that GRL-08513 and GRL-08613 showed no cytotoxicity against these cell lines when these agents were added up to 100 μM and incubated for 72 h (Table 10); viabilities of HKBMM, KG-1-C, and TM-31 cells after 72 h of incubation with 100 μM GRL-08513 were 102.0, 95.5, and 102.1%, respectively. As for 100 μΜ of GRL-08613, viabilities of HKBMM, KG-1-C, and TM-31 cells were 99.3, 91.7, and 102.9%, respectively.

TABLE 10.

Cytotoxicity of GRL-08513 and GRL-08513 against CNS-derived cell lines^a

Cell line	CC50 (μM) after 72 h of incubation
	GRL-08513	GRL-08513	SQV	APV	ATV	LPV	DRV	EFV	RAL
HKBMM	>100	>100	>100	>100	>100	>100	>100	29.6 ± 3.3	>100
TM-31	>100	>100	93.4 ± 5.9	>100	>100	>100	>100	61.0 ± 0.1	>100
KG-1-C	>100	>100	49.6 ± 2.2	>100	>100	>100	>100	47.2 ± 0.7	>100

^aHKBMM (human malignant meningioma), TM-31 (human astrocytoma) (73), and KG-1-C (human glioma) (74) cells, which were originally isolated from patients’ brain tissues, were cultured in the presence of various concentrations of each compound for 72 h, and the CC₅₀s were determined by the MTT assay. All assays were conducted in duplicate, and the data represent mean values (±1 SD) derived from the results of two independent experiments.

Ma et al. reported that the HIV-1 integrase inhibitor elvitegravir showed a stable CSF concentration (5.47 ± 1.25 ng/mL) that exceeded the 50% inhibitory concentration (IC₅₀) for wild-type HIV-1 (IC₅₀ = 0.76 ng/mL) and showed good antiviral effect in the CNS, but tenofovir entirely did not reach therapeutic concentration in the CSF, in a 24-week, single-arm, open-label study of treatment-experienced adults living with HIV-1 and receiving a treatment regimen including tenofovir alafenamide fumarate and elvitegravir (55). In our present study, GRL-08513 and GRL-08613 showed P_app values 2.57-fold and 2.51-fold higher than that of elvitegravir in our in vitro BBB permeability assay (Table 7). Although there is an experimental limitation because our evaluations were conducted only in vitro, it is possible that combination of GRL-08513/GRL-08613 and ART drugs which show stable CSF concentrations, such as elvitegravir, may lead to a prohibition of the emergence of drug-resistant HIV-1 in the CNS and might be a counterplan for neurosymptomatic CSF viral escape, which is characterized by neurologic symptoms attributed to CNS infection by HIV-1 and detectable CSF viral load, despite suppression of plasma virus (56, 57).

