ABSTRACT
During HIV-1 reverse transcription, there are increasing opportunities for nucleos(t)ide (NRTI) or nonnucleoside (NNRTI) reverse transcriptase (RT) inhibitors to stop elongation of the nascent viral DNA (vDNA). In addition, RT inhibitors appear to influence the kinetics of vDNA synthesis differently. While cell-free kinetic inhibition constants have provided detailed mechanistic insight, these assays are dependent on experimental conditions that may not mimic the cellular milieu. Here we describe a novel cell-based strategy to provide a measure of the intrinsic inhibition efficiencies of clinically relevant RT inhibitors on a per-stop-site basis. To better compare inhibition efficiencies among HIV-1 RT inhibitors that can stop reverse transcription at any number of different stop sites, their basic probability, p, of getting stopped at any potential stop site was determined. A relationship between qPCR-derived 50% effective inhibitory concentrations (EC50s) and this basic probability enabled determination of p by successive approximation. On a per-stop-site basis, tenofovir (TFV) exhibited 1.4-fold-greater inhibition efficiency than emtricitabine (FTC), and as a class, both NRTIs exhibited an 8- to 11-fold greater efficiency than efavirenz (EFV). However, as more potential stops sites were considered, the probability of reverse transcription failing to reach the end of the template approached equivalence between both classes of RT inhibitors. Overall, this novel strategy provides a quantitative measure of the intrinsic inhibition efficiencies of RT inhibitors in the natural cellular milieu and thus may further understanding of drug efficacy. This approach also has applicability for understanding the impact of viral polymerase-based inhibitors (alone or in combination) in other virus systems.
INTRODUCTION
The HIV-encoded reverse transcriptase (RT) enzyme catalyzes the initiation, elongation, and termination of viral DNA (vDNA) synthesis through an ordered multistep process known as reverse transcription (Fig. 1A) (1, 2). This process, by which single-stranded HIV-1 RNA is converted to double-stranded HIV-1 DNA, follows a series of successive events whereby the product of each nucleoside incorporation serves as a substrate for the following reaction until the end of the genomic template is reached. First, minus-strand vDNA synthesis is initiated inefficiently from a primer-binding site (PBS) by a cell-derived
Cell-based approach for monitoring reverse transcription vDNA products of various length by quantitative real-time PCR. (A) Schematic representation of HIV-1 reverse transcription stages. Minus-strand DNA synthesis (indicated in blue) is initiated from a
Due to its essential role in HIV-1 replication and the lack of a cellular homolog, RT has been a major target for antiviral agents. Both nucleos(t)ide (NRTIs) and nonnucleoside (NNRTIs) reverse transcriptase inhibitors (RTIs) have proven clinical benefit of sustained efficacy and long-term safety and further represent a current standard of care for treatment for HIV-infected individuals (6). NRTIs structurally resemble natural 2′-deoxynucleosides in which the 3′-hydroxyl moiety on the deoxyribose sugar or pseudosugar has been removed or modified. Upon conversion into their triphosphate or diphosphate forms by cellular kinases, NRTIs compete with the natural deoxynucleoside triphosphates (dNTPs) for incorporation into the nascent elongating vDNA, resulting in chain termination. Eight N(t)RTIs have been approved for clinical use, in the following order: zidovudine (AZT), didanosine (ddI), dideoxycytidine (ddC), stavudine (d4T), lamivudine (3TC), abacavir sulfate (ABC), tenofovir disoproxil fumarate (TDF), and emtricitabine (FTC) (7). In comparison, NNRTIs are hydrophobic compounds with diverse chemical structures that do not require intracellular metabolism for activation. NNRTIs primarily act as noncompetitive RTIs which bind to and induce conformational changes in RT that result in diminished catalytic efficiency without affecting dNTP binding (8, 9). In addition, NNRTIs may also inhibit other key steps in the viral replication cycle through multiple diverse mechanisms (10). Five NNRTIs have been approved for clinical use, in the following order: nevirapine (NVP), delavirdine (DLV), efavirenz (EFV), etravirine (ETR), and rilpivirine (RPV) (7).
Under conventional steady-state conditions, RT enzyme kinetic assays have permitted derivation of inhibition efficiency parameters, Ki/Km and 50% inhibitory concentration (IC50). In the case of NRTIs, Ki/Km measures the ability of RT to selectively bind to and incorporate a dideoxynucleoside triphosphate (ddNTP) analog relative to the corresponding natural dNTP substrate (11, 12). For further resolution, pre-steady-state kinetic assays have permitted derivation of a catalytic efficiency parameter for single turnover and nucleotide incorporation, kpol/Kd, where kpol is the turnover rate constant for a two-step process resulting in the transfer of dNMP to the 3′ end of the primer and Kd is the dissociation constant for dNTP from the RT-template/primer (T/P) complex. Comparison of kpol/Kd between the natural dNTP and ddNTP analog defines discrimination efficiency (selectivity) of RT (13, 14). While informative, these kinetic parameters of inhibition determined under cell-free conditions can vary widely depending on the type (RNA or DNA), nature (homopolymeric or heteropolymeric), sequence, and length of T/P complex used (12, 15–19). Moreover, such conditions that allow polymerization from multiple start sites, monitor complete and incomplete DNA products, and/or use variable ratios of T/P to RT concentrations all do not emulate conditions found in the natural cellular milieu. For these reasons, it can be difficult to relate biochemical measures of inhibition efficiency with those expected in vivo or to directly compare NRTIs and NNRTIs using a shared inhibition parameter.
