Required concentration index quantifies effective drug combinations against hepatitis C virus infection

Successful clinical drug development requires rational design of combination treatments based on preclinical data. Anti-hepatitis C virus (HCV) drugs exhibit significant diversity in antiviral effect. Dose-response assessments can be used to determine parameters profiling the diverse antiviral effect during combination treatment. In the current study, a combined experimental and mathematical approaches were used to compare and score different combinations of anti-HCV treatments. A “required concentration index” was generated and used to rank the antiviral profile of possible double- and triple-drug combinations against HCV genotype 1b and 2a. Rankings varied based on target HCV genotype. Interestingly, multidrug (double and triple) treatment not only augmented antiviral activity, but also reduced genotype-specific efficacy, suggesting another advantage of multidrug treatment. The current study provides a quantitative method for profiling drug combinations against viral genotypes, to better inform clinical drug development.


Introduction
Newly approved antiviral drugs rely upon dosage, treatment period, and drug combinations established during clinical trials. Trials require large cohorts of patients, significant cost, extensive time and strict management of ethics and compliance: Different dose regimens, treatment times and drug combinations are evaluated during trials [1,2]. Additional trials are needed to establish drug efficacy against different viral genotypes [3][4][5][6]. Despite the significant effort placed in clinical trials, escalation of dosage, increased treatment period, and combination therapy, significant improvement in efficacy have not always been realized.
Drug concentrations which achieve 50% virus reduction (IC 50 ), can be used to characterize drug activity. Lower IC 50 means that antiviral effects are achieved with lower concentrations of drug [7]; however, a lower IC 50 does not necessarily translate to higher antiviral effect. Antiviral effect depends on the Hill coefficient (m), in addition to IC 50 . A higher m value exponentially increases antiviral activity at higher doses [8][9][10][11][12][13][14]. We have previously shown that m is unique to each anti-hepatitis C virus (HCV) drug, and that augmentation of antiviral activity with escalation of drug dose is quite diverse among the types of anti-HCV drugs [14]. Multidrug treatments also result in diverse effects depending on the drug combination. In-depth profiling of drug antiviral effects can be useful in designing a treatment protocol with maximal antiviral efficacy. Such profiling could result in significant savings in clinical trials. To date, antiviral efficacy variances between different anti-HCV drugs and drug combinations has not been characterized in detail.
HCV infection is a leading cause of liver cirrhosis and hepatocellular carcinoma, serious public health problems affecting approximately 170 million people worldwide [15]. Recently, the development of new antiviral drugs known as direct acting antivirals (DAAs), have greatly improved treatment outcomes [16,17]. Commercial interests restrict the combinations which have entered clinical trials as the combinations are all company specific rather than based on any assessment of what would be the best combination for all available agents. Further evaluation of HCV DAA effects could help identify the "best" available therapy and assist with optimizing combination treatments. A new quantitative method could also support evaluation of next generation anti-HCV treatments that could lead to the eradication of HCV. In the current study, we compare antiviral profiles of different classes of anti-HCV drugs to understand diversity of effects.
We recently developed a cell culture system combined with a mathematical model for quantifying anti-HCV drug efficacy at any concentration and multidrug combination [14]. We systematically evaluated and compared the intrinsic anti-HCV activity of 15 antiviral agents and their combinations against HCV genotype 1b. In the current study, we evaluate intrinsic anti-HCV activity in both genotype 1b and 2a. We create an "effectiveness" ranking for HCV replication inhibition in mono-and multi-drug cultures following exposure to high drug dose ranges. Significant diversity was observed between the antiviral activity profiles of different drugs. Thus, it is necessary to carefully select multidrug combinations to increase drug efficacy. We have demonstrated that the developed ranking index is able to delineate the advantages of past firstin-line anti-HCV treatment choices [14]. Thus, in the current study, we use the combined cell culture plus mathematical modeling approach to quantify efficacy of diverse antiviral drug combinations. This framework could be applied to other diseases requiring multidrug treatment, such as tuberculosis and cancer.

