Preexisting drug-resistance mutations reveal unique barriers to resistance for distinct antivirals

Abstract

Clinical trials of direct-acting antiviral agents in patients chronically infected with hepatitis C virus (HCV) have demonstrated that viral resistance is detected rapidly during monotherapy. In patients, HCV does not exist as a single, genetically homogenous virus but rather as a population of variants termed “quasispecies.” Preexisting variants resistant to specific antiviral drugs, overlooked in traditional hit-to-lead discovery efforts, may be responsible for these poor clinical outcomes. To enable real-time studies of resistance emergence in live cells, we established fluorescent protein-labeled HCV replicon cell lines. We validated these cell lines by demonstrating that antiviral susceptibility and the selection of signature resistance mutations for various drug classes are similar to traditional replicon cell lines. By quantifying the kinetics and uniformity of replication within colonies of drug-resistant fluorescent replicon cells, we showed that resistance emerged from a single cell and preexisted in a treatment-naive replicon population. Within this population, we determined the relative frequency of preexisting replicons capable of establishing foci during treatment with distinct antivirals. By measuring relative frequency as a function of dose, we quantitatively ranked distinct antiviral molecules on the basis of their distinct barriers to resistance. These insights into RNA virus quasispecies structure provide guidance for selecting clinical drug concentrations and selecting antiviral drug combinations most likely to suppress resistance.

Resistance to antiviral drugs is a significant worldwide health problem. Although we typically discuss drug resistance in language suggesting that resistance “emerges” after administration of an antiviral drug, classical microbial genetics suggests that drug-resistant mutants preexist in viral populations. The Luria and Delbruck experiment (1) provided the first evidence that drug-resistance mutations could preexist. In infected patients, hepatitis C virus (HCV), a positive-strand RNA virus, is composed of a swarm of genetically heterogeneous variants termed “quasispecies.” A central question is whether this diversity is so pronounced that drug-resistance mutations preexist in a drug-naïve HCV population. Despite the importance of drug resistance in driving therapeutic outcomes, identification of rare variants within polymorphic virus populations has been fundamentally difficult in vitro and in vivo.

Although many new classes of direct-acting anti-HCV drugs are in clinical development, viral resistance to these agents generally is detected within a few days of monotherapy (2, 3). Further supporting the model of preexisting resistance, ultradeep sequencing experiments identified mutations, before drug exposure, that confer resistance to NS3 protease inhibitors (4, 5). Although drug-discovery campaigns traditionally have prioritized leads and lead classes on the basis of potency and toxicity, quantitative comparisons of the frequency of drug-resistance mutations (preexisting or otherwise) have not been described. Nonetheless, resistance to direct-acting antivirals will limit the efficacy of these drugs when they are combined with current standard-of-care agents and will be even more critical as reduction or replacement of IFN is explored.

HCV replication can be achieved in vitro with replicons (6) or genotype 2a JFH-derived viruses (7), and these in vitro tools have been accepted by the US Food and Drug Administration as the standard for assessing potency and toxicity. However, the mechanisms by which resistance emerges in these systems have not been characterized previously. In addition to baseline frequencies of preexisting drug-resistance mutations, relative losses in susceptibility to antiviral agents and changes in fitness conferred by drug-resistance mutations may influence the clinical efficacy of drugs (8). Multiple reports suggest the replicon system can be used to monitor the loss of susceptibility and changes in fitness (9–13), but it is not clear that drug-resistance mutations preexist in these in vitro systems. A model that robustly and rapidly captures all three phenomena would guide target drug concentrations in plasma and rational drug combinations by elucidating the relationship between dose and the prevalence of drug resistance, termed the “barrier to resistance.”

As described here, we developed, validated, and applied an in vitro system for measuring resistance barriers for HCV antiviral drugs. We first created fluorescent protein-labeled genotype 1 HCV replicon cell lines that enable kinetic, cell-by-cell observation of viral replication. We monitored the kinetics of resistance emergence to demonstrate that drug-resistance mutations for multiple antiviral classes do preexist. We then quantified the relative frequency of these preexisting drug-resistance mutations (i.e., the number of focus-forming replicons in the presence or absence of antiviral drugs) and developed an assay for side-by-side comparison of resistance barriers in different classes of antivirals. This comparison showed that clinical-stage drug candidates exhibit distinct abilities to suppress drug resistance.

