Development of non-nucleoside reverse transcriptase inhibitors (NNRTIs): our past twenty years

Abstract

Human immunodeficiency virus (HIV) is the primary infectious agent of acquired immunodeficiency syndrome (AIDS), and non-nucleoside reverse transcriptase inhibitors (NNRTIs) are the cornerstone of HIV treatment. In the last 20 years, our medicinal chemistry group has made great strides in developing several distinct novel NNRTIs, including 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT), thio-dihydro-alkoxy-benzyl-oxopyrimidine (S-DABO), diaryltriazine (DATA), diarylpyrimidine (DAPY) analogues, and their hybrid derivatives. Application of integrated modern medicinal strategies, including structure-based drug design, fragment-based optimization, scaffold/fragment hopping, molecular/fragment hybridization, and bioisosterism, led to the development of several highly potent analogues for further evaluations. In this paper, we review the development of NNRTIs in the last two decades using the above optimization strategies, including their structure–activity relationships, molecular modeling, and their binding modes with HIV-1 reverse transcriptase (RT). Future directions and perspectives on the design and associated challenges are also discussed.

Graphical abstract

Human immunodeficiency virus (HIV) is the primary infectious agent of acquired immunodeficiency syndrome (AIDS), and non-nucleoside reverse transcriptase inhibitors (NNRTIs) are the cornerstone of HIV treatment. In this review, we summarize the development of NNRTIs in our past two decades using the multiple optimization strategies.

1. Introduction

Acquired immunodeficiency syndrome (AIDS) is considered as a pandemic caused by human immunodeficiency virus (HIV) infection. According to the latest data from the Joint United Nations Programme on HIV/AIDS (UNAIDS), 36.9 million people globally were living with HIV, and 21.7 million people were receiving antiretroviral therapy in 2017¹. Remarkable reduction in AIDS-related mortality has been made possible by the current combination therapy, termed highly active antiretroviral therapy (HAART). Typically, HAART targets multiple viral replication cycles and includes two nucleoside reverse transcriptase inhibitors (NRTIs), a non-nucleoside reverse transcriptase inhibitor (NNRTI), and a protease inhibitor (PI) or an integrase inhibitor (INSTI)2, 3, 4. The NNRTIs are essential components in the drug combination therapies due to their unique antiviral activity, high specificity, and low toxicity5, 6, 7. To date, more than 50 chemotypes have been identified as NNRTIs, including 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymines (HEPTs), dihydro-alkoxy-benzyl-oxopyrimidines (DABOs), diarylpyrimidines (DAPYs), etc.7, 8, 9. Among these, U.S. Food and Drug Administration (FDA) approved six NNRTIs for HIV-1 treatment (Fig. 1A), namely nevirapine (NVP, 1996), delavirdine (DLV, 1997), efavirenz (EFV, 1998), etravirine (ETR, 2008), rilpivirine (RPV, 2011), and doravirine (DOR, 2018)^3,10. NNRTIs inhibit reverse transcriptase (RT) by binding to an allosteric site, namely NNRTI binding pocket (NNIBP), located about 10 Å away from the DNA polymerase active site (Fig. 1B). Although the NNRTIs are structurally different and unrelated, they fit within the NNIBP and their binding modes are in a similar conformation to be considered as a “butterfly”, “horseshoe”, or “U” mode with a central scaffold and two “wings” (Fig. 1A). This feature provides valuable information for the discovery of NNRTIs and lead optimizations.

Figure 1.

(A) The chemical and three-dimension structures of FDA-approved NNRTI drugs obtained from the crystal complexes with HIV-1 RT showing similar binding conformations: a “butterfly”, “horseshoe”, or “U” mode with a central scaffold and two “wings” (B) Structure of HIV-1 RT (P66/P51) in complex with RNA/DNA and a NNRTI, NVP, are highlighted in different colors (PDB entry: 4PUO).

However, a growing number of drug-resistant HIV-1 strains in NNRTI-treated cells and patients with HIV-1 are being reported¹¹. Mutations within or near the NNIBP dramatically reduce the efficacy of NNRTIs, resulting in treatment failures. Therefore, drug resistance poses a considerable challenge in anti-HIV-1 drug development. In the last 20 years (1999–2019), our medicinal chemistry group tackled the development of novel anti-HIV-1 drugs using multiple optimization strategies involving the classical NNRTIs, including structure- and fragment-based drug design, scaffold hopping, molecular/fragment hybridization, and bioisosterism to identify novel NNRTIs with high antiviral activity against wild-type (WT) and mutant viruses. With ongoing efforts, many derivatives have been obtained; in particular, two DAPY analogues have advanced to preclinical evaluations. In this article, we summarize our efforts on structure optimizations and development strategies, structure–activity relationship (SAR) studies, and computational analysis. Future directions and perspectives on the design and development of anti-HIV agents and associated challenges are also discussed.