The advent and application of quantitative real-time PCR (qPCR) technology have furthered our understanding of HIV-1 reverse transcription, and thus the impact of RTIs, in cell-based assays (20, 21). These early studies showed that high concentrations of RTIs (AZT, ddI, d4T, and NVP) minimally inhibit the generation of short vDNA products yet block intermediate to full-length vDNA products as a function of vDNA chain elongation (21–24). More recently, other studies have suggested that several RTIs may influence the kinetics of minus-strand vDNA synthesis differently (25). Despite these technological advances, we currently do not have an objective measure (parameter) of inhibition efficiency inherent to the cellular milieu that would allow for intra- and interclass comparison of RTIs on common ground. Goody et al. proposed that the elucidation of such a theoretical parameter in vivo, based on stochastic cumulative incorporation of substrate in the presence of an RTI, would overcome the limitations of previous measures of inhibition efficiency (26). In particular, these authors argued that the relative significance of chain termination versus dead-end complex formation toward inhibition potency would vary with the length of the template and the ratio of RT to primer, suggesting mechanistic differences under cell-free and cell-based conditions. Since inhibition potency alone cannot predict clinical drug efficacy, a cell-based system that could assess true inhibition potential would provide an additional valuable parameter for consideration.
Here we report the establishment of a cell-based assay and strategy to determine the basic probability, p, of chain termination (stopping) by a given RTI, which is a measure of its intrinsic inhibition efficiency per stop site. This assay uses established qPCR methodology to determine inhibitory potencies (50% effective inhibitory concentrations [EC50s]) across reverse transcription vDNA products of different lengths in acutely infected cells. A functional relationship was established between the experimental EC50s and the basic probability which enabled determination of the latter by successive approximation. This measure of intrinsic inhibition efficiency is uniquely independent of vDNA product length, the number of potential stop sites, the concentration of RTI, and the mode of inhibition. Using this assay, we tested and compared two NRTIs (TFV and FTC) and one NNRTI (EFV) commonly administered in current combination treatment regimens. We found, based on their values of p, that the tested RTIs exhibited differential inhibition efficiencies as follows: TFV ≥ FTC > EFV. Finally, we propose several applications of this strategy, including the clinical and nonclinical characterization of viral polymerase inhibitors for HIV-1 and other viruses.
MATERIALS AND METHODS
Compounds.Tenofovir (TFV), emtricitabine (FTC), and efavirenz (EFV) were synthesized at Gilead Sciences, Inc. (Foster City, CA).
Cells, vectors, and transfection.MT-2 cells were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program (Germantown, MD) and maintained in RPMI 1640 medium supplemented with 100 μg/ml of penicillin G, 100 μg/ml of streptomycin (Life Technologies, Grand Island, NY), and 10% (vol/vol) heat-inactivated fetal bovine serum (FBS). HEK (human embryonic kidney) 293T cells were obtained from American Type Culture Collection (ATCC; Manassas, VA) and maintained in Dulbecco modified Eagle medium (DMEM) supplemented with antibiotics and 10% (vol/vol) FBS.
Plasmid pKSHX.Luc is an NL4-3-based vector derived from pKS13.Luc (27); it carries the HXB2 sequence of reverse transcriptase and integrase, inactivating mutations in env and vpr, and a codon-optimized firefly luciferase gene introduced in place of nef (27). Plasmid pHCMV-VSVG encodes the vesicular stomatitis virus G protein (VSV-G) and was purchased from Cell Biolabs Inc. (San Diego, CA). To generate an endogenous control standard template for qPCR, the gene encoding β-globin (1,524 bp) was PCR amplified from MT-2 cells and cloned into pUC19 (New England BioLabs, Ipswich, MA). Pseudotyped HIV-1 capable of a single cycle of replication was produced by transfection of HEK 293T cells with pKSHX.Luc and pHCMV-VSVG using TransIT-293 according to the manufacturer's protocol (Mirus Bio LLC, Madison, WI). At 6 h posttransfection, vector DNA was thoroughly washed from the culture plates with phosphate-buffered saline supplemented with 1% (vol/vol) FBS. Virus-containing culture supernatants were harvested at 48 h posttransfection, clarified by filtration through a 0.22-μm-pore membrane (Millipore, Billerica, MA), and stored at −80°C until use. Virus titers inMT-2 cells were determined by luciferase-based expression to evaluate the median tissue culture infective dose (TCID50/ml) 2 days postinfection.
