Effect of Mutations at Position E138 in HIV-1 Reverse Transcriptase and Their Interactions with the M184I Mutation on Defining Patterns of Resistance to Nonnucleoside Reverse Transcriptase Inhibitors Rilpivirine and Etravirine

Abstract

Impacts of mutations at position E138 (A/G/K/Q/R/V) alone or in combination with M184I in HIV-1 reverse transcriptase (RT) were investigated. We also determined why E138K is the most prevalent nonnucleoside reverse transcriptase inhibitor mutation in patients failing rilpivirine (RPV) therapy. Recombinant RT enzymes and viruses containing each of the above-mentioned mutations were generated, and drug susceptibility was assayed. Each of the E138A/G/K/Q/R mutations, alone or in combination with M184I, resulted in decreased susceptibility to RPV and etravirine (ETR). The maximum decrease in susceptibility to RPV was observed for E138/R/Q/G by both recombinant RT assay and cell-based assays. E138Q/R-containing enzymes and viruses also showed the most marked decrease in susceptibility to ETR by both assays. The addition of M184I to the E138 mutations did not significantly change the levels of diminution in drug susceptibility. These findings indicate that E138R caused the highest level of loss of susceptibility to both RPV and ETR, and, accordingly, E138R should be recognized as an ETR resistance-associated mutation. The E138K/Q/R mutations can compensate for M184I in regard to both enzymatic fitness and viral replication capacity. The favored emergence of E138K over other mutations at position E138, together with M184I, is not due to an advantage in either the level of drug resistance or viral replication capacity but may reflect the fact that E138R and E138Q require two distinct mutations to occur, one of which is a disfavorable G-to-C mutation, whereas E138K requires only a single favorable G-to-A hypermutation. Of course, other factors may also affect the concept of barrier to resistance.

INTRODUCTION

The reverse transcriptase (RT) of human immunodeficiency virus type 1 (HIV-1) is an important target for anti-HIV drugs. RT is a multifunctional enzyme possessing each of RNA and DNA polymerase activities as well as an RNase H activity (1, 2) and is responsible for converting the single-stranded viral RNA genome into double-stranded DNA (dsDNA), which becomes integrated into host cell DNA. Currently, two types of RT inhibitors have been approved for treatment of HIV-1 infection, i.e., nucleoside reverse transcriptase inhibitors (NRTIs) and nonnucleoside reverse transcriptase inhibitors (NNRTIs). NRTIs, which act as competitive inhibitors and cause chain termination of the growing viral DNA chain, include zidovudine (AZT, ZDV), didanosine (ddI), stavudine (d4T), dideoxycytosine (ddC; zalcitabine), lamivudine (3TC), emtricitabine (FTC), abacavir (ABC), and a nucleotide, tenofovir disoproxil fumarate (TFV). NNRTIs, which act allosterically by binding to the NNRTI binding pocket (BP) located 10 Å from the polymerase active site (3), include the earlier drugs nevirapine (NVP), delavirdine (DLV), and efavirenz (EFV) and newer products, such as etravirine (ETR) and rilpivirine (RPV). Both NRTIs and NNRTIs are key components of highly active antiretroviral therapy (HAART), which has been effective in suppressing HIV-1 replication, reducing HIV-1-associated complications, and prolonging the lives of treated patients (4, 5). However, both classes of drugs can be compromised by drug resistance (6, 7), which, in the case of NNRTIs, is due to mutations within the NNRTI binding pocket (8).

One major characteristic of the earlier NNRTIs is that they have a low genetic barrier for resistance, and cross-resistance among NNRTIs is common (9, 10), such that the sequential use of earlier NNRTIs is excluded in treatment-experienced patients. Recently, efforts to discover novel NNRTIs have led to new compounds with improved potency and resistance profiles that have overcome some of the limitations associated with earlier NNRTIs. Two newer NNRTI diarylpyrimidine (DAPY) drugs are ETR (11, 12) and RPV (13), which have been approved for use in treatment-experienced and drug-naive patients, respectively (14, 15). One major characteristic of these newer NNRTIs is their inherent binding flexibility that helps to maintain activity against viruses containing a wide range of resistance mutations associated with the earlier NNRTIs NVP and EFV (16). Nonetheless, both ETR and RPV, like all other antiretrovirals, can also select for drug resistance mutations that can compromise their antiviral activities.

Structural studies have revealed a similar binding mode within the NNRTI binding pocket for ETR and RPV (16–18), yet currently available data indicate that ETR and RPV might have different resistance profiles. Many data sets regarding ETR resistance have been derived from the phase III DUET studies (19–22), and 20 mutations, including V90I, A98G, L100I, K101E/H/P, V106I, E138A/K/G/Q, V179D/F/T, Y181C/I/V, G190A/S, and M230L, are considered to be ETR resistance-associated mutations (RAMs) (23–26). In the DUET studies, the most frequently emerging RT mutations were V179F/I and Y181C, although frequent changes were also observed at positions K101 and E138 (27). In the case of RPV, 15 mutations, including K101E/P, E138A/G/K/Q/R, V179L, Y181C/I/V, H221Y, F227C, and M230I/L, have been recognized to be RAMs (24). In the phase III clinical trials that led to the approval of RPV (ECHO and THRIVE), E138K was the most frequent mutation to emerge, in most cases together with M184I, in patients who failed RPV-based therapy and who also received two nucleos(t)ides, most commonly emtricitabine and tenofovir disoproxil fumarate (TDF) (15, 28, 29).

We and others recently showed that the E138K and M184I mutations can mutually compensate for each other to restore the enzymatic fitness that is compromised by either mutation alone (30–32). Due to structural similarities between ETR and RPV, it is not surprising that overlapping resistance mutations were observed for these two newer NNRTIs (24), with mutations at E138 seeming to play an important role (16). Although the phenotypic susceptibility of E138A/G/K/Q to RPV has been established (33), detailed enzymatic and virological analyses of RPV resistance for mutations at position E138 in HIV-1 RT have not been reported.