2. Development of NNRTIs

2.1. Naphthyl-HEPT (HENT) analogues

In 1989, the first discovered NNRTI, HEPT (Fig. 2) was reported by Miyasaka et al.¹² as a novel anti-HIV-1 lead compound, showing promising antiviral activity (EC₅₀ = 7 μmol/L) and selectivity index (selectivity index, SI = 106). Subsequent optimizations led to a more potent ethoxymethyl analogue MKC-442, also named emivirine (Coactinon), which was selected as a candidate drug for phase III clinical trials in AIDS patients (NCT00002413). However, the clinical trial by Triangle Pharmaceuticals in 2002 was terminated due to lower drug potency compared to other antiretrovirals¹³. Besides, the drug resistance mediated by the NNIBP mutations limited the use of HEPT analogues¹³.

Figure 2.

Structure optimization workflow for the HEPT analogues developed by our group. The modified parts are colored.

As shown in Fig. 3A, the MKC-442-RT co-crystal complex showed that (1) MKC-442 was constrained by a strong hydrogen-bonding interaction between the 3-NH of the pyrimidine ring with the carbonyl oxygen of K101 and water-mediated hydrogen bonds with K101 and E138; (2) the 6-benzyl ring of the inhibitor, inserting into the hydrophobic region consisting of Y181, Y188, and W229 residues, were conformationally different from those in the unliganded RT structure and formed π–π stacking interactions with these aromatic residues¹⁴. Steric bulky groups could not replace the isopropyl group located at the E138 site due to limited space in the Y181 and E138 channel. The ethoxymethyl tail was situated in the F227, Y318, and P236 region. Therefore, the hydrophilic hydroxyl group was replaced by a hydrophobic group to improve the antiviral activity¹⁵.

Figure 3.

(A) Crystal complex of MKC-442 with RT (PDB: 1RT1); and molecular docking models of RT with compounds 1 (B), 2 (C), 4 (D) and 9 (E). The figures were generated by PyMOL.

Based on the structural biology information, two parts of HEPT or MKC-442 have been optimized, and the representative structure optimization workflow is summarized in Fig. 216, 17, 18, 19, 20, 21, 22, 23, 24. (1) Replacement of the 6-benzyl ring by a naphthyl ring could enhance π-stacking interactions in the NNIBP¹³. (2) Introducing an aromatic ring such as a phenyl group at the ethoxymethyl tail aimed at increasing the hydrophobic interactions with the Y318/P236 channel. Consistent with the co-crystal information, our 3D-QSAR study also revealed that the 6-benzyl ring of MKC-442 could be replaced by the α-naphthyl ring (abbreviated as HENT)^21,23. The resulting compound 1 exhibited an EC₅₀ of 17 nmol/L against HIV-1 WT strain in MT-4 cells (HEPT, EC₅₀ = 7 μmol/L; MKC-442, EC₅₀ = 20 nmol/L)¹³ and relatively low cytotoxicity (SI = 2229). Compounds 2 and 3 with phenyl and benzyl groups at the ethoxymethyl tail had similar potency and cytotoxicity as compound 1²². Molecular docking studies showed that the naphthyl group located at the hydrophobic region strengthened the π–π stacking interactions with Y181, Y188, and W229 (Fig. 3B and C), and the phenyl ring brought an additional hydrophobic interaction with the Y318 (Fig. 3C).

Carbon to sulfur variation is observed commonly in drug development using a bioisosterism strategy. A new series of naphthylthio HEPT (S-HENT) was synthesized based on the original HEPT and bioisosteric replacement¹⁷. Compound 4 with α-naphthyl had an EC₅₀ of 48 nmol/L while the analogue with β-naphthyl (compound 5) decreased the anti-HIV-1 activity by 13.5-fold, indicating limited space or flexibility for the naphthyl group at the hydrophobic pocket (Fig. 3D). Enhanced π–π stacking interactions by introducing an α-nitro group in compound 6 were probably due to the reduced electron density. Consequently, the antiviral activity of compound 6 improved by 10-fold with an EC₅₀ of 65 nmol/L. However, the amino substitution (7) with a different electronic effect dramatically decreased the interaction, with an EC₅₀ of 3650 nmol/L. A carbonyl group between the naphthyl and central pyrimidine ring was tolerated at the binding pocket (Fig. 3E)¹⁸. However, the antiviral activity decreased by 6–8 fold for compounds 8 and 9, potentially due to lack of π–π interaction with Y181 by the flipped naphthyl ring. Further SAR studies illustrated that the ester group presenting on the phenyl side chain played a critical role in the antiviral activity. The replacement of the β-oxygen of the N-1 side chain with a carbonyl group (e.g., compound 10) significantly suppressed the anti-HIV-1 activity²⁰.

2.2. DABO analogues

The DABO analogues were first reported in 1992 as novel NNRTIs²⁵. Chemically, the DABOs bearing the 4-pyrimidinone scaffold, are structurally similar to HEPTs with the N-1 substitution changed to the C-2 position (Fig. 4), thus leading to reasonably high anti-HIV activity^26,27. The molecular docking study (Fig. 5) showed that (1) the central pyrimidinone scaffold was located at the E138, K101 and K103 region forming two hydrogen bonds with K101; (2) at the C-5 position, the hydrophobic cavity of V179 and K103 residues had limited space, suggesting use of small chemical groups such as methyl, ethyl and isopropyl group at this site; (3) optimization of the side chain at the C-2 position could be attempted because it was pointing to a hydrophobic channel surrounded by V106, F227, and P236; (3) the oxygen could be replaced by sulfur using a bioisosterism strategy and a hydrogen-bond acceptor (e.g., –OCH₃) could be introduced to improve the activity.

