A comparison of the ability of rilpivirine (TMC278) and selected analogues to inhibit clinically relevant HIV-1 reverse transcriptase mutants

Abstract

Background: The recently approved anti-AIDS drug rilpivirine (TMC278, Edurant) is a nonnucleoside inhibitor (NNRTI) that binds to reverse transcriptase (RT) and allosterically blocks the chemical step of DNA synthesis. In contrast to earlier NNRTIs, rilpivirine retains potency against well-characterized, clinically relevant RT mutants. Many structural analogues of rilpivirine are described in the patent literature, but detailed analyses of their antiviral activities have not been published. This work addresses the ability of several of these analogues to inhibit the replication of wild-type (WT) and drug-resistant HIV-1.

Results: We used a combination of structure activity relationships and X-ray crystallography to examine NNRTIs that are structurally related to rilpivirine to determine their ability to inhibit WT RT and several clinically relevant RT mutants. Several analogues showed broad activity with only modest losses of potency when challenged with drug-resistant viruses. Structural analyses (crystallography or modeling) of several analogues whose potencies were reduced by RT mutations provide insight into why these compounds were less effective.

Conclusions: Subtle variations between compounds can lead to profound differences in their activities and resistance profiles. Compounds with larger substitutions replacing the pyrimidine and benzonitrile groups of rilpivirine, which reorient pocket residues, tend to lose more activity against the mutants we tested. These results provide a deeper understanding of how rilpivirine and related compounds interact with the NNRTI binding pocket and should facilitate development of novel inhibitors.

Figure 1

Chemical structures of approved NNRTIs rilpivirine (1), etravirine (2), nevirapine (3), efavirenz (4) and delaviridine (5).

Figure 2

Chemical structures of rilpivirine analogues. Analogues differed from rilpivirine (1, boxed) by the addition of an exocyclic moiety to the pyrimidine ring in either of its ‘flipped’ conformations (6–9), the replacement of the pyrimidine ring with a 2,6-purine ring system in ‘flipped’ conformations with or without a protecting group (10–13), replacing the 4-benzonitrile moiety in addition to replacing the pyrimidine ring with a 2,6-purine (14–25), or replacing the pyrimidine ring with a 2,9-purine ring system (26, 27). Differences between each analogue and rilpivirine are indicated in red.

Figure 3

Activities and cytotoxicities of rilpivirine (1) and its analogues. EC_50s, as measured by reduction of luciferase reporter activity, are reported as nM values, CC_50s are reported as μM values. Standard deviations are indicated in parentheses. Shading of EC₅₀ values indicates a log scale: no shading = EC₅₀ <1.0 nM, light gray = EC₅₀ between 1.0 nM and 9.9 nM, dark gray = EC₅₀ between 10 nM and 99nM, black = EC₅₀ ≥100 nM. Therapeutic Index (T.I.) is the ratio of CC₅₀/EC₅₀.

Figure 4

Comparison of EC₅₀ values of selected analogs against an expanded set of mutants. Values for mutants not shown in Figure 3 are included in Additional file 1: Table S1.

Figure 5

Overlay of crystal structures and models of 1 and 16 bound to RT. Co-crystal structures of WT RT bound to 1 (gray) and 16 (green) show only minimal differences in the position of the compound and the conformation of the binding pocket. This is consistent with the similar activity of the two compounds against WT RT (0.2nM for 1 and 0.4nM for 16). Models were generated for each compound bound to Y188L RT (1: magenta, 16:cyan). Both models show the compound positioned approximately one angstrom further into the binding pocket than in the respective WT co-crystal structure. This results in a repositioning of the F227 and W229 side chains and a shift in the overall positioning of the β12-β13 hairpin (shown as a ribbon). These differences in the modeled interactions are consistent with the observed difference in antiviral activities against vectors using Y188L RT (2.3nM for 1 and 29nM for 16).

Figure 6

NNRTI-binding pocket superposition of the crystal structures of K103N-Y181C mutant RT/21 (yellow protein and green ligand) and K103N/Y181C mutant RT/rilpivirine (PDB ID. 3BGR; blue protein and gray ligand). The hydrogen bonds are represented as dotted lines, and the most significant structural differences are indicated by red arrows.

Figure 7

Models of analogues 7 (cyan) and 9 (magenta) bound to K103N/Y181C RT. (A) Overlay of the two compounds as modeled in the binding pocket of K103N/Y181C HIV-1 RT. (B) The exocyclic amine in 9 cannot H-bond with the backbone NH of K101, as seen in crystal structures of 1. The N3 and linker NH both bind a water molecule that forms a bridge to the backbone carbonyl of E138. (C) The H-bonds with the backbone of K101 are restored in 7. In addition, the exocyclic amine and linker NH both H-bond a water molecule in a different orientation than is seen with 9.