Methods
Anti-HCV effect of each drug against genotype 1b or genotype 2a was evaluated with subgenomic replicon systems. As a genotype 1b model, LucNeo#2 (LN2) cells were employed that carry a dicistronic subgenomic replicon including open reading frames (ORFs) for the firefly luciferase-neomycin phosphotransferase fusion protein (translated by HCV 5′-untranslated region) and the NS3-NS5B region of HCV genotype 1b strain NN (translated by encephalomyocarditis virus (EMCV) internal ribosome entry site) [18]. Huh-7.5.1 cells transfected with a subgenomic replicon that included the ORFs for the NS3-NS5B region of HCV genotype 2a strain JFH-1 and the firefly luciferase gene (SGR-JFH1/Luc) were used for a genotype 2a model [19]. These cells were seeded at 7 × 10 3 cells per well and treated with indicated concentrations of various drugs. Following 72 h of incubation, cells were lysed and cellular luciferase activity was measured to evaluate the HCV replication activity with a Luciferase Assay System (Promega) per manufacturer's protocol [18].
Fourteen anti-HCV drugs were evaluated as single treatments. Eleven of these were direct-acting antivirals . The other 3 drugs tested were host-targeting agents (HTAs) including interferon-alpha (IFNα) and cyclophilin inhibitors [Cis: cyclosporin A (CsA) and SCY-635. For multidrug studies, cells were treated with combinations of two or three drugs prior to evaluation of activity. All anti-HCV agents were purchased or kindly provided as described [14]. Fig. 1a provides a schematic of the combined experimental and mathematical system that we previously developed for quantifying anti-HCV activity of drug(s) [14]. In the previous study 14 anti-HCV agents were evaluated in mono and combination treatments against HCV genotype 1b [14]. In the current study the same 14 drugs (Table 1) were tested against HCV genotype 1b ( Fig. 1b) and HCV genotype 2a (Fig. 1c). Antiviral activity results from mono and combination treatments were used to develop a novel ranking index, the "required concentration index" or RCI (see below). Note that 14 anti-HCV agents include 11 direct-acting antivirals

Ranking anti-HCV mono-drug treatments
As shown in Fig. 1b, c, the antiviral profile of drugs against HCV genotypes 1b and 2a vary widely, suggesting that anti-HCV drugs exhibit strain-dependent effects. The typical dose-response curves of a single antiviral drug can be analyzed using the following hill function [14] (Fig. 1d): Here, f u represents the fraction of infection events unaffected by the drug (i.e., 1 − f u equals the fraction of drug-affected events). D is the drug concentration, IC 50 is the drug concentration that achieves 50% inhibition of activity, and m is the slope of the dose-response curve (i.e., Hill coefficient) [14]. Dose-response curves for drugs with higher m values show stronger antiviral activity at the same normalized drug concentration so long as the drug concentration is higher than IC 50 (Fig. 1d). Least-square regression analysis was used to fit Eq.(1) to dose-response curves (Fig. 1b, c) and estimate IC 50 and m values. Estimated values for each drug against each HCV genotype are summarized in Table 1. The hill function may not accurately fit the dose-response curve at lower drug concentrations (Fig. 1c, especially for doses lower than IC 50 ). Typical clinical drug  [18,19]. Cells were incubated for 72 h with or without drug(s) then harvested and luciferase activity detected. Inhibition of HCV replication was quantified as the luciferase activity in drug-treated cells, relative to untreated cells. b Log-Log plots of dose-response curves normalized by IC 50 (x-axis), determined from HCV genotype 1b subgenomic replicon assay of NS3 protease inhibitors (PIs; TPV, DPV, ASV, SMV), nucleoside-type NS5B polymerase inhibitor (NI; SOF), non-nucleoside-type NS5B polymerase inhibitors (NNIs; VX, DAS, NSV, TGV), NS5A inhibitors (NS5AIs; DCV, LDV), interferon (IFNα), and cyclophilin inhibitors (CIs; CsA, CSY). Each point represents the mean of three experiments. Least-square regression analysis was used to fit Eq.(1) to the corresponding dose-response curve for estimation of IC 50 and m value for each drug against HCV genotype 1b. c Log-Log plots of dose-response curves from HCV genotype 2a subgenomic replicon assay of PIs (ASV, SMV), NI (SOF), NNIs (DAS), and NS5AIs (DCV, LDV). Each point represents the mean of three experiments. Least-square regression analysis was used to fit Eq. [1] to the corresponding dose-response curve for each drug against HCV genotype 2a. d Dose-response curves for hypothetical drugs with m = 1 and 5. Drugs with a higher m value show stronger antiviral activity at the same normalized drug concentration concentrations are around 10-to 100-fold of IC 50 , therefore it is generally possible to quantify effectiveness of anti-HCV drug(s) with this method especially for such a high drug concentration. As discussed in recent publications [8][9][10][11][12][13][14], both IC 50 and m values are needed to accurately estimate antiviral drug potency, though only IC 50 is widely used in the drug development field. Since estimated values for each drug differ relative to target HCV genotype, it is important to optimize mono and combination therapy against each genotype.
To characterize efficacy of drugs, we calculated a "required concentration index" (RCI) for each anti-HCV drug against genotype 1b and 2a. Assuming 1 − f u = x inhibition of viral replication, the RCI x represents the critical fold increase of IC 50 requiring x inhibition of viral replication. Solving Eq.(1) for D/IC 50 , then RCI x is represented as follows: Here, D x is the drug concentration required to suppress x of viral replication. Drugs with small RCI x values are more efficient inhibitors of HCV replication than drugs with high RCI x . Interestingly, high m tends to be associated with smaller RCI x . By substituting estimated IC 50 and m parameters and setting x to 0.95 in Eq.(2), we calculated the RCI x required for 95% inhibition of HCV replication (i.e., RCI 95 ). We summarize RCI 95 values of each drug against genotypes 1b and 2a in Fig. 2a, b, respectively. It should be noted that SOF, a nucleoside-type polymerase inhibitor used as a key agent in current and past DAA combinations, was effective in both genotype 1b and 2a, which is  [20].