Figure 4.

Structure optimization workflow for the DABO analogues developed by our group. The modified parts are colored.

Figure 5.

Molecular docking models of RT with compounds 11 (A), 13 (B), 18 (C), 17-S (D), 17-R (E) and the superposition mode of 17-S and 17-R (F). The figures were generated by PyMOL.

Thus, the structure-based drug design (SBDD) was conducted to obtain several potent analogues. Compounds 11 and 12 showed great potency in the nanomolar range and high SI values compared with the micromolar-range DABOs (Fig. 4)^25,28. Similar to HEPT, the dimethylbenzyl ring was inserted into the hydrophobic tunnel of Y181, Y188, F227, and W229 residues (Fig. 5A). A naphthyl group was introduced to obtain compounds 13 and 14 with an EC₅₀ of 30 and 45 nmol/L, respectively (Fig. 5B)29, 30, 31.

Subsequently, the linkers between the central pyrimidinone and side chains were modified. First, we focused on the left part of the molecule. The carbonyl group of the C-2 alkylthio chain was changed to a thioether (compound 15, EC₅₀ = 120 nmol/L), and the amide was generated by introducing a 4-chlorinated aniline (compound 19 with a decreased EC₅₀ of 210 nmol/L), proving the importance of the carbonyl group of the C-2 alkylthio chain^32,33. Nevertheless, the carbonyl group did not show any hydrogen-bonding interaction with the surrounding residues (Fig. 5A and B). In contrast, the linker on the molecule's right part was also changed. Reports^34,35 show suitable modifications of the methylene bridge between the aryl moiety and pyrimidine might favor the π-stacking interactions between the electron-deficient pyrimidine ring and the electron-rich phenyl ring of RT Tyr188 or Tyr181. Mai et al.³⁴, encouraged by the conformational restriction of a methyl linker, replaced the methylene by a carbonyl group to obtain compound 16, which exhibited an EC₅₀ of 290 nmol/L in HIV-1 infected cells³⁵. However, it was less potent than the methyl analogue probably due to an unfavorable orientation of the 3-F-benzyl moiety and weaker hydrophobic interactions with V106, P225, and F227.

Next, we changed the methyl to a cyano group, commonly considered as a bioisostere of carbonyl, hydroxyl, carboxyl, and halogen groups³⁶. The newly obtained compounds 17 and 20 had better potency (EC₅₀ ∼90 nmol/L). Meanwhile, a chiral center was introduced at the CH–CN position. Based on molecular docking analysis, the two isomers of compound 17 (Fig. 5D and E) showed similar superpositioned binding modes (Fig. 5F) as the cyano group was acceptable in the hydrophobic tunnels of V179/E138 and V106/F227³⁷. Finally, we undertook structural simplifications of the cyano analogues³⁸. When the p-methoxylbenzyl was removed, the alkyl groups, e.g., isopropyl, were accepted in the V106/F227 tunnel (Fig. 5C). Compound 18 had the best EC₅₀ (2.0 ± 0.2 nmol/L) and SI value (4600) in this series. However, higher cytotoxicity (CC₅₀ = 0.8 μmol/L) was observed compared to benzylthio compounds.

2.3. DATA analogues

The DATA, bearing a triazine amine scaffold, is an important family of HIV-1 NNRTIs (Fig. 6)^39,40. The co-crystal complex of compound 21 (R129385) with RT (PDB: 1S9E) provided detailed insight for structure-based optimization⁴¹. As shown in Fig. 7A, compound 21 interacted with RT in a typical “butterfly” conformation. The 4-cyanophenyl was inserted into the hydrophobic pocket containing F227/P236 residues. The central triazine amine was located at the entrance channel of NNRTI binding pocket, a largely open solvent-exposed region in front of K101, E138, and V179 residues. The amides and N-3 formed hydrogen bonds with residues E138 and K101. The 2,6-dichlorophenoxy group was inserted into the aromatic-rich W229 subpocket.

Figure 6.

Structure optimization workflow for the DATA analogues developed by our group. The modified parts are colored.

Figure 7.

Co-crystal complex of compound 21 with RT (PDB: 1S9E) (A); and molecular docking models of RT with compounds 22 (B), 23 (C), and 24 (D). The figures were generated by PyMOL.