The combination of RPV, TDF, and FTC, coformulated as a single-tablet regimen (Complera), is approved for use in first-line antiretroviral therapy. In view of the fact that RPV selects for mutations at position E138 while FTC selects for M184I, it is important to study the effects of various mutations at position E138, alone and in combination with M184I, on RPV and ETR resistance. Cross-resistance between RPV and ETR might also limit the use of ETR as part of salvage regimens in patients who have failed RPV-containing regimens.

The current study was performed to investigate the impact of mutations at E138 (A/G/K/Q/R/V), alone and together with M184I, on RPV susceptibility, enzyme properties, and viral replication capacity. We also wished to provide mechanistic insight into reasons for the favored selection of E138K versus other E138 mutations in the ECHO and THRIVE clinical studies. E138V was also chosen for study because it had been observed in patients who failed ETR-containing therapy in the DUET studies (23) and in cell culture under ETR pressure (34, 35) and has not previously been characterized.

MATERIALS AND METHODS

Chemicals, cells, and nucleic acids.

Etravirine (ETR) and rilpivirine (RPV) were gifts of Janssen Pharmaceuticals (Titusville, NJ). Emtricitabine (FTC) was kindly provided by Gilead Sciences (Foster City, CA). Lamivudine (3TC) was a gift of GlaxoSmithKline (Greenford, United Kingdom).

Cord blood mononuclear cells (CBMCs) were obtained through the Department of Obstetrics, Jewish General Hospital, Montreal, QC, Canada. The HEK293T cell line was obtained from the American Type Culture Collection (ATCC). The following reagents and cells were obtained through the NIH AIDS Research and Reference Reagent Program: the infectious molecular clone pNL4-3 from Malcolm Martin and TZM-bl (JC53-bl) cells from John C. Kappes, Xiaoyun Wu, and Tranzyme Inc.

The pNL4.3PFB proviral DNA was a generous gift from Tomozumi Imamichi, National Institutes of Health, Bethesda, MD. The plasmid pRT6H-PROT was a generous gift from Stuart F. J. Le Grice, National Institutes of Health, Bethesda, MD.

An HIV-1 RNA template ∼500 nucleotides (nt) in size spanning the 5′ untranslated region (UTR) to the primer binding site (PBS) was transcribed in vitro from AccI-linearized pHIV-PBS DNA (36) by using an Ambion T7-MEGAshortscript kit (Invitrogen, Burlington, ON, Canada) as described previously (37). The oligonucleotides used in this study were synthesized by Integrated DNA Technologies Inc. (Coralville, IA) and purified by polyacrylamide-urea gel electrophoresis. For 5′ end labeling of oligonucleotides with [γ-³²P]ATP, an Ambion Kinase Max kit was used, followed by purification through Ambion NucAway spin columns, according to protocols provided by the supplier (Invitrogen, Burlington, ON, Canada).

SDM and preparation of virus stocks.

To construct HIV-1 RT expression plasmids and HIV-1_NL4-3 variants harboring desired mutations in the RT gene, site-directed mutagenesis (SDM) reactions were first carried out using a QuikChange II XL site-directed mutagenesis kit (Stratagene, La Jolla, CA) on an HIV-1 RT expression plasmid, pbRT6H-PROT (37). To make recombinant HIV-1_NL4-3 containing the desired RT mutations, we first amplified fragments spanning RT amino acids (aa) 25 to 314 from the pbRT6H-PROT variants by PCR; the resultant 871-bp mutant DNA after restriction digestion was used to replace the corresponding 871-bp fragment of pNL4.3PFB proviral DNA (38). DNA sequencing was performed to verify the absence of spurious mutations and the presence of any desired mutation. Recombinant HIV-1_NL4-3 wild-type and mutant viruses were generated by transfection of the corresponding proviral plasmid DNAs into HEK293T cells using Lipofectamine 2000 (Invitrogen, Burlington, ON, Canada) according to the manufacturer's instructions. Viral supernatants were harvested at 48 h posttransfection, centrifuged for 5 min at 800 × g to remove cellular debris, filtered through a 0.45-μm-pore size filter, aliquoted, and stored at −80°C. Levels of p24 in the viral supernatant were measured by a PerkinElmer HIV-1 p24 antigen enzyme-linked immunosorbent assay (ELISA) kit. Virion-associated RT activity was measured as described previously (1) with 50 μl of an RT reaction mixture containing 10 μl of culture supernatants, 0.5 U/ml of poly(rA)/p(dT)_12-18 template/primer (T/P; Midland Certified Reagent Company, Midland, TX) in 50 mM Tris-HCl (pH 7.8), 75 mM KCl, 5 mM dithiothreitol (DTT), 5 mM MgCl₂, 0.05% Triton X-100, 2% ethylene glycol, 0.3 mM reduced glutathione, and 5 μCi of [³H]dTTP (70 to 80 Ci/mmol, 2.5 mCi/ml). Following a 240-min incubation at 37°C, the reaction was quenched by adding 0.2 ml of 10% cold trichloroacetic acid (TCA) and 20 mM sodium pyrophosphate and the mixture was incubated for at least 30 min on ice. The precipitated products were filtered onto Millipore 96-well MultiScreen HTS FC filter plates (MSFCN6B) and sequentially washed with 200 μl of 10% TCA and 150 μl of 95% ethanol. The radioactivity of incorporated products was analyzed by liquid scintillation spectrometry using a PerkinElmer 1450 MicroBeta Tri Lux microplate scintillation and luminescence counter.

Recombinant RT expression and purification.