Comparative structural analysis of PR_WT–GRL-08513, PR_WT–GRL-08613, and X-ray crystal structures with PR_WT–GRL-084-13 (PDB code 6D0E) and GRL-087-13 (PDB code 6D0D) (40) revealed that the overall binding profiles of all inhibitors support their subnanomolar anti-HIV-1 activities against wild-type HIV-1 (Table 1). Differences in the P1 moieties (GRL-08513 versus GRL-08613) and P2′ moieties (GRL-08513/GRL-08613 versus GRL-084-13/GRL-087-13) of the inhibitor binding profiles in the active site of PR_WT explain, in part, their differential anti-HIV-1 activities against laboratory-induced HIV-1 variants and clinically isolated HIV-1 strains (Tables 3 to 5). As reported previously (40), the P2′-O-methoxy benzene moiety of GRL-084-13 and GRL-087-13 showed better antiviral activity than the P2′-amino benzene moiety of DRV. However, with a modified P2′ moiety, both GRL-08513/GRL-08613 showed far superior antiviral activity than the aforementioned. As shown in Fig. 5, the total number of vdW contacts in the S2′-binding pocket of PR_WT was enhanced at least twice compared to the case with GRL-084-13/GRL-087-13. Along these lines, the antiviral activities of GRL-08513 (EC₅₀: 0.003 μM) and GRL-08613 (EC₅₀: 0.006 μM) were 4- and 8-fold higher than those of their analogs GRL-084-13 (EC₅₀: 0.012 μM) (38) and GRL-087-13 (EC₅₀: 0.047 μM) (40), respectively, against HIV-1_APV^R_5μM (Table 5). This difference in the antiviral activities primarily owes to the P2′ moiety modification in this study compared to the previously reported analogs and confirms that the P2′ modification in GRL-08513 and GRL-08613 is resistant against the structural changes in the S2′-binding pocket of PR_WT due to the amino acid substitutions listed in Table 2. However, the P1-difluoro (GRL-08513) versus P1-para-fluoro (GRL-08613) also contributed to the antiviral activity. For example, GRL-08513 (EC₅₀: 0.0034 μM) and GRL-084-13 (EC₅₀: 0.003 μM) both contain P1-difluoro moiety and were equipotent against HIV-1_LPV^R_5μM in spite of the differences in their P2′ moieties. Similarly, GRL-08613 (EC₅₀: 0.006 μM) and GRL-087-13 (EC₅₀: 0.047 μM) both contain P1-para-fluoro moiety but showed a 10-fold difference in their antiviral activities against HIV-1_LPV^R_5μM, with GRL-08613 being a better compound. It is noteworthy that although ATV exhibited favorable antiviral activity along with DRV, the EC₅₀s of GRL-08513, GRL-08613, and ATV against HIV-1_ATV^R_5μM were 0.0036 μM, 0.004 μM, and >1.0 μM, respectively. Finally, both GRL-08513 and GRL-08613 showed enhanced antiviral activities against HIV-1_MDRmixDRV^R_20P, with EC₅₀s of 0.0033 μM and 0.009 μM, whereas all the inhibitors, including DRV, failed within the same assay (Table 3). These observations in part confirm that both P1 and S2′ moieties in GRL-08513 and GRL-08613 exhibit cooperativity to enhance and preserve the binding affinity even when there are multiple amino acid substitutions as seen in laboratory strains as well as the clinical isolates.

In conclusion, the present data suggest the possibility that minimum administration of the novel CNS-targeting PIs, GRL-08513 and GRL-08613, could be sufficient for the treatment/prevention of drug-resistant HIV-1 infection and HIV-1-associated CNS complications compared to other currently available FDA-approved anti-HIV-1 drugs. However, it should be noted that much longer-term evaluation is definitely required to confirm the safety profiles of such CNS-targeting agents in future clinical trial settings.

MATERIALS AND METHODS

Cells and viruses.

MT-2 and MT-4 cells were grown in RPMI 1640-based culture medium supplemented with 10% fetal calf serum (FCS; JRH Biosciences, Lenexa, MD), 50 U/mL of penicillin, and 100 μg/mL of kanamycin. These cells were obtained from NCI/NIH. The following HIV-1 strains were employed for the drug susceptibility assay: HIV-1_LAI, HIV-1_NL4-3, HIV-1_ERS104pre, clinical HIV-1 strains isolated from drug-naive patients with AIDS, and four HIV-1 clinical strains which were originally isolated from patients with AIDS who had received 9 or 10 anti-HIV-1 drugs over the past 34 to 83 months and were genotypically and phenotypically characterized as multi-PI-resistant variants (41). HIV-1_97UG029, HIV-1_92UG037, HIV-1_97ZA003, and HIV-1_Ba-L were provided by the NIH AIDS Reagent Program. All primary HIV-1 strains were passaged once or twice in 3-day-old phytohemagglutinin-activated peripheral blood mononuclear cells (PHA-PBMCs), and the virus-containing culture supernatants were stored at −80°C until use as sources of infectious virions.

Antiviral agents.

Roche Products Ltd. (Welwyn Garden City, United Kingdom) kindly provided saquinavir (SQV). Amprenavir (APV) was received as a courtesy gift from GlaxoSmithKline, Research Triangle Park, NC. Lopinavir (LPV) was kindly provided by Japan Energy Inc., Tokyo, Japan. Atazanavir (ATV) was a contribution from Bristol-Myers Squibb (New York, NY). 3′-Azido-2’,3′-dioxythymidine (AZT) was purchased from Sigma-Aldrich (St. Louis, MO). Darunavir (DRV) was synthesized as previously described (58). Abacavir (ABC) and raltegravir (RAL) were provided by the NIH AIDS Reagent Program. Nevirapine (NVP), efavirenz (EFV), and elvitegravir (EVG) were gifted from Kenji Maeda (NCI, NIH, USA).