Results

Construction and Validation of Fluorescent Protein-Labeled Replicon Cells.

To enable real-time assessment of HCV replication levels at single-cell resolution, we designed and constructed subgenomic replicons with fluorescent reporter genes. Although a genotype 1b replicon with GFP inserted into domain III of NS5A has been reported previously (14), we constructed replicons with the GFP fused to the neomycin phosphotransferase II gene in the first cistron. This placement eliminates the possibility that HCV replication or antiviral susceptibility is affected by the substantial insertion in the NS5A protein. Stable cell lines with the genotype 1b GFP-encoding replicon were established by transfecting Huh7-Lunet cells with 1b-GFP replicon RNA (Fig. 1A). Similarly, stable cell lines with the genotype 1a RFP-encoding replicon were established by transfecting a genotype 1a-permissive cell line, 51C (15), with 1a-RFP replicon RNA (Fig. 1A). Individual G418-resistant colonies were isolated and expanded, and cells exhibiting robust replicon replication were enriched by fluorescence-activated cell sorting (FACS) (Fig. 1B). The final, optimized genotype 1b and 1a replicon cell lines were termed “Lunet-GFP-1b” and “51C-RFP-1a,” respectively.

Fig. 1.

Creation of fluorescent protein reporter replicon cell lines. (A) GFP-1b (Upper) and RFP-1a (Lower) replicon constructs. Replicons with fluorescent proteins (eGFP or tdTomato) in the first cistron were constructed. (B) Isolation of Lunet-GFP-1b (Upper) and 51C-RFP-1a (Lower) cell lines. Genotype 1b and 1a replicon constructs were transfected in Lunet or 51C cells, respectively, and were enriched by iterative rounds of FACS.

Validation of Fluorescent Reporter Replicons in Antiviral Assays.

To ensure that fluorescent protein levels reflect viral replication levels accurately, we measured the EC₅₀ values of known antivirals in 51C-RFP-1a cells. The potency of NS3 protease inhibitors (BILN-2061, INTM-191, VX-950, and MK-7009), allosteric NS5B polymerase inhibitors (A-782759 and HCV-796), nucleoside NS5B polymerase inhibitors (MK-608 and 2′C-methyl adenosine), and an inhibitor of host-cell function (cyclosporine A [CsA]) was determined by treating cells with multiple concentrations of inhibitor for 3 d, fixing the cells, and staining them with Hoechst dye to visualize the nuclei (Fig. 2A). Quantitative image analysis determined the percentage of cells with RFP fluorescence at each drug concentration, and data were fit to a dose–response curve to calculate EC₅₀ values (Fig. 2B). To compare the EC₅₀ values obtained by this method with those generated in replicon cell lines encoding a more traditional reporter, we performed parallel antiviral assays using the luciferase-encoding genotype 1a replicon cell line 51C-RFP-1a (15). We observed a significant correlation (R² = 0.88, P = 0.0001) in EC₅₀ values generated with luciferase and fluorescent protein reporter replicons (Fig. 2C) and thus concluded that RFP fluorescence is an accurate measure of HCV replication in the 51C-RFP-1a cell line.

Fig. 2.

Validation of fluorescent protein reporter replicon cell lines. (A) Assessing the HCV replicon response to HCV-796 by fluorescent microscopy. Images of the replicon levels (RFP expression, Upper) and nuclei (Hoechst staining, Lower) are acquired at increasing concentrations of HCV-796. (B) Determination of HCV-796 potency by quantitative image analysis. The percentage of fluorescent cells in each image is quantified (n = 4), and data are fit by nonlinear regression to a sigmoidal curve. (C) Fluorescent protein reporter levels accurately reflect replication levels. EC₅₀ values for a number of known antivirals targeting the NS3 protease (black), NS5B polymerase (red), or CypA/NS5A (blue) correlate with our historical values obtained using luciferase-labeled replicons.