Recombinant RTs in heterodimeric form were expressed from plasmid pbRT6H-PROT (37) and purified as described previously (39, 40) with minor modifications. In brief, RT expression in Escherichia coli M15(pREP4) (Qiagen, Mississauga, ON, Canada) was induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at room temperature. The pelleted bacteria were lysed under native conditions with BugBuster protein extraction reagent containing Benzonase endonuclease (Novagen, Madison, WI) according to the manufacturer's instructions. After clarification by high-speed centrifugation, the clear supernatant was subjected to the batch method of Ni-nitrilotriacetic acid (NTA) metal affinity chromatography (QIAexpressionist; Qiagen, Mississauga, ON, Canada). All buffers contained complete protease inhibitor cocktail (Roche, Mississauga, ON, Canada). Hexahistidine-tagged RT was eluted using an imidazole gradient. RT-containing fractions were pooled, passed through DEAE-Sepharose (GE Healthcare, Mississauga, ON, Canada), and further purified using SP-Sepharose (GE Healthcare, Mississauga, ON, Canada). Fractions containing purified RT were pooled, dialyzed against storage buffer (50 mM Tris-HCl [pH 7.8], 50 mM NaCl, 50% glycerol), and concentrated to 4 to 8 mg/ml with a Centricon Plus-20 30-kDa-molecular mass cutoff membrane (Millipore, Etobicoke, ON, Canada). Aliquots of proteins were stored at −80°C. Protein concentration was measured by a Bradford protein assay kit (Bio-Rad Laboratories, Saint-Laurent, QC, Canada), and the purity of the recombinant RT preparations was verified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The RNA-dependent DNA polymerase (RDDP) activity of each recombinant RT preparation was evaluated as described previously (41) using various concentrations of RT and a synthetic homopolymeric poly(rA)/p(dT)_12-18 T/P (Midland Certified Reagent Company, Midland, TX).

RT inhibitor susceptibility assays with recombinant RT enzymes.

Susceptibility to RT inhibitors was assayed using recombinant RT enzymes and a heteropolymeric HIV-1 PBS RNA T/P system as described previously (37). Briefly, RT reaction mixtures containing 50 mM Tris-HCl (pH 7.8), 6 mM MgCl₂, 60 mM KCl, deoxynucleotide triphosphates (dNTPs; 5 μM each) with [³H]dTTP as a tracer, 30 nM heterogeneous HIV-1 RNA template/primer, the same activity of RT enzymes incorporating ∼200,000 cpm radioactivity, and variable concentrations of RT inhibitors were included in 50-μl reaction volumes. Reaction mixtures were incubated at 37°C for 30 min, the reactions were terminated by adding 0.2 ml of 10% cold TCA and 20 mM sodium pyrophosphate, and the mixtures were incubated for at least 30 min on ice. The precipitated products were filtered through a 96-well MultiScreen HTS FC filter plate (Millipore, Etobicoke, ON, Canada) and sequentially washed with 200 μl of 10% TCA and 150 μl of 95% ethanol. The radioactivity of the incorporated products was analyzed by liquid scintillation spectrometry using a 1450 MicroBeta Tri Lux microplate scintillation and luminescence counter (PerkinElmer, Woodbridge, ON, Canada). The 50% inhibitory concentrations (IC₅₀s) of ETR and RPV were determined by nonlinear regression analysis using the GraphPad Prism (version 5.01) program (GraphPad Software, San Diego, CA).

Phenotypic drug susceptibility assays with recombinant HIV-1.

Phenotypic susceptibility analysis of RT inhibitors was performed with recombinant HIV-1_NL4-3 clones in a TZM-bl cell-based in vitro assay as described previously (42). Briefly, RT inhibitors at variable concentrations were added to TZM-bl cells (2 × 10⁴ cells/well) grown in 100 μl Dulbecco's modified Eagle medium (DMEM; Invitrogen, Burlington, ON, Canada) supplemented with 10% fetal bovine serum (Invitrogen, Burlington, ON, Canada), 1% penicillin-streptomycin, and 1% l-glutamine (Invitrogen, Burlington, ON, Canada) in 96-well plates. Immediately after drug addition, cells were infected with wild-type (WT) or mutant viruses. Standardization of virus infections was determined by adding virus at a p24 concentration that would yield ∼160,000 ± 20,000 relative light units (RFU), as detected by luminescence. After 48 h, 100 μl of Bright-Glo reagent (Promega; Fisher Scientific, Nepean, ON, Canada) was added to 100 μl of infected TZM-bl cells. Cell lysates were then transferred to a white, opaque, 96-well plate (Corning; Fisher Scientific, Nepean, ON, Canada). Luciferase assay reagent (Promega; Fisher Scientific, Nepean, ON, Canada) was added to each well, and the numbers of RLU/well were measured by a PerkinElmer 1450 MicroBeta Tri Lux microplate scintillation and luminescence counter. Drug efficacy was determined by quantifying luciferase activity as a measure of viral replication. The 50% effective drug concentration (EC₅₀) was calculated using the GraphPad Prism (version 5.01) program (GraphPad Software, San Diego, CA).

Processivity assays.

The processivities of recombinant RT enzymes were analyzed as described previously using heteropolymeric HIV-1 PBS RNA template in the presence of a heparin enzyme trap to ensure a single processive cycle, i.e., a single round of binding and of primer extension and dissociation (30). The T/P was prepared by annealing the HIV-1 PBS RNA with the 25-nt DNA primer D25 labeled with ³²P at the 5′ end at a molar ratio of 1:1, denatured at 85°C for 5 min, and then slowly cooled to room temperature to allow specific annealing of the primer to the template. RT enzymes with equal amounts of activity (800 nM to 1,600 nM) and 40 nM T/P were preincubated for 5 min at 37°C in a buffer containing 50 mM Tris-HCl (pH 7.8), 50 mM NaCl, and 6 mM MgCl₂. Reactions were initiated by the addition of 0.5 μM dNTPs and heparin trap (final concentration, 3.2 mg/ml), and the reaction mixtures were incubated at 37°C for 30 min; subsequently, 2 volumes of stop solution (90% formamide, 10 mM EDTA, 0.1% each xylene cyanol and bromophenol blue) were added to arrest the reaction. Reaction products were denatured by heating at 95°C and analyzed using 6% denaturing polyacrylamide gel electrophoresis and phosphorimaging. The effectiveness of the heparin trap was verified in control reactions in which the trap was preincubated with substrate before the addition of RT enzymes and dNTPs.