Single-replication cycle antiviral activity assay.The susceptibility of HIV-1 to RTIs was determined using an integration-based luciferase reporter assay as previously described, with some modifications (28). Appropriate 2-fold serial dilutions of tested RTIs were prepared in triplicate in 96-well white solid-bottom plates, followed by addition of MT-2 cells at a final density of 3 × 105 cells per well. Infections were then carried out by addition of 10-fold-diluted virus-containing supernatant at 50 μl per well for a final assay volume of 200 μl. After 2 days of incubation at 37°C, MT-2 cells were resuspended using a Bravo automated liquid handling platform (Agilent Technologies, Santa Clara, CA), and prewarmed Bright-Glo luciferase assay system substrate (Promega, Madison, WI) was added at a 1:2 ratio. Luminescence was measured on an Envision multilabel plate reader (PerkinElmer, Waltham, MA), and luciferase activity was normalized to infections performed in the absence of RTI. Antiviral inhibition potency (EC50) of each RTI was determined from triplicate experiments by best-fit nonlinear regression analyses of luciferase activity using Prism software (GraphPad Software, La Jolla, CA).
Infection and determination of vDNA copy number by quantitative real-time PCR.To analyze the effects of RTIs on HIV-1 reverse transcription, MT-2 cells were maintained in the presence of each RTI before, during, and after infection. MT-2 cells were plated at a density of 1 × 106 cells per well in 6-well plates and incubated for 24 h in medium containing RTIs at various concentrations. Experiments included treatment conditions of 1-fold or 10-fold EC50 concentrations of each RTI, predetermined in a single-cycle antiviral activity assay. Additional experiments used 2-fold-decreasing concentrations of each RTI to permit construction of 18-point dose response curves. On the day of infection, virus stocks were thawed and treated with 100 U of DNase I (Life Technologies, Grand Island, NY) plus 5 mM MgCl2 for 1 h at 37°C to remove any vector DNA carried over from the transfection. MT-2 cells were infected with high titer of virus (multiplicity of infection [MOI] = 10) by spinoculation at 1,900 × g for 2 h at room temperature, as previously described (29). Each experiment contained appropriate controls, including the absence of RTI and infection with a previously heat-inactivated virus (68°C for 30 min) so that accurate vDNA quantities could be determined from infections. After infection, cells were washed twice with phosphate-buffered saline supplemented with 1% (vol/vol) FBS to remove residual virus before being incubated in fresh medium containing the same concentration of RTI as before. At 6 h postinfection (or every 1 to 2 h in the case of the initial time course study), cells were harvested and total cellular DNA was extracted using the EZ1 DNA tissue kit on an EZ1 Advanced XL Biorobot (Qiagen, Valencia, CA).
Conventional primers and probes used for qPCR detection of specific minus-strand vDNA products corresponding to early (RU5 and U3) and late (GAG and PBS) stages of reverse transcription have been described (20, 30). Additional primers and probes used for detection of vDNA products of intermediate length between 1,000 and 8,000 nucleotides (nt) and endogenous cellular β-globin have also been described (22, 31). All primers and dually labeled probes were synthesized by Integrated DNA Technologies, Inc. (San Diego, CA). Dually labeled probes were end labeled with 5′ 6-carboxyfluorescein (FAM) or 5′ hexachlorofluorescein (HEX) and 3′ 6-carboxytetramethylrhodamine (TAMRA). For each experiment, a standard curve of each amplicon was measured from 102 to 107 linearized copies of each vector plus a no-template control, all diluted into an equivalent amount of 500 ng of sheared salmon sperm DNA (Life Technologies, Grand Island, NY). Each amplicon from each extracted DNA treatment sample was measured in quadruplicate. Quantitative real-time PCR mixtures contained 1× Roche LightCycler 480 probe master mix kit (Roche Diagnostics Corporation), 300 nM forward primers, 300 nM reverse primers, 100 nM probe primers, and 500 ng of DNA in a 25-μl volume. After initial incubations at 50°C for 2 min and 95°C for 10 min, 40 cycles of amplification were carried out at 15 s at 95°C, followed by 1 min at 60°C. Multiplexed PCRs were analyzed after color compensation using the Roche LightCycler 480 sequence detection system (Roche Diagnostics Corporation). Copy numbers were normalized by endogenous β-globin levels and represented as a percent total copy number of the no-drug infected control (set to 100%) for each amplicon measured. For time course experiments, copy numbers were represented as percentage of the no-drug infection control after 10 h postinfection. The 50% effective inhibitory concentration (EC50) of each vDNA product was determined from triplicate experiments by best-fit nonlinear regression analyses of normalized percent vDNA copy numbers using Prism software (GraphPad Software, La Jolla, CA).