Ranking anti-HCV multi-drug treatments
Using the replicon system, the antiviral activity of doubleand triple-drug combinations (Fig. 3 & Fig. 4) were investigated using consistent ratios of drug concentrations (i.e., 0.25 × IC 50 , 0.5 × IC 50 , 1 × IC 50 , 2 × IC 50 , and 4 × IC 50 ). Inhibitory activity was evaluated for 43 double drug combinations against HCV genotype 1b, and, 9 double drug combinations against genotype 2a. Results are shown in Fig. 3a, b, respectively. Here, D a , D b , …, D i are defined as the concentration of drug a, b, …, i and IC a 50 , IC b 50 , …, IC i 50 refer to the corresponding IC 50 . Combined drug concentration in these experiments is described as is the constant ratio to IC 50 of each combined drug (x-axis of dose-response curves). As shown in Fig. 3c, a similar hill function can be fit to dose-response curves of drug combinations [14]: Here, f com u is the fraction of infection events unaffected by the drug combination, IC com 50 is the constant ratio that inhibits HCV replication by 50%, and m com is the Hill coefficient [14]. In Table 2, we summarize estimated parameters, IC com 50 and m com , for double-drug combinations.
Similar to mono treatments, the required concentration index for drug combinations is derived as The RCI 95 required for 95% inhibition of HCV replication is extrapolated from the point at which the curve intersects f com u ¼ 0:05 (dashed line in Fig. 3c). Note that the critical constant ratio,D c , satisfying Eq.(4) can be uniquely determined. The RCI 95 values for double-drug combinations against genotype 1b and 2a are summarized in Fig. 3d, e, respectively. RCI 95 varies depending on drug combination. For genotype 1b, RCI 95 ranged from 1.56 to 5.14, for genotype 2a RCI 95 ranged from 1.05 to 2.28. The drug combination with the best anti-HCV profile against genotype 1b is SMV plus IFNα; Fig. 3d❶. This combination used to be the first-in-line anti-HCV drug prior to the development of DAA treatments [17]. Combinations including a non-DAA are presented as gray bars with black number designations. Combinations with DAA-only double treatments are plotted in light pink to blue and designated with white   (Fig. 3d). For the DAA-only combinations, one of the most effective treatments against genotype 1b was the combination of SMV and SOF (Fig. 3d③), a primary treatment choice in the early era of DAA-only treatment [16]. A long term first-in-line DAA combination, SOF and LDV (Fig. 3d , e⑥), ranked in the mid-range of efficacy against both genotype 1b and 2a. Most other drug combinations ranked differently against genotype 1b and genotype 2a. ASV plus LDV (Fig. 3d , e⑤) was the least effective DAA-only combination against genotype 1b, but fell in the mid-range for effectiveness against genotype 2a. SOF plus DAS (Fig. 3d , e⑨) ranked in the mid-range against genotype 1b, but ranked lowest against genotype 2a. These trends suggest an overall difference in drug effect depending on the target HCV genotype, and indicate the importance of profiling drugs against each genotype. Eight triple-DAA treatments were profiled against HCV genotype 1b and 6 triple-combinations were evaluated against genotype 2a (Fig. 4a, b). Triple combination assessments included NS3 protease inhibitor (SMV, ASV) with NS5A inhibitor (DCV, LDV) and NI NS5B polymerase inhibitor (SOF), or NS5A inhibitor with NI NS5B polymerase inhibitor and NNI NS5B polymerase inhibitor (VX, DAS). IC com 50 and m com for triple-drug combinations are summarized in Table 3. We need to note that our experimental assay can detect the range of 0:005 < f com u < 0:01 in Fig. 4a, b, whereas it is difficult to measure f com u < 0:005 in areas of higher drug concentration, reaching to the detection limit of the assay. RCI 95 values of triple-drug combinations against genotype 1b and 2a are summarized in Fig. 4c, d, respectively. RCI 95 values ranged from 1.21 to 2.33 for genotype 1b and 0.58 to 0.98 for genotype 2a. Triple combination treatment with SOF, LDV and SMV was most effective against genotype 1b (Fig. 4c①), and least effective against genotype 2a (Fig. 4d⑥). SOF plus DCV and SMV (Fig. 4c②) was also significantly effective against genotype 1b, consistent with the reported clinical efficacy of this triple combination [21,22]. These results show the optimal combination of drugs to suppress viral replication in vitro, and shed light on the promising drug combinations for improving clinical outcome.
The correlation in ranking between the required concentration index and clinical data suggest that this method could assist with the search for drugs that achieve an efficient antiviral inhibition with different HCV genotypes.