Drug susceptibility assay.

The susceptibility of laboratory wild-type HIV-1_LAI to various drugs/compounds was determined as previously described (43, 44). Briefly, MT-2 cells (10⁴/mL) were exposed to 100× 50% tissue culture infectious dose (TCID₅₀) of HIV-1_LAI in the presence or absence of various concentrations of drugs/compounds in 96-well microculture plates and were incubated at 37°C for 7 days. After incubation, 100 μL of the medium was removed from each well and 3-(4,5-dimetylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution (10 μL; 7.5 mg/mL in phosphate-buffered saline) was added to each well in the plate, followed by incubation at 37°C for 2 h. After incubation to dissolve the formazan crystals, 100 μL of acidified isopropanol containing 4% (vol/vol) Triton X-100 was added to each well and the optical density was measured in a kinetic microplate reader (Vmax; Molecular Devices, Sunnyvale, CA). All assays were performed in duplicate. In some experiments, MT-2 cells were chosen as target cells in the MTT assay, since these cells undergo greater HIV-1-elicited cytopathic effects than MT-4 cells. To determine the sensitivity of primary HIV-1 isolates to drugs/compounds, PHA-PBMCs (10⁶/mL) were exposed to 50 TCID₅₀s of each primary HIV-1 isolate and cultured in the presence or absence of various concentrations of drugs/compounds in 10-fold serial dilutions in 96-well microculture plates. For determination of the drug susceptibility of certain laboratory HIV-1 strains, MT-4 cells were employed as target cells as previously described (43), with minor modifications. In brief, MT-4 cells (10⁵/mL) were exposed to 100 TCID₅₀s of drug-resistant HIV-1 strains in the presence or absence of various concentrations of drugs/compounds and were incubated at 37°C. On day 7 of culture, the supernatants were harvested and the amounts of p24 (HIV-1 capsid [CA]) Gag protein were determined by using a fully automated chemiluminescent enzyme immunoassay system (Lumipulse f; Fujirebio Inc., Tokyo, Japan) (59). Drug/compound concentrations that suppressed the production of p24 Gag protein by 50% (50% effective concentration [EC₅₀]) were determined by comparison with the p24 production level in drug/compound-free control cell culture. All assays were performed in duplicate. Each drug/compound’s EC₅₀ shown in the current report represent average values of the data obtained from two to four independently conducted experiments. PHA-PBMCs were derived from a single donor in each independent experiment. Thus, for obtaining the data, two to four different healthy donors were recruited.

Determination of nucleotide sequences.

The determination of the nucleotide sequences of HIV-1 strains passaged in the presence of anti-HIV-1 agents was performed as previously described (59). In brief, high-molecular-weight DNA was extracted from HIV-1-infected MT-4 cells by using the InstaGene matrix (Bio-Rad Laboratories, Hercules, CA) and was subjected to molecular cloning, followed by sequence determination. The primers used for the first round of PCR with the entire Gag- and protease-encoding regions of the HIV-1 genome were LTR F1 (5′-GAT GCT ACA TAT AAG CAG CTG C-3′) and PR12 (5′-CTC GTG ACA AAT TTC TAC TAA TGC-3′). The first-round PCR mixture consisted of 1 μL of proviral DNA solution, 10 μL of premix Taq (Ex Taq version; TaKaRa Bio Inc., Otsu, Japan), and 10 pmol of each of the first PCR primers in a total volume of 20 μL. The PCR conditions used were an initial 3 min at 95°C and then 35 cycles of 40 s at 95°C, 20 s at 55°C, and 2 min at 72°C, with a final 10 min of extension at 72°C. The first-round PCR products (1 μL) were used directly in the second round of PCR with primers LTR F2 (5′-GAG ACT CTG GTA ACT AGA GAT C-3′) and KSMA2.1 (5′-CCA TCC CGG GCT TTA ATT TTA CTG GTA C-3′) under the PCR conditions of an initial 3 min at 95°C and then 35 cycles of 30 s at 95°C, 20 s at 55°C, and 2 min at 72°C, with a final 10 min of extension at 72°C. The second-round PCR products were purified with spin columns (MicroSpin S-400 HR columns; Amersham Biosciences Corp., Piscataway, NJ) and subjected to sequencing with a model 3130 automated DNA sequencer (Applied Biosystems, Foster City, CA).