Our 3D-QSAR study⁴² and the subsequently disclosed co-crystal complex⁴¹ suggested that increasing the volume of the substituent(s) on the C-6 of triazine ring might favor strengthening the π–π stacking interactions between the ligand and the highly conserved hydrophobic region (W229)⁴³. In addition, the entrance channel was another suitable site for ligand optimization⁴⁴. The newly synthesized compound 22 with a β-naphthyl and N-allyl group showed promising anti-HIV-1 activity (EC₅₀ = 34 ± 1 nmol/L), low cytotoxicity (CC₅₀ = 220 μmol/L), and a good SI value (6475). As expected, the molecular docking analysis showed that the naphthyl group strengthened the π–π stacking interactions with Y181, Y188, and W229, which could be improved dramatically by the electron-withdrawing bromo group. The N-allyl group occupied the entrance channel and maintained the hydrogen bond with residue E138 (Fig. 7B). However, the interaction was not critical to the activity (Fig. 7C) as replacement of allyl by a methyl (compound 23) group improved activity with an EC₅₀ of 9 ± 3 nmol/L and higher SI value (15,385)⁴⁵. Finally, the effect of the α/β-naphthyl was determined. Compound 24 with the α-naphthyl group had an EC₅₀ value of 93 ± 40 nmol/L, which was 10-fold lower than that of the β-naphthyl compound 23. Similar to DABOs, the π-stacking interactions weakened as α-naphthyl was flipped, placing one of the phenyl rings outside the hydrophobic pocket (Fig. 7D). In addition, a series of fluorine-containing DATA analogues (structure not shown) was obtained with no significantly improved antiviral activity⁴⁶. Further optimization by introducing biphenyl to replace the naphthyl ring enhanced the π-stacking interactions with W229 (see details in section 2.4.). Compound 25 with a pyrrolidinol significantly improved the activity against not only WT but also mutant strains except F227L + V106A double mutants. However, its high cytotoxicity (CC₅₀ = 4.5 ± 2.6 nmol/L) made it not suitable for further in vivo evaluation⁴⁷.

2.4. DAPY analogues

The DAPY, first reported in 2001, is a novel class of NNRTIs with highly potent antiviral activity, and ultimately yielded ETR and RPV approved by the FDA⁴⁸. They form a flexible “butterfly” conformation, minimizing the loss of binding stability^41,48. However, a growing number of drug-resistant mutants, especially E138K, are recognized from the failure of RPV or ETR treatments in the clinic^49,50. Besides, the absolute oral bioavailability of ETR in humans is still unknown, due to poor solubility and lack of suitable intravenous formulation^51,52. The high cytotoxicity (CC₅₀ ∼5 μmol/L) of ETR and RPV cannot be ignored. Thus, we focused our efforts to find novel NNRTIs using ETR or RPV as the lead compound, aiming at overcoming drug resistance, and improving safety and pharmacokinetic (PK) profiles (Figure 8, Figure 9, Figure 10).

Figure 8.

Structure (left “wing”) optimization workflow for the DAPY analogues developed by our group. The modified parts are colored.

Figure 9.

Molecular docking models of RT (PDB: 3MEC) with compounds 26 (A), 27 (B), 28 (C) and 33 (F); RT (PDB: 2ZD1) with 29 (D), and 30 (E). The figures were generated by PyMOL.

Figure 10.

Linker optimization workflow for the DAPY analogues developed by our group. The modified parts are colored.

Similar to previous HENT, DABO, and DATA scaffolds, the naphthyl group was also introduced into the left “wing” on the DAPY C4-position based on the co-crystal complex of ETR with RT (Fig. 8)⁵³. Compound 26 with α-bromo-β-naphthyl showed an IC₅₀ value of 2.4 nmol/L against HIV-1 infected MT-4 cells and low cytotoxicity (CC₅₀ = 154 μmol/L) with a high SI of 65,591. However, the activity against the double mutants (K103N + Y181C) was weak (EC₅₀ = 6570 nmol/L). Molecular docking analysis showed that: (1) the central pyrimidine and the NH bridge of the cyanophenyl wing formed two hydrogen-bonding interactions with K101 (Fig. 9A); (2) the naphthyl and cyanophenyl groups inserted into the hydrophobic pockets formed strong π–π interactions (Fig. 9B). Further optimizations of the naphthyl group by introducing a cyano group at C6-position (e.g., compound 27) exhibited higher potency with an extremely high SI value (181,247) against WT HIV-1 but also against the K103N + Y181C double mutant than for compound 26. The calculated results showed enhanced π-stacking interaction between the inhibitor and W229 with an interaction energy (E_int) of −5.55 kcal/mol, much higher than the non-cyano analogue (E_int = −1.76 kcal/mol)⁵⁴. The SAR study showed that the dual-substituted compounds at C1- and C3-positions of the naphthyl ring exhibited more potent activity against the mutant viruses. The di-methoxy group (28) was introduced to constrain the conformation and improve the π-stacking interaction (Fig. 9C), and effectively inhibited WT HIV-1 (EC₅₀ = 5 nmol/L) and K103N + Y181C double mutant (EC₅₀ = 160 nmol/L), showing low cytotoxicity (CC₅₀ > 118 μmol/L) and high SI (>25,000).