Discussion
Our study shows that the concentration of drug (calculated as fold of IC 50 ), that achieves 95% virus inhibition (RCI 95 ), highly varied depending on the type of drug and combination with other drugs. RCI 95 of drugs in mono treatment ranged as much as 4.2 fold in antiviral activity against HCV genotype 1b (Fig. 2a, RCI 95 = 5.08-21.4). This diversity in RCI 95 indicates the importance of characterizing more than just the IC 50 of drugs when predicting antiviral efficacy in clinical settings. In double-drug combinations, RCI 95 values decreased (Fig. 3d, e) compared with mono treatments (Fig. 2a, b), indicating elevated antiviral activity resulted from combination treatment. The RCI 95 values of DAA-only double combinations ranged from 1.70 (SMV & DAS) to 5.14 (ASV & LDV) in genotype 1b and from 1.05 (SOF & DCV) to 2.28 (SOF & DAS) in genotype 2a. Thus, the diversity in RCI 95 is different among genotypes. Genotype differences are probably due differences in replication activity and the varied dependency on target [23,24].
Triple DAA treatments have become the final strategy for improving treatment outcomes, especially with difficult-to-treat HCV. Triple combinations are also used as a means to shorten treatment periods. Understanding the activity of triple DAA combinations is important in advancing towards worldwide eradication of HCV virus [25][26][27][28]. Consistent with ongoing clinical trials which show higher treatment efficacy of triple-drug combinations, triple combinations reduced RCI 95 beyond double-

Conclusion
In an era of rapidly progressing anti-HCV treatments, selection of the "best" combination treatment is critical to establishing the next generation of anti-HCV treatments against difficult-to-treat HCV and eventually eradicating HCV. We have developed an integrated experimental and mathematical method to evaluate the efficacy of anti-HCV dugs against HCV genotype 1b and 2a. The method was used to score mono-and multidrug treatment regimens against HCV. This scoring could be used to optimize multidrug treatment regimens prior to clinical entry.