Determination of partition and distribution coefficients of the CNS-targeting PIs using the shake flask method.

On day −1 of the experiment, saturation of 1-octanol [CH₃(CH₂)₆CH₂OH] (Nacalai Tesque, Kyoto, Japan) with water and Tris-buffered saline (TBS; 10× working solution; 20 mM Tris [pH 7.4] and 0.9% NaCl; Sigma-Aldrich, St. Louis, MO) took place. Four different flasks were used. One contained 50 mL of water plus 100 mL of 1-octanol (water saturated with octanol), and another flask contained 1-octanol saturated with water by the addition of 50 mL of 1-octanol and 100 mL of water. For the other two flasks, the same ratio and volumes were kept for 1-octanol saturated with TBS and TBS saturated with 1-octanol. The flasks were sealed and placed into a bioshaker at room temperature for 24 h at 90 rpm. Simultaneously, dilutions of GRL-08513, GRL-08613, and DRV were performed from a 20 mM dimethyl sulfoxide (DMSO) stock to a final concentration of 100 μM using distilled water (dH₂O), TBS, and 1-octanol as solvents. Successive dilutions were made to obtain concentrations of 10 μM, 1 μM, and 0.1 μM. A standard curve was generated in a light spectrophotometer (DU series 700; Beckman Coulter, Fullerton, CA) at an absorbance wavelength of 230 nm.

On day 1 of the experiment, the lipid and liquid interfaces were separated, and drugs/compounds were diluted again from 20 mM DMSO to 100 μM using the 1-octanol, water, and TBS obtained from the shake flask assay. The resulting diluted drugs/compounds were then added to separate serum tubes containing equal proportions of (i) 1-octanol and water and (ii) 1-octanol and TBS. The solution was hand-shaken for 5 min, then centrifuged at 3,500 rpm, and kept at room temperature for 20 min. Finally, the drug/compounds were recovered from the 1-octanol, TBS, and water interfaces and then measured on a light spectrophotometer.

The values for log P and log D were obtained by applying the following mathematical formulas:

$$\log P_{\text{octanol/water}}=\log \left(\frac{{[\text{compound}]}_{n-\text{octanol}}}{{[\text{compound}]}_{\text{water}}}\right)$$

$$\log D_{\text{octanol/water}}=\log \left(\frac{{[\text{compound}]}_{n-\text{octanol}}}{{[\text{compound}]}_{\text{ionized}}+{[\text{compound}]}_{\text{neutral}}}\right)$$

where [compound]_n-octanol, [compound]_water, [compound]_ionized, and [compound]_neutral represent the concentrations of compound in n-octanol, water, ionized water, and neutral water, respectively.

Determination of apparent permeability blood-brain barrier coefficients of CNS-targeting PIs using a novel in vitro model.

A novel in vitro BBB model (BBB Kit; PharmaCo-Cell Ltd. Nagasaki, Japan) incorporating a triple culture of rat-derived astrocytes, pericytes, and monkey-derived endothelial cells (48) was used to determine the apparent permeability BBB coefficient (P_app) of GRL-08513, GRL-08613, SQV, APV, LPV, ATV, DRV, AZT, ABC, NVP, EFV, RAL, EVG, or caffeine and sucrose.

The BBB Kit was kept at −80°C until thawing on day 0 of the experiments. Nutritional medium was added to both brain and blood sides of the wells. This solution consisted of Dulbecco modified Eagle medium (DMEM) F-12 medium with 10% (vol/vol) FCS, heparin at 100 μg/mL, basic fibroblast growth factor (bFGF) at 1.5 ng/mL, insulin at 5 μg/mL, transferrin at 5 μg/mL, sodium selenite at 5 ng/mL, hydrocortisone at 500 nM, and gentamicin at 50 μg/mL. Fresh medium was added 3 h after thawing following the manufacturer’s instructions and 24 h later. The plates were incubated at 37°C until day 4 of the experiment, when the condition of cells was checked under a light microscope. Following this, the integrity of the collagen-coated membrane was verified by measurement of the transendothelial electrical resistance (TEER) using an ohmmeter. As TEER increased over the days, reaching optimal values between days 4 and 6 of the experiment, the apparent permeability BBB coefficient determinations were done during this period. Membranes were tested individually, and those collagen-coated membranes displaying TEER values greater than 150 Ω/cm² were suitable for the execution of the drug/compound BBB penetration assay. Detailed information regarding the components of the BBB kit as well as its mechanisms can be obtained by accessing the manufacturer’s website (http://www.pharmacocell.co.jp/en/bbb/index_e.html).