Selection of Viral Resistance Using Fluorescent Reporter Replicon Cell Lines.

We next investigated whether fluorescent protein-labeled replicon cells could be used to identify, isolate, and characterize drug-resistant mutants. After serially passaging 51C-RFP-1a replicon cells in medium containing 200 nM HCV-796 and G418 for 40 d, highly replicative drug-resistant cells were isolated by FACS. The resulting cells exhibited high levels of replication at concentrations of HCV-796 that inhibited wild-type 51C-RFP-1a cells (Fig. S1A). The EC₅₀ value of HCV-796 in these resistant cells was 3.5 μM, 166-fold higher than the wild-type EC₅₀ of 21 nM (Fig S1B). Sequencing of the pooled 51C-RFP-1a replicon RNA indicated the presence of the C316Y NS5B mutation in these cells. A similar selection for resistance was conducted with the host-factor inhibitor CsA in the 51C-RFP-1a replicon cells, and the expected NS5A D320E mutation, known to confer resistance to both CsA treatment and cyclophilin A (CypA) knockdown, was observed (16). These experiments demonstrate that selections for drug resistance performed with 51C-RFP-1a cells yield populations of cells that maintain high levels of replicon and reporter gene expression and exhibit properties of genotypic and phenotypic resistance consistent with other replicon systems.

Evidence of Preexisting Drug Resistance.

To determine if drug-resistance mutations preexist in populations of treatment-naïve fluorescent replicon cells, we first monitored HCV replication levels in real time during drug treatment. We adapted a classical genetic approach, based on the Luria and Delbruck experiment (1), to determine if drug-resistance mutations are preexisting (“spontaneous”) or arise after application of selective pressure (“induced”). Because the replicon cells divide approximately once every 24 h, each isolated individual cell should expand to ∼16 cells by day 5. If a drug-resistant replicon copy existed in the cell before the first division (which we would define as “preexisting” resistance), then a uniformly fluorescent colony of ∼16 colonies would be observed by day 5 (Fig. 3A). If, instead, resistance emerged after the first division, half of the cells in the colony (or less than half of the cells, if the resistance emerged after the second cell division) would be fluorescent on day 5 (Fig. 3B). Stable replicon cell lines contain thousands of copies of replicon per cell, and the replicon population within cells is likely to be predominantly wild-type, with rare cells containing drug-resistant mutants at very low copy numbers. For this reason, an initial decrease in fluorescence levels between days 1 and 3 would be expected as the wild-type replicon is inhibited. As the resistant replicon increases in copy number within in each cell, the fluorescence level per cell should increase as well. A cell containing a copy of replicon that encodes a preexisting drug-resistant mutant therefore would decrease in fluorescence initially before becoming uniformly fluorescent on day 5.

Fig. 3.

Kinetic evidence of preexisting drug resistance. (A) A kinetic model to determine if drug-resistance mutations preexist. A cell harboring at least one replicon copy encoding the drug-resistant mutation (R) that existed before the first cell division would expand to uniformly replicating (i.e., uniformly fluorescent) drug-resistant colonies. S, cells containing only drug-sensitive variants of the HCV replicon. (B) Mutations that arise after the first division would yield resistant colonies in which half, or fewer, of the cells are fluorescent. (C) Replication kinetics indicates preexisting drug resistance. A resistant 51C-RFP-1a colony that emerged in the presence of a high dose (200 nM) of HCV-796 exhibits replication kinetics that closely follows the theoretical model for preexisting resistance described in A.

The observed kinetics of HCV replication through resistance emergence resembled the pattern shown in Fig. 3A, suggesting that the drug-resistant replicon copy was present before, not after, the first cell division (Fig. 3C). The RFP levels in 51C-RFP-1a cells were monitored across a large field of sparsely plated cells grown in medium containing 200 nM HCV-796 (more than 10-fold in excess of the EC₅₀) but without G418. Although most isolated cell colonies were reduced to background fluorescence within 24 h, rare drug-resistant colonies originating from a single cell were observed (Fig. 3C). Continued tracking of RFP levels in these colonies for 11 d showed that these replicons were capable of sustained replication in 200 nM HCV-796.