To extend the length of the π–π stacking conjugated system in the left “wing”, the biphenyl ring was utilized for fragment-based drug design of biphenyl-DAPYs^55,56. The Alexandre group and our QSAR study^57,58 both demonstrated that interaction of a polar cyano group with W229 was particularly beneficial to improve mutant sensitivity to drugs. The resulting compound 29 with a 4′-cyano group on the biphenyl ring, displayed EC₅₀ values of 1.0, 1.3, 0.8, 1.5, 11.0, 2.0, and 10.0 nmol/L against WT HIV-1, L100I, K103N, Y181C, Y188L, E138K, and K103N + Y181C mutants, respectively⁵⁵. The docking modes showed that the biphenyl ring markedly strengthened the π–π stacking effect by inserting deeply into the W229 pocket as proposed (Fig. 9D). However, 29 exhibited a high cytotoxicity value CC₅₀ of 2.1 μmol/L and a relative low SI of 2059. The aqueous solubility and liver metabolic stability were still unsatisfactory.

The thiophene [3,2-d]pyrimidine scaffold first reported by Liu and Zhan's groups⁵⁹, exhibited excellent safety, low toxicity, and high SI values. We further introduced a thiophene ring into our biphenyl-DAPYs by a scaffold-hopping strategy aiming at overcoming the issue of high cytotoxicity of biphenyl DAPYs (Figure 8, Figure 9E). In this series, compound 30 with the di-fluoro group on the biphenyl ring had much lower cytotoxicity (CC₅₀ = 216.9 μmol/L) and higher SI of 16,094 than compound 29, maintained marked inhibitory activity (EC₅₀ = 13.5, 9.4, 17.0, 52.0, and 58.2 nmol/L) against WT, K103N, E138K, L100I, and Y181C mutants, respectively, and moderate activity towards double mutants⁵⁶. To further explore the conformationally constrained effects, substituents at different positions of the biphenyl ring have been introduced⁶⁰. As a result, the 2′-methyl group, attributed to enlarge the dihedral angle of biphenyl rings and compound 31 with dimethyl groups, exhibited potent biological activity against WT HIV-1-infected MT-4 cells (EC₅₀ = 14 nmol/L) and adequate activity against mutant variants with moderate cytotoxicity (CC₅₀ = 66 μmol/L). Due to high cytotoxicity and metabolic potential of the dimethyl phenyl analogues, a series of non-dimethylphenyl-diarylpyrimidines were also developed. Compound 32 with di-fluoro biphenyl moiety showed good antiviral activity against not only WT (EC₅₀ = 1.3 nmol/L) but also mutant strains (EC₅₀ = 10.8 nmol/L, L100I; 2.6 nmol/L, K103N; 6.1 nmol/L, Y181C; 130 nmol/L, Y188L; 1.9 nmol/L, E138K), and RT enzyme (EC₅₀ = 4.7 nmol/L). The hydrochlorate form of compound 32 exhibited improved water solubility of 5.6 μg/mL much higher than ETR (<<1 μg/mL) and 29, better liver metabolic stability, and a great oral bioavailability of 44% as a drug candidate. In addition, no apparent toxicity was observed in an acute toxicity assay (2 g/kg) and HE staining did not show any damage to the mouse organs.

In contrast, the saturated rings, e.g., cycloalkyl rings, were developed based on bioisosterism principle⁶¹. The left “wing” of DAPYs and three linkers (O, NH or S) between the cycloalkyl and the central pyrimidine were maintained (Fig. 8). The SAR demonstrated that the antiviral activity generally improved in series O < NH < S. The bulky cycloalkyl groups affected the activity as follows: cyclopropyl < cyclopentyl < cyclohexyl, indicating the importance of the left “wing”. Compound 33 with a cyclohexyl group had moderate anti-HIV-1 activity (EC₅₀ = 1160 nmol/L) and no significant improvement was observed with nitrogen replacement (compound 34, EC₅₀ = 960 nmol/L). Compound 35, with a sulfur linker, showed a marked increase in antiviral activity against WT HIV-1 (EC₅₀ = 55 nmol/L, SI = 7290) in cells. The activity in cells might be attributed to the effect on the physicochemical properties but not by strengthening the π–π interaction with W229 pocket (Fig. 9F).

Encouraged by the series of compounds 33–35, the linker variations between the left “wing” and the central pyrimidine moiety are other important factors for optimization. Additionally, the conformational flexibility and positional adaptability of the linker, are vitally important in the design of NNRTIs⁶². The hydroxyimino, hydrazinyl, alkylamino, hydroxyl, halogen, and cyano groups were introduced on the methylene linker (Fig. 10).

The best compound 36 in the hydroxyiminate series inhibited HIV-1 in infected cells with an IC₅₀ of 6 nmol/L but lacked activity against double mutants⁶³. Predicted binding modes explained the activity (Fig. 11). The molecule was in the “horseshoe” conformation, and a new hydrogen-bonding interaction between the hydroxyimino and E138 was observed. However, the compound did not increase the interaction with the W229 in the hydrophobic region (Fig. 11A). Similarly, hydrazinyl linker (37) was suitable to DAPYs with an EC₅₀ of 9 nmol/L and SI value of 39,599⁶⁴. With the introduction of a cyanovinyl group, compound 38 increased the antiviral activity (EC₅₀) to 1.8 nmol/L and SI to 106,367, probably due to enhanced interactions of the cyanovinyl group with Y181, Y188, F227, and W229 residues (Fig. 11B)⁶⁵. The linker, when replaced by different alkylamino groups led to lower potency and activity in the series isopropylamino > n-propylamino > methylamino > ethylamino ≈ cyclopropylamino⁶⁶. The hydrogen–bonding interaction of hydrazinyl with E138 disappeared compared with 38, the cyclopropyl group inserted into the Y181/E138 channel, and the di-fluoro groups limited the rotational freedom of the phenyl ring (Fig. 11C), collectively leading to compound 39 with an EC₅₀ of 99 nmol/L.