Once the conditions of cell viability and membrane integrity were met, drug/compound dilutions were performed from 20 mM DMSO stocks of GRL-08513, GRL-08613, SQV, APV, LPV, ATV, DRV, AZT, ABC, NVP, EFV, RAL, and EVG, while caffeine and sucrose were used as positive and negative controls, respectively. Standard curves were generated for each drug/compound on a light spectrophotometer as previously described (36). A 100 μM concentration of each drug/compound was added to the luminal (blood side) of the wells and incubated at 37°C for 30 min, and then the amount of drug/compound that crossed the in vitro BBB (abluminal; brain side) was collected and measured under a light spectrophotometer at an absorbance wavelength of 230 nm.

P_app (in centimeters per second) was calculated using the following mathematical formula:

$$P_{\text{app}}=\frac{V_A}{A\times {[\text{compound}]}_{\text{luminal}}}\times \frac{\Delta {[\text{compound}]}_{\text{abluminal}}}{\Delta t}$$

where V_A is volume of the abluminal chamber (0.9 cm³), A is membrane surface area (0.33 cm²), [compound]_luminal is initial luminal compound concentration (micromolar), Δ[compound]_abluminal is abluminal compound concentration (micromolar), and Δt is time of the experiment (seconds).

Protein expression and preparation.

Expression and purification of protease were carried out as previously described (60). Briefly, Rosetta (DE3) pLysS strain (Novagen) was transformed with an expression vector (pET-30a), which contained the genes of wild-type HIV-1_NL4-3-PR (PR_WT), using heat shock. The culture was grown in a shake flask containing 30 mL of Luria broth plus kanamycin and chloramphenicol (LB^Km+/Cp+) at 37°C overnight. For the expression of PR_WT, 20 mL of the grown culture was added to 1 L of ZYM-10052 (1.0% N-Z amine, 0.5% yeast extract, 25 mM disodium hydrogen phosphate, 25 mM potassium dihydrogen phosphate, 50 mM ammonium chloride, 5 mM sodium sulfate, 1.0% glycerol, 0.05% glucose, 0.2% α-lactose, 2 mM magnesium sulfate) plus kanamycin and chloramphenicol (ZYM-10052^Km+/Cp+). The ZYM-10052^Km+/Cp+ culture was further continued at 37°C for 20 to ∼22 h (61). Then the culture was spun down for pellet collection, and the obtained pellets were stored at −80°C until use. For purification of PR_WT, the pellet was resuspended in buffer A (20 mM Tris, 1 mM EDTA, and 1 mM dithiothreitol [DTT]) and lysed by sonication. The cell lysates were separated into a supernatant fraction and an inclusion body fraction by centrifugation. The PR_WT was confirmed to be present in the inclusion body fraction, which was washed five times with buffer A containing 2 M urea and then with buffer A without urea. The twice-washed pellet was solubilized and PR_WT was unfolded with 100 mM formic acid (pH 2.8) (62). The unfolded PR_WT was purified using a fast protein liquid chromatography system (ÄKTA pure 25; GE Healthcare) and separated using a reverse-phase chromatography column (RESOURS RPC 3 mL; GE Healthcare) using the gradient of buffer B (1.0% formic acid, 2.0% acetonitrile) and buffer C (1% formic acid, 70% acetonitrile). The flow rate was set to 1.0 mL min⁻¹, and the column was equilibrated with 75% buffer B and 25% buffer C. Then the amount of buffer C was increased to 75% over a 30-min period (10 times the column volume). PR_WT was eluted with 35 to ∼50% buffer C. After elution, the buffer C amount was increased to 100% in 6 min and returned to the starting condition over the next 6 min. The peak fractions including PR_WT were collected and diluted three times with buffer B. The diluted PR_WT solution was injected into the ÄKTA pure 25 again, and the targeted PR_WT was purified using the same purification step as described above. The collected fractions containing PR_WT were subjected to desalting (HiTrap desalting; GE Healthcare), and the eluted solution was equilibrated using 100 mM formic acid and stored at −80°C until use.

Crystallization.

The unfolded PR_WT was refolded with the addition of neutralizing buffer A (100 mM ammonium acetate [pH 6.0], 0.005% Tween 20), making the final pH 5.0 to 5.2. The PR_WT-containing solution was run through Amicon Ultra-15 10K centrifugal filter units (Millipore), giving a solution containing PR (1 to ∼2 mg/mL) in 10 mM ammonium acetate (pH 5.0) and 0.005% Tween 20. Occasionally, fifth-greater concentrations of a test compound were used for crystallization. After centrifugation, the supernatants were collected and subjected to crystallization using the hanging-drop vapor diffusion method. The Nextal Tubes ProComplex suite (Qiagen) was used for the first screening to determine the optimum crystallization condition. PR_WT–GRL-08513 complexes were formed in 0.1 M HEPES (pH 7.5), 12% (wt/vol) polyethylene glycol 8000, and 0.2 M sodium chloride. PR_WT–GRL-08613 complexes were formed in 0.1 M sodium citrate (pH 5.5) and 15% (wt/vol) polyethylene glycol 6000. Crystals of PR_WT complexed with GRL-08513 or GRL-08613 were retrieved, immersed in a reservoir containing cryoprotective solution plus 30% glycerol, and flash-frozen in liquid nitrogen.

X-ray diffraction data collection and processing details.

The X-ray diffraction data for the cocrystals of PR_WT–GRL-08513 and PR_WT–GRL-08613 were collected at SPring-8 BL41XU with an X-ray wavelength of 1 Å under a 100 K cold stream using a PILATUS3 6M detector (Dectris Ltd.). The distances from crystal to the detector were 250 mm for PR_WT–GRL-08513 and 260 mm for PR_WT–GRL-08613. All data were collected with a rotation step of 0.5°/frame and an exposure time of 0.5 s/frame. Data were processed by using HKL-2000 software (63). A summary of X-ray diffraction data processing details is given in Table 9.

Structure solutions and refinement.

With the help of solvent content and Matthew’s coefficient values determined using the CCP4 (64) suite of programs, the number of PR_WT molecules per asymmetric unit was estimated. The X-ray crystal structure of PR_WT taken from PDB code 4HLA (65) was used as a search model to obtain structure solutions through MOLREP (66). One PR_WT dimer was fit per asymmetric unit. Structure solutions were directly refined using REFMAC5 (67) through CCP4. GRL-08513 and GRL-08613 were fit using ARP/wARP (68)-Ligands (69) through CCP4. The initial coordinates for both GRL-08513 and GRL-08613 were prepared by modifying the coordinates of GRL-084-13 (PDB code 6D0E) and GRL-087-13 (PDB code 6D0D), respectively (40). Solvent molecules were fit using ARP/wARP-Solvent through CCP4. The models were then refined using the simulated annealing protocol in phenix.refine through the Phenix suite of programs (70). Refinement libraries for GRL-08513 and GRL-08613 were optimized using the eLBOW-AM1 (71) method and were manually checked through REEL. A summary of structure solution refinement statistics is given in Table 8.

Structural analysis.

The X-ray crystal structures of PR_WT–GRL-08513 and PR_WT–GRL-08613 were processed through the Protein preparation wizard in Maestro (Schrödinger LLC, New York, NY), and the hydrogen bonding patterns were optimized through PROPKA (50) by sampling the solvent. Hydrogen bonds were calculated by using a 3.0-Å cutoff for the distance between the two heavy atoms and angle cutoff values of 90° (donor) and 60° (acceptor). Hydrogen bonds of <3.0 Å were considered strong bonds and those of >3.0 Å were considered weaker bonds. The van der Waals (vdW) contacts were calculated by using a distance cutoff value of 5 Å. Halogen bonds were calculated using a 4.0-Å distance cutoff with angle ranging between 120° and 180°. Halogen bonds of <4.0 Å were considered strong bonds and those of >4.0 Å were considered weaker bonds.

Accession number(s).

The final refined coordinates of PR_WT–GRL-08513 and PR_WT–GRL-08613 along with their corresponding X-ray data were deposited in the worldwide Protein Data Bank (PDB) with accession codes 6UWB and 6UWC, respectively.