Quantifying the uniformity of fluorescence across multiple drug-resistant colonies indicated that the preexisting phenotype (Fig. 3A) was more common than the adaptive phenotype (Fig. 3B). By staining with Hoechst 33342 to visualize the nuclei, we were able to find isolated colonies 5 d after treatment with HCV-796. Because of their significant physical separation from nearby cells, each of these colonies probably originated from a single cell. We then quantified the fluorescence of each cell in each colony (Fig. 4A). We were able to identify 12 isolated RFP-positive colonies in this experiment, and all exhibited fluorescence levels uniformly above background (Fig. 4B). There was, of course, some intracolony variation in fluorescence, but the fluorescence of all cells in all identified colonies was more than eight SDs above background. Isolated colonies in which 25–50% of the cells fluoresced were not observed. This result suggests that preexisting resistance is much more frequent than adaptive drug resistance.

Fig. 4.

Uniform replication across drug-resistant colonies supports preexisting drug resistance. (A) Quantifying fluorescence of each cell in a drug-resistant colony. By locating isolated clusters of DAPI-stained nuclei 5 d after drug addition, we identified colonies likely to have arisen from a single cell (Left). Replication subsequently was visualized, by overlaying the RFP signal with the Hoechst signal, to determine if that colony was capable of replicating in the presence of HCV-796 (Center). By quantifying the fluorescence intensity of each cell within the colony, we determined the uniformity of replication in that drug-resistant colony (Right). (B) Uniformity of replication levels across 12 distinct colonies. Each vertical cluster represents a single colony, and each data point represents a single cell. By plotting the fluorescence of each cell in each colony, we show that all identified colonies fit the model of preexisting resistance. As controls, a colony in the absence of drug (red) has uniformly high replication levels, and a colony in the presence of a 500-fold excess of INTM-191 (blue) has uniformly low replication.

Quantification of Relative Frequencies of Preexisting Drug-Resistant Mutants.

We next determined the relative frequency of preexisting drug-resistance mutations in replicon cell populations. Stable 51C-RFP-1a replicon cells were allowed to divide for more than 20 d posttransfection to generate significant genetic diversity in the replicon population. Total RNA then was isolated and transfected into Cured 51 cells. By counting foci of RFP-positive cells 5 d posttransfection, we determined the number of RFP-positive focus-forming units (FFU) in the total replicon RNA pool (Fig. 5A). To determine the fraction of RFP-positive FFU that contained drug-resistance mutations in the total RNA pool, the number of FFU in the absence of drug was compared with the number of FFU in the presence of high doses (10× EC₅₀) of the NS5B polymerase inhibitors HCV-796 or A-782759. Ten micrograms of total RNA from treatment-naïve replicon cells contained an average of 5,993 RFP-positive FFU; 96 of these FFU (1.5%) were resistant to 2,400 nM A-782759, and 66 FFU (1.1%) were resistant to 150 nM HCV-796 (Fig. 5B).

Fig. 5.

Quantification of the relative frequencies of preexisting drug-resistance mutations. (A) A pool of drug-naïve RFP-labeled replicons, isolated from stable replicon cells to generate genetic heterogeneity, is transfected into 51 Cured cells in the presence or absence of drug. The resulting foci are counted. (B) In the absence of drug, an average of 5,993 fluorescent replicon foci is observed. An average of 66 (1.1%) of these foci are resistant to 150 nM HCV-796, and 96 (1.5%) are resistant to 2,400 nM A-782759 (n = 2).

Quantification of Barriers to Resistance for Multiple Antivirals.