Figure 11.

Molecular docking models of RT (PDB: 3MEC) with compound 36 (A), 38 (B), 39 (C), 40-R (D), 40-S (E) and 43 (F). The figures were generated by PyMOL.

In 2011, we also designed and synthesized a series of DAPYs bearing a hydroxyl on the methylene linker⁶⁷. They possessed excellent activities against WT HIV-1 at nanomolar concentrations and moderate activities against the double mutant K103N + Y181C. The racemic compound 40 showed an EC₅₀ of 9 nmol/L and SI of 4115. The chiral center on the hydroxymethyl linker was taken into consideration, and supercritical fluid chromatography (SFC) afforded both compound (+)-40 and (−)-40 with enantiomeric excess (ee) value > 99% and purity >99%⁶⁸. Their absolute configurations were assigned as R and S, respectively, based on the experimental electronic circular dichroism (ECD) calculations. They exhibited different antiviral activity: the (+)-(R)-40 had an EC₅₀ of 5 nmol/L, which was 12-fold more potent than the (−)-(S)-40 (EC₅₀ = 65 nmol/L). Docking studies showed that they targeted a similar RT conformation as other DAPYs. A hydrogen–bonding interaction between the hydroxyl group and Y188 was observed in the (+)-(R)-40-RT model (Fig. 11D) while not seen in the (−)-(S)-40-RT model (Fig. 11E), thus interpreting the biological result. Further alkyl modification of the linker produced new CR₂(OH)-DAPYs and the best compound 41 in this series had an EC₅₀ of only 67 nmol/L⁶⁹. The halogen substitution was optimized to obtain compound 42 with an EC₅₀ of 5 nmol/L⁷⁰. For the positive influence of the cyano substitution on the phenyl ring, the chloride was replaced by a cyano group to obtain compound 43 with superior antiviral activity (EC₅₀ = 2 nmol/L) and high SI value (>118,595)⁷¹. As shown in Fig. 11F, the R/S enantiomers of compound 43 displayed similar binding conformations with the cyano group inserted in the Y181/E138 tunnel, which might be a key feature for the activity. However, compound 43 was inactive towards the K103N + Y181C double mutant.

2.5. Molecular hybrids

Molecular hybridization or fragment-based optimization is a highly efficient strategy in drug discovery⁷². This method is also applied widely in the design of next-generation NNRTIs to overcome drug resistance, improve safety, and/or physicochemical properties (Fig. 12)^7,72.

Figure 12.

Molecular hybridization workflow for the DAPY analogues developed by our group. The modified parts are colored.

In 2015, a series of hybrids were designed by hopping the thioacetanilide moiety onto the right “wing” of DAPY, based on the superposition of lower-energy binding conformations of ETR and VRX-480,773⁷³. Compound 44 displayed moderate inhibitory activity against WT HIV-1 RT with an EC₅₀ value of 240 nmol/L and bound to the NNRTI binding pocket in a “U” mode (Fig. 13A). The SAR study showed an encouraging activity of the electron-withdrawing substituents on the right “wing”. The antiviral activity was lost when the amide group was removed, indicating its importance⁷⁴. The sulfide of compound 44 was oxidized to the sulfone, and the strong electron-withdrawing nitro group introduced to obtain compound 45, had 180-fold better antiviral activity and a much higher SI value of 45,830⁷⁵. Molecular docking showed that the compound bound well to RT (Fig. 13B) and super-positioned with compound 44 (Fig. 13C); two hydrogen-bonding interactions between the sulfone and K101 and the nitro, and F227 (Fig. 13B) were observed. Inspired by the promising activity of the above hybrids, a family of piperidin-4-yl-aminopyrimidine derivatives was obtained by combining ETR-VRX-480,773 hybrids and piperidine-linked aminopyrimidines, another family of NNRTIs⁷⁶. Compounds 46 and 47 displayed highly potent activity against WT HIV-1 with an EC₅₀ of 7.0 and 1.9 nmol/L. However, compound 47 showed high toxicity (CC₅₀ = 1.6 μmol/L), low SI value (845) and lacked potency against the double mutant strains K103N + Y181C, thus needing further optimization.

Figure 13.

Molecular docking models of RT (PDB: 3MEC) with compounds 44 (A), 45 (B), 44 and 45 (C), 48 (D), 50 (E) and 52 (F). The figures were generated by PyMOL.