Curve fitting to determine the basic probability of chain termination.Curve fitting and plots were performed using SigmaPlot scientific data analysis and graphing software (Systat Software Inc., San Jose, CA). A successive approximation scheme (see Fig. S1 in the supplemental material) was devised to determine the basic probability, p, of chain termination (i.e., a measure of inhibition efficiency per stop site) based on a relationship between the probability of chain termination, P(non_end), and experimental EC50s determined by qPCR (see the supplemental material for equations and derivation of the functional form of p). Each iteration was composed of 3 steps. Step 1 was approximation of 1 − p using F(EC50n) for the largest n value, followed by calculation of the F(EC50n) curve for the smaller n values using equation VIII. Step 2 was plotting of experimental EC50n values against calculated F(EC50n) values, followed by curve fitting using equation IIIc to solve for parameters A and m. In step 3, parameters A and m determined in step 2 were used with the mean experimental EC50n values for the two largest n values to approximate a new value of F(EC50n) for the largest n. This new value of F(EC50n) for the largest n was then used in step 1 of the second iteration to approximate a new value for 1 − p and calculate the F(EC50n) curve for smaller n values. Steps 1 through 3 of this successive approximation are repeated until convergence of the values for p.
RESULTS
Inhibition of reverse transcription in acutely infected cells as a function of time.To generally compare the impacts of TFV, FTC, and EFV on HIV-1 reverse transcription and establish the best time postinfection for more detailed investigation, we first performed time course experiments using 10× EC50 drug concentrations (which may inhibit viral replication by >95%). The accumulation of vDNA products, detected using either early or late primer-probe sets (Fig. 1), were expressed as a percentage of the total copy number of products that accumulated at the 10-h time point. In untreated cells, the production of short (RU5) vDNA products increased rapidly up to 6 h postinfection, whereas long (PBS) vDNA products increased gradually over the entire time course (Fig. 2A). After 10 h, short (RU5) vDNA products had accumulated to ∼58% of control levels in the presence of TFV and ∼40% of the control levels in the presence of FTC or EFV. In comparison, accumulation of long (PBS) vDNA products was severely attenuated in the presence of any RTI (Fig. 2B). These results indicate that TFV, FTC, and EFV more prominently inhibit the production of late vDNA products as a function of time, consistent with previous observations of early approved RTIs (22, 25). In addition, 6 h postinfection appeared to be an appropriate time point to capture and compare all vDNA products during rapid accumulation.
Kinetics of reverse transcription during a single cycle of infection. MT-2 cells were pretreated (−24 h) with 10-fold-adjusted (10×) EC50 levels of TFV (▽), FTC (□), or EFV (○) prior to infection, along with a no-drug control (⧫). At indicated time points postinfection, cells were lysed, and vDNA product copy numbers quantified in replicate by qPCR were expressed as a percentage of the response in the no-drug control-treated cells measured at 10 h postinfection and set to 100%. The impact of RT inhibitors was assessed for early reverse transcription of short vDNA products by monitoring for minus-strand strong stop (RU5) (A) and late reverse transcription of long vDNA products by monitoring for second-strand transfer (PBS) (B). Data represent the means ± standard errors of the means of two independent experiments.
Inhibition of reverse transcription in acutely infected cells as a function of vDNA product length and drug concentration.To compare the inhibitory potentials of TFV, FTC, and EFV as a function of vDNA product length and drug concentration, the following conditions were subsequently utilized. Acutely infected MT-2 cells were treated with 1× or 10× the EC50 of each RTI and total DNA was extracted at 6 h postinfection. Using 9 different primer-probe sets to detect vDNA products that spanned early to late stages of reverse transcription (Fig. 1B), we found that vDNA product accumulation gradually decreased with increasing product length in the presence of TFV, FTC, or EFV (Fig. 3). At drug concentrations 1× the EC50, early stage vDNA products less than 1,000 nt accumulated to 90% of control levels, indicating 10% inhibition, while the same concentration of RTIs diminished the accumulation of late-stage vDNA products greater than 8,000 nt by ∼50%. At drug concentrations 10× the EC50, vDNA products greater than ∼2,000 nt in length were severely diminished, to 9 to 12% of control levels, indicating 88 to 91% inhibition. In contrast to the case with TFV or FTC, the accumulations of both short and long vDNA products were similarly attenuated in the presence of EFV.
Relative accumulation of vDNA products of different lengths. MT-2 cells treated with 1-fold- or 10-fold-adjusted EC50s of TFV, FTC, or EFV were harvested at 6 h postinfection. Each vDNA product was quantified in replicate by qPCR and expressed as a percentage of the response of the same product quantified in the absence of drug, set to 100%. Data represent the means ± standard errors of the means of two independent experiments, fit by nonlinear regression using a single-exponential decay function. Best-fit parameters and values are shown in Table 1.
To more aptly compare the inhibitory effects of TFV, FTC, and EFV on the amount and length of vDNA product generated, these results were fit to a single-exponential decay function using the following equation, where Yo is the initial relative vDNA product, H is the inhibition plateau, X is the product length in nucleotides, and K is a decay constant per nucleotide (nt−1): Y = (Yo − H)e−KX = H. A similar approach was previously used to assess the drug concentration-based impacts of other RTIs (AZT, 3TC, and NVP) on vDNA product accumulation (18, 22). Best-fit values for TFV, FTC, and EFV are shown in Table 1. Although limited data across all possible vDNA product lengths were available, most curves exhibited a reasonable goodness of fit, with r2 values of >0.90. As a surrogate measure for inhibition efficiency, the exponential decay constants (K) for TFV, FTC, and EFV were compared pairwise. At drug concentrations 1× the EC50, the decay constants of each RTI were at most 2-fold different from each other. However, at 10× the EC50, the decay constant of EFV was ∼10-fold higher than that of either TFV or FTC. Together, these differences indicate a drug concentration dependency in utilizing this methodology for comparing relative inhibition efficiencies among RTIs.
Predicted decay constants for relative diminution of vDNA products of different lengtha
Inhibition of reverse transcription in acutely infected cells as a function of inhibition potency (EC50).To complement and extend the above-described findings, we developed a cell-based strategy to permit determination of an “inhibition efficiency” parameter for TFV, FTC, and EFV as a quantitative measure of inhibition across all stages of reverse transcription, independent of vDNA product length, RTI concentration, or the number of potential stop sites. The first part of our strategy required a determination of (i) experimental inhibition potencies (EC50s) across multiple vDNA products of different anticipated lengths and (ii) the potential number of chain termination stop sites within each vDNA product.
Acutely infected MT-2 cells were treated with various concentrations of each RTI, total DNA was extracted at 6 h postinfection, and the same vDNA products (ranging from 200 to 9,000 nt) were quantified by qPCR, as before. The mean effective drug concentrations of TFV, FTC, and EFV needed to inhibit 50% accumulation (inhibition potency; EC50) of each of the 9 vDNA products were then determined by nonlinear regression analysis (Table 2). As expected, inhibition potencies of the tested RTIs followed the order of EFV > FTC > TFV. However, all 3 RTIs exhibited low inhibition potencies (high EC50s) on short vDNA products, i.e., less than 1,000 nt, during early stages of reverse transcription relative to their respective EC50s determined after 48 h using a single-cycle reporter-based antiviral assay (Table 2). Coincidentally, these values were analogous to IC50s reported for biochemical assays, in which short templates are typically used (32–34). Between early and late stages of reverse transcription, inhibition potency between short (RU5) and long (PBS) vDNA products correspondingly increased 13-fold for TFV and FTC and 4-fold for EFV. In contrast to early-stage EC50s, late-stage EC50s were highly comparable with those determined after vDNA integration at 48 h using a single-cycle reporter-based antiviral assay (Table 2). A further comparison of the EC50s between RU5 and U3 and between GAG and PBS suggested that TFV, FTC, and EFV do not have any direct impact on template switching efficiencies, consistent with previous findings using other RTIs (AZT, ddI, and 3TC) (23).
Antiviral susceptibilities and relative inhibition potencies against accumulation of vDNA products of different lengths
Finally, since RTIs can potentially stop at any number of different stop sites along a template, an intrinsic measure of inhibition efficiency must be independent of these differences in addition to the overall length of the vDNA products themselves. For TFV and FTC, which are NRTI analogs of adenosine and cytidine, respectively, the corresponding number of sites for substitution in each vDNA product was resolved by sequencing analysis (Table 3). For EFV, an NNRTI capable of stopping reverse transcription at any site, the total number of nucleotides in each vDNA product was considered.
Number of potential stop sites for RTIs along vDNA products of different lengths
Intrinsic inhibition efficiencies of TFV, FTC, and EFV.Because different RTIs can potentially stop vDNA polymerization at any number of different sites, the inhibition potency (EC50) alone is not the best comparator on a purely cellular mechanistic level. For this reason, we set out to devise a method for the comparison of RTIs using their basic probability, p, of chain termination, which is a measure of their intrinsic inhibition efficiencies. We define p as the probability for RT to get stopped by the inhibitor at any stop site just before reaching it. This basic probability is per stop site and therefore is independent of vDNA product length, the number of potential stop sites, the drug concentration, and the mode of inhibition (competitive versus noncompetitive).
To illustrate the stochastic process of RT inhibition, a simple scenario of vDNA polymerization along a template with 3 potential stop sites (susceptible to chain termination), where p is the basic probability of the RT enzyme being stopped by an RTI at any site just before reaching it, was considered (Fig. 4).
Stochastic process of RT inhibition at potential stop sites. A set of events {s1, s2, s3} for RT to be stopped at sites 1, 2, and 3 and another set of events {ns1, ns2, ns3} for RT not to be stopped at sites 1, 2, and 3 are defined.