To determine if different antivirals have distinct resistance barriers, we measured the frequency of drug-resistant foci as a function of dose for two molecules with similar potencies but very different mechanisms of action. A-782759 inhibits HCV replication directly by binding the palm domain of the NS5B polymerase. In contrast, CsA inhibits HCV indirectly by binding the host enzyme CypA and preventing its participation in the HCV replicon. Despite different mechanisms of action, these molecules have similar EC₅₀ values (271 and 378 nM, for A-782759 and CsA, respectively, in 51C-RFP-1a cells). A-782759 selects highly fit mutations at NS5B residue M414 that confer a large decrease in drug susceptibility (17, 18). In contrast, a single HCV mutation that confers a more than fivefold shift in potency has not been reported despite a large body of literature on CsA (19–21). When a large (10⁵) population of 51C-RFP-1a replicon cells was treated with each of these drugs, a dramatic difference in the frequency of resistance foci was observed (Fig. 6A). At 10 times the EC₅₀ (3 μM), the number foci resistant to CsA approaches the lower limit of detection, but more than 1,000 A-782759–resistant foci are present. These dose–response relationships show that drugs with similar potencies can exhibit quantitatively distinct barriers to resistance.

Fig. 6.

Antivirals with similar potencies present distinct barriers to resistance. The number of foci (n = 2) is plotted as a function of drug concentration. (A) Much higher concentrations of A-782759 (red) than of CsA (blue) are required to suppress resistant foci. (B) NS3 protease inhibitors. At concentrations normalized to the EC₅₀, the acyclic VX-950 (black) suppresses resistance more effectively than the macrocyclic BILN-2061 (green), MK-7009 (blue), or INTM-191 (red) (n = 2). (C) NS5B polymerase, NS5A, and CypA inhibitors. Normalized to potency, the nucleoside analog MK-608 (tan) suppresses resistance more effectively than the nonnucleoside NS5B inhibitors A-782759 (red) and HCV-796 (green). Normalized to potency, CsA (blue) suppresses resistance more effectively than the NS5A inhibitor BMS-790052 (black) (n = 2). To avoid jackpot effects, replicates were initiated from distinct pools of replicon cells.

We quantified the resistance barriers for other known HCV inhibitors. To account for varying potencies of inhibitors from different mechanistic classes, we normalized all concentrations to multiples of EC₅₀. We first compared four NS3 protease inhibitors: VX-950, MK-7009, BILN-2061, and INTM-191. MK-7009, BILN-2061, and INTM-191 demonstrated a plateau in the number of resistant foci (10–100) despite treatment with large multiples of EC₅₀ (up to 200-fold). In contrast, VX-950 showed significantly fewer resistant foci when treated with fivefold or greater multiples of EC₅₀ (Fig. 6B). When assessing NS5B inhibitors, we found that the number of resistant foci dropped steeply as the HCV-796 dose increased (Fig. 6C) but then reached a plateau at 10–100 resistant foci despite large increases in the EC₅₀ multiple, similar to the experience with MK-7009, BILN-2061, and INTM-191. A-782759 showed a different profile, with a less steep decline but a relatively linear decrease in the formation of resistant foci as the dose increased. The nucleoside MK-608 and the CypA inhibitor CsA showed very sharp declines in the formation of resistant foci at relatively low EC₅₀ multiples, indicating a high barrier to resistance. The NS5A inhibitor BMS-790052 demonstrated a plateau with high numbers (100–1,000) of resistant foci observed despite very high multiples of EC₅₀ (to more than 500-fold). Although the potencies and toxicities of these clinically relevant compounds have been reported, this approach directly compares their resistance barriers.

Conclusions

To study the origin of HCV drug resistance in vitro, we constructed replicon cell lines expressing fluorescent protein reporters. The sensitivity (EC₅₀) of the 51C-RFP-1a cell line to multiple classes of antivirals agreed with values reported in the literature as well as with values generated in a parallel assessment using a replicon cell line with a luciferase reporter. After serial passaging in the presence of the NS5B polymerase inhibitor HCV-796, FACS selected a highly drug-resistant cell population. Sequencing this resistant cell line revealed the presence of the C316Y mutation, previously shown to confer resistance in vitro and identified in patients during treatment with HCV-796 (22, 23). These results show that the fluorescent reporter gene is an accurate indicator of replicon levels and can be use to select drug-resistant mutants.

We next used these cell lines to show that drug resistance preexists in replicon cultures. Measuring the kinetics and uniformity of fluorescence in resistant replicon colonies indicated that drug resistance preexisted in the replicon cells that founded individual resistant colonies. Drug-resistance selections are a common and essential tool in antiviral drug discovery, and these results show that the resistance observed during in vitro replicon studies results from the outgrowth of preexisting mutations rather than from the generation of mutations after the onset of antiviral suppression. Clinical identification of drug-resistance mutations within days of therapy initiation suggests preexisting resistance in vivo as well (2).

To assess the relative frequency of preexisting drug-resistance mutations, we compared the number of foci resistant to distinct antivirals. Because the number of replicon copies per cell is not known, it is unlikely that these values reflect the absolute frequency of drug-resistance genomes. However, these studies indicated that 1.1% and 1.5% of preexisting replicons were competent to form foci in concentrations of HCV-796 or A-782759, respectively, that were more than 10-fold in excess of EC₅₀. Previously described genotypic techniques such as deep sequencing (5) identified preexisting mutations at slightly lower (0.3–2%) frequencies. Together, these results shed light on the composition of the HCV replicon population.

To extend the limit of detection and improve throughput, we developed a streamlined assay to quantify the relative frequency of drug-resistance mutations in our stable 51C-RFP-1a cells. If estimates of 1,000–5,000 copies of replicon per stable replicon cell are accurate (24), the starting pool of 10⁵ replicon cells in this assay contained more than 10⁸ replicons. We previously showed that mutations resistant to 150 nM HCV-796 were present at a relative frequency of 1.1%, and that same concentration produced more than 800 distinct resistant foci in this assay. This result represents a significant extension of the limit of detection as compared with our previous approach and the most sensitive previously described techniques. Using a high-content imaging platform allowed us to automate the quantification of foci and assess many compounds and drug concentrations in parallel. Thus, we were able to measure efficiently the number of resistant foci produced by multiple antivirals (nine) at multiple drug concentrations (four to seven) in replicate assays. With few other tools available to investigate the diversity of an RNA virus population, these analyses provide a highly sensitive glimpse into the structure of HCV replicon quasispecies.

The relationships between the frequency of drug-resistant foci and dose described in Fig. 6 are supported by the existing knowledge of both fitness and phenotypic resistance levels reported for drug-resistance mutations. The in vitro resistance barrier of the covalent protease inhibitor VX-950, for example, appeared to be greater than those of the noncovalent inhibitors MK-7009, BILN-2061, and INTM-191. Although all these protease inhibitors lose potency against A156T and R155K mutations, the overall fold of resistance conferred by R155K is smaller for VX-950 (∼10 fold) than for MK-7009, BILN-2061, and INTM-191 (70- to 250-fold) (9, 25). VX-950 also retains activity against mutations at position D168 that confer very high levels of resistance to MK-7009, BILN-2061, and INTM-191 (up to >100-fold) (9, 26–28). Because the R155K and D168V mutations are more fit than A156T, they probably are present in the replicon quasispecies at higher frequencies (9). The dose–response curves in Fig. 6C capture the complex interplay between known alterations in replicon fitness and antiviral potency conferred by NS3 protease mutations.

Antivirals that select low-fitness drug-resistance mutations benefit from both the reduced replication rates of drug-resistant mutants and the reduced frequency of drug-resistance mutations in the treatment-naïve population. This effect is particularly relevant to the resistance barriers for the NS5B inhibitors in Fig. 6C. The S282T mutation is the only substitution reported to confer resistance to the nucleoside MK-608 and has significantly impaired fitness (11, 29); accordingly, this compound showed the greatest barrier to resistance for any of the direct acting antivirals we tested. In contrast, multiple mutations can be selected by both A-782759 (H95Q, N411S, Y448H, and M414L/T) and HCV-796 (C316S/F/Y/N, S365A/T, L392F, and M414I/T/V), and these mutations range in fitness from significantly impaired to highly fit (17, 18, 23, 29). Consistent with this knowledge, the resistance barrier curves for these nonnucleoside inhibitors demonstrate significant numbers of resistant foci persisting at high multiples of EC₅₀.

Finally, two compounds targeting neither NS3 nor NS5B directly, CsA and the NS5A inhibitor BMS-790052, show perhaps the most distinct resistance profiles (Fig. 6C). Mutants conferring >1,750-fold resistance to BMS-790052 are strikingly more frequent than mutations conferring resistance to a mere 10-fold excess of CsA. Although host factor inhibitors, such as CsA, are thought to present the highest barrier to resistance, we observed a comparable dose–response for the nucleoside analog, MK-608. Together, these data provide a quantitative comparison of resistance barriers in these classes of HCV inhibitors.

Although we quantified drug-resistance foci at multiple drug concentrations, the pharmacokinetics of individual inhibitors ultimately will dictate which concentrations are clinically relevant. For example, despite administration of a high dose of VX-950 three times per day, the concentration achieved in the clinic was insufficient to suppress resistance (2). The plasma binding-adjusted trough (C_min) concentrations were, in fact, less than twofold greater than the EC₅₀ (30, 31). Therefore the R155K mutation and other mutations that confer relatively low-level resistance (e.g., T54N and V36A/M) were detected (27). These pharmacokinetic constraints would have to be considered in concert with the dose–response curves in Fig. 6 to estimate which antivirals or antiviral combinations have the maximum likelihood of suppressing resistance in the clinic.

Quantifying relative frequencies of drug-resistance mutations within an HCV population may have practical implications for antiviral dose selection. Target-dose estimation for early-stage clinical trials, for example, typically is determined by potency and safety (but not resistance-suppression) metrics. Because estimates of total viral load suggest the presence of as many as 10¹² copies of virus in an HCV-infected patient (32), mutations with frequencies too low to detect by allele-specific PCR or deep sequencing may, nonetheless, be relevant in these clinical trials. If target-dose estimates were based, additionally, on the relationship between dose and preexisting resistance frequencies, clinical outcomes might improve.

Although preclinical antiviral drug discovery has relied heavily on quantitative and side-by-side comparisons of small-molecule potency (EC₅₀) and toxicity (the mean 50% cytotoxic concentration, CC₅₀), equally quantitative and side-by-side comparisons of barriers to resistance have not been reported. This report shows that distinct antiviral compounds indeed exhibit unique barriers to resistance independent of their potency or toxicity. The assay described here enables the ranking of antiviral molecules based on the prevalence of preexisting drug-resistance mutations. In combination with existing replicon-based tools for measuring mutation fitness and the loss of antiviral susceptibility, a thorough in vitro assessment of an antiviral's resistance-barrier profile is possible. This additional preclinical assessment could guide the selection of antivirals (or antiviral combinations) less likely to suffer from the adverse clinical risks driven by a drug's poor barrier to resistance.

Methods

Quantification of the Relative Frequencies of Preexisting Drug-Resistance Mutations.

Total RNA was isolated from 51C-RFP-1a cells using the RNeasy kit (Qiagen). Ten micrograms of total RNA was added to 400 μL of cells. After electroporation using a Gene Pulser (Bio-Rad) and incubation for 10 min at room temperature, cells were resuspended in 30 mL of medium. Two milliliters of transfection suspension were mixed with 8 mL of medium plus the drug and plated in a PetriWell dish (Genetix). Cells were fixed in 1% paraformaldehyde plus 20 μg/mL Hoechst dye on day 5 and were imaged using the ImageXpress Micro (Molecular Devices). Replicon colonies above a size and intensity threshold (to ensure that they were preexisting) were counted using MetaXpress Software (Molecular Devices). To avoid jackpot effects, replicates were initiated from distinct pools of total RNA from replicon cells.

All other experiments were performed with standard techniques and are described in detail in SI Methods.