Then, we focused on fusing pharmacophores to the left “wing” of DAPYs. The diaryl ethers were reported as a class of NNRTIs with excellent oral bioavailability and anti-HIV potency⁷⁷. Compound 48 combining the important pharmacophores of DAPYs and diaryl ethers, displayed good activity against WT HIV-1 with an EC₅₀ of 13 nmol/L⁷⁸. The newly introduced 2-fluorophenylamide group was located in the W229 hydrophobic pocket (Fig. 13D). Similarly, compound 49 combined the structural features of DAPY and GW678248 (structure shown in Fig. 15) led to a reduced anti-HIV activity (EC₅₀ = 790 nmol/L)⁷⁹. We attempted to synthesize hybrid DAPY with the integrase antiviral inhibitor, the first quinolone-based anti-HIV drug, elvitegravir⁸⁰. The newly designed diarlyprimidine-quinolone hybrid compound 50 displayed an EC₅₀ value of 280 nmol/L against WT HIV-1 infected cells. An enzymatic assay with the compound showed an EC₅₀ of only 7.2 ± 0.5 μmol/L, indicating the highly polar quinolone 3-carboxylic acid moiety was not suitable for binding to the hydrophobic pocket (Fig. 13E). The follow-up work showed that the length of the N-substituted alkyl group and the introduction of an iodine atom on the quinolone backbone affected the activities. Compound 51 displayed a significant EC₅₀ value of 9.6 nmol/L against WT HIV-1 and of 0.98 μmol/L against K103N + Y181C⁸¹. This strategy of integrating pharmacophores of different types of inhibitors might lead to compounds with dual targeting effects, needing further characterization.

Figure 15.

Other antiviral inhibitors developed by our group. The modified parts are colored.

Finally, we focused our attention on the entrance channel by fusing a phenyl ring (similar as compound 30) based on the superposition model of DPC083 or EFV, and ETR in the binding pocket of HIV-1 RT^82,83. For instance, compound 52, with the two “wings” of ETR unchanged, had impressive activity against WT HIV-1 with an EC₅₀ value of 1.8 nmol/L, and with an SI value of 111,954. It was also highly active in the nanomolar range against a broad range of viral mutants including L100I (EC₅₀ = 18 nmol/L), K103N (EC₅₀ < 3.6 nmol/L), E138K (EC₅₀ = 11 nmol/L), Y181L (EC₅₀ = 31 nmol/L), Y188L (EC₅₀ = 36 nmol/L), F227L + V106A (EC₅₀ = 107 nmol/L), and K103N + Y181C (EC₅₀ = 60 nmol/L)⁸². As proposed, the fused phenyl ring was located at the entrance channel of E138/K103/K101 residues (Fig. 13F), probably maintaining the conformational flexibility to compensate for the resistant mutants’ effect.

Very recently, we developed a series of novel diarylbenzopyrimidine analogues by hybridizing FDA-approved drugs ETR and EFV⁸³. Introduction of the important fragment of RPV, cyanovinyl group and substituent modifications on the left “wing” resulted in the development of new derivatives with the combined strength of the two drugs. Compound 53 showed promising activity towards the EFV-resistant K103N mutant (EC₅₀ = 10.2 nmol/L) but was not better against E138K (EC₅₀ = 17.7 nmol/L) than ETR. The PK profile was acceptable with a rat oral bioavailability of 15.5%. Substitution optimizations on the fused phenyl ring led to compound 54 with much higher activity than ETR and EFV against the WT, E138K, and K103N variants (EC₅₀ = 3.4, 4.3, and 3.6 nmol/L, respectively) and a rat oral bioavailability of 16.5%, which should be better than that of ETR. Molecular docking (Fig. 14) analysis showed that three hydrogen bonds stabilized the “butterfly” conformation and the fused phenyl ring occupied the E138/K101 channel. In the mutants, the orientation of the new K138 was altered due to the existence of the adjacent K101, making the space more accessible to the fused ring. For the N103 mutant, the 2-methyl-6-nitro group restricted the conformation, and it was oriented towards new N103, potentially resulting in a water-mediated hydrogen bond with the NH of amide. Collectively, the binding features explain the improved antiviral activity and guide further DAPY optimization to improve the activity against WT and the mutants.

Figure 14.

Molecular docking models of compounds 53 and 54 with the HIV-1 WT and E138K/K103N mutant RT (A) WT with 53 (B) E138K mutant RT with 54 (C) K103N mutant RT with 53 (D) WT with 54 (E) E138K mutant RT with 54 (F) K103N mutant RT with 54. The figures were generated by PyMOL.

2.6. Others

Alternatively, we attempted to hybridize the pharmacophore of DAPY into early generation NNRTIs^84,85. For instance, compound 55 was developed by introducing the 4-cyanophenyl group into EFV, with an EC₅₀ value of 0.84 nmol/L against the WT HIV-1 strain, and 3.5 and 66.0 nmol/L against E138K and K103N + Y181C mutants. However, the cytotoxicity was high (CC₅₀ = 1.9 μmol/L)⁸⁴. The B-ring of GW678248 was replaced by naphthyl ring as NNRTIs⁸⁵. However, they showed only moderate activity against WT HIV-1 and mutated viruses. Compound 56, the best one in this series, showed an EC₅₀ of 4.8 nmol/L against WT HIV-1 and 2.1 μmol/L against the HIV-1 double mutant K103N + Y181C, exhibiting much less potency than the parent GW678248.