During chain elongation in the presence of an RTI, the probabilities of RT getting stopped at sites 1, 2, and 3 are P(s1), P(s2), and P(s3), respectively, with the following expressions: P(s1) = p, P(s2) = p(1 − p), and P(s3) = p(1 − p)2, where s1 is site 1, s2 is site 2, and s3 is site 3. The probability of RT reaching the end of the template, P(end), has the following expression: P(end) = (1 − p)3. The probability of RT not reaching the end of the template, P(non_end), is the sum of the probabilities of getting stopped at sites 1, 2, and 3 and has the following expression:
To better compare TFV, FTC, and EFV, which can stop reverse transcription at any number of different sites, a successive mathematical approximation strategy (see Fig. S1 in the supplemental material) was devised to determine their basic probability, p, of chain termination (i.e., a measure of inhibition efficiency per stop site). This strategy was based on a relationship between the probability of chain termination, P(non_end), and experimental EC50s determined by qPCR for each vDNA product with n potential stop sites (see the supplemental material for equations and derivation of the functional form for p). After 12 iterations of successive approximation, it was determined that (i) p values had converged to a single value and (ii) the two functional expressions of P(non_end) had become superimposable (see Fig. S2 in the supplemental material). On a per-stop-site basis, TFV exhibited the greatest inhibition efficiency, p, followed by FTC and then EFV (Table 4). While inhibition efficiencies varied only 1.4-fold between TFV and FTC, as a class, these NRTIs exhibited an 8- to 11-fold-greater inhibition efficiency per stop site than EFV, an NNRTI. As expected, the probability, P(non_end), of not polymerizing to the end of a given vDNA product increased with the number of potential stop sites (Table 4). P(non_end) for inhibition of the shortest vDNA product (RU5) varied 1.7-fold between TFV and FTC and 1.9- to 3.2-fold between these NRTIs and EFV. Similarly, P(non_end) for inhibition of the longest vDNA product (PBS) varied 1.03-fold between TFV and FTC and ∼1.2-fold between these NRTIs and EFV. These results support the concept that inhibition of reverse transcription is a cumulative process of greater opportunities for inhibition with increased vDNA product length. As shown in equation II in the supplemental material, the relationship between P(non_end) and the number, n, of potential stop sites is not linear. As the number of potential stop sites between RU5 and PBS increased 48-, 66-, 64-fold for TFV, FTC, and EFV, the probability of not reaching the end, P(non_end), increased 13-, 22-, 34-fold, respectively. Taken together, these results show that NRTIs (TFV and FTC) exhibit greater probabilities of inhibition per stop site than does an NNRTI (EFV), but these RTIs exhibit more similar overall probabilities of vDNA product noncompletion when considering the entire viral RNA template.
Probabilities of reverse transcription inhibition per stop site and per vDNA product lengtha
DISCUSSION
Conventional measures of cell-free RTI activity, including Ki/Km and kpol/Kd, are limited in their capacity to fully describe antiviral inhibition efficiency. Moreover, since clinical concentrations of antiretroviral drugs are typically used well above established EC50s, antiviral potency alone cannot predict drug efficacy. In this study, we established a cell-based strategy to estimate and compare the intrinsic inhibition efficiency of mechanistically different HIV-1 RTIs (TFV, FTC, and EFV) on common ground.
As has been demonstrated in biochemical assays, the efficiency of DNA polymerization is dictated by the processivity of HIV-1 RT, which depends on the relative rates of the polymerization reaction (Kd and kpol) and the rate-limiting dissociation of RT (koff) from the elongating template/primer (T/P) complex (8, 14, 18, 35–37). While cell-free kinetic assays are important in establishing a clear mechanism of action(s), experimental conditions may limit their applicability to provide a global and natural representation of RTI-based inhibition. First, Ki/Km does not permit comparison of NRTIs and NNRTIs on common ground, since NNRTIs do not prevent dNTP or T/P binding but rather induce a conformational change in RT that reduces dNTP incorporation (kpol). Thus, unlike NRTIs, NNRTI-based inhibition cannot be overcome by increasing concentrations of the natural substrate (10, 38). Second, kinetic inhibition parameters can vary depending on the type (RNA or DNA), nature (homopolymeric versus heteropolymeric), sequence, and length of T/P (short versus long) used (12, 15–19). Third, RT polymerization is initiated in steady-state assays from multiple sites, leading to both full-length and incomplete products, whereas minus-strand vDNA synthesis is initiated biologically from a single start site, leading to a full-length product. Fourth, the concentration ratio of RT to T/P employed may impact the relative contribution of diminished processivity versus chain termination toward inhibition efficiency by RTIs. Under steady-state conditions this ratio is low, and a few RTs moving between many T/Ps and initiation sites may be more apt to diminish processivity by forming irreversibly terminated dead-end RT-T/P complexes (DECs) (39–41). Under pre-steady-state conditions this ratio is high, and inhibition is driven by chain termination in the form of single nucleotide extension. In comparison, RT is in excess (n = 20 to 50) relative to T/P (n = 2) in the natural biological milieu (42), and despite the removal of some RTs due to DECs, many opportunities for chain termination exist along the lengthy vRNA template. Finally, cell-free measures of inhibition efficiency do not globally account for the possibility of reversible chain termination (NRTI excision) or binding (NNRTI dissociation). For these reasons, it is at times difficult to compare and validate inhibitory results from cell-free biochemical analyses with cell-based analyses.
Clinical drug efficacy is measured in terms of the percentage of patients that achieve plasma HIV-1 RNA levels of <50 copies/ml, and therefore, cell-based inhibition potency (EC50) has served as a surrogate to describe inhibitory potential. Determining inhibition at clinical concentrations can, however, be limited by the need for a method to extrapolate from EC50s to higher concentrations. In addition, these antiviral activity assays measure cell death or reporter-based expression long after reverse transcription is complete and thus do not permit a mechanistic assessment of intrinsic inhibition efficiency during active vDNA polymerization. To properly assess such a parameter, a cell-based system certainly provides for the intricacies of reverse transcription initiated from a single
As expected, 1-fold and 10-fold EC50s of TFV, FTC, and EFV preferentially blocked the accumulation of vDNA products as a function of vDNA chain elongation, similar to previous findings using other RTIs (22, 23, 25). However, curve fitting as a measure of inhibition efficiency appeared to be limited by the accuracy of nonlinear regression, the RTI concentration used, and an unaccounted number of chain termination stop sites in each vDNA product. Building upon these observations, we then devised an improved strategy to determine an intrinsic measure of inhibition efficiency that would be independent of such factors as vDNA product length, the number of potential stop sites, the concentration of the RTI, and the mechanism of inhibition. We reasoned that chain termination (by NRTIs) and/or stopping reverse transcription (by NNRTIs) could be thought of as in stochastic terms (outlined in Fig. 4). Moreover, a relationship between inhibition potency per vDNA product (EC50), the number of potential stop sites, and a basic probability, p, of chain termination (stopping) per stop site could be established in a manner that would allow for determination of p by successive approximation (see Fig. S1 in the supplemental material). Using this strategy, we found that TFV exhibited the greatest basic probability of chain termination per stop site, followed by FTC and then EFV, which together reflects different intrinsic inhibition efficiencies. However, the overall probabilities, P(non_end), of not completing full-length reverse transcription (PBS, 9,103 nt) were more comparable between all RTIs (∼1.03- to 1.2-fold), which supports their clinical effectiveness in combination therapy. Convergence of the two expressions of P(non_end) at the final iteration (see Fig. S2) validated the functional relationship we derived between experimental EC50 and basic probability of chain termination.
While large differences in inhibition potency (EC50) between TFV, FTC, and EFV might imply corresponding differences in inhibition potential, an examination of their intrinsic inhibition efficiencies would suggest otherwise. In truth, the decay rate of viremia following initiation of combination antiretroviral therapy and the clinical outcome of drug efficacy are influenced by many complex factors that move beyond simple measurements of EC50. Such biological and clinical factors include drug dose, half-life, metabolism, distribution, toxicity, and tolerability, as well as drug-drug interactions, adherence, and genetic barriers to resistance. Much like recent assertions that dose-response curve slopes influence class-specific inhibitory potential of antiretroviral drugs (44), intrinsic inhibition efficiency parameters may provide an additional dimension for understanding the clinical impact among different RTIs routinely administered in combination. Nonetheless, our determination of p and P(non_end) as measures of the inhibition efficiency of RTIs represents an advancement over previous cell-based methods of analysis that rely solely on inhibition potency for comparison. Our strategy is also unique in that it allows for comparison of competitive (NRTIs) and noncompetitive (NNRTIs) inhibitors on common ground by reducing the complex mechanism(s) of reverse transcription inhibition to the basic fundamental level of stochastic chain termination.
There are currently no other cell-based data in the literature that can be used for comparison to our measures of intrinsic inhibition efficiency. While p and P(non_end) provide global biological estimates of the ability to stop polymerization per stop site and per given template length, respectively, cell-free parameters (Ki/Km and kpol/Kd) provide resolution of the impact of RTIs on catalytic and chemical aspects of (d)dNTP incorporation and chain elongation under predefined conditions. Reported Ki/Km values for TFV-DP (0.44 to 0.50) and FTC-TP (2.8 to 3.2) have suggested that TFV-DP has a higher capacity for binding/incorporation relative to dATP than FTC-TP does relative to dCTP (45–49). Similarly, our ranking of their basic probability of chain termination agrees with the kpol/Kd values for TFV-DP (0.58 to 1.44 μM−1 s−1) and FTC-TP (0.06 to 0.09 μM−1 s−1) that have suggested a higher catalytic efficiency of TFV incorporation relative to FTC-MP (13, 50–52). Not to be directly compared, EFV exhibits a remarkably low kpol/Kd (0.06 μM−1 s−1) for incorporation of dTTP (38) and a low dissociation constant (koff) (53), which together may explain its observed inhibition efficiency.
Our estimates of inhibition efficiency, represented by p, are associated with the accuracy of EC50 determinations of inhibition potency across all stages of reverse transcription: 9 vDNA transcripts of different lengths. Therefore, the determined value of p is an average value for all potential stop sites during reverse transcription. It is unclear whether additional experimental measurements on vDNA products less than 1,000 nt would have significantly improved our estimations of p, since short transcripts are inherently more difficult to inhibit and thus prone to larger EC50 variability. Since EC50n cannot have a theoretical upper limit, the function F cannot be chosen such that EC50n as a function of F(EC50n) plateaus. Therefore, we recommend at least the current set of primers and span of measurements in determinations of p. By pretreating the cells prior to infection, our assay was also designed to ensure the presence of adequate drug levels. This was particularly important, as EFV is immediately active upon administration, while both TFV and FTC require intracellular metabolic conversion to their fully phosphorylated forms. Compared to the T-lymphoblastoid cells used in this study, it would be expected that EC50s obtained using primary cells would be significantly different. This is particularly true for NRTIs, for which increased potency is associated with endocytic transport pathways and the activation state of the cells (54). Although this was not tested, we would expect a similar range of increased potency across all vDNA transcripts, leading to a similar estimate of intrinsic inhibition efficiency, which, as mentioned earlier, is independent of effective drug concentration.
To permit mechanistic assessment of inhibition efficiency in the absence of complicating biological factors, our cell-based assay limited viral replication to a single-cycle format. It is possible that multiple-round replication experiments led to greater levels of inhibition by multiplication of inhibition efficiency. Our assay also cannot resolve possible differences between DNA- and RNA-dependent DNA polymerization inhibition efficiency, since plus-strand vDNA products are initiated discontinuously from multiple sites during reverse transcription (20, 25). In our functional form derivation of the basic probability of chain termination, we presupposed that RTIs do not significantly impact the interactions between RT and the T/P. Recent evidence has nevertheless suggested that NNRTIs trap HIV-1 RT in a polymerase-independent RNase H competent mode, preventing it from binding to the T/P in a polymerase-dependent mode (33, 55). This may partially inhibit or accelerate RT polymerase-dependent or -independent RNase H activity, respectively (56–58). Also, NRTIs do not alter the context dependencies per stop site or the pattern of template-based pause sites due to stalling or dissociation but may increase pausing at some sites over others (33, 59). These and other attributes of viral polymerase-based inhibition, including ATP-mediated excision and NNRTI-based dissociation, may not be experimentally measured in our assay but are certainly captured in the final determined global estimates of p.
The novel methodologies developed in this study to determine intrinsic inhibition efficiencies of HIV-1 RTIs may have several applications. Resolution of the basic probability of chain termination of different analogs of the same dNTP could indirectly support the frequency of their utilization in clinical practice and add another dimension to understanding drug efficacy. Moreover, it would be interesting to assess the intrinsic inhibition efficiency of other NRTIs and NNRTIs relative to those of RTIs under development that act by further distinct mechanisms. Such RTIs would include translocation-defective RTIs, delayed chain terminator RTIs, dinucleotide tetraphosphates, nucleotide-competing RTIs, pyrophosphate analogs, RT-associated RNase H function inhibitors, and dual-activity inhibitors (60). While the impact of NRTI or NNRTI resistance-associated mutations on p remains to be determined, such approximations could add yet another factor to understanding virologic failure with resistance. It may also be possible to predict how different RTIs may be synergistic in combination therapy by using these basic probabilities of chain termination. Accordingly, such studies would involve applying the general rule of probability additivity to fixed-dose combinations of RTIs with identical, complementary, or overlapping stop sites. Finally, this new assay and strategy could be used to characterize viral polymerase-based inhibitors in other virus systems such as respiratory syncytial virus (RSV), influenza A virus, or cytomegalovirus (CMV) in which qPCR is used to monitor virus replication (61, 62). Wide applicability may, however, be limited by differences in the timing, processivity, and continuity of viral replication.
In conclusion, we developed a cell-based strategy with application of successive approximation to effectively determine the basic probability of chain termination of mechanistically different HIV-1 RTIs (i.e., a measure of the intrinsic inhibition efficiency). We found that the tested RTIs exhibited differential inhibition efficiencies as follows: TFV ≥ FTC > EFV. Since inhibition potency alone is limited in its capacity to describe efficacy of RTIs, our results suggest that a consideration for intrinsic inhibition efficiency may additionally be important. This study is the first to successfully quantify this parameter for HIV-1 RTIs. These results may have application across multiple classes of polymerase inhibitors and multiple virus systems.
ACKNOWLEDGMENTS
We thank Joy Feng and Bruno Marchand for their critical review of the manuscript.
FOOTNOTES
- Received 25 August 2014.
- Returned for modification 26 September 2014.
- Accepted 11 November 2014.
- Accepted manuscript posted online 17 November 2014.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.04163-14.
- Copyright © 2015, American Society for Microbiology. All Rights Reserved.