RNA-dependent DNA polymerase activity assay.

The same HIV-1 PBS RNA template and D25 primer labeled with ³²P at the 5′ end described above were used to assess primer extension efficiency in processive DNA synthesis by recombinant RT enzymes in gel-based time course experiments (30, 43). The final reaction mixtures contained 20 nM T/P, 400 nM RT enzyme, 50 mM Tris-HCl (pH 7.8), and 50 mM NaCl. Reactions were initiated by adding 6 mM MgCl₂ and dNTPs at 200 μM, and the reaction mixtures were mixed with 2 volumes of stop solution at various time points. Reaction products were separated by 6% denaturing polyacrylamide gel electrophoresis and analyzed by phosphorimaging.

Replication capacity in TZM-bl cells.

The relative replicative capacities of recombinant wild-type HIV-1_NL4-3 clones E138A, E138G, E138K, E138R, E138Q, E138V, E138A/M184I, E138G/M184I, E138K/M184I, E138R/M184I, E138Q/M184I, and E138V/M184I were evaluated in a noncompetitive infectivity assay using TZM-bl cells, as previously described (44, 45). Twenty thousand cells per well in 100 μl DMEM (Invitrogen, Burlington, ON, Canada) supplemented with 10% fetal bovine serum (Invitrogen, Burlington, ON, Canada), 1% penicillin-streptomycin, and 1% l-glutamine (Invitrogen, Burlington, ON, Canada) were added in triplicate into a 96-well culture plate. Viral stocks for both wild-type and mutant viruses were normalized by p24, and recombinant viruses were serially diluted 2-fold from viral stock suspensions. After 4 h, 50 μl of DMEM was removed from the wells and replaced by 50 μl of virus dilution; a control well did not contain virus. Virus and cells were cocultured for 48 h, after which 100 μl of Bright-Glo reagent was added and luciferase activity was measured in a luminometer as described above. The viral replication level was expressed as a percentage with reference to the number of RLU for wild-type virus.

RESULTS

Purification of recombinant HIV-1 RT enzymes.

The amino acid residue E138 in HIV-1 RT is part of the β7-β8 loop in the p51 subunit at the p66/p51 interface, which is a key structural element for RT dimerization and constitutes the floor of the NNRTI binding pocket (46–48). It was shown previously that the E138K mutation alone and E138K/M184I in tandem did not interfere with either heterodimer formation or enzyme purification (30, 49, 50). However, introducing other mutations at E138 alone or with M184I in tandem might potentially lead to such negative physical characteristics or otherwise affect enzymatic efficiency. Therefore, 12 recombinant heterodimeric (p66/p51) RT enzymes containing the E138A, E138G, E138K, E138R, E138Q, E138V, E138A/M184I, E138G/M184I, E138K/M184I, E138R/M184I, E138Q/M184I, or E138V/M184I substitution(s) in both RT subunits as well as WT RT were purified to >95% homogeneity, as assessed by Coomassie blue staining in SDS-polyacrylamide gels (Fig. 1). The RT p66 and p51 subunits were processed to similar molar ratios in each case. This verifies that the mutations introduced into the recombinant HIV-1 RT molecules in this study did not affect proteolytic cleavage, p66/p51 heterodimer formation, or RT enzyme purification.

Fig 1.

Purification of recombinant HIV-1 RTs. Purified heterodimer RTs were visualized by Coomassie brilliant blue staining of gels following 8% SDS-PAGE. Purification of heterodimeric RT enzymes was achieved by attachment of a His₆ tag at the C terminus of the p66 subunit through immobilized metal affinity chromatography (IMAC). Lanes: 1, WT; 2, E138A; 3, E138G; 4, E138K; 5, E138Q; 6, E138R; 7, E138V; 8, E138A/M184I; 9, E138G/184I; 10, E138K/M184I; 11, E138Q/M184I; 12, E138R/M184I; 13, E138V/M184I. The positions of purified recombinant RT heterodimers p66/p51 are indicated on the right.

The RT preparations were titrated by standard RNA-dependent DNA polymerase (RDDP) activity assay (37, 41), and all of the mutant RTs showed activities similar to the activity of WT RT, except for E138V, which exhibited decreased activity (∼50%) relative to that of WT RT (data not shown).

Inhibitory activities of RPV and ETR against WT and mutant recombinant HIV-1 RTs.

To evaluate the impact of mutations at position E138 in a background of WT HIV-1 RT and HIV-1 RT with the M184I mutation on susceptibility to RPV and ETR, we employed a filtration assay, as described above. The IC₅₀s of RPV and ETR for mutant RTs were determined and the fold changes (FCs) compared to the value for the WT are presented in Table 1. All of the mutations introduced at E138, except for E138V, the virus with which remained fully susceptible to both ETR and RPV, displayed FCs ranging from 1.6 to 3.1. Among these mutations, E138R (3.0-fold) and E138Q (2.9-fold) caused the greatest decreases in levels of RPV susceptibility, followed by E138G (2.2-fold) and E138A (2.1-fold), while E138K caused the lowest resistance level (1.8-fold). In regard to ETR, E138R was also associated with the highest level of resistance (3.1-fold), followed by E138Q (2.5-fold), E138K (2.3-fold), and E138G (2.2-fold), while virus with E138A exhibited the lowest level of resistance (1.6-fold). These data confirm that mutations E138A/G/K/Q/R decreased susceptibility to both ETR and RPV and that E138K is not the mutation that results in the highest level of loss of susceptibility to RPV and ETR, even though it is the mutation that was the most frequently observed in the ECHO and THRIVE clinical trials. In regard to the doubly mutated RT enzymes, viruses with E138Q/M184I and E138R/M184I displayed the highest levels of resistance to RPV (3.0- to 3.1-fold), followed by viruses with E138G/M184I and E138K/M184I (2.5- to 2.6-fold) and E138A/M184I (2.2-fold). In contrast, virus with E138V/M184I was hypersusceptible to RPV (0.4-fold). In regard to ETR resistance, virus with E138R/M184I exhibited the highest level of resistance (3.3-fold), followed by viruses with E138Q/M184I, E138G/M184I, and E138K/M184I (2.8- to 3.0-fold) and E138A/M184I (2.5-fold), while E138V/M184V did not confer resistance to ETR (0.7-fold). Thus, the E138A/G/K/Q/R mutations also decreased susceptibility to both ETR and RPV when M184I was also present, and E138K/M184I is not the combination that resulted in the highest level of loss of susceptibility to RPV and ETR. E138R is not currently considered to be an ETR RAM, but the present enzymatic and virological data (see below) indicate that it should be.

Table 1.

Susceptibilities to ETR and RPV of HIV-1 recombinant WT and mutated RT enzymes containing mutations at codon 138 alone or with M184I in tandem^a

Mutation	IC50 (nM)
	ETR	RPV
WT	70.3 ± 5.7	40.5 ± 6.7
E138A	112.5 ± 15.8 (1.6)	85.1 ± 12.0 (2.1)
E138G	154.7 ± 19.3 (2.2)	89.1 ± 12.4 (2.2)
E138K	161.7 ± 17.4 (2.3)	72.9 ± 12.6 (1.8)
E138R	218.0 ± 23.3 (3.1)	121.5 ± 16.0 (3.0)
E138Q	175.8 ± 19.3 (2.5)	117.5 ± 21.8 (2.9)
E138V	77.3 ± 8.3 (1.1)	38.5 ± 7.6 (1.0)
E138A/M184I	175.8 ± 20.9 (2.5)	89.1 ± 12.6 (2.2)
E138G/M184I	203.9 ± 23.1 (2.9)	105.3 ± 18.1 (2.6)
E138K/M184I	196.8 ± 16.2 (2.8)	101.3 ± 19.7 (2.5)
E138R/M184I	232.0 ± 26.2 (3.3)	121.5 ± 18.9 (3.0)
E138Q/M184I	211.0 ± 21.9 (3.0)	125.6 ± 18.3 (3.1)
E138V/M184I	48.9 ± 6.2 (0.7)	16.2 ± 3.8 (0.4)

^aIC₅₀s were determined in recombinant RT assays using heteropolymeric HIV-1 PBS RNA template. Data represent the means ± standard deviations (SD) of at least 3 independent experiments. Values in bold type differ significantly from the values for WT (P < 0.05, by analysis of variance using Turkey's multiple-comparison test). Data in parentheses are the fold change in resistance.

We also tested the susceptibility of recombinant WT and mutant RT enzymes to FTC and 3TC. None of the single mutations at position E138 conferred resistance against these drugs, while the double mutants containing M184I showed high-level resistance to FTC and 3TC similar to that of mutants containing M184I alone (data not shown).

Drug susceptibilities in cell culture phenotypic assays.

We also performed cell-based assays using recombinant WT virus and mutant HIV-1 to evaluate the impact of mutations at position E138 alone and in a background of M184I on the antiviral activities of RPV and ETR. The results in Table 2 show that all the E138 mutations, except for E138V, decreased susceptibility to RPV, with FCs ranging from 2.2 to 4.3. In contrast, virus with E138V was fully susceptible to ETR (1.1-fold) and RPV (0.9-fold). E138Q resulted in the highest level of loss of RPV susceptibility (4.3-fold), followed by E138G (3.7-fold) and E138R (3.5-fold), while virus with E138K displayed a lower level of decrease of susceptibility (2.7-fold) than did virus with E138A (2.2-fold). These findings are in agreement with the data obtained in the recombinant RT assays described above. Notably, the mutant RTs E138G/Q/R caused a higher level of loss of susceptibility to RPV than did the E138K RT. In regard to ETR resistance, virus with E138R exhibited the highest level of decrease in susceptibility (4.2-fold), followed by E138Q (3.7-fold), E138K (3.3-fold), E138A (3.2-fold), and E138G (3.1-fold), confirming that mutations E138A/G/K/Q/R decreased susceptibility to both ETR and RPV and that E138K is not the mutation associated with the highest level of RPV and RPV resistance. In combination with M184I, virus with E138R/M184I and E138Q/M184I displayed the highest levels of resistance to RPV (4.7- to 4.8-fold), followed by virus with E138G/M184I and E138K/M184I (4.3- to 4.4-fold) and E138A/M184I (3.0-fold), while virus with E138V/M184I remained susceptible to RPV (0.8-fold). In regard to ETR resistance, virus with E138R/M184I exhibited the highest level of resistance (4.9-fold), followed by virus with E138Q/M184I (4.4-fold), E138G/M184I (4.1-fold), E138K/M184I (4.1-fold), and E138A/M184I (3.1-fold), while E138V/M184I did not confer ETR resistance (0.8-fold). Thus, the E138A/G/K/Q/R mutations together with M184I decreased susceptibility to both ETR and RPV, and E138K/M184I was not the combination associated with the highest levels of RPV and ETR resistance. Our data further confirm that the E138K/M184I double mutation results in a higher level of loss of RPV and ETR susceptibility than E138K alone (31, 51).

Table 2.

Drug susceptibilities for recombinant HIV-1_NL4-3 WT and site-directed mutant viruses containing various E138 mutations alone or together with M184I assessed in TZM-bl cell cultures^a

Mutation	EC50 (nM)
	ETR	RPV
WT	1.5 ± 0.1	0.9 ± 0.1
E138A	4.8 ± 0.5 (3.2)	2.0 ± 0.2 (2.2)
E138G	4.7 ± 0.7 (3.1)	3.3 ± 0.5 (3.7)
E138K	5.0 ± 0.2 (3.3)	2.4 ± 0.3 (2.7)
E138R	6.3 ± 0.9 (4.2)	3.2 ± 0.7 (3.5)
E138Q	5.6 ± 0.8 (3.7)	3.9 ± 0.5 (4.3)
E138V	1.7 ± 0.6 (1.1)	0.8 ± 0.1 (0.9)
E138A/M184I	4.7 ± 0.9 (3.1)	2.7 ± 0.2 (3.0)
E138G/M184I	6.2 ± 1.1 (4.1)	4.0 ± 0.7 (4.4)
E138K/M184I	6.2 ± 0.9 (4.1)	3.9 ± 0.6 (4.3)
E138R/M184I	7.4 ± 0.8 (4.9)	4.2 ± 0.8 (4.7)
E138Q/M184I	6.6 ± 0.9 (4.4)	4.3 ± 0.7 (4.8)
E138V/M184I	1.2 ± 0.1 (0.8)	0.7 ± 0.1 (0.8)

^aData are the means ± standard deviations (SD) of 3 independent experiments. Values in bold type differ significantly from the values for WT (P < 0.05, by analysis of variance using Turkey's multiple-comparison test). Data in parentheses are the fold change in resistance.

We also tested the susceptibility of recombinant WT and mutant viruses to FTC and 3TC and found that none of the single mutations at E138 conferred resistance to FTC and 3TC but that the double mutants containing M184I showed high-level resistance similar to that shown by virus with M184I alone, similar to data obtained in the recombinant RT assays (data not shown).

Mutations E138K/Q/R restore the enzyme processivity of RT containing M184I.

HIV-1 RT drug resistance mutations that confer resistance to NRTIs can often affect RT enzymatic fitness by decreasing enzyme processivity (52–54), which is defined as the number of nucleotides incorporated in a single round of binding, elongation, and dissociation. Earlier studies showed that diminished HIV-1 RT processivity is the major determinant of impaired viral replication capacity associated with the M184I/V mutations, especially at low dNTP concentrations. In previous reports, we showed that E138K restores the enzyme processivity of RTs containing either M184I or M184V at low dNTP concentrations (30, 43) but that the Y181C mutation in RT can diminish the processivity of E138K-containing RT. Indeed, interactions between different drug resistance mutations, whether compensatory or antagonistic, seem to constitute the molecular basis for the favored or disfavored development of certain drug resistance patterns. Hence, we wished to determine how various mutations at E138 alone or in the presence of M184I would affect RT processivity and thus performed single-cycle processivity assays with recombinant RT enzymes at low dNTP concentrations, as described previously (30, 43). The results showed that the E138K/R/Q-containing mutant enzymes, alone or in combination with M184I, had higher processivity than M184I alone or E138A/G/V alone, indicating that E138R and E138Q, like E138K, can compensate for the diminished enzyme processivity associated with M184I (Fig. 2). In contrast, the E138G and E138G/M184I RTs displayed diminished processivity compared to WT. The E138V and E138V/M184I enzymes were barely able to extend the primer, indicating severe impairment of RT processivity under these conditions. The E138A and E138A/M184I mutant RTs exhibited processivity similar to that of WT. Thus, E138Q and E138R, like E138K, are able to function as compensatory mutations for M184I, as was also confirmed in the cell-based replication capacity assays whose results are presented below.

Fig 2.

Comparative analysis of enzyme processivity of WT RT and RT enzymes containing mutations at E138 alone and with M184I in tandem. The processivity of purified recombinant RT enzymes was analyzed using the D25 primer labeled with ³²P at the 5′ end annealed to a 471-nt HIV-1 PBS RNA template as the substrate; the resulting full-length DNA is 471 nt in length. Processivities were determined by the size distribution of DNA products in fixed-time experiments at low concentrations of dNTPs (0.5 μM) in the presence of a heparin trap. The sizes (in nucleotide bases) of some fragments of the ³²P-labeled 25-bp DNA ladder (Invitrogen, Burlington, ON, Canada) are indicated on the left. All reaction products were resolved by denaturing 6% polyacrylamide gel electrophoresis and visualized by phosphorimaging. The positions of the ³²P-labeled D25 primer (³²P-D25) and the 471-nt full-length (FL) extension DNA product are indicated on the right. A control reaction to verify the efficiency of the heparin trap by preincubation with substrate prior to addition of WT RT is also shown. A representative image from one of three independent experiments from which similar results were obtained is shown.

Effects of interactions of mutations at E138 in RT with M184I on polymerization efficiency of processive DNA synthesis.

In a previous report, we demonstrated mutual compensatory effects between E138K and M184I on the efficiency of initial polymerization in a dNTP concentration-dependent manner (30). At a high dNTP concentration, we showed that M184I can compensate for the defect of E138K in polymerization, while at low dNTP concentrations, E138K can restore the polymerization efficiency of the enzyme with M184I. These mutual compensatory effects constitute the molecular basis for competent replication of the doubly mutated virus at variable dNTP concentrations, i.e., in cells with both large and small dNTP pools. Now we wished to determine whether other mutations at E138 compensate for the defect in the efficiency of polymerization in processive DNA synthesis caused by M184I. We first performed RNA-dependent DNA polymerase reactions with all of the singly and doubly mutated enzymes and compared the processivity with that obtained with WT RT in time course experiments at high dNTP concentrations (200 μM) (Fig. 3A). RT molecules were used at an ∼20-fold excess over the amount of substrate, so that any RTs that dissociated from the primer terminus during synthesis would be rapidly replaced. In this case, the rate-limiting step would be nucleotide addition (55). The efficiency of polymerization was assessed by comparison of the length of the extension products (indicated by arrows) synthesized at 60 s (Fig. 3A). Each of the enzymes with E138K, E138Q, and E138R displayed similar efficiencies of DNA synthesis, i.e., impaired compared to that of WT RT, while E138A and E138V had no impact. The highest efficiency of polymerization was observed with E138G mutant RT. Among the doubly mutated RT enzymes, the polymerization efficiencies for each of the enzymes with E138Q, E138R, and E138K were restored by M184I, while each of the other RT enzymes tested displayed an efficiency of processive DNA synthesis that was similar to or even higher than that of WT RT.

Fig 3.

(A) Comparative analysis of enzyme efficiency of processive DNA synthesis of HIV-1 WT RT and RT enzymes containing mutations at E138 with M184I in tandem at high dNTP concentrations (200 μM). The D25 primer labeled with ³²P at the 5′ end (³²P-D25) was annealed to the 471-nt HIV-1 PBS RNA template, and primer extension assays were performed at an excess of recombinant RT enzymes at high dNTP concentrations (200 μM). Reactions were stopped at 30 s (30”) and 60 s (60”). The longest extension products generated at 60 s are identified by arrows and indicate differences in the efficiency of polymerization. A representative image from one of three independent experiments that yielded similar results is shown. (B) Impact of interactions of M184I with mutations at HIV-1 RT E138 on the efficiency of processive DNA synthesis at low dNTP concentrations (0.5 μM). Reactions were stopped at 30 s (30”), 60 s (60”), and 240 s (240”). The longest extension products generated at 240 s are identified by arrows and indicate differences in the efficiency of polymerization. A representative image from one of three independent experiments from which similar results were obtained is shown.

We also previously showed that at low dNTP concentrations E138K can compensate for the diminished polymerization efficiency of the enzyme with M184I (25), which is defective in regard to dNTP usage (52, 55), even though E138K on its own leads to a lower catalytic efficiency at high dNTP concentrations. Now we wished to test whether other mutations at E138 might also compensate for this deficit at low dNTP concentrations. The data in Fig. 3B show that E138K/M184I, E138Q/M184I, and E138R/M184I restored the efficiencies of DNA synthesis, while all the other double mutant enzymes exhibited a similar efficiency of DNA synthesis as the M184I enzyme. These compensatory effects of E138K/Q/R together with M184I were further validated in cell-based replication capacity assays, as shown below.

The E138K/Q/R mutations compensate for the impaired viral replication capacity of HIV-1 containing M184I.

Using a TZM-bl cell culture-based viral infectivity assay, we previously showed that the E138K mutation can restore the replication capacity of HIV-1 that harbors the M184I mutation (30). Now we wished to investigate the impact of interactions between M184I and mutations at E138 on viral replication. Hence, we infected TZM-bl cells with serially diluted stocks of various viruses normalized by the amount of p24 antigen. The infectivity of the WT and mutant viruses was determined by measuring luciferase activity at 48 h postinfection. As shown in Fig. 4, the relative replication ability of viruses containing M184I alone was decreased by 3-fold compared to that of the WT virus, while the replication capacity of the E138K/M184I double mutant virus was equal to that of the WT virus, in agreement with our previously published data (30). This compensatory effect on viral replication capacity was also observed with the E138Q/M184I and E138R/M184I double mutant viruses, indicating that E138Q and E138R can also function together with M184I to restore viral fitness. In contrast, each of the other mutant viruses, i.e., those with the E138G/M184I, E138A/M184I, and E138V/M184I mutations, exhibited lower viral replication capacity than WT, indicating that E138G, E138A, and E138V cannot compensate for the M184I mutation in regard to viral replication. These findings are in agreement with the enzymatic data described above that show that E138Q and E138R, like E138K, can function as compensatory mutations for M184I in regard to RT enzymatic fitness.

Fig 4.

The HIV-1 RT mutations E138K/Q/R compensate for the impaired viral replication capacity of M184I. Viral stocks of the wild-type HIV-1_NL4-3 clone and clones containing the indicated mutations were normalized for p24 and used to infect TZM-bl cells. Luciferase activity was measured at 48 h postinfection as an indication of viral replication. The relative infectivity of WT virus compared to that of mutant viruses is shown on the y axis, while the x axis denotes the input of p24. The figure is representative of two independent experiments. Error bars represent standard deviations of the mean.

DISCUSSION

The newer NNRTIs RPV and ETR were developed to overcome the low genetic barrier for resistance characterized by the earlier NNRTIs, since only a single mutation is sufficient to confer diminished susceptibility to both NVP and EFV. Although it had been reported that resistance to both ETR and RPV generally required an accumulation of several RT mutations (11, 56), recent data from cell culture studies and clinical trials indicate that mutations at position E138 can emerge under RPV pressure to cause virological failure. Current resistance-associated mutation (RAM) lists for ETR and RPV assign E138A/G/K/Q as mutations associated with resistance against both ETR and RPV, while E138R is designated a RAM only for RPV (24). In general, however, limited data have been available in regard to the role of these mutations, either alone or in combination with others, in conferring drug resistance and on RT enzyme function.

In view of the fact that patients in the ECHO and THRIVE trials commonly developed tandem E138K and M184I mutations when failing RPV-based therapy, the current virological study examined the contribution of various mutations at position E138, alone or in combination with M184I, on resistance to RPV and ETR. We show here that each of E138A, E138G, E138K, E138Q, and E138R, but not E138V, decreases susceptibility to both RPV and ETR. Virus with E138R and E138Q displayed the highest levels of resistance to both RPV and ETR. This cross-resistance effectively rules out the use of ETR in patients who have failed an RPV-containing regimen. It should be noted that only modest levels of decrease of drug susceptibility in vitro were observed for most of these mutations; however, the drug levels reached during treatment in vivo might often be insufficient to suppress replication of HIV-1 harboring relevant resistance mutations.

Our enzymatic and virological data are in agreement with previous work that showed that recombinant HIV-1 clones containing E138R and E138Q, rather than E138K, showed the highest levels of drops in susceptibility to ETR (23). Others have shown that clones with the E138Q/G/A mutations displayed higher levels of RPV resistance than those with the E138K mutation (26), even though E138K was the most frequent mutation to emerge, together with M184I, in the RPV arm of the ECHO and THRIVE clinical trials (15, 28, 29). In addition, E138K was the first mutation to emerge in cell culture selection studies performed with each of three DAPY compounds, i.e., ETR (44), RPV (34), and dapivirine (DPV) (42), and was first identified in selection experiments that employed the structurally distinct NNRTI 2′,5′-bis-O-(tert-butyldimethylsilyl)-3′-spiro-5″-(4″-amino-1″,2″-oxathiole-2″,2″-dioxide) (TSAO) compounds (57–62). These studies also reported that E138K mutant virus and enzyme displayed higher-level resistance to the TSAO compounds than other mutants. In contrast, the present study shows that neither E138K nor E138K/M184I caused the highest drops in levels of susceptibility to RPV and ETR. Thus, the favored selection of E138K in cell culture by DAPY compounds or the emergence of E138K/M184I in patients failing RPV-containing regimens cannot be explained solely by the levels of drug resistance associated with the E138K mutation. Rather, the compensatory effects associated with E138K and the fact that its emergence requires only a single G-to-A hypermutation, as discussed below, may be key determinants.

E138K has previously been shown to compensate for low RT enzyme processivity and polymerization efficiency associated with M184I/V mutations at low dNTP concentrations, while M184I/V can compensate for defects in polymerization efficiency associated with E138K at high dNTP concentrations (30). These mutually compensatory effects make E138K/M184I RT catalytically more efficient than singly mutated RT, and this conclusion is also supported by analyses based on transient kinetics (32). This mutual compensation also resulted in a restoration of viral replication capacity. Others have also shown that E138K RT is more catalytically efficient than other E138 mutant RTs (51). We now extend previous observations on this topic to show that each of E138K, E138Q, and E138R can mutually compensate for M184I in regard to RT enzyme processivity and polymerization efficiencies. In addition, we have shown here that viruses containing the mutations E138K/M184I, E138Q/M184I, and E138R/M184I, but not the mutation E138A/M184I, E138G/M184I, or E138V/M184I, exhibited wild-type replication capacity. Thus, considerations of viral replication capacity alone cannot explain the favored selection of E138K/M184I over E138Q/M184I and E138R/M184I, especially since the last two combinations caused the highest level of losses in susceptibility to RPV. Differences in FTC/3TC resistance of all the M184I-containing enzymes and viruses tested were not demonstrated in our assays. Although a shortcoming of this work is that we did not assess viral replication in competitive fitness assays (31, 44, 45, 51), our viral replication data are in agreement with the results of our enzymatic analyses that also show that E138K/M184I does not have a significant advantage over E138Q/M184I and E138R/M184I.

Cellular dNTP pools play an important role in HIV-1 replication and in the development of drug resistance mutations (63–66). The genome of HIV-1 is significantly rich in adenine nucleotides due to the G → A hypermutation in the HIV-1 RNA genome that is driven by both the error-prone nature of HIV-1 RT and by dCTP/dTTP pool imbalances during reverse transcription (67–69). Under selection pressure with 3TC or FTC, M184I (ATG → ATA) usually emerges first because of the G → A hypermutation and is only eventually replaced by M184V (ATG → GTG) due to the relative fitness advantage of the latter mutation (65, 70–72). The G → A transition that occurs in E138K (GAG → AAG) also derives from the G → A hypermutation. Evidence for the roles of ratio bias of dCTP/dTTP and the low fidelity of the RT enzyme in determining HIV-1 resistance to TSAO compounds was obtained by increasing the ratio of cellular endogenous dCTP/dTTP pools (65) in a manner that altered the characteristic E138K (GAG → AAG) mutation to E138G (GAG → GGG). Considering that E138K and M184I are mutually compensatory in regard to viral fitness (30, 31), the favorable G → A hypermutation associated with both M184I and E138K may contribute to and help explain the predominance of the E138K/M184I but not other mutations at E138, even though both E138Q and E138R can similarly compensate for M184I. Of course, host factors may also play a role in the predominant appearance of E138K/M184I, and it is known, as an example, that human leukocyte antigen (HLA)-B*51-restricted cytotoxic T-lymphocyte (CTL) pressure can impact the occurrence of NNRTI resistance, including the E138K mutation (73). In addition, the host restriction factor APOBEC3 plays a key role in regard to G-to-A hypermutability, by virtue of its cytidine deaminase function, and therefore also contributes to the spontaneous emergence of both E138K and M184I (74).

Finally, the preferential emergence of E138K over E138R (GAG → AGG) and E138Q (GAG → CAG) is also probably due to the fact that the former requires only a single favorable G → A nucleotide change, while each of the last two requires two distinct mutations and a single G → C nucleotide change, respectively. This explains why E138K preferentially emerged over E138Q/R in the ECHO and THRIVE trials, in spite of the fact that the last two mutations, either alone or in the presence of M184I, resulted in higher levels of loss of susceptibility to both RPV and ETR than did the former.