3. Conclusions and future perspectives

In the past 20 years, several commonly used medicinal chemistry strategies, including structure-based drug design, fragment-based optimization, scaffold/fragment hopping, molecular/fragment hybridization and bioisosterism, have been utilized in our NNRTI drug development. Progress in structural optimizations of HEPT, DABO, DATA, and DAPY scaffolds have been made, leading to highly potent derivatives against not only WT HIV-1 but also single/double mutant variants. However, the growing incidence of resistance to nearly all current drugs is investigated continually. Resistance-associated mutations occur within or near the NNIBP, resulting in decreased efficacy of NNRTIs86, 87, 88. Thus, more novel compounds to treat HIV infections are urgently needed. Nevertheless, novel and high-quality lead compounds, especially towards mutants remains a significant challenge in the current anti-HIV drug discovery. There are some perspectives and directions to be considered for the development of next-generation NNRTIs.

• (1)The number of rotatable bonds is essential for antiviral activity. Except for DLV, the FDA approved drugs having more freedom of rotation in the binding site had better activity against RT mutants³. For instance, RPV showed significantly improved activity in various mutants with the introduction of the cyanovinyl group. In addition, the cyanovinyl group inserted into the W229 hydrophobic tunnel strengthened the π–π stacking interactions. Our compounds 53 and 54 with cyanovinyl showed good anti-HIV-1 WT, E138K and K103N activities. The biphenyl derivative compounds 29 and 30, and the methylene linker optimized compounds (36–43) with more rotatable bonds showed much higher potency and antiviral profiles.
• (2)Forming additional hydrogen-bonding interactions with the targeted protein is a common strategy to enhance the binding affinity. In the NNRTI development, this method has been used widely to improve the activity against WT HIV-1 or prevent efficacy reduction towards mutants⁸⁹. For instance, the sulfone of compound 45 formed additional hydrogen-bonding interactions with K101, possessing 180-fold better antiviral activity than the corresponding sulfide compound 44⁷⁵.
• (3)The W229 hydrophobic region has to be occupied for improving π–π stacking interactions with surrounding aromatic residues, e.g., the introduction of naphthyl, biphenyl, cyanovinyl group, or others. The halogen interaction⁹⁰, methyl effect⁹¹, and cyano group³⁶ could be applied for further optimizations.
• (4)The optimization of the PK profiles will be addressed in the future. For instance, ETR has very poor water solubility and lacks an intravenous formulation leading to unknown absolute oral bioavailability^51,52,83. High plasma protein binding (>95%) of DLV, EFV, ETR, and RPV was observed, and >60% NVP binding with a long half-life, should be considered during clinical treatments³. Introducing halogen, especially F, CF₃ group, sulfur, phosphor, silicium, and others might increase the antiviral activity and PK profiles. Introducing solubilizing groups at the entrance channel and other suitable sites, and design of prodrugs are other strategies to improve the physicochemical properties. Compound 31 with fluoro substitutions exhibited an improved PK profile as well as reduced cytotoxicity. Liu and Zhan et al.^52,92, 93, 94, 95, 96 succeeded in enhancing the PK, water solubility, and oral bioavailability by structural optimizations of multiple NNRTIs.
• (5)Clinically, the use of NNRTIs is always in conjunction with two NRTIs as HHART³. The efficacy, especially the safety of several drug regimens, is of great importance in the clinical evaluation. From the medicinal chemistry view, it is possible to make a NNRTI and NRTI conjugate act as a bifunctional inhibitor due to the proximity (∼10 Å, Fig. 1B) of the NRTI and NNRTI binding sites in RT. In 2013, the bivalent NNRTI and NRTI inhibitors had been developed by Anderson et al.⁹⁷, and the representative compound (the triphosphate of d4T-4PEG-DAPY analogue conjugate, structure not shown here) inhibited RT polymerase at low nanomolar concentrations, which was more potent than the two parent drugs. Besides, a prodrug design of these agents or other modern strategies, such as proteolysis targeting chimeras (PROTACs) to degrade HIV-1 RT, are prospected. Recently, a class of small molecule antivirals were developed to induce proteasomal degradation of HCV viral proteins to overcome resistant viral variants. This work highlights reference to HIV targeted protein degradation and a new paradigm for the development of molecules with superior resistance profiles⁹⁸.

The above-discussed directions will undoubtedly attract further interest in the coming years and predict that novel NNRTI-based drugs will be discovered using modern drug discovery strategies and new chemical sources. Phenotypic screening, fragment-based drug design, structure/ligand-based virtual screening, DNA-encoded library generation and screening, conjugation chemistry, prodrugs, PROTACs, etc. will be applied widely for the rapid development of future antiviral drugs against HIV infection.

Author contributions

Chunlin Zhuang and Fener Chen generated the manuscript draft. Christophe Pannecouquec and Erik De Clercq edited and revised the manuscript. All authors reviewed and approved the final version of the manuscript.

Conflicts of interest

The authors have no conflicts of interest to declare.

Appendix A. Supplementary data

The following is the Supplementary